Sam's Laser FAQ, Copyright © 1994-2024, Samuel M. Goldwasser, All Rights Reserved.
I may be contacted via the Sci.Electronics.Repair FAQ Email Links Page.

  • Back to Sam's Laser FAQ Table of Contents.

    Laser Instruments and Applications

    Sub-Table of Contents

  • Back to Laser Instruments and Applications Sub-Table of Contents.

    Introduction to Laser Instruments and Applications

    When the laser was invented in 1960, it was amazingly, a solution looking for a problem. While the laser's weapons potential was clear, most of the uses of lasers that have changed the World were not foreseen even by the so-called experts of the time. In this chapter, we touch on perhaps one tenth of one percent of those where lasers are now indispensable, or at least have the potential to be in the future.

    But lasers are not the solution to every problem. There are applications where lasers are not useful and probably never will be. Among the short list of idiotic proposals for lasers are (in no particular order): grass and tree trimming, insect extermination, and advertising on the moon. For more details and a few chuckles, see the section: Laser Humor.

  • Back to Laser Instruments and Applications Sub-Table of Contents.


    Using a Laser to Measure Distance, Position, or Speed

    There are a variety of ways of using lasers to measure distance. The precise 3-D shape or profile of solid objects can be determined using laser scanning techniques. Common approaches include: Laser Atlanta Optics is an example of a company that specializes in laser based distance and speed measureing technology.

    Manufactures/suppliers of devices used in laser rangefinders include: E-O Devices and Analog Modules.

    Optical Rangefinders

    This is the basic principle used in 35 mm rangefinder cameras and other devices where you view the distance scene and turn a knob to line up two images that are either superimposed or split top/bottom half. In the case of the camera, turning the lens focus ring adjusts the angle of mirror A below.

              To distant scene.
              ^               ^
              |               |
              |       C/------/D    
              |A       |      
              \--------\       (B is partially silvered or a half mirror to
             adjust   B|        permit viewing of both sides from the scene.)
             angle     ^
                   view here
              |               |
              |<- baseline -->|

    The further apart the mirrors are (size of baseline), the greater the useful range. Adjust the angle of mirror A or D until the images are superimposed. Calibrate the angular setting to distance.

    The distance from A to the scene is then: tan(angle A) * baseline.

    For long distances, C and D can be eliminated - they compensate for the difference in path lengths of the two views - else the sizes would not be the same. (Even this doesn't work perfectly in any case. Can you figure out why?)

    You can add telescopes and other optics if you like - this is just the basics.

    Look Ma, no electronics. :-)

    Note that SLR cameras do NOT use this approach as they are entirely optical (meaning that adjusting the focus only controls the lens - nothing else!). With SLRs, a pair of shallow prisms oriented in opposite directions (or many in the case of a 'microscreen' type) are cemented onto a clear area of the ground glass. When the image is precisely focused onto the ground glass, the prisms have no effect. However, when the image is in front or behind, they divert the rays such that the two halves of the image move apart (or the image breaks up in the case of the 'microscreen').

    There were some "Amateur Scientist" articles in Scientific American a few decades ago on constructing several types of optical range finders. These were included in the book, "Light and Its Uses". See the section: A HREF="laserclt.htm#cltsi">Scientific American Articles on Lasers and Related Topics.

    Simple Laser Rangefinder Based on Triangulation

    (Portions from: Mike Cimorosi (

    My students construct a simple laser rangefinder using a few basic parts:


    Basic procedure:

    1. Place the laser to the left of the optical bench. Follow standard safety procedures for using 1/2 mW lasers.

    2. About 3 inches to the right of the laser aperture (opening), place the beam splitter at an angle of 45 degrees with respect to (wrt) the incident beam. This will split the beam into two different paths. Most of the beam will pass through the splitter. Some will be reflected at a right angle wrt the incident beam.

    3. About 6 feet to the right of the splitter, place the rotary table with the mirror on it and face it toward the beam that passes through the splitter.

    4. Now, before you turn on the laser, make sure you have a safe place to aim the beam for the distance you want to determine.

    5. Now fire up the laser. Note where the first reflected beam strikes the target (a wall maybe?). Now, slowly and carefully rotate the rotary table until the beam reflected from the mirror coincides with first reflected beam. You now have formed a right triangle made of laser light! Pretty neat! Remember to respect the beam, especially with respect to your eyes!!!

    6. Finally, you can use the trig relation: distance = 6 ft x tan(angle) to determine the distance. How's your trig? :-)

    7. It's not the most precise rangefinder - i.e., the equation is pretty sensitive to the angular precision of the rotary table. However, it does demonstrate the basic principle. Maybe the diagram below will help with setting up the laser rangefinder.

    Rough diagram of rangefinder setup:

                  To wall                        To wall
                     ^                               ^
                     |                                \ 
            distance | first reflected beam            \ second reflected beam
                     |                                  \
                     |                             angle \
        Laser -------/------------------------------------/
                Beamsplitter                    Rotary table with mirror
                     |<------------- 6 feet ------------->|

    Of course, you can make the non-laser version of this type of rangefinder (but this is a laser FAQ! --- sam). My students also make that one as well. Both are pretty neat and demonstrate the power of trig to determine distances!

    Comments on Laser Rangefinders

    (From: Andrzej Hanczak).

    I am just finishing the development of a range finder based on the TOF (pulse-Time-Of-Flight) measurement method. There are also different methods like phase-shift method which compares the phase shift between outgoing modulated beam and reflected light.

    The Pulse TOF method has some advantages which make it very useful: you can use relatively high pulse power and still be in the Class I safety range.

    While building such a range finder there are two crucial components which have influence on its accuracy: the time measurement circuits and the receiver. Our aim was to build a laser scanner with the resolution of 1 cm which means that you have to be able to measure the time with the resolution of 67 ps. The range of the scanner should be approx. 30m. We are not ready yet but there are some results.

    For the first prototype we used a 1.25 GHz oscillator and special microstrip design to get the resolution of 70 ps. In the current prototype we use a special prototype IC which should deliver 50 ps resolution.

    The problems are on the receiver side, a relatively large jitter (which I'm fighting now) destroys my high time measurement precision. The jitter on the input results in the distance differences of approximately 10 cm). This can be filtered out by averaging of a number of measurements and that is what we are doing now. Our measurement frequency is at present 100 kHz, but we will probably perform the averaging over 10 measurements so that effective measurement rate will be 10 kHz.

    (From: jfd (

    The problem is getting simultaneous long standoff range and extremely accurate range. You can phase detect with accuracies in the sub-inch range using direct detected RF modulated LIDARS or you can use an interferometric technique with a reference to get sub-micron distances.

    (From: Robert (

    For much better resolution than would be possible with simple sampling while still maintaining low cost, digital TOF rangefinders can combine a precision analog temporal interpolator with say a CMOS system running at 100 MHz. The analog circuitry to accomplish this is in many production units (for different applications) - but 5 ps resolution has been achieved with low-cost components and in production for 15 years from at least one manufacturer. The idea is interpolate between the digital count periods with a precision time-to-voltage converter which is then sampled by microcontroller and combined with the digital counter results.

    (From: Bill Sloman (

    You may be able to achieve this at low unit cost, but getting a precision analog temporal interpolator to work well next to CMOS running at 100 MHz isn't something I'd describe as easy.

    We developed a system of this sort at Cambridge Instruments between 1988 and 1991 using a mixture of 100K ECL and GigaBit Logic's GaAs for the digital logic. Any digital signal going to or from the analog temporal interpolator was routed as a balanced pair on adjacent tracks, and we were very careful about the layout, but we still had to work at getting the noise on the interpolator output down to the 60 picosecond jitter on our 800 MHz master clock (getting a better master clock was the next priority).

    Current-steering logic (like ECL and GaAs) is a lot quieter than voltage-steering logic (like TTL and CMOS), which is why very fast DACs and ADCs use ECL interfaces. Precision analog interpolators are no less sensitive.

    Do you know who has actually achieved that 5 ps resolution and for what application? Tektronix and time domain reflectometers come to mind, though Tektronix isn't exactly cheap. IIRR Triquint was originally their in-house analog foundry and I think Tektronix has been using GaAs ASICs in their faster gear for quite some time now.

    The hybrid approach certainly isn't new, but getting it to work is a fair test of one's analog skills.

    Of course, using phase-shift not only makes for easier circuit design, but also lets you run your LED at a 50% duty cycle, giving you a lot more reflected photons to work with than the 0.01% you get with TOF.

    (From: Lou Boyd (

    The Texas Instruments book "Optoelectronics: Theory and Practice" published by McGraw-Hill had a chapter (23) on the design of an LED/Si Diode rangefinder with schematics of the transmitter, receiver, and timing section. This was a phase modulated design but obsolete by todays standards. Low cost modern rangefinders like those by Leica or even Bushnell are far more advanced in the detection circuit than that in the TI book. Most eye-safe commercial rangefinders use phase modulated techniques. This gives good accuracy but limited range, usually less than 1 kilometer with measurement times typically 1/10 second.

    Most military rangefinders use a much higher power transmitter with a time of flight method. A time of flight rangefinder just sends a single pulse and receives it. Some use multiple pulses for improved resolution and range but that typically isn't necessary. A counter is started on the rising edge of the transmitted pulse and stopped when the rising edge of the receive pulse is detected. If the counter is measuring a 150 MHz (approx) clock the range will be displayed in meters. Unfortunately that fast of counter requires at least a few high speed chips beyond the capability of standard CMOS or TTL logic. Since the round trip takes only 6.667 microseconds per kilometer you don't even need blanking on the displays. They can be attached directly to the counters or just read by a computer. A four or five digit counter suffices for most purposes. There is a little added complexity on sophisticated units for making the sensitivity of the receiver increase with time after the pulse is transmitted. This is sometimes done by charging a capacitor attached to a gain control which increases the gain with the square of time out to the maximum the unit is capable of. These rangefinders tend to be expensive because of the technology but the electronics is simple in concept. Ranges are limited only by the transmit power which can be extremely high using solid state Q switched lasers.

    Surplus lasers and the associated electronics from military rangefinders have been showing up on the surplus market in the $300 range. Unfortunately the receivers have not.

    For some insight on the level of complexity involved look at the boards sold by E-O Devices These are time of flight pulsed laser rangefinder components designed for use primarily with LED's or diode lasers. Also check Analog Modules for examples of state of the art variable gain rangefinder receivers. If you want one of their modules plan on spending between $1,000 and $2,000. :-(

    Phase shift methods allow achieving high precision in distance resolution with lower power and lower speed circuitry. That equates to lower cost and higher precision. Which type is best depends on what properties are needed.

     Parameter      Single Pulse           Phase Shift
     Range          100 m to 100 km        1 m to 10 km
     Resolution     1 m any target         1 mm corner cube to 1 m any
     Cost           $5000 and up           $100 and up
     Power level    10 w to 1 MW           1 mW to 1 W
     Time to read   sub-ms                 0.01 to 10 seconds
     Applications   artillery, navigation  surveying, hunting

    Single pulse rangefinders typically use YAG or erbium lasers while most of the phase shift type use diode lasers.

    (From: Don Stauffer

    Which type to use depends a bit on what range resolution you are looking for. If you want high resolution, you will be working with a high modulation frequency. Then you may find many circuits designed for receiving audio modulation may not provide enough bandwidth.

    Also, there is the range ambiguity problem. If you go high enough in frequency, you may find some range ambiguity.

    You will also likely be needing very accurate phase measurement circuits if you are using moderate modulation frequency, so study carefully high accuracy phase detectors. These are not trivial circuits. In order for them to work well, you need a pretty good SNR.

    (From: A. E. Siegman (

    Adding to what others have said, hand-held laser rangefinders using low-power RF-modulated CW lasers (a.k.a. diode lasers) together with phase-detection techniques are simpler, cheaper, smaller, *much* more battery efficient, and much safer; and are more or less replacing the pulsed hand-held versions of yore.

    These techniques are also moderately old. Coherent (maybe Spectra also) were making widely used laser surveying instruments ("Geodolite"?) that worked this way a couple decades or more ago (and there may have been incoherent light source versions even further back).

    I suppose that compared to TOF, one disadvantage is that it takes longer to integrate up the signal to get a range finding, and if you're in a tank battle and want to get off the first shot before alerting the enemy that you're illuminating him and giving him a chance to duck, the pulsed type may still be better.

    Do some web searching: You can buy binoculars with a built-in diode laser rangefinder from Amazon, and use it to measure the distance to the pin on your next golf outing.

    (From: Louis Boyd (

    Prior to laser diodes (1960's) there were optical geodimeters which used a tungsten lamp, a Kerr shutter (which modulates light at multi-megahertz rates using polarizers and high voltage rf driven nitrobenzene), and photomultiplier receivers. These could measure distances to a few centimeters at ranges of several kilometers. They were large, expensive, and a bi*ch to calibrate. They used phase shift techniques similar to modern diode rangefinders, but without the aid of microprocessors. They switched modulation frequencies to resolve phase ambiguities.

    Modern rangefinders often use pseudorandom modulation and cross-correlation computation to give the round-trip delay which is proportional to distance. Distance resolution can be much finer than the length of the shortest pulse.

    With modern geodimeters the distance accuracy is primarily limited by uncertainty of light propagation velocity in the air since it's not practical to measure the pressure and humidity at all points along the path, but can be accurate to better than 1 part in 10^6 with care. Tape and chain is difficult to get better than 1 part in 10^3 which is the typical accuracy of $200 pocket laser rangefinders.

    (From: Mike Poulton (

    Using pulses is not very practicable - if you want to achieve a resolution of a few mm over a distance of 100 m or so, you find that you'd need extremely short pulses (recall that 1 ns corresponds to 30 cm or 12 inches, approximately, so you's need pulses of a few ps); you could do this with a W-switched SS laser, but those little hand-held devices, who do have a resolution in this order of magnitude, cannot work in this way. They use a RF-modulated CW signal from a laser diode, say with 100 MHz, and measure the phase shift of the 100 MHz signal between outgoing and incoming beams. This phase shift can be very accurately measured by first converting the 100 MHz down to a few 100 kHz (like a superheterodyne receiver).

    Some while ago I had been interested in such a circuit myself (for measuring optical path lengths) but didn't find anything useful on the web.

    (From: Repeating Rifle (

    Equipment of this ilk is called *distance measuring equipment* or DME and has all but replaced the use of chains in surveying practice. Various implementations have been used. Some use high frequencies to obtain precision and lower frequencies for range ambiguity resolution. Others use inconmensurate frequencies that are not all that different from one another. I you match the filtering to the transmission, you pretty much get the same signal to noise ration for all kinds of devices. The broad-band pulses mentioned above use short pulses. The CW devices use narrow band filters.

    The first items of this nature used RF directly without light.

    Trade names that come to mind quickly are tellurometer and geodimeter.

    For the military rangefinders that use high power pulses, signal processing is less than optimum. An error of 5 meters will usually not be a big deal. For surveying, that kind of error will usually be unacceptable. In both cases extended (in range) targets will introduce error.

    Almost all of the inexpensive hand-held rangefinders on the market use a simplified form of phase detection with relatively low modulation rates. Phase sensing rangefinders uses a variable pulse width modulated laser diode. It would use use thousands of on/off transitions in determining each distance measurement by comparing the modulation pattern to the returned signal using cross-correlation techniques. Resolution is a function of measurement time, speed and size of the registers, and instrument stability. Single pulse TOF rangefinders on the other hand are generally used for very long ranges (several km and up) with very high pulse power (kilowatts to megawatts peak) and range resolution rarely better than a meter. Low power single pulse rangefinders are rare as the expense of the detection circuits isn't justified for the low resolution.

    The accuracy of quality surveying distance meters is limited primarily by the uncertainty of the velocity of propagation of light through the atmosphere. That varies of with air pressure and humidity which can't easily be determined over the entire path. Still, they're orders of magnitude better than a tape or chain.

    (From: Phil Hobbs (

    Modulated CW measurements also allow you to use very narrow measurement bandwidths very easily (e.g. with a PLL), which helps the SNR very much. In shorter range units, sinusoidal modulation can also be used to prevent back-reflections from causing mode hopping. You choose delta-f so that the phase modulation of the back-reflection (in radians) is at a null of the zero-order Bessel function J0. This can make a huge difference (3 orders of magnitude) in the back-reflection sensitivity.

    Building a Time-of-Flight Laser Rangefinder

    The following is what I would suggest for a relatively low cost approach achieving 15 to 50 cm resolution and 100 meter or more range. However, also see the next section for a much simpler approach that may be adequate.

    A Q-switched solid state laser will give you short pulses with minimal fuss. A unit like the small surplus Nd:YAG laser (SSY1) described in chapter: Solid State Lasers was originally part of the M-1 tank rangefinders and thus should be ideal. It is quite trivial to build a suitable power supply these laser heads since a passive Q-switch is used and this doesn't require any electrical control.

    A few mJ should be sufficient. (SSY1 is probably in the 10 to 30 mJ range using the recommended pulse forming network.) With a Q-switched laser, the required short pulse if created automagically eliminating much of the complexity of the laser itself.

    Diode laser assemblies from the Chieftain tank rangefinder are also available on the surplus market but you probably would have to build a pulsed driver for them which would be more work.

    For the detector, a PIN photodiode or avalanche photodiode (APD) would be suitable. The preamp is the critical component to get the required ns response time. You need to sample both the pulse going out and the return since the delay from firing the flashlamp (if you are using a solid state laser) to its output pulse is not known or constant.

    15 cm resolution requires a time resolution of about 1 ns (twice what you might think because the pulse goes out and back). GHz class counters are no big deal these days.

    However, approaches that are partially analog (ramp and A/D) which don't require such high speed counters are also possible. In fact, if your digital design skills aren't so great, this is probably the easiest way to get decent resolution, if possibly not the greatest accuracy/consistency. All you need is a constant current source and an A/D (Analog to Digital converter). This can be as simple as a FF driving a transistor buffer to turn the voltage to charge the capacitor on and off with a transistor set up with emitter feedback for as a constant current source. Or, it can just be an exponential charge with non-linear correction done in software. The A/D doesn't need to be fast as long as its output word has enough bits for your desired resolution. For a typical exponential charging waveform, add 1 bit to the required A/D word size. For example, determining distance over 100 meters to to 5 cm resolution would require that the full voltage ramp be about 700 ns in duration (a bit over maximum round trip time, cut off sooner if there is a return pulse) and then sampled with a 12 bit A/D.

    Another even simpler way of doing this is to charge the capacitor as above but then discharge it with a much longer time constant and determine how long it takes to reach a fixed voltage. By making the discharge time constant sufficiently large, any vanilla flavored microprocessor could be used for control and timing.

    All in all, these are non-trivial but doable projects.

    See the previous sections on laser rangefinders for more info.

    Here is a Web site that appears to go into some detail on the design of TOF laser rangefinders:

    (From: Anonymous.)

    A laser phase shift distance meter can be constructed by analog modulation of the laser and measuring the phase shift of the return signal. With some filtering you can do multiple frequencies at the same time. Also, the feedback diode in a semiconductor laser can be used as the sensor (in which case the circuitry gets interesting). High precision can be accomplished relatively easily. I'm trying to get better than 0.1 mm (preferably better than 0.01 mm) over short distances (a couple of meters).

    Resonant Time-of-Flight Laser Rangefinder

    This is a slightly modified approach and may be made to work with relatively simple inexpensive circuitry. The idea is to use a normal IR or visible laser diode (e.g., such as from a CD or DVD player) in conjunction with a common photodiode to form an oscillator whose frequency will depend on the path delay between them - i.e., the distance to the "target". Basically, the laser diode is turned on which sends out a leading edge of a light pulse. The light hits the target and is reflected back into the photodiode, which turns the laser diode off. The loss of signal then turns the laser diode on and the cycle repeats continuously. The oscillating frequency is then equal to 1 over (4 times the distance to the target plus 2 times the internal circuit delay). A simple frequency to voltage converter drives an analog meter. No really high speed components are needed.

    This was seen as a project in a Dutch book: "Lasers in Theorie en Praktijk: Experimenten - Meten - Holografie", by Dirk R. Baur, Uitgeverij Elektuur/Segment B.V., Postbus 75, 6190 AB, Beek (L) The Netherlands.

    I'm not convinced that the circuit as presented works - there is at least one part value (C4, 100 uF) which would appear to be much larger than desired inside the feedback loop. The principle appears valid though.

    Time-of-Flight Laser Rangefinder using CCD Camera

    Each pixel of a CCD-based image sensor accumulates charge proportional to the light intensity and shutter open or "gate time". For normal video, the electronic shutter is open for a duration which is a large fraction of a video frame to maximize sensitivity and minimize aliasing in moving images. For stop motion photography, much shorter shutter open times are used. If it were possible to synchronize the electronic shutter with the generation of a light pulse illuminating the scene, then the amount of charge in each CCD cell would also depend on how long it takes for the light to reach the CCD (since the shutter would close before the light from more distant points returned). One problem, of course, is that this is possible only under very special conditions. A way to get around this would be to do the measurement in two steps:

    In order for this to be implemented with a normal CCD camera, either direct control of the electronic shutter is needed, bypassing any synchronous logic, or a "sync" output from the camera must be available. Also note that the charge integration times involved - 10s or 100s of ns - are orders of magnitude smaller than those normally used on all but very specialized CCD cameras, even with a fast shutter. So, sensitivity is going to be very low. A high power pulsed laser may be needed to generate adequate photons and even then, the CCD may not be able to supply enough charge.

    However, there are CCD image sensors that have been designed specifically for this application. They include logic on each pixel to enable the arrival time to be determined and stored. This permits an entire depth map to be captured with a single TOF pulse. See, for example: CSEM Optical Time-Of-Flight Imaging - A Technology for Multiple Applications.

    Using a CD or DVD Optical Pickup for Distance Measurements

    The simplist way of doing this may be to use the existing focusing mechanism of the pickup. Focus in a CD or DVD device depends on a reflection from a relatively flat smooth surface (the metalized information layer of the disc/k) to produce an elliptical spot back at the photodiode array. The major axis of the ellipse lies on a diagonal (45 or 135 degrees) and depends on the distance above or below optimal focus - at that point, it is a perfect circle. A four quadrant photodetector takes the difference of the amplitude of the return signals from the two pairs of diagonally opposed quadrants to determine the focus error. See the document: Notes on the Troubleshooting and Repair of Compact Disc Players and CDROM Drives for more on how optical pickups actually work.

    If the surface is smooth and flat over a scale of 5 to 10 um, this could work as a way of determining distance to the pickup. In other words, the dominant return from the surface has to be a specular reflection back to the source in order for the focus servo to lock properly. (The width and depth of the pits/lands of the CD or DVD disc is small compared to the beam so they are mostly ignored by the focus servo.) I don't know how much angular deviation could be tolerated.

    The output would be an analog voltage roughly proportional to focus error which could be mapped to lens height (assuming the device is in a fixed orientation with respect to gravity - more complex if you want to do this while on a roller coaster or in microgravity!). The total range would be 1 to 2 mm with an accuracy of a few um.

    Also see the section: Can I Use the Pickup from a CD/DVD Player or CD/DVDROM Drive for Interferometry?, which would be even more precise but more complex. The practical issues of using the guts of these devices are also discussed there.

    Using a CD or DVD Optical Pickup in a Precision Position or Angle Encoder

    Conventional optical encoders - whether they are the dirt-cheap variety inside your computer mouse or the precision type found in industrial robots and other machine tools - consist of a light source or sources, some means of interrupting or varying the light intensity based on linear position or rotation angle, and photodetectors to convert the light to an electrical signals. By using various patterns on film or glass strips or discs, relative (2 bits) or absolute (many bits) measurements can be made with a computer or dedicated logic calculating position or angle, speed or rotation rate, acceleration, and so forth from this data. Through clever design and careful manufacturing, extremely high resolution is possible using conventional LEDs or incandescent lamps for the light source(s). However, lasers can be used as well with some potential advantages - even higher precision and stand-off (some distance between the moving parts) operation.

    Since the 'stylus' of a CD player has an effective size of around 1 um (DVD would be even less), it could in principle be used to implement a very high resolution optical encoder for use in linear, rotary, or other sensing application. The stand-off distance (from objective lens to focal point) can be a couple of mm which may be an advantage as well. While this is probably somewhat less difficult than turning a CD player into an interferometer (see below), it still is far from trivial. You will have to create an encoder disc or strip with a suitable reflective pattern with microscopic dimensions. Without access to something like a CD/DVD mastering unit or semiconductor wafer fab, this may be next to impossible. Your servo systems will need to maintain focus (at least, possibly some sort of tracking as well) to the precision of the pattern's feature size. To obtain direction information, the 'track' would need to have a gray code pattern similar to that of a normal optical encoder - but laid down with um accuracy in such a way that the photodiode array output would pick it up. (Implementing an absolute encoding scheme would probably require so many changes to the pickup as to make it extremely unlikely to be worth the effort.) Of course, you also need laser diode driver circuitry and the front-end electronics to extract the data signal. Not to mention the need for a suitable enclosure to prevent contamination (like lathe turnings) from gumming up the works. And, with your device in operation, any sort of vibration or mechanical shock could cause a momentarily or longer term loss of focus and thus loss of your position or angle reference.

    If you are still interested, see the section: Can I Use the Pickup from a CD/DVD Player or CD/DVDROM Drive for Interferometry? since some of the practical issues of using the guts of these devices are discussed there.

    Measuring Speed with a Laser

    Speed is just the rate of change of position so any of the approaches that measure position can be adapted for speed measurements by simply taking a pair of readings and computing their difference with respect to time. More direct methods using CW lasers depend on using some form of the doppler shift of the reflected beam, usually of a subcarrier imposed on the the laser beam by amplitude modulation.

    For example, if the outgoing laser beam is modulated at 1 GHz and the reflected beam is combined with this same reference 1 GHz in the sensor photodiode or a mixer, for relative speeds small compared to c (the velocity of light), the difference frequency will be approximately 1 Hz per 0.5 foot/second.

  • Back to Laser Instruments and Applications Sub-Table of Contents.

    General Interferometers

    Basics of Interferometry and Interferometers

    The dictionary definition goes something like:
    "INTERFEROMETER: An instrument designed to produce optical interference fringes for measuring wavelengths, testing flat surfaces, measuring small distances, etc."
    As an example of an interferometer for making precise physical measurements, split a beam of monochromatic coherent light from a laser into two parts, bounce the beams around a bit and then recombine them at a screen, optical viewer, or sensor array. The beams will constructively or destructively interfere with each-other on a point-by-point basis depending on the net path-length difference between them. This will result in a pattern of light and dark fringes. If one of the beams is reflected from a mirror or corner reflector mounted on something whose position you need to monitor extremely precisely (like a multi-axis machine tool), then as it moves, the pattern will change. Counting the passage of the fringes can provide measurements accurate to a few nanometers!

    Basic Michelson Interferometer shows a simple implementation that's the underpinning of a wide variety of applications:

    1. The laser produces a coherent monochromatic beam which is expanded and collimated by a pair of lenses (not shown).

    2. Part of the laser beam is reflected up by the Beam-Splitter (partially reflecting mirror), bounces off of Mirror 1 and back down. A portion of this passes through the Beam-Splitter to the Detector.

    3. The remainder of the laser beam passes through the Beam-Splitter and bounces off of Mirror 2. Part of this is reflected down by the Beam-Splitter to the Detector.

    4. The two beams combine at the Detector resulting in an interference pattern of light and dark fringes or a full field varying between light and dark as the path length is changed. A screen, magnifier, microscope, or other optical imaging system for a human observer or electronic sensor may be provided to view or analyze the fringe pattern in more detail or provide input to an electronic measurement system.

    In a perfectly symmetric Michelson interferometer, the fringe pattern should uniformly vary between bright and dark (rather than stripes or concentric circles of light) depending on the phase difference between the two beams that return from the two arms. A circular pattern is expected if the two curvatures of the wavefront are not identical due to a difference in arm-lengths or differently curved optics. Stripes (straight or curved) in any direction) would be an indication of a misalignment of some part of the interferometer (i.e. the beams do not perfectly overlap or one is tilted with respect to the other).

    In the basic Michelson interferometer, about 50 percent of the light gets reflected back toward the laser and is wasted. When perfectly aligned, the return path will take exactly the same path as the outgoing laser beam, and may destabilize laser action. HeNe lasers are particularly susceptible. Both of these problems can be easily dealt with by, for example, changing the mirrors to retro-reflectors (cube-corners) or roof prisms so that the outgoing and return beams are offset and follow different paths.

    A microscopic shift in position or orientation of either mirror will result in a change to the pattern. Thus, for example, Mirror 1 may be mounted on some equipment like a disk drive head actuator that is being tested or calibrated. Its position can then be determined or controlled down to nanometer precision. For these "metrology" applications, the interferometer is set up to produce a fringe pattern with at least two sensors to determine direction and velocity in a sophisticated version of the A-B quadrature decoder used in your typical computer mouse. :) Much more on this topic may be found in the sections starting with Interferometers for Precision Measurement in Metrology Applications.

    A long coherence length laser producing a TEM00 beam is generally used for this application. HeNe lasers have excellent beam characteristics especially when frequency stabilized to operate in a single longitudinal mode. However, some types of diode lasers (which are normally not thought of as having respectable coherence lengths or stability) may also work. See the section: Interferometers Using Inexpensive Laser Diodes. Even conventional light sources (e.g., gas discharge lamps producing distinct emission lines with narrow band optical filters) have acceptable performance for some types of interferometry.

    Such a setup is exceedingly sensitive to EVERYTHING since positional shifts of a small fraction of a wavelength of the laser light (10s of nm - that's nanometers!) will result in a noticeable change in the fringe pattern. This can be used to advantage in making extremely precise position or speed measurements. However, it also means that setting up such an instrument in a stable manner requires great care and isolated mountings. Walking across the room or a bus going by down the street will show up as a fringe shift!

    Interferometry techniques can be used to measure vibrational modes of solid bodies, the quality (shape, flattness, etc.) of optical surfaces, shifts in ground position or tilt which may signal the precursor to an earthquake, long term continental drift, shift in position of large suspended masses in the search for gravitational waves, and much much more. Very long base-line interferometry can even be applied at cosmic distances (with radio telescopes a continent or even an earth orbit diameter apart, and using radio emitting stars or galaxies instead of lasers). And, holography is just a variation on this technique where the interference pattern (the hologram) stores complex 3-D information.

    NASA has some information on interferometry oriented toward cosmic measurements at the NASA Interferometry Page. And you can try your hands at aligning a Michelson interferometer at the NASA Interactive Interferometer Page.

    This isn't something that can be explained in a couple of paragraphs. You need to find a good book on optics or lasers. Here are some suggestions for further study:

  • Gordon McComb's: "The Laser Cookbook" [1} and the Scientific American collection: "Light and its Uses [5]" include various type of interferometers which can be built with (relatively) readily available parts.

  • Keysight Technologies (formerly HP, then Agilent, among others) manufacture 'Laser Interferometry Measurement Systems' based on these techniques. Information and application notes are available by searching for the key words: "Laser" or "Dimensional Measurement". For Agilent in particular, searching for "5501" or "5517" will find information on their specific systems.

  • The Astroweb Internet Resources for Astronomy Web site (and others). There are links to people interested in designing, building, and operating various types of laser interferometers. Much of the information relates directly to the testing of optical components for astronomical telescopes but there should be much of general interest as well.

    Where Does All the Energy Go?

    Suppose we have a Michelson interferometer (see the section: Basics of Interferometry and Interferometers) set up with a perfectly collimated (plane wave source) and perfectly plane mirrors adjusted so that they are perfectly perpendicular to the optical axis (for each mirror) and the beamsplitter is also of perfect construction and oriented perfectly. In this case, there won't be multiple fringes but just a broad area whose intensity will be determined by the path-length difference between the two beams. Where this is exactly 1/2 wavelength (180 degrees), the result will be nothing at all and the screen will be absolutely dark! So, where is all the energy going? No, it doesn't simply vanish into thin air or the ether, vacuum, the local dump, or anywhere else. :-)

    Your initial response might be: "Well, no system is ideal and the beams won't really be perfectly planar so, perhaps the energy will appear around the edges or this situation simply cannot exist - period". Sorry, this would be incorrect. The behavior will still be true for the ideal case of perfect non-diverging plane wave beams with perfect optics.

    Perhaps, it is easier to think of this in terms of an RF or microwave, acoustic, or other source:

    Hint: From the perspective of either of the two signals, how is this different (if at all) than imposing a node (fixed point) on a transmission line? Or at the screen of the interferometer? After all, a nodal point is just an enforced location where the intensity of the signal MUST be 0 but here it is already exactly 0. For the organ pipe, such a nodal point is a closed end; for the string, just an eye-hook or a pair of fingers!

    OK, I know the anticipation is unbearable at this point. The answer is that the light is reflected back to the source (the laser) and the entire optical path of the interferometer acts like a high-Q resonator in which the energy can build up as a standing wave. Light energy is being pumped into the resonator and has nowhere to go. In practice, unavoidable imperfections of the entire system aside, the reflected light can result in laser instability and possibly even damage to the laser itself. So, there is at least a chance that such an experiment could lead to smoke!

    (From: Art Kotz (

    We don't have to to think all that hard to figure out where all the energy is dissipated in a Michelson interferometer. Nor do we have to refer to imperfect components either. The thought experiment of perfect non-absorbing components still renders a physically correct solution.

    To summarize a (correct) previous statement, in a Michelson interferometer with flat surfaces, you can get a uniform dark transmissive exit beam. The power is not dissipated as heat. There is an alternate path that light can follow, and in this case, it exits the way it came in (reflected back out to the light source).

    In fact, with a good flat Fabry-Perot interferometer, you can actually observe this (transmission and reflection from the interferometer alternate as you scan mirror spacing).

    In the electrical case, imagine a transmitter with the antenna improperly sized so that most of the energy is not emitted. It is reflected back to the output stage of the transmitter. If the transmitter can't handle dissipating all that energy, then it will go up in smoke. Any Ham radio operators out there should be familiar with this.

    (From: Don Stauffer (

    Many of the devices mentioned have been at least in part optical resonators. It may be instructive to look at what happens in an acoustic resonator like an organ pipe or a Helmholtz resonator.

    Let's start with a source of sound inside a perfect, infinite Q resonator. The energy density begins to build up with a value directly proportional to time. So we can store, theoretically, an infinite amount of acoustic energy within the resonator.

    Of course, it is impossible to build an infinite Q resonator, but bear with me a little longer. It is hard to get an audio sound source inside the resonator without hurting the Q of the resonator. So lets cut a little hole in the resonator so we can beam acoustic energy in. Guess what, even theoretically, this hole prevents the resonator from being perfect. It WILL resonate.

    No optical resonator can be perfect. Just like in nature there IS no perfectly reflecting surface (FTIR is about the closest thing we have). Every time an EM wave impinges on any real surface, energy is lost to heat. With any source of light beamed at any surface, light will be turned into heat. In fact, MOST of the energy is immediately turned to heat. By the laws of thermodynamics, even that that is not converted instantaneously into heat, but goes into some other form of energy, will eventually turn up as heat. You pay now, or you pay later, but you always pay the entropy tax.

    (From: Bill Vareka (

    And, something else to ponder:

    If you combine light in a beamsplitter there is a unavoidable phase relation between the light leaving one port and the light leaving the other.

    So, if you have a perfect Mach-Zehnder interferometer like the following

                +-------+      BS          M
                | Laser |=====>[\]---------\
                +-------+       |          |         M = Mirror
                                |          |        BS = Beamsplitter
                                |       BS |
                              M \---------[\]---->A
    If you set it up so that there is total cancellation out of, say, port A, then Port B will have constructive interference and the intensity coming out port B will equal the combined intensity coming in the two input ports of that final beamsplitter. This is due to the phase relation between the light which is reflected at the beamsplitter. That which is reflected and goes out port A will be 180 degrees out of phase with that which is reflected and goes out port B. The transmitted part of port A and port B are the same. Hence the strict phase relationship between the light from the two output ports. This is an unavoidable result of the time-reversal symmetry of the propagation of light.

    (From: A. Nowatzyk (

    A beam-splitter (say a half silvered mirror) is fundamentally a 4 port device. Say you direct the laser at a 45 degree angle at an ideal, 50% transparent mirror. Half of the light passes through straight, the rest is reflected at a 90 degree angle. However, the same would happen if you beam the light from the other side, which is the other input port here. If you reverse the direction of light (as long as you stay within the bounds of linear optics, the direction of light can always be reversed), you will see that light entering either output branch will come out 50/50 on the two input ports. An optical beam-splitter is the same as a directional coupler in the RF or microwave realm. Upon close inspection, you will find that the two beams of a beam-splitter are actually 90deg. out of phase, just like in an 1:1 directional RF coupler.

    In an experiment where you split a laser beam in two with one splitter and then combine the two beams with another splitter, all light will either come out from one of the two ports of the second splitter, depending on the phase. It is called a Mach-Zehnder interferometer.

    Ideal beam-splitters do not absorb any energy, whatever light enters will come out one of the two output ports.

    Interference between E/M Radiation of Different Wavelengths

    We all know that light from a single coherent source can create interference patterns and such. What about arbitrary uncorrelated sources?

    There will be interference but you won't see any visible patterns unless the two sources are phase locked to each-other since even the tiny differences in wavelength between supposedly identical lasers (HeNe, for example) translate into beat frequencies of MHz or GHz!

    (From: Charles Bloom (

    The short answer is yes.

    Let's just do the math. For a wave-number k (2pi over wavelength), ordinary interference from two point-like apertures goes like:

    Psi = (e^(ik(L+a).) + e^(ik(L-a).))/2
        = e^(ikL) * cos(ka)
    I = Psi^* Psi = cos^2(ka)
    (a is actually like (x-d)^2/L where 2d is the slit separation, and x is the position along the screen; L is the distance from the center of the slits to our point on the screen).

    Now for different wavenumbers:

    Psi = ( e^(ik(L+a).)+ e^(iK(L-a).))/2
    I = Psi^* Psi = 1/2 [ 1 + Re{ e^(i ( k(L+a) - K(L-a) ).)} ]
    	      = 1/2 [ 1 + cos( L(k-K) + a(k+K) ) ]
    	      = cos^2[ 1/2( L(k-K) + a(k+K) ) ]
    This is almost a nice interference pattern as we vary 'a', but we've got some nasty L dependence, and in the regime L >> a where our approximations are valid, the L dependence will dominate the a dependence (unless (k-K) is very small; in particular, we'll get interference roughly when a(k+K) ~ 10 and L(k-K) ~ 1 , and L >> a , which implies |k-K| << |k+K| , nearly equal wavelengths.)

    The L dependence is the usual phenomenon of "beats" which is also a type of interference, but not the nice "fringes" we get with equal wavelengths (the L dependence is like a Michelson-Morely experiment to compare wavelengths of light, by varying L (the distance between the screen and the sources) I can count the frequency of light and dark flashes to determine k-K.

    What about Hobbyist Interferometry?

    Building something that demonstrates the principles of interferometry may not be all *that* difficult (see the comments below). However, constructing a useful interferometer based measurement system is likely to be another matter.

    So you would like to add a precision measurement system to that CNC machining center you picked up at a garage sale or rewrite the servo tracks on all your dead hard drives. :) If you have looked at Agilent's products - megabucks (well 10s of K dollars at least), it isn't surprising that doing this may be a bit of a challenge. As noted in the section: Basics of Interferometry and Interferometers, a high quality (and expensive) frequency stabilized single mode HeNe laser is often used. For home use without one of these, a short HeNe laser with a short random polarized tube (e.g., 5 or 6 inches) will probably be better than a high power long one because it's possible only 2 longitudinal modes will be active and they will be orthogonally polarized with stable orientation fixed by the slight birefringence in the mirror coatings. As the tube heats up, the polarization will go back and forth between the two orientations but should remain constant for a fair amount of time after the tube warms up and stabilizes. Also see the section: Inexpensive Home-Built Frequency or Intensity Stabilized HeNe Laser.

    The problem with cheap laser diodes is that most have a coherence length that is in the few mm range - not the several cm or meters needed for many applications (but see the section: Can I Use the Pickup from a CD Player or CDROM Drive for Interferometry?). There may be exceptions (see the section: Interferometers Using Inexpensive Laser Diodes) and apparently the newer shorter wavelength (e.g., 640 to 650 nm) laser pointers are much better than the older ones but I don't know that you can count on finding inexpensive long coherence length laser diodes. Even if you find that a common laser diode has adequate beam quality when you test it, the required stability with changes in temperature and use isn't likely to be there.

    The detectors, front-end electronics, and processing, needed for an interferometer based measurement system are non-trivial but aren't likely to be the major stumbling block both technically and with respect to cost. But the laser, optics, and mounts could easily drive your cost way up. And, while it may be possible to use that $10 HeNe laser tube, by the time you get done stabilizing it, the effort and expense may be considerable.

    Note that bits and pieces of commercial interferometric measurings systems like those from HP do show up on eBay and other auction sites from time to time as well as from laser surplus dealers. The average selling prices are far below original list but complete guaranteed functional systems or rare.

    (From: Randy Johnson (

    I'm an amateur telescope maker and optician and interferometry is a technique and method that can be used to quantify error in the quality of a wavefront. The methods used vary but essentially the task becomes one of reflecting a monochromatic light source, (one that is supplied from narrow spectral band source i.e., laser light) off of, or transmitting the light through a reference element, having the reference wavefront meet the wavefront from the test element and then observing the interference pattern (fringes) that are formed. Nice straight, unwavering fringe patterns indicate a matched surface quality, curved patterns indicate a variation from the reference element. By plotting the variation and feeding the plot into wavefront analysis software (i.e., E -Z Fringe by Peter Ceravolo and Doug George), one can assign a wavefront rating to the optic under test.

    The simplest interference test would involve two similar optical surfaces in contact with each other, shining a monocromatic light source off the two and observing the faint fringe pattern that forms. This is known as a Newton contact interferometer and the fringe pattern that forms is known as Newton's rings or Newton's fringes, named for its discoverer, you guessed it, Sir Issac Newton. If you would like to demonstrate the principle for yourself, try a couple of pieces of ordinary plate glass in contact with each other, placed under a fluorescent light. Though not perfectly monochromatic, if you observe carefully you should be able to observe a fringe pattern.

    Non-contact interferometry is much tougher as it involves the need to get a concentrated amount of monochromatic light through or reflected off of the reference, positioning it so it can be reflected off of the test piece, and then positioning the eye or imaging device so that the fringe pattern can be observed, all this while remaining perfectly still, for the slightest vibration will render the fringe pattern useless.

    (From: Bill Sloman (

    An interferometer is a high precision and expensive beast ($50,000?). You use a carefully stabilized mono-mode laser to launch a beam of light into a cavity defined by a fixed beamsplitter and a moving mirror. As the length of the cavity changes, the round-trip length changes from an integral number of wavelengths of light - giving you constructive interference and plenty of light - to a half integral number of wavelengths - giving you destructive interference and no light.

    This fluctuation in your light output is the measured signal. Practical systems produce two frequency-modulated outputs in quadrature, and let you resolve the length of a cavity to about 10 nm while the length is changing at a couple of meters per second. The precision is high enough that you have to correct for the changes in speed of light in air caused by the changes temperature and pressure in an air-conditioned laboratory.

    Hewlett-Packard invented the modern interferometer. When I was last involved with interferometers, Zygo was busy trying to grab a chunk of the market from them with what looked liked a technically superior product. Both manufacturers offered good applications literature.

    (From: Mark Kinsler (

    You can get interferometer kits from several scientific supply houses. They are not theoretically difficult to build since they consist mostly of about five mirrors and a lens or two. But it's not so easy to get them to work right since they measure distances in terms of wavelengths of light, and that's *real* sensitive. You can't just build one on a table and have it work right. One possible source is: Central Scientific Company.

    (From: Bill Wainwright (

    Yes, you can build one on a table top. I have done it. I was told it could not be done but tried it anyway. The info I read said you should have an isolation table to get rid of vibrations I did not, and even used modeling clay to hold the mirrors. The main problem I had was that the image was very dark and I think I will use a beamsplitter in place of one of the mirrors next time. The setup I had was so sensitive that lightly placing your finger on the table top would make the fringes just fly. To be accurate you need to take into account barometric presure and humidity.

    (From: Sam.)

    And check out my range of interferometer kits on eBay under user ID: siliconsam. Sorry for the plug. ;-)

    Interferometers Using Inexpensive Laser Diodes

    The party line has tended to be that the coherence length of diode lasers is too short for interferometry or holography. (See the sections beginning with: General Interferometers.) While I was aware of CD laser optics being used with varying degrees of success for relatively short range interferometry (a few mm or cm - see the section: Can I Use the Pickup from a CD Player or CDROM Drive for Interferometry?), the comments below are the first I have seen to suggest that performance using some common laser diodes may be at least on par with that of a system based on a typical HeNe laser (though not a high quality and expensive frequency stabilized single mode HeNe laser).

    While I don't know how to select a laser diode to guarantee an adequate coherence length, it certainly must be a single spatial (transverse) mode type which is usually the case for lower power diodes but those above 50 to 100 mW are generally multimode. So, forget about trying to using a 1 W laser diode of any wavelength for interferometry or holography. However, single spatial mode doesn't guarantee that the diode operates with a single longitudinal mode or has the needed stability for these applications. And, any particular diode may operate with the desired mode structure only over a range of current/output power and/or when maintained within a particular temperature range.

    (From: Steve Rogers (

    I have been involved with laser diodes for the last 15 years or so. My first was a pulsed (only ones available at that time) monster that peaked 35 watts at 2 kHz with 40 A pulses! It was a happy day when they could operate CW and visible to say the least. Anyway, in the course of my working travels, I have built numerous Twymann-Green double pass interferometers for the wave front distortion analysis of laser rods, i.e., Nd:Yag, Ruby, Alexandrite, etc. The standard reference light source for this instrument has always been the 632.8 nm HeNe laser. Good coherence length and relatively stable frequency was its strong suit.

    When visible diode lasers came out I often wondered aloud about their suitability as a replacement for the HeNe. I despise HeNe lasers. They are bulky and I have been shocked too many times from their power supplies.

    I assumed that since CD player laser diodes at 780 nm could have coherence lengths on the order of tens of centimeters or into the meters (!!, see, for example: Katherine Creath, "Interferometric Investigation of a Diode Laser Source", Applied Optics (24 1-May-1985) pp. 1291-1293), Visible Laser Diodes (VLDs) could make excellent replacements. As it turned out, VLDs tend to have coherence lengths which are considerably shorter according to the latest technical literature and I held off on experimenting with them. Last week, I went through my shop and found enough mirrors, beamsplitter, assorted optics to throw together my own double-pass interferometer for home use. This coincided with my acquisition of a 635 nm 5 mw diode module - a good one from Laserex.

    To make a longer story shorter, I assembled said equipment with the VLD and WOW! excellent fringe contrast (a test cavity of four inches using a .250" x 4.0" Nd:Yag rod as the test sample.) When a HeNe laser was substituted for the VLD, virtually no difference in the manual calculation of wave front distortion (WFD) and fringe curvature/fringe spacing. The only drawback with the VLD is that it produces a rectangular output beam. When collimated you have a LARGE rectangular beam rather than a nice round HeNe style beam. My interferometer now occupies a space of 10" x 10" and is fully self contained. It probably could even be made smaller. Not only that, but it runs on less than 3 V!!!

    I am just as surprised as you are with the results that I achieved. This is one reason why it took me so long to attempt this experiment (something like 4 to 5 years). I have always assumed that a HeNe laser would be FAR superior in this configuration than a VLD would be. Perhaps others may know more about the physics than I do. One thing is certain, these are "single mode" index guided laser diodes and typically exhibit the classic gaussian intensity distribution which is not so evident with the "gain guided" diodes. This in turn implies a predominant lasing mode which in turn would imply a (somewhat) stable frequency output. Purists would note that this VLD has a nominal wavelength of 635 nm +/- 10 nm while the HeNe laser is pretty much fixed at 632.8 nm. This variable could account for extremely minor WFD differences.

    (From: W. Letendre (

    There's an outfit in Israel selling a diode based laser interferometer enough cheaper than Zeeman split HeNe units to suggest that they are using a laser diode in the 'CD player' class, or perhaps a little better. They are able to measure, 'single pass' (retro rather than plane mirror) over lengths of up to about 0.5 m, suggesting that as an upper limit for coherence length.

    Can I Use the Optical Pickup from a CD/DVD Player or CD/DVDROM for Interferometry?

    With the nice precision optics, electromechanical actuators, laser diode, and photodiode array present in the mass produced pickup of a CD/DVD player, CD/DVDROM drive, or other optical disc/k drive, one would think that alternative uses could be found for this assembly after it has served for many years performing its intended functions - or perhaps, much earlier, depending on your relative priorities. :-) (Also see the section: Using a CD or DVD Optical Pickup in a Precision Position or Angle Encoder.

    People sometimes ask about using the focused laser beam for for scanning or interferometry. This requires among other things convincing the logic in the CD/DVD player or CD/DVDROM drive to turn the laser on and leave it on despite the possible inability to focus, track, or read data. The alternative is to remove the optical pickup entirely and drive it externally.

    If you keep the pickup installed in the CD player (or other equipment), what you want to do isn't going to be easy since the microcontroller will probably abort operation and turn off the laser based on a failure of the focus as well as inability to return valid data after some period of time.

    However, you may be able to cheat:

    Where such a feature is not provided:

    CAUTION: Take care around the lens since the laser will be on even when there is no disc in place and its beam is essentially invisible. See the section: Diode Laser Safety before attempting to power a naked CD player or simlar device.

    It may be easier to just remove the pickup entirely and drive it directly. Of course you need to provide a proper laser diode power supply to avoid damaging it. See the chapter: Diode Laser Power Supplies for details. You will then have to provide the focus and/or tracking servo front-end electronics (if you need to process their signals or drive their actuators) but these should not be that complex.

    Some people have used intact CD player, CDROM, and other optical disc/k drive pickup assemblies to construct short range interferometers. While they have had some success, the 'instruments' constructed in this manner have proven to be noisy and finicky. I suspect this is due more to the construction of the optical block which doesn't usually take great care in suppressing stray and unwanted reflections (which may not matter that much for the original optical pickup application but can be very significant for interferometry) rather than a fundamental limitation with the coherence length or other properties of the diode laser light source itself as is generally assumed.

    In any case, some of the components from the optical block of that dead CD/DVD player may be useful even if you will be substituting a nice HeNe laser for the original laser diode in your experiments. Although CD optics are optimized for the IR wavelength (generally 780 nm), parts like lenses, diffraction grating (if present and should you need it), and the photodiode array, will work fine for visible light. However, the mirrors and beamsplitter (if present) may not be much better than pieces of clear glass! (DVDs lasers are 635 to 650 nm red, so the optics will be fine in any case.)

    Unfortunately, everything in a modern pickup is quite small and may be a bit a challenge to extract from the optical block should this be required since they are usually glued in place.

    If what you want is basic distance measurements, see the section: Using a CD or DVD Optical Pickup for Distance Measurements which discusses the use of the existing focusing mechanism for this purpose - which could be a considerably simpler approach.

    Also see the section: Basics of Interferometry and Interferometers.

  • Back to Laser Instruments and Applications Sub-Table of Contents.

    Interferometers for Precision Measurement in Metrology Applications

    Interferometer-based techniques are used in all types of systems requiring precision measurement of position, velocity, angle, straightness, and many other parameters. These are part of a class of what are called "metrology" applications. Examples include semiconductor wafer steppers in photolithography systems, hard drive and CD/DVD/Blu-ray disc mastering, optical diamound turning and other high performance CNC machines, general machine tool calibration, and many more. Measurements can be made over 10s of meters with resolution down to nanometers using the wavelength of a known wavelength of laser light as the meter (or yard) stick. Before discussing systems using two-frequency lasers, we need to back up.

    Note that all the techniques being discussed are for measuring displacement (or position change), not absolute position. Absolute measurements using interferometers are possible using lasers but require additional techniques that are beyond the scope of this discussion.

    There are two classes of measurement interferometers called "homodyne" and "heterodyne". They have much in common including the general configuration and use of similar or identical optics. However, the lasers and detection electronics differ substantially and each method has it benefits and drawbacks. Most, if not all, utilize optical configurations that are variations on the Michelson interferometer. See Basic Michelson Interferometer. In short, a laser beam is split into two parts which are bounced off of a pair of reflectors and recombined at a detector. Any change in the relative path lengths of the two "arms" formed by the reflectors results in a phase shift between the waves in the two beams, which can be measured and converted to displacement (change in position) down to nanometer precision.

    These techniques generally require a high quality specular reflector like a planar mirror, cube-corner (trihedral prism, also called a retro-reflector), "cat's eye" lens system, or something similar that returns a beam with a high signal-to-noise ratio. They will not work with diffuse reflectors or multi-level reflectors in the beam or where the beam may move over abrupt changes in the position of the reflective surface relative to the interferometer. At least not without much more effort. So forget about measuring plant growth unless it is possible to hang a cube corner on one of the stems. ;-)

    Interferometers using Single Frequency Lasers

    In its simplest form, the homodyne approach compares the phase of a reference beam (returned from a fixed reflector) and a measurement beam (returned from the target whose position is to be measured) directly to compute displacement (change in position). All these systems are generally based on a variation of the Michelson interferometer, though other types can be used depending on the specific needs. The light source is most often a stabilized single frequency (also known as a Single Longitudinal Mode or SLM) HeNe laser. This is a well developed technology dating back over 50 years which still cannot be matched in an affordable way. The single frequency laser assures that coherence is maintained over a large change in the Path Length Difference (PLD) between the reference and measurement beams. In principle this can be 10s or even 100s of meters or more, though in practice it is generally much much less. For applications where the change in PLD is small - typically a few cm - a less expensive unstabilized Multi-Longitudinal Mode (MLM) HeNe laser can be used.

    A typical configuration is shown in Interferometer Using Single Frequency HeNe Laser. The laser can be any of the (single) frequency stabilized HeNe lasers described in the chapter: Commercial Stabilized HeNe Lasers. The linearly polarized beam must be oriented at a 45 degree angle with respect to the Polarizing Beam-Splitter (PBS). Half of it then gets polarized horizontally (into the plane of the diagram) by the PBS and is returned from the retroreflector of the "Fixed Arm" as the reference beam (REF) while the other half gets polarized vertically passing through the PBS and is returned from the retroreflector of the "Test Arm" as the measurement beam (MEAS). They are recombined in the PBS as a single beam that has two components whose relative phase depends on the relative position of the two retroreflectors, and this changes as the Test Arm is moved. Imagine a pair of sinusoida combs moving with respect to each-other. Some fraction of the combined beam goes to an "Intensity" photodiode that produces an output proportional to the beam power. This is needed to keep track of the actual signal level. The remainder is split into two parts which go to separate photodiodes (PDs), one of which has its phase shifted by 90 degrees to generate sin and cos signals in quadrature (offset from each other by 90 degrees). These are sufficient to determine displacement (consisting of both distance change and direction) using digital hardware only slightly more complex than a common up-down counter. This is the same type of hardware used with optical encoders based on parallel lines or gratings, but with the interferometer approach, using the wavelength of light itself. Interferometer Setup using Teletrac Laser and Plane Mirror Interferometer shows the components of a rig I put together to test the use of a homodyne interferometer with the µMD1 readout. It consists of a Teletrac 150 laser with a built-in optical receiver, a Teletrac Plane Mirror Interferometer on a precision micrometer linear stage, and Atmega 328 Nano 3.0 interface. Oh, and sorry, the laser is pointing the wrong way in the pic (lasers should output to the right to be happy) but that's so I could reach the knob. ;-) Specific information can be found in the section: Teletrac Model 150 Stabilized HeNe Laser 4 and the section before it for the optical receiver.

    One benefit of the homodyne approach is simplicity and low cost (at least in a relative sort of way as none of these systems is exactly inexpensive!). The laser can be built very inexpensively (despite what it probably costs the end-user!). And for some applications, the performance is more than adequate. Another benefit that is often overlooked is that the native resolution (without fancy interpolation) of the homodyne interferometer is four times better (i.e., 1/4 the increment) than for heterodyne. For the Plane Mirror Interferometer (PMI), that's ~40 nm compared to ~158 nm. And slew rate is limited by the electronics/data processing, not by the laser as it is with heterodyne. With the common The basic measurement processing is little more than what keeps track of the position of a computer mouse. Companies offering homodyne systems tout the benefits of their approaches including the ability to perform interpolation to achieve higher resolution, which nearly all use to increase the effective resolution to below 1 nm. A nice introduction can be found in the Motion X MX Interferometer Manual.

    Many types of interferometer optics can be used with homodyne systems in addition to the "Linear Interferometer" in the example above. In fact, they are identical to those for the heterodyne interferometers discussed below. So see the section: Optics for Interferometers Using Two-Frequency Lasers.

    However, there are several deficiencies that make these systems undesirable (or at least much more difficult to implement reliably) for more sophisticated applications. Since they are comparing the phases of the REF and MEAS beams directly, the result at any given time is a DC level that depends not only on the relative phase, but also on the actual output power of the laser and optical losses elsewhere in the system, drift in the electronics, and even very slight changes in optical alignment. But the signal processing does tend to be simpler and unlike the two-frequency systems, the only upper limit on velocity is one of optical detection and processing speed, not the value of the "split" frequency of the laser. (More on this below.)

    Electronics for Interferometers Using Single-Frequency Lasers

    To compute a simple displacement from a homodyne interferometer, a quadrature decoder is needed to both count fringes and generate direction - essentially similar to the function of a the quad-A-B output of a linear or rotary encoder. And for basic measurements, it doesn't need to be much more complex than this, at least in principle.

    Quadrature Decoders for Homodyine Interferometers

    Many approaches can be used to generate displacement signals in homodyne interferometers. A search on the Internet will return dozens of complex convoluted schemes that may be appropriate for research projects, but not those that are used most often in commercial systems - which are usually quite simple. The goal of a quadrature decoder in a homodyne interferometer is accept the combined REF and MEAS otherogonally linearly polarized components from the interferometer optics and to generate a pair of electrical signals shifted by 90 degrees with respect to each-other to be used to compute the position and direction of motion of the "Tool". Interferometer Using Single Frequency HeNe Laser shows the overall organization with one specific type of quad sin/cos decoder. The same type of quadrature signals are used in common optical rotary or linear encoders but here they depend on the interference of light. For metrology applications, one of the optical components is generally fixed and called "REF" (reference) while the other is the return from the Tool and called "MEAS" (measurement). As noted above, REF and MEAS are linearly polarized and orthogonal, aligned vertically and horizontally, though which is which depends on the specific implementation. The only effect of swapping them (with everything else fixed) is to flip the sign of the motion direction. And for differential measurements, REF will also be changing. The same decoders may also be used in other applications like Ring Laser Gyros (RLGs) where both the CCW (counterclockwise) and CW (clockwise) beams are orthogonally polarized. Then REF and MEAS are replaced with CCW and CW. However, for most RLGs, both beams have the same polarization. So one will need to have its polarization rotated by 90 degrees using a Half WavePlate (HWP) at 45 degrees before merging in the NPBS.

    Basic Homodyne Laser Interferometer Quadrature Decoders shows several possible schemes. These all depend on using a Quarter WavePlate (QWP) to shift the phase of the horizontal and vertical polarized components so that the resulting signals differ by 90 degrees in phase. Several of these schemes have been found in commercial homodyne optical receivers as noted. All should produce similar results except possibly for the sign of the phase shift between A and B (+90 or -90 degrees). Types 1 and 2 are essentially equivalent so there's no reason not to use Type 1 as it's simpler.

    Type 3 has the benefit that the QWP and LP at 45 degrees are exactly what are present in LCD screen contrast enhancing circular polarizing sheets. These are really inexpensive and simplify assembly by not requiring a separate and possibly delicate and/or expensive QWP. Enough CP material to construct several dozen of these decoders is under $3 on eBay. ;-) The alignment even works out so that they can be cut up into tiny rectangles all aligned with the edges. While arbitrary adjustment of the phase is a bit more limited with the Type 3 decoder, there will probably be enough range to get to 90 degrees by fine tuning the angle of the CP with the piece of QWP facing the beam before glueing. If not, a piece of almost any clear household plastic - or even the protective film that comes with the CP sheets - can be used in front of the B photodiode to adjust it. In fact, both PDs can have LPs (only) at 45 degrees glued to their faces and a plastic sheet oriented in front of one to provide a precise 90 degree phase shift and perfact quadrature behavior.

    In short, it's almost trivial to do this, but then there would be no way to get those convoluted schemes published. ;-)

    Whenever the orthogonal REF and MEAS aligned with the X and Y axes are passed through a QWP at 45 degrees, the result is counter-rotating field vectors so that the relative phase can be tuned by adjusting the angle of the LP and thus the precise phase shift between Channels A and B can be fine tuned. This would be of importance if taking advantage of all the state changes to multiply the basic number of counts by a factor of 4, and then to analyze the actual waveforms for interpolation to achieve even higher resolution.

    In the diagram, the Intensity channel generally present in these systems is NOT shown since it would be identical for all of them with just a non-polarizing beam-splitter providing a fraction of the total power to a photodiode. The Intensity signal is needed to keep track of the actual signal level to be used in interpolation calculations and to compensate for the normal decline in laser power with use and effects of alignment and/or contamination of the interferometer optics. Also, the precise sign of the angles of the QWPs in the diagram may not agree with the actual implementation since it's not straightforward to determine these from simple tests. And there could also be (offsetting) differences among specific models of systems from the same manufacturer. In other words, your mileage may vary and adjustments may be required to get these to work as desired in an experimental implementation. ;-)

    In attempting to visualize what's going on, think of REF as a fixed sinusoid with MEAS being similar but shifting with respect to REF as the Tool moves. Then pick the case where they are both in phase and thus identical coming from the interferometer optics figure out what the QWP(s) and LP(s) to do them. Fancy calculations are not needed for an understanding of what's going on. Or if it's easier to visualize of a pair of sinusoidal combs, that works as well. :-)

    There are two types of approaches shown in the diagram:

    One assumption with all of these is that the NPBS does not alter the relative phase of the the optical components (either linear or circular) due to the behavior of its dielectric coating. Being able to compensate for this could be a benefit of Type 2, which places dual QWPs after the NPBS to add additional degrees of freedom in adjustment. So there will be cost trade-offs when selecting among these implementations.

    Types 1 and 2 are functionally equivalent. Compared to the third scheme, they can have up to twice the signal amplitude for the same laser power, as well as more flexibility in setting the precise phase shift by simply adjusting the relative orientation of the linear polarizers. So this could be why all the commercial implementations I've analyzed are designed like this.

    But Type 3 has the advantage of being easier to understand without visualizing rotating E/M field vectors. ;-) So it was used for my RLG out of expediency since a nearly complete beam sampler assembly was available to repurpose with an NPBS, LPs, and PDs, and a sliver of a QWP just had to be slipped inside of it. ;-) See the section: Sam's Home-Built Ring Laser Gyro 1.

    Where there is plenty of laser power and a QWP providing a precise 90 degree phase shift (or where this accuracy doesn't matter), there's no particular advantage to one scheme over the other. And there may be other variations that produce similar signals. However, a Web search for "homodyne decoder" or the like had generally proven pretty useless as what mostly turned up were much more complex obscure convoluted schemes, probably from post-graduate theses or esoteric research projects, that may or may not even be useful for their intended purpose, let alone for basic interferometry. Now that this document other others of mine describing quadrature decoders are on the Web, these searches have a better chance of being useful. ;-)

    Two Simple Quad-Sin-Cos Decoders and Scope Display of Reflector Movement shows implementations of the Type 1 and Type 3 versions using a variable attenuator as the NPBS. The reflector is mounted on a mini-woofer driven with a sine-wave from a function generator. The opposite phase shift determined by the direction of movement is clearly visible on my "Continuum Laser Zapped Scope" but a close examination shows that it is slightly more than 90 degrees for the Type 3 decoder. The phase shift can be fine-tuned for the Type 1/2 detectors by rotating either LP while monitoring the detected signals using the scope's X-Y mode. The resulting Lissajous display should be an ellipse with its principal axes aligned with X and Y or a perfect circle if the X and Y sensitivities are adjusted so the sizes are equal. A phase shift of other than 90 degrees will produce an ellipse with its principal axes at an angle. Direction will be indicated by clockwise or counter-clockwise motion of the spot on the scope. Lissajous Display of Quadrature Decoder Signals for Oscillatory Motion - 90 Degrees shows the display when the relative phase is near optimal and Lissajous Display of Quadrature Decoder Signals for Oscillatory Motion - Less than 90 Degrees shows when it's misadjusted.

    The simplest way to put these together with bits of the CP sheet uses the adhesive already present on the QWP-side that is to be stuck to something like the PD. For the LP-side where there is no adhesive, 5 minute Epoxy or UV-cure index matching optical cement can be used. The UV-cure stuff ends up being less messy but more expensive. I use Norland 65 UV-cure cement from Thorlabs and a $1 1W 365 nm LED to cure it. No need for the $2,583 UV cure gizmo they sell (or much higher cost for the same thing that dentists buy). ;-) Similar UV-cure cement is also available for replacing smart phone screens and is less expensive.

    And one laser manufacturer's Power Point tutorial on homodyne interferometers calls the QWP a "Special Optic". This I assume is to protect their valuable intellectual property from grand theft, as though basic optical components are somehow company proprietary, and they assume their intended audience is too stupid to figure this out. :) Every first-year physics student should know in their sleep that the "Special Optic" is a QWP. ;-)

    Micro Measurement Display 2 (µMD2)

    µMD2 is an inexpensive system for precision readout of displacement (change in position), angle, straightness, and more in metrology applications. It is designed and optimized for HeNe laser (homodyne) interferometry as well as devices like linear and rotary encoders with Quad-A/B or up/down pulsed signals, but also supports heterodyne applications. It is based on a Teensy 4.0 microcontroller along with a hand-full of other components mounted on the custom SG-µMD2 PCB. Virtually everything is built-in as shown in Typical Homodyne Interferometer Setup using µMD2.

    A version that can be replicated based on this approach is now available. The total parts cost for the display electronics excluding the PC should be well under $50. The Teensy firmware and Windows Graphical User Interface (GUI) software are available free for non-commercial or research applications. Laser and interferometer not included. :-) (The GUI is the same one used for µMD1, a similar system for heterodyne interferometers using two frequency lasers.)


    And sorry, the "1" and "2" are logically swapped because the system for two frequency lasers was developed first, live with it. ;-)

    Interferometers using Two-Frequency Lasers

    The interferometers described in the previous sections and found in physics labs (assuming such topics are even taught with hands-on experience!) all use CW lasers and look at the fringe shifts as the relative path length of the two arms is changed. While this works in principle and has been used widely, modern commercial measurement systems based on interferometry often use more sophisticated techniques to reduce susceptibility to signal amplitude changes and noise, and improve measurement accuracy, stability, and convenience. These are called "heterodyne" systems in which the laser beams are in essense carriers for a lower "split" frequency in the MHz range provided by the two-frequency laser. The split frequency is detected optically, but then can be manipulated using straightforward electronics totally in the AC domain. If you're totally confused by now, never fear. There is much more below. ;)

    The microchips in virtually all modern electronics (including the CPU and memory inside the PC, MAC, tablet, or Smarkphone you're reading this on) were likely produced on photolithography systems incorporating wafer steppers using two-frequency interferometers for multiple axes of ultra-precise motion control. Based on a scientifically proven metric - the availability of used equipment on eBay :-), heterodyne systems are in much wider use than homodyne systems, by at least an order of magnitude.

    Interferometer-based measurements systems typically use some type of low power stabilized helium-neon laser to produce the "yardstick" beam of light. By stabilizing the laser with reference to the neon gain curve, the accuracy of the optical frqeuency/wavelength can easily be known to better than +/-0.1 ppm (parts per million). As noted above, a basic system may use such a laser in a Michelson or similar interferometer, with a quadrature (sin/cos) detector to count fringes representing changes in path length as described above. Problems with such a system are that changes in light intensity will result in measurement errors, alignment is very critical to obtain adequate fringe contrast, and they are more susceptible to noise.

    In two-frequency interferometers, a special stabilized HeNe laser is used that produces a beam consisting of two very slightly different frequencies (and corresponding wavelengths) of light simultaneously. This may be achieved by various techniques. HP/Agilent lasers employ a special tube which uses a magnet to perform Zeeman splitting resulting in useful difference frequencies being limited to around 4 MHz due to the Physics. Zygo uses an external acousto-optic modulator to produce a 20 MHz split frequency. As above, both types of lasers are locked in such a way that the optical frequency is very precisely known. A higher split frequency is desirable because it ultimately limits the maximum stage slew rate. But too high a split frequency and subsequent processing for measurement or control becomes complex.

    A diagram of the general approach is shown in Interferometer Using Two Frequency HeNe Laser.

    U.S. Patent #3,656,853: Interferometer System outlines the overall approach in dry patent legaleze. :) Being a patent, it doesn't really apply directly to any real system, not even the original HP-5500A system. And in this case, doesn't even appear to have one in mind. What's below is more reader-friendly.

    The following description applies to the HP/Agilent implementation using Zeeman splitting to create the two frequencies. With Zygo, the method of generating the them differs, but their use in the interferometer is the same.

    In the Zeeman split approach, the two-frequency laser consists of a HeNe laser tube surrounded by permanent magnets which produce a constant axial magnetic field. The laser tube is short enough that without a magnetic field, only a single longitudinal mode will normally oscillate if it is located near the center of the neon gain curve. (Those on either side will not see enough gain.) The net result of the magnetic field is that instead of a single longitudinal mode, two modes are produced that differ very slightly in frequency and have right and left circular polarization. The difference between the two frequencies is typically in the 1.5 to 4 MHz range (though some go up to 6 MHz or more), which makes the resulting signals extremely easy to process electronically. The actual difference frequency is determined by the strength of the magnetic field, length of the internal laser cavity, and other physical details, as well as the exact place on the Zeeman-split neon gain curve where the laser has been locked.

    To stabilize the laser, there is a piezo element and/or heater to precisely adjust cavity length. A feedback control system is used to adjust the cavity length to maintain the position of the Zeeman-split frequencies - and thus the wavelengths - constant. The feedback is generally based on the simple approach of forcing the orthogonally polarized outputs to be equal, which results in the most stable optical frequency.

    The wavelength of the laser is the measurement increment ("yardstick") and will remain essentially unchanged for the life of the instrument. For example, with the doppler broadened gain curve for the HeNe laser being about 1.5 GHz FWHM (1 part in about 300,000 with respect to the 474 THz optical frequency at 633 nm) and a 1 percent accuracy within the gain curve, the absolute wavelength accuracy will then be better than 1 part in 30 million! Not too shabby for what is basically a very simple system. In practice it's even better. :) The laser tube is not much different than the type that was used by the 100s of thousands in grocery store barcode scanners in the 1980s.

    Note that the exact value of the difference frequency does not need to be very precisely controlled over the long term. Rather, it is the difference between the reference difference frequency and the measurement difference frequency that matters, and the latter only depends on the motion of the target reflector - and the speed of light. Thus, the exact beat frequency of each laser need not be precisely controlled or even precisely measured and recorded or used anywhere in the calculations.

    Since the output of the laser is a beam consisting of a pair of circularly polarized components, a Quarter-Wave Plate (QWP) and Half-Wave Pate (HWP) are used to separate these into two orthogonal linearly polarized components called F1 and F2, and to orient them such that they are parallel to the horizontal or vertical axes.

    The beam consisting of F1 and F2 is split into two parts with a non-polarizing beam-splitter: One part goes through a polarizer at 45 degrees (to recover a signal with both F1 and F2 linearly polarized in the same direction) to a photodiode which generates a local copy of the reference frequency (REF, the difference between F1 and F2) for the measurement electronics; the second is the measurement beam which exits the laser. The return beam is called MEAS.

    The purpose of the remainder of the interferometer is essentially to measure the path length change between two points called displacement. In a typical installation, the beam consisting of F1 and F2 is sent through a polarizing beamsplitter. F1 goes to a cube-corner (retro-reflector) on the tool whose position is being measured and F2 goes to a cube-corner fixed with respect to the beamsplitter and laser. However, differential measurements could be made as well using F2 in some other manner. Various "widgets" are available for making measurements of rotary position, monitoring multi-axis machine tools, etc. But they all ultimately result in the same sort of change in basic displacement readings - but perhaps with different scale factors or more complex calculations.

    The return from the object corner reflector is F1+ΔF1 which is recombined with F2 and sent to an "optical receiver" module - a photodiode behind a polarizer at 45 degrees and preamp which generates a new difference frequency, F2-(F1+ΔF1). This signal, called "MEAS" is compared with REF to produce an output which is then simply ΔF1. The "Signal Processing" block might be implemented with digital logic like counters and subtractors, a fast microprocessor, or combination of the two. A change in the position of the object by 316 nm (1/2 the laser wavelength) results in ΔF1 going through a whole cycle. By simply keeping track of the number of complete cycles of ΔF1, this provides measurements of object position down to a resolution of a few hundred nm with an accuracy of +/-0.02 ppm! And the typical implementation will either multiply the REF and MEAS frequencies by 16 or 32 or more using a pair of phase-locked loops, or perform interpolation using sub-cycle phase compison of the REF and MEAS signals to produce a corresponding improvement in resolution down to a few nanometers or better!

    The primary disadvantage of heterodyne systems is that the maximum velocity is limited in the direction that would reduce MEAS since going through 0 Hz would be confusing at best. So, one of the key specifications for these lasers is the (minimum) split frequency. For example, the HP-5517B has a split frequency range of 1.9 to 2.4 MHz with typical samples being 2.20 MHz. But the minimum is the critical value and for 1.9 MHz, the maximum velocity will be around 0.5 m/s using the simplest (linear) interferometer. Zygo lasers have a 20 MHz split frequency so the velocity can be over 10 times higher. But even 0.5 m/s is adequate for many motion control systems.

    More information on the two-frequency HeNe laser can be found in the sections: Hewlett-Packard/Agilent Stabilized HeNe Lasers and Two Frequency HeNe Lasers Based on Zeeman Splitting. Searching on the Agilent Web site will yield product specific information and application notes on two frequency interferometers. A comprehesive but not too hairy description of the two frequency approach can be found in the Hewlett-Packard`Journal, February, 1976. Yes, this is an old technique (actually much older)! Searching at HP Archive for "Interferometer" and similar terms will turn up many more interesting articles. For an excellent introduction written by Agilent insiders :), see "A Tutorial on Laser Interferometry for Precision Measurements", Russell Longhridge and Daniel Y. Abramovitch, 2013 American Control Conference (ACC), Washington, DC, June 17-19, 2013. While this is an IEEE conference paper, an on-line version may be found by using the search string: a tutorial on laser interferometry for precision measurements agilent.

    Zygo, another manufacturer of interferometer measurement systems using two-frequency lasers had an excellent tutorial called "A Primer on Displacement Measuring Interferometers" but it seems to have disappeared from their Web site. But never fear, I archived it at Sam's Copy of Zygo's Primer on Displacement Measuring Interferometers. And, an even more extensive introduction can now be found in the PowerPoint presentation at Introduction to Displacement Measuring Interferometry and Sam's Copy of Introduction to Displacement Measuring Interferometry. (The actual title found by Google is the same as the previous one, "A Primer on Displacement Measuring Interferometers".)

    And if you find info on a laser claiming a 618 to 658 MHz difference frequency, run the other way. ;-) As of 2019, REO (Research Electro-Optics) is offering directly and on Newport's Web site what they dub "OPIS" or "Orthogonally Polarized Interferometer System" which appears to be largely MARKETING HYPE. OPIS appears to be simply a bog-standard (for any other manufacturer) 3 mW 633 nm unstabilized random polarized HeNe laser head with power supply. But the list price is about twice that of an almost identical 3 mW laser without the OPIS label. This is no doubt partly due to the MARKETING HYPE, but also because REO has to use a more complex expensive tube to do what comes naturally to other companies. More info may be found in the section: Research Electro-Optics Stabilized HeNe Lasers. I am not aware of any practical use for such a laser that would make it advantageous for heterodyne interferometry. The primary reason is that it is not stabilized, as all metrology lasers must be to have decent wavelength accuracy. It would be 50 to 100 times worse than a typical HP/Agilent/Keysight laser. And at times, 3 longitudinal modes may be present, and due to mode pulling, the difference frequencies between them will vary slightly representing a further complication. And performing the digital processing obtain the displacement information at 600+ MHz would be a serious challenge even if a custom chip could be designed, and probably impossible or at least highly impractical without one. Ironically, REO's stabilized HeNe (having different Marketing hype) would be suitable if (1) the polarizer that normally blocks the second mode at its output were removed and (2) the custom chip could be designed. ;-)

    Lasers for Interferometers Using Two-Frequency Lasers

    There are generally low power HeNe lasers with either specially designed (and expensive) laser tubes or an external acouto-optic modulator (also expensive) to produce the two (relatively) closely spaced optical frequencies with orthogonal polarization. Depending on technique, the difference frequency can be anywhere from a few hundred kHz to 20 MHz or more. Since the beat frequency between the reference and measurement signals decreases with one direction of motion and can't go below 0 Hz (or at least becomes confusing as it passes through 0 Hz), a higher difference frequency translates into higher maximum speed of position change in the measurement system. Therefore, depending on the specific application, a higher difference frequency may be essential.

    Optics for Interferometers Using Two-Frequency Lasers

    The following discusses the various types of optical components, mostly those supplied by Hewlett Packard (then Agilent, and now Keysight) for measurement of position (or more accurately, displacement or change of position) or velocity (rate of change of position). There are also other optical configurations for measurement of angle, flatness, straightness, squareness, and more. But in essense, all of these convert a change in the measurement parameter into a change in position. So, the basic principles of operation are the same. Optics from other companies like Excel Precision and Zygo are similar.

    CAUTION: In most instances these optics are quite robust (or at least as robust as more common high quality optical components) and can be cleaned if necessary without problems. However, it seems that under some conditions, the AR-coatings can, as they say, come unglued. :( :) Whether this is simply due to a bad coating run, or some environmental factor wherever they had been used, cleaning - even while taking great care and using accepted practices for precision optics - can result in the AR coatings simply disappearing. Although the lack of an AR coating doesn't usually have much impact on performance, it is unsightly.

    While the description below deals with "AC" or "heterodyne" systems using a two-frequency laser, the same optical configurations are also applicable to "DC" or "homodyne" systems using a single-frequency laser. Aside from the type of laser, the optical receivers (and subsequent processing) will also differ. Teletrac (now Axsys) was one manufacturer of these generally lower performance (and lower cost) systems.

    The most basic application (for a single axis measurement) will consist of the following optical components:

    Since the optical frequency/wavelength is being used as the "measuring stick" in these systems, it must be known to a high degree of precision and anything that affects it must also be taken into account. In particular, the temperature, pressure, and humidity of the air must be factored into the measurement calculations. Or, if part or all of the measurement setup is in a vacuum, this will affect it. These corrections can be done at least partially automatically with sensors, or by manually entering parameters into the measurement computer. See Refractive Index of Air and Wavelength Correction Calculator (NIST).

    These systems generally allow a single laser to be used with installations where the motion of multiple access needs to be measured. So in addition to the actual measurement optics, there will be components to split and redirect the original beam from the two-frequency laser to each axis.


    The heart of all of these systems are the interferometers. The three most configurations used most often are shown in Most Common Hewlett Packard/Agilent Interferometers. (This diagram applies directly to two-frequency lasers like the 5517 where F1 (the lower optical frequency) is oriented horizontally. Where F2 (the higher optical frequency) is oriented horizontaally as with the 5501A/B, simply swap F1 and F2 in the diagram.) There can be various permutations of the individual components that are optically and functionally equivalent. Combinations of multiple interferometers mounted on a common platform are also available for compact multi-axis applications. What HP calls the "interferometer" consists of all the components in the center of each diagram - the Polarizing Beam Splitter (PBS), 1 or 2 Retro Reflectors (RRs), and 0, 1, or 2 Quarter-Wave Plates (QWPs), and 0 or 1 Plane Mirror (PM). There will also be a RR or PM on the "tool" whose position is to be measured, and an Optical Receiver (OR) for the return beam. The Two-Frequency Laser (TWL) can be shared among all the axes of the machine by distributing its beam using non-polarizing beam splitters and 45 degree mirrors ("Beam Benders", but all beam orientations must be a multiple of 90 degrees to the original TFL).

    As noted, all of these interferometers contain a high quality Polarizing Beam Splitter (PBS) cube as their central component. (In principle, a non-polarizing beam-splitter could be used instead as in a traditional Michelson interferometer. However, there would be a loss of 50 percent or more in efficiency and a reflected beam would be directed back into the laser, which could make it unhappy. The common interferometer configurations return nearly 100 percent of the power of the laser back to the optical receiver, only limited by optics losses. And there is virtually no light directed back toward the laser.) What gets added on to the PBS depends on the specific type and may include Retro-Reflectors (RRs, which are solid cube-corners), and/or Quarter-Wave Plates (QWPs), and/or a Plane Mirror (PM). Please refer to the diagram, above. For HP-5517 lasers, F1 is the lower frequency component and is horizontally polarized, while F2 is the higher frequency component and is vertically polarized. (For reasons not known to anyone on this planet, HP-5501 lasers are the opposite, but the only effect is a sign change in the measurement calculation.)

    The Agilent and (and likely Zygo) prices are also interesting in that they are at least 5 times greater than what similar optical components would cost from a supplier like Newport. I do not know how much - if any - of this is due to the required optical quality compared to less demanding applications. Most of the cost is likely due to the specilized precision mounting and the relatively low production volume of interferometer optics. Fortunately for hobbyists and experimenters, the common HP/Agilent interferometers are readily available surplus, often at very reasonable prices. And non-HP/Agilent optics should be just fine for non-critical applications.

    See Links to Agilent Laser and Optics User's Manual for general information on the current Agilent lasers and interferometer optics (including the more exotic configurations for angle, flatness, straightness, and others). For much more on these systems, go to Keysight Technologies and search for "Interferometers".

    It should be noted that the exact same optical configurations are used with single frequency (homodyne) interferometers. With those, f1=f2, usually produced by orienting the linearly polarized input beam at 45 degrees or using a laser whose output is circularly polarized.

    One might ask why some obvious simple configurations are never used. While all the commercial interferometers are really variations on the basic Michelson design, none use it precisely. There are at least two fundamental reasons why that would be undesirable. First, when a Michelson interferometer is perfectly aligned, a portion of the outgoing laser beam is reflected directly back into the laser's aperture. Most lasers including HeNes are prone to being destabilized by such back-reflections. In addition, using a non-polarizing beam-splitter with the output of a two-frequency laser is very inefficient, with the usable signal at the optical receiver cut by 75% or more. However, by using a polarizing beam-splitter (PBS), both of these deficiencies are nearly totally eliminated. Except for coating losses, a PBS results in virtually no back-reflections and nearly 100 percent efficiency. (Strictly speaking, there needs to be a linear polarizer at 45 degrees in front of the optical receiver which cuts the power by 50 percent, but that's only once; for multipass configurations like the PMI there would be huge additional losses without the polarization optics.) That's why all the commercial interferometers have a PBS cube as their central component. But they also all include one or more precision retro-reflectors (RRs, cube corners) which are both bulky and expensive. There is good reason for this in many - possibly most - applications, but it is essential? What about something along the lines of No Retro-Reflector Interferometer (NRRI) which uses plane mirrors in both the reference arm and measurement arm - there are no retro-reflectors anywhere. It would have the same resolution as the SBI (and LI) since there is only a single pass to the remote plane mirror (unlike the normal PMI with two passes). (The NRRI configuration should not be confused with the HP/Agilent C01-10705A conversion, which modifies the SBI by removing a QWP and adding a RR to turn it into a configuration similar to that of the 10706A PMI, though a non-polarizing beam splitter is required to obtain the OR signal.) The NRRI would be even more compact than the normal 10705A SBI, as well as potentially less expensive using a plane mirror in the reference arm instead of a RR. This arrangement is perfectly acceptable theoretically. However, it suffers in practice by requiring much more precise alignment. Having RRs provides some automagical compensation for alignment changes which is probably why every commercial interferometer configuration has an RR internally in the reference path. A small change in orientation of an RR by an angle Θ results in at most a very small translation in the return beam where the beams are offset in the RR, and essentially none for a centered beam. However, a similar change in orientation of a plane mirror resuilts in a change in the angle of the return beam by 2*Θ and a position shift of approximately D*2*Θ. Even a very small change in the parallelism of the F1/F2 beams at the optical receiver will affect the signal quality.

    As a test, a NRRI was constructed by removing the reference retro-reflector from an HP-10705A SBI and installing an aluminized first surface mirror attached to a plate, with 4 screws and split washers to permit fine alignment. This is shown in HP-10705A SBI Modified to be No Retro-Reflector Interferometer. The NRRI replaced the HP-10706A PMI in my setup for testing HP/Agilent/Keysight lasers. See Diagram of Two-Frequency Interferometer Laser Tester with the optical receiver repositioned since the return beams are in a different location. The NRRI was attached to an adapter plate which also provided some fine pan/tilt adjustment via 4 screws and an O-ring between it and the base. Initial alignment was done to assure that the beams from the NRRI to the optical receiver were coincident (and thus parallel) at the NRRI and 1 meter away. Then, fine tuning was performed while watching an oscilloscope display of REF and MEAS to obtain a stable MEAS signal with the remote mirror in motion. And it does work, though the change in displacement over which alignment could be remained with this rig was less than 1 mm. However, the mirror on the voice coil (loudspeaker) actuator is known not to be mounted perfectly perpendicular to its axis of motion or to remain perfectly parallel over its range of travel. And indeed, translating the entire speaker assembly on its linear stage maintained alignment and a stable signal over its full range (greater than 25 mm). But with the 10706A, there are no problems using the voice coil positioner. Thus the conclusion is that a configuration like the NRRI is usable as long as alignment can be maintained. However, a retro-reflector could be used in place of the remote plane mirror, greatly reducing issues with respect to Tool alignment, which is the most critical especially over a long range. The reference arm of the interferometer is usually very stable, so it could still use a plane mirror in place of the retro-reflector.

    Cube-Corner Trihedral Prism Retroreflectors and Polarization

    A Cube-Corner (CC) has the property that over a wide range of input angles, an incident beam is returned parallel to the original with at most a translation in position. Reflection of an image is also preserved but with a mirror flip due to the three reflections inside the prism. The CC has 3 planar surfaces at right angles to each-other and may be fabricated as 3 exposed front-surface mirrors, or the internal surfaces of a trihedral prism. The prism is by far the more common implementation due to physical stability, ease of fabrication, and immunity from contamination. CCs are often colloquially called "corner cubes" but this doesn't even make sense! Cube corner is the perfect description since they are effectively the corner (or apex) of a cube that is sliced perpendicular to the long diagonal.

    CCs are used in laser interferometer-based measuring systems as Retro-Reflectors (RRs) for the reference and/or measurement paths. In these applications, the polarization state of the beam is usually critical and must be maintained as close as possible through the optical path to maximize signal level and minimize cross-talk. (Retroreflectors are also widely used in surveying and other applications where polarization isn't a major consideration.)

    The first specification of interest for a CC (besides size) is that of angular precision of the facets, which is usually defined in arc seconds (arcsec, 1/3,600th of a degree). Typical values range from 1 to 10 arcsec or more. For reference, 1 arcsec is approximately 1 part in 5x10-6 based on sin(1 deg/3,600). That would be for a single reflective surface. Since there are 3 surfaces and the change in reflected angle is double the change in incident angle, the angular change in a reflected beam would be up to 6 times the arcsec error. So, for a typical commercial-grade CC with a specification of 3 arcsec, the pointing error could approach 1 part in 10-4. At 10 meters, this would be a shift in the return beam of up to around 1 mm assuming the reflected beam doesn't straddle a vertex. If it does, there could be other issues. The above analysis is back of the envelope and among other things, ignores the change in angle at the glass surface, so there are very likely significant errors. :( :)

    How polarization is affected when a beam passes through a CC is also critical for interferometer applications. Not all CCs are suitable due to changes in polarization from the multiple reflections. There is a whole body of technical literature devoted to the effects of reflection on polarization. But the quick answer appears to be that even though one might think that Total Internal Reflection (TIR) is occurring inside the prism, the reflective facets must be metal-coated to minimize the effects on polarization. The reflections are really from the metal coating, not the glass in contact with it. Even very high quality uncoated CCs result in poor and variable performance in a simple test setup consisting of a linearly polarized laser, CC under test, linear polarizer, and detector/screen. More below.

    Thorlabs has an informal slide show describing the effects of a single sample of an uncoated (both front and back) retroreflector on polarized light. See Thorlabs Retroreflectors Lab Facts or Polarization Change After Propagation Through a Prism Retroreflector (which is linked from there). My results with uncoated retroreflectors are consistent with their presentation.

    Basic tests were performed by reflecting a linearly polarized HeNe from the RR under test, back through a linear polarizer oriented to block the orthogonal orientation, and to a white screen. Tests were done with the input polarizatoin vertical, horizontal, and at roughly 45 degrees. With no change in polarization, the result would be perfect extinction. The optimal orientation of the polarizer for minimum transmission, as well as the approximate beam intensity at minimum, was observed at multiple orientations of the retroreflector ratated on a V-block.

    With an uncoated retroreflector, the polarizer orientation for a minimum transmitted beam varied widely and the minimum transmission would at times be quite large indicating a conversion to elliptical polarization as indicated in the Thorlabs presentation.

    However, with metal-coated back faces, the behavior is fundamentally different and far superior where maintaining the polarization state is critical. With a linearly polarized input, the reflected beam's polarization is rotated very slightly, but it remains almost perfectly linearly polarized as indicated by very little transmission through the polarizer. The orientation for minimum transmission tends to flip between a positive angle and a negative angle depending on which combination of facets are in the reflected path, with that angle estimated to be around 10 degrees for all the metal-coated samples tested.

    The behavior was similar for inexpensive Far East imported copper-coated prisms and those from HP/Agilent (likely silver-coated). The HP/Agilent prisms had AR-coated front surfaces while the others did not, but this only affected the losses, not the polarization. Uncoated prisms were mediocre with respect to changes in the polarization, though for unknown reasons, a Far East import appeared to be better (have less effect on polarization) than a similar retroreflector from a very expensive USA-made instrument.

    Another benefit of a coating even if polarization is of no concern is that it prevents fingerprints or other contamination on the back surfaces from affecting performance. And it's way too easy to mess up the back surfaces with grubby (or even perfectly clean) fingers during normal handling! :-) Also, any air-born vapors that condense there will affect the reflections.

    The reflection efficiency is a result of losses from two traversals of the front surface and three internal reflections from the back surfaces. The following data are based on tests of several prisms at 633 nm:

                                           Effect on    Reflection
        Back Surface      Front Surface   Polarization  Efficiency
        Uncoated (1)        AR-Coated        Large         93%
        HP 10703A (2)       AR-Coated        Small         90%
      Agilent 10703A (2)    AR-Coated        Small         89%
        Silver Coated        Uncoated        Small         86%
        Copper Coated        Uncoated        Small         79%
      Aluminum Coated (3)    Uncoated        Small         55%


    1. The prism with the uncoated rear surfaces has the highest efficiency because it relies on true total internal reflection. In fact, the mystery is why it isn't even better. However, one possibility is that since this came from an instrument that was used at IR wavelengths, the AR coating may not be that great at 633 nm.

    2. Next to the uncoated prism, the HP and Agilent 10703As are best and the difference between them and the silver-coated one can be explained by the lower losses from their AR-coated front surface.

    3. It's not clear why aluminum is so poor. While aluminum does have a slightly lower reflectance than copper and silver, that can't be the entire explanation. It's not *that* bad. This particular prism was also much larger but that shouldn't have made a noticeable difference.

    Interferometers Using Two-Frequency Lasers with Non-Orthogonal Axes

    In virtually all common applications, the laser, interferometers, and remote reflectors can be arranged to be at orthogonal orientations. In fact, HP/Agilent/Keysight specifically states that all angles must be multiples of 90 degrees. But suppose it is desired to control motion where this is not the case? For example, controlling a hexapod or adjusting a multi-segment telescope mirror. Of course a separate laser could be used for each axis but that seems unwieldy and possibly just a bit silly (not to mention expensive). The important thing to keep in mind is that what's important is that the f1/f2 polarizations as seen by the interferometer need to be orthogonal. However, the overall geometry does not have this restriction.

    To redirect a beam to an arbitrary orientation and position in three dimensions requires at most 2 bounces if there are no space constraints. One bounce is required to direct the beam so it intersects the optical axis normal to the remote reflector and a second bounce to direct it to the remote reflector. Depending on the specific geometry, additional bounces may be required to minimize the size of the setup. The beam then passes through a suitable interferometer oriented so that it is aligned with the polarization (f1/f2) axes of the beam or a Half WavePlate (HWP) may be used to rotate them to align with the interferometer. The return beam from the remote reflector can be sent directly to an optical receiver aligned with the interferometer or passed through a linear polarizer at 45 degrees and sent to a remove optical receiver via an optical fiber.

    One complication is the effects multiple reflections will have on the polarization. For metal-coated mirrors (most often protected aluminum or protected silver), there should be no change. But for dielectric mirrors, the reflection coefficient depends on the Angle of Incidence (AoI). For total internal reflection from inside a prism for example, even metal coatings affect the polarization, though much less than with non-metallic coating or no coating at all. For normal incidence (AoI=0), there is no change. But as the AoI increases, the ratio of the S and P polarizations will change. So for example, linear polarization may be converted to elliptical polarization. While even at a 45 degree AoI the change is not large, it is still better to minimize the AoI as this will reduce the orthogonality of the f1/f2 components. One option is to rotate the polarization using a HWP before each reflection so the S and P oriented to avoid these changes. Then the only effect will be to slightly change the ratio of f1 to f2, which only results in lower AC signal strength.

    It's all done with mirrors and HWPs. ;-)

    Electronics for Interferometers Using Two-Frequency Lasers

    The most basic requirement is to convert the phase shift between REF and MEAS to a count or position. Conceptually, this is nearly trivial, being simply the different between the total cycles of the REF and MEAS signals. And in fact, the original HP-5505A display unit did this by brute force with a pair of counters and a subtractor. As will be shown below, this turns out to be a clean, but rather hardware intensive solution. In general, care needs to be taken in the design of the processing hardware and/or software to avoid possible errors ultimately resulting from the Uncertainty Principle.

    Possible approaches

    1. Dual counters with subtractor: As noted above, this was the original method used by HP in the 5505A display, which went with the 5500C laser. To handle a +/-1 meter range (without resolution extension) requires approximately +/-1.6 million counts corresponding to 21 bits plus sign, or 6-1/2 digits plus sign. This is a fair amount of hardware is implemented with discrete CMOS or TTL, but should fit nicely in a modern FPGA.

      With 2's complement or 9's complement arithmetic, as long as the difference remains less than half the maximum (i.e., the sign doesn't flip), the result should be correct for a subtract. Since both counters can be triggered directly from REF and MEAS (with suitable input filtering and limiting, there are no arbitration or sampling issues with the counters. However, readout of the subtracted result must be done with care to avoid the chance of catching the result during a transition. One approach would be to read twice in rapid succession and only accept the value if the results match. In addition, to avoid single-count oscillation even when nothing is moving, the read should be referenced to either REF or MEAS, but not done with a totally independent clock, and/or successive values should be compared to previous values and only updated when there are two successive counts in the same direction.

    2. Digitally sampled REF and MEAS drive up/down counter: This is the simplest for a strictly hardware-based solution. However, just generating Up and Down clock pulses from the REF and MEAS signals can have potential problems where they occur very close together. Commercial up/down counter chips are not designed to produce unambiguous results if the two clocks occur so close together as to violate their setup and hold specifications. Sampling and synchronization to a system clock is required. But whenever an independent signal is sampled by a periodic clock in something like a D flip-flip, there is also no way to guarantee it will satisfy the setup and hold time requirements of the device. There will be a (hopefully) small window where the flip-flop can enter a metastable state where Q and ~Q are equal and not recover for an arbitrary unbounded amount of time which may as long as the sampling period. This is contrary to what many logic designers assume. In fact, there has been considerable research published in scholarly journals on this topic! It's theoretically impossible to eliminate this potential problem entirely, though with careful design, the probability can be made so low as to be of no concern. Which devices are more susceptible to metastable behavior is not something found in the datasheets. For example, when the common 74xx74 D flip-flip was tested, it was found that not only was the behavior dependent on the type (e.g., LS, AS, F, HC, etc.), but on the particular manufacturer as well! The issue is that should the device enter the metastable state, its output may only slowly recover, and when used as an input to subsequent logic, not meet their setup or hold requirements! An error that occurs once in a billion samples may not seem of consequence, but we're talking about measurements that may be made over hours with clocks running at MHz rates.

      And, as with the first approach, some hardware or software should be included to eliminate the single-count oscillation issue.

      I built a pulse converter that is based on this approach. It includes single-ended inputs for REF and MEAS, the provision for up to 4 stages of shift register to minimize the probability of metastability occurring, and single-count oscillation elimination. It even has a pair of monostables driving LEDs so I can watch when even single Up and Down pulses occur! It's truly amazing how sensitive a system that measures in micrometers is to any vibration!

    3. Digital Signal Processor based: With a suitably fast programmable system, input conditioning, sampling, counting and other arithmetic can be performed in firmware providing both a hardware-efficient design as well as much more flexibility. With modern technology, 100 or 200 MIPS - or faster - DSPs are inexpensive and should be able to handle the required tasks and many more without working too hard. :)

    4. High performance microcontroller: The plummeting cost of really fast versatile single chip computers like the Microchip PIC32 family makes developing a basic measurement display almost trivial. The simplest implementation could use a pair of internal counters clocked by REF and MEAS, extended in software to at least 32 bits if necessary. Simply taking their difference with an appropriate scale factor will provide displacement. Angle, straightness, squareness, and other variations simply require different scale factors and units readout. (And, of course, suitable interferometer configurations!) Computing the change using a real-time clock provides velocity. The actual display can be a touch-screen LCD or via USB to a host PC. Since there is so much free documentation and development support available along with inexpensive hardware for microcontroller-based systems, this is probably the ideal approach for the hobbyist and experimenter or even a "real" application. :)

    Counting cycles of the phase difference is fine for education and demos and will provide resolution of a fraction of a wavelength, but real systems almost always implement some type of resolution enhancement scheme like frequency multiplication using a Phase-Locked Loop (PLL) or ultra-high speed sampling. The spec sheets can then claim a resolution that is not simply a fraction of the wavelength of the HeNe laser, but down to a few nanometers or even better, the order of magnitude of the diameter of a hydrogen atom. :) It is not known to what extent these spectacular specs are realizable and repeatable in practice over the life of the system taking into consideration the normal decline in laser power, accumulated dust on the optics, changes in alignment, and other factors which result in increased noise in the optical signal.

    Of course, this isn't exactly rocket science. Besides HP/Agilent, Zygo, Excel, and the other major players, these things have been stuffed into LSI ICs for a long time. One example of an almost single chip solution is from Laser Metric Systems, Inc.. They have a much more limited line of metrology systems but do have a PC ISA or PCI bus card (PRM-004 series) whose brains is a single Altera FLEX FPGA that can handle the performance requirements of just about any two-frequency laserinterferometer with a claimed resolution down to 0.1 nm. Well, maybe. :) See the product info on their Web site.

    And check out this (open access) paper: "FPGA-Based Smart Sensor for Online Displacement Measurements Using a Heterodyne Interferometer", Sensors 2011, 11, 7710-7723. They digitize the analog REF and MEAS signals in dual 14 bit 20 MHz flash A/Ds. So, in addition to the basic computation taking the difference of (wrap-corrected) accumulators for REF and MEAS, they use the analog waveforms to refine the measurement and estimate the actual phase difference (partial wavelength) claiming a resolution of 3.4 nm over a range of 3 m.

    Sam's Measurement Displays for Two-Frequency Interferometers

    I was tired of searching for something inexpensive on eBay and anyhow, wanted something interesting to do. I also miss the days of 0s and 1s a bit (no pun...) and my drawer of ancient TTL chips was getting rather lonely. :)

    I decided on the approach using the up/down counter due to its simplicity. The schematic for testing except for the up/down counter is shown in Sam's Pulse Converter for Two-Frequency Interferometer. It consists of:

    While the laser is warming up and there is no REF signal, only the green LED is lit. Once REF appears but if the MEAS beam is misaligned or blocked, only the RED LED is lit. Once everything is stable and aligned, neither is lit if there is no movement. I also breadboarded a version using a PAL to generate the control signals. This eliminates most of the discrete logic chips.

    Before I added the display, I had to be content watching the green and red LEDs. Any vibration - including a moderately loud radio - result in flickering. I'm not sure what type of music they prefer, but a sustained tone can result in quite impressive activity. :-)

    The multi-digit up/down counter ICs I had found tended to be marginal with respect to their maximum frequency count capability. So I decided to go with the separate chip "brute force" approach for the first version. Most of the control logic is in a single 22V10 PLD. This does use proper differential line receivers (UA9637 or UA9639) to be compatible with virtually any laser. The prototype shown in Photo of Sam's Measurement Display 1 in Action has been tested and works as expected. In fact, it accepts signal levels and REF/MEAS frequencies that would choke the 5508A. I always have it running in parallel with the 5508A in my interferometer test rig since it will work with the newest high-REF Agilent/Keysight 5517E/F/G lasers that the 5508A completely ignores.

    But I have NOT included a schematic, PCB layout, or parts list for SGMD1 only because it is NOT how I would recommend building a measurement display in the 21st Century (or even later part of the 20th Century), if for no other reason than to remain sane in terms of soldering 1,000+ pins. :( :) However, if someone did insist, I could provide those. Perhaps even a mostly complete set of parts (including blank PCB) to make one (1) copy.

    Having said all that, as noted, the one and only original TTL-based system is still connected in parallel with the HP-5508A (and sometimes µMD1 or µMD2 as well) because the up/down LEDs can provide useful information where signal quality is marginal that would be lost when the more sophisticated systems simply display "Error". :( :)

    Micro Measurement Display 1 (µMD1)

    The next version of a home-built measurement display is based on a microChip PIC32 microprocessor, originally part of the Digilant chipKIT DP32, a when that ceased production, the custom SG-µMD1 PCB, which is functionally equivalent. Virtually everything is built-in as shown in Typical Heterodyne Interferometer Setup using µMD1. The only other part required is a dual line receiver IC for REF and MEAS. (It may be possible to even forgo this but the isolation makes it virtually impossible for the signal wiring to damage the hardware.) The chipKIT DP32 is under $25 and can readily handle REF and MEAS frequencies well beyond 4 MHz, suitable for any of the common HP/Agilent lasers. Just add an obsolete Windows PC (almost anything with a USB port) to complete the user interface. A version that can be replicated based on this approach is now available. The total parts cost for the display electronics excluding the PC should be well under $50. The DP32 firmware and Windows Graphical User Interface (GUI) software are available free for non-commercial or research applications. Laser and interferometer not included. :-)


    With changes to the front-end hardware and firmware, the same µMD1 GUI may be used with homodyne interferometers. For rudimentary applications, the microcontroller can be as simple as a $3 Atmega 328 Nano 3.0 board. The addition of a quadrature-to-count logic to the chipKit DP32 would provide excellent performance. However, with the existance now of µMD2 optimized for homodyne interferometers - which for all practical purposes has no upper limit on slew rate - doing would be silly except as an experiment. :-) (See the info on homodyne interferometers, above.) And µMD2 also supports heterodyne interferometers, but without the interpolation (at least for the foreseeable future).

    Measuring Very Small Displacement Changes

    During the development of the interpolation firmware for µMD1 (see the previous section), detecting movement on the nanometer scale was required. The stability of my normal test setup was totally inadequate. Any disturbance - even while tapping on my computer table two floors up while monitoring the display using Remote Desktop resulted in detectable fluctuations in the displacement. And vibrations from the fan in the laptop near the interferometer swamped any changes that were being measured.

    Thus, it was necessary to attach a PZT with a plane mirror directly to the measurement arm of the HP 10706A interferometer as shown in Mirror on PZT attached to HP 10706A Plane Mirror Interferometer. The PZT is a $2 beeper from Digikey and requires only a few volts to move over a micron. A Wavetek function generator can be set to provide anywhere from below 1 nm to several microns of movement p-p depending on whether the 2 or 20 V output is selected, and the setting of the variable output level control. While not totally immune to external vibrations, it is several orders of magnitude less sensitive. In fact, the environmental noise floor is now so low that a 10 nm p-p displacement waveform comes out clean.

    Using a combination of interpolation and averaging, detecting changes well under 1 nm is now possible. For reference, 1 nm is about the width of 8 hydrogen atoms sitting patiently side-by-side. ;-)

    Sources of Measurement Error in HP/Agilent Metrology Systems

    While it sounds really impressive to be basing precision measurement on the wavelength of light, and HP/Agilent lists the nominal wavelength of the 5501 and 5517 lasers to 9 significant figures, there are many environmental and installation factors that impact what is actually useful. The following is the briefest of summaries of accuracy issues. A large portion of the operation manuals for these systems is devoted to this topic. More than you could ever hope to know can be found in the links in the section: Agilent Laser and Optics User's Manual. The following is a brief summary.

    Laser wavelength accuracy

    A common question that comes up with respect to these systems is: "Since these are based on two-frequency lasers, which frequency is used as the wavelength specification?". The quick answer is: It's not clear. :)

    A phase change of 360 degrees of the difference frequency of the measurement beam with respect to the reference beam represents a change in position of of the moving part of the measurement system (e.g., the "tool") by 1/2 wavelength (linear or single beam interferometer) or 1/4 wavelength (plane mirror interferometer). Thus the component (F1 or F2) that goes to the tool is the actual "yardstick" wavelength. If both F1 and F2 are used in a differential measurement, then each contributes to the measurement based on its wavelength. So, strictly speaking, one should use those wavelengths in the calculation. But as a practical matter, it really doesn't matter as the difference in frequency between the two components F1 and F2 is so small compared to the optical frequency, that the error introduced by using one or the other is way below the accuracy specification for even the military calibrated versions of these systems.

    It's not clear (at least to me) where the value of 632.991372 nm for the 5501B and 5517A/B or 632.991354 nm for the 5517C/D comes from. My assumption would be that it's the theoretical lasing line center of the Zeeman-split neon gain curve. Various sources list other slightly different values for HeNe lasers. Wikipedia has a page on the Meter Measuring Unit that gives a value of λHeNe=632.99139822 nm. And an HP Journal article on the 5528A gives yet another value of 632.991393 nm for the 5518A, but that tube should be essentially identical to the tube in the 5517A! The actual fill pressure, ratio of He:Ne and their isotopic mix, temperature, and other factors will affect the exact wavelength. But why the different values even for essentially similar lasers from HP/Agilent including lasers that are still in production? The corresponding difference in optical frequency between 632.991372 nm and 632.991354 nm is about 13.5 MHz, so it's way beyond the error due to whether line center or either F1 or F2 is used. Perhaps, it has something to do with the newer tubes being filled with pure Ne20 or Ne22 and the older ones having a mixture to guarantee compatibility in legacy applications. How's that for a wild guess? :) The difference is still under 0.03 ppm, so it should generally not be a huge issue in any case.

    The normal HP/Agilent lasers have a long term stability specification of +/-0.1 ppm. The laser stabilization depends on the condition of the electronics and may result in a small variation in this wavelength. Even one power cycle to the next does not result in a precise return to the exact same conditions due to the particular temperature, and thus cavity mode number, at which lock occurs. But any variation will only be a few MHz at most, well below the +/-0.1 ppm specification.

    The Agilent info indicates that they will certify a laser to MIL STD-45662 for long term stability of +/-0.02 ppm. MIL STD-45662 requires traceability to a (wavelength/frequency) reference, which I would assume to be something like an iodine-stabilized laser. However, this doesn't sound like the laser design is necessarily any different, just that the optical frequency of the specific laser is measured precisely, and the exact value to even more significant digits than the those given above, is included in the calibration report. With that number known, +/-0.02 ppm corresponding to about +/-9.4 MHz, really shouldn't be hard to maintain.

    Velocity of light

    Ambient temperature, air pressure, and humidity all have a very significant effect on the measurement. Approximately a 1 part-per-million change in the velocity of light and thus measurement wavelength will result from:

    Since 1 ppm is 10X of the basic measurement specification for accuracy, it is clear that these factors must be taken into account. The HP/Agilent measurement systems have sensor options to allow this to be done automatically, or the corrections can be entered manually.

    And, if the "tool" is in a vacuum, that's roughly a 3 part in 10,000 error due to the difference in the index of refraction of a vacuum compared to air!

    See Refractive Index of Air and Wavelength Correction Calculator (NIST).

    Material effects

    Depending on the construction of the equipment on which the interferometer optics are mounted, the change due to thermal expansion and other effects can be very significant resulting in serious errors if not taken into consideration.

    Determining the Exact Laser Wavelength or Frequency

    So, how accurately can the wavelength or frequency of one of these lasers be determined outside of a NIST calibration laboratory or Agilent test facility? It all really comes down to having a standard of reference. Commercial instruments called wavelength meters available from companies like TOPTICA Photonics AG can have very high accuracy, some down to 2 MHz (0.0000027 nm) or better. But that is only if a more accurate reference is used for calibration not to long before making measurements of the unknown laser. 2 MHz isn't very much when it comes to optical frequency! So, even if you could afford one of these expensive instruments or could borrow one, it would still need to be calibrated against some reference!

    The most common reference of relevance for this testing would be a mode stabilized HeNe laser like a Spectra-Physics 117A. Its frequency will have a long term stability of 10 MHz or less, but whether the absolute accuracy can be nailed down to better than 50 or 100 MHz - about 0.1 to 0.2 parts-per-million (ppm) - is questionable due to factors like the exact neon isotope(s) in the gas fill and the precise point on the neon gain curve where the laser is locked. So, a higher precision reference would be needed to calibrate that!

    However, some of the HP/Agilent lasers themselves may have their frequency known to extremely high accuracy. For the normal commercial versions of the 5517 lasers, the spec'd accuracy is *only* +/-0.1 ppm, which is equivalent to +/-0.0000633 nm (+/-47.4 MHz in optical frequency) with a nominal wavelength of 632.991354 nm (for the 5517C/D; 632.991372 nm for the 5501B and 5517A/B and some others even though the tube is virtually identical in all of these lasers).

    But there are some 5517 lasers that come with a "pedigree" - a report including the wavelength of the specific laser measured to very high accuracy under controlled environmental conditions.

    The report for one particular 5517D laser included the following: Temperature and humidity test conditions of 22.8 °C and 36.2%, respectively, a locked output power of 409 uW (spec is 180 uW minimum), a split (REF) frequency of 2.75 MHz (spec'd range is 2.4 to 3.0 MHz), and the actual vacuum wavelength of 632.9913662 nm (nominal is 632.9913540 nm).

    That difference of 0.0000122 nm is actually a rather large error, equivalent to about 9.4 Mhz in optical frequency or about 0.02 ppm.

    I don't know if it is even remotely possible to obtain information for a specific serial number laser from Agilent unless you're the original buyer, and I rather doubt it. After all, your good fortune on eBay is hardly something that Agilent is likely to care much about! :) But, it may be possible to obtain the information from the original buyer if they can be tracked down. I was able to have access to one of these lasers that I was repairing, and they provided the above data from the calibration report. With this data, I was hoping to be able to measure the exact wavelength/frequence of one of my HP/Agilent lasers and then use it as the reference when I return the other one. However, it turned out that the wavelength report didn't apply to this particular sample but a similar one. Oh well.

    However, age and use of the laser will affect the optical frequency by enough to matter. So unless the laser is relatively new, any data may be unreliable. The drift may be predominantly due to changes in the He and Ne partial pressure. These both alter the width of the Doppler broadened Ne split gain curves, and shift the center optical frequency. End-of-life lasers - the type often found on eBay - would be most prone to such effects. But even relatively young lasers will see a drop in pressure that may be significant. One tip-off to a potential discrepancy could be the REF frequency. From my observations and the comment by an engineer who tests these lasers, the REF frequency tends to increase slightly with use. This is consistent with a broadening of the gain curves which will increase the mode pulling effect. (But, at least the change is in the beneficial direction - the one that increases the maximum measurement velocity specification, assuming the processing electronics can handle the higher REF frequency!)

    One reference is: "Frequency stability measurements on polarization-stabilized He-Ne lasers", T. M. Niebauer, James E. Faller, H. M. Godwin, John L. Hall, and R. L. Barger, Applied Optics, vol. 27, no. 7, 1 April 1988, pp. 1285-1289.

    However, assuming you found a new-in-box Agilent laser with complete documentation of optical frquency:

    Some relevant numbers: 633 nm is about 474 THz based on the speed of light of 299,792,458 m/s. So, 1 nm is 0.749 THz or 749 GHz and 0.0000001 nm is 74.9 kHz. 100 kHz is 0.000000134 nm or 1 MHz is 0.00000134 nm. The difference between 632.9913540 nm and 632.9913662 nm is then around: 9.137 MHz.

    Comparing the Optical Frequencies

    With one of these lasers in hand, it's really quite straightforward at least in principle to determine the exact frequency offset of another similar laser by beating (heterodyning) the two outputs in a high speed photodiode and measuring the difference frequency. Since frequency doesn't depend on environmental conditions, nothing that happens outside the laser will affect it. The resulting optical frequency value can then be divided into the speed of light in the relevant medium (e.g., air at STP or vacuum) to compute the exact wavelength.

    My Two-Frequency Interferometer Laser Tester (See Diagram of Two-Frequency Interferometer Laser Tester and Photo of Two-Frequency Interferometer Laser Tester) was modified to permit the beam from a second HP/Agilent laser to be combined and sent to an optical receiver or biased photodiode. A diagram is shown in Diagram of Test Setup for HP/Agilent Laser Optical Frequency Comparison. A Polarizing Beam-Splitter (PBS) was added, at first only on a somewhat adjustable mount (normally used for Beam Splitters or Beam Benders in these systems) and could easily be installed and removed without tools. However, this proved to have nowhere near the estimated 1/10th of a mR precision needed to align the beams to obtain a beat signal. So, a Newport MM1 kinematic mount with the PBS attached to it was installed in its place. The second laser itself is on a fully adjustable platform (3 screws), so the beams can be lined up precisely.

    The distance from each laser to the PBS is relatively closely matched, so if they have the same optics (e.g., 6 mm), the wavefront curvature should be similar resulting in minimal alignment issues and decent signal signal amplitude. How much mismatched optics (e.g., 6 mm with 9 mm) will affect this is not known though. But in the far field, the more significant issue would probably be the loss of available optical power from the size mismatch since the wavefronts should be quite close to planar.

    A Half-Wave Plate (HWP) which may be installed in front of either laser will rotate the polarization if needed to pass it through the PBS, with the PBS itself serving to eliminate the unwanted F1 or F2 component. Then, F1 or F2 from Laser 1 can be beat with F1 or F2 from Laser 2. However, not knowing the range of possible difference frequencies likely to exist for any given pair of lasers, the highest bandwidth HP/Agilent 10780 may not be adequate to capture the difference frequency without a lot of luck, especially if comparing lasers with different spec'd nominal optical frequencies like the 5517A and 5517C (12 MHz difference). So my original intent was to replace the HP optical receiver with a Zygo 7080 which is used with their lasers having a REF frequency of 20 MHz. Assuming a maximum of a +/-0.1 ppm frequency offset with respect to nominal for each laser, the difference could be up to 95 MHz, though I highly doubt even half of this is at all likely. Nonetheless, some higher speed photodiode detector may be needed in general. And my poor old 10 MHz oscilloscope which had been dedicated to monitoring of HP/Agilent REF and MEAS frequency signals was augmented with another equally old 50 MHz scope. :) (As it turned out, for the tests I actually performed on two specific 5517B lasers, the original 10780A was quite adequate.)

    Since F1 (the lower optical frequency) and F2 (the higher optical frequency) have known orientations, being able to select each one will make it possible to unambiguously determine whether the difference frequency between the two lasers represents Laser 1 or Laser 2 being the one that is higher in frequency.

    Another approach would be to monitor the difference frequency as the cube-corner or plane mirror ("tool") in the interferometer is moved; whether it decreases or increases for a given direction of motion will also unambiguously determine which laser's frequency is the higher one.

    For maximum accuracy, both lasers need to warm up from a cold start (not restarted) for several hours in an environment with a fairly constant temperature. But of course, the activity during warmup is in itself quite exciting. :)

    Note that the stray magnetic field from one laser will change the REF frequency of the other laser if they are close together, especially if they are parallel to each-other. The frequency offset that is introduced is only order of 1 or 2 percent of the reference frequency at most. It may in fact not affect the optical frequency very much, if at all. And even if it does, the magnitude should be of negligible consequence. But any disturbance is still worth minimizing. A bit more on this below.

    There may also be second-order effects of external magnetic fields that are not aligned with the laser's magnetic field. This will not only change the strength of the axial magnetic field, but will also introduce a transverse component to the magnetic field with unknown consequences.

    Although it turned out that I don't have a laser with a known optical frequency and had to return the one that was probably close, I used a pair of healthy 5517Bs to perform the experiment. But nothing is perfect! :) A few of the issues:

    Using a Half-Wave Plate (HWP), it was possible to select F1 or F2 of Laser 1 (L1) to beat with F2 of Laser 2 (L2). The difference in optical frequency changed from 5.0 Mhz (no HWP, L2F2-L1F1) to 2.7 MHz (HWP oriented at 45 degrees for a 90 degree rotation, L2F2-L1F2), proving that Laser 1 has the lower absolute optical frequency. (Laser 1's REF or split frequency is 2.3 MHz.) It sure feels good when the physics cooperates! :) (For convenience in trying to keep track of things, the difference frequencies here are referenced to F1 of both lasers, rather than the average of F1 and F2, which strictly speaking may be more accurate. In most cases, this results in a shift of less than 1 MHz.)

    The difference frequency between F1 of Laser 1 and F1 of Laser 2 (L2F1-L1F1) after coming Ready can be anywhere from approximately -2 MHz to +6 MHz depending on whether one or both lasers was started from being cold. (Of course, before locking, the difference frequency can be up to 1.6 GHz or so - the FWHM width of the neon gain curve!) The beat then drifts by 4 or 5 MHz over the next hour or so. Typically, if both lasers are started at the same time, having been off for more than an hour, the difference frequency (L2F1-L1F1) just after locking is around -1 MHz, goes down to -2 MHz, and then climbs to a final difference frequency of around +2.6 MHz over the course of an hour or so. Pretty impressive for systems running at 474 THz. That 2.6 MHz is less than 1 part in 100,000,000!

    The temperature of the laser - or more likely the temperature of the control electronics - also affects the optical frequency by slightly shifting the lock point on the split HeNe gain curve. The data above was with the lasers undressed. Installing their covers resulted in the difference frequency dropping to under 1 MHz after an hour but then it went through 0 Hz and seems to have finally settled at around -2.3 MHz after 5 hours. Removing the cover over the PCB on Laser 1 caused the difference frequency to climb back up to +2.6 MHz, so a change of almost 5 MHz. I swapped in the Control PCB from another 5517 laser just for grins and giggles, The optical frequency after 10 hours from a cold start moved to about -1 MHz with at least as wide a variation during warmup as with the original Control PCB. I don't presently have a 5517B with a newer digital Control PCB so I can't compare its stability to that of the common analog Control PCB.

    Finally, I swapped the beam sampler assemblies including the LCD switch on Laser 2 and this resulted in a significant change in the difference frequency - almost 5 MHz. It's possible that an LCD panel that is starting to delaminate or degrade in some other way could do this, but I have no reason to suspect that one of these was bad.

    Given how small the difference frequency between two randomly selected 5517Bs is, could it be that these 5517B lasers really are close to the nominal spec'd value of 632.991372 nm or 473.612234 THz? Possibly, since the least significant digit of 632.991372 or 0.000001 nm is about 0.75 MHz. Where's a NIST calibration lab when you need one?

    I also did some very basic experiments with magnetic fields applied to one of the lasers. Placing another similar laser along-side Laser 1, the difference in optical frequency could be dramatically changed as the distance or orientation of the two lasers was altered. But, most - but not all - of the frequency change was eliminated in a few seconds as the control loop readjusted the locked position. The initial response didn't seem to be instantaneous, so perhaps it was something related to a slight change in laser tube current or something else affecting the temperature of the tube. Given that both the strength and symmetry of the magnetic field was being affected by repositioning, it's not at all clear what was actually changing. But everything else should be intuitively obvious. :-)

    Not content to permit one of these lasers to rest wherever it pleases, I've built a simple network to introduce a small offset into the error signal driving the laser tube heater power amp to fine tune the optical frequency. The circuit consists of a pair of 1K ohm resistors feeding 4.7 V zener diodes from +15 VDC (TP8) and -15 VDC (TP10) to ground (TP1). (These testpoints represent convenient locations to attach the circuit.) The regulated +/-4.7 VDC are probably not really needed but won't hurt. A 25 turn 25K ohm pot connects between +/-4.7 VDC with its wiper feeding a "gain control resistor", whose other end is the offset output. The "Power Amp" test jumper on the Control PCB was removed and replaced with a connector having a 2K ohm resistor to partially isolate the driving op-amp (test header Pins 1 and 2, assuming pin 1 is on the left), and the offest is introduced to pin 4 (which is shorted to pin 2 on the PCB). With a 33K ohm gain control resistor, one turn of the pot changes the optical frequency by about 1 MHz, providing a range of more than +/-10 MHz. A positive offset reduces the optical frequency on the 5517 laser. (It would probably be the opposite for the 5501B.)

    There's no need to bother with this for the 5501A as it has a "Photodiode Offset" pot (R4) on the Lock Reference PCB, which essentially performs the same function. (It's the square pot in the corner.) Normally, the pot is adjusted to maximize the REF/split frequency, which automagically centers the lasing position on the split neon gain curve. But it has quite a wide range - at least +/-50 MHz, possibly much more.

    It would be a simple extension to lock two of these lasers together at 0 Hz with a PLL using the REF frequency of Laser 1 (L1F2-L1F1) as the reference input to a phase/frequency detector (a flip-flip and some simple logic), and the difference frequency between the lasers (L1F2-L2F1) as the "VCO" input. The phase/frequency detector output would be the offset error signal fed into the power amp. The locked state would then have L1F1 equal to L2F1.

    Even simpler and actually more flexible would be to use a monostable and RC filter to convert the L1F2-L2F1 difference frequency to a voltage, and compare that with a set-point value to generate the offset error signal. This would allow the difference frequency to be adjusted over a wide range with closed loop control, including the case where L1F1 equals L2F1 (or with trivial modification, where L1F2 equals L2F2).

    Both of these schemes are left as exercises for the student. :)

    There is still another annoyance. It is a slow variation in the difference frequency with a period of a few seconds. It looks like sort of a dance with the control loops of one or both lasers hunting back and forth a few hundred kHz to and much as +/-1 MHz. (The values for the difference frequencies above were averages.) The period is between 2 and 3 seconds and appears to remain relatively constant. When I first noticed this hunting behavior, I didn't know if was a peculiarity of the locking electronics or from some external influence like the DC power supplies, vibrations, drafts, or aliens attempting to communicate with Earth. :) There is a digital clock with a period of 2.56 seconds on the laser Control PCB but I didn't think it should be doing anything once the laser locks. And it wasn't something associated with the original Control PCBs in either laser since swapping them with Control PCBs from other lasers didn't noticeably change the behavior. Nor did swapping one of the HeNe laser power supplies on the off chance that the switching frequencies were interacting in some peculiar way.

    I also substituted a 5501B for Laser 2, so call it Laser 3. (The final resting place for the frequency difference was around 2.5 MHz for L3F1-L1F1.) The variation in difference frequency was still present, but its deviation seems to be much lower, perhaps +/-100 kHz. The other thing that changed was the power supply since I had to use a different one for the 5501B. So it was possible that power supply fluctuations, origin unknown, could be the cause. However, then I swapped in the power supply originally used for Laser 2 to power Laser 1 and that made no difference. Thus, not the power supply. I then substituted Laser 2 for Laser 1 and the large fluctuations returned. So, they must be either associated with the tube itself or the LCD switch, since everything else of relevance (Control PCB and HeNe laser power supply) had already been swapped with no effect on the frequency fluctuations. And note that the combination of Laser 1 and Laser 3 did have some of this, just not nearly as much with Laser 2. But, even though it isn't pretty, this could still very likely be considered normal since even the much greater variations of Laser 2 are below the allowable specifications for these lasers. However, I then swapped the tube from Laser 2 into another case which meant the HeNe laser power supply and Connector PCB were different. At first, it looked like the difference frequency variation was way down, below +/-50 kHz, but over time as both lasers reached equilibrium, it climbed back up to around +/-300 kHz. Possibly somewhat lower than before but nothing conclusive. The difference frequency had also shifted down by about 5 MHz (L3F1-L2F1 of 6 MHz). Swapping the control PCB made little difference, but swapping in the original beam sampler assembly restored the difference frequency to its previous value of 1.5 MHz but also restored the large fluctuations! So, then yet another beam sampler assembly was installed and now the difference freuquency is around 0.5 MHz, but the fluctuations are also way down at +/-50 kHz and seems to be staying that way. So, it may be that the beam sampler has the most impact on both of these issues. Seriously strange.....

    In frustration, I finally did some of the things I should have done originally - changing the digital clock speed, looking at some control PCB signals and - gasp! - actually reading the service manual in more detail! :) It's easy to select clock speed on 5517B/C/D lasers - there is a convenient jumper for "Normal" and "High". Switching to High immediately caused the hunting to be much faster. Parallelling some key resistors to change the oscillator speed also had a similar effect. So, the hunting was related to the digital circuitry but how? Checking some key test points immediately revealed that my assumptions were totally wrong and that the cause is a direct result of the HP/Agilent implementation of the feedback loop using the LCD switch to alternately select each of the two polarized modes, rather than incorporating a polarizing beam sampler with separate photodiodes. Before realizing the cause, I had assumed that the LCD switch was alternating polarization states at 50 Hz. It's not. The two states are (1) when there is no drive to the LCD and both sides are at the same potential (passive state, polarization rotated 90 degrees) and (2) when the LCD is driven by a 50 Hz squarewave of opposite polarity on each side (active state, polarization unchanged). Apparently, the response of the LCD is slow enough that the active state is essentially DC - it doesn't see the 50 Hz ripple. (At least that's the theory. Given that the behavior of these lasers in terms of exact locked optical frequency and slow speed oscillation in optical frequency is affected by the specific beam sampler with its LCD that is installed, I wonder if there is enough variation in the residual response to be the cause.) The LCD switches from the passive to the active state with a period of 2.56 seconds! Just before each state change, the appropriate sample-and-hold is latched with the photodiode voltage corresponding to that state. This happens every 1.28 seconds. So the hunting is a direct result of this digital artifact in an otherwise analog system. It should be possible to eliminate or greatly reduce the effect. A few of the possibilities are to simply run the system at a higher speed once it's locked, to improve the sample-and-hold circuits (e.g., bigger caps!), or to replace the LCD and S&H circuits entirely with a normal polarizing beam sampler and dual photodiodes.

    At least one mystery remains and it has to do with the REF (or split) frequency behavior after locking. Starting at that time there is a fairly long period of an hour or more where the REF frequency oscillates by 1 or 2 percent up and down (period of several minutes), with a smaller oscillation of a few tenths of a percent period of 20 or 30 seconds). For example, on Laser 1, it starts between 2.279 MHz and 2.308 MHz (change of about 1.25 percent) and the deviation of the faster oscillation is 0.002 MHz (change of about 0.09 percent). The deviation of these oscillations gradually declines until it becomes small or non-existent, or the periods become very long so the changes are not easily seen. The mystery is why there is no obvious corresponding oscillation of the difference frequency between two lasers! One might expect that the origin of the REF frequency oscillations is the optical frequency varying about its nominal value causing the REF frequency to change based on the lasing position on the split gain curve. But while the difference frequency drifts during warmup, there is no obbious correspondance, by eye at least, with the REF frequency variations.

    So, both of these will probably remain unresolved for now.

    See the section: HP/Agilent Laser Wavelength/Optical Frequency for measurements of other lasers and additional comments.

    I then ran a pair of these lasers for many hours to get an idea of the longer term stability. In all cases they are Lasers 2 and 3 from the above link, a healthy 5517B and 5501B. The first set of data is for both from a cold start:

      Lasers      Balanced
      Time on     Frequency
      (hh:mm)    Difference
        0:00     -8.50 MHz*
        0:15     -6.75 MHz
        0:30     -6.65 MHz*
        0:45     -6.55 MHz
        1:00     -3.95 MHz
        1:15     -2.55 MHz
        1:30     -2.05 MHz
        1:45     -1.75 MHz
        2:00     -1.75 MHz
        2:15     -1.69 MHz
        2:30     -1.67 MHz
        2:45     -1.66 MHz
        3:00     -1.61 MHz
        3:15     -1.57 MHz
        3:30     -1.55 MHz
        3:45     -1.57 MHz
        4:00     -1.45 MHz
        5:00     -1.44 MHz
        6:00     -1.32 MHz
        7:00     -1.44 MHz*
        8:00     -1.45 MHz

    The "Balanced Frequency Difference" is the mean optical frequency for Laser 2 [(L2F1+L2F2)/2] minus the mean optical frequency for Laser 3 [(L3F1+L3F2)/2]. * Denotes interpolated values, not measured. Read: Guessed, because I missed recording the value for that time slot! :)

    So, approximately 2 hours was required to get within 0.5 MHz of the final value. However, these are the average of several 10 second measurements and there is an uncertainty of several hundred kHz for each one. There is high frequency jitter at the switching frequencies of the HeNe laser power supplies (40 kHz tyical), as well as at the PWM frequency of the 5501B controller (50 kHz typical), and slow hunting back and forth from the control loops of both lasers (a few seconds or more).

    Next, I turned off Laser 2 for 2 hours to let it return to room temperature without touching Laser 3, and then started Laser 2 up again:

      Laser 2     Balanced
      Time on     Frequency
      (hh:mm)    Difference
        0:00     -8.15 MHz
        0:15     -6.60 MHz*
        0:30     -5.05 MHz
        0:45     -3.52 MHz
        1.00     -2.48 MHz
        1:15     -1.84 MHz
        1:30     -1.61 MHz
        1:45     -1.44 MHz
        2:00     -1.35 MHz
        2:15     -1.20 MHz
        2:30     -1.21 MHz
        2:45     -1.26 MHz*
        3:00     -1.32 MHz
        3:15     -1.39 MHz
        3:30     -1.39 MHz
        3:45     -1.39 MHz

    Finally, I turned Laser 3 off and let Laser 2 run all night:

      Laser 3    Balanced
      Time on    Frequency
      (hh:mm)    Difference
        0:00     -3.25 MHz
        0:05     -1.45 MHz
        0:15     -1.65 MHz
        0:30     -1.55 MHz
        0:45     -1.20 MHz
        1:00     -0.85 MHz
        1:15     -0.97 MHz*
        1:30     -1.10 MHz
        1:45     -1.15 MHz*
        2:00     -1.20 MHz*
        2:15     -1.26 MHz
        2:30     -1.31 MHz*
        2:45     -1.37 MHz
        3:00     -1.46 MHz
        3:15     -1.52 MHz
        3:30     -1.51 MHz
        3:45     -1.46 MHz
        4:00     -1.36 MHz
        8:00     -1.56 MHz
       10:00     -1.95 MHz
       12:00     -1.85 MHz
       24:00     -2.02 MHz
       25:00     -2.05 MHz
       26:00     -1.75 MHz
       29:00     -1.69 MHz
       30:00     -1.94 MHz
       32:00     -1.68 MHz
       33:00     -1.79 MHz
       34:00     -1.79 MHz
       36:00     -1.76 MHz

    So, it would appear that most of the frequency drift with warmup is due to Laser 2, and Laser 3 has very little drift after the first few minutes and for several hours after that. I don't know if this is a characteristic of 5501Bs in general, or whether Laser 2 is simply particularly slow to reach equilibrium. But based on previous testing, other 5517s were not much better. The 5501B control scheme does differ in subtle ways from that of the 5517s, though not in anything fundamental. The cause of the increase in optical frequency difference after about 8 hours is not clear but it does seem to have settled in at the new location. It's unlikely to be anything related to the lasers themselves reaching equilibrium as that should have happened well before this time. The ambient temperature may be changing as I have little control over that in my, um, lab. :) The limit cycle seems to have a total deviation of +/-150 kHz or about 0.0003 ppm.

    Recall that these are averages over multiple 10 second intervals. My next objective was to reduce the short term variations, initially those due to the switching noise from the HeNe laser power supplies in the lasers. So, I fabricated adapters so that the tubes in both lasers could be powered from Spectra-Physics 248 exciters, which are AC line powered HeNe laser power supplies with linear regulators. The residual ripple in the tube current is below the scope noise floor and undetectable, compared to 3 percent p-p from the VMI power supplies HP used. Based on the previous test where I substituted an SP-248 in one of the lasers, I assumed this would clean it up. Well, not quite.

    The result was slightly less messy, but there was still serious frequency modulation with deviation in the MHz range, now seeming to be at a bit over 50 kHz. Since I was beating a 5517B with a 5501B, the last possibility for switching noise of this type was from the 5501B, which uses Pulse Width Modulation (PWM) to drive the heater rather than a linear power amplifier. It runs at around 50 kHz nominal, but that's not very precise as it is simply determined by an RC network.

    So, I swapped in another 5517B (Laser 1) and Voila! The beat is now a really nice sinusoid with only a hint of fuzz. This is what a heterodyne signal should look like! No wonder HP did away with the PWM driven heater in the 5517 lasers! Exactly how the ~50 kHz noise makes its way into the optical frequency is not at all obvious. It would seem to be too high a frequency to be a mechanical effect but the heater coil is bifilar-wound so there should only be a very small magnetic field from it. But perhaps, very small is still enough!

    As further insurance, I have modifed a pair of SP-248s to reduce the already essentially undetectable power line ripple by at least another order of magnitude. For details, see the section: Reducing SP-248 Current Ripple I also plan to add a protection circuit to the SP-248s that will shut the supply down if either the current exceeds 4 mA due to a fault or the current drops out. I don't want a power supply failure to fry a tube!

    Photo of Test Setup for HP/Agilent Laser Optical Frequency Comparison is really ugly but quite functional. The scope on the right shows the actual beat frequency of about 3 MHz between F1 of one 5517B facing left-right and F2 of the 5517B pointing toward the back of the photo, in this case with their covers off. The Variac is powering the two SP-248s sitting under it to reduce the AC input voltage until I add the addtional RC filter. (Although they work at normal line voltage, the voltage across the regulator pass treansistors is pushing their limits at the low current.) The Thorlabs DET-110 can just barely be made out in place of the HP interferometer. In front of that also sitting on the butcher block is a polarizer wheel that may be installed in either beam. And the HWP to select F1 or F2, and bounce mirror for alignment, are on the table in front of the butcher block. The HP-5508A is displaying "PA Error" (Path Error) because it isn't getting the signals it normally expects.

    Most of the high frequency FM modulation is gone but the slow variation persists with the frequency difference still varying by at least 300 kHz and sometimes much more over the course of a few seconds. This was very puzzling. I have substituted a healthy 5517A for each of the 5517Bs with no change and even put the AC power line to the entire setup on a Sola constant voltage transformer, again with no change. Later I realized that the LCD is selecting H and V polarization with a period of around 2.56 seconds. And the specific LCD has at least some effect. The behavior appears somewhat chaotic because the two lasers are doing it independently at slightly different frequencies. (The master clock on each is an RC oscillator, not crystal controlled.

    Causes of Variation of the Optical Frequency

    These lasers were apparently designed to be good enough for the metrology applications, but are far from what would be considered to have low noise or a super narrow line-width. The following are only some of the possible causes resulting in changes to or frequency modulation of the optical frequency by as much as 1 MHz or even more, over various time scales. The following applies directly to HP/Agilent 5517 and 5501B lasers except as noted, but much of it also applies to other HeNe Metrology lasers:

    Not included here are normal variations in the optical frequency that are a result of the lasing process itself.

    Refractometers for Directly Measuring the Index of Refraction of Air

    Most commonly, these types of metrology systems use the measured or manually entered values for temperature, pressure, and humidity to compute the effective index of refraction of air. However, there can always be errors in the measurements used to determine these inputs. One alternative is use an instrument called a refractometer to precisely measure the index by comparing the wavelength in air to that in a vacuum using an interferometer setup similar to that of a wavemeter. But rather than comparing the wavelength of a reference laser (usually a 633 nm HeNe) to that of an unknown laser, it is compared to the same laser but with a part of the beam passing through a vacuum cell whose length can be changed using a bellows arrangement. When the path lengths in air and the vacuum cell are changed simultaneously, the resulting two sets of fringes can be used to compute the index of refraction of air in near real-time. Since this is a direct comparison, in principle, calibration of these refractometers is never required. The only way the readings could be inaccurate is if the vacuum cell were to leak.

    Spindler & Hoyer LR1 Refraktometer (Refractometer)

    At least two companies are known to have developed this technology independently - Spindler & Hoyer, and Zygo. While the implementations differ in details, they are based on essentially the same approach. Two of the relevant patents on which the LR1 is based are: German Patent DE3703086C2: Laserinterferometer-Refraktometer and European Patent EP0277496B1: Laserinterferometer-Refraktometer. The European Patent contains the claims section of the patent also in English and French. (For reference, two of the Zygo patents are: US Patent 4,733,867: Apparatus for the Measurement of the Refractive Index of a Gas and US Patent 4,685,803: Method and Apparatus for the Measurement of the Refractive Index of a Gas.)

    The actual implementation of the LR1 is close to what's shown in the patents but not identical. See Optical Organization of Spindler & Hoyer LR1 Laser Refraktometer. A few liberties have been taken in the diagram. Some parts have been moved to enable them to be shown without requiring a 3-D perspective rendering. :) And some (hopefully) non-essential details have been guessed because the relevant parts could not be seen without extensive disassembly. However, the optical organization should be equivalent.

    The beam from a JDSU 1103P 2 mW polarized HeNe laser is split into two reference beams and two measurement beams as in the patent but what happens after that differs. In particular, the front of the vacuum cell (closest to the laser) is fixed while the far end moves back and forth so that both beams pass through it:

    The index of refraction of air then = (IFC + DFC) / DFC. More below.

    D = Distance over which fringes are measured, VWL = Vacuum WaveLength, AWL = Air WaveLength. The factor of 2 is there due to the total path being up and back.

    These are equivalent to linear interferometers for each of the two optical channels sharing a common retro-reflector for the fixed arm.

    The displacement is the distance up and back on each pass over which fringe data is captured. However, while the number of fringes for the Displacement channel is several thousands the number of fringes for the Index channel - which is the difference between two large numbers that are very close together - is less than 10. For an active displacement of 0.5 cm, there would be around 15,800 full fringe cycles but only 4 or 5 Index fringe cycles. Therefore, the computation for Index must be done with more sophistication than for the displacement to achieve an ultimate accuracy better 1 part in 300 in the 3 significant digits in the display of 1.000xxx. This could be accomplished with a variety of techniques including analog electronics (e.g., multipliers, gain-controlled amplifiers, zero crossing detectors, etc.) or a microprocessor if samples of the Index waveform could be captured at a high enough resolution. Fortunately, computation speed is not a high priority here. ;-) The actual implementation used in the LR1 is not known.

    In the diagram, BS1 splits the input beam into two equal parts. BS2 splits these pairs into the measurement beams for D and I, and the reference beam. BS3 combines the return beams from the retro-reflectors. The beam-splitters are non-polarizing. The direction of polarization for all beams is normal to the plane of the diagram. Interestingly, the reference retro-reflector is not an entire round cube-corner, but a slice vertically about 1 cm thick. Since there is plenty of space, this is strange. :) Perhaps mounting was thought to be simpler. And, an apex/edge line is smack in the center and would appear to be unavoidable even with the narrow beams. :( Also of note is that the bottom turning mirror (TM2) is actually some type of solid prism, but it's too far recessed to examine closely. It might even be another beam-splitter though I don't see what purpose it would serve. The details of the front and rear vacuum seals are also not known but drawing them as optical windows will work. ;-) Normally, the length of the vacuum cell stainless steel bellows changes by about 1 cm via the linear motor during each pass, but unfortunately as is described below, it can expand by much more with sufficient force.

    The extended location of the Reference retro-reflector (compared to how it is depicted in the patents next to the beam-splitters) is believed to serve two purposes: The first is to minimize mechanical jitter and noise, particularly for the Index measurement. The Index retro-reflector and Reference retroreflector are rigidly mounted to the same aluminum block and any vibration would tend to cancel in the interferometer. The other purpose may be to reduce the path length difference between the Reference arm and Measurement arm of the interferometer to maximize fringe contrast, particularly for the Index measurement which is more critical. The Reference path length (BS2 to RR-R) and Index path length (BS2 to RR-I) are almost exactly the same. If RR-R was mounted close to the beam-splitters as in the patent, the path length difference would be close to 3/2 the laser tube cavity length, a null point where fringe contrast goes to zero with a multi-longitudinal mode laser.

    Unfortunately, the unit I acquired must have had a close encounter with a hard surface. While there was no visible external damage and nothing obvious even when the covers were removed, it turns out that the small cube-corner (RR-D) was seriously smashed when it hit the center of the large cube-corner (RR-I) which may even have been pushed slightly out of position and suffered an large unsightly ding. Neither of these would have affected its performance, but the damage to RR-D was enough that there was no return beam from it at all. While there appeared to be a hard end-stop to carriage movement which would have kept the two optics safely separated (spaced at least 2 cm), it turned out that only a stiff spring (possibly only the vacuum bellows) prevents movement beyond this. So, with enough G-forces, they were able to smacke into one-another. By my calculations, a drop on its face from less than one meter onto a carpeted floor would have been suffient for bad things to happen.

    It was possible to repair the unit by replacing RR-D with one I had sitting around. ;-) After that it was just optics alignment. First, RR-D was aligned by positioning it for maximum displacement fringe amplitude and tightening its screws. TM-1 was adjusted using its 3 screws to further peak the RR-D fringe signal. Then RR-I was installed (it had to be removed to replace RR-D), positioned to peak its fringe signal, and secured. Both signal amplitudes are now much greater than needed for the machine to display a valid value.

    But the axial position of the carriage had also shifted slightly forward from the trauma resulting in it making an annoying "clank" against the end-stop on each cycle, which also limited its range, and the consistency of the display. After spending over an hour attempting to reposition the frame by loosening bolts (one of which was and still is totally frozen), I found the obvious position adjustment ring on the linear motor and rotated it a couple turns. :( :)

    This LR1 is now reliably fully functional displaying values that differ by less than 1 percent in the final 4 digits compared to the NIST Index of Refraction of Air Calculator based on Ciddor Equation. The error is always low, so it could be that the vacuum has leaked out (in?) by the same percentage since the value in those 4 digits is inversely proportional to the pressure inside the vacuum chamber. But I have absolutely no intention of doing anything about it. ;-)

    If the firmware encounters a bad measurement or sufficient number of them due, for eaxmple, to corrupted fringe counts, the leading "1" will change to a "0" but the lower digits will retain their previous value and remain locked until a RESET.

    There is an access panel on top of the LR1 along with a tool to permit the position of the carriage to be adjusted from the outside. So that's what the unidentified tool is for. ;-) If I had realized that earlier, it would have eliminated a lot of wasted time and effort in eliminating the clanking. :( :) Apparently, the adjustment is need to compensate for major changes in altitude, which would shift the axial position by enough to matter. For example, that might happen when moving it from Tampa to Denver - a change in air pressure from around 1 atm (760 mm/Hg) to 0.85 atm (620 mm/Hg). (But the shift in this unit went the wrong way, so that's not the explanation for the need for the position adjustment unless it was being used at the bottom of an oil well.)

    There is also a trim-pot on the main PCB for carriage movement amplitude, and a trim-pot on the detector PCB, probably for gain.

    Apparently the firmware is rather finicky about the number of Index and/or Displacement cycles as setting the a carriage movement ampitude too small or too large results in an error condition (a 0 in the leading digit, or a blank display if reset).

    I do not know how many LR1s still exist, but I bet it's a small number. I doubt more than a few dozen were ever built. And there are fewer still in working condition.

    Here are some photos:

    Now here are some comments from the original designer of the LR1 in response to my questions with regard to what I was observing with the damaged unit:

    (From: Dieter Frolich.)

    By pure chance, a guy at Zygo Corporation invented essentially the same thing at essentially the same time. We did not know of each other, and the patent applications occurred during the periods when the other application was not yet published. So in the end, I as well as Zygo each had a patent for practically the same thing. I sold my patent (together with my company) to S&H, so S&H was the owner when Zygo`s patent popped up. Instead of fighting each other, we (easily at that time, if I remember right, no lawyers were involved) reached an agreement that each company could act as if the other patent did not exist. In case you like American patent slang instead of reading the European patent, the Zygo patents have a similar description. (Though as noted, the details of the implementation differ. --- Sam) It was hard to find them: They do not contain the word "refractometer".

    A problem of the system was the vacuum, the tightness of the interface between the optical parts (window / corner cube), and you almost certainly experience this problem with your device. Even if you have a loss of vacuum, the system should display the refractive index or what it thinks the refractive index is IF(!) a complete measurement cycle is detected. The LR1 should achieve an accuracy of better than 0.000,001 in the index of refraction. This is less demanding than it might seem. The LR1 directly measures the difference between 1.0003xx and 1.000000, so this is a practical accuracy of 1 part in 300 in measuring the period of one interferometric cycle. At 633 nm, one interferometric cycle has a mechanical length of approximately 1 mm. We measured the length of a few cycles because it is difficult to detect exactly the start and the end point of one cycle. I believe, we used approximately 5-10 cycles, but I am not sure about this (in my German patent I wrote 4, but we have probably gone to the safe side). And the LR1 will only generate a value for the index of refraction if this minimum number of cycles has been detected. So: You will need 5-10 mm of travel before the LR1 displays anything.

    As said above, your cell probably has a leak. If you have a wide open leak, then the internal volume contains exactly the same as the surrounding medium, and not a single interferometric cycle will be generated because air flows into and out of the cell during each measurement cycle. With a wide open leak, the stiffness is the same as with a tight cell, but the movement will have an offset (constant, but different, pressure in the cell). The carriage will bump into one of the ends, but even if you change the offset (you could do that by modifying the placement of the compensating spring(s)), you will never get a signal. If you have a very narrow leak, so narrow that there is practically no exchange of gas between the inside and the outside during a single cycle, then you have a functional system in principle. However, you will have an offset in position, the same as in the case above, and in addition, you have a change in stiffness: The compression of the gas is effectively an additional spring, but I think this additional spring is quite weak compared to the already existing stiffness. If you change the offset, you should be able to generate an interferometric signal. Interestingly: If the cell were filled with an ideal gas instead of vacuum and if it were gas tight, then you would measure exactly the index of refraction of the surrounding air (as desired) despite the fact that you do not have any vacuum in the system. However, air is not an ideal gas, and therefore what you measure will not be exactly the index of refraction of the surrounding air but something different. But at least, you might get a number displayed. So I propose: Try to change the offset such that the cell moves freely between the ends - try to move the sping(s), use different springs or try something else, there are no really delicate parts in the system, all the accuracy results from the interferometric nature. If you get free motion, but no signal, then try to improve the air tightness of the cell by adding some epoxy around the window / corner cube and the end of the evacuation tube. (Remember: If that helps, you will nevertheless not measure the true index of refraction.)

    Dielectric Coating Damage or Degradation

    Several types of dielectric (non-metallic) coatings are used in interferometers. The most common are the Anti-Reflections (AR) coatings on the surfaces of optics in the beam paths. This is followed by the polarization-sensitive coating the on the diagonal of the Polarization Beam Splitter (PBS), which is at the heart of most interferometers configurations.

    While modern AR coatings are quite robust, they can be affected by both physical abuse during cleaning as well as by chemicals in the environment. The result can be micro-scratches, fading, or mottling. Minor damage of this sort is generally of little consequence even if the appearance is quite gross. But serious scratches or other physical damage can render them unreliable or totally inoperative.

    The PBS cubes - which are critical to most metrology interferometers - utilize a dielectric coating along the hypotenuse (diagonal) the separates the P and S polarized components. A healthy PBS will be almost perfect but one that has degraded may not provide good separation, or may have a limited angle of acceptance. The best PBS will have almost perfect extinction of the undesired polarized component over a decent range of incident angles. One that has degraded may result in this being marginal or even unacceptable. The cause of this degradation is not obvious at that coating is quite well protected but Excel PBS cubes (even apparently new/NOS ones) appear to be particularly susceptible. But even Hewlett Packard PBS cubes may be marginal.

    The simplest way to test the PBS uses a linear polarized 633 nm HeNe, or the 5517 laser (or other metrology laser) with a linear polarizer (LP, which can be one for a camera but NOT a circular polarizer). Remove everything from the interferometer - cube corner(s) and quarter wave plate(s) if present - and place it in the beam with the label at the top and an edge parallel to the beam. Position the LP in the beam between the laser and PBS with its axis of polarization vertical. If the PBS is good, there will be almost no light coming out the front with most being reflected to one side. (If the axis of polarization for the LP is not labeled, rotate it until almost no light gets through. For a defective PBS, there may be no such orientation even if the incident beam is perfectly perpendicular to the face of the PBS.) Rotate the PBS a few degrees either way around the vertical axis and very little light should still get through. Rotate the PBS 90 degrees around the vertical axis and repeat. The behavior should be similar except that the reflected beam will exit from the opposite face. Defective PBS will allow too much light to get through even when the beam is perfectly aligned. Marginal ones will just barely cut off the beam under these conditions. They would still be acceptable but alignment is more critical.

    Mystery Two-Axis Laser Interferometric Displacement Sensor (LIDS2)

    This beautifully engineered assembly showed up on eBay with no information other than what was on the unit itself. It consists of a platform with a block of prisms and separate cube corner mounted in a massive stainless steel case. The precise alignment of the platform can be fine tuned using set-screws from the sides and then locked down. It has two inputs which are fed from Polarization-Maintaining (PM) fibers, and corresponding fiber outputs (non-PM) for their associated detectors. Although they aren't labeled as inputs, putting PM fibers at the outputs would be plain silly. And while the beam paths are mostly symmetric between input and output, but not entirely. And the not entirely becomes non-sensible when tracing them.

    The actual displacement is measured using free-space beams. This thing apparently operates at a wavelength around 1,300 nm as the OZ Optics fiber connectors all have designations for 1,300 nm, strange. Strange because there are no common low cost laser sources which are both single longitudinal mode and optical frequency accurate around 1,300 nm. The most likely might be a DFB diode but once stabilized, the cost would be similar or higher than a similar HeNe laser. And resolution at 1,300 nm all other factors being equal will be about half of that at 633 nm. And, no, it's not just the fiber connectors. It definitely doesn't work correctly with a 633 nm HeNe.

    The manufacturer is unknown - scraped off before the seller acquired these. But it's likely not many of these were ever made as the serial number of one I have is "6". Not even 106 or 1006, just 6. ;-) So perhaps it wasn't entirely successful. :( :)

    Here are some photos of a typical unit (courtesy of Mike Read):

    By using a 633 nm (red) fiber-coupled HeNe, it was possible to confirm the beam paths for the two possible inputs:

    But with no access to a polarized 1,300 nm laser, any polarization coatings do not do anything meaningful. :( 633 nm is probably about the worst wavelength to use to determine anything about reflective coatings or polarization behavior, being close to 1/2 of 1,300 nm. So, while these beam paths are fairly certain, exactly how the unwanted beam paths are suppressed - if the are - is not known. For example, what is the purpose of the diagonal inside the input prism? The reflection from there hits the wall of the case for both the L and H inputs. Perhaps it acts as a polarizing beam-splitter and guarantees that the unwanted polarization does not get any further. ;-) But there shouldn't be any unwanted polarization from the polarization-maintaining fiber. And the block under the cube corner is suspected to be a Quarter WavePlate (QWP) but that cannot be confirmed either.

    However, at this point my thought is that this was a failed experiment whose design is fundamentally flawed. There are several reasons for this but the biggest one is that the use of a PM fiber for the input and non-PM fiber for the output as implemented does not permit displacement information to be determined regardless of whether a Single Frequency (SF) or Two Frequency (TF) laser is used. Both of these require the polarization state of the merged beam to be fixed and orthogonal. With an SF laser, a "quadrature decoder" produces electrical sin and cos signals that are 90 degrees from each other. With a TF laser, the beat signal between the two orthogonal polarized components is used, but this requires maintaining the polarization state until the detector also. A non-PM fiber will not do either. Further, as noted above, there is no real TF laser option at around 1,300 nm.

    Another curiosity is the apparent requirement for two lasers (or at least two laser inputs) to the interferometer blocks. Normally, a single laser is used and split to feed each axis. Doing that internally would make a lot more sense compared to requiring a fiber splitter, two expensive fibers, and two expensive PM fiber couplers.

  • Back to Laser Instruments and Applications Sub-Table of Contents.

    Scanning Fabry-Perot Interferometers


    While the interferometers described in the previous sections have many applications in diverse areas, the Scanning Fabry-Perot Interferometer (SFPI) is specifically designed to make measurements of the longitudinal (axial) mode structure of CW lasers. The SFPI rates it's own set of sections both due to its importance and because it is possible to construct practical SFPIs at low cost without the need for a granite slab or optical table for stability.

    The longitudinal mode structure of a laser is one of those concepts that is often explained but not so often demonstrated. There are a number of indirect ways of showing that it exists including monitoring the beat frequencies between modes and looking at the fringe patterns in a Michelson or other conventional interferometers. But one of the clever ways of actually being able to display the modes as they would appear in a textbook is to use an instrument called a "Scanning Fabry-Perot Interferometer" (SFPI). The term "Laser Spectrum Analyzer" (LSA) may also be used for some of these instruments, but an LSA should not be confused with an "Optical Spectrum Analyzer" (OSA), which is generally - but not always - based on a very different technology, the scanning monochromator. And for the most part, SFPIs cannot be used to determine a laser's frequency or wavelength. Most SFPIs accept a free-space laser beam. However, some - mostly designed for telecom applications - may use a fiber-optic connector for input - a fiber collimator inside the instrument generates a free-space beam for the actual SFPI.

    When most people see or hear "Scanning Fabry-Perot Interferometer" - even those familiar with lasers - what probably comes to mind is a device with planar mirrors at each end. While these do exist and are used for specific applications, the vast majority of instruments for laser mode analysis have a pair of identical curved (spherical) mirrors positioned at a fixed spacing called "confocal" equal to where their focal points would meet, which is also equal to their Radius of Curvature (RoC). These are among several alternatives to planar mirrors that have significant advantages as will be discussed below.

    What follows provides informations on several different types of SFPIs including some not found in most optics or laser texts, along with examples of both home-built and commercial instruments.

    While an SFPI is conceptually simple and actually quite straightforward to construct (at least in principle), even a basic system can display detail in the longitudinal mode structure of a laser that represents about 1 part in 50,000,000 compared to the optical frequency of oscillation or wavelength of the laser. For a red HeNe laser, the "resolvance" of such an instrument would be on the order of 10 MHz (out of 474 THz) or 0.013 picometers (0.000013 nm, out of 633 nm)! It's all done with mirrors! :-)

    However, an SFPI can only show the spectrum of a laser's output over a limited range of wavelengths (determined by the SFPI's mirrors) modulo a much smaller value called the "Free Spectral Range" (FSR). The FSR is determined by the mirror spacing and is typically not more than a few hundred times the resolution. An SFPI cannot measure the absolute optical frequency or wavelength of a laser. That requires an instrument like an Optical Spectrum Analyzer (OSA) or optical wavemeter, or comparison (heterodyning) with a known reference laser. So, an SFPI is like a microscope that can display a very small region of a laser's spectrum at very high resolution. And depending on the specific application, the SFPI display may only make sense if there is already at least some idea of what to expect. :)

    An SFPI can be used to view the mode structure of lasers where the gain bandwidth is less than its FSR such as a HeNe laser (~1.6 GHz) or ion laser (~5 GHz). A typical SFPI display of a 5 mW HeNe laser is shown in SFPI Mode Display of Melles Griot 05-LHR-151. Two sets of 3 modes are visible due to the 1.75 GHz FSR of the SFPI. For the red (633 nm) HeNe laser, 1.75 GHz is sufficient to cover all lasing modes.

    However, an SFPI can usually be used to determine if a laser is Single Longitudinal Mode (SLM) regardless of its gain bandwidth since the chances of multiple modes being both stable in height and frequency, and falling on top of one-another (modulo the FSR) so that the display shows a single peak is very small. And if more than one SFPI is available with different FSRs, and the displayed frequency offset of a laser with multiple longitudinal modes is the same on both, then that's an indication - though no guarantee - that the distance between modes is less than the smaller FSR.

    A number of companies currently (as of 2010) offer SFPIs including Coherent, Thorlabs, and Toptika. For only a few thousand dollars, one of these instruments can be yours. Alternatively, it's possible to build something with very respectable performance for a couple bucks. Companies like Spectra-Physics, and TecOptics no longer manufacturer SFPIs but their instruments (as well as those that are still in current production) may show up surplus at very affordable prices. SFPIs may also be called "Laser Spectrum Analyzers". But they should not be confused with even more expensive "Optical Spectrum Analyzers" which are generally scanning monochromator-based instruments that do not even come close to the resolving power of a typical SFPI.

    Principles of Operation

    An SFPI uses the optical transmission characteristics of a specially designed Fabry-Perot (F-P) resonator as a very selective filter to scan across the optical spectrum of the laser. Any F-P resonator will have a transmission behavior that has peaks and valleys based on optical frequency (or wavelength). The peaks will be located where the distance between mirrors is an integer multiple of one half the laser wavelength. As the reflectivity of the mirrors approaches 100 percent, the peaks become increasingly narrow and the valleys increasingly flat and close to zero transmission. This characteristic looks like that of a "comb" filter which is very selective.

    An SFPI consists of a pair of mirrors with relatively high reflectivity (90% to 99.9% or more is typical) mounted in a rigid frame. In most SFPIs, the laser under test (LUT) is aimed into one end and a photosensor is mounted beyond the other end. Depending on the type of SFPI, the coarse spacing and alignment of the mirrors may be adjusted by micrometer screws, by some other means, or set precisely and fixed at the time of manufacture. The axial position of one of the mirrors can also be varied very slightly (order of a few half-wavelengths of the LUT) by a linear PieZo Transducer (PZT). (Other methods of moving the mirror can and have been used but the PZT is most popular.) By driving the PZT with a ramp waveform and watching the response of the photosensor on an oscilloscope, the longitudinal modes of the LUT can be displayed in real time. In essence, the comb response of the SFPI is used as a tunable filter (by the PZT) to analyze the fine detail of the optical spectrum of the LUT. As long as the FSR (c/2*L except under certain (but very useful) conditions, described below) of the SFPI is larger than the extent of the lasing mode structure of the LUT, the mode display will be unambiguous. Where this condition isn't satisfied, the mode display will wrap around and may be very confusing. For example, the common helium-neon (HeNe) laser has a gain bandwidth of about 1.5 GHz and longer HeNe laser tubes will generally operate with multiple longitudinal modes covering much of this range. Thus the FSR of an SFPI to be used with such a laser must be greater than 1.5 GHz, corresponding to a SFPI cavity length of less than about 100 mm (assuming c/2*L for a planar cavity), as in the photo, above. For Nd:YAG, the gain bandwidth is about 150 GHz, which results in a required SFPI cavity length of less than 1 mm! However, in practice, lasers don't necessarily lase over their entire gain bandwidth, especially if specific steps have been taken to assure single or dual mode operation (also called single or dual frequency operation). For those - which include many useful lasers - the requirement can be relaxed such that the FSR of the SFPI only needs to be larger than the width of the expected mode structure. And for a single mode laser, this would be only the width of the lasing line itself. Therefore, in these cases, a long cavity low FSR SFPI will result in the highest resolution.

    Commercial scanning Fabry-Perot interferometers usually cost thousands of dollars - or more! But it's possible to construct an SFPI that demonstrates the basic principles - and can be even quite useful - for next to nothing, and one that rivals commercial instruments for less than $100.

    The resolution ("resolvence") of a Fabry-Perot (FP) interferometer is determined by the wavelength, mirror reflectance, mirror spacing, and incidence angle of the input beam. For the following, we assume normal incidence (which will be satisfied in most practical situations).

    Consider a plane-plane mirror FP cavity with a mirror spacing (d) of 80 mm and reflectance (R) of 99 percent at a wavelength (λ) of 632.8 nm (red HeNe laser):

                  λ2 * (1-R)         4*10-13 * 0.01
     Delta-λ = --------------- = --------------------- =
                2*d*π*sqrt(R)     0.16 * 3.14 * 0.995
      ~8*10-15 m = 0.000008 nm or about 6 MHz.
    (A wavelength of 633 nm corresponds to an optical frequency of 474 THz, so 0.000008/633*474 THz is approximately 6 MHz.)

    Another measure of the performance of an interferometer or laser cavity is the "finesse". This dimensionless quantity is the ratio of the FSR to the resolution. In essence, for the SFPI, finesse determines the how much fine detail is possible within one FSR. The reflectance finesse for a plane-plane cavity is equal to π*sqrt(R)/(1-R) where R is the reflectance of each mirror (which are assumed to be equal). For R near 1 as would be the case in a useful SFPI, this reduces to π/(1-R). For most discussions of finesse that follow, high R mirrors are assumed so this equation is valid and finesse is then inversely proportional to (1-R). (But if you're a stickler for precision, feel free to use the exact equation!) So, with a reflectivity of 99 percent for both mirrors, the theoretical maximum finesse will be roughly 300. If the FSR is 1.875 GHz as in the example above, the resolution will be approximately 6 MHz.

    However, to simplify setup and improve usability, practical SFPIs are often built with cavity configurations other than plane-plane known as "mode-degenerate", which typically have a theoretical finesse that is lower by a factor of 2 or 4. Even for these, alignment and mode matching of the input beam to the SFPI cavity still impact finesse. And various aspects of the mirrors themselves such as coating losses, surface finish, dust, and contamination can further reduce the realizable finesse and limit resolution, especially for high performance instruments. Much more below.

    Transmission of Fabry-Perot Resonator versus Optical Frequency is a composite plot that shows how finesse affects F-P behavior. The Transmission of Fabry-Perot Resonator Slide Show has a separate plot for each value of finesse, which is less confusing when finesse is high. In an SFPI, the cavity spacing rather than optical frequency is varied, but the SFPI display of a single frequency (single longitudinal mode) laser for a given value of finesse will look similar to these plots. Where the laser has more than one frequency as is typical of common (unstabilized) HeNe lasers, the display will essentially be a summation of shifted and scaled versions of these plots (much more below). Low finesse F-P cavities (often even lower than a value of 1!), usually called etalons, may be used to select specific lasing lines due to their effect on intra-cavity gain. However, to be useful for an SFPI, a high finesse is desired to be able to resolve lasing lines that are close together. The finesse of a general purpose SFPI will typically range from 100 to 500. Higher values are possible but require better quality more expensive mirrors. As they say, it's all done with mirrors. :)

    A few other relevant equations can be found at the bottom of the Vintage Spectra-Physics Model 450 SFPI Brochure Page.

    Other factors will conspire to reduce the useful resolution of a practical SFPI. At modestly high mirror reflectivity (e.g., R=99%), these include alignment, input beam diameter, and input beam collimation. As R is pushed closer to 100%, the quality of the mirrors, their cleanliness, and internal losses become increasingly important. But for the example above, even if the actual finesse is worse by an order of magnitude compared to the theory, it will still be possible to easily resolve the individual modes of any common HeNe laser and probably even the nearly 2 meter long Spectra-Physics model 125 (177 cm resonator, mode spacing of 85 MHz). This is a factor of better than 1 part in 10,000,000 comparing resolution to optical frequency!

    However, note that while textbooks will tell you that the peaks should get through with little attenuation, this is probably not going to be true with practical high finesse SFPIs. (At least not those you're likely to see!) The amplitude of the peaks will depend critically on the quality of the mirrors and of course, on the alignment. For "laser quality" dielectric mirrors, I've gotten as high as 5 to 10 percent peak transmission for a high finesse SFPI using mirrors with a reflectivity of 99.8%. I'm sure this can be improved upon but even so, for a 1 mW laser, there is still more than enough optical power at the output of the SFPI to produce a nice display on most scopes using a 1:1 probe without a preamp.

    An on-line calculator of Fabry-Perot behavior can be found at Light Machinary Etalon Designer. However, the finesse is extremely sensitive to the entry for "SUrface Finish". Put in a value of "0.01 nm" to get textbook results. :)

    (From: A. E. Siegman (

    In evaluating the effect of losses in Fabry-Perot mirrors you really have to distinguish between internal losses (or loss-equivalent effects, like scattering) that are physically located "inside" the mirrors (i.e., inside the effective reflection plane of each end mirror), and external losses that are physically located "outside" the effective reflection plane, but still within the physical layer of the mirror.

    Losses that are outside the mirrors are effectively just additional external transfer losses in the system - they have the same effect as if they were separate from the FP, so that they don't affect the FP itself but just weaken the light before or after the FP.

    Losses inside the mirrors (aka "internal" losses) are more serious because they are exposed to the higher-intensity resonant fields inside the F-P and therefore can significantly affect the finesse and peak transmission of the FP.

    Just measuring the net reflectivity and net transmission of the mirror itself won't clearly distinguish between these internal and external losses. Also, how you'd describe a situation where the losses are distributed through a moderately thick mirror layer is something I've never thought through; doing this would require a slightly more sophisticated wave calculation of forward and backwave wave propagation inside the finite-thickness partially absorbing mirror layer itself.

    (Too bad I'm no longer actively teaching laser courses; this calculation would make a nice homework problem to torment -- sorry, educate -- students.)

    Mode-Degenerate Fabry-Perot Interferometers

    A major disadvantage of the general spherical F-P cavity is that super precise alignment and control of the input beam size and collimation, along with an intracavity aperture, may be needed to suppress higher order transverse modes in the SFPI resonator. Though not present in a TEM00 laser, exciting higher order modes are almost unavoidable in the SFPI cavity and may in fact dominate the display and render it completely useless. Even if the time consuming steps required to eliminate the higher order modes are taken, there will always be uncertainty as to what is actually being seen. The flat-flat cavity doesn't have this problem but suffers from disadvantages of its own, mainly in the need for a well collimated input and very precise mirror alignment to achieve high finesse and as a result, reflection of the input back directly back into the laser, which may destabilize many types of lasers.

    One way to eliminate the transverse mode problem is to use a cavity configuration called a Mode-Degenerate Interferometer (MDI) in which the higher order transverse modes have the same frequency/wavelength as the TEM00 (longitudinal) modes and thus simply fall on top of them in the display. Even though each peak in the display representing a longitudinal mode of the input laser may actually be built up of contributions from multiple transverse modes excited in the resonator of the interferometer, the characteristics of the individual longitudinal mode components in each of these transverse mode are the same so the accuracy of the resulting display isn't affected. (This should not be confused with the very different situation of a laser having multiple transverse modes in its output where the frequencies, phases, amplitudes, and polarizations of the corresponding longitudinal modes in each transverse mode may differ.)

    Two practical arrangements that satisfy this condition are the (1) spherical cavity (d=2*r) and (2) confocal cavity (d=r). (However, as will be described below, thare are many more, some of which are even useful.) The confocal cavity has the larger finesse and is thus usually employed in SFPIs since the finesse is a measure of Q-factor with respect to the FSR or mode spacing, and thus higher finesse results in better resolution. A planar cavity (r of infinity) doesn't support higher order modes at all and its theoretical finesse is double that of the confocal cavity, but is generally a less desirable configuration due to alignment and other issues (see below).

    Note that the term "confocal" actually refers to any cavity where the focal points of the two mirrors are coincident. However, only the case where d=r is stable and thus generally useful for the MDI SFPI. This means that the two mirrors must have an identical Radius of Curvature (RoC) for optimal performance, and thus should be from the same manufacturer and production run if possible. (SFPIs constructed using mirrors with different RoCs may not be totally useless, but should be avoided for most applications.)

    The frequencies of the transverse modes of a symmetric cavity Fabry-Perot resonator are given by the following equation:

              c            1                              d
      fmn = ------ * [q + ---- * (1 + m + n) * cos-1(1 - ----)]
            2 * d          π                              r


    So, the equation can be rewritten as a base frequency plus an offset:

             c       c                      1               d
      fmn = --- + {------ * (1 + m + n) * [--- * cos-1(1 - ----)]}
             λ     2 * d                    π               r

    Thus the first term is simply the optical frequency of the laser while the second term consists of three parts: the longitudinal mode spacing or FSR, the integer mode numbers, and a correction factor (<1) that depends on the mirror RoC and spacing.

    The interferometer will be mode-degenerate when there are TEM00 modes that have the same frequency/wavelength as some of the transverse modes. The requirement for this to be satisfied is for the inverse cosine term in the equation above to be equal to π divided by an integer, l. Then there will be "l" types of modes with one type - where (1+m+n) is equal to 1, modulo(l) - having the same frequencies/wavelengths as some TEM00 modes. When (1+m+n) is not equal to 1, modulo(l), that mode will fall in between the TEM00 modes in locations depending on (m+n)/l, modulo(l). So, the SFPI display will be similar to that of a non-MDI setup where only the TEM00 modes are excited except that the FSR will be reduced by a factor of l. But the resonances will actually be mostly for higher order transverse modes (of the interferometer) unless the alignment of the input beam is near perfect - and that usually doesn't happen by accident. (More on this below.) And for the user, good performance is achieved with non-critical alignment.

    For the following, "d" is the mirror spacing and "r" is the RoC of the mirrors as above:

    A diagram comparing the most common configurations is shown in Three Scanning Fabry-Perot Cavities with Equal FSR. The mirror reflectance is assumed to be high enough that the finesse approximation is valid. But other factors will reduce the realizable finesse.

    For the confocal cavity, half of the transverse modes are not mode-degenerate when an on-axis input beam is used as there are two types of modes depending on whether the quantity (1+m+n) is even or odd:

    This seems a bit strange that the TEM00 modes (m+n=0) have non-integer mode numbers but the equation has been confirmed from at least two different sources.

    As noted, with two sets of peaks, the FSR is effectively cut in half to c/(4*d). Rearranging the equation above with the new FSR of c/(4*d) out in front, one sees that the various transverse modes (those that differ in m+n) result in a frequency difference of c/(4*d). However, integer differences in q corresponding to the longitudinal modes, still have an FSR of c/(2*d). Where a paraxial beam (one parallel to the optical axis) enters the confocal cavity off-center, the beam path repeats itself after two traversals of the cavity (in a zigzag pattern) and the FSR is easily seen to be c/(4*d) rather than c/(2*d). However, if the beam is very well aligned and centered, the FSR will be c/(2*d) since only some symmetric modes will be excited. However, the finesse is still the same (with respect to the 4 pass round trip cavity).

    When adjusting the mirror distance of a confocal cavity SFPI to be precisely confocal as it needs to be, there will be many positions where the SFPI may appear to work but which aren't quite confocal. This is especially true of short confocal cavities - the type most commonly found in commercial instruments. Depending on the specific distance, non-degenerate higher order modes will result in multiple and/or ghost peaks and/or a variation in the amplitude of the lasing modes depending on their position on the voltage ramp drive signal. The amplitude will also be lower overall. However, when the correct distance is approached, all of these ghosts will collapse into the desired high amplitude display. Don't be fooled! Thus it's best to know or determine the exact RoC for the mirrors before installing them in the SFPI so the initial distance can be set reasonably precisely. However, those other bogus resonances have been exploited to investigate a variable FSR mode-degenerate SFPI - it's a feature, not a bug! :) See the next section.

    Planar mirrors may also be used since a true flat-flat cavity does not support any stable higher order modes, degenerate or otherwise, but it is the most difficult to align. And, although the theoretical finesse is double that of the confocal cavity, the realizable finesse is usually lower and is also much more dependant on the alignment than with the confocal or with other non-planar configurations. Also, with optimal alignment, the incident beam is reflected directly back into the laser which may result in instability for many types of lasers. So, it's often necessary to use an optical isolator of some type (Faraday or polarizing beam-splitter with Quarter-Wave Plate (QWP), or at least an optical filter to reduce the intensity of the back-reflected beam (by the square of the transmission coefficient). However, where the distance between the mirrors of the SFPI is adjustable as in some general purpose instruments like the TecOptics FPI-25 or the (likely) custom Burleigh Triple-Pass Scanning Fabry-Perot Interferometer described below, there is no choice. Both of these enable the distance between the mirrors to be varied from almost touching (for an FSR of 100 GHz or more) to at least 15 cm (for an FSR of of 1 GHz or less). (Intracavity etalons also usually use planar mirrors but the finesse of these does not generally need to be very high so the alignment is not nearly as critical.)

    Some useful things to keep in mind:

    Selectable FSR Mode-Degenerate Fabry-Perot Interferometers

    While the confocal and spherical MDI configurations are the best known and the confocal SFPI is probably the most widely used for analyzing a laser's characteristics, it's possible to make use of symmetric cavities having values of d/r other than 1 or 2 and they may be useful for certain applications. Check out the short paper: K. Kernera, S. M. Rochestera, V. V. Yashchuka, and D. Budkera, "Variable Free Spectral Range Spherical Mirror Fabry-Perot Interferometer". However, the term "variable FSR" is not strictly accurate: Unlike a plane-plane cavity, the FSRs here are restricted to specific values determined by the RoC of the mirrors. Thus "selectable FSR" is a more accurate description. :)

    The equation that must be satisfied for resonance in a cavity with spherical mirrors having identical RoCs is:

          d             k*π
         --- = 1 - cos(-----)
          r              N


    (My apologies for changing the names of variables. I've maintained my own if possible. But "k" replaces "l" both because l may be similar to the number "1" in many type fonts like Courier, and so as not to be mistaken for l in the MDI equation, above, which now becomes "N". Are you confused yet?)

    The FSR will be equal to c/(2*N*d) and the amplitude of each peak will scale as 1/N. Plugging in N=2 and k=1 results in the confocal cavity; N=1 and k=1 is the normal spherical cavity. The number of spots on each mirror (if not perfectly aligned on-axis) will be equal to N.

    Here are four examples (including the true spherical and confocal):

    The following table lists all valid cavity configurations for N from 1 to 10, for a spherical cavity SFPI as well as a planar cavity with a spacing of d=r (the RoC of the mirrors for the spherical cavities). The values of FSR, FWHM, and Finesse are all relative to those of the planar cavity where finesse is π/(1-R) assuming R is close to 1:

                    <---- Relative ---->  <----- Example (5) ----->
                    (2)     (3)    (4)      d     FSR    FWHM  Fin-
      N  k   d/r    FSR    FWHM  Finesse  (cm)   (GHz)   (MHz) esse  Notes (1)
      1  0  0.000  1.000   1.000  1.000   3.331  4.500   15.00  300  Planar 1-0
      1  1  2.000  0.500   0.500  1.000   6.662  2.250    7.50  300  Spherical 1-1
      2  1  1.000  0.500   1.000  0.500   3.331  2.250   15.00  150  Confocal 2-1
      3  1  0.500  0.667   2.000  0.333   4.995  3.000   30.00  100  Spherical 3-1
      3  2  1.500  0.222   0.667  0.333   1.665  1.000   10.00  100
      4  1  0.293  0.854   3.414  0.250   0.975  3.841   51.21   75
      4  3  1.707  0.146   0.586  0.250   5.685  0.659    8.79   75  Spherical 4-3
      5  1  0.191  1.047   5.236  0.200   0.636  4.712   78.54   60
      5  2  0.691  0.289   1.447  0.200   2.301  1.302   21.71   60
      5  3  1.309  0.153   0.765  0.200   4.395  0.688   78.54   60
      5  4  1.809  0.111   0.553  0.200   6.024  0.498    8.29   60
      6  1  0.134  1.244   7.474  0.167   0.446  5.598  111.96   50
      6  5  1.866  0.089   0.536  0.167   6.214  0.402    8.03   50
      7  1  0.099  1.443  10.098  0.143   0.330  6.491  151.47   43
      7  2  0.377  0.379   2.656  0.143   1.254  1.707   39.84   43
      7  3  0.777  0.184   1.286  0.143   2.589  0.827   19.29   43
      7  4  1.223  0.117   0.818  0.143   4.071  0.526   12.27   43
      7  5  1.623  0.088   0.616  0.143   5.406  0.396    9.24   43
      7  6  1.901  0.075   0.526  0.143   6.330  0.338    7.89   43
      8  1  0.076  1.642  13.137  0.125   0.253  7.390  197.06   38
      8  3  0.617  0.202   1.620  0.125   2.056  0.911   24.30   38
      8  5  1.383  0.090   0.723  0.125   4.604  0.407   10.85   38
      8  7  1.924  0.065   0.520  0.125   6.407  0.292    7.80   38
      9  1  0.060  1.842  16.582  0.111   0.201  8.291  248.73   33
      9  2  0.234  0.475   4.274  0.111   0.779  2.137   64.11   33
      9  4  0.826  0.134   1.210  0.111   2.752  0.605   18.15   33
      9  5  1.174  0.095   0.852  0.111   3.908  0.426   12.78   33
      9  7  1.766  0.063   0.566  0.111   5.881  0.283    8.49   33
      9  8  1.940  0.057   0.516  0.111   6.459  0.258    7.73   33
     10  1  0.049  2.043  20.432  0.100   0.163  9.194  306.48   30  Max FSR/FWHM
     10  3  0.412  0.243   2.426  0.100   1.373  1.092   36.39   30
     10  7  1.588  0.063   0.630  0.100   5.287  0.283    9.45   30
     10  9  1.951  0.051   0.513  0.100   6.497  0.231    7.69   30  Min FSR


    1. These are the named cavities, and the two described in the text, above. The planar cavity isn't strictly part of this series but is included for completeness, and as a baseline for comparison. Its mirror RoCs would be infinite but its mirror spacing (d) is equal to the RoC (r) used for all the spherical cavities. Configurations for N above 10 would have larger maximum values for the FSR and FWHM (k=1), and smaller minimum values for the FSR (k=N-1), with the FWHM approaching that of the normal spherical cavity (1-1).

    2. FSR takes into account the actual value of d as the cavity spacing is varied. The values shown are relative to that of the planar cavity (with mirror spacing equal to RoC=r of the mirrors used in the spherical cavities). Then, FSRr(N)*d/r=1/N, or FSRr(N)=r/(d*N).

    3. FWHM is the width of the spectral peaks relative to that of the planar cavity based on the mirror reflectivity (R) and mirror spacing (d).

    4. Finesse is with respect to the effective FSR, relative to that of the planar cavity and scales as 1/N.

    5. This example shows the resulting mirror spacing (d), FSR, FWHM of the spectral peaks, and finesse for each spherical cavity with mirror RoC=r compared to a planar cavity with a mirror spacing of d=r=3.331 cm for an FSR of 4.5 GHz with a finesse of 300 (mirror reflectivity of approximately 99 percent).

    So, most of these are actually the resonances that one avoids when setting the spacing of a normal confocal SFPI! :) Since the spectral width of each peak is determined only by the mirror reflectivity (R) and mirror spacing (d), and r (RoC) would be fixed for a selectable FSR spherical SFPI like this, d (and thus FSR and width) can vary quite dramatically for any given value of N. For example, with spherical cavities 4-1 and 4-3, these differ by a factor of almost 6! But the width is never narrower than that of the normal spherical cavity (1-1), so the resolution can be no better as well. If a very small FSR is selected, the finesse is also lower. For example, when N=10, the finesse within 1 FSR is only 1/10th of what it would be for the normal spherical cavity or 1/5th that of the normal confocal cavity! The main benefit seems to be the capability of being able to select a large FSR using mirrors that would result in a much smaller confocal FSR. In the example above, it would be possible to have an FSR of over four times the confocal FSR with cavity 10-1 (9.194 GHz), but the finesse then would only be 1/5th as large (30). Though cavity 4-1 with almost double the confocal FSR, would be less than 1/3rd as long and still 1/2 the finesse (75). There could be other special situations where a selectable FSR and/or non-confocal SFPI might be useful, for example if the RoC of the only mirrors available would result in a confocal cavity too long to be practical. One specific example might be to implement a compact SFPI as a reference for a stabilized HeNe laser. For this, low finesse is quite acceptable and space could be critical. Using (n,k) = (9,2), the mirror spacing would be on 0.779 cm for an FSR of 2.137, more than adequate for the 1.6 GHz gain bandwidth of the HeNe laser. Using the hemispherical version (see below), the mirror spacing would be only 0.39 cm.

    Since the higher-N cavities require N times less movement of the mirror on the PZT to cover 1 FSR, they require less drive voltage as well. Using the typical beeper element PZT, at N=7, only around 2 V p-p will cover more than 1 FSR. But the down-side is that the sensitivity is so high that random vibration from cooling fans and such - and even sound from speaking or a radio results in jitter or worse. For these, using a smaller diameter PZT or adding some ballast like 5 minute Epoxy or hot-melt glue to stiffen it and reduce its sensitivity may be desirable.

    Note that commercial confocal SFPI heads cannot readily be used as their adjustment range is quite limited. Thus, either a custom-built cavity or a one of the totally general and very expensive commercial SFPIs normally fitted with planar mirrors would be required. So it's not clear when being able to actually select among the unusual spacings would be generally desirable enough to warrant the added mechanical complexity, effort, and cost involved in setting and fine tuning them beyond the intellectual challenge of an academic exercise. ;-) But it is straightforward, if not exactly inexpensive, to set one up using some standard parts from companies like Thorlabs. See the sections starting: Sam's Selectable FSR Hemispherical and Spherical Mode-Degenerate Scanning Fabry-Perot Interferometers, and also starting with Sam's Selectable FSR High Resolution Mode-Degenerate Scanning Fabry-Perot Interferometer 1.

    Hemispherical versions of all of the above - half the length with a planar mirror substituted for one of the curved mirrors - should also be valid, but with a further halving of the finesse if using mirrors with the same reflectivity, so I couldn't resist building one. :-) See the section: Sam's Selectable FSR Hemispherical and Spherical Mode-Degenerate Scanning Fabry-Perot Interferometers.

    But note that one mirror really does need to be planar - close is probably not useful since the cavity will no longer be mode-degenerate. Even mating a 1 meter RoC (rather than planar) front mirror with a 4.3 cm RoC rear mirror (a ratio of over 20:1) will result in compromised performance with ghost or double peaks showing up at times even for the most basic (almost) hemispherical confocal mirror spacing. Perhaps with a 5 or 10 meter RoC, the degradation would be small enough to be acceptable but planar is still best.

    Other approaches have been investigated using what might be termed "semi-mode-degenerate" cavities, where the first few undesirable transverse modes merge with the TEM00 mode, but others (which have much lower amplitudes) may not. Some examples can be found in U.S. Patent #5,418,641: Fabry-Perot Optical Resonant Cavity Systems. granted to Newport Corporation in 1995. This patent (which is definitely a great cure for insomnia!) covers just about every imaginable cavity configuration (including all those above). But with one exception, I am not aware of any commercial SFPIs employing a cavity other than the planar or confocal, even from Newport (who now has none at all), and none other than planar where the FSR can be changed with a knob. That one exception is the very old Tropel 2440 using the hemispherical confocal cavity. See the section: Tropel 2440 Scanning Fabry-Perot Interferometer.

    Further investigation of these special cases is left as an exercise for the determined student. :)

    An Excel spreadsheet is available that will generate the relavent parameters for N from 1 to 12 based on two example RoCs and the planar finesse. A screenshot is shown in Screen Shot of Excel Spreadsheet for Calculating Selectable FSR SFPI Parameters. This lists those for spherical configuration with RoCs = 4.2 cm and 100 cm and a planar finesse of 300. The entries at the top in italics may be entered to change those as well as to select the hemispherical configuration (HS = 2, but the annotation won't change). Contact me via the Sci.Electronics.Repair FAQ Email Links Page if interested.

    Mode Degenerate Scanning Fabry-Perot Interferometers with High Values of N

    Adjusting the mirror spacing for the common confocal cavity SFPI can be challenging enough, but obtaining a perfect display using very high orders of N, especially with long radius high-R / high-finesse mirrors even with equal RoCs may not be possible. Unavoidable small errors will get amplified resulting in the sub-peaks not able to be perfectly superimposed. For example, attempts to use 1 meter RoC green HeNe 544 nm OC mirrors (R>99.8%) with N=10 (mirror spacing of 4.9 cm) resulted in a display that seemed really disappointing at first. These mirrors were installed in the frame of Sam's Selectable FSR Hemispherical and Spherical Mode-Degenerate Scanning Fabry-Perot Interferometers. I was hoping that a compact setup with some 1 m RoC mirrors that were available could be offered for testing green lasers to confirm Single Longitudinal Mode (SLM) / single frequency operation. Two different sets of mirrors with supposedly identical RoCs were tried with what seemed at the time to be disappointing results. But it turns out that I misinterpreted the display and thought that what were actually legitimate peaks were sub-peaks that could not be superimposed no matter how carefully the mirror spacing was adjusted. Even if that were the case, the objective would still be possible, but just not as clean as using mirrors with a short RoC and N=2 (confocal). But then I recalled that the ramp sensitivity versus FSR multiplies by N. So these were actual peaks that were close together on the ramp.

    Mode Degenerate N=10 SFPI Display of C215M 532 nm SLM Laser shows the actual appearance on a mediocre DSO. Remarkably, even with the division factor of N=10, the effective finesse is still 100 or more. The FSR is 306 MHz for a resolvance of ~3 MHz. The peak amplitudes do vary significantly on the DSO and this trace was selected for a instance where it wasn't too bad. :( ;-) I originally thought the variation was due to the high value of N or SFPI alignment, but that is probably not the case as indicated by the fact that the amplitudes differ significantly on either side of the ramp valley. If it were an interferometer issue, they should be symmetric. The cause is more likely simply randomness in sampling of the very narrow peaks without a proper antialiasing filter. The peaks cover only 1 or 2 columns of pixels on the DSO screen. The variation was much less of an issue using a "proper" vintage analog scope. ;-) And note the amplitude of the ramp waveform: Scanning 1 FSR requires only around 0.4 V! This is due to the sensitivity multiplication factor of N and is probably also why the ramp is not clean - noise in the vertical input of the scope or from the function generator. A related side effect of the high N is the same increase in sensitivity to unintended vibrations of the "drum-head" PZT. The CPU cooling fan in a laptop at the other end of the workbench and even sounds are detectable by this "interferometer microphone".

    If small radius mirrors usable for green (532-544 nm) like those for yellow-red (590-650 nm, 4.2 cm RoC) were available, confocal (N=2) and modestly higher values of N (3,4) could be used for confirming SLM operation of a green laser. Even a "Nano SFPI" with a cavity length of 1.23 cm (N=4) would have acceptable performance. But those mirrors do not have high enough R for green wavelengths to display more than lumps.

    Folded Mode-Degenerate Scanning Fabry-Perot Interferemoters

    In order to construct high resolution SFPIs that are more manageable in terms of size, the intra-cavity beams can be reflected by one or more high quality planar mirrors without a major impact on performance, though structural stability will be more critical and alignment more challenging. With a single fold mirror, the output beam may be near the input beam but can be offset in both position and angle; with two fold mirrors, the output beam can be at the opposite end as in the normal configuration. The fold mirrors must be planar so as not to interfere with the mode-degenerate requirements and with R near 100%, low losses, low surface scatter, and high flatness. But the leg lengths do not need to be the same. Unlike the ring cavity SFPI described below, this is still a Fabry-Perot geometry and obeys the normal rules based on total cavity length. And the input and thus intrcavity beams will be normal to the curved mirrors avoiding asymmetry in the effective RoCs in X and Y that would result due to reflections from tilted mirrors.

    For example, using mirrors with a 1 meter Radius of Curvature (RoC), a single fold would reduce the length to just over 50 cm while maintaining the same FSR and nearly the same finesse. With two folds, it would be only 33.3 cm.

    While alignment would be a bit trickier than with a simple linear cavity, relatively crude adjustable mounts should suffice for all but possibly one of the end mirrors since they can be set up once and then left alone. However, as the number of fold mirrors increases, so does the challenge of achieving good alignment in finite time. One of the mounts will need to be "user-friendly" to facilitate initial alignment and with three adjusters for cavity length tuning. Each additional fold mirror adds a degree of freedom and it may be important to assure that the intracavity beams are all coaxial - retrace their steps. Otherwise, the requirement that they be normal to the curved cavity mirrors will not be met, though how much - if at all - this will matter with long radius mirrors is not known..

    As an alternative to purchasing very expensive mirrors from major optics companies, high quality planar mirrors like this for 633 nm may be salvaged from defunct HeNe laser tubes, which often use planar mirrors for the HR.

    Note that a spherical mirror SFPI with a single fold is NOT equivalent to one with a hemispherical SFPI having a curved mirror and planar mirror with the same reflectance in a linear cavity, even in the limit where the angle between the arms approaches zero. The hemispherical SFPI will have the same FSR but approximately 1/2 the finesse. While the full cycle (round trip) number of reflections will be the same, the planar mirror(s) are not HR and will add losses. An example would be a folded confocal SFPI with RoC = 10 cm and R of 99% for both curved mirrors and a planar HR fold mirror compared to a hemispherical confocal SFPI with R = 10 cm and planar mirror, both with R = 99%.

    Also see the section: Sam's Folded High Resolution Confocal Scanning Fabry-Perot Interferometer.

    Scanning Ring Cavity Interferometer

    A linear cavity is only a small subset of resonant cavity interferometers. Rather than starting an entire separate description of this special case, it's lumped in here, sorry. ;-)

    A ring configuration with 3 or more mirrors can provide certain advantages for use as a tunable etalon or laser spectrum analyzer. And the "ring" doesn't need to have non-intersecting beam paths but could be something like a bow-tie. But there are also complications.

    One major advantage of the Scanning Ring Cavity Interferometer (SRCI) is that even with perfect alignment, there can be NO back-reflections directly into the source since the intra-cavity beam is unidirectional and the entrance mirror will not generally be normal to the incident beam. Disadvantages include increased size and complexity, and more difficulty in alignment.

    The FSR for a ring cavity with a perimeter length L is c/L, double what it would be for a linear cavity with a mirror spacing of L. However, reflections at other than normal incidence from curved mirrors introduce astigmatism resulting in different effective RoCs for the in-plane and perpendicular axes. For this reason, while curved mirrors may be used, mode-degenerate configurations for the SRCI may not work well unless the geometry is very close to linear. However, note that the beam path in a confocal SFPI for an off-axis beam is analogous to a bow-tie configuration as shown in Confocal Scanning Fabry-Perot Interferometer but the beams with respect to the mirrors are very close to normal incidence so the result is only a small decrease in finesse.

    The effective focal length of a mirror as a function of angle is given by:

     ef = RoC*(1 - --------)


    And the Effective Radius of Curvature (eRoC) would be twice ef. A typical 3 mirror equilateral ring cavity where the incidence angle is 30 degrees will use one curved mirror and two planar mirrors. Then the in-plane RoC will be approximately 15 percent smaller than the perpendicular eRoC. For stability, both the RoC and eRoC must be greater than one half the perimeter L.

    More Information on SFPI Theory and Practice

    In addition to what is present in the sections below, check out the following links:

    SFPI Types and Capabilities

    Any electrically adjustable fabry-perot resonator can be used in a variety of ways including as a tunable filter or etalon. However, our interest here is mostly with respect to their application as a measurement instrument: The SFPI used as a laser spectrum analyzer. Nowadays, there are even F-P devices based on MEMS (Micro Electro-Mechanical Systems) technology. A complete device may be extremely small, but as such, they will be restricted to larger FSRs. Their flexibility will also be limited since all adjustments *must* be done electrically. Excluding these presently more exotic techniques, there are three common types of SFPIs that one is likely to find in an optics lab. These have been around and substantially unchanged for several decades:

    1. General purpose Fabry-Perot interferometers.

    2. Reconfigurable confocal cavity Fabry-Perot interferometers.

    3. Fixed configuration confocal cavity Fabry-Perot interferometers.

    In most commercial instruments, a separate ramp generator is used to drive the PZT and an oscilloscope is used for the display. They will typically include the ramp generator since the PZTs in most SFPIs require several hundred volts of drive to cover a few FSRs. The 'scope must be provided by the user. The ramp generators can be of various levels of sophistication, some including fine tuning of alignment for the general purposes SFPIs with triple-PZTs. Some drivers include temperature controllers for those interferometers with heaters to maintain constant (average) cavity length and even automatic locking to keep the display stationary even if the laser modes drift. A few SFPIs like those from Thorlabs have a requirement of only 5 V/FSR. Home-built SFPIs can be of any of the three types, but (2) is easiest to construct if appropriate mirrors are available. When using a beeper element as the PZT, an analog function generator (with at most a simple op-amp circuit to boost the voltage) can serve as the ramp generator since their drive requirements are modest. Cheap digital "DDS" function generators may not have a clean enough ramp/triangle waveform. However, Sam's Mini Laser Mode Analyzer 1 uses a low cost Arduino-compatible microcontroller and color LCD to combine the ramp generator and display into a single compact low cost unit compatible with low voltage drive SFPI heads. While it won't replace a rig with a nice scope, the performance may be adequate for many applications including determining if a laser is SLM and suitable for interferometry or holography.

    Subsequent sections cover home-built as well as common commercial SFPIs.

    Using an SFPI to Test a Laser for SLM Operation

    One of the common uses for a general purpose SFPI (also known as a Laser Spectrum Analyzer or LSA) is to test a laser to determine or confirm that it is running with a Single Longitudinal Mode (SLM). The most common LSAs are confocal cavity SFPIs with a Free Spectral Range (FSR) between 1.5 and 10 GHz.

    The following also applies to SFPI setup for any type of TEM00 (single spatial mode) laser, except that the display will be more complex. Only in special cases can an SFPI make sense of a laser that is running with a significant component of higher order spatial modes (non-TEM00) though it may be used to detect them.

    Testing a laser to confirm that it is running SLM - sometimes referred to as single frequency - is one of the common uses for an SFPI or LSA. For the following, the use of a somewhat generic commercial confocal cavity SFPI is assumed. However, the same approach applies to a home-built SFPI with similar characteristics except that the controls and settings will differ.

    Where the laser output power is more than 10 or 20 mW, attenuation must be added to limit the power to the SFPI. Since testing a laser for SLM is usually done at up to full power, only a small portion of the original beam must be sampled for the SFPI. This may be done with a combination of one or more beam samplers (beam-splitters) and optical filters. For this purpose, an uncoated glass plate such as a microscope slide or cover slip will suffice as a beam sampler. Each reflection will cut the power by around a factor of 10. An optical filter can then be used instead of another beam sampler to attenuate the beam further. The use of at least one beam-sampler will enable the original optical setup to remain essentially undisturbed, especially if it's very thin like a cover slip and doesn't shift the beam significantly. Or, it can enable the laser output power to be monitored simultaneously.

    CAUTION: Though the SFPI specifications may allow for more power, limiting the input to 10 or 20 mW will minimize the chance of any damage to the mirrors. Why? Because inside the high Q cavity, the peak power at the mirrors can be more than 2 orders of magnitude greater than the input power. 2 or 3 mW would be optimal.

    Note that the output from the laser is assumed to be more or less collimated. Even if the SFPI head has an adjustable input lens, it probably doesn't have enough range to accommodate a widely diverging beam. Thus the beam should be reasonably well collimated at the laser.

    Also, the SFPI will generally perform better with a narrow beam regardless of how large its input aperture is. Resolution isn't that important for determining SLM but it usually is for other uses.

    The basic connections for a typical setup are:

    Double check connections if the same connector type is used for multiple functions. Accidentally driving the photodiode with the HV ramp will not do anything good.

    The SFPI controller should be set approximately as follows:

    The scope should be set approximately as follows:

    Now comes the fun part. ;-)

    1. Turn on the laser, at low power if possible during setup.

    2. Use a beam-splitter like a microscope slide in a "third hand" or optics mount to pick off a small portion of the beam so that it can be directed into the SFPI's input. If the power is still over 20 mW, add either a second beam-splitter or optical filter to reduce it to a few mW.

    3. The SFPI should be aligned so that the beam hits near the center of the SFPI's input aperture and reflects back almost but not quite into the source.

    4. With the controller and scope powered, there should be some sort of display on the screen that correlates with the scan. Adjust the the SFPI alignment to maximize the height of the peaks and minimize their width. Adjust the SFPI alignment and input lens (if present) for the best scope display.

      The initial goal should be to display 2 or 3 Free Spectral Ranges (FSRs) on the screen. If the laser is is SLM, there will then be only 2 or 3 peaks.

    Adjusting the centering or offset control will move the peaks left and right. The dispersion or expansion selector switch or control can be used to zoom in on what's located near the center of the screen. The scan rate can usually be set at the maximum unless the limited bandwidth of the photodiode preamp is causing the peak amplitude to decline significantly.

    The display for a well behaved laser should be fairly stable in amplitude and position. As noted above, for an SLM laser, there will be single peaks separated by the FSR of the SFPI. And they will be essentially the same regardless of the type of laser as long as its SLM. However, for Multi-Longitudinal Mode (MLM) lasers, how many peaks there are, how far away they are from each other, and even being able to make any sense out of the display at all will depend on the mode spacing of the laser and FSR of the SFPI. For the display to actually represent the mode structure of the laser, the lasing bandwidth must be less than the SFPI FSR. For general purpose LSAs, that's likely to only be true of gas lasers, usually meaning HeNe and ion lasers. Solid state and diode lasers that aren't forced to be SLM usually have lasing bandwidths of 100s of GHz - wider than the FSR of all but very specialized SFPIs.

    A small amount of jitter is to be expected, due to vibrations and the SFPI itself. However, multiple sets of peaks (usually of greatly differing amplitude) and/or a noisy pattern would be an indication that the laser is running multi-longitudinal mode. A mode hop would be a sudden shift in the position of the peaks. (There will always be a slow drift due to thermal changes, but this would be an instantaneous change.)

    If it isn't possible to obtain a meaningful display on the scope, see the info below. The SFPI may be incomaptible with the laser, misadjusted, or broken.

    Using an SFPI to Compare the Mode Behavior of Two (or More) Lasers

    Displaying two lasers simultaneously on an SFPI may be desirable to determine the stability of their optical frequencies relative to one-another.

    This is very simple, if a bit tricky. Simple because it just a matter of beam combining. Tricky because the SFPI can only be aligned to one of the lasers using its adjustable mount; the second laser must be aligned to the SFPI directly and/or using one or more reflections. And note that the lasers do not even need to be the same wavelength as long as the SFPI mirror set supports them. The following assumes a pair of lasers, the extension to n lasers should be obvious:

    Complications include matching the beam power, diameter, and divergence of the beams to the SFPI. But this is not important if all that matters is monitoring the relationship of the lasers' optical frequencies over time. Then the quality of the SFPI peaks doesn't matter and lumps are just fine. ;-)

    Problems with Confocal Cavity Scanning Fabry-Perot Interferometers

    (This section deals with the interferometer itself. Electronic problems with ramp generators are not common, and when they do occur, relatively easy to diagnose and repair.)

    Commercial SFPI heads are generally quite reliable but problems can arise from misadjustment, abuse, or even poor environmental conditions over time. Symptoms can range from "no output from photodetector" to "rattles when shaken". :( :) The following assumes the system is set up correctly, the driver and scope are functional, and their controls are set properly. Refer to operation manual(s). :)

    Adjusting the Confocal Cavity Spacing

    The following also applies to other mode-degenerate SFPIs but confocal is by a wide margin the most common commercial general purpose instrument.

    The confocal cavity SFPI requires that the mirrors be spaced precisely at their RoC. For a short SFPI like those used as general purpose LSAs, this is order of a few microns or 1/10th the width of a human hair. For commercial SFPIs, the spacing will have been set at the factory and should not change over time unless messed with by someone who either didn't know what they were doing, or for SFPI heads with interchangeable mirror sets, replaced the mirrors and never succeeded in adjusting the spacing correctly.

    A single longitudinal mode narrow beam laser is best for this as it eliminates many of the possible ambiguities in mirror spacing. However, a laser with a small number of modes that fit within the FSR of the SFPI can also be used. For example, a 5 mW red HeNe laser for a 2 GHz FSR SFPI with a compatible mirror set.

    The following assumes mirror spacing adjustment from scratch as would be the case if swapping mirror sets. Where the SFPI produces a recognizable display, only some peaking will be required.

    1. If possible, the mirror spacing should be set close to the confocal spacing based on physical measurements. Where the spacing isn't listed, for the confocal SFPI, it will be L = c/4f (L = mirror spacing in meters, c = speed of light = 299,800,000 meters/second, f = FSR in Hz). For example, if the FSR is 2 GHz, the mirror spacing will be close to 3.75 cm or 1.5 inches.

      Where measurement isn't possible in a commercial SFPI, set it to roughly the mid-point of the adjustment range.

    2. Connect the PZT to the ramp output of the SFPI controller.

    3. Remove the photodiode assembly from the back of the SFPI head and mount it on-center a few inches behind the SFPI head in a way that it can be easily positioned to line up with the beam that makes it through the mirrors.

    4. Connect the photodiode to its preamp input. CAUTION: If the connectors are same for both the PD and PZT, make sure they aren't swapped as that can blow photodiode and/or ramp driver.

      If there is no preamp, the following circuit will have decent performance if the laser power is high enough. It may be noisier than the matched preamp but should be satisfactory for setup at least. Generally, 1 or 2 mW should be sufficient:

                                  Shielded Cable
             R Protect    PD1         Center
         +-----/\/\-------|<|-----------------------+----o Scope input (direct)
         |    1K ohms            +--------------+   |
         |  (Optional)           |    Shield    |   /
         |                       |              |   \ R Load
         |                       |              |   / 10K ohms
         |                       |              |   \
         |     +| | -            |              |   |
         +------||||-------------+              +---+----o Scope Gnd
             B1 | |

      This is similar to what's in a typical biased photodiode like the DET110 or DET36A2 from Thorlabs.

      Confirm that the PD is responding to light. Room light will suffice, but a better test source is a dimmable LED flashlight. These use Pulse Width Modulation (PWM) to chop the light output and a pulsed waveform should be clearly visible on the scope.

    5. Trigger the scope externally using the trigger signal from the ramp generator, or the ramp itself if none is available.

    6. Set up the test laser so it is aimed precisely into the center of the input mirror. If there is a lens, it should probably removed at this time as it may make things more confusing.

    7. If the reflected beam is fairly small, align the SFPI head and/or test laser so that the reflected beam just misses the laser's aperture. If it's a blob, some back-reflections into the laser will be unavoidable for now.

    8. Drive the PZT with the ramp at about 1/2 max and any range switch set to X1.

    9. With the scope's vertical sensitivity turned up, watch for any signal from the photodiode that is synchronized to the ramp. If using your own photodiode preamp or none and the blips go negative, reverse the PD polarity.

    10. If your cavity distance and mirror alignment were perfect, the result scanning through two FSRs for a laser with 3 longitudinal modes would look similar to the what's shown in SFPI Display of Melles Griot 05-LHR-151 5 mW HeNe Laser.

      For a single mode laser, there would be one peak separated from its neighbor by the FSR of the SFPI.

      More likely, the peaks will be smeared out or composed of multiple small blips as in the sequence of graphics shown in Typical Confocal Cavity Scanning Fabry-Perot Interferometer Mirror Spacing Adjustment Displays. This shows the general appearance of the SFPI display of a single longitudinal mode laser with a scan spanning approximately 2 FSRs for a 4.2 cm confocal cavity as the mirror spacing approaches optimum. The cavity length error (from left-to-right, top-to-bottom) is order of: 0.3 mm, 0.15 mm, 0.07 mm, 0.03 mm, 0.015 mm, 0 mm, a cavity length change of <1 percent for the typical SFPI LSA. But it will also depend on mirror RoC (longer has a higher tolerance), finesse (higher has a lower tolerance), mode order (confocal, half confocal, etc.), and other factors. In other words, your mileage may vary. :) The amplitude of the peaks would actually increase by a much larger amount than shown at the optimal setting. Which side the "crud" is on depends on the relationship of the ramp voltage to cavity length, swap if backwards. :) Some of these diagrams were originally from the Toptica SFPI 100 manual but have been heavily edited, I hope they won't mind. :)

      If the spacing is too far off, there may be just some tiny blips only visible with the photodiode gain turned all the way up. Adjust the spacing of the mirrors in small increments Slowly and then let the display settle down. With any movement, the display will become quite scrambled, so be patient. If going one way makes it worse, go the other way. :) For the typical LSA SFPI head, if the initial cavity spacing was within about 0.5 mm of being optimal, there should be only one place close by where it resolves into a beautiful display.

    11. If there is no evidence of any response related to the laser, confirm that the PZT is actually being driven - set the function generator to a few Hz and listen to the PZT! There should be a very audible clicking from the return of the sawtooth ramp. Check ramp output with the scope and connections if there is none. CAUTION: It may be several hundred V p-p! And check that the photodiode responds to light and the laser. Confirm that laser and mirror wavelength is compatible. And that the mirrors have the correct RoC.

    12. Once it's near optimal make very small changes in the spacing adjustment and allow it to settle out. Only a 1 degree rotation of the ring may go from perfect to a mess. This will depend on the FSR, finesse, and thread pitch. Once perfection is achieved, if there is a way to secure the lock ring so it won't move on its own, do it! But some commercial SFPI heads depend on special grease that doesn't flow. Kind of the opposite of Silly Putty™. ;-) All this may require 3 or 4 hands. :)

    First time users do not appreciate how precise the spacing needs to be. But for these short cavity SFPIs, it's less than 1/10th the width of a human hair - a few microns.

    It's also essential to avoid back-reflections into the laser, which will likely destabilize it and create chaos in the display. Initially, back-reflection to the laser may be unavoidable. But once mostly tuned up, the alignment should be adjusted such the the reflections from the SFPI (mostly the front mirror) do NOT enter the laser's aperture. With the confocal cavity SFPI, a slight offset will not significantly affect resolution. Ideally, an optical isolator could be used but they are pricey. With the lens in place (if available), the reflected spot may be much smaller and easier to direct off-axis.

    Constructing Inexpensive Scanning Fabry-Perot Interferometers

    I have used commercial general purpose Scanning Fabry-Perot Interferometers (SFPIs). For example, the TecOptics FPI-25 is an example of a very solidly constructed precision instruments with adjustments for just about everything. However, being so general, in some sense it is not optimal for anything! (See the section: The TecOptics FPI-25 Scanning Fabry-Perot Interferometer.) There are somewhat less flexible but easier to use SFPIs from companies like Coherent, Thorlabs, and Toptica Photonics. These provide the following:

    They also have a price tag to match - those from Thorlabs start at around only $2,400 not including the driver box (around $800), others are even more expensive. (Several are covered in subsequent sections, below.) You don't want to ask about the prices of the very flexible SFPIs. :)

    My original challenge was to prove that an SFPI could be constructed that would at least demonstrate the basic principles and possibly even be useful. The results are described in this and the following sections. The first few cost me absolutely nothing (except time) but that wouldn't sound as credible as $1.00 or $2.00 or $3.00. :) Yet many aspects of their performance are comparable to the multi-$k commercial SFPIs. And in some cases, far superior. For later ones, some parts from Thorlabs were used to simplify the construction. Then this endeavor snowballed over the years so that well over a dozen SFPIs have been constructed ranging in size from a 1 inch cube for use with the Mini Laser Mode Analyzer to more than 1/2 meter in length designed to achieve high to ultra-high resolution. These include confocal, selectable FSR mode-degenerate, both spherical and hemispherical, a folded implementation to save space, and others. All are described below.

    The heart of the SFPI is its two mirrors. For longer visible wavelengths (i.e., 600 to 700 nm), the mirrors can be the OCs salvaged from a pair of dead red (632.8 nm) HeNe laser tubes. For other wavelength ranges, mirrors from green (532 nm) DPSS lasers, green or blue ion lasers, HeCd, and other lasers may be useful. While some of these mirrors may have a relatively broad band reflectance, this cannot be counted on. More often than not, the reflectance falls off dramatically beyond 10 or 20 nm from the spec'd wavelength. And, obtaining proper single mode performance of the SFPI without great pain may require that mirrors with specific reflectances and RoCs not normally found in common lasers be used. Of course (gasp!), suitable mirrors can be also be purchased. For common wavelengths, they may be available from companies like CASIX at very reasonable prices. But in general, obtaining the optimum mirror might require ordering a set of custom mirrors. It's not the ground and polished mirror glass itself that will cost a lot of money. They can often be standard concave lenses with suitable curvature available from places like Edmund Industrial Optics or Melles Griot. It's the custom coating, which can easily exceed $1,000, and it doesn't matter that much whether the lot is 2 mirrors or 200 mirrors as what counts is the coating machine time. So, find 99 friends who want to build the same SFPI and the per-mirror cost could still be quite low. :)

    For a short RoC confocal cavity SFPI (more below), the only readily available mirrors I know of are either the misfits I'm using in my $3 SFPI for HeNe lasers (also more below) or mirrors from flowing dye lasers. Unfortunately, the latter tend to have ground, but not polished, outer surfaces. However, since the outer surfaces aren't critical, simply using some index-matching fluid, optical cement, or even common oil or water, between the ground surface and a piece of glass like a microscope slide or cover slip is know to work well enough. It's the coated mirror surface that's important.

    As far as attempting to coat your own mirrors - in two words: Forget it. :) Unless you have access to a dielectric mirror coating machine and know how to use it (and are permitted to use it!), there is no way to produce coatings that will do anything more than provide a hint of what's possible. And these do not turn up on eBay very often. :-) Metal (aluminum, silver, gold) coated mirrors do not work well since their maximum reflection coefficient is around 94 to 97 percent and they have high absorption losses. Thus finesse will be poor and the photodetector signal will be very small. And except for gold, the coatings degrade (tarnish, oxidize) in air without a protective layer, with silver being the worst. For good quality dielectric mirrors, absorption losses only become a major concern for very high reflectivities (perhaps above 99.9%) and modern coatings do not degrade significantly under normal conditions as long as they are not subject to physical abuse or improper cleaning techniques.

    When specifying the mirror RoC (r) for a particular application, it usually makes sense to base it on the maximum frequency range over which there will be action, not simply on the gain bandwidth of the laser(s) being observed. Not only will this result in the best resolution, but doing otherwise may simply not be practical. For common gas lasers like the HeNe and argon ion which have longitudinal modes filling most of their gain bandwidth, (1.5 GHz and 5 GHz, respectively) there's no choice if the display is to be unambiguous. But where the modes have already been limited by an etalon or some other means, only the range of the modes that are present need to fit into the SFPI's FSR. For example:

    I have a variety of inexpensive mirrors suitable for 633 nm SFPIs available on Sam's Classified Page and on eBay under my eBay ID: siliconsam (it's historical). These include short cavity mirrors that result in a truly spectacular finesse at 633 nm. (FSR of 1.78 GHz, finesse may exceed 500!)

    The other major components of the SFPI include the PieZo Transducer (PZT) to move one of the mirrors a few microns, and a photodiode to monitor the output beam.

    High quality PZTs can be purchased at exorbitant cost. But the beeper from a digital watch or similar device will work nearly as well and has the advantage that it runs on much lower voltage than most other types. You never did like that alarm anyhow. :) But there is no need to discombobulate your watch as these piezo elements can be purchased from electronics distributors or surplus places for around $1.00. :) While they aren't quite as linear or have as good a frequency response as the high priced units, these deficiencies don't really matter much for an SFPI. And since they will move several microns on only 50 V, a high voltage amplifier isn't needed as with many commercial SFPIs. The 10 or 20 V p-p output of a typical analog function generator (e.g., "Wavetek") or simple op-amp circuit is quite adequate.

    The photodiode can be almost anything since it needs neither a large area or high frequency response. The salvaged photodiode from a barcode scanner with a 10K ohm resistor load and 10:1 or 1:1 scope probe is often adequate. Where more sensitivity is needed as with very high-R mirrors or low power lasers, a trans-impedance amplifier with high gain can be added since frequency response isn't that critical. Almost any common op-amp will suffice, expecially if multiple stages having modest gain (e.g., 5 to 10) are used.

    Everything else is hardware. The structure and mirror mounts are easily home-built. However, one area where it may be hard to compete with commercial SFPIs is in minimizing the effects of temperature. They typically construct the main support as a cylinder or set of rods made from Invar, a low coefficient of thermal expansion alloy. Some designs further compensate for residual effects by balancing them against those of the PZT resulting a near zero net change in FSR with respect to temperature and/or may include a heater in a closed-loop temperature stablization system. Invar stock is available or can be salvaged from various dead lasers. Some people build SFPIs by mounting the back mirror and PZT in an Invar tube, positioning the front mirror using a 5-axis lab stage, and then gluing it in place permanently when the optimal mirror spacing and alignment has been determined. But glue tends to be too permanent for my taste. :) Constructing the SFPI using Invar rods is nearly as good. But simply enclosing a non-Invar based SFPI in an insulating box will go a long way in reducing temperature effects. And using lower cost "cage" parts from companies like Thorlabs can greatly simplify construction and represents an acceptable compromise.

    Where the objective is to achieve a high finesse and maintain it, then enclosing the entire SFPI to prevent mirror contamination is essential, if not during operation at least for storage. While testing my high resolution SFPI, the finesse had been steadily declining over a few days which turned out to be dust collecting on the mirrors even though their coated surfaces were vertical. Any cleaning of high quality mirrors is to be avoided. Even when using the proper techniques for cleaning of laser mirrors, some permanent degradation of the dielectric coatings is virtually unavoidable. With many cleanings - or only one if the proper techniques are NOT followed - the damage will be enough to result in a noticeable decline in performance. (For information on laser mirror cleaning techniques, see the section: Cleaning of Laser Optics.)

    A triangle (or sawtooth) wave source (it can be a simple circuit constructed for this purpose or a general purpose analog function generator) and oscilloscope (preferably dual trace and/or with an X-Y display mode) will be required to view the scan but needn't be dedicated to the SFPI, so they don't count toward the cost!

    The following sections include general descriptions and photos of over a dozen home-built SFPIs. Schematics for both a photodiode preamp and simple function generator are provided later in this chapter, along with the Mini Laser Mode Analyzer which can replace the function generator and scope where a basic low cost system is desired.

    (From: A. E. Siegman (

    When thinking about producing small and not too fast mechanical motions or pressures, consider also magnetic methods.

    After University Labs in Berkeley introduced the first really low-cost lasers in the early 1970s (priced at circa $300 each rather than the prevailing several thousand dollars and up), it also produced a really neat and equally inexpensive little scanning F-P interferometer with plastic end plates and the scanning mirror driven by what was in essence a miniature loudspeaker coil.

    One of the advantages of the magnetic versus piezoelectric approach is low voltage, higher current drive circuitry, perfectly adapted to IC or semiconductor electronics. Another advantage is wider range of motion.

    Sam's $1.00 Scanning Fabry-Perot Interferometer

    This is the first of several SFPIs I've constructed, differing mostly in the mirrors and their spacing. It uses curved mirrors but is not mode-degenerate, having been built before I knew about such things. :)

    The basic design is shown in Home-Built Scanning Fabry-Perot Interferometer 1. My prototype uses the OC mirrors from a couple of dead Aerotech 1 mW HeNe laser tubes. The PZT is the beeper from some sort of musical greeting card with a 4 mm hole drilled in the center. The photodiode is from a barcode scanner. The frame and mounts are a bit different than those shown in the diagram, above. They were made from the platter clamping plates from some ancient 5-1/4 inch harddrives, hex spacers, and miscellaneous scrap metal. The circular plates are nice because they have predrilled holes with 6-fold symmetry thus simplifying construction. See Photo of Sam's $1.00 Scanning Fabry-Perot Interferometer. (For the mirrors, /V denotes Concave, /P denotes Planar.) Here is a summary:

    The front mirror is removable so other reflectances or RoCs can be tried. The rear mirror is glued to the PZT. The hole was made by placing the PZT on a hard surface (e.g., an aluminum plate) and drilling through it slowly with modest pressure using a normal metal bit in a drill press. The piezo material is more of a compressed powder than a true ceramic so it's possible to grind it away (using the metal drill) with minimal chipping. Thin flexible wires were already attached but if they aren't, solder the top lead near the edge to leave room for the mirror and to minimize any change in elasticity of the top surface. Once soldered, Secure the wires mechanically with a drop of flexible adhesive like 5-minute Epoxy. Also note that the metallization tends to disappear with even modest heat or stress so solder quickly. Conductive paint or silver Epoxy can be used to touch up bare spots if needed but use as thin a layer as possible as it may increase stiffness and reduce response sensitivity in that area. For this reason, DO NOT coat the entire surface with adhesive of any type!

    To perform initial alignment, I used a yellow-orange HeNe laser thinking it would be easier since the mirrors are less reflective away from the 632.8 nm design wavelength. The scatter off of the mirror surfaces was used as the initial means of setting alignment, by minimizing the size of the line or blob formed by the multiple reflections. With a pair of concave mirrors, not only do they have to be aligned with respect to the input beam, they also have to be aligned with respect to each other. In other words, their optical axes must coincide which requires walking them until the scatter pattern is minimized. When misaligned, it will be a line or circle and no amount of adjustment of only one mirror may improve it. Once the initial alignment was done, the PZT could be driven and the output of the photodiode used to fine tune it. In retrospect, using the funny color HeNe laser wasn't necessary as enough red light gets through to be easily seen for alignment purposes. And the display of the modes of that multi-wavelength and multi-transverse mode laser was definitely strange.

    The preliminary results using a Melles Griot 05-LHR-911 HeNe laser were also confusing. This is a 2 mW laser using a tube with about 165 mm between mirrors, corresponding to a mode spacing of 883 MHz. The scope trace in Sam's SFPI Display of Melles Griot 05-LHR-911 HeNe Laser - Initial Attempt shows a jumbled mess due to many transverse modes being excited in the SFPI. The trace on the left should cover a span of approximately three FSRs of the SFPI - about 19.5 GHz. Three clumps that look about the same are clearly visible but the complexity isn't real. The trace on the right is an expanded region of the one on the left. A hint of the modes of the laser can be seen but only a hint. The 05-LHR-911 should have 2 or 3 longitudinal modes at most but the short cavity of the SFPI using long radius mirrors is resonating with multiple transverse modes.

    There is also some hysteresis in the PZT response. It's barely visible on the display as the pattern differs slightly on the positive and negative slopes of the triangle driving function. Using X-Y mode on the scope would show up the hysteresis more clearly. Reducing the sweep speed slightly virtually eliminates the hysteresis. (A 20 trace/second display has minimal hysteresis and is still quite usable. Of course, this wouldn't be an issue with a digital scope

    The overall linearity of the PZT is around 5 to 10 percent over a range of +/-20 V, corresponding to 5 or 6 FSRs of the SFPI. I've actually tested several PZTs (another one was from a digital clock for which the alarm was more of a nuisance than useful!). The response of one is compressed more toward the upper end of the voltage range; the other is slightly compressed at both ends. Within a single FSR, the linearity is probably better than 2 percent and a range of a single FSR provides all the information usually needed. For a system of this type where qualitative information is most important, perfect linearity, especially over multiple FSRs, really isn't a major issue in any case as long as it is known and doesn't change over time. A third PZT was quite linear but had a range of only around 1 FSR of the SFPI - probably due to the excessively thick layer of silver Epoxy I used to cover some bald spots on the piezo disk.

    To confirm that transverse modes were the cause of the complex display and to partially remedy the situation, I aligned the SFPI more carefully by adjusting the front mirror so that the 05-LHR-911 beam bounced directly back to the source with dancing interference patterns, then aligned the rear mirror for maximum amplitude of the displayed signal, and added an aperture about 0.3 mm in diameter (a pin hole in a piece of aluminum foil) inside the SFPI cavity. The aperture was mounted on a micropositioner but could be installed permanently so that doesn't blow my budget. :) The results are shown in Sam's SFPI Display of Melles Griot 05-LHR-911 HeNe Laser. The sequence of the six traces show the modes of the 05-LHR-911 cycling over time as they move under the HeNe gain curve. The horizontal scale is the same as in the jumbled mess trace, above, but the transverse modes have been almost entirely eliminated. The distance between similar peaks (2.2 boxes on the screen) is the FSR of the SFPI - about 6.5 GHz. The distance between longitudinal modes (0.3 boxes) is the 883 MHz FSR of the 05-LHR-911. The math even works. :) So, this represents success of sorts but alignment of everything is super critical and any vibrations - even the audio from a radio - create havoc with the display. There is also a quasi-periodic fluctuation in amplitude of all the displayed modes with no corresponding power fluctuations in the laser. I suspect this to be due to residual mode competition in the SFPI as the frequency of the modes changes relative to the SFPI cavity, possibly a side effect of the aperture.

    Sam's SFPI Display of a Melles Griot 05-LHR-151 HeNe Laser shows the result using the same setup for a longer laser with more closely spaced modes - 436 MHz compared to 833 MHz for the 05-LHR-911. With this higher power laser, there are still some non-TEM00 modes just visible in the display but they are fairly low level. Sam's SFPI Display of Vertically Polarized Modes of Melles Griot 05-LHR-151 HeNe Laser shows the effect of inserting a polarizing filter into the beam. Since adjacent modes tend to be of orthogonal polarization in randomly polarized HeNe lasers, every other mode on the display has disappeared.

    Finally, I tried a Spectra-Physics model 117A HeNe laser head, which when used with its mating controller is a frequency or intensity stabilized (single longitudinal mode) laser. I'm running it on an SP-248 so it's not stabilized but the modes are a bit interesting. The mode spacing is around 600 MHz which is consistent with a 2 to 3 mW HeNe laser. However, as the modes cycle, there isn't a smooth progression through the gain curve. It almost seems as though having exactly 2 modes is enhanced somehow and that it's very unlikely to see 1 or 3 modes. When 1 or 3 modes would be expected to pop up, they might appear very briefly, or be skipped entirely in favor of the 2 modes one of which is on the opposite side of the gain curve. The polarizations of the modes also appear to be of the "flipper" variety, changing suddenly rather than staying with a particular mode. I don't know if this behavior is by design. However, since orthogonally polarized modes are sensed by a pair of photodiodes in the laser head and used for stabilization, strong mode pairs could be beneficial.

    After determining experimentally that an aperture helped but didn't totally eliminate the transverse mode problem, a Post Doc in our lab wrote a simple Matlab program to calculate Hermite Gaussian transverse mode profiles given the mirror RoCs and the distance between mirrors. Plugging in the long radius SFPI cavity configuration revealed that the TEM00 and TEM10/01/11 modes have a high degree of overlap regardless of axial position. So, any aperture that suppresses them very effectively would also result in unacceptable attenuation of the TEM00 mode. So, on to plan B. :) I hope to have a compiled version of this program available in the near future as it appears to be quite useful for visualizing cavity modes in general.

    Here is a summary of the configurations I've tried so far on the $1.00 SFPI:

    Of these, the first is probably the best choice unless super high resolution is needed. All except the flat-flat required an aperture inside the SFPI cavity to suppress non-TEM00 (transverse) modes.

    Sam's $2.00 Scanning Fabry-Perot Interferometer

    Well, it wasn't actually $2.00. :) I found some small radius mirrors originally intended for a research project that is now defunct. These should work well in a confocal configuraion in the green region of the spectrum free of those annoying transverse modes!

    The mirrors were actually Melles Griot plano concave lenses custom coated (along with a batch of microchip laser crystals) for 1,540 nm. Now, it's perhaps a not so well known fact that a dielectric mirror coated for a wavelength of X nm will also perform reasonably well at a wavelength around X/3 nm (think of a stack of 3/4λ layers instead of 1/4λ layers). The actual reflectance function will depend on the design of the original mirror including the number of layers and uniformity of the layer thickness. The reflectance at the new wavelength will almost certainly be lower and the losses may be slightly higher. But with luck, these mirrors will be useful in a wavelength range centered around 513 nm (1,540/3).

    I had two types available: Those that were supposed to be 98 percent as OC mirrors and those that were supposed to be HR mirrors, both at 1,540 nm. Here are how they performed at the two green wavelengths of interest:

                             Reflectivity at       Reflectivity at
         Mirror Type       532 nm (Green DPSS)   543.5 nm (Green HeNe)
       OC (98%@1,540nm)          97.8%                    88%
       HR  (HR@1,540nm)          99.8%                    99%

    For 532 nm, neither is really ideal. The "OC" is a bit low - I would have preferred around 99% to achieve a higher finesse. However, 97.8% is still decent. The reflectance of the "HR" - which could be even higher than the measured 99.8% since the 0.02% transmission measurement was not very accurate - might be too high to get a decent signal but could result in a very high finesse. But at 543.5 nm, the "HR" mirror seems to be perfect.

    The only thing not wonderful about these mirrors is that the planar side isn't AR coated. (Since they were intended only for some tests, we saved money by not having AR coating!) But, if they are slightly tilted, hopefully, this won't be a major problem.

    There are also several radii to choose from. For the first version, I used the longest RoC which is a Melles Griot 01-LPK-01. This is a 10 mm diameter BK7 lens with a focal length of -20 mm which has a RoC of about 10.3 mm. (For BK7, the RoC of a plano-concave lens is -0.517 of the focal length.) This results in an FSR of about 7.8 GHz. Note that the FSR is c/(4*d) for the confocal cavity, one half that of the long radius or planar SFPI cavities. See the previous section. So, these will be good for all green HeNe lasers and longer cavity single mode green DPSS lasers like the C315M and C532, as well as that Far East disaster described in the section: Reconstruction of an 80 mW Green DPSSFD Laser. However, short cavity DPSS lasers including green laser pointers, the Uniphase uGreens, MCA based DPSS lasers, and possibly the Transverse TIM622 will require a shorter SFPI cavity. The other sets of mirrors go down to around a 5 mm RoC so another version may be built with a set of these.

    However, note that since the gain bandwidth of Nd:YAG and Nd:YVO4 is over 150 GHz and the SHG green conversion also doubles the frequency between modes, multimode solid state lasers may have frequencies which greatly exceed the FSR of these medium length SFPI cavities. Unambiguous display of their modes may require an SFPI with an FSR of more than 300 GHz - a cavity length of 0.25 mm for the confocal configuration! It's simply not impractical to grind and polish mirrors with very small RoCs. The limit is about 2.5 mm for an FSR of 30 GHz. Fortunately, what's often most important is to confirm single or maybe dual longitudinal mode performance so a much smaller FSR is adequate and desirable for maximum resolvance. With a bit of care in interpretation, almost any FSR will be fine for this purpose.

    The mechanical configuration is similar to the $1.00 SFPI except that the rear mirror mount can be moved along the optical axis on threaded rods to match the mirror distance to the RoC of the mirrors. A diagram along the lines of the simple design of the $1.00 SFPI is shown in Home-Built Scanning Fabry-Perot Interferometer 2. Again, mine was constructed of cast off disk drive parts and other miscellaneous junk. :) The first photodiode I used for this SFPI was a $2.00 part from Digikey - which would have been my total cost if it hadn't already been in one of my random stuff drawers. :) And, the frame is a bit shorter since the RoC of all of these mirrors is so small. Please see: Photo of Sam's $2.00 Scanning Fabry-Perot Interferometer.

    For the initial test, I am using the HR mirror set with an 05-LGR-151 green HeNe laser head. Since this is a less than 0.5 mW output laser and the sensitivity of silicon photodiodes at 543.5 nm is somewhat lower than at 632.8 nm, detection is more difficult.

    Furthermore, in order for the SFPI to be mode-degenerate, the mirror spacing really has to be quite close to the RoC for the confocal configuration. Since these were originally lenses and not mirrors, the exact RoC is not really known. OK, the real story is that I didn't locate the part numbers of the lenses until after I did the initial construction and wrote this paragraph! There are many ways to determine the actual RoC of the mirrors. A collimated beam can be reflected from the mirror at a slight angle. The focal point will be at a distance of one half the RoC. Alternatively, a point source like a bare visible laser diode can be imaged back onto itself from the mirror. Then, the RoC is the distance to the mirror. However, any such measured RoC is only approximate. For the SFPI to be mode-degenerate, it needs to be quite precise and this can only be determined experimentally.

    The mirror alignment itself isn't super critical. It's best to have a way of changing mirror distance without affecting alignment very much but simple three-screws adjusters work just fine. The laser used for the alignment should have a known spectrum if possible, preferably a single longitudinal mode. As the correct distance is approached, the little peaks from all the modes of the not quite confocal cavity - which may indeed be very small or undetectable - will gradually merge into one peak whose amplitude will increase and width will decrease dramatically.

    Note that the MDI doesn't eliminate higher order transverse modes. It only assures that they will appear at the same locations on the display as the TEM00 modes. If the distance between the mirrors isn't close to the RoC, there will be higher order modes at essentially random frequencies relative to the TEM00 modes. The result will be very low fringe contrast in the output as the PZT voltage is varied and lumps all over the place in the display. However, as the correct distance is approached, these will approach the TEM00 modes. Visually, if the distance between the mirrors is moved slowly with the PZT around the optimal distance, the output beam from from the SFPI (going to the photodiode) will flash on and off uniformly across its entire width, while on either side there will be concentric rings of light and dark sweeping from center to edge or vice-versa. It's actually quite remarkable that varying the PZT voltage by hand (ramp turned off), the output of the SFPI can be tuned to all light or all dark very precisely when the distance is just right. In addition, alignment of the SFPI relative to the laser is very easy. The reference I am using is to adjust the the reflection from the planar surface of the front mirror to be just below the output aperture of the laser, then adjust the position of the beam (without changing its angle) to center the reflected blob from the curved rear surface of the front mirror.

    After some fiddling, I am able to see the modes of the 05-LGR-151, though the signal is extremely low level and the finesse is poor. In addition, the modes appear to be somewhat distorted - possibly due to the distance between the mirrors not being quite correct. Switching the function generator to DC output mode and adjusting the voltage through the modes of the HeNe laser shows a very complex transverse mode pattern which is clearly not degenerate even when the mirror distance is very close to optimal. I don't know if this is due to the distance still not being perfect (commercial SFPIs are set to within a few um) or due to poor accuracy in the spherical shape of the mirrors. Focusing the beam improves the resolution and amplitude of the signal somewhat or just due to the nonuniformity of the coating which results in the reflectance decreasing from center to edge. A modest size aperture (perhaps 1 mm) will probably help to eliminate many of the higher order mode since they are quite spread out.

    Up to this point, my conclusions were mixed. Yes, the jumbled peaks were gone. And, alignment is definitely much less critical - once the distance of r is found, any two of the three rear mirror mount nuts or mirror adjusters can easily peak the output in no time flat. But, the resolution is lower than my $1.00 SFPI - between 50 and 100 MHz, compared to better than 25 MHz. Whlte the larger FSR means that the resolution will not as fine for the same finesse, another factor may be the quality of the mirrors (or lack thereof, actual specs unknown). A focusing lens (see below) and modest size intracavity aperture will help somewhat. And a photodiode preamp will help make alignment easier. As long as the reflections from the various front optics don't return to the HeNe laser, the modes are quite stable. However, very obvious instability results if a major portion of the reflected HeNe beam hits the laser's output mirror. Then, wild mode fluctuations appear in the SFPI display - some modes may momentarily double in amplitude or disappear entirely. And visible power fluctuations are also visible in the beam and interference patterns.

    The next step will be to add a proper focusing lens as shown in the $2.00 SFPI diagram (there is none in the one in the photo). Presently, the curved surface of the front mirror results in a large diverging effect on the input beam. Using a long focal length lens helps somewhat. But in a test using a short focal length positive lens mounted in a spring clothspin on a micropositioner helps even more. This cancels out the negative curvature of the front mirror and adds some additional focusing to match the TEM00 mode of the confocal cavity. The signal amplitude increases by at least a factor of 2 and the resolution also improves.

    Eventually, I will probably construct a preamp for the photodiode to provide an adjustable gain of up to 1,000 using a couple of op-amps. This will greatly ease alignment since the height of the signal on the scope on its most sensitive setting with a 10X probe now is only about 1/2 cm at best using the low power green HeNe laser. A possible design is shown in Adjustable Gain Photodiode Preamp. (Frequency compensation capacitors which may be needed for stability are not shown.) The gain is variable from 0.1 to 1,000 compared to the bare phododiode feeding a 10K ohm load. A gain of 10 would be sufficient so this should have enough headroom for other lower output power lasers and/or higher reflectance mirrors.

    However, for now, I just replaced the 10K phododiode load resistor with a 100K pot and substituted a 1X probe for the 10X probe. This resulted in more than enough sensitivity even for the low power green laser while maintaining adequate frequency response.

    Finally, I installed a 9 mm focal length focusing lens as shown in the diagram. This results in a collimated input beam coming to a focus inside the cavity (the focal length of the lenses being used for the mirrors is -20 mm).

    And then it was perfect. :) Well not quite perfect - the finesse isn't much better but it is quite stable, there is no evidence of unwanted ghost frequencies, it is easy to align, and all in all, works quite well. With careful alignment and centering of the input beam, I was even able to achieve the situation where the FSR became c/(2*d) or 14.6 GHz. In this case, every other mode display per sweep of the SFPI nearly disappeared with the remaining ones almost doubling in amplitude.

    The finesse is probably not as terrible as I'm implying. For my 99 percent mirrors, the theoretical finesse is a bit over 150. So, 14.6 GHz divided by 150 is about 100 MHz which is close to what I've measured. And, as noted, it's quite possible the mirrors are actually somewhat less reflective than the 99 percent being used for the finesse calculation.

    This SFPI can be used to easily test most DPSS green (532 nm) CW lasers for single frequency operation. It's easier to set up than a commercial SFPI with a separate ramp generator/preamp box as my Wavetek function generator is always there on top of the scope. :) The high reflectivity of the mirrors for 532 nm turned out to not be a problem. The ~14.6 GHz FSR is large enough to display unique modes for the C215M, C315M, C532. While the cavities of the uGreen and LWE-142 lasers are very short and have a higher FSR, it's still possible to detect spurious non-single frequency operation since the extra modes will not be stable or have a fixed relationship to the primary mode.

    As a free bonus, the same SFPI can also be used for 1,5XX nm lasers by swapping the photodiode. When I became obsessed with the desire to look at the longitudinal modes of a Melles Griot 05-LIR-150 1,523 nm HeNe laser, there didn't seem to be too many options. None of my other SFPIs (home-built or commercial) would work beyond 900 nm. But then it occurred to me that I already had this SFPI using mirrors coated for 1,540 nm. Being HR at that wavelength (and probably close to HR at 1,523 nm), getting a signal might be quite a challenge, but aside from the near impossibility of lining everything up with the <1 mW beam from the IR laser, it was worth a shot.

    I did have an IR photodetector for a Newport power meter, so Simply removing the existing PD board would allow the beam to exit the back of the SFPI. To have the most flexibility, the PD preamp from an SP-476 SFPI driver was used (but not the HV scan output, though that could also have been used in place of the function generator with a resistor divider to reduce the maximum voltage). After a bit of fiddling with room lights out (any bit of fluorescent light overwhelmed the signal and/or added 120 Hz ripple), a really messed up display was obtained with all sorts of ringing and garbage (technical term!). However, it was possible that this was due to the detector being designed for a laser power meter with a smoothing capacitor or something else inside.

    I had just rediscovered the use of cut-open germanium transistors as sensors for 800 to 1,800 nm. With one of those, despite it's somewhat low sensitivity, the modes of the 1,523 nm laser appeared in all their glory. The finesse is only between 100 and 150, but it is more than adequate for displaying IR HeNe laser modes. I don't do Telecom. :) Since I now have a SP-470-03 which works from 532 nm through at least 650 nm, that's what's used for routine laser testing and characterization. So, I'll probably leave the IR PD permanently in the $2 SFPI and dedicate it for IR lasers.

    In 20:20 hindsight, the mechanical design of a confocal SFPI can be considerably simpler than what I created. With careful alignment of the mirrors during mounting and glueing (for the one on the PZT, which has wedge), no adjustable alignment is really necessary. So, it becomes a pair of plates on threaded rods. The front mirror would simply be clamped or glued to the front plate and the back mirror would be glued to the PZT, which is clamped or glued to the back plate. After initial setup setting the precise confocal spacing, alignment is done by X/Y (pitch/yaw) adjustments of the mounting for the SFPI head. This can be a fancy kinematic or spherical mount, or simply a platform with adjustable screws for feet.

    Sam's $3.00 Scanning Fabry-Perot Interferometer

    About a year after building my $2 SFPI, I came across some other short radius mirrors:

    At first I thought these were for some Spectra-Physics dye laser. But thinking about it, I'm now inclined to believe they were a HeNe laser mirror goof. The specifications called for 43 cm RoC mirrors and someone dropped a factor of about 10 between design and manufacturing. (My measurement may be off by a couple of mm, so they could indeed be 45 mm mirrors.) Hey, if NASA can goof up units, so can a laser company! How else to explain that there were literally thousands of these available surplus at one time. SP never sold that many dye lasers, but production runs of thousands of HeNe laser tubes for barcode scanners at the peak of their popularity would not have been unusual. Also, SP's dye laser pump mirrors with short RoC mirrors tended to have the non-reflective side fine ground (not polished and AR coated as with these). Also, the SP dye laser mirrors I've seen have an RoC of about 50 mm, not 43 mm. Regardless of the origin, I'm not complaining. The person I got the mirrors from insists they are HeNe mirrors and will even send me a laser tube that uses them if he can find one. In principle, I suppose that is possible but it would have to be a very peculiar resonator configuration with a focal point inside. I won't hold my breath in anticipation. :) He had been selling them on eBay (sorry, no more available from there!) and had so many that he was using them as decorative stones in his tropical fish tank. Transgressions like that really need to be punished! :-)

    The mirrors were installed in a slightly stretched frame to enable the longer 43 mm resonator length as shown in Home-Built Scanning Fabry-Perot Interferometer 3. It was then a simple matter to get this rig to work with much better finesse. That is, after I realized two things:

    1. The focusing lens from the $2.00 SFPI had too short a focal length for the much longer cavity and was smearing out and reducing the amplitude of the response.

    2. The confocal distance was indeed 43 mm and not 38 or 40 mm as I originally thought. At 38 mm, the SFPI initially appeared to work but the display wasn't stable at all, mode amplitudes varied depending on where they were on the ramp voltage, and the photodiode signal was quite weak. Once it was adjusted at 43 mm, the display looked very much like the one in a textbook. :)

    The only problem with this SFPI for use with HeNe lasers is that the Free Spectral Range (FSR) for the mode-degenerate confocal configuration is c/(4*d), which is only about 1.75 GHz for the 43 mm cavity. This is just barely more than the Doppler broadened gain bandwidth of the HeNe laser, about 1.5 GHz. So, there can be some confusion when interpreting lasing lines on the tails of the gain curve, though this is minor. However, a benefit is that the 1.75 GHz FSR provides nearly the largest useful resolution by almost filling the FSR with the HeNe laser modes.

    I have a set of basic parts available for building a similar SFPI. Sorry, it will cost more than $3 though. :) More information can be found at Sam's Classified Page.

    See W's Scanning Fabry-Perot Interferometer Page for an SFPI using these same mirrors (as well as others for other wavelengths). His mechanical setup uses parts that are a bit more professional and several orders of magnitude more expensive than mine though. Yet, he complains about instabilities that my resonator frames constructed from recycled harddrive parts and Home-Depot hardware don't have. :)

    As with the $2 SFPI, simplification of the mounting is indeed possible.

    I later more fully tested one sample of these mirrors for reflectance at various wavelengths:

                    Measured     Predicted
      Wavelength   Reflectance    Finesse
        544 nm        60%?         >2
        594 nm        99.4%        >250
        633 nm        99.7%        >500
        655 nm        99.9%        >1000
        680 nm        99.0%        >150

    However, there could be major variations in the reflectance from one mirror to the next between lots or even within a lot, especially for other than the design wavelength of 633 nm.

    A finesse of more than 600 has been observed using these mirrors at 633 nm more than 250 at 594 nm. Performance is abismal at 544 nm (and even worse at 532 nm), but may still be adequate to confirm that a laser is single longitudinal mode (single frequency).

    Sam's Plane Mirror Scanning Fabry-Perot Interferometer

    The planar-planar (P-P) cavity is a configuration that will not support higher order transverse modes at all. However, it is only borderline stable for the TEM00 mode and extremely difficult to set up and align as the two mirrors must be parallel to a very high precision AND the input beam must be orthogonal to the input mirror to get decent performance. The latter requirement is particularly troublesome in that without an optical isolator, that any reflections will go directly back to the source. Since the SFPI mirrors are nearly 100 percent reflecting most of the time (except when in resonance), this means a nearly total return. And most lasers get modestly to really annoyed when any of the output beam is reflected back into their cavity, resulting in a variety of instabilities and changes in mode structure. However, where it is desired to either have an FSR for which confocal mirrors are not available - which is most FSRs without custom mirrors - or to be able to adjust the FSR for various applications, there is no choice.

    I built a P-P SFPI using a similar structure to that my others inexpensive home-built "instruments". Unfortunately, Murphy took no days off and ALL of these problems were present. I selected an FSR of around 5 GHz (25 mm mirror spacing) so it would unambiguously display modes of a two-frequency Zeeman HeNe laser like an HP-5517B. (Due to the Zeeman splitting, these lasers can potentially have lasing modes over a much wider bandwidth than the normal 1.6 GHz or so of the Doppler-broadened neon gain curve. Think of a pair of neon gain curves that are shifted by several hundred MHz with respect to each-other, thus making the available range larger.) None of my other SFPIs would be entirely suitable. My home-built one for 633 nm had an FSR of only 1.7 GHz and the Spectra-Physics 470-03 (see below) has an FSR of 2.0 GHz. A Zeeman laser could easily have lasing modes over a bandwidth of 3 GHz or even more, especially one that has been rebuilt improperly. Where the behavior is not what is expected, the aliasing in a narrow range SFPI would make interpretation of what's actually going on very difficult.

    The planar mirrors used were near-HR (99.8%) at 633 nm with no AR coating but ground with wedge so the reflections from the uncoated surface would not interfere with the SFPI operation. But the home-built mirror adjusters that were perfectly adequate for the confocal SFPI were barely marginal for this one, requiring very careful tweaking for best response. And they didn't want to remain aligned for more than a few minutes. But worst of all, without an optical isolator, the modes of the usually well behaved 05-LHP-151 laser head being used for testing were jumping all over the place due to the back-reflections. And there were never more than 3 modes present, generally only 2, and sometimes only a single mode. Normally, there would be 4 or 5 modes at all times for this 5 mW laser.

    During those periods where it behaved, the performance was quite acceptable easily resolving the 05-LHP-151's longitudinal modes (spacing of 438 MHz) with a factor of 3 or 4 to spare. This would have been more than enough for use with the short laser tubes (mode spacing of 1 GHz or more). But the needs for super-precise alignment and virtually unavoidable back-reflections makes this impractical for my intended application of easily analyzing the modes of a variety of HP/Agilent lasers and home-built equivalents.

    Sam's Hemispherical Confocal Scanning Fabry-Perot Interferometer

    When I first acquired a Tropel 2440 SFPI head with a planar front mirror and curved rear mirror, I had my doubts that such a configuration was valid. All other commercial SFPI heads I'm aware of are built with a pair of identical curved mirrors, nearly always in a normal confocal configuration. Whlie some general purpose instrumentsdesigned for masochists :) use planar mirrors, they don't have separate self-contained SFPI heads. Perhaps it was simply a short hemispherical cavity and precise alignment was necessary to avoid higher order modes in the display. Or, perhaps someone had replaced the curved front mirror with a planar mirror for reasons unknown. But that seemed extremely unlikely. Based on what could be seen of the curved rear mirror after removing the front planar mirror, the cavity length looked about right to be one half its RoC. In fact, the Tropel 2440 is what got me thinking about the Hemispherical Confocal (HC) cavity SFPI in the first place - then having serious doubts about its functionality.

    Although the HC cavity would have a resonance as a result of the reflection from two passes up and back to the curved mirror via the planar mirror, there would also be the normal resonance between the front and rear mirror. So, why wouldn't this one dominate and screw up the behavior? If valid, the FSR should be the same as for a normal confocal configuration with mirrors having an RoC equal to d and spaced d apart (FSR=c/(4*d), but the HC cavity would be half as long and the PZT sensitivity should double, requiring only half as much voltage to cover one FSR.

    As a test, I replaced the front mirror in my home-built plane mirror SFPI (above) with a 4.3 cm RoC mirror like those used in my normal confocal SFPIs (see the sections: Sam's $3.00 Scanning Fabry-Perot Interferometer and Scanning Fabry-Perot Interferometer Head Frame using Commercial Parts), as well as in the SFPI kits available on Sam's Classified Page. (The rear mirror is a salvaged internal mirror HeNe HR.) And this actually works quite beautifully. :) It was quick and easy to set the cavity spacing and obtain a clean display. Alignment is non-critical and similar to that of the normal confocal configuration. As predicted, the FSR remained unchanged at around 1.7 GHz and the PZT sensitivity (V/FSR) appears to have doubled, tuning through up to 8 or 10 FSRs using a Wavetek function generator with a maximum output of around 20 V p-p. Initially, no mode matching lens was used so the finesse wasn't as high as it should be based on theory but if the planar mirror were in front, mode matching would be less critical and a lens would probably not be required. But since this was really only done as a test, replacing the front mirror was much simpler, not being glued to the PZT. And dusting off the planar mirror on the original SFPI assembly that had been sitting in a dusty closet for several years made a dramatic improvement. Can you believe that? :-) With a 2 inch focal length lens positioned near the front mirror, the finesse is at least 100, perhaps over 150, which still may be limited by the recycled planar mirror. See Sam's Hemispherical Confocal Scanning Fabry-Perot Interferometer. Note the spacing of the mirrors of around 0.85 inches (one half of the 43 mm RoC). (As a reference, the distance between the two front plates is approximately 1 inch.) A "third hand" holds the randomly selected mode matching lens. With the rear mirror being HR, the photodiode signal is slighly weaker than when a lower reflectance mirror is used but not by that much. The photodiode preamp of the SP-476 is being used, but on its least sensitive setting, but as with the other open-frame SFPIs, a cover may be necessary to block room light (not shown).

    When alignment is near perfect (close to being centered and aligned with the optical axis), instead of every other FSR dropping out and the others doubling in amplitude as with the normal confocal SFPI, the pattern is over 4 FSRs or pairs of FSRs and even 3 of the 4 dropping out and the remaining one quadrupling in amplitude. Other than that, it would be difficult to tell the two apart from general behavior. All in all, quite fascinating. ;-)

    Sam's Selectable FSR Hemispherical and Spherical Mode-Degenerate Scanning Fabry-Perot Interferometers

    I figured it would be interesting to build a system of the type described in the section: Selectable FSR Mode-Degenerate Fabry-Perot Interferometers. However, using the usual 4.3 cm (1.7 inch) RoC mirrors would require the spacing to vary over almost a 3.5 inch range, with fine adjustment at any given position. This could be done with a 4 inch linear slide and separate micrometer positioner. I had even dragged out some salvaged LaserDisc player slider bearing rails that could actually be used to build a nice linear slide with a 4 inch travel. An interferometer-based positioning system could be used to set the spacing down to the angstrom, but that's a bit more complexity than I had in mind! ;-) Then it occurred to me that using the hemispherical version of the mode-degenerate cavity would only require half the range. This was attractive because: (1) I just happened to have a compact Newport linear positioner with a 2 inch travel and (2) the hemispherical cavity differs from the one presented in the paper and conceivably has not been built before. This would enable me to at least pretend that I'm doing something new! ;-)

    Here are all the resonances up to N=10 for a selectable FSR hemispherical cavity SFPI with a planar mirror and 4.2 cm RoC curved mirror, both having a reflectance of approximately 99% corresponding to a planar finesse of 300. The values of FSR, FWHM, and Finesse are all relative to those of the planar cavity where finesse is π/(1-R) assuming R is close to 1:

                      <---- Relative ---->  <--- 4.2 cm RoC (5) ---->
          Num         (2)     (3)    (4)      d     FSR    FWHM  Fin-
      N k Rep  d/r    FSR    FWHM  Finesse  (cm)   (GHz)   (MHz) esse  Notes (1)
      1 0  1  0.000  1.000   1.000  1.000   4.200  3.569   11.9  300  Planar 1-0
      1 1  2  1.000  0.500   2.000  0.500   4.200  1.785   11.9  150  HS (1-1)
      2 1  4  0.500  0.500   4.000  0.250   2.100  1.785   23.8   75  HSC (2-1)
      3 1  6  0.250  0.667   8.000  0.167   1.050  2.379   47.6   50
      3 2  6  0.750  0.222   2.667  0.167   3.150  0.793   15.9   50
      4 1  8  0.146  0.854  13.657  0.125   0.615  3.046   81.2   38
      4 3  8  0.854  0.146   2.343  0.125   3.585  0.523   13.9   38
      5 1 10  0.095  1.047  20.944  0.100   0.401  3.738  124.6   30
      5 2 10  0.345  0.289   5.789  0.100   1.451  1.033   34.4   30
      5 3 10  0.655  0.153   3.056  0.100   2.749  0.545   18.2   30
      5 4 10  0.905  0.111   2.211  0.100   3.799  0.395   13.2   30
      6 1 12  0.067  1.244  29.856  0.083   0.281  4.440  177.6   25
      6 5 12  0.933  0.089   2.144  0.083   3.919  0.319   12.8   25
      7 1 14  0.050  1.443  40.391  0.071   0.208  5.149  240.3   21
      7 2 14  0.188  0.379  10.624  0.071   0.791  1.354   63.2   21
      7 3 14  0.389  0.184   5.145  0.071   1.633  0.656   30.6   21
      7 4 14  0.611  0.117   3.272  0.071   2.567  0.417   19.5   21
      7 5 14  0.812  0.088   2.464  0.071   3.409  0.314   14.7   21
      7 6 14  0.950  0.075   2.104  0.071   3.992  0.268   12.5   21
      8 1 16  0.038  1.642  52.548  0.063   0.160  5.861  312.6   19
      8 3 16  0.309  0.202   6.480  0.063   1.296  0.723   38.5   19
      8 5 16  0.691  0.090   2.893  0.063   2.904  0.323   17.2   19
      8 7 16  0.962  0.065   2.079  0.063   4.040  0.232   12.4   19
      9 1 18  0.030  1.842  66.327  0.056   0.127  6.576  394.5   17
      9 2 18  0.117  0.475  17.097  0.056   0.491  1.695  101.7   17
      9 4 18  0.413  0.134   4.841  0.056   1.735  0.480   28.8   17
      9 5 18  0.587  0.095   3.408  0.056   2.465  0.338   20.3   17
      9 7 18  0.883  0.063   2.265  0.056   3.709  0.225   13.5   17
      9 8 18  0.970  0.057   2.062  0.056   4.073  0.204   12.3   17
     10 1 20  0.024  2.043  81.727  0.050   0.103  7.292  486.1   15
     10 3 20  0.206  0.243   9.704  0.050   0.866  0.866   57.7   15
     10 7 20  0.794  0.063   2.519  0.050   3.334  0.225   15.0   15


    1. These are the named cavities: Planar, HemiSpherical (HS), and HemiSpherical Confocal (HSC). The planar cavity isn't strictly part of this series but is included for completeness, and as a baseline for comparison. Its mirror RoCs would be infinite but its mirror spacing (d) is equal to the RoC (r) used for all the spherical cavities. Configurations for N above 10 would have larger maximum values for the FSR and FWHM (k=1), and smaller minimum values for the FSR (k=N-1), with the FWHM approaching that of the normal hemispherical cavity (1-1).

    2. FSR takes into account the actual value of d as the cavity spacing is varied. The values shown are relative to that of the planar cavity (with mirror spacing equal to RoC=r of the mirrors used in the spherical cavities). Then, FSR(N)*2*d/r=1/N, or FSR(N)=r/(2*d*N).

    3. FWHM is the width of the spectral peaks relative to that of the planar cavity based on the mirror reflectivity (R) and mirror spacing (d).

    4. Finesse is with respect to the effective FSR, relative to that of the planar cavity and scales as 1/(2*N).

    5. This shows the resulting mirror spacing (d), FSR, FWHM of the spectral peaks, and finesse for each hemispherical cavity with mirror RoC=4.3 cm compared to a planar cavity with a mirror spacing of d=r=4.3 cm for an FSR of around 3.5 GHz and a finesse of 300 (mirror reflectivity of approximately 99 percent).

    The completed unit is shown in: Sam's Selectable FSR Hemispherical Mode-Degenerate Scanning Fabry-Perot Interferometer. The ~4.3 cm RoC 99.5%@633nm rear mirror is glued to a PZT attached to the front of a 1/2" kinematic mount with the photodiode mounted on a piece of plastic on the back. This assembly is fastened to a Newport 422-1S Miniature Ball Bearing Linear Stage, enabling the mirror spacing to be varied over a 2 inch range with a thumbscrew. (The actual continuous range with the thumbscrew is only about 1.5 inches, so a 1/2" block must be placed between the tip of the screw and carriage to achieve the final 1/2". The specifications for the 422-1S actually only lists a 1 inch travel but that's probably for the guaranteed alignment tolerance - it is more than acceptable over the entire 2 inch range for the SFPI.) The planar 99.8%#633nm ("near HR") stationary front mirror is in an adapter in a similar 1/2" kinematic mount. The adjustable mounts aren't absolutely essential but they were available and too ugly to use for most other applications. :) And having fine control of mirror alignment should help to optimize performance. Everything is on a heavy baseplate with 3 adjustment feet. A scale on the baseplate enables the coarse position to be set easily. Space has been reserved in front of the stationary mirror for a mode matching focusing lens, but this will hopefully not be required due to the hemispherical cavity.

    The results are encouraging. Setup was very easy with alignment using the baseplate screws and fine tuned with the mirror mounts, and a shim placed under the rear mirror mount to match the heights of the mirrors. Near optimal alignment can be maintained over the entire travel of the movable mirror.

    However, while quite usable, the finesse is still rather disappointing, being less than one half of the value predicted by theory. The only way that seems to improve it somewhat is to add a lens to focus the beam directly at the surface of the front mirror. For the specific case of the hemispherical confocal cavity length of r/2, the first bounce back from the curved mirror should do exactly that. But for some reason, it's not as good. So, I considered the possibility that there needs to be another factor of two reduction in finesse thrown in that is not being accounted for by the equations. But this doesn't appear all that likely. Consider the following thought experiment starting with the normal confocal cavity: Take one mirror and rotate it 90 degrees so it is at a right angle to the other mirror. Now add an ideal planar mirror at 45 degrees in between them and adjust the total cavity length to be the same as before (r=RoC). Then the logical geometry really hasn't changed and the behavior should be identical. And with the normal off-axis alignment of the input beam, it should be possible to further rotate the curved mirror from 90 to 180 degrees so the two curved mirrors merge into one to create the hemispherical confocal cavity, again without changing the geometry. The only thing the ideal planar mirror does is to flip the "handedness" of the intracavity beam, but it's not clear how this could affect anything related to finesse unless eliminating the independence of the two curved mirrors somehow causes input alignment to be super critical. Now that'sreally grasping for optical straws! :)

    I even tried a real planar HR to see if that would help. But after fighting with alignment due to the low signal, there was no dramatic improvement in finesse beyond what could be accounted for with 99.9+%R compared to 99.8%R.

    And to confirm that I wasn't just unlucky having selected mediocre mirrors, the front planar mirror was then replaced with one identical to the rear mirror (~4.3 cm RoC, 99.5%R@633 nm) thus converting this rig into a selectable FSR spherical SFPI. The range of travel is not sufficient to allow for much above the confocal cavity length, but does permit all the shorter ones to be selected. And as a practical matter, most of the possible FSRs with longer than confocal spacing are not all that useful anyhow (or shall we say are even less useful than the others).

    Here are all the resonances up to N=10 for a selectable FSR spherical cavity SFPI with 4.2 cm RoC mirrors having a reflectance of approximately 99% corresponding to a planar finesse of 300. The values of FSR, FWHM, and Finesse are all relative to those of the planar cavity where finesse is π/(1-R) assuming R is close to 1:

                      <---- Relative ---->  <---- 4.2 cm RoC (5) --->
          Num         (2)     (3)    (4)      d     FSR    FWHM  Fin-
      N k Rep  d/r    FSR    FWHM  Finesse  (cm)   (GHz)   (MHz) esse  Notes (1)
      1 0  1  0.000  1.000   1.000  1.000   4.200  3.569   11.9   300  Planar 1-0
      1 1  1  2.000  0.500   0.500  1.000   8.400  1.785    5.9  300  Sphere 1-1
      2 1  2  1.000  0.500   1.000  0.500   4.200  1.785   11.9  150  Confoc 2-1
      3 1  3  0.500  0.667   2.000  0.333   2.100  2.379   23.8  100
      3 2  3  1.500  0.222   0.667  0.333   6.300  0.793    7.9  100
      4 1  4  0.293  0.854   3.414  0.250   1.230  3.046   40.6   75
      4 3  4  1.707  0.146   0.586  0.250   7.170  0.523    7.0   75
      5 1  5  0.191  1.047   5.236  0.200   0.802  3.738   62.3   60
      5 2  5  0.691  0.289   1.447  0.200   2.902  1.033   17.2   60
      5 3  5  1.309  0.153   0.764  0.200   5.498  0.545    9.1   60
      5 4  5  1.809  0.111   0.553  0.200   7.598  0.395    6.6   60
      6 1  6  0.134  1.244   7.464  0.167   0.563  4.441   88.8   50
      6 5  6  1.866  0.089   0.536  0.167   7.837  0.319    6.4   50
      7 1  7  0.099  1.443  10.098  0.143   0.416  5.149  120.1   43
      7 2  7  0.377  0.379   2.656  0.143   1.581  1.354   31.6   43
      7 3  7  0.777  0.184   1.286  0.143   3.265  0.656   15.3   43
      7 4  7  1.223  0.117   0.818  0.143   5.135  0.417    9.7   43
      7 5  7  1.623  0.088   0.616  0.143   6.819  0.314    7.3   43
      7 6  7  1.901  0.075   0.526  0.143   7.984  0.268    6.3   43
      8 1  8  0.076  1.642  13.137  0.125   0.320  5.861  156.3   38
      8 3  8  0.617  0.202   1.620  0.125   2.593  0.723   19.3   38
      8 5  8  1.383  0.090   0.723  0.125   5.807  0.315    8.6   38
      8 7  8  1.924  0.065   0.520  0.125   8.080  0.232    6.2   38
      9 1  9  0.060  1.842  16.582  0.111   0.253  6.576  197.3   33
      9 2  9  0.234  0.475   4.274  0.111   0.983  1.695   50.9   33
      9 4  9  0.826  0.134   1.210  0.111   3.471  0.480   14.4   33
      9 5  9  1.174  0.095   0.852  0.111   4.929  0.338   19.1   33
      9 7  9  1.766  0.063   0.566  0.111   7.417  0.225    6.7   33
      9 8  9  1.940  0.057   0.516  0.111   8.147  0.204    6.1   33
     10 1 10  0.049  2.043  20.432  0.100   0.206  7.292  243.1   30
     10 3 10  0.412  0.243   2.426  0.100   1.731  0.866   28.9   30
     10 7 10  1.588  0.063   0.630  0.100   6.669  0.225    7.5   30


    1. These are the named cavities: Planar, Spherical, and Confocal. The planar cavity isn't strictly part of this series but is included for completeness, and as a baseline for comparison. Its mirror RoCs would be infinite but its mirror spacing (d) is equal to the RoC (r) used for all the spherical cavities. Configurations for N above 10 would have larger maximum values for the FSR and FWHM (k=1), and smaller minimum values for the FSR (k=N-1), with the FWHM approaching that of the normal hemispherical cavity (1-1).

    2. FSR takes into account the actual value of d as the cavity spacing is varied. The values shown are relative to that of the planar cavity (with mirror spacing equal to RoC=r of the mirrors used in the spherical cavities). Then, FSRr(N)*d/r=1/N, or FSRr(N)=r/(d*N).

    3. FWHM is the width of the spectral peaks relative to that of the planar cavity based on the mirror reflectivity (R) and mirror spacing (d).

    4. Finesse is with respect to the effective FSR, relative to that of the planar cavity and scales as 1/N.

    5. This shows the resulting mirror spacing (d), FSR, FWHM of the spectral peaks, and finesse for each hemispherical cavity with mirror RoC=4.3 cm compared to a planar cavity with a mirror spacing of d=r=4.3 cm for an FSR of around 3.5 GHz with a finesse of 300 (mirror reflectivity of approximately 99 percent).

    Here are the available spherical resonances up to N=10 sorted by increasing mirror spacing. Mirror spacings of 1.0 or less are often more useful being able to either reduce the size of an instrument or provide a larger FSR than the confocal spacing. Mirror spacings of more than 1.0 may be useful to provide smaller FSRs. The Planar cavity 1-0 at the top with the same mirror spacing as the confocal cavity (d=RoC=4.2 cm) would have an FSR of 2.0 x CFSR, but is much more difficult to use.

                                            <------ 4.2 cm RoC ----->
          Num         <---- Relative ---->    d     FSR    FWHM  Fin-  FSR Rel to
      N k Rep  d/r    FSR    FWHM  Finesse  (cm)   (GHz)   (MHz) esse   Confocal
      1 0  1  0.000  1.000   1.000  1.000   4.200  3.569   11.9   300  2.00 x CFSR
     10 1 10  0.049  2.043  20.432  0.100   0.206  7.292  243.1   30
      9 1  9  0.060  1.842  16.582  0.111   0.253  6.576  197.3   33
      8 1  8  0.076  1.642  13.137  0.125   0.320  5.861  156.3   38
      7 1  7  0.099  1.443  10.098  0.143   0.416  5.149  120.1   43
      6 1  6  0.134  1.244   7.464  0.167   0.563  4.441   88.8   50
      5 1  5  0.191  1.047   5.236  0.200   0.802  3.738   62.3   60
      9 2  9  0.234  0.475   4.274  0.111   0.983  1.695   50.9   33
      4 1  4  0.293  0.854   3.414  0.250   1.230  3.046   40.6   75
      7 2  7  0.377  0.379   2.656  0.143   1.581  1.354   31.6   43
     10 3 10  0.412  0.243   2.426  0.100   1.731  0.866   28.9   30
      3 1  3  0.500  0.667   2.000  0.333   2.100  2.379   23.8  100
      8 3  8  0.617  0.202   1.620  0.125   2.593  0.723   19.3   38
      5 2  5  0.691  0.289   1.447  0.200   2.902  1.033   17.2   60
      7 3  7  0.777  0.184   1.286  0.143   3.265  0.656   15.3   43
      9 4  9  0.826  0.134   1.210  0.111   3.471  0.480   14.4   33
      2 1  2  1.000  0.500   1.000  0.500   4.200  1.785   11.9  150  Confoc 2-1
      9 5  9  1.174  0.095   0.852  0.111   4.929  0.338   19.1   33
      7 4  7  1.223  0.117   0.818  0.143   5.135  0.417    9.7   43
      5 3  5  1.309  0.153   0.764  0.200   5.498  0.545    9.1   60
      8 5  8  1.383  0.090   0.723  0.125   5.807  0.315    8.6   38
      3 2  3  1.500  0.222   0.667  0.333   6.300  0.793    7.9  100
     10 7 10  1.588  0.063   0.630  0.100   6.669  0.225    7.5   30
      7 5  7  1.623  0.088   0.616  0.143   6.819  0.314    7.3   43
      4 3  4  1.707  0.146   0.586  0.250   7.170  0.523    7.0   75
      9 7  9  1.766  0.063   0.566  0.111   7.417  0.225    6.7   33
      5 4  5  1.809  0.111   0.553  0.200   7.598  0.395    6.6   60
      6 5  6  1.866  0.089   0.536  0.167   7.837  0.319    6.4   50
      7 6  7  1.901  0.075   0.526  0.143   7.984  0.268    6.3   43
      8 7  8  1.924  0.065   0.520  0.125   8.080  0.232    6.2   38
      9 8  9  1.940  0.057   0.516  0.111   8.147  0.204    6.1   33
      1 1  1  2.000  0.500   0.500  1.000   8.400  1.785    5.9  300  Sphere 1-1

    With a mode matching lens, the finesse is indeed much higher, but that mode matching becomes annoying as the optimal lens or at least its optimal location depends on the FSR that is selected. Thus, it would be useful to have the lens on a linear slide as well.

    All the same resonances are available for the selectable FSR spherical and hemispherical cavity SFPIs with the latter being 1/2 the mirror spacing for each. Although the hemispherical cavity halves the finesse for mirrors with identical reflectance, the reflectance of the planar HR is much higher so the finesse should not be much lower. However, as noted above, for reasons not entirely clear, this was never achieved.

    For some spacings corresponding to higher orders of N, the amplitudes of the peaks corresponding to the various FSRs tend to differ, sometimes quite dramatically, even when alignment is far from on-axis. This was not totally unexpected though where a large number of distinct higher order modes need to be excited to average out for the appearance of them to be equal.

    A computer-controlled interferometer-based positioning system with smart parameter calculation software would be a definite plus here. :) It's often difficult to determine which exact table entry is being displayed, especially where usable mirror spacings are close together. The mediocre repurposed microscope mechanical stage scale really isn't adequate (aside from it reading from 40 to 105 instead of 0 to 65, which would be more useful).

    Eventually it was upgraded with a piece of a machinist's scale glued down so 0 actually corresponds to where the mirrors would actually be 0.000000 mm apart. With that setup, it was then possible to measure the RoC of the mirrors quite accurately. For these samples, it comes out to be very close to 4.2 cm, not 4.3 cm as had been listed everywhere else here until recently. ;( :) And in fact the scale is accuarate enough that the SFPI can now be quickly set to any desired mode-degenerate order from within the range of the positioner with only quick fine tuning to peak the response.

    When setting the mirror spacing to a specific value of N/k, it's useful to keep in mind that the number of repeated peak-clumps ;-) for a fixed ramp drive voltage scales as N and the average peak amplitude scales as 1/N. So this makes it easy to recognize when a particular distance setting is selecting the incorrect values for N. The peak amplitudes may vary dramatically with respect to the ramp voltage regardless of alignment when an incorrect spacing is selected. Further, if not perfectly aligned, then the number of distinct spots on the mirror surfaces will be equal to N, though this may be difficult to see. And as with all mode-degenerate SFPIs, the accuracy with which the spacing must be set for maximum resolution is inversely proportional to RoC. So long SFPIs are less sensitive to cavity length errors than short ones.

    As another test, I installed a 1 meter RoC 99.4%R mirror in place of the planar mirror. This results in a highly asymmetric spherical cavity. As expected, this is not pure MDI so while similar major resonances are present, ghost and double peaks come and go with changes in the input laser's longitudinal mode frequencies and SFPI cavity length. Perhaps a 5 meter RoC mirror would be close enough to planar to reduce these artifacts to acceptable levels but I don't have one of those handy.

    I had originally dismissed this selectable FSR SFPI as an academic exercise. However, there is at least one application where this could prove useful to the kind of experiments on which I spend way too much time: Observing the mode sweep of Zeeman-split HeNe lasers. While the resolution is not high enough to actually see the split Zeeman-split lasing mode, the larger FSRs easily dialed up using this system such as 3.046 GHz (Cavity 4-1) would easily enable the overall mode sweep behavior to be observed, while the higher finesse default FSR of 1.78 GHz would suffice for most everything else. Pretty lame, huh? :)

    Another related application would be to actually observe the Zeeman split modes using a much longer instrument. See the sections starting with: Sam's Selectable FSR High Resolution Spherical Mode-Degenerate Scanning Fabry-Perot Interferometer 2.

    While the rig described above is certainly adequate as a "proof of concept", constructing a more robust and stable selectable FSR SFPI would require a more elaborate structure. Even playing a radio (remember those?) at modest volume causes the display to jump around. One possibility for quickly setting coarse mirror spacing might be to use a mechanism similar to that of a zoom lens where an inner element is moved via a knurled ring. Instead of zoom factors, it would be labeled in FSRs! Fine tuning would be via a separate threaded ring similar to the mirror spacing adjustment of some normal confocal SFPIs. Both would be capable of being locked in place. Yes, I know, it's more likely that there will be a sighting of pink flying pigs before anything like this gets built! ;-)

    However, as a compromise, I am constructing a dumb one using a linear slide. Here is a suggested parts list:

     Function            Part Model/Description
     Cavity Mirrors      42 mm RoC, >99%R at 590-633 nm.
     Mirror Mounts       Thorlabs KMSS or KMS, KM05FL, KM05FR, MK05, KS05, K05,
                          etc., or Newport U50-A, U50-P, U50-P1, etc.
     Translation Stage   Thorlabs DT12 or T12X.
     Linear Slide        MGN9C on 150 mm rail.
     Locking mechanism   Thumbscrew via custom plate.
     Mounting Base       Aluminum plate 3/8 x 3 x 6 inch.
     Adjustment Screws   6-32 x 1-1/2" cap head, optionally ground to a point.
     Calibration         10 cm piece of machinists' scale.
     Miscellaneous       Spacers, adapters for the 7.75 mm mirrors, etc.

    And of course, in place of the translation stage and basic linear slide, one could be that incorporated a linear motor and encoder using microprocessor control with auto optimization firmware. :) Not. ;-)

    A Non-Confocal Cavity SFPI Using Transverse Mode Confinement?

    Well, that was the idea anyhow. The rational was that rather than finding a matched set of confocal mirrors for a high finesse SFPI which has proven to be rather challenging (or at least expensive, see the next section), why not build a cavity with concave mirrors to simplify alignment but use the bore of a (defunct) HeNe laser tube to suppress higher order modes. It works in the HeNe laser, right? Then, the exact cavity length wouldn't be as critical, and it could actually be longer than the confocal length (to decrease FSR). In addition, if it could be forced to operate only in the TEM00 mode, there would be no loss of a factor of 2 in finesse as there is with the confocal configuration.

    To test this idea, an already dead (up to air) Spectra-Physics 088 HeNe laser tube was sacrificed by removing the cathode-end (HR) mirror. Actually, the entire end-cap assembly came apart at the glass-to-metal seal when, after scoring the metal tip-off, a pair of pliers was used to try to break it off. But there was no damage to the remainder of the tube including the entire glass envelope. (This might have been a result of a hairline crack already being present at the seal and the reason for the leak.) So, the anode-end mirror attached to its mount, bore, tube envelope, and centering spider was installed in my laser test jig, normally used with one-Brewster HeNe laser tubes. This tube is supported by 4 Nylon screws in two places to permit fine adjustment of centering and alignment. With this scheme, by slightly loosening two pairs of Nylon screws, the tube could be easily moved over a range of FSRs of about 400 to 500 MHz, hopefully without totally losing alignment. A mirror with similar characteristics to that of the SP-088 OC was glued to a PZT beeper element and attached to the adjustable mirror mount.

    Since the OC mirror is known to be properly aligned, its reflection could be used to align the test laser (initially a Melles Griot 05-LHR-911). Then when that was close, the function generator and scope were activated and fine alignment of the adjustable mirror could commence.

    This entire exercise turned out to be easier than I had expected but the first results were somewhat under-whelming: I was able to obtain a finesse of nearly a value of... 2, and just barely recognize what I assumed to be the modes of the 05-LHR-911 laser as lumps. :)

    Now, I was expecting and hoping for a finesse of 200 or 300 to be able to resolve the split modes of HP/Agilent metrology lasers, a few MHz apart. Clearly some more work was called for.

    My initial thought was that the first problem was that the 05-LHR-911 beam diverges as a fast enough rate that it clips the bore so diffraction losses are very high and doesn't even create a stable intracavity mode volume. So, the next step was to at least confirm this as one problem by first moving the test laser closer to the SFPI to reduce the beam diameter. A long focal length positive lens was also added to focus the laser beam into the SFPI cavity.

    These did help a bit but didn't produce any eureka moment. Then something happened. My memory of the exact sequence of events is somewhat fuzzy, but then the finesse jumped to a much more reasonable value. Part of the problem was that the photodiode output wasn't terminated except using the 1M ohm input impedance of the scope. This was both resulting in saturation at higher light levels and seriously low pass filtering the response. A 10K ohm resistor took care of that. However, I don't believe this was the entire problem because I had tried terminating the signal with no significant improvement. But then, fiddling with the alignment resulted in a very dramatic increase in finesse and output level. So, it was probably a combination of factors. Possibly the original response was not even due to a direct path down the bore with a couple reflections off the side-wall of the bore. Not that likely, but possible. The finesse is now consistently at least 50 and likely over 100 at times. Not great, but more respectable than 2! :)

    That's the good news.

    The bad news is that this scheme still has problems. For one, higher order modes are still present. Not as many as with my original short cavity SFPI, but enough to be annoying and confusing. Their amplitude can be anywhere from 10 to 100 percent of the height of what I believed to be the TEM00 modes. But the relative heights of the modes can be varied by any change in alignment, even pressing gently on the mirror mount. With the FSR of the SFPI being much less than the neon gain bandwidth, interpreting what was going on became even more difficult.

    I have attempted to more closely match the input beam to the mode volume of the SFPI, but so far, this is not been very productive. And even if this did work, requiring such painstaking setup for each test would be impractical.

    In short, although not a total failure, this approach has significant difficulties of its own, so I intend now to go back to Plan A, which is to build a long confocal SFPI. :)

    Sam's High Resolution Scanning Fabry-Perot Interferometer

    This would be a nice long confocal SFPI capable of resolving the two lines of Zeeman-split two-frequency HeNe lasers such as those from HP/Agilent and Excel. :) The typical separation of the two frequencies (called F1 and F2) is between 1.6 and 4.0 MHz. So, an SFPI with a resolvance of 1-2 MHz would be required. This will need both a combination of larger mirror spacing and decent finesse. However, the basic design would be similar to that of my other SFPIs.

    One possibility for mirrors would be the OCs from deceased low to medium power HeNe laser tubes. The type that could be satisfactory would be 99%@633nm with an RoC of 60 cm. For the confocal configuration, the FSR would be 125 MHz with a finesse of about 150 producing a resolvance of about 0.83 MHz. And the Gods of Dead Lasers know I have many of these mirrors. :) However, one deficiency is their small diameter. Most are around 7 mm if bare, but only 3 or 4 mm if mounted in the original mirror mount stems, somewhat less than desired. Removing mirrors intact is a hit or miss proposition, especially for those like Melles Griot having a thick bead of glass frit. My success rate has been very low. You don't want to see the results! An alternative would be the OC mirrors from Spectra-Physics 084 tubes which have a similar RoC and reflectance. However, they are in limited supply, though I could probably dig up a pair. But a somewhat higher reflectance would be even better, say 99.5% or even 99.8%, which would boost the finesse. Now it turns out that I have found suitable mirrors in a somewhat strange place - the HR mirrors of tube from Spectra-Physics 117A stabilized HeNe lasers! (This is the same as the Melles Griot 05-STP-901 and Melles Griot has made the tubes for these lasers for many years.) Unlike most modern HeNe laser tubes, these are curved (rather than being planar) with an RoC of 60 cm. They also have an AR coating to minimize back-reflections. In addition, they are probably have a slightly lower reflectance than a true HR (but still much higher than an OC) since the beam sampling is done through the HR mirror. But I've already totally destroyed one trying to get it off the mount - it fractured from scraping the frit, not even particularly vigerously. And my supply is rather limited - 3 or 4 more at most are available from dead SP-117A tubes.

    But the required length of the SFPI using 60 cm RoC mirrors would be somewhat unwieldy, so what I'd really like would be a smaller RoC to make the instrument more compact - say 30 cm - and with a reflectance of 99.5% to 99.8%. Unfortunately, this combination or RoC and reflectance is rather hard to find. Melles Griot does offer some curved HR mirrors, but they cannot guarantee that the reflectance wouldn't be so high as to be useless for an SFPI. And while not that expensive as these things go, they wouldn't be free or cheap enough to try out.

    I also have several New Focus kinematic mounts for 1" optics. These are a bit large, but hey, you use what you have! :) And the nice thing about these is that they have 3 screws (rather than two adjustment screws and a ball bearing for the pivot). So the mirror can be translated precisely along the Z axis to fine tune the spacing. Then, all that's required is a rigid frame. I considered using the L-bar resonator from from a defunct Spectra-Physics 124 laser. Its length would be ideal for use with 60 cm RoC mirrors and one of the mirrors could be installed in the existing mount at one end. But simply disassembling that laser would be a fair amount of work and I'd have to schedule an appointment with the Gods of dead lasers to determine what else might be needed in the way of chants and incantations. :)

    In the end, I decided to build the smaller instrument, at least for now. It would use a pair of 30 cm RoC OC mirrors salvaged from some unknown 1 or 2 mW HeNe laser tubes. I have several of these, already removed from the tubes (no further special approvals required), have tested their reflectance, and selected a pair with the highest, about 99.0 and 99.2 percent. (Their average reflectance, 99.1, will be close enough for calculation purposes.) These mirrors will result in an FSR of around 250 MHz (for the confocal condition) and theoretical finesse (with respect to the confocal condition) of π*sqrt(0.991)/(1-0.991)/2=174 and a theoretical resolvance of about 1.41 MHz - sufficient to view the split modes of any HP/Agilent two-frequency metrology laser.

    I dug up a length of extruded aluminum stock to serve as a rail. The two New Focus mounts were attached directly to it with the relevant surfaces of the mirrors spaced exactly 30 cm apart. The adjustment screws provide approximately a 10 mm total range in distance to allow for a normal tolerance in RoC. My determination of RoC is not very precise, based on the location of the focal point of a collimated HeNe beam reflected from the mirror. So I'm assuming that it is a nice round number of cm! The final determination would be made by adjusting the SFPI for the mode-degenerate condition. Should the RoC turn out to be in error by more than 10 mm, the hole by which one of the mounts is attached could easily be elongated or moved. However, I believe the RoC is very close to 30 cm. :)

    The parts went together easily with the two mounts positioned such that the distance between the surfaces of the mirrors was 30 cm. One mount is secured in an elongated hole just in case. :) The 99.2% mirror was glued to a $1 PZT beeper from Digikey and the 99% mirror was sandwiched between a pair of small hard drive platter clamping plates so it could easily be swapped with something else if needed. All the knobs were removed from the New Focus mounts so that a hex driver is needed for adjustment, making inadvertent screwups less likely. (The knobs can be easily put back on, another benefit of the New Focus mounts, but that will probably never be needed.)

    Please see: Photos of Sam's High Resolution Scanning Fabry-Perot Interferometer 1. The frame is a recycled chassis rail. The New Focus mirror mounts are ideal for this application since they have three adjustment screws so that the cavity length can be fine tuned. They also have four 8-32 tapped holes conveniently positioned to hold stuff. :) The front mirror (lower left photo) is gently clamped between a pair of metal plates with a nylon washer for cushioning, and then the entire assembly is clamped to the New Focus mount with 4 screws. The back mirror (lower middle photo) is glued to the PZT beeper, which is clamped to the New Focus mount with 4 Nylon screws over a plastic sheet to insulate it from the chassis. The insulation makes it possible to drive the PZT with my SGSF1 ramp generator, which can be configured for a 50 V p-p signal from a pair of opposite polarity outputs. (As it turned out, this was not necessary as even the 15 or 20 V p-p available from a function generator was more than enough to span several FSRs.) The photodiode has an approximate active area of 5x8 mm (because that's what I had available) plugs into a pair of socket pins pressed into a black plastic sheet (lower right photo). A 10K ohm load resistor provides appropriate termination as long as the maximum transmitted optical power is less than about 0.1 mW. For higher power lasers, either optical filters can be added to drop the power, or a proper transimpedance (zero voltage offset input) amplifier can be used (one of these is part of SGSF1). Filters have the added benefit of reducing back-reflections into the laser, which most lasers do not like very much. Though not as much of a problem with a confocal SFPI as with a plane-plane SFPI, they are still possible. To minimize stray light into the photodiode, a cylindrical shield (removed for the photo) about 2 inches in length normally surrounds the back mirror/PZT assembly. The feet (2 at the front and 1 at the rear) are 4-40 machine screws turned down to a point allowing for fine alignment of the overall SFPI with respect to the laser.

    Of all the SFPIs I've experimented with (both home-built and commercial), there is no doubt that this one is by far the easiest to align. Practically just placing it in the general vicinity of a laser results in modes on the scope. :) The nice New Focus mirror mounts help but their contribution is very minor. Fundamentally, there is a wide range of alignment - both of the mirrors and of the SFPI with respect to the laser - over which a usable signal with decent amplitude and resolution is produced. In fact, there was a credible mode display the first time it was switched on. The only alignment done prior to that was to set up the laser so its beam went approximately down the axis of the SFPI, and then adjusting the mirrors to get a clump of reflected spots on both mirrors. With most optical systems "long" and "easy alignment" are oxymorons. Not so with this configuration, a benefit of the stable confocal singularity and the ease with which the incoming beam can be aligned with the SFPI mirrors. After some minor touch-up of the mirror alignment and distance, the performance is approaching expectations. There's no doubt that it is running at the confocal distance (or very close to it). The display is quite stable with only the two modes of the 05-LHR-911 laser present, changing during mode sweep in the normal manner. Of course with an FSR of only 250 MHz and a longitudinal mode spacing of 883 MHz for the 05-LHR-911, the peaks don't have the normal spacing. It is in fact 883 MHz modulo 250 MHz, so they are actually located (relative to one of the modes) at 0 MHz, 133 MHz (883-3*250 MHz), 250 MHz, 383 MHz, etc. However, their amplitudes vary exactly as they would with a larger FSR, but they move across the screen really quickly due to the FSR (250 MHz) being much smaller than the neon gain bandwidth (~1.6 GHz). The peaks are clean and symmetric, there are no smaller ghost modes, none of the random variations in mode heights that are usually present when the distance is way off, or the peculiar slow mode height variations not correlated with laser modes seen with the non-confocal barcode scanner tube SFPI described above. It's about as close to a textbook display as possible without being in a textbook. :) The finesse may still be a bit lower than what calculations predict - perhaps 100 instead of 174 (with respect to the confocal FSR of 250 MHz). But even with a finesse of 100, it should be possible to resolve the split modes of an HP/Agilent 5517D or 5517E, and probably a 5517C. But there's still room for improvement and fine tuning of alignment and cavity length help somewhat. A long focus positive lens didn't have any dramatic effect one way or the other, though while fiddling with it (stuck on a "third hand") I stumbled on the condition of perfect alignment where the FSR doubles - every other set of modes vanishes and the remaining modes double in height, with the effective finesse then becoming 200 with respect to the 500 MHz FSR. But that was probably a coincidence as it occurred again later without the lens.

    What I hadn't realized initially was that the sensitivity of the display with respect to offset from the confocal cavity length is much lower for a longer SFPI, with more than one turn of the 80-pitch screws being needed to produce a noticeable change in the display. Having just fought with the 30 GHz Coherent SFPI's 5 mm cavity where 1 or 2 degrees of rotation was significant (and difficult to do consistently), this was most welcome. The only annoyance is the frame. Aluminum is not exactly the most thermally stable structural material. But covering the SFPI with a cardboard box helps. :-)

    The next test was to try a Hewlett Packard 5517C, with a split frequency of around 2.4 MHz. (I had performed my "soup can mod" to bring the split frequency down to near the lower end of the HP-5517C range, 2.4 MHz.) The laser was set up on an adjuatable platform with an HP beam expander oriented backwards to reduce the beam diameter to around 1 mm. And sure enough, once everything was arranged on appropriate high tech wood blocks :) to get their height approximately equal within the range of the adjustments of the laser platform and SFPI, it was immediately obvious that when the 5517C was locked, there was only a single peak per FSR since it is single longitudinal mode, but that it was jagged at the top, not a normal mode. (While the laser was warming up, two clean modes could be seen over part of the mode sweep cycle and the jagged one being present during a small portion, mostly by itself.) With careful adjustment of alignment and some fine tuning of cavity length, the existence of the two Zeeman modes was clearly visible. The peaks weren't distinct and totally separate, but a 10 to 20 percent dip could clearly be seen between them. Based on a Matlab simulation, this indicates that the finesse is probably quite close to 100 (relative to the 250 MHz confocal FSR). Not surprisingly, the resolution is often, but not always, highest just about when the optimal alignment is achieved such that the FSR doubles and every other mode disappears (or at least gets much smaller). Interestingly, the actual resolution of the larger and smaller modes are not generally the same (even accounting for their difference in height). The larger peak usually has a more pronounced dip but not always. More below. It was fairly easy to obtain a finesse of about 125 with a dip of around 25 percent. I also tried the laser without the beam reducer (its beam is only 3 mm in diameter), but the results were somewhat worse. Rather than include all the boring details, suffice it to say that in the end (or at least what passes for the end so far), the beam reducer at its narrowest setting along with careful alignment of the laser, beam reducer, and SFPI, as well as positioning the SFPI closer to the laser, and last but not least - cleaning the mirrors - produced the best results with a finesse of around 160 and a dip of around 45 percent. This is very close to the theoretical finesse of 174 shown in Simulation of Sam's High Resolution SFPI 1 Display of HP-5517C Modes which includes plots of more than a full FSR (400 MHz range) and a zoom-in of the Zeeman-split peak (20 MHz range. The plots for the measured finesse of 160 are barely distinguishable from the these except that the dip is about 5 percent less deep. The real-time display looks very similar to the simulation except that every other peak is smaller as noted above. However, it's just much easier to capture the results of the simulation than to photograph a jittery expanded scope trace! Better cleaning would probably recover the missing finesse. The mirrors must collect dust even being vertical surfaces because I know the finesse had been declining over the last few days. I need to make a cover. :)

    Recall how I said this SFPI was easy to align? Well, that was to get some sort of display. To actually approach the theoretical finesse required significant effort and time. And it's still not clear what combination of conditions actually results in the best finesse, but it always occurs near to point of perfect alignment but not precisely at it. One set of peaks is about 50 percent larger than the other set. The larger peaks have the high finesse while the smaller peaks may be half of that or even less even with the cavity length set optimally for the confocal condition. But the finesse is actually lower at perfect alignment when one set of peaks disappears. My speculation is that this would be the optimal condition if the cavity were perfectly mode-matched to the incoming laser. The finesse of that larger peak also seems marginally higher if the cavity is a bit short compared to the confocal length by somewhere around 0.5 mm. Getting all the way to the finesse of 175 predicted by theory (at which point the dip should extend down to approximately half the peak) should still be possible but may require sacrifices to the Gods of Laser Instruments. :) Or, as noted, it may simply be a matter of cleaner mirrors, which would still require those sacrifices given how much I detest cleaning mirrors. But a few nice crunchy dead HeNe laser tubes would probably be acceptable sacrifices. :) However, there's little point in spending yet more time removing every last molecule of contamination from the mirrors unless the SFPI is covered to keep them clean! Of course, it's also possible that my measurements of the mirror reflectances weren't accurate, as hard as this may be to believe. :) But if the reflectance of both mirrors were 0.99 or one was 0.989 and the other was 0.91, then the theoretical finesse is only around 149 - so it's already way past that!

    The Zeeman-split modes were also not quite equal in amplitude, but that's in the laser and due to the modified control PCB in this 5517C. Unlike the standard one, I added a mode balance pot which was not set correctly. It was interesting to watch the relative mode amplitudes change as that pot was adjusted. The relative split mode heights also vary slightly, which may be due to back-reflections even through the ND filter. And, this high mileage laser had significant amplitude and/or frequency ripple due either to a defective HeNe laser power supply or plasma oscillations which were showing up in the display and confusing the interpretation. These were eliminated by running on a different HeNe laser power supply at a slightly higher tube current. I'll have to fix that. :) For more on HP/Agilent metrology lasers, see the section: Hewlett-Packard/Agilent Stabilized HeNe Lasers.

    And it turns out that this home-built SFPI actually has an even better theoretical resolution than the fancy long Coherent SFPI that I have wanted. The Coherent SFPI has an FSR of 300 MHz with a finesse of 200 for a resolution of 1.5 MHz (compared to 1.41 MHz for mine). Of course it does have the advantage of using a proper sealed low expansion cavity with a nice adjustable mount. And then there is the small difference in cost. :)

    Next up: An SFPI with several times the resolution! This wouldn't be as spectacular as the ultra-high resolution SFPI described below, but would be along the lines of something you or I could build with a bit of lucky scrounging or real money. It would have the following specifications:

    All that would be required are a pair of high quality mirrors with an RoC of 50 cm and a reflectance of 99.7 percent. An actual plot of the two modes of an HP-5517A laser (1.65 MHz separation) made with such an instrument can be found in High Resolution Scanning Fabry-Perot Interferometer Display of HP-5517A Modes 1. However, this was not made by me, as suitable mirrors have yet to materialize.

    But never fear, those described in subsequent sections come close or even exceed these specs. ;-)

    Scanning Fabry-Perot Interferometer Head Frame using Commercial Parts

    Building SFPIs from scratch can be quite rewarding, but it does become somewhat boring after the 15th or 20th. :) This section deals with using commercial parts to largely eliminate the need for custom parts when constructing an SFPI with short RoC mirrors (e.g., 42 mm/1.78 GHz) generally intended to be used for a Laser Spectrum Analyzer (LSA) or similar application. And the same structure can accommodate selectable FSR mode-degenerate cavity spacings up to at least the mirror RoCs. Examples of Scanning Fabry-Perot Interferometers using Thorlabs Cage System Parts shows several that are described in detail below. To construct a similar structure suitable for specialty SFPIs with mirror spacings up to one half meter or more, see the section: Generic Scanning Fabry-Perot Interferometer Testbed. While these will not be as stable as a several thousand dollar instrument using 1 inch diameter Invar rods, their performance is quite satisfactory.

    For around $100 in readily available parts, many of the headaches of constructing the most common type of SFPI can be eliminated. Of course, this does assume the availability of suitable mirrors. The following assumes the use of "30 mm Cage" parts from Thorlabs:

    Thorlabs parts except as noted

     Qty  Part #  Description
      1   Lens mount:
           CP32 (CP11)  SM05-Threaded 30 mm Cage Plate, 0.35", 2 Retaining Rings
      1   Front mirror mount:
           CP33T (CP02T)  SM1-Threaded 30 mm Cage Plate, 0.50", 2 Retaining Rings
           SM1AD8  Externally SM1-Threaded Adapter for 8 mm Optic
      1   Rear PZT/PD mount (only one of the following required):
           CP33 (CP02)  SM1-Threaded 30 mm Cage Plate, 0.35", 2 Retaining Rings
           CP35  30 mm Cage Plate with 1" Double Bore, 8-32 Tap
      1   ER4-P4  Cage Assembly Rod, 4" Long, 6 mm, 4 Pack
      1   LA1213  BK7 Plano-Convex Lens, 1/2", f=50.0 mm, AR: 350-700 nm
      1   L14148  Surplusshed 12.4 mm, f=52 mm, AR-coated (less expensive)

    Note: Thorlabs part numbers have recently changed. CP33 was CP02, CP33T was CP02T, and CP32 was CP11. The only obvious difference is that the grub screws that lock the plates to the rods are now M4 instead of 4-40. A driver for an 8-32 grub screw works but an 8-32 screw will not fit. This might be more secure but if one is lost, it could be a problem to find a replacement. Also, they come in paper envlopes inside carboard protectors instead of plastic bags to help the environment. ;-)

    The CP33T and the SM1AD8 is used to mount the front mirror while the CP33 is used to mount the PZT on one side and the photodiode on the other. The CP32 is for mounting a focusing lens. This can be the Thorlabs LA1213 listed (around $30 in 2018) or something similar and less expensive from a place like Surplus Shed. The 4 cage rods allow for quick and easy coarse adjustment of cavity and lens spacing. Rotation of the SM1AD8 is then used for fine cavity spacing, locking in place with one of the retaining rings. This isn't quite "plug and play" as some gluing or drilling and tapping is required, as well as a few other odds and ends to complete the SFPI assembly as shown in SFPI Frame Using Thorlabs Cage Parts and Most Other Components. But the effort will still be an order of magnitude less than building one of these from salvaged harddrives and scrap aluminum. :) Depending on the rod length (4 inches in this case), almost any confocal cavity can be accommodated with the appropriate RoC mirrors and focusing lens.

    Other parts

     Qty   Description
      2    7.75 mm diameter cavity mirrors, RoC of 42 mm, 99.5%R@590-650 nm
      1    27 mm diameter PZT two-wire beeper element, drilled with center hole
      1    PZT mount (requires one of the following):
            1.25 inch OD x 1/8 inch thick washer for PZT mount if using CP33 (CP02)
            1.0 inch OD x 3/8 inch long cylinder for PZT mount if using CP35
      1    Aluminum plate (2x5x3/8" inch) for adjustable mount
      3    4-40 x 1 inch cap-head screws with pointed tips for height adjustments
      1    Piece of Perf. board for photoiode and connectors
      -    Miscellaneous mounting screws and other hardware

    To simplify initial alignment, a KC1 cage-compatible 3-screw kinematic mount can be substituted for the CP32, but this does roughly double the cost and increases the size and bulk. The prototype of such a rig is shown in SFPI Head Using Thorlabs Cage Parts with Adjustable Front Mirror. It's kind of a waste though since once the SFPI is set up, the adjustable mount is of little value and possibly a liability as its settings can change. However, being able to precisely align the two mirrors so their optical axes coincide may result in slightly better performance. An alternative to a KC1 would be to mount the PZT on a plate with 3 springs or split washers so that its alignment could be fine tuned. Added cost: $0.00.

    Having said all that, the simpler SFPI head design really does work just fine. The completed unit mounted on a 3-screw adjustable platform is shown in Completed SFPI Laser Spectrum Analyzer Head Using Thorlabs Cage Parts. Coarse mirror spacing is set by moving the CP33T with the front mirror. This can be done by first setting its position based on the known RoCs of the mirrors, and then by maximizing the envelope of the SFPI display of a well behaved laser like an 05-LHR-151. Fine tuning is then performed by rotating the SM1AD8 adapter holding the front mirror using an improvised spanner made from a large paper clip. :) A tapped 4-40 hole added on the side of the CP33T for a soft-tipped set-screw that is snug but still allows movement would enable smooth adjustment and lock the setting in place. An opaque cover should be added to prevent ambient light from getting to the photodiode and protect the mirrors from dust and other contamination. The 8-32 tapped hole in the CP33T can be used to attach the SFPI head to an adjustable platform or multi-axis mount as in the setup, above. A photo of an actual scan from this unit is shown in SFPI Mode Display of Melles Griot 05-LHR-151.

    Both of these are quite stable and worth every penny. ;-) The only problem seems to be a ringing in the $1 PZT at the beginning of the scan due to the abrupt change in direction, resulting in distortion of the display over the first 10 percent of the scan when running at 100 Hz, with a proportionally lower percentage at lower scan rates. This is an issue with all of the SFPIs using the thin beeper PZTs. Using a similar PZT but with a higher resonance frequency, a modified waveform that starts more gradually, or applying damping material to the PZT would help. Partially coating the back of the PZT with hardware store acrylic caulk did in fact reduce the duration, though a lossier material would probably be better. Of course, in the grand scheme of things, a bit of distortion over less than 10 percent of the scan is something that is probably tolerable. If not, simply delay the scope trigger by 1 or 2 ms. ;-)

    A longer Thorlabs cage assembly can be used with other mirror spacings and this has been tested up to over 0.5 meter (22 inches cage rod length) with adequate stability even when built with rods in three sections of 6,8,8 inches.

    Home-Built SFPI using Thorlabs Mirrors - NOT

    I was hoping to find an alternative to the limited supply of suitable SFPI laser spectrum analyzer mirrors (42 mm RoC, >99%R). However, this has so far proven frustrating. The most promising candidates from Thorlabs were thought to be their CM126P-0xx-E02: "1/2 Inch Broadband Dielectric Concave Mirrors (400-750 nm), Back Side Polished". While not exactly inexpensive at $92 a pop, if usable, they would provide coverage for the most common laser wavelengths from blue to red. Unfortunately, tests of two samples of these mirrors has shown that it would be difficult or impossible to make them work because the %T is too low. The graphs on the Thorlabs Web site page for these mirrors show %R at 633 nm (as an example) to be between 99% and 99.6%. Normally, that would imply a %T of 1.0% to 0.4%, but the measured transmission through a single mirror is virtually undetectable and probably below 0.01%. It is not known whether this is because the spec is a worst case and %R could be much closer to 100%, or because of the broadband coatings, or something else. But even a sample of a "Coherent Ultimate Broadband" HR mirror has higher transmission. Getting a usable output from an SFPI with two mirrors in series and no photons would be quite a challenge

    Too bad. No light, no action, no good. ;( :)

    Sam's Selectable FSR High Resolution Spherical Mode-Degenerate Scanning Fabry-Perot Interferometer 1

    Long SFPIs can in principle be built to improve resolution given a constant mirror reflectivity as suggested at the end of the previous section. But using the principle of the selectable FSR SFPI, a much shorter and more manageable instrument is possible. For example, mirrors with R above 99.5 percent between at least 590 nm and 650 nm for yellow (594 nm), orange (605 and 612 nm), or red (633 nm) would be ideal, but only two RoCs are available: 4.3 cm and 100 cm. (In the trivial triviality department, both of these were almost certainly costly goofs by the mirror manufacturers which is how they became available at affordable prices. It's almost certain that the 4.3 cm RoC mirror should have been 45 cm RoC, and the 100 cm RoC mirror had too high a reflectance from 600 to 650 nm.) And while the finesse of a confocal SFPI using some samples of these mirrors would be greater than 500, that's still not sufficient to resolve down to 1 MHz with the 4.3 cm RoC mirrors. Using the 100 cm RoC mirrors in a confocal SFPI is possible but it would be, well, over 1 meter long, and quite unwieldy. :)

    Fortunately, a Fabry-Perot cavity set to one of the low order mode-degenerate spacings can be much shorter than the RoCs of the mirrors can provide a reasonable compromise for FSR and resolution. A selectable FSR SFPI has been built using Thorlabs cage parts enabling a mirror spacing up to around 32.5 cm as shown in Sam's High Resolution Selectable FSR Scanning Fabry-Perot Interferometer 1. Counterclockwise from the bottom-left are the iris diaphragm variable size input aperture, adjustable home-built front mirror mount, rear mirror glued to PZT glued to piece of pill bottle secured in Thorlabs KC-1 adjustable mount, KC-1 overall, and Perf. board with PZT drive and photodiode connectors. The Thorlabs KC-1 adjustable mount at the rear-end is used for fine alignment. The front mirror is mounted on a home-built adjustable plate secured with three 2-56 cap screws backed by split washers for the restoring force, It is aligned initially but doesn't require further tuning. The iris really isn't essential but came with the Thorlabs rig on eBay :) so it's available to aid in initial alignment as well as to help block the reflected beam from entering the laser aperture. The photodiode is on the hidden side of the piece of Perf. board screwed to the back of the KC-1, along with connectors for the ramp drive and photodiode signals.

    And if you're curious (or really bored), the larger aluminum plate is from an OEM version of an HP 5517B laser which is milled rather than molded, and the smaller one is from some other unidentified laser. ;-)

    Based on the chart in the section: Selectable FSR Mode-Degenerate Fabry-Perot Interferometers, order (N/k) 4/1 should be the best given the length of the cage setup. At that mirror spacing of 29.3 cm, the FSR is around 128 MHz according to the chart and underlying equations and That is close to what it is set to in the photo. But when first built, the results appeared very strange. In fact, instead of ~128 MHz, the FSR looked like is was close to 6.66 GHz. That's only off by nearly a factor of 50. Surely close enough for government work. ;-) The first laser used for testing was a Melles Griot 05-LHP-071, a 2.5 mW polarized red HeNe. See Whacky Mode Display of Melles Griot 05-LHP-071. This laser usually has 2 modes 636 MHz apart with three modes present only when 2 are straddling the dominant mode. And that was exactly how it appeared. The spacing between the two modes in the display was thought to be 636 MHz. In fact, these modes behaved beautifully during mode sweep moving smoothly through the neon gain curve, except that they were moving much too quickly and were very unstable with any vibration or air currents. The behavior was so consistent with a normal mode sweep display that the first suspicion was that the equations must be wrong. Except that the predicted and experimental behavior for order 3-1 matched perfectly with the 4.3 cm RoC mirrors. (See the section: Sam's Mini Laser Mode Analyzer 1.) So the equations were probably correct. The next suspicion was that this was a case of aliasing - wrap-around due to the small FSR relative to the gain bandwidth of the laser. So a test was done with a Melles Griot 05-LHR-911 laser, a 2 mW random polarized red HeNe with a longitudinal mode spacing of 883 MHz. But the behavior appeared to be qualitatively very similar and still consistent with a 6.66 GHz FSR. At which point the Universe flipped upside-down because there were no other options. :( :) But it wasn't time to give up just yet. I thought it unlikely that a pair of lasers selected more or less at random would result in a similar whacky display, but that could not be ruled out. So the SFPI was tested with a Melles Griot 05-LHR-151, a 6 mW random polarized red HeNe with a spec'd mode spacing of 438 MHz. See Whacky Mode Display of Melles Griot 05-LHR-151. This was interesting. Now the display was definitely different with the 3 to 4 modes of the 05-LHR-151 further apart and moving in a counter-intuitive way, which is more along the lines of what was originally expected if aliasing was in play. The two large peaks roughly centered on the screen and the two smaller ones between them represent the 4 lasing modes of the 05-LHR-151. The ones on either side are repeats.

    But the clincher was to view the split modes of an Agilent metrology laser. The particular unit used is a 5517 laser with an N1211A tube, which has a lower split frequency than most others and is thus a more demanding test of SFPI resolution. The results are quite impressive. Split Modes of Agilent N1211A Two Frequency Laser shows the display. Each of the two large peaks is separated by the effective FSR of the SFPI, 128.9 MHz. Within each peak, the two frequency components of the split modes are clearly visible, 1.4 MHz Apart. At the resolution obtainable, they only go down to about 1/2 the peak value between them but that still represents an effective resolution better than 1 MHz with a finesse of over 160. Estimates of finesse use the data in the chart at Transmission of Fabry-Perot Resonator versus Optical Frequency. Finesse then is equal to the FSR divided by the FWHM of the peaks. For this case where the dip is about 50 percent, the actual FWHM is decreased by an additional factor of ~0.575 due to the overlap. With better alignment of both front and back mirrors, a mode-matching lens, and mirror cleaning, the resolution may be significantly better.

    Sam's Selectable FSR Mode-Degenerate Scanning Fabry-Perot Interferometer 1 Displays shows the 05-LHP-071, 05-LHR-151, and N1211A Lasers side-by-side.

    As with the confocal cavity, when perfect alignment is approached, some of the peaks change in amplitude or disappear entirely. Ultimately, only one would remain revealing the raw finesse of the cavity, in this case over 600. But the reflections (primarily) from the front mirror also go directly back to the laser and tends to destabilize it causing erratic behavior and possibly lose of lock. It settles down if an optical isolator is added with its input polarizer orientated at 45 degrees so that equal amounts of both the H and V polarized components pass through it. But the performance improvement is not enough to justify the added hassle of the isolator and noisier signal due to the reduction in beam power. Simply adding a ~1 mm aperture close to the laser to block much of the reflected beam (which is larger) was almost as effective.

    To calculate the expected locations of the modes for the 05-LHP-071 and 05-LHR-151 lasers requires taking mod(n*MS,FSR) for (n=1;n To account for the slight discrepancy between the spec'd mirror RoC of 100 cm which would result in an N/k of 4/1 length of 29.3 cm and the actual value of 29.063 cm, the SFPI FSR becomes 128.94 MHz. The "correction" was which changes the mirror RoC to 99.19 cm was done to make the calculated and measured values for the 05-LHR-151 agree. And the 29.063 cm spacing is within measurement uncertainty of the actual value for best resolution determined experimentally. But then the actual RoC of these mirrors must be only 99.2 cm. My complaints have been relayed to the manufacturer. ;-) The results are:

                   Mode        Mode Locations
        Laser     Spacing      M1         M2         M3        M4
      05-LHP-071  636 MHz  120.2 MHz  111.5 MHz  102.8 MHz    -
        Calculated Differences    8.7 MHz    8.7 MHz
          Measured Difference    11.5 MHz
      05-LHR-151  438 MHz   51.2 MHz  102.3 MHz   24.6 MHz  75.7 MHz
        Calculated Differences   26.6 MHz  24.6 MHz  26.6 MHz
        Measured Differences     26.3 MHz  24.6 MHz  26.3 MHz

    Note: The differences for the 05-LHR-151 are not necessarily between adjacent values.

    It is not known why the measured difference for the 05-LHP-071 is so far off. But if the actual mode spacing was 633 instead of 636 MHz, it would be very close. Or perhaps the 438 MHz mode spacing of the 05-LHR-151 is not quite correct. Or both. And since they depend on the cavity length of the laser, a small error could also account for the discrepancy. I have not confirmed either since the tubes are inside head cylinders and no RF spectrum analyzer is available for measuring the mode spacing beat frequencies.

    Other mode-degenerate settings that might make sense are orders (N/1) of 5/1, 6/1, or 7/1. These would not have as high a resolution but would allow for shorter instruments to be constructed using the same mirrors.

    Sam's Selectable FSR High Resolution Spherical Mode-Degenerate Scanning Fabry-Perot Interferometer 2

    A week or so later, the "instrument" described above was rebuilt with 8 inch rod extensions so that it could operate with N/k=3/1 at a cavity length of close to 49.6 cm, half the confocal spacing. Rather than try to locate a suitable slab to mount this, plates were fabricated for the alignment screws to attach to the cage plate at the front and the adjustable mount at the rear as shown in Sam's High Resolution Selectable FSR Scanning Fabry-Perot Interferometer 2. The only disadvantage is that not being as massive, it likes to walk around too easily. ;-) Construction could also have been done with a single set of 24 inch Thorlabs cage rods - or an Invar or quartz tube. The FSR from determining previously that the mirror RoCs slightly less than the 100 cm spec and closer to 99 cm is around 101 MHz. The finesse is over 225 with a resolution of better than 750 kHz. Even lasing lines closer than 500 kHz apart would be easily detectable.

    The suggested Thorlabs parts would be similar to what I used:

    The SP33 would be used for the front plate allowing for the optional SM1D12 iris to be installed. The CP32 is for the front mirror with its home-built adjustable mounting plate. (The actual implementation differs slightly from the photo which shows a CP33 in that location.) However, after constructing and testing the ultra-high res SFPI below, this one was rebuilt using its frame (extended to 22 inches) with the front mirror on the Newport U50-A mirror mount and Thorlabs T12 linear translation stage, which made setting the mirror spacing an order of magnitude quicker.

    Alignment with the longer cavity is even more critical and best performance was only achieved close to the optimal setting when sometimes 2 of the 3 peaks were reduced in amplitude. But the display of a laser with a linewidth (including split lines) less than the 101 MHz FSR would still be unambiguous. It did cooperate for the photo though. ;-) And the result was close to the calculated resolution tested with an Agilent "tuned" (magnetic field-reduced) to produce a split frequency of around 750 kHz. It looks basically similar to the display using the first version of this SFPI with N/k=4/1 and the laser with a 1.4 MHz split frequency. In both cases, the dip between the split modes extends to around 1/2 the peak amplitude. With the laser adjusted to produce a 2.0 MHz split frequency, the two lines are nearly totally resolved with the dip going down to around 10 percent of the peak value as shown above. For lasers like the 5517D (3.4 MHz minimum split frequency), they would appear totally separate.

    Sam's Selectable FSR Mode-Degenerate Scanning Fabry-Perot Interferometer 2 Displays shows the 750 kHz and 2.0 MHz split frequency lasers, along with what happens when the Zeeman magnetic field is pushed too high in an effort to boost the split frequency. The small "rogue" modes on either side of the split mode are actually separated from it by approximately 1.18 GHz, the normal longitudinal mode spacing of the laser tube. But they appear only a few MHz away from the split modes due to aliasing of the high resolution SFPI with an FSR of only around 100 MHz. The split frequency has increased slightly to around 2.2 MHz but will not go much higher regardless of the magnetic field, and the rogue modes will increase in size at the expense of the desired split mode. If they become large enough, the laser will no longer be able to lock, and errors may result in measurements. The change in peak amplitude is probably reasonably accurate as thhe power in the split mode does decrease at higher magnetic fields and/or as rogue modes appear. (A normal SFPI or LSA would also display the rogue modes and at the correct location.)

    Sam's Selectable FSR Nano Mode-Degenerate Scanning Fabry-Perot Interferometer

    From the 1/2 meter high resolution SFPI, we now take you to the other end of the size spectrum - an SFPI head just over 1 inch in total length. It is built using Thorlabs 16 mm cage parts and the same 4.2 cm RoC mirrors as used in my other short home-built SFPIs. To fit in the 1 inch frame, the cavity parameter N must be greater than or equal to 4 with K=1. For N of 4, the mirror spacing is around 1.23 cm. Since the PZT voltage to scan one FSR is inversely proportional to N, it is able to use a smaller less sensitive PZT beeper 20 mm in diameter, which required less trimming to fit the 16 mm cage frame. And there is no dual polarization detector (for now at least) as that would add to its length. :) See Sam's Nano Selectable FSR Mode-Degenerate Scanning Fabry-Perot Interferometer. The Thorlabs assembly is essentiall a 1 inch cube. ;-) Initial mirror spacing was first set with a machinists' scale for the 1.23 cm of N=4. Then with the cage assembly just sitting on a lab jack, it was fine-tuned by loosening the grub screws on one side and pushing the cage plate with the front mirror forward or backward. The free play permitted a movement of a few microns either way with each iteration. And aside from it weighing too little to really stay put, just positioning it on the lab jack by hand sufficed to optimize the horizontal alignment. A fancy pan/tilt kinematic mount for the SFPI head is not really needed, just something like an adjustable height post with a ball joint attachment.

    The suggested Thorlabs parts for the frame are:

     Qty  Part #    Description
      1   SR1.5-P4  Compact Cage Assembly Rod, 1.5" Long, Ø4 mm, 4 Pack
      2   SP02      Compact 16 mm Cage Plate with SM05 Thread, Two Retaining Rings
      1   LMRA8     1/2 inch Lens Adapter for 8 mm Optics

    The display shows the modes of a Melles Griot 05-LHR-151 633 nm ~5 mW HeNe. There are four longitudinal modes present - the two tall ones and the two short ones on either side. The additional small blip is believed to be a higher order spatial mode that appears from time-to-time with this laser. To double check, it was tested using an SP-470 SFPI. A similar artifact was present, although its location differed due to the relative FSRs of the two SFPI. As with aliasing of the high resolution SFPIs, it just happens to look like a fifth longitudinal mode.

    The performance is quite good even without any focusing lens. One could be added but it would have to be glued to the front - there is no space for an additional cage plate or for the lens to be squeezed in with the front mirror adapter. Testing with a ~35 mm focal length lens did not make any obvious improvement, though no great effort was made to optimize its position. However, one benefit was that the reflection from the front mirror changed from a large blob to a small spot, which could perhaps more easily be aimed to avoid entering the laser aperture and destabilizing it. What also might help would be an aperture in front of the photodiode, but that isn't going to happen because it would require disassembling the entire back section with the PZT. The spots on the mirror appear to be tight though so it may not matter much.

    While the Nano SFPI won't compete with one using the mirrors with the confocal spacing due to lower finesse coupled with the wider FSR, it is more than adequate for testing common lasers and for use with the Mini Laser Mode Analyzer (mLMA1) display.

    Sam's HP/Agilent Laser OC Mirror Mid-Size Scanning Fabry Perot Interferometer

    This one isn't that spectacular in terms of high resolution or difficulty in construction :) but it does deserve a place here. Its purpose was actually to determine if an iodine cell could be placed inside the SPPI cavity to achieve a multiplication in sensitivity to absorption without requiring it to be inside the *laser* cavity, requiring a HeNe Brewster tube. Those experiments were inconclusive though as the finesse dropped by a factor of 5 to 10 due to factors unknown. But the SFPI survives.

    The mirrors from nearly all HP/Agilent/Keysight lasers from the 5500A through the 5517GL use OC mirrors with an RoC close to 13.6 cm. The reflectance varies from around 98.8% (5501B/5517A) to less than 97% (5517GL). Higher R is generally desirable for an SFPI used for display (rather than as an etalon). 5501B and 5517A tubes have the 98.8% R and unusable or dead lasers are generally available at modest cost. While the tubes are not healthy, the mirrors are generally fine and can be repurposed for an SFPI. However, extensive chants and incantations to the gods of dead lasers are essential PRIOR to discombobulation to have any chance of success. Additional possibly disastrous consequences are not known. ;-)

    This mid-size SFPI uses three-screw New Focus mounts for both mirrors which provide for precise cavity length tuning. A Thorlabs 10 mm to 1 inch adapter is used to hold the input mirror and the typical PZT beeper element was used for the output mirror. See Sam's HP/Agilent Laser OC Mirror Mid-Size Scanning Fabry Perot Interferometer. This rig works quite well as an SFPI with decent finesse and ease of alignment. It was also useful to confirm that the mirror RoC was quite close to 13.6 cm. ;-) Of course, with a confocal spacing of 13.6 cm, the FSR is only ~551 MHz meaning that aliasing would occur if attempting to display the modes of a red (633 nm) HeNe laser with a gain bandwidth of 1.5 to 1.6 GHz. However, knowing that would still enable it to be useful as a laser spectrum analyzer. And regardless, to confirm single longitudinal mode operation of a stabilized HeNe.

    Sam's Folded High Resolution Confocal Scanning Fabry-Perot Interferometer

    Not being content with less than the absolute best resolution from these ~1 meter RoC mirrors, yet unwilling to deal with a 1 meter+ long instrument, an SFPI was built using most of the parts of the one above, but with addition of another ~50 cm arm and planar HR fold mirror. See Sam's Folded High Resolution Confocal Scanning Fabry-Perot Interferometer (Diagram) Originally, I was considering a double fold, but since the losses from the planar mirrors on each pass are approximately twice the mirror transmission - and this would impact resolution - I knew I would not be content with that sacrifice even if it were undetectable. So this uses a single 633 nm HR planar fold mirror which was tested to have a transmission at 633 nm of less than 0.005% over at least 10 degrees with respect to normal incidence. It's virtually impossible to measure the actual %R and it's not equal to 1-T due tolosses within the mirror coating, but it should not be far off. Even if 10X of that value, it's still only order of 10 percent of transmission of the main SFPI cavity mirrors. The spec on these mirrors is R of 99.95% worst case compared to ~99.5% for the cavity mirrors, a 10X difference, so even that is a factor of 10.

    See Sam's Folded High Resolution Confocal Scanning Fabry-Perot Interferometer (Photo). The main cage frame has the same KC1 adjustable mount at one end with its PZT and cavity mirror. Mounted next to it on a 1.5 inch long frame is the other SFPI cavity mirror on its home-built mount with another cage plate for the optional iris diaphragm. At the other end is the planar HR on a small Newport (sorry Thorlabs, it's what I had available!) adjustable mount. The only additional issue during alignment compared to a linear cavity is to make sure both the outgoing and return intra-cavity beams are coincident on the planar mirror. This will guarantee that they hit the cavity mirrors with perfect normal incidence, thus avoiding astigmatism and differences in the effective X and Y RoCs for the mirrors, which would potentially compromise the mode-degenerate condition and limit resolution. Even worst case, the effect would probably be very small, but when attempting to achieve better than 250 kHz resolution, every little bit counts. ;-)

    Initial alignment was done using a Melles Griot 05-LHR-151 ~5 mW random polarized HeNe. From the two selectable FSR high resolution SFPIs, above, the actual RoC of the mirrors has been determined to be very close to 99 cm. I just use 100 cm in the photo caption because that is their spec'd value and is a nice round number to keep things simple. ;-) But knowing that it's really 99 cm, the initial setting of the mirror spacings could be done quite accurately. Then the trick is alignment. It's definitely trickier with the fold mirror, but by progressing from input to the far cavity mirror and back, it was possible to obtain a signal relatively easily. Fine tuning to tighten up the spot arrays on all the mirrors took more time goning back and forth. Tweaking an adjustment, then peaking all the others to see if the result was better or worse. It should be possible to do the X and Y separately as with mirror walking external mirror lasers, though that wasn't obvious from the behavior. At this point the finesse appears to be at least 200. The spots on the mirrors appear to be fairly tight so a focusing lens probably won't help. But a finesse of 500 or so should be possible. But a mirror dust-off could as they've been sitting around quite awhile.

    Even the crude home-built mirror mount for the input cavity mirror had more than enough sensitivity and stability. So the Newport mount could have been replaced with a similar one.

    Next up was the Agilent N1211A two frequency laser used to test the other high resolution SFPIs. It has a split frequency that can be adjusted from around 2.2 MHz down to less than 750 kHz.

    Using this laser, it was fairly easy to achieve performance better than with the mode-degenerate SFPI at 1/2 the confocal spacing using the same cavity mirrors, actual RoC = 99 cm. Tweaking it to approach the theoretical performance required careful alignment ending up nearly perfectly aligned such that every modes for every other FSR were partially or almost totally suppressed. That only required several more hours of fiddly work. :( :) It also involved a few mirror cleanings. But the resolution does indeed approach 250 kHz. Sam's Folded High Resolution Confocal Scanning Fabry-Perot Interferometer Displays shows the 500 kHz (with a span of around one and two FSRs) and 1.6 MHz split frequency lasers.

    Sam's Ultra-High Resolution Scanning Fabry Perot Interferometer 1

    Having built several large SFPIs, two of them being over half a meter long, I decided do a more manageable one 30 cm in length of even better resolution using the highest-R mirrors readily available at no cost - selected HeNe 633 nm HR mirrors. However, not all HeNe HRs are created equal, even for the same wavelength and radius of curvature. Some have lower efficiency and thus higher losses, which reduce finesse. But the problem is that by definition an "HR" mirror is supposed to reflect 100 percent at the design wavelength. While they usually aren't quite perfect, a typical cherry flavored (red) HR will block 99.99% or more at 633 nm allowing 0.01% or less light to pass through. That would be OK if the mirror were specifically designed as an ultra-high performance reflector for an interferometer, but these generally don't come close as there can be losses within the coating as well as scatter. Those can contribute way more than the 0.01% transmission, thus reducing the SFPI finesse. Super-polished HR mirrors for maximizing intra-cavity power in particle counter and similar applications may qualify. The typical reflectance is >99.99% at 633 nm and they are designed for 10s of WATTs of intra-cavity power.

    This SFPI was designed to used 60 cm RoC super polished HR mirrors and is currently set up for 30 cm, 1/2 confocal mode-degenerate spacing. The spec for the mirrors is >99.99%R at 633 nm and they were tested to have a transmission of around 0.008%. They came from adjacent cells in the original package so are likely to have very close to identical RoCs. Assuming 99.99%R, the theoretical finesse for the 1/2 confocal spacing is ~10,000 with an FSR of 167 MHz. However, coming anywhere close to that finesse would require that the mirrors be absolutely perfectly clean and the input beam be mode-matched to the cavity. Based on past experience with the high resolution SFPIs, mirrors cannot be expected to remain perfectly clean for any length of time unless the entire instrument is totally enclosed, which is not how it is presently. And the requirements for this SFPI are much more stringent.

    In terms of alignment, the R of >99.99% means that with the beam from a 5 mW laser incident on the SFPI's input mirror, less than 0.5 µW gets through making adjustments more than a bit of a challenge. And with the N1211A laser used for testing of split frequency resolution, it's less than 0.05 µW, which is essentially impossible to even see except in a pitch-black room. So swapping from the 05-LHR-151 to N1211A meant trusting that the alignment of the mirrors did not change and only alignment of the SFPI with respect to the laser needed to be performed by aiming the reflections directly back to the laser, adjusting XYPT of the SFPI while looking for any sign of a signal, and hoping the laser remained locked.

    The structure is based on Thorlabs cage parts to improve mechanical stability using a frame around 14 inches long with a three-screw adjustable mount at back-end and another small ajustable mount on a micrometer linear stage at the front-end to simplify cavity length tuning. Shorter spacings would also be possible using the same frame but that seems silly. ;-)

    Sam's Ultra-High Resolution Scanning Fabry Perot Interferometer 1 shows the result. Before you get real excited, although the resolution is quite impressive, alignment is a total pain in the you-know-what and the signal is very low until optimal alignment is achieved. Even then, it's much lower than that of a similar "normal" SFPI. And the thing really needs to be clamped down to a rigid surface along with the laser. Having said that, based on the chart at Transmission of Fabry-Perot Resonator versus Optical Frequency, the finesse is estimated to be at least 1,500 and could be more than 3,000. The testing was done with an Agilent N1211A laser with a split frequency of around 520 kHz. The appearance of the split mode was somewhere between the curves for 5 and 10. Then the SFPI's finesse is approximately that finesse multiplied by the ratio of the FSR (~167 MHz for the 1/2 confocal spacing of 30 cm for the 60 cm RoC mirrors) to the split frequency (520 kHz). The resolution is therefore at least 50 kHz and could be below 25 kHz. With this long SFPI, back-reflections with optimal alignment tend to cause the laser to unlock. So, alignment may not be quite optimal. And as noted above, with the mirrors exposed to the environment, they will collect dust and every little bit will reduce finesse, with a much more degrading effect than with the 99.5% or lower R mirrors of most of the other SFPIs.

    Capturing photos of the displays for this beast was quite a challenging and lengthy endeavor. And the results are still quite ugly, but here they are: Sam's Ultra-High Resolution Scanning Fabry Perot Interferometer 1 Displays. View and weep. ;-) As can be seen, faithfully showing the display of the Spectra-Physics 117A stabilized single frequency laser with a span of 1 FSR or more is quite futile on the scope as the peak would be a hair's-width spike - 1 part in more than 1,500 of the screen width. Once locked, the SP-117 is very stable so the single output lasing mode does not move around much and the short term line-width is no more than 1 kHz and may be much smaller and thus does not really contribute to the width of the displayed peaks. But with the analog scope, the spot size is more that 5 times the FWHM of the peak based on the finesse of the SFPI. :-) And the bandwidth of the detector widens it further. (The difference in peak amplitudes is due to near perfect alignment suppressing 2 of the 3 FSRs for the half confocal SFPI.) And it was barely possible for the SP-117A to maintain lock with the back-reflections from the SFPI even with an optical isolator and additional attenuation. But it cooperated long enough for the photos. Obtaining the expanded photos of the display of the split frequency of the Agilent N1211A two frequency laser was even more involved. Unlike the SP-117A, the short term frequency stability is not very good for these in that the optical frequency of the lasing line moves back and forth by up to 1 Mhz or more on a time scale of seconds. What do you expect from a $10,000 laser?! :) The only way it was even possible to get anything recognizable was to trigger the scope on the photodiode signal. Using the ramp or sync from the ramp driver (function generator) was even more unstable. And then it only required order of 50 shots to capture one that was usable. However, I do believe this is an accurate rendition of what was going on. The spacing between the bumps really did change with as the split frequency was tuned between 1.6 MHz and 520 kHz. Really. ;-)

    The signal was so weak initially that the photodiode preamp gain had to be fairly high and the noise was greater than the signal. :( :) This also meant the bandwidth was so small that it wasn't possible to scan at a rate of more than a few Hz with the high finesse before the amplitude of the peaks declined way below the noise floor. Even after more-or-less optimal alignment where the gain could be reduced to the lowest setting (with the original ~5 mW Melles Griot 05-LHR-151 laser) and the preamp bandwidth is then much higher, this was still a major issue. The capacitane of the photodiode will limit response as well. And with the Agilent N1211A at less than 1/10th the power, a higher preamp gain had to be used making the response worse. Unlike the ultra-high resolution SFPI using "proper" ultra-high quality mirrors designed for such purposes (and ultra-high $$$$ as well) described in the section Ultra-High Resolution Scanning Fabry Perot Interferometer, I don't think the speed of light is significant here in reducing signal level, though that can't be ruled out with a finesse in the thousands. In all fairness though, the perfomance of this SFPI is closer to the one using the high priced mirrors than I had expected, probably within a factor of 2 in finesse, though the efficiency is probably lower but not wildly so. I recall not so fondly fighting with alignment there as well.

    Generic Scanning Fabry-Perot Interferometer Testbed

    The assembly used for the ultra-high resolution SFPI above is essentially a general purpose setup that can accommodate various types of mirrors and mode-degenerate spacings and is essentially an enhanced version of the one introduced in the section: Scanning Fabry-Perot Interferometer Head Frame using Commercial Parts. While it doesn't have the stability of an instrument using 1 inch Invar bars, the stainless steel rods provide a decently rigid structure with minimal effects of temperature changes in the short term. Many different SFPIs can be constructed from the same basic set of parts as shown in Examples of Scanning Fabry-Perot Interferometers using Thorlabs Cage System Parts. The photos are all approximately to the same scale. All were built from Thorlabs 30 mm cage system parts except for the smallest (2) that somehow made it into this photo. :) Here is a list for components available from Thorlabs to construct (5) except as noted:

    Thorlabs parts

     Qty  Part #  Description
      1   Iris diaphragm (one of the following):
           SM1D12  SM1 Lever-Actuated Iris Diaphragm, 0.8-12 mm
           SM1D12D	SM1 Ring-Actuated Iris Diaphragm, 0.8 - 12 mm 
      1   Front plate:
           CP33 (CP11)  SM1-Threaded 30 mm Cage Plate, 0.35", 2 Retaining Rings
      1   Mid plate (optional):
           CP35  30 mm Cage Plate with Ø1" Double Bore, 8-32 Tap (optional)
      1   Translation stage (one of the following):
           DT12  1/2" Dovetail Translation Stage, 8-32 Taps
           T12X  Miniature 1/2" Translator, X Config, 2-56 Mounting Holes
      1   Front mirror mount (one of the following):
           KMSS  Compact Kinematic Mirror Mount, Hex Adjuster, 8-32 Taps 
           MK05  Mini-Series Kinematic Mirror Mount for 1/2" Optics, 4-40 Taps
           K05  Polaris Ø1/2" Mirror Mount, 3 Low-Profile Hex Adjusters
      1   Back PZT/mirror mount:
           KC1     Kinematic 30 mm-Cage-Compatible Mount for 1" Optic
      1   Rods for SFPI frame (selection among the following):
           ER4-P2  Cage Assembly Rod, 2" Long, 6 mm, 4 Pack
           ER4-P4  Cage Assembly Rod, 4" Long, 6 mm, 4 Pack
           ER4-P8  Cage Assembly Rod, 8" Long, 6 mm, 4 Pack

    Some of these parts like the T12X are quite pricey (almost $300!), over 3 times that of the DT12 and is essentially equivalent for this purpose. The T12X is included because it may turn up on eBay at a nice price. Other front mirror mounts than the MK05 or KMSS are equally suitable such as the Newport U50-A which is what I used since it was available. :) All will either require minor modifications or an adapter to attach the 7.75 mm diameter mirrors. However, some like the older Newport MM1 or the Thorlabs KM05 are too large to fit within the frame without trimming. Using 1 set each of the 2 inch and 4 inch cage rods, and 2 sets each of the 8 inch cage rods, will enable a frame of arbitrary length to be assembled in increments of 2 inches up to 22 inches. Much longer and it would probably be too wobbly. If the size is known and fixed, it may be possible to use different sets of cage rods requiring fewer pieces at lower cost.

    In addition some simple parts fabricated from scrap aluminum and common hardware will be required for the leveling feet and spacers, as well as the mounts for the PZT and photodiode:

    Other parts

     Qty   Description
      2    7.75 mm diameter cavity mirrors depending on your desires :)
            Example: 100 cm RoC, 99.5%R@590-650nm.
      1    7.75 mm adapter for front mirror to 1/2" mount
      1    7.75 mm adapter for back mirror to PZT
      1    27 mm diameter PZT two-wire beeper element, drilled with center hole
      1    1 inch OD x 3/8 inch long aluminum cylinder for PZT mount (optional)
      1    Aluminum plate (2x5x3/8" inch) for stage mount and front foot
      1    Short (e.g., 1x3/8x2 inch) aluminum bar for rear foot
      1    Spacer for translation stage mount
      1    3/4 inch Spacer between front plate and cage plate
      3    6-32 x 2 inch cap-head screws with pointed tips for height adjustments
      1    Piece of Perf. board for photoiode and connectors
      -    Miscellaneous mounting screws and other hardware

    First of course are the mirrors which are typically 7.75 mm in diameter. An adapter for the 7.75 mm mirrors to fit the 1/2" mirror mount will be needed. While Thorlabs has these for 1 inch mounts, there is nothing similar for 1/2 inch mounts. I made one from a short length of 1/2" aluminum rod stock drilled ~8 mm with a soft-tipped grub screw, making the front mirror easily replaceable. A similar adapter can be glued to the PZT since mass is not really critical with µm-size movement.

    For high resolution SFPIs with mirror reflectivity approaching 100%, a dust cover will also be essential to minimize contamination. And an opaque shield surrounding the photodiode will be needed to block room light especially when using higher gain settings on the photodiode preamp.

    With these basic components, a variety of SFPIs can be quickly constructed with mirror spacings from a few mm to a meter or more in length by sliding the front cage plate along the rods, and/or by replacing or augmenting the rods. And with the front mirror mount on a linear stage, optimizing mirror spacing is simpler than using the three screws of either mirror mount, though readjustment of the rear or front mirror alignment screws may be necessary after a few turns of the stage micrometer. The iris comes in handy when centering the beam and to reduce the effects of the interferometer formed by the output mirror of the laser and the input mirror of the SFPI.

    Taken together, this makes it trivial to experiment with different mirror sets or mirror spacings. That is, aside from the required 37 hours of fiddly alignment in a pitch-black room if using HR mirrors in a meter long frame. ;-)

    Sam's Scanning Fabry-Perot Interferometer Driver 1 (SG-SF1)

    I have also now designed a stripped down function generator especially for driving the PZT of these SFPIs. See Sam's Scanning Fabry-Perot Interferometer Driver 1. This unit generates a variable frequency triangle (approximately 5 to 200 Hz) or sawtooth (approximately 10 to 400 Hz) with a full range adjustable offset. The output amplitude may be set from 0 to to over 25 V p-p, with both non-inverted and inverted outputs. Where neither side of the PZT is grounded, it may be connected between the two outputs to provide a voltage range of up to more than 50 V p-p. The output may also be set to DC and adjusted over the full range using the offset control for initial setup of the SFPI. The sawtooth has a slew-rate limited falling edge so there is no risk of damage to the PZT or excessive ringing at the start of the scan. The adjustable amplitude implements sweep expansion to enable examination of the fine detail of the laser spectrum. A photodiode preamp is built in to SG-SF1 completely eliminates the need for anything beyond an oscilloscope and +/-15 VDC power supply.

    Using better op-amps than the jelly bean LM358s might increase the maximum output voltage range slightly but at these frequencies, won't make much difference in any other respect. Of course, it would be trivial to modify this circuit for a different frequency or voltage range. But, as drawn, it will cover the needs of most SFPIs using "drum head" type PZTs, as well as the Thorlabs SA200 and SA210.

    Blank PCBs are now available for the combined PZT driver and photodiode preamp. The PCB for SGSF1 is just under 2x2.5 inches as shown in Photo of Sam's Scanning Fabry-Perot Driver 1. Power requirements are regulated +/-15 VDC at 50 mA. A suitable dual DC power supply would be trivial to construct using a wall adapter putting out 14 to 16 VAC, a pair of diodes, a pair of 1,000 uF, 25 V filter capacitors, and 7815 and 7915 (or similar) IC regulators. SGSF1 may be used as shown, or built into a project box with front panel controls in place of the switch(es) and trimpots. A selector switch and fixed resistors may be substituted for the pot to provide calibrated expansion factors (e.g., 1X, 2X, 5X, etc.). And the offset control could be a 10 turn pot. For more info, please go to Sam's Classified Page or contact me via the Sci.Electronics.Repair FAQ Email Links Page. Note that the parts on these PCBs are labeled differently than on the schematic above. Please refer to: Sam's Scanning Fabry-Perot Interferometer Driver 1 (PCB Version). And, no, I haven't the slightest intention of explaining why since it makes no obvious sense. ;-)

    Dual Polarization Detectors for Scanning Fabry-Perot Interferometer

    For displaying the longitudinal modes of some types of lasers, it is desirable to be able to view the polarized modes separately. This would probably be most useful for common random polarized HeNe lasers and Zeeman-split HeNe lasers, though it could also be used with other types of lasers that are not polarized. (These's no benefit with linearly polarized lasers.)

    All that's required is to replace the normal sensor with a polarizing beam-splitter and a pair of photodiodes (and possibly separate pre-amps). The polarizing axes of the detector will need to be oriented to be the same as those of the laser, but most SPFIs allow for the sensor to be rotated in its mount. For axial Zeeman lasers producing circularly polarized outputs, a Quarter-Wave Plate (QWP) would be required to convert to linear polarization. A QWP is usually part of a commercial Zeeman laser tube assembly. However, if you're rolling your own, then an external QWP will need to be added.

    This is most dramatic with a fancy digital scope so the polarized modes can be shown in living colors in real time. :) Not real fancy, but see SFPI Dual Polarization Display of Melles Griot 05-LHR-911 Modes on DSO-Quad™ Miniscope. However, any old 2 channel scope will work. Or, the SFPI can be swept at low speed if necessary with the PD outputs fed to two channels of a PC data acquisition system. Where the polarizing beam-splitter is wavelength sensitive, a separate unit may be needed for each mirror set/wavelength range of interest.

    As an example, for the SP-470, the normal sensor assembly can be replaced with a short piece of 1/2 inch PVC pipe filed down to fit into the SPFI housing into which a 4 or 5 mm PBS cube and inexpensive photodiodes are mounted. The PD signals can be fed directly to two scope channels with only 10K load resistors, or through suitable preamps depending on the laser power involved. The one I built worked reasonably well, but there was between 5 and 10 percent crosstalk, due to a combination of the PBS cube not having perfect separation, not being optimally oriented, or reflections off the faces of each photodiode getting into the other channel due to their close proximity. In fact, even the best polarizing beam-splitter cubes may have substantial residual P-polarization in the reflected S-polarized output, though the transmitted P-polarization is generally quite pure. (A non-polarizing beam-splitter with pieces of sheet polarizer mounted in front of each photodiode might actually have better separation.) So, rather than ripping the PD assembly apart and then finding that the crosstalk was still present even after extensive fiddling and frustration, I built a two stage preamp which includes a means to suppress it by subtracting a small (adjustable) amount of channel 1 from channel 2 and vice-versa. The trans-impedance (first) stage needed to be a high speed op-amp, a MC33077 because it was handy. With the original jelly-bean LM358, there was serious overshoot and ringing at all but the lowest gain settings. And with small capacitors (e.g., 10 pF) across the feedback resistor, the waveform still had artifacts if the amplitude was more than a few hundred mV. While a larger (100 pF) capacitor resulted in a clean output, the frequency response was reduced to where the SFPI resolution would be compromised at the normal scan rate. The second (subtracting) unity gain stage was perfectly happy with an LM358. The crosstalk cancellation works quite well when SFPI alignment is close to optimal. Otherwise, there may be a small phase shift between two channels making perfect cancellation impossible.

    A similar crosstalk cancellation technique may be applied elsewhere to allow for the use of less expensive polarizing optics or where alignment cannot be easily optimized. The cross-gain for each channel should be set up with a pure linearly polarized beam aligned with its axis while nulling the output of the other axis assure optimum orthogonality.

    Dual Polarization SPFI Display of HeNe Laser with Higher Order Spatial Modes shows a series of screen shots from my antique Tektronix 465B analog scope (sorry, no color). All the peaks with a height of more than about one screen division are the normal longitudinal modes and they behave, well, normally. However, the smaller peaks appear to belong to a weak non-TEM00 higher order spatial mode. This isn't necessarily a laser where the dual mode display provides a critical benefit, but being able to easily see how the rogue spatial modes behave with respect to polarization is somewhat illuminating. Sorry, pun intended. :-) While the dual polarization display doesn't help to reveal the existence of the higher order spatial modes, it does shed light on how they associate themselves with normal modes, or actually run away from them. Originally (with the normal SFPI), I had assumed that the little blips closest to a larger normal mode were of the same polarization, not orthogonal to them, which is the actual situation.

    And I couldn't resist creating a complete design including the Sam's Dual Polarization Photodiode Preamp 1 Schematic and Sam's Dual Polarization Photodiode Preamp Front Panel Layout. A populated PCB is shown in Sam's Dual Polarization Photodiode Preamp 1 Populated PCB. It hasn't been fully tested though. This same design can also be used for general data monitoring or capture using a data acquisition system. For use with an SFPI, power would most conveniently be provided by a 14 to 16 VAC wall adapter feeding the AUX PWR connector, J3. But for other applications, the the power pins on the output connector can be used. The BLANK PCB is available.

    SG-DP1 Specifications

    These assume the component values shown in the schematic:

    Most common random polarized red (633 nm) HeNe lasers produce only two sets of orthogonally polarized longitudinal modes, generally fixed with respect to the tube orientation. However, this is not always true and there can be modes at arbitrary angles. One example is of a Zeeman-split laser that was intended to be a clone of an HP-5517D. However, it used an HeNe laser tube that was too long, resulting in a pair of low level rogue longitudinal modes in addition to the normal Zeeman modes. In that case, the resulting modes (after the waveplates) were linearly or elliptically polarized and orthogonal to each-other, but oriented at 30 degrees with respect to the primary modes. (The normal linearly polarized output modes result from circularly polarized tube modes, but for the rogue modes to be off-axis may mean that they were elliptically polarized tube modes.) The dual polarized mode SFPI would have easily identified those since they would show up on both channels and the detector assembly could then be rotated to maximize channel separation. Where pure linearly polarized modes could not be achieved, the degree of elliptical polarization and orientation could be easily determined. Hmmmm, the detector assembly will need to be mounted on a calibrated rotation stage. ;-)

    Note that the success of the dual polarization detector assumes that the SFPI itself is polarization insensitive. An SFPI has no polarizing components, but thin film mirror coatings may have a very slight polarization anisotropy even at normal incidence. This would not be a problem for most applications, but where hundreds or thousands of reflections are involved as in an interferometer, even a small polarization anisotropy in one of both mirrors could add up and result in a polarization preference or other peculiar behavior. So, it's worth checking the SFPI with a linearly polarized input and confirming that the output polarization remains pure and tracks the input as its orientation is changed. I don't know if this is ever an issue with commercial SFPIs though since a polarization asymmetry would also result in a response that depends on the laser polarization or SFPI orientation even with the normal polarization insensitive detector - something that should not generally be present.

    Another possible implementation of a dual polarization detector that would have no (new) issues with mirrors or require constructing a totally new sensor assembly would be to use the normal detector with a switchable polarization selector in front of the SFPI (like the LCD switch in HP/Agilent lasers but possibly faster). The polarization would be oriented horizontal and vertical on alternate scans with appropriate scope triggering, or with a pushbutton for manual selection. To avoid long term damage to the LCD, the average voltage applied should be 0 and thus AC. This was tested with a 555 timer driving the LCD switch and DC power to it simply being turned on and off manually. Initially the frequency was set around 100 Hz, similar to what's used in the laser. This slow speed resulted in the display bouncing around quite a bit and allowed some residual wrong polarization to get through. Increasing it to 1 kHz helped. But swapping polarity once per scan or on alternate presses of the button might work better and still keep the LCD reasonably happy. Adding a second 555 timer to do the switching was irresistible and Polarization Switch 1 was the result. The prototyping board has the two 555 timers with adjustments for switching speed and LCD drive frequency. The output to the LCD is capacitively coupled to assure that there is no DC. The black object is the top half of the beam sampler assembly from a 5517A laser with the photodiode removed and a hole drilled in its PCB for the beam to pass.

    While I assume that every patent for SFPI technology has something along these lines of: "Although specifically described with respect to a single detector, anyone skilled in the Art will recognize that this approach can be used with multiple characteristics of laser light such polarization or wavelength each with its own detector.". Nonetheless, should a major company read this section and decide to market a product, I DO expect royalties. OK, right, pigs will need to fly first on both counts. ;-)

    Kinematic Micro-Aligner for Planar Scanning Fabry-Perot Interferometer

    Cavities using two planar mirrors are much more difficult to align than those with at least one spherical mirror. Commercial high performance SFPIs with planar cavities typically use a set of three PZTs to both control cavity length and adjust the fine alignment. For home-built planar SFPIs may benefit from being able to fine tune the alignment of at least one mirror, though it is not essential. But it is simple and inexpensive using the same type of beeper element PZTs as used for the cavity length control.

    Initially, I thought about replicating the triple-PZT arrangement of the commercial SFPIs, but only for micro-alignment, while maintaining the same single PZT for cavity length control. But then I realized that is separate, a kinematic arrangement with two PZTs independently driving the angle in X and Y would be simpler to build and easier to use. The three PZT arrangment for micro alignment in commercial SFPIs is just an artifact of using three PZTs for the cavity length control. The result is Kinematic PZT Micro-Aligner for Planar Scanning Fabry-Perot Interferometer which shows the major parts and sequence of steps for assembly. The main "holey" PZT (27 mm) is is on attached to a spacer ring that contacts two smaller (20 mm) PZTs near their centers and a fixed pivot via "bumps" - essentially small semi-flexible spacers. The net result is similar to a common kinematic mirror mount but with micro motion. With a pair of potentiometers each prividing 0-24 VDC via a voltage divider, the total range should be more than 0.1 mR in each axis. These PZTs can safely be driven with a higher voltage for a wider range if desired. And it will fit on a Thorlabs KC1 adjustable mirror mount in a 30 mm cage assembly so can easily be substituted for the "normal" back mirror on its PZT. A short 1 inch diameter cylinder can be secured to the back of the plate and that would then be locked into the KC1. But since the actual motion is rotated 45 degrees with respect to the adjustment axes of the KC1, a slight disadvantage if going back and forth between them.

    Combined Longitudinal Mode and SFPI Dual Polarization Optical Head

    To the best of my knowledge, this provides capabilities not available in any commercial system: Namely the ability to display both the amplitude of the longitudinal modes during mode sweep AND the SFPI display of their individual amplitudes simultaneously. And with dual polarization for both. :)

    The prototype was built using Thorlabs 16 mm cage parts to be ultra-compact but a more normal size would work at least as well. (It's not as small as the on described in the section: Sam's Selectable FSR Nano Spherical Mode-Degenerate Scanning Fabry-Perot Interferometer though.) See Dual Mode Dual Polarization Optical Head 1. For the longitudinal mode display, a around 20 percent beam sampler (more or less polarization independent) was installed between the focusing lens and front mirror, Its reflected beam feeds a dual polarization detector. Another dual polarization detector is mounted behind the rear mirror for the SFPI display. These are simply a 4x4x4 mm Polarizing Beam-Splitter (PBS) cube with 3x3 mm area silicon photodiodes attached to two of its faces with index-matching optical cement.

    This particular head has the mirrors mounted at one half their RoCs which reduces the finesse somewhat but results in 4/3rds of the confocal FSR, around 2.324 GHz, which is better for the red HeNe with its 1.5 to 1.6 GHz gain bandwidth. See the section: Sam's Mini Laser Mode Analyzer 1 for more on its use.

    Sam's Proposed Stand-Alone Scanning Fabry-Perot Interferometer Controller

    The basic setup of the SFPI hasn't changed in any significant way in over 40 years: A ramp generator, photodiode preamp, and (user supplied) oscilloscope. While this is certainly adequate, what about alternatives made possible by modern low cost digital technology.

    One possibility would entail converting to a (PC or Mac) USB interface with both the user controls and display being on the computer. However, I would caution against rushing full speed into such a project as it's extremely difficult to implement a software driven user interface on a conventional personal computer that approaches the utility and convenience of real knobs! But this would open up the possibility of providing additional measurement capabilities as well as more advanced functions like closed-loop etalon control. And the full color display would add a touch of class. :)

    Or, how about fully custom hardware? While I rather doubt that the development of such a device could ultimately be justified based on expected profits due to a relatively limited market, it never hurts to dream. See Sam's Proposed Microprocessor-Based Stand-Alone Scanning Fabry-Perot Interferometer Controller. This would use a high resolution color LCD display with 10 soft keys and 4 soft knobs for all user interaction. The objective with the GUI would be to have the most commonly used functions expected in a conventional SFPI always available without descending into endless menus. In addition, capabilities not found in any commercial SFPI like sweep magnification (in addition to sweep expansion) can easily be provided, which zooms in on only a variable portion of the central area of the sweep. And when not being used for an SFPI, your kids will love the color Etch-A-Sketch™ app, included free for a limited time only. ;-)

    A university student has sent me their Senior Project proposal for a microcontroller-based SFPI system based on a commercial confocal SFPI head. Stay tuned for the exciting developments.

    Scientific Connections has apparently developed (or at least developed the specifications for) an advanced SFPI called EagleEye™ which interfaces via USB and in addition to the normal SFPI functions, can compute linewidth well below the resolution limit determined by the finesse and FSR of the confocal cavity. This is done by controlling the drive voltage while measuring the photodetector response to locate the FWHM points on a mode. They claim a raw resolvance of 1 MHz but a computed linewidth limit of 20 kHz. This may all be vaporware though - I have asked about it via their Website.

    In fact, EXFO/Burleigh had developed a similar concenpt, though like the system above, it may never have been fully commercialized, and is now officially discontinued. The NuView FPS-250 consisted of a USB controller and EXFO NuView Laser Spectral Analysis Software. Together, these would provide control of a conventional SFPI, and display and analysis on a PC.

    Here's another possibility: The "DSO-Quad™" is a pocket 4 channel digital color storage oscilloscope which is based on open source hardware and software. Search for "DSO-Quad" and you'll get more than you ever wanted to know about it, but the features relevant for an SFPI are 2 analog input channels with a combined sample rate of at least 36 Ms/sec. Add a dual channel photodiode preamp and this could represent a nice compact instrument. I was hoping the built-in function generator could provide the ramp signal, though an external amplifier would be required if using an SFPI head requiring high voltage (but perhaps not if one of mine or Thorlabs). Unfortunately, I found that the triangle output of the DSO-Quad™ is quite pathetic with a dozen or so steps :( at best, but it could provide a trigger for an external ramp generator or its squarewave could drive an external integrator to generate a triangle waveform. And mastering the rather strange user interface has so far been a challenge for me at least. But I haven't given up yet as I believe the DSO-Quad™ still has potential and while there are limitations, it is usable. SFPI Dual Polarization Display of Melles Griot 05-LHR-911 Modes on DSO-Quad™ Miniscope is a screen capture (built in "Save BMP"). However, this had to be done at a an SFPI scan rate of 1 Hz. Any faster and the sampling would be hit or miss for narrow peaks so their amplitude would appear to vary dramatically when in fact they were nearly constant.

    Sam's Proposed Tablet PC-Based Scanning Fabry-Perot Laser Spectrum Analyzer

    A tablet PC would be an ideal platform to implement the ultimate SFPI controller and display without the need to design special purpose hardware other than for the ramp driver, photodiode preamp, and their USB interface. An optimized touch-screen-based GUI could enable all functions to be carried out in a natural way while eliminating the costs (as well as retro appearance) of hardware knobs and buttons! And most of the software infrastructure will already be present allowing for natural finger gestures to perform most common functions. For example, spread fingers to control expansion or magnification; touch and drag to change position (centering), or pull on selected mode peak to change preamp gain or use soft arrows. The tablet PC screen layout could look similar to what's in the diagram above, but with all interaction via the touch-screen and thus no need for physical knobs and buttons. What what a great excuse to buy a tablet PC (or better yet, have your company buy one for you!). Who could resist? :)

    Here are proposed specifications for a complete system, tentetively called the LSA100, consisting of an SFPI head with built-in USB interface attached to a touch screen-based tablet computer as depicted very roughly in Sam's Proposed Tablet PC-Based Scanning Fabry-Perot Laser Spectrum Analyzer. No other equipment would be necessary to perform display and analysis of laser mode structure, with export of the resulting data to other apps. The following lists the common specifications for all versions not dependent on FSR or mirror wavelengths.

    Sam's Mini Laser Mode Analyzer 1

    Mini Laser Mode Analyzer 1 (mLMA1) was originally intended as a proof of concept for implementing an SFPI with any head that uses low voltage PZT driver. But mLMA1 turned out so well that it could have practical value, especially where a portable low cost instrument is desired for basic checking of the mode behavior of lasers intended for applications like holography. ;-)

    No function generator or oscilloscope is required as the "instrument" is totally based on a low cost Arduino-compatible microprocessor platform, LCD or OLED display, and a hand-full of equally low cost electronic components. Assuming the availability of suitable mirrors, the additional cost of SFPI head can range from $0 if built from junk-box parts to around $100 if using nice components from Thorlabs. In some ways it's a minimalist version of Sam's Proposed Tablet PC-Based Scanning Fabry-Perot Laser Spectrum Analyzer.

    While mLMA1 will probably not replace a $5,000 SFPI, it would be more than adequate to confirm SLM behavior of lasers intended for holography or interferometer, evaluate HeNe lasers for flipperitis and polarization behavior,

    In addition to the SFPI function, mLMA1 also provides a mode sweep display and rudimentary laser power meter. And these all have dual polarization capability: The P and S mode orthogonally polarized modes of a laser can be display separately or combined.

    There are two versions of mLMA1: V1 was the original proof of concept and is very basic with only a 128x64 pixel OLED display. V2 has more functions with a 240x240 pixel color LCD and its user interface is more polished. A kits of parts with custom PCB is available for V2. For more information, see the Mini Laser Mode Analyzer 1 (mLMA1) Manual.

    The $99 Scanning Fabry-Perot Interferometer

    While my $1 SFPI can be made to work, the choice in the types of mirrors that are typically available surplus or from salvage are severely limited. Alignment becomes extremely critical and an aperture is needed to suppress non-TEM00 modes. In addition, reflections back to the laser under test may be destabilizing. Though this is probably not a major issue with typical HeNe lasers or green DPSS lasers with an IR-blocking filter in their output, It could be significant for stabilized HeNe lasers and IR DPSS and other non-frequency converted lasers.)

    I lucked out for my $2 SFPI in just happening to have short radius mirrors that could be pressed into service, but most people wouldn't have this option.

    For my $3 SFPI, I do have mirrors available but there are really only useful for red HeNe lasers, for which they are nearly ideal.

    Being able to specify the mirror radius of curvature, wavelength, and reflectance, would greatly expand the possibilities and still result in an instrument for under $100. That's really not too bad considering it should have almost the same performance as a $9,999 commercial SFPI.

    (From: Christoph Bollig (

    Just some comments on the SFPI resonator options: The confocal configuration has the big advantage that it can be used at an angle or an offset! Most single-frequency lasers outputting at the fundamental (not frequency converted) don't like it if they get reflections straight back, and especially when those reflections are from a high reflectivity mirror and well aligned to go back into the laser. And that's exactly what you need to do with a plane-plane interferometer or even with most other non-confocal ones.

    With the confocal interferometer, the best choice would probably be to come in along the optical axis but with a slight offset. The back-reflection will then be at an angle. Since such an arrangement will need two round trips to reproduce, the second mirror can be HR and the "transmission" will be through the same mirror as the incoming beam, just at a different angle as shown in Confocal Scanning Fabry-Perot Interferometer. As you can see, there are no reflections back into the laser.

    Another advantage is that since the second mirror is high reflector, no hole is needed in the PZT. :)

    We have considered different options for the mirrors for use with near-IR lasers, but one of the more likely scenarios is to use a 50 mm RoC output coupler from CASIX with either 98 or 94 percent reflectivity (NDO0205, $50). see CASIX Nd Laser Optics. These are also available in other curvatures down to 25 mm). For the high-reflector on the PZT one could use one of the CASIX standard HR mirrors from the DPSS series (quite a few from CASIX Diode Pump Laser Optics Kits would do. For example, the DPO1301 or DPO1302 ($45) (or the green laser output coupler from Roithner, also 50 mm radius). Or the DPO1303 (HR at both 1,064 nm and 532 nm) which would then be useful for green DPSS lasers as well.

    Ultra-High Resolution Scanning Fabry Perot Interferometer

    John Barry at Yale University builds ultra-high resolution SFPIs for his work in laser cooling of molecules. (See, for example: Laser Makes Molecules Super-Cool.) If there's a scientific instrument for laser nerds to drool over, this would be a good candidate as it puts everything else described here to shame:

    (Portions from: John Barry.)

    The cavity is a simple confocal design consisting of a quartz tube, two 1018 steel end-caps, a small ring PZT, and two mirrors (RoC of 500 mm, and reflectivity of ~99.983% at 632.8 nm) based on the specifications from the manufacturer at Layertec Optics HR Mirror (Low Loss, >99.97%, 620-680 nm). While the specifications list 99.97% reflectivity, this is at the wavelength extremes (620 nm and 680 nm) and is lowest there. At 633 nm, the plot shows it to be at least 99.983%, and since they are conservative, 99.99% or even higher is not out of the question. One mirror is glued to a threaded steel piece for rough adjustment of the cavity length. The other mirror sits behind the PZT for fine cavity length adjustment. See Photo of Ultra-High Resolution Scanning Fabry-Perot Interferometer. The closeups at the bottom show the PZT with its mirror hidden inside on the left and other mirror glued to the steel piece on the right.

    The cavity is designed to be athermal. As temperature increases, the quartz expands and lengthens the cavity mirror spacing. Simultaneously the 1018 steel expands relative to the ends of the quartz where it is attached, thereby acting to decrease the mirror spacing. By balancing these two effects and accounting for smaller corrections (expansion of mirrors and PZT), the distance of the cavity can be made to be largely independent of temperature. Here are the specifications:

    The theoretical values are based on a mirror reflectance of 99.983% obtained from the plot on the Layertec Web page, above. Since this is probably a lower bound, potential performance may be much higher.

    Now this isn't exactly the sort of SFPI head you attach to a pan-tilt mount on a table top. It's normally located in a temperature-controlled environment inside a vacuum chamber with feedback to eliminate outside sources of error.

    With this spectacular finesse/resolvance, a plot of the Zeeman-split modes of an HP-5517A laser, separated by 1.65 MHz, looks like the SFPI display of a normal HeNe two mode laser where the modes are several hundred MHz apart! See Ultra-High Resolution Scanning Fabry-Perot Interferometer Display of HP-5517A Laser Modes. The two Zeeman-split longitudinal modes (F1 and F2) are 1.65 MHz apart with the span from 0 to 0.01 seconds being about 3.8 MHz (out of the 150 MHz FSR). The measured performance (based on careful analysis of this plot - counting pixels on a blown up version in MS Paint!) are as follows:

    The two values correspond to the widths of the left and right peaks on the plot. There may be a 5 percent uncertainty. The smaller peaks are most likely higher order transverse modes - evidently the length of the cavity is not exactly confocal. The reason the spacing of those higher modes is not constant relative to the two lowest order modes may be due to the cavity vibrating. This is one reason this 150 MHz cavity is no longer used. The newer design is both shorter and uses quartz that is twice as thick.

    (From: Sam.)

    Here, we have a resolvance described in terms of kHz when most other SFPIs can't even achieve the equivalent number of MHz! While I'm fighting to get a resolvance of a few MHz, this one does more than 100 times better. Of course, it might cost 10,000 times as much to build and definitely lacks something in the portability department! :-)

    So, I emailed John and asked about borrying it - after all he did say they don't use it there anymore! Within an hour, I received the tracking number! As they say, "be careful what you wish for....". This will now require real work. :) My intention being to see how well it performs with an HP/Agilent 5517B laser in air (I'm not going to put everything in a vacuum the way they do for their experiments).

    The test setup is shown in Ultra-High Resolution SFPI Test Setup:

    The SP-476 and DC power supplies for the laser can be seen behind everything else sitting on foam blocks for vibration isolation.

    I originally thought that the most difficult aspect will be initial alignment of the SFPI to the 5517 laser to get any detectable signal. A common SFPI has a finesse measured in the low hundreds with mirror reflectivities of around 99 percent. Enough laser light gets through the mirror (around 1 percent) and there is enough scatter from imperfect mirrors, that the position of the beam can be seen at both ends to provide a rough guide to alignment. With 99.98% reflectivity, it might be possible to see the position of the beam on the input mirror, but *nothing* worth writing home about will get through the second mirror unless the FP cavity is in resonance. In fact, nothing could be seen on the first mirror either.

    However, initial alignment turned out to be relatively easy by simply setting the X and Y position of both ends of the SFPI tube to approximately line up with the input beam. At the front, this is done trivially using the beam itself. For the back, the photodetector is centered on the beam and locked in place, and it then acts as the reference. The, a small amount of searching would produce recognizable blips.

    Originally, the SFPI quartz tube was almost entirely exposed using short pieces of head cylinder. Ahhh, a naked SFPI! :-) The result was that the room had to be completely dark or else the ambient light (and 120 Hz hum from the fluorescent lamps) overwhelmed the photodetector. Since it wasn't possible to see any scatter from the mirrors anyhow, fully enclosing the quartz tube inside a HeNe head cylinder didn't sacrifice anything, and this totally eliminated the problem.

    But the display was bouncing around due to vibrations. Both the DC power supplies for the Agilent laser and the SP-476 (visible in the photo, above) produce line frequency vibration. Since I don't have a vibration-damped optical table with overhead racks for equipment, I put both of the trouble-makers on soft foam blocks and this seemed to calm things down considerably.

    The next issue was that the display had lumps with multiple peaks. This was due to the distance between the mirrors not satisfying the confocal condition of the RoC of 50 cm. The input mirror is mounted on a large threaded cylinder. Some careful adjustment in increments of 360 degrees (since the mirror isn't quite perfectly centered or perpendicular to the optical axis) tuned the distance to be quite close. However, due to the backlash and slop in the machining of the threaded cylinder and the outer cylinder into which it mates, fine adjustments were, to put it mildly, hit or miss. Originally, I was simply moving it by hand but even after building a tool out of a pill bottle, the consistency wasn't much better. And with the high finesse, the confocal distance is extremely critical. I'm contemplating adding some means of apply permanent pressure to the threaded cylinder to both eliminate the free play and keep it from changing position.

    But then the mode spacing appeared to be around 10 MHz, not the 2.2 MHz expected from the laser. At first I thought: "Wow, that finesse isn't too bad...". The problem turned out to be back reflections into the laser forcing it to be tuned offset from where it is supposed to be. I'm not exactly sure how it comes up with a clean spacing of 10 MHz, unless it has something to do with aliasing of multiple longitudinal modes with the 150 MHz FSR. But by adding a pair of ND filters, the Zeeman-split mode spacing now is correct for the 5517B laser based on the ratio of the distance between the twin peaks and the FSR (2.3 MHz to 150 MHz).

    Another problem was that the (temporal) frequency response was not high enough for these narrow peaks to be displayed accurately when run at a convenient repetition rate with a span of a full FSR or more. At first I thought this was limited by the PD preamp in the SP-476 with a Thorlabs DET-110 photodiode. The rep rate*span product then needed to be reduced by a factor of 10 or more to get a decent display. Switching to a Thorlabs PDA55 amplified photodiode helped slightly, but then I realized that the high finesse - or equivalently, Q-factor - of the FP cavity could be even more significant by acting like a low pass filter. The response can be modeled as an RC filter with a time constant equal to the cold cavity decay time (Tcc), how long it takes for light inside the cavity to decay to 1/e of its original value with no outside input. Tcc = [1/(Mirror Transmission) * (Cavity Length)/c)] = [1/(1-0.99983) * 1.67 ns] = 9.8 µs. Then the 3 dB bandwidth is f3dB = 1/(2*π*Tcc) = ~16.24 kHz. Clearly, the scan rate needs to be greatly reduced to get decent resolution with a large span. Only if scanning in a narrow range (as in the plot above) can a relatively high scan rate produce an accurate display. Due to both vibrations and temperature variations conspiring to prevent a stable display, the only way to do this using the oscilloscope without some fancy locking scheme is to trigger on the actual peaks rather than the scan ramp or blanking pulse from the HV driver.

    With a fair amount of fiddling of the confocal spacing and alignment, the finesse is now somewhere around 4,000 (!!) based on the Matlab simulations, not quite up to what should be possible but not too shabby either. See Display and Simulation of Ultra-High Resolution SFPI. The multiple peaks in slightly different positions captured while the camera "shutter" was open are probably due to the way the scope was synced and a combination of PZT non-linearity and vibration. It's possible that the response is still limited by the SFPI's temporal bandwidth and slowing down the scan rate would still result in some improvement. The scope time-base is set at roughly 0.4 ms/div which means that the pulse width shown in the photo (about 0.15 divisions) is about 60 µs representing a temporal bandwidth within a factor of 2 of that of the SFPI/SP-476/PDA155 combination. And further testing shows that the resolution does improve slightly at a slower scan rate. There's something rather stramge about a simple physical device whose frequency response is being seriously limited by the speed of light! ;-)

    I doubt the performance now is quite up to what might be possible, but it's probably close. The three areas that might produce some improvement would be (1) setting the confocal distance more precisely, (2) better mode-matching of the input, and (3) putting the entire thing in a vacuum. None of these are easy (especially the last!) or likely to happen so at this point, it's probably as good as it's going to be! :)

    But one relatively simple way to stabilize the display would be to lock the peaks to the scan ramp. Then, a much narrower span and higher repetition rate could be used. This could be done with not much more than three monostables and an op-amp integrator. A retriggerable monostable would be clocked by the PD output to produce a pulse wider than the distance between the pair of peaks (PD Pulse). A second monostable would be triggered by the start of the scan ramp to produce a pulse to position the lock point at the desired location on the scope screen (Delay Pulse), and its trailing edge would then trigger a third monostable to produce a pulse with the same width as the PD Pulse, called the Scan Pulse. Then, if the PD Pulse preceeded the Scan Pulse, the integrator would be incrementally decreased, and if the Scan Pulse preceeded the PD pulse, it would be incrementally increased. The output of the integrator would be added into the PZT voltage. The logical OR of the Delay Pulse and Scan Pulse would act as a gate so that garbage during scan retrace would not confuse the locking scheme. This probably won't handle serious vibration but should compensate for environmental changes in temperature and pressure. Someday, maybe. :-)

    Commercial Scanning Fabry-Perot Interferometers and Drivers

    Most of the following are no longer in production. The only exceptions are some of those from Coherent, and apparently they are being phased out. However, if you're inclined to actually buy an SFPI new, several companies still do offer complete systems and components including Thorlabs and Toptika.

    Spectra-Physics 470 Scanning Fabry-Perot Interferometer

    The SP-470 Scanning Fabry-Perot Interformeter head along with the SP-476 controller provides similar capabilities to my $2 SFPI for only an additional $4,998. :) Actually, I don't know what the selling price was but these are typically $5,000 or more. The SP-470 came in several flavors depending on the wavelength range of the mirror set (450 to 550 nm or 550 to 650 nm) and Free Spectral Range (FSR, 2 GHz or 8 GHz with 20 MHz or 40 MHz resolution). The finesse is 200 for all versions. More info may be found under Vintage Lasers and Accessories Brochures and Manuals at the end of the section for Spectra-Physics in the "High Bandwidth Scanning Interferometer Brochure".

    One that I have is the SP-470-3, 550 to 650 nm with a 2 GHz FSR. This is absolutely ideal for all common visible HeNe lasers including the green HeNe at 543.5 nm. (There was no obvious reduction in resolution at 543.5 nm, though I didn't do any precise measurements. And, even for a 532 nm DPSS laser, the finesse was still at least 50.) When I first acquired this unit, the cavity length was all messed up so I had to set it for the confocal condition. This was done using a low power red (632.8 nm) HeNe laser which has only 2 or 3 longitudinal modes at most. After chasing my tail for quite awhile, I found the sweet spot. The adjustment is by turning the mirror cell at the detector end. Being recessed, a plastic "tool" was needed to get at it without fear of damaging the mirror. It's spring-loaded so should stay put, but there is no way to lock it in place. An external detector (Thorlabs DET-110 was set up beyond the end of the SFPI head, which was on a kinematic pan-tilt mount, and that was clamped down so it would not move relative to the HeNe laser. Once the confocal condition was achieved, it was relatively easy to jog the adjustment one way or the other to fine tune the cavity length. And then it really did work like the diagrams in textbooks, and almost as well as my $2 SFPI. :) OK, it is actually better in certain respects: The solid massive resonator virtually eliminates any drift due to the short term effect of temperature on the cavity length and also results in much reduced sensitivity to vibration.

    In fact, it appears as though the resolution may actually be much better than the 20 MHz listed in the specs.

    For a simple display of the modes of a HeNe laser, this high resolution really doesn't matter and may actually be a distraction. I need to find a laser that will do it justice!

    However, the 2 GHz FSR is too small to display the modes of a Zeeman-split HeNe laser without aliasing. When an axial magnetic field is applied to a HeNe laser tube, the neon gain curve splits into two similar curves, one shifted up and the other shifted down by several hundred MHz. Thus, the effective split gain curve can be much wider than the nominal 1.6 GHz or so of the normal HeNe laser. Only when the alignment between the SFPI and laser is essentially perfect and the FSR doubles is the display unambiguous.

    The destabilizing effect of any back-reflections from the SFPI into the laser is also very evident as random noise superimposed on the mode display. So, either an optical isolator must be used, or the beam aimed at an angle so the none of the reflection of the beam enters the laser aperture.

    The SP-470 allows for interchangeable mirrors and detector, unlike the SP-450 where everything is fixed. There were also optional lenses, polarizers, and apertures that can be screwed into the front of these SFPI heads. The lenses can improve the resolvance under some conditions, but not always. The aperture does generally help and also makes alignment even easier, at the expense of signal level.

    As a practical matter, swapping mirrors still requires precise adjustment of their spacing, so it's not quite as easy as described in the SP brochure. So, it's best to have a separate SFPI head for each wavelength range.

    I've also tested a 470-02, 450 to 550 nm, 8 GHz FSR, ideal for most of the common wavelengths of the argon ion laser as well a the 532 nm DPSS laser.

    The sensor is just a photodiode with a lens to focus the center 3 mm or so of the transmitted beam onto the detector. For a narrow beam like that from a HeNe laser, almost any photodiode will work. However, performance does seem to be better with the lensed photodiode compared to a cheap non-lensed one for large diameter beams, perhaps simply because the capture area of the lensed PD is smaller. I'm not sure exactly what else, if anything, is inside the detector housing beside the PD socket. What I believe to be a genuine SP detector measures 1.5K ohms across the PD socket, but no other components are shown in the SP-476 schematic. And, the 1.5K across the PD results in power line frequency hum, which is particularly annoying on the more sensitive ranges. This is apparently due to ground loops inside the SP-476 as rearranging the wiring inside the supposedly shielded PD preamp section affects it. My home-built detectors using cheap photodiodes mounted inside filed-down 1/2 inch PCV pipe couplings work fine without the resistor and do not have noticeable hum. (Where the resistor is present, the amplitude of the hum can be reduced by about 75 percent by removing the black ground wire between the PD Input BNC connector and the PD preamp PCB, and rearranging some of the cables. But I wasn't able to get it to go away. I don't believe there is any fault in this SP-476. It's just not built using good analog design practices. Perhaps there's an ECO for later versions. It might also be possible to remove the resistor inside the detector housing, though that appears difficult.)

    Spectra-Physics 450 Scanning Fabry-Perot Interferometer

    The SP-450 Scanning Fabry-Perot Interferometer head along with a suitable controller has similar performance to the SP-470 but is in a smaller package with a single permanently attached cable for PZT drive and photodiode output. It terminates in a strange plug that doesn't mate with the SP-476. There is no mention of that in the SP documentation that I've seen but an adapter cable to BNCs apparently exists. I simply cut off the plug and installed a pair of well marked coaxes for PZT and PD. The pins are as follows:

       Pin   Wire Color   Function
        A      Black      PD Anode
        B       Red       PD Cathode
        C      Bare       Shield/Ground
        D      White      PZT HV+
        E      Green      PZT HV-

    PD is simply a silicon photodiode with no other circuitry. Normally, PD Cathode would be grounded with PD Anode being the signal input. The polarity for HV has frequency increasing to the right on the display (PZT shrinks with voltage). Reverse to have wavelength increasing to the right.

    Like the SP-470, the SP-450 is also a confocal design but with a fixed approximately 1 mm hole on the input side. But this makes it very easy to use. Just about any beam that enters the hole roughly on-axis will result in a decent display with at most fine tuning of alignment needed for it to be perfect.

    See Components of SP-450 Scanning Fabry-Perot Interferometer Head. The double convex lens that was glued to the front of the PZT/optics assembly can be seen to the left of it. The photodiode is mounted within a Nylon, just visible behind the white and green wires.

    Access to the internal components is remarkably straightforward. After removing the rear cover and cable (3 itty-bitty set-screws), use a thin tool to carefully pry out the large red O-ring. The entire PZT/optics assembly can then be pulled free. It's only anchored with a similar large red O-ring at the front. There may be a focusing lens glued to the front optics holder. The PZT HV+ (white) wire is fastened with a set-screw at the front whose head is possibly blocked by the focusing lens. It's real easy for this wire to break off since the set-screw tends to mangle the strands. I installed a larger solid wire so the set-screw could be tightened more securely and then soldered to that, with heatshrink for protection over the splice. The PZT HV- (green) wire is secured with a screw into the rear of the PZT/optics assembly.

    The cavity should never need adjustment but if someone before you decided the outer ring at the rear was loose and screwed it down tight, I believe that's what does it. :) (The inner ring secures the mirror.) And while the description of the SP-450 didin't offer the option of swapping mirror sets, it really should be possible.

    Spectra-Physics 476 Scanning Fabry-Perot Interferometer Driver

    While a low speed function generator with a maximum output of 20 to 40 V p-p will work with my PZT beeper-based SFPIs, most commercial instruments like the SP-470 described above require 100 V or more to provide enough sweep span. The SP-476 Scanning Fabry-Perot Interferometer Driver is basically a dedicated high voltage ramp generator designed specifically for this purpose. It produces a variable frequency sawtooth ramp with selectable and adjustable amplitude and offset. The maximum output is switch selectable between 300 and 1,000 V, which really specifies the approximate maximum range including the centering or offset voltage, the actual p-p output of the sawtooth is somewhat less than these values). A photodiode preamp with 5 gain ranges, as well as a temperature controller (for SFPI setups with thermal control capability) are also included.

    Note that the SP-481 and SP-481A Dye Laser Etalon Controllers have most of the functions of an SP-476 and some additional ones including a set of slow speed ranges and a front panel temperature control. (I don't know what the difference is between the SP-481 and SP-481A.) They have a bunch more stuff as well for the dye laser etalon control application including a separate HV output for the cavity. (Either the etalon or cavity HV output may be used for the sweep, but not both at the same time.) None of this interferes with its use for an SFPI. The only relevant difference seems to be that the SP-481/A lacks a selection for 300 V or 1,000 V maximum output - it's always 1,000 V. But since there is a knob to adjust the amplitude from 0 to maximum anyhow, that's no great loss as long as care is taken not to exceed the voltage rating of your PZT.

    Both the SP-476 and SP-481/A show up on eBay quite frequently, generally going for less than $200. And they provide more features than most other SFPI drivers including adjustable outputs for blanking/scope trigger and (scaled) ramp, a heater controller, and the ability to be used as a high voltage amplifier. Although they are quite old, except for the power transformer which is custom (and identical for both models), the circuitry isn't very complex and uses common readily available components should repair be needed:

    However, one nice thing about the SP-476 (and presumably the SP-481/A as well) is that since the high voltage driver is essentially a voltage controlled shunt regulator, even a continuous short circuit of the output to ground will do no harm, apparently unlike in many other designs! :)

    A description and photo of the SP-476 can be found under Vintage Lasers and Accessories Brochures and Manuals at the end of the section for Spectra-Physics on page 5 of the "High Bandwidth Scanning Interferometer Brochure".

    Spectra-Physics 422 Scanning Fabry-Perot Interferometer Head with SP-430 Driver

    This may be one of the earliest Spectra-Physics SFPIs available. The system I tested is SN: 102, which is probably the 2nd one made (with SP SNs starting a 101). The 1st is probably in the Spectra-Physics museum. ;-) See: Spectra-Physics 422 SFPI Head with 430 Driver. The SP-422 SFPI head appears to be plane-plane with an FSR of around 8 GHz. It requires super-precise alignment using a pinhole-size entrance aperture. With optimum alignment, reflections go directly back to the laser typically causing it to become unhappy. The SP-430 driver is bare-bones with knobs for Scan Time, Dispersion, and Center Frequency. The outputs are Ramp Monitor and Signal (which is just connected to the raw photodiode in the SP-422 head). Auto/Manual selects the free-running ramp or one controlled by the Center Frequency knob. The label on the SP-422 head lists only 500 nm but these typically cover a 100 nm range, so probably good - or at least usable - from 450-550 nm (blue-green). The model number for the 550-650 nm (green-red) version is not known.

    Unfortunately, the diameter of the head is not compatible with any mounts available. So for testing, it was jammed in a V-block of sorts with foam and carefully adjusted for optimal alignment. With the narrow beam from a 544 nm HeNe laser, the finesse was at best between 50 and 100. With a proper XYPT mount, it might be slightly better but not dramatically so.

    The ramp on this one (both the high voltage to the SFPI head and Ramp Monitor) is not linear with the shape of a simple RC charging circuit. It is not known whether this is a feature or bug. The system may have been intended to be used with an X-Y display on a scope of the era, in which case the non-linearity would be less important. But it is probably broken and I have no intention of fixing it or even tracing the circuit. See: Interior of Spectra-Physics 430 SFPI Driver. I did check the single large ancient metal can electrolytic capacitor and that appears to be fine for both value and ESR. The active devices consist of 3 transistors and JFET. None are shorted. That's as far as I intend to go!

    This is at best a collectible now - or an SFPI of last resort. I wonder if there was ever a SN: 103. ;-)

    Burleigh/EXFO SA Plus Scanning Fabry-Perot Interferometer

    Burleigh, since taken over by EXFO, manufactured a variety of optical instruments including wavemeters and interferometers. (Though as of 2010, they have discontinued this product line.)

    The SA Plus is a system similar to the Spectra-Physics 476 controller with 450 or 470 SFPI head, but of more modern construction and some nice features compared to those vintage instruments. A brochure/spec sheet can be found at Burleigh/EXFO SA Plus Laser Spectrum Analyzer Brochure. The same head is used for all versions but the FSR can be either 2 GHz or 8 Ghz, determined by the mirrors and mirror holders. Mirror sets are available covering wavelengths from 550 nm to 1,800 nm, with a finesse of either 200 or 300 depending on the wavelength range, higher for IR:

       ID       Wavelength     Comments
       07     550 -   650 nm   Green - Red
       08     650 -   750 nm   Red - Near-IR
       09     750 -   890 nm   Near-IR
       10     850 -   990 nm    "   "
       11     980 - 1,145 nm    "   "
       12   1,150 - 1,345 nm    "   "
       13   1,300 - 1,550 nm    "   "
       14   1,425 - 1,675 nm    "   "
       15   1,550 - 1,800 nm    "   "

    The ID numbers correspond to the suffixes used when ordering. The lack of any standard mirrors below 550 nm is interesting, despite the obvious lack of IDs 01 to 06. :) (The operation manual for the earlier version of the Burleigh confocal SFPI lists mirrors for 450 to 550 nm, 550 to 650 nm, 760 to 850 nm, or 1,010 to 1,110 nm.)

    The standard mount for the SA Plus SFPI head has both pan and tilt, and X-Y adjustments. The X-Y greatly simplifies setup as the laser then doesn't need to be on an adjustable platform. The only problem is that the mount is quite HUGE!

    The cavity length can be easily fine tuned and then locked in position without going inside, unlike the SP-470 which requires removing the photodiode and using a tool to turn the mirror holder, then replacing the PD and checking if the adjustment helped or hurt. Or, mount the PD externally and angle a tool inside to turn the holder without scratching the mirror. There's also no way to lock the mirrors in the SP-470 against vibrations messing up the distance. (The SP-450 is adjusted at the factory and cannot be changed.)

    The ramp generator doesn't have as many bells and whistles as the SP-476 but it is adequate and also adds a nifty little bar-graph display to show approximately the amplitude location of the ramp and relative to the maximum voltage swing available.

    The photodiode preamp is a separate little box powered by a pair of 9 V batteries. The PD itself slips into the back of the SFPI head and is held in place by a pair of magnets.

    The SA Plus I tested has the mirror set for 550 to 650 nm. At least I think it does since there were no markings on the mirror holders and I wasn't about to remove the actual mirror glass to check them. The mirrors had that silvery broad-band appearance in reflection and deep purple in transmission.

    It works well at 633 nm and 543.5 nm, and probably even at 532 nm though I didn't do a complete test there. I believe it will also be useful well beyond 650 nm. Setup is very easy once I mounted the SFPI head assembly on a Newport post holder screwed to a wooden plank. :) It took under a minute from a totally misadjusted condition to find and fine tune the cavity length for optimal confocal response. (I, of course, had disassembled it to see what was inside and try to determine the part numbers on the mirrors!)

    Although the controller only has a single-turn pot for centering, that seems to be enough. The photodiode preamp saturates at around 1 mW in each mode, but that could be partially because the batteries were somewhat weak at 7.5 V instead of 9 V. :) An ND filter takes care of that and also helps to reduce backreflections to the laser.

    I also have the head (only) from what is probably an earlier version - perhaps the "SA" without the "Plus". :) The manufacturer's sticker had been removed so I do not know the precise model. It has a holder with a threaded lock-ring for the back mirror and a plate with the mirror glued into it for the front mirror. But it is otherwise similar with a large locking ring and fits a standard 2 inch mount.

    The mirrors in mine are for an IR wavelength range (as yet to be determined), but it was fairly easy to replace them with 633 nm 1.7 GHz FSR high finesse mirrors. The back mirror dropped right in being the same size as the Burleigh mirror. But rather than attempting to remove the Burleigh mirror from it's glue and possibly ruining it, a quick and dirty adapter plate was fashioned out of the end-cap from a dead 3/4 inch diameter HeNe laser tube. This also enabled the cavity to be lengthened by approximately 0.2 inches to accommodate the difference in FSR (1.7 versus 2.0 GHz for the Burleigh mirrors). The 633 nm mirrors worked quite well. It was easy to locate the optimal spacing and lock it in place and the achievable finesse was quite decent, tested using an Agilent 5517 laser with a ~2 mm beam. However, the "sweet spot" for alignment was quite small possibly indicating that the two mirrors are not quite aligned with their optical axes coincident. So, I will probably have a proper adapter plate machined eventually, and at the very least, it will look a lot spiffier compared to the re-purposed HeNe laser tube end-cap!

    Coherent/Tropel 240 Scanning Fabry-Perot Interferometers

    These are another line of Scanning Fabry-Perot Interferometers (SFPIs) similar to the instruments from Spectra-Physics and Burleigh, although Coherent calls it a "Laser Spectrum Analyzer". These were originally developed by Tropel, which then became part of Coherent (for this as well as many other products). But unlike the other SFPIs I've come across, this ones from Coherent were a current product (as of 2010) but have now (2014) been "phased out of the product portfolio" as Coherent puts it. I suppose that does sound better than "terminated" :( :). There were four versions with FSRs of 300 MHz, 1.5 GHz, 7.5 GHz, and 30 GHz. I suppose HeNe lasers are no longer of high priority to be analyzed as none of these FSRs is really optimal for a 1.6 GHz gain bandwidth! However, even the original Tropel versions had the same FSRs.

    An old Coherent 240 manual only listed three wavelength ranges - and in Angstroms (A) which really dates it, but this may simply have been edited to change the name from Tropel:

       ID      Wavelength      Comments
        1    4,500 -  5,500 A   Blue - Green
        2    6,000 -  7,000 A   Yellow/Orange - Deep Red
        3   10,500 - 11,500 A   Near-IR

    Except for the 300 MHz FSR head, they all use the same body (spacer tube and lens assembly, part number 33-2492) so only the mirror sets differ. The one with a 300 MHz FSR has a body that looks more like a telescope or flashlight being about 14.5 inches long (part number 33-2502) that expands at the front. Eventually under the Coherent name, there were 14 standard mirror sets available covering wavelengths from 337 nm to 1,625 nm (except for the 30 GHz FSR version which lacks the 5 shortest wavelength rangess.) But there are some gaps in the IR wavelength coverage. (This may be more a matter of specifications than anything else as it's likely that mirrors on either side would still have decent performance in the gaps. They are listed below.) The finesse is spec'd at 200 for all except the 30 GHz FSR, for which it is only 100. If you'd like to order one from Coherent and have a working time machine the price in 2010 was around $6,500 with the ramp driver. :-) As noted, Coherent has now phased out the entire product line. Go to Coherent and search for "laser spectrum analyzer" and you'll be able to confirm the bad news. I have saved the glossy product brochure at Coherent Laser Spectrum Analyser System. Since this is an obsolete product, I'm hoping Coherent won't mind. :) (The 240 model designation is historical. Coherent used fancier harder to remember model numbers.)

    One example is the 33-6438-001 with the standard Coherent gimbal mount as shown in Typical Coherent Scanning Fabry-Perot Interferometer Head and Mount. This is representative of all units except the stretch version with the 300 MHz FSR.

    The short SFPI heads have an adjustable focusing lens in front which enables the focal point to be optimally positioned in the center of the cavity. The photodiode on a bayonet mount (looks like an oversized BNC) which makes it more secure with alignment that is more precise and consistent. The mount that comes with the SFPI head uses a gimbal design which means that the head pivots about a common point near the center of the mount. (A kinematic mount pivots about a point near one corner. As a practical matter, this doens't really matter very much for an SFPI.) The mount has pan and tilt adjustments and is on a post that slides into a massive base, adjustable in height. However, unlike the Burleigh SA mount, there is no side-to-side adjustment. Fine X and Y adjustment would be highly desirable, especially for the more finicky heads with the 30 GHz FSR (more below). The 300 MHz version fits the same mount with an adapter ring (since it's narrower over most of its length), but it's not at all clear if pan and tilt with its axis near the center of the cavity is really optimal for ease of adjustment with this long head. The ramp driver has a maximum output of 250 V which is sufficient for viewing 2 FSRs, but not with a lot of margin. It appears generally similar to the one for the Burleigh SA, but has the PD preamp built in.

    The first Coherent SFPI head I acquired had the nice mount, but no driver, so I'am using it with an SP-476 driver and that works fine. However, since Coherent specs the maximum voltage to be 500 V, the 1,000 V setting of the SP-476 should not be used; the 300 V setting is enough for much more than 1 FSR. This SFPI head has an FSR of 30 GHz with a wavelength range of 550 to 650 nm. That might be useful for some types of diode lasers but is certainly far from ideal even for short HeNe lasers with a large mode spacing.

    With the mirrors having a Radius of Curvature (RoC) of only about 2.5 mm (1/10th inch) spaced an equal distance apart, alignment to the laser even with the confocal cavity becomes much more tricky. The person who sent it to me couldn't get any response at all from the photodiode, and at first, I had the same problem. But finally, after very careful alignment, it started to behave more like the other confocal SFPIs I've tested, except that the finesse is poor. With a spec'dfinesse of only 100, the resolvance under ideal conditions (adjustment, alignment, and beam focusing) is only about 300 Mhz. So far, I've only been able to achieve a finesse of around 50 (resolvance of 600 MHz) which can barely resolve the longitudinal modes of a Melles Griot 05-LHR-911 HeNe laser (mode spacing of 883 MHz). I have not yet found the cause. It is very likely at least in part due to the beam diameter being too large, as well as the mirror spacing not being precise enough, which seems quite likely since the previous owner may have attempted to adjust it after not being able to obtain a signal. And it is very fincky! But there could conceivably be damage to the mirrors as a result of abuse it may have been subjected to in its previous life. The specs may simply be overly optimistic for real World conditions.

    Disassembly for adjustment or removal of the mirrors is supposed to require special optical spanner wrenches so it's more difficult to do anything inside (or mess it up!) than with the Spectra-Physics or Burleigh SFPI heads. But mirror sets are intended to be replaceable by the user, so it shouldn't be that bad. :) Since my Coherent tool set hasn't arrived, at first I was using an improvised pair of filed-down needle-nose pliers, but this required removing the photodiode assembly and there wasn't enough clearance to mount it externally since it would have to be relatively close due to the small RoC of the mirrors and divergence of the transmitted beam. So, observing the display in real-time was not practical. But this enabled the rear mirror cell to be removed for inspection. Both mirrors appear fine. It's amazing how small they are - a 2.5 mm RoC and nearly a complete concave hemisphere. How do they even grind the substrate for a mirror like that? Finally, threatening the SFPI with a dental pick did the trick. Angling the dental pick into one of the holes in the adjustment ring with an external photodiode permitted very fine movement while observing the SFPI display. It is still VERY finicky. Even with careful adjustment of cavity length - which is still a pain - and perhaps a bit of fudging of the data, the finesse is close to the spec'd value of 100 based on the FWHM, though the lower portion of each peak is more spread out than would be expected based on the textbook plots, and there are still some artifacts in the display. Then again, perhaps all those assumptions made in calculating finesse simply don't work very well with a finite beam size and such a small mirror RoC! Something about serious spherical abberation conspiring to mess it up. Perhaps, the beam really has to be a smaller diameter or better collimated or something. And various other things conspire to make this more difficult than it could have been with better design and more precise machining. The problem with the adjustment is that the required setting must be accurate to well under +/-1 degree of rotation, and even a slight change affects other aspects of alignment due to tolerances in the machining, fit, and polish. :) For example, the two very tiny mirrors are held inside long tubes that protrude from mounts on screw threads at opposite ends of the spacing tube. Expecting their focii to line up within a small fraction of 1 mm is asking a lot. And, there is no backing spring - apparently the special sticky grease is supposed to keep everything in line. So, it's not clear how stable this is over time - just a lot of fine thread area and grease. Too bad they didin't include at least a set-screw for locking.

    I don't really have a good use for a 30 GHz, 550 to 650 nm SFPI, though it's quite possible one could turn up. However, the Coherent model I would really have liked would be their part number 33-6305: 300 MHz FSR, 550 to 650 nm. Then I wouldn't have had to build my own SFPI to resolve the split line of two-frequency HeNe lasers from HP/Agilent and Excel. However, Coherent may not have offered this particular combination anymore since there was no link to it on their Web site (even as of 2010). But I would have settled for a used one. Perhaps, Coherent might have accepted a trade. :) (As it turned out, I did end up building my own with an FSR of 250 MHz and finesse exceeding 150. See the section: Sam's High Resolution Scanning Fabry-Perot Interferometer.)

    Now here's the peculiar effect of the week: I wanted to put a neutral density filter in front of the laser to reduce the maximum possible intensity of any back-reflections. This is desirable to minimize the chance of laser instabiliity that can result from mutually coherent light re-entering its cavity. The transmitted power would be the input power times the filter's transmission coefficient, so the amplitude of the display would be reduced, but there is plenty of gain! However, any back-reflections would be reduced by the square of the transmission coefficient, even if there was 100 percent reflection. For example, with a filter coefficient of 0.25, the transmitted power would be 25 percent of the laser power, and the maximum reflected power would be no more than 6.25 percent. I tried several filters that are basically amber-colored glass. Three out of four behaved as expected: The optical power reaching the SFPI had the expected value and amplitude of the display was reduced. Some adjustment of the SFPI alignment was required to optimize the display if the glass plate was at an angle, but the resulting amplitude was very close to what would be expected based on the transmission coefficient of the filter. However, the forth piece of glass behaved, well, strangely. While it reduced the power as measured by eye and with a laser power meter by the expected amount (75 percent), the SPFI display disappeared entirely regardless of the orientation of the filter or adjustment of the SPFI alignment for an amplitude of exactly zero or 0.0000000 on the 'scope even with the gain controls set at maximum! Not even any tiny bumps. In addition, under normal conditions when optimally aligned, the back-reflection from this SFPI is a large more or less uniform disk of light. That was totally absent as well.

    And to add to the strangeness, there was no similar effect using the same laser and filter with a Spectra-Physics 8 GHz SFPI.

    It turned out that the cause was very simple: The one problematic filter has a rather large wedge - probably 1 or 2 degrees. The others have little or no wedge. It's not enough to detectably divert the beam but must be tilting the wavefront so the required interference cancels out. And in fact with the wedge producing a deflection horizontally, careful realignment including repositioning the SFPI head horizontally finally was able to restore the display to normal. The extremely small cavity of the 30 GHz SPFI is much more finicky about (laser) alignment in general so it must be much more sensitive to the wavefront as well. Interesting..... :)

    As noted, these started out as Tropel so not surprisingly, the construction of the Coherent versions is almost identical. While they look very similar and the heads do fit the same mount, there are some subtle differences. At the very least, the thread diameter for mounting the photodiode assembly differs by just enough on some versions that they are not interchangeable. Go figure. :) I acquired a Tropel 240 with the mirror set for green/blue lasers, which is probably equivalent to Coherent part number 33-6206 which has an FSR of 1.5 GHz. And of course the first thing I did was to attempt to confirm that the confocal mirror spacing was optimal. It probably was and so this became a BIG mistake. The special Tropel grease had congealed over the eons so while the mirror barrel initially seemed to be movable, it jammed solid and required total disassembly to be able to apply enough torque to free it - and that was almost impossible. I nearly broke the official Coherent mirror spacing adjustment tool in the process and had to use filed down snap-ring pliers to rotate the mirror barrel with the threaded metal flange of the inner cylinder clamped in a (cushioned) vice! However, there were two benefits to this: Primarily, it allowed the internal construction to be documented. :) See Tropel Model 240 Scanning Fabry Perot Interferometer Head Components. The only parts not shown are the 4 screws that secure the BNC connector. Disassembly (and reassembly) is straightforward except that the BNC connector needs to be unsoldered from short stubs of the red and black wires before the inner assembly can be removed. That's not so bad. Going the other way is the pain. :( :) The Coherent photodetector assembly is shown; I do not have one from Tropel but assume they looks similar even if the thread size differs by enough to be annoying. And now the mirror spacing adjustment is smooth as silk. I didn't have the special Tropel or Coherent grease, so I used the tiniest bit of high vacuum grease. It may not stay put quite as well but I can add a dab of 5 Minute Epoxy (removable should the need arise) to prevent any drift.

    Another short Coherent head I acquired had 30 GHz FSR mirrors for 900 to 1,070 nm. This would be useful for Nd:YAG and other similar lasers, but they are kind of boring. :) So, it occurred to me that with small adapter rings, the mirrors I provide in my SFPI kit for 633 nm could be made to fit the Coherent SFPI head. The mirrors would need to poke slightly beyond the normal mounting surface since their 43 mm RoC is 7 mm shorter than the 1.5 GHz FSR of the SFPI body. (The 7.5 GHz and 30 GHz Coherent/Tropel mirrors do something similar.) At first, I was going to cobble something together but in the end decided to have them professionally machined. The mirrors are glued to the adapter rings and then they are easily installed like the Coherent/Tropel mirrors. And their performance is actually quite phenomenal. For a 6 mm diameter beam from a Zygo 7701 laser, the finesse exceeds 350; for a 3 mm beam, it is over 550, and probably slightly higher for a smaller beam! (Theory predicts it can exceed 600 based on mesaured mirror reflectances.)

    The performance at 633 nm is considerably better than that of the standard Coherent/Tropel mirrors. The two-frequencies 20 MHz apart are resolved as nearly independent peaks. See Scanning Fabry-Perot Interferometer Display and Simulation of Zygo 7701 Laser Spectrum. The photo on the left is the unretouched screen shot with a span of about 2 FSRs. The photo in the middle is one of the twin peaks expanded by a factor of 10 on the scope. The plot on the right is a Matlab simulation of a Zygo spectrum with an SFPI finesse of 550.

    Tests using a REO tunable HeNe laser show a finesse of between 400 and 450 at 604/612 nm and between 250 and 300 at 594 nm. I need to test at longer wavelengths but don't have a convenient tunable laser for that. If the reflectance function is symmetric, the finesse at 670 nm should be in the 250 to 300 range. However, since the coatings of HeNe laser mirrors are often designed with a reflectivity fumction that is a "cliff" with respect to wavelength, it's possible that the useful range above 633 nm may not extend that far.

    I acquired a Coherent model 216 - the nice long one with an FSR of 300 MHz. (The "216" designation is a carryover from Tropel and seems to have zero correlation with the more recent Coherent part numbers.) Construction is generally similar to the shorter Coherent (and Tropel 240) SFPIs with mirror mounts glued to a PZT spacer tube. For the long SFPI, the PZT is the same length, but a glass cylinder extends it the required distance. I say "glass" but it may be Zerodur or something like that. The SFPI cavity is sandwiched between the front and back end-plates, with only a rubber O-ring as cushioning at the front (to allow for the PZT to do its thing). The end-plates have aluminum-on-aluminum threads, which meant that I once again had to fight with metal-lock. But this did allow me to inspect the inside. Someone may have repaired this unit at some point in the past as the glue jobs on the PZT look suspect and the rear mirror mount actually came off the glass tube, but some 5 minute Epoxy solved that. :)

    Unfortunately, the wavelength range of the mirror set was not labeled. They have a slightly bluish tint in reflection and a slightly yellowish appearance in transmission.

    Here is a list of all the Coherent standard wavelength ranges:

       ID       Wavelength     Comments
        1     275 -   305 nm   UV
        2     305 -   337 nm   ""
        3     307 -   365 nm   Near-UV
        4     365 -   405 nm   Near-UV - Violet
        5     405 -   450 nm   Violet - Blue
        6     450 -   550 nm   Blue - Green
        7     550 -   650 nm   Green - Red
        8     650 -   775 nm   Red - Deep Red
        9     690 -   830 nm   Deep Red - Near-IR
       10     790 -   930 nm   Near-IR
       11     900 - 1,070 nm    "   "
       12   1,000 - 1,100 nm    "   "
       13   1,250 - 1,450 nm    "   "
       14   1,400 - 1,625 nm    "   "

    (The IDs are my arbitrary designation. All of these were available for the 1.5 GHz, 7.5 GHz, and 300 MHz FSR heads. The 30 GHz FSR version lacked the UV/violet options, IDs 1 through 5. More details can be found in the product brochure at Coherent Laser Spectrum Analyser System.) IDs 6 and 7 were ruled out both by their appearance and by testing with HeNe red and DPSS green lasers. The yellowish tint in transmission means that the wavelength for peak reflection must be in the deep blue or 3 times that wavelength in the near-IR.

    Having initially concluded (perhaps incorrectly) that the original mirrors weren't likely to be that useful, even once their wavelength range could be determined, I decided to installed mirrors of my own, at least as a test. Fortunately, swapping mirrors in all of the Tropel/Coherent SFPI heads is relatively easy and low risk, each being held in by a threaded ring with a rubber O-ring for cushioning. The required RoC is 25 cm and the mounts have a diameter of 12 mm diameter. I did have some HeNe mirrors that I thought had an RoC of 25 cm, and they were the right size so I popped them in. But it turned out that their RoC was actually 30 cm, so the results were, well, a bit strange. They did work but with optimal adjustment of the SFPI cavity, the display had an effective FSR of 60 MHz (1/5th of the expected 300 MHz), but with a finesse referenced to 300 MHz. And the reflectance of these mirrors is about 98.5 percent - somewhat lower than optimal. Thus, the resolvance was not very good. But nonetheless, it could be seen that a Zeeman-split HP-5517C laser produced a pair of wavelengths, though their separation could not be measured. (This was partly due to not attempting to optimally mode match the widely diverging beam from an HP laser without the normal collimator to the SFPI. However, this did prove that the SPFI was operational. So the finesse wasn't actually too terrible and in fact, such a system could still be useful to confirm that a laser is single longitudinal mode (SLM). Although the multiple peaks weren't as consistent in amplitude as they would be with the proper mirrors, non-SLM behavior would still be apparent. But this wasn't really what I Wanted. One option would be to make adapters so that standard 6 mm HeNe mirrors can be dropped in, of which I probably have several suitable candidates with a somewhat higher reflectaace, but perhaps not enough of an improvement to warrent the effort.

    Later, I did some more testing of the original mirrors. They passed 405 nm from a diode laser like a sieve so that eliminated IDs 4 and 5 (above). Testing with a 1,064 nm laser tossed ID 11 and 12. Based on the appearance, the most likely mirrors would be IDs 13 or 14. Finally, I dug up a Lightwave Electronics model 120-02 laser that operates at 1,319 nm. Bingo! The mirrors appear to have very low tranmission and high reflectance at 1,319 nm. Using an IR detector card, the transmission was essentially zero. The next step was to reinstall these mirrors and test the SPFI in its original configuration, except using one of my cut open germanium transistor photodetectors (since the original sensor did not come with the 216).

    I expected it to be fairly easy to get going at 1,319 nm based on my experience with this SFPI at 633 nm and my home-built high resolution SFPI with a slightly longer cavity length. However, alignment turned out to be a pain in the you-know-what. This was partially due to my desire to do the initial test before I had a proper adjustable mount for the SFPI (or the laser) and the cavity spacing of the SFPI was almost certainly not optimal for the new mirror set, but also for other reasons that will become obvious.

    The laser I used for testing was the LWE-120-02 (about 4 mW) with a single frequency output at 1,319 nm (or at least that what the specs say). I started with a Newport IR detector intended for a power meter like the 835. I know this to be reliable but with a mediocre frequency response. I could always reduce the scan rate to check for display quality. And once any display was achieved, I would switch to one of my cut-off germanium transistors which have a nice small detector area, low capacitance, and thus decent frequency response. But it took quite awhile to obtain anything resembling an SFPI trace. I was even at the point of suspecting the installation of the mirrors and checked both to be sure they were in the correct way around and not cocked in their mounts. They were fine. Occasionally, there would be some very low level blips that correlated with the SFPI scan but it was very difficult to maintain these. Finally with enough duct tape and modeling clay (just kidding), a trace appeared that was stable for long enough to be able to approximately peak the SFPI cavity length. Even with the Newport detector, it became obvious that this was no ordinary mirror set. And with a stable setup, switching to the better cut-off transistor detector was a snap.

    Not only does this thing work well at 1,319 nm, but the finesse of at least 500 and possibly approaching 1,000 even without worrying about the laser's beam size or divergence. My conclusion is that these are probably not a standard mirror set. And recall that the finesse of a confocal cavity SFPI is cut in half, so the mirrors must be even better. With the scope set so that one FSR spans the complete screen, a peak is essentially a thin vertical line, much narrower than what's typical with the off-the-shelf mirrors or found in the brochure (above). Only when expanded using the SPFI controller or 10X sweep on the scope can the width be resolved. To achieve an effective (confocal) finesse of 500 requires mirrors with a reflectance of 99.7%. Since the finesse here could be 1,000 (or more), the reflectance may be much higher. Now I know why it was more of a pain to get going at 1,319 nm then at 633 nm with my 98.5% mirrors. These are closer to HRs. And of course working with a low power laser at 1,319 nm didn't help.

    Not surprisingly, the LWE-120 must not like back-reflections. There is a second mode that comes and goes but is more likely with the SFPI peaked for maximum resolution. Misalign the SFPI sufficiently and the second mode goes away entirely. On the display it appears ~50 MHz away, but who knows where it actually is due to the aliasing of the SFPI's FSR. Because of the way everything is set up now, I can't really move the laser any distance away - it's nose-to-nose with the SFPI - so back-reflections are inevitable. And I don't have an optical isolator for 1,319 nm.

    The standard mount that is supposed to also be used with this SFPI (that looks more like a HeNe laser head cylinder) seems to be far from ideal. And although I have one for the other Coherent SFPI head, I don't have the required adapter ring as the barrel diameter is smaller than for the short SFPIs. So, I may build a platform mount with three adjustable feet for it.

    In 2014, I finally - FINALLY - was able to test a Coherent 216 with the 550 to 650 nm mirrors. :) As luck would have it, I had to repair it first - the front mirror mount assembly had broken off of the spacer tube. But it was straightforward to glue it back in place. :-) Alignment using an Agilent 1211A two-frequency laser was easy in so far as obtaining a display, but similarly fiddly as my home-built high resolution SFPI to optimize it. Even with the N1211A's narrow beam, a pinhole was needed to maximize resolution. And interestingly, any back-reflections seem to shift the laser's lock point slightly so the amplitude of the two peaks changes. The pinhole reduced that effect as well. (An optical isolator could have been used to prevent back-reflections, with the beam first passed through a polarizer at 45 degrees to include both frequency components and be linearly polarized.) I should have taken a photo but neglected to do so, and I'm not setting it up again simply for that purpose! So, a simulation will have to do. :) See Simulation of Coherent 216 SFPI Display of Agilent N1211A Modes. The difference frequency is around 1.6 MHz and the actual display had at least as much resolution as the simulation using a finesse of 400.

    More to come.

    TecOptics SA Scanning Fabry-Perot Interferometer

    TecOptics is currently a manufacturer of custom optics including (fixed) Fabry-Perot etalons. However, they used to offer a variety of standard products including the FPI-25 general purpose plane mirror SFPI with adjustable FSR, as well as confocal cavity SFPI systems similar to those from Burleigh, Coherent, Spectra-Physics, and others.

    The SA series of confocal SFPIs consists of an SFPI body which includes a fixed lens at the front and fine-thread adjustment ring at the back into which the Si or Ge photodiode module is installed. The adjustment ring is always accessible to enable the mirror spacing to be fine tuned, and may be locked in place so the setting shouldn't change. The mirror sets are mounted on plates mounted via two screws to the PZT assembly inside.

    The HV connection is a miniature coax and the photodiode uses a 3-pin DIN so they can't be accidentally swapped, at least not at the head-end!

    Here are some photos of an SA-7.5 which has an FSR of 7.5 GHz (mirror spacing of about 10 mm):

    The mirror set in the unit I was given has a wavelength range of 850 to 920 nm, which doesn't overlap any of the lasers I really care about. It might be good for use with a Ti:Sapphire laser, which regrettably, I don't happen to have available at the moment. :) It could also be used with single spatial mode diode lasers. But as yet, I have not actually tested this unit, though I see no reason why it shouldn't behave like all the others. I have thought about making suitable mounting plates for a pair of mirrors with the same reflectivity as those in my $2 SFPI, approximately 99 percent at 532 nm, custom coated for a defunct project on substrates that are Melles Griot plano-concave lenses. But since I already have an SP-470-02 (8 GHz, as well as my $2 SFPI), this wouldn't really provide anything fundamentally new. So, I'll simply have to find a suitable laser!

    Thorlabs SA200 and SA210 Scanning Fabry-Perot Interferometer

    The SA200 and SA210 series of SFPI heads from Thorlabs comes in several wavelength ranges: 350 to 535 nm, 535 to 820 nm, 820 to 1,275 nm, 1,275 to 2,000 nm, and 1,800 to 2,500 nm. (The mirror sets are not intended to be swapped by the user, if at all.) The SA200 has an FSR of 1.5 GHz while the tThe SA210 has an FSR of 10 GHz. Google will come up with the relevant Thorlabs info by searching for "Thorlabs Scanning Fabry-Perot" or by searching on their Web site.

    The SA200 and SA210 are of a modern design with a PZT requiring only around 20 V rather than several hundred V to cover approximately 2 FSRs. (This is similar to my PZT beeper-based home-built SFPIs!) So, while Thorlabs does offer the SA201 control box, a normal function generator will suffice to drive the PZT. (The SA201 also include a PD preamp.) CAUTION: The rated input of the PZT is only 150 V, so take care if using a high voltage ramp driver like the SP-476; add a voltage divider or zener or neon lamp clamp to limit the maximum voltage! Another nice feature is iris diaphragms at both the input and output (photodetector) ends of the head. Reducing the apertures aids in alignment and enables the effective resolution to be maximized. A small aperture also tends to reduce the likelihood of back-reflections which may destabilize the source laser. Neither head has an input focusing lens built-in like the SP-470s, so an external lens with its focus approximately at the center of the SFPI cavity may need to be added for best performance.

    The SA201 Spectrum Analyzer Controller includes a ramp generator and photodiode preamp. The ramp driver has a 10 turn pot with a nice large knob for sweep offset, a 7 position switch for 1x, 2x, 5x, 10x, 20x, 50x, or 100x sweep expansion (which actually selects sweep time), a button to select sawtooth or triangle waveform, and a recessed trimpot for sweep amplitude. There is no provision for true sweep expansion, which would maintain the same total sweep time (or equivalently, sweep frequency), but with its amplitude reduced by factors of 1, 2, 5, 10, 20, 50, or 100 around the center of the sweep. Having both sweep expansion and sweep time (as with the SP-476) would be desirable, though hardly essential. The sawtooth has a fixed limited slew rate for its fall (retace) so there is no risk of harming the PZT. The photodiode preamp may be set to a gain of 10k, 100k, or 1M V/A, and there is a recessed trimpot to adjust risetime. The PZT output and scope trigger connectors are on the front panel while the PD input and output connectors are on the rear panel. It has a nice basic set of features and would work well with my home-built SFPIs. :) The design is based on a Lattic CPLD for the waveform generation with virtually no analog timing components, so it should be quite stable over time. My only complaint is with respect to size and weight: With modern components and a wall adapter for power, the entire controller could easily be 1/10th as large and weigh 1/10th as much. And the power button apparently controls a relay producing a resounding CLUNK, as though it's using 10 kW of AC power! :)

    The complete operation manuals for the SA200, SA210, and SA201, may be found on the Thorlabs Web site.

    I wanted an SFPI with the smallest FSR available (from Thorlabs) to check the line-width of a stabilized line-narrowed diode laser at 685 nm. My SP-470-03 is only spec'd from 550 to 650 nm so there was some question as to whether what I was seeing there was accurate. My home-built high resolution SFPI has mirrors that would have been suitable, but the FSR is too small.

    The head I tested was an SA200-5B which has the 535 to 820 nm mirror set. The 1.5 GHz FSR is a bit small for unambiguously monitoring the modes of red (633 nm) HeNe lasers as outlying modes may alias on the display - 2 to 2.5 GHz would be preferred. But for most modern :) applications like checking for single longitudinal (single frequency) performance of diode and DPSS lasers, this isn't a problem. Although only spec'd down to 535 nm, not surprisingly, the performance at 532 nm is still exceptional. In fact, without careful measurements, it would be impossible to tell the difference between the resolvance at the three wavelengths I used for testing: 532 nm (DPSS), 633 nm (HeNe), or 685 nm (diode). I suppose that spec'ing 535 nm may necessary to take into account worst case variations in mirror coating wavelength coverage. However, given the prevalence of 532 nm, it would have been nice to be included within the spec'd wavelength range.

    As it turned out, again not unexpectedly, SP's specs are also conservative so the SP-470 display was very similar to that of the SA200-5B at 685 nm and 532 nm, though the SA200-5B may have had slightly better resolvance. Both these are well outside the 550 to 650 nm range of the SP-470, but 532 nm is very close to the low-end spec of the SA200-5B and 685 nm is near the middle of its spec'd range.

    The lack of a focusing lens also didn't seem to have any detrimental effect on performance with any of these lasers, but they all have a small beam size. The apertures could always be reduced (laser power permitting) for lasers with a larger beam diameter.

    The PD preamp has a wide dynamic range so it's easy to find a combination of gain setting and scope sensitivity that fills the screen without clipping, distortion, or excessive noise. The lack of a sweep voltage knob was not really of any consequence. (There is the recessed trimpot.) It came from the factory set at a full span of approximately 2 FSRs, which would be my default setting anyhow. The sweep expansion switch provides more zoom than could be ever be needed.

    And the Lab Snacks Thorlabs included with the loaner unit were definitely yummy. ;-)

    Of course, even a perfect product can be improved, so here are some suggestions (Thorlabs, are you listening?):

    Burleigh Triple-Pass Scanning Fabry-Perot Interferometer

    This thing showed up on eBay with a listing title of "Burleigh Laser Exciter with Prisms (UNTESTED)". It had what appeared to be cube-corners at both ends and a bunch of wires coming out of it. Even though the seller provided many photos, none was very good at revealing what the device really was. If it was indeed a laser exciter (I have no idea where they got that name), then perhaps there was a fancy laser inside! But the 6 skinny wires hanging out of it terminated with pin-plugs didn't seem right for a laser and would be more appropriate for a bomb trigger. :) It was listed several times over a couple weeks with no takers and with the "Buy-It-Now" price getting lower and lower and lower until....It finally became inexpensive enough that I HAD to satisfy my curiosity!

    Well, the "Burleigh Thing" as I've been calling it turned out to be a Fabry-Perot interferometer with a triple PieZo Transducer (PZT) for fine adjustment of mirror spacing. The mirrors are planar and their separation is adjustable by changing the position of the PZT assembly, so it can be set for a wide range of FSRs. (Free Spectral Range - the extent over which optical frequencies are unique.) This device could certainly serve as a Scanning Fabry-Perot Interferometer (SFPI) for displaying laser modes, though it's more likely to have been used as a tunable etalon or narrow-band optical filter. The normal beam path goes through the mirrors three times, which is like putting three SFPIs, etalons, or filters in series. One thing is for sure: This must have cost someone (probably the U.S. Taxpayer!) a fortune. :)

    Here are a few photos.

    The design wavelength of the mirrors is not known, but their reflectivity is about 75%@633nm and 90%@532 nm, which would be rather mediocre if used with a single pass interferometer. They appear greenish in reflection and weak pink/purple in transmission. Based on appearance and that the reflectance is decent at 532 nm, I suspected they may have been intended for 1,5XX nm (roughly 3x532 nm) and have a higher reflectance there. However, the transmission functions through the interferometers multiply when they are used in series. So with three passes, this thing may still be quite impressive at these wavelengths, even if they aren't optimal. If one pass suppresses out-of-band wavelengths by a factor of 10, with three passes, it would be a factor of 1,000. And specs of standard Burleigh triple pass SFPIs show 93% as a nominal reflectivity. So perhaps I should not have ripped the mirrors out - more below.

    (From: John Barry.)

    "I found that the best way to consider this problem involved thinking about the transmission function for a single Fabry Perot, given below from the Wikipedia entry for "Fabry-Perot Interferometer". (Note: I derived these myself just to check that it is applicable.)

                 (1-R)2                  1
      Te = ------------------- = -----------------
            {1+R2-2*R*cos(δ)}     [1+F*sin(δ/2)2]

    where δ = 4*π*distance between F-P mirrors/wavelength and F is the coefficient of finesse, which is related to, but not identical to the actual finesse:

                 finesse = ---------------------
                  1            π
     2*arcsin(---------) = ---------
               sqrt(F)      finesse
                  1             π
       arcsin(---------) = -----------
               sqrt(F)      2*finesse
                   1                π
               --------- = sin(-----------)
                sqrt(F)         2*finesse
                 sqrt(F) = sin(-----------)
                       F = [sin(-----------)]-2
    For Three F-P interferometers in series, the transmissions should multiply (as you said) giving the final transmission to be Te3 where Te is given above. Unfortunately this curve of Te3 has a different functional form than Te so the Finesse is not really well defined."

    Compare Transmission of Fabry-Perot Resonator versus Optical Frequency with Transmission of Triple-Pass Fabry-Perot Resonator versus Optical Frequency. Note how much narrower and deeper the peaks are with three passes even for very low finesse. It would appear that while multiple passes doesn't do exactly the same thing as increasing the finesse by using higher reflectance mirrors, the overall effect is similar. For three passes, the resolvance has increased by a factor of between 2 and 3. Comparison of Transmission of Single-Pass and Triple-Pass Fabry-Perot Resonator versus Optical Frequency show specific examples for cavity finesse equal to 1, 5, and 20. Even more dramatic than the improvement in finesse is the change in contrast between the peak and the minimum, which can increase by multiple orders of magnitude. (For a finesse of 20, it will be larger by a factor of more than 1,000.)

    Fabry-Perot Interferomter (Univeritat Osnabruck) and Sam's copy of Fabry-Perot Interferomter (Univeritat Osnabruck) has a discussion of the multipass interferometer technique with a triple-pass example (as well as other useful info). But this lecture note doesn't really address the relative merits of multiple passes versus higher reflectance mirrors.

    Burleigh apparently did have optional CCAs for their standard adjustable Fabry-Perot interferometers. The RC-22 (50 mm) fits the RC-110, RC-140, and RC-150, while the RC-27 (70 mm) fits the RC-170. Masks for 3 or 5 pass operation were included to aid is setting beam position and size. The recommended mirror sets for multipass operation had lower reflectivity, at least in part because aligning with the normal high-R mirrors would be virtually impossible. They were: 3 pass only - 93%; 5 pass only - 88%, and if alternating between 3 and 5 pass - 90% as a tradeoff. But the Burleigh Thing is almost certainly a custom job.

    While I doubt that anyone will ever see another one of these beasts, for reference, here is the wiring info for the PZTs. The polarity is arbitrary - I don't know whether a positive voltage expands or shrinks the PZT bars:

         Color          Function
       Red/Orange      PZT1+/PZT1-
       Black/White     PZT2+/PZT2-
       Blue/Green      PZT3+/PZT3-

    The operation manual for a Burleigh RC-44 Ramp Generator lists color codes that agree with these. The "+" is assigned to the HV Ramp, and the "-" to the HV Bias (electrical adjustment for fine tuning of mirror alignment, which I intend to add eventually). So, this color code must be used for other Burleigh SPFIs.

    With the cube-corner assemblies removed (single pass), and set for an FSR of about 6 GHz, I aligned the mirrors using a red HeNe laser and attached the PZTs (in parallel) to my SP-476 controller. And the Burleigh Thing does work, sort of. The finesse for a single pass at 633 nm is truly mediocre, but it is able to resolve the longitudinal modes of a Melles Griot 05-LHR-911 HeNe laser (883 MHz mode spacing) - barely, as overlapping lumps rather than peaks. The finesse is probably optimistically around 8. This poor performance was expected given the low reflectance of the mirrors at 633 nm. With a single frequency 532 nm DPSS laser (Coherent C215M), the display is much better with distinct peaks and a finesse of at least 15. :) Theory predicts a finesse of about 30, but the discrepency could be due to the fact that the beam of the C215M laser is slightly diverging. I'm not sure I have the determination to get it working triple-pass but it may come to that! However, driving all three PZTs using the single HV output of the SP-476 works well enough with no evidence of a change in amplitude over several FSRs. The PZT/Mirror assembly does make quite a racket though - a loud ticking sound - with the SP-476's sawtooth drive, much more so than the small SP "normal" SFPI heads. Or the near silence of my home-built SPFIs! But being able to control the drive voltages (or at least the bias voltages) independently would provide fine tuning of mirror alignment, which is difficult with the large clunky adjustment screws. More on this in the next section.

    And no, I didn't attempt to get the Burleigh Thing to work in triple-pass mode and probably never will!

    I had a really fuzzy vague recollection of being shown an SFPI similar or identical to the Burleigh Thing (but without the CCAs) a few years ago at a local college, gathering dust in the corner of an undergrad physics teaching lab. I mentioned to the instructor who pointed it out to me that the mirrors didn't look like they were good for red, but I didn't realize that they also weren't very good for green, and he probably had no clue either. I might have also commented on how much larger and more massive it was than the SFPIs I build. Then again, perhaps all this was just my imagination. Or, as someone else put it: "When I was a child, I had a fleeting glimpse of something out of the corner of my eye. When I turned to look, it was gone and I cannot put my finger on it now." :) Now totally coincidentally, I had the opportunity to visit the teaching lab because the SFPI was a bit sick. And wouldn't you know, it really didn't look anything like the Burleigh thing! See the section: The Tropel 350 Scanning Fabry-Perot Interferometer.

    Installing High Finesse Mirrors in the Burleigh Triple-Pass Scanning Fabry-Perot Interferometer

    The original mirrors in this thing result in mediocre performance for any lasers I care about (and that's being generous!) with a single pass, though they may have been very good triple pass. It's possible they were designed for 1.5 µm and produce a display barely recognizable as a spectrum at 532 nm or 633 nm. So, I decided to attempt to replace them with mirrors that would be useful for visible lasers in a single pass. (In retrospect, I probably should have tested the thing in triple pass with the cube corners installed and left well enough along, mirror-wise.) Now I just happened to acquire a pair of nice planar mirrors. While also likely designed for 1.5XX µm, they were probably intended to be HRs there and the reflectivity at 532 nm is around 99 percent. As we now know based on my $2 SFPI, an HR at 1.5xx µm will have a lower reflectance at one third of 1.5 µm or approximately the 500 to 533 nm range. I figured that 99 percent at 532 nm (and probably slightly less at 543.5 nm but close enough for Government work!) would be ideal, but that was based on a confocal cavity, not plane-plane for the Burleigh thing. I did not realize how difficult alignment would be with the planar mirrors and not so fantastic Burleigh mirror adjusters.

    Removing the original mirrors wasn't going to be easy because they were glued into recessed tight fitting holders. And in fact, for various reasons, they did not survive very well. So, there was no going back. I had to make the replacement mirrors work.

    The original mirrors were about 1.5 inches in diameter and the replacements were 2 inches in diameter. My plan was to glue them to the lips of the original holders with 5 minute Epoxy. Hopefully, the resilience of the Epoxy would minimize any stresses on the mirrors that might distort them. And these mirrors were about half the thickness of the originals. This approach worked nicely for the back mirror on the PZT because that had plenty of clearance. But what I had not anticipated was that the front mirror fit through a hole in the end-plate attached to the SFPI body and that hole was only slightly larger than the mirror holder, and was definitely less then 2 inches! So, the plate had to come off. But then the larger diameter mirror would prevent the mirror mount from ever being removed without ungluing the mirror. That might be an annoyance in the future but of more immediate concern was the fact that the mirror mount blocked access to the screws securing the end-plate to the SFPI body when it was installed. It seemed like a catch-22 situation that would require some tiny robots to go inside and install the mirror after the end-plate and mount were screwed back together. But as it turned out, there was just enough leeway to permit a ball-end hex wrench to get in between them to tighten the screws to reattach the end-plate even with the mirror attached.

    So, both replacement planar mirrors were installed without incident. That was the easy part.

    Mirror alignment is always a royal pain with a plane-plane Fabry-Perot (F-P) cavity and even more critical with a high finesse F-P cavity. The mirror adjusters on this thing are to put it mildly, not ideal for fine control. OK, not putting it mildly, they stink. :( :) The construction is simply three screws pulling the mirror mount plate toward the end-plate attached to the SFPI body, and another three screws behind the mirror mount plate holding it back. No springs or even springy washers! So, once the alignment is close, it's a matter of tightening each pair of screws against each-other while attempting to do the fine tuning. Optimizing alignment is a matter of observing the scatter off the mirror from the multiple reflections and forcing them into as tight a splotch as possible with these clunky adjusters, which also tend not to be totally independent of each-other. Admittedly, this was designed to be aligned once and locked in place forever. Then fine-tuned would be done by adjusting the bias voltages to the 3 PZTs. (More on this below.) So, fancy mirror adjusters weren't justified. And 1,000 tpi screws or a differential screw scheme would be required to get the sort of resolution needed to really make manual alignment less of a royal pain.

    As if this weren't enough, in a plane-plane SFPI, the laser must be set up so that its beam is perpendicular to the input mirror so that its beam is reflected precisely back on itself. Now, most lasers get mighty unhappy when this is done and make their hurt feelings known by becoming unstable and mode hopping all over the place, even off the optical table. :) But if the beam is at an angle to the input mirror, not only is the finesse achievable much lower, but there is no way to judge when the mirrors are parallel to each-other, and when they aren't, the scatter pattern becomes a curved rather than a straight series of spots, making it much more difficult to even figure out which way to turn the alignment screws.

    So, the laser must be aligned to the input mirror hoping it doesn't complain too much. An optical isolator can be quite effective but is also quite expensive (BIG $$$), although a polarizing beam-splitter and Quarter WavePlate (QWP) might work if the laser is polarized. An optical filter can also be used to reduce the intensity of the return beam at the expense of usable output power. At first, I was using a 5 mW green laser pointer that had a real on-off switch and was designed for more or less continuous operation. But the transmission through two ~99 percent mirrors was so low that only in a totally dark room, was it possible to begin to see anything coming through. Even so, that did enable the SFPI to be more or less aligned and to produce a viewable signal. But the pointer was probably multi-mode on its own and really complaining with constant mode hops, oscillation, and noise. However, these mirrors are very narrow band and almost useless for a green HeNe laser at 543.5 nm, which I tried next. So, I switched to a ~20 mW Coherent C215M laser. Its higher power made life a lot easier and for the most part, it seemed immune to the back-reflected beam, though there was some jitter and the occasional mode hop (which may simply have been due to the C215M case not being temperature-controlled).

    The two SPFI mirrors must be aligned parallel to each other to a very small fraction of a mR - much higher precision than even for a large frame (narrow bore) HeNe laser. And with the mirror adjusters locked tight, just pushing on the massive mirror plate still resulted in a detectable change in the signal. I ended up using rubber wedges between the mirror mount plate and backing plate to fine tune it since it was impossible to use the adjusters alone to peak the signal. There was too much friction and backlash to make precise enough tweaks. A ramp generator with individual bias settings for each PZT would have helped with the fine alignment, but I didn't have one handy.

    So, I built a little bias box for use with the SP-476 (or any other single output ramp generator). It has a built-in 200 VDC power supply (salvaged from a particle measuring system, probably for an avalanche photodiode) so it can be used with any ramp generator. Three pots provide independent bias to the normally grounded ends of the PZTs. There are 100K ohm resistors in series with the PZTs for protection against shorts in the PZTs or wiring, and 0.1 uF capacitors between the pot wipers and ground to bypass any feed-through of the drive voltage via the capacitance of the PZTs. (The operation manual for the Burleigh RC-44 Ramp Generator shows the PZTs being composed of separate ramp and bias sections with the center point grounded. But the Burleigh thing only has single PZT elements. Thus the bypass capacitors may be needed.) With the pots more or less centered, the mechanical alignment can be performed so that the display looks decent with tall narrow peaks, but it doesn't have to be the absolute best possible. Then, the three pots provide convenient and repeatable optimization.

    Since even with near perfect alignment, there is inevitable walk-off and spreading after a zillion reflections, a small photodiode or a small aperture in front of the photodiode also helps to improve resolution. In the end, I used a photodiode with an active area of about 1x1mm.

    With careful setup (laser, mirror, and photodiode alignment), the Burleigh thing is now capable of a finesse of 200 to 300 at a wavelength of 532 nm. For some reason, when adjusted for best resolvance the display has a skinny peak sitting on top of a wider pedestal, with the FWHM point being well above the wide section. The pedestal is assymetric though, more spread out on one side than the other. But this is not an electrical issue such as a slow photodiode preamp - the spread out side swaps if the polarity of the PZT drive is reversed. If a positive input to the PZTs decreases cavity spacing (by increasing the length of the PZT element), the wide side is on the slope of the peak going towards smaller mirror spacing. It almost has the appearance of one or more additional lasing modes just at the limit of resolvance, but this laser has been tested with another SFPI and is known to be pure SLM. And multiple lasing modes so close together that remain stationary with respect to the main mode as these do would be extremely unlikely or impossible anyhow. So, this is very likely simply a result of imperfect alignment as the peak can be increased to a decent height relative to the pedestal with enough fiddling. :) The higher ratio of peak to pedestal also translates into better finesses. One of the better results so far (with the replacement mirrors) is shown in Burleigh Thing Plane-Plane SFPI Display of 532 nm SLM DPSS Laser. The unequal heights of the two peaks is likely due to differences in sensitivity of the three PZTs, which has not been corrected since they are driven in parallel. But this does show how sensitive the finesse (and thus peak height) of a plane-plane F-P cavity is to mirror parallelism: A miniscule change in alignment over a movement of less the 1 µm (with a mirror spacing of over 37 mm) is enough to produce a sizable effect. The relative heights of multiple peaks in the display can also be changed at will with the mirror bias pots.

    The main benefit of a plane-plane SFPI is that the FSR can be varied over a wide range by changing the distance between the mirrors. But even with the non-adjustable mirror (and PZT) mounted in cylinder that's a precise fit to the SFPI body, moving it any significant distance will mess up alignment, requiring going through mechanical alignment all over again!

    So the conclusions of this exercise (and my plane mirror SFPI) may be that attempting to make a high finesse plane-plane SFPI is not worth the effort unless it's absolutely needed, usually to achieve a higher FSR than what's practical in a confocal SFPI due to the very small highly curved concave mirrors that are required. The highest FSR for a commercial confocal SFPI I know of is 30 GHz from Coherent. Those mirrors, which have an RoC of only 2.5 mm, look like the interior of a pea that's been cut in half with each being a good portion of a full hemisphere. How are high quality mirrors like that even ground and polished? And confocal SFPIs from most other manufacturers only go to 8 or 10 GHz. A plane-plane SFPI can go up to over 1 THz! But if this is required, start with a precision instrument like the TecOptics FPI-25 or Tropel 350 described below. As a practical matter, though, a monochromator-based Optical Spectrum Analyzer (OSA), while pricey, is probably a better and easier instrument to use with high bandwidth lasers.

    Burleigh TL Series Plane-Plane Scanning Fabry-Perot Interferometer

    This small head has no model number and originally I thought it might have been custom and intended as a tunable etalon though it was in the standard Burleigh 4-axis gimbal mount as shown in Burleigh Plane-Plane Scanning Fabry-Perot Interferomter Head in Gimbal Mount. And at first I assumed it was a normal confocal SFPI head with IR (1,5xx nm) mirrors (based on appearance) but the behavior was strange: It was not possible to even obtain "lumps" for a display using a green (532 nm) DPSS laser (which should have worked at least marginally at 1/3rd the design wavelength). The reason became obvious once I realized that the mirrors were planar - my ramp generator was driving only one of the three PZT stacks so the mirror was tilting rather than translating. Several view of the head alone are shown in Burleigh Plane-Plane Scanning Fabry-Perot Interferomter Head. The bottom views are looking into the head with the rear focusing lens and photodetector assembly removed.

    Testing with a Melles Griot 05-LIR-151 1,523 nm HeNe laser and ramp driver wired to all three PZTs in parallel, resulted in a good enough display to deduce the FSR. The distance between longitudinal modes of the 05-LIR-151 is 438 MHz appearing at 1/32nd of the distance between where the display repeats for an FSR of around 14 GHz. In the interest of round numbers, I'm guessing it is actually 15 GHz. :) It may that the mirror spacing is adjustable but I have no idea how.

    Three screws at the back of the head provide for coarse mirror parallelism adjustment with the PZT bias settings doing the fine alignment. Then the Pan and tilt of the mount are used to set the beam to be exactly normal to the mirrors. This, of course, results in a strong back-reflection directly into the laser, so ideally some type of optical isolator is required. However, simply attenuating the beam with a neutral density filter allowed for acceptable stability. The finesse is poor - optimistically 75 - but this may be due to imperfect mirror parallelism, the slightly expanding beam of the laser, and the coatings not being optimal for 1,523 nm.

    Some time later, I found the Burleigh/EXFO TL Series Laser Spectrum Analyzer Brochure which seems to fit. :) So, this is probably the TL-15-15-14-99 laser spectrum analyzer (15 GHz FSR interferometer body with TL-115-14-99 1,450 to 1,650 nm mirror set). Now I need to find a proper RC-93 controller, or better yet, FPS-250-NuView PC USB controller and EXFO NuView Laser Spectral Analysis Software. ;-)

    Burleigh RC-42 Ramp Generator

    The Burleigh RC-42 is a high voltage ramp generator includes the usual ramp amplitude and speed controls as well as three bias controls and three gain trim-pots. These provide for electrical fine tuning of mirror alignment and compensation for unequal PZT sensitivity on interferometers with triple PZT stacks like the Burleigh triple-pass SFPI described above. It can also be used with single-PZT instruments by ignoring 2 of the 3 outputs.

    Here are some relevant specifications, determined by testing. All values are approximate:

    Here are two photos:

    The RC-42 is both larger and heavier than it appears in the photos. At first I was wondering if there might be vacuum tubes inside! The power transformer may be from the tube era - a Stancor 8418 (230-0-230 V rated 50 mADC, 6.3 V at 2.5 A) but it is all solid state and even has several ICs, though a few of them are TO5 cans jammed into 8 pin DIP sockets for some reason! As it turns out, much of the weight is in the thick steel chassis and cover. That doesn't help it to look any smaller though. :-)

    Here is the pinout for the DB9M PZT connector:

      Pin   Function
       1    PZT Drive 1
       2    PZT Bias 1
       3    PZT Drive 2
       4    PZT Bias 2
       5    PZT Drive 3
       6    PZT Bias 3
       7    NC ??
       8    Chassis Ground
       9    125K ohms to Ground

    There is a wire attached to pin 7 so it may actually do something, but it was not obvious as there was no voltage on it and infinite resistance to Ground.

    TecOptics FPI-25 Scanning Fabry-Perot Interferometer

    The TecOptics FPI-25 is a general purpose instrument with a plane-plane cavity that can be set for an FSR of more than 300 GHz (mirrors almost touching) to a much lower FSR when the mirrors are at their maximum extent. They accommodate easily interchangeable mirror sets covering several wavelength ranges. There's no reason the FPI-25 couldn't also be configured as a confocal SFPI with suitable mirrors. Of course, like other voltage-controlled Fabry-Perot etalons, it may also be used in other ways in addition to as an SFPI for CW lasers. With a slow ramp, it can generate the time-averaged spectrum of a quasi-CW laser, and with a fixed (or feedback-controlled) drive voltage, it can act as a tunable optical filter.

    The FPI-25 is similar in basic capabilities to the "Burleigh Thing" described above (when used with only a single pass) but has a somewhat better set of adjustments designed to be used for optics experiments. Cavity spacing is varied by a knob turning a fine-threaded rod that moves the massive Invar resonator assembly. It may be locked in place once the desired position is reached. Three knobs which turn differential screws are used for extremely precise mirror adjustment, though this is considered "coarse" compared to what the PZTs can provide. The triple PZT with its mirror is fixed to the base. Changing cavity spacing still may require fine tuning of the mirror alignment, but not by much so that individual control of the three PZT offset voltages may be enough for this. The entire instrument sits on a pan-tilt mount so aligning it to the input source is quite easy. It may also be removed from the base and mounted on a (sturdy) post.

    Based on operation manuals I've seen, some versions of the FPI-25 were sold for awhile by Melles Griot, possibly after TecOptics gave up or sold the interferometer product line. The manual for systems that appear identical to those from TecOptics lists three models differing only in the maximum cavity spacing/minimum FSR: W1000 (60 mm/2.5 GHz), W2000 (100 mm/1.5 GHz), and W3000 (150 mm/1 GHz). There were at least two ramp generators. The basic version (FPZ-1-RG, photo below but not referenced by Melles Griot) doesn't have individual controls for bias and offset as does the fancier FPZ-3-RG. Melles Griot also offered an updated version of an optical head similar (but not identical) to the W1000 (50 mm/3 GHz) with their own redesigned ramp generator - the 13-FPC-001. (At least it has a more stylish front panel!) The main change to the optical head aside from reducing the FSR slightly seems to be that the resonator with the adjustable mirror and photodiode assembly is attached to the base and the fixed mirror on the PZTs is what moves to vary the FSR. However, as of 2010, there are no references to either system or their components on the Melles Griot Web site. But here are the two manuals (with permission from Melles Griot):

    Both manuals also include nice sections on basic interferometer theory.

    I have used an original TecOptics FPI-25 with the FPZ-3-RG, but don't actually own one. That system had a separate photodiode preamp box, the DA-1 (mentioned in the Melles Griot manual but with no description). Here are some photos of a TecOptics FPI-25 from an eBay auction. This would be equivalent to the Melles Griot W3000. I should have bought it but wasn't an SFPIs enthusiast back then. :) (I'll be happy to acknowledge the source of the following photos if the owner will come forward):

    Tropel 350 Scanning Fabry-Perot Interferometer

    Tropel was a manufacturer of a variety of laser-related equipment including interferometers and stabilized HeNe lasers, but apparently sold off these products to Coherent, possibly in the late 1970s. I don't know if the Tropel metrology division of Corning is a descendent of the same Tropel.

    The Tropel 350 is a real beauty, but a monster. OK, a beautiful monster! Or, would be if it was cleaned of years of neglect and given a nice polish. :) It has a heavy 4-bar (probably Invar) frame roughly 8x8x12 inches overall, and weighs in at over 30 pounds. Everything is visible which make it extremely useful in a teaching lab (which is where this one is located). It has a large planar mirror mounted on one end-plate and a large planar mirror that can be positioned along the length of the resonator by loosening some set-screws and moving it on the rods. The fixed mirror has three very smooth precise micrometer adjustments. The plate on which the movable mirror is mounted is attached to it by three cylindrical PZTs (about 1-1/4 inches long by 3/4 inch in diameter) that are fully exposed. The mirrors are 2 inches in diameter.

    The controller is fairly basic with 10 turn pots for scan amplitude and offset, a selector for speed, and three bias adjustments for fine mirror alignment. A Thorlabs photodetector with focusing lens (of course not original equipment!) was mounted externally. (Though there really is no "inside" to this SFPI!)

    Although there was a Burleigh manual with the setup, this appears to be Tropel through and through (unless the two companies had some connection in the past). There were no photos or diagrams of the equipment in the manual, so it was rather generic.

    The design wavelength of the installed mirrors is not known but it is definitely not for red and not for green. The reflectivity for 633 nm (red HeNe) is probably less than 75 percent and the reflectivity for 543.5 nm (green HeNe) is probably around 90 percent. (It might be slightly better at 532 nm.) Using a JDS Uniphase 1674 green HeNe laser results in really dreadful finesse - optimistically maybe 20. With the mirror spacing set for an FSR of 3 GHz (about 2 inches), it can barely resolve the longitudinal modes spaced 325 MHz apart. The display looks more like hands with upward pointing fingers than nice narrow peaks. :)

    So, the mirrors might be designed for argon ion blue wavelengths but there was no suitable laser to try. Or, as with my Burleigh Thing, they may be designed for 1.5 µm. However, this instrument probably predates the telecom age (and any need for 1.5 µm) by a few decades!

    I've offered to find a set of 98 to 99 percent 633 nm mirrors that could be installed (possibly with an adapter). These would result in a much more respectible finesse - 150 to 300 under optimal conditions, but at least half of this with ease. Anything over 1/2 inch in diameter is probably adequate, though 1 inch would be desirable simply to prevent the monster from looking totally silly with tiny mirrors. :-)

    Here are some photos (coming probably never but here are the descriptions):

    I even found a reference to the use of the Tropel 350 as a triple-pass SFPI like the one described in the section: The Burleigh Triple-Pass Scanning Fabry-Perot Interferometer, with Tropel RC22 cube corners and RC70-B4 2" mirrors (whatever those are!). Unfortunately, the authors don't mention why this configuration was used. (And the rest of the paper is absolutely boring.) Google easily found several other papers referencing the Tropel 350 so it must have been fairly popular (as these things go) at some point in the past.

    The reason I was able to play with the Tropel 350 was that I received an urgent email from the lab instructor (who I had sold a one-Brewster laser to a few years ago) that the SFPI seemed to be misbehaving. It was occasionally making an electrical arcing sound and then ceasing to move the mirrors. You probably never realized that laser doctors make house calls! :-) In the end, all I could find was that perhaps there was some dirt or some other contamination causing an intermittent short circuit in one of voltages to the PZTs. It sounded like it was originating at the PZT itself, but it could have been in the HV connector or elsewhere since the resulting rapid change in voltage would make the PZT and mirror act like a loudspeaker of sorts. And unplugging and replugging the HV connector and jiggling wires seemed to make it go away, at least temporarily. So, I recommended carefully cleaning the exterior with alcohol including the HV connector and to email me in the morning. :) This would also greatly approve the appearance for the photos I requested! If external cleaning didn't help, then it would be necessary to disassemble the movable mirror mount to get inside the PZTs for cleaning, but that too should be very straightforward.

    Later I found a question about this same instrument with the same arcing problem 7 years ago on a teaching apparatus list server! Here is the only relevant reply:

    "The one that we used to use was modified for an MHV terminal for the high voltage piezo-drive to correct arcing from the previous connector. This also allowed us to use our own driver (that one died).

    There are special spring-like washers and insulators that keep the cell from shorting, perhaps the piezo is touching somewhere in the tube? Or the drive lead has detached from the ceramic?

    If it is like our old one, you must make a tool, like a hollow 3/16" tube with a 1/16" "tee" pin at the end, (shaped like a T) to tighten and adjust the cell after disassembly. This is for a lens ring that held it in, but also lets the beam through.

    Good luck, S. Anderson"

    The instrument described in this posting sounds like it has only a single PZT, but the advice about replacing the connector is even more valid for one with three PZTs! The original 6 pin connector is definitely not rated for 1,000 V and would be a prime suspect, especially given that unplugging and replugging it seemed to make a difference. The replacement doesn't really need to be an official high voltage connector. A Molex or AMP multi-pin nylon shell would have the needed dielectric strength, especially if only every other position were used. In addition, I'd suggest adding a resistor in series with each PZT to protect the driver should a short occur in the PZT assemblies. Something around 100K ohms should provide adequate current limiting without excessively distorting the drive waveform. (This doesn't really matter if it uses shunt regulators like the SP-476.)

    Burleigh CFT-500 Scanning Fabry-Perot Interferometer

    This one looks more like a fat 20 mW HeNe laser than an SFPI. :) There was no model number on the unit I have but based on photos and a spec sheet, it is probably a CFT-500 with an FSR of 150 MHz and minimum finesse of 125. See "CF Series" under Burleigh Brochures. However, it's also possible this was a custom unit and may have been intended as a tunable etalon (filter) rather than as an SFPI, though the difference is simply in application. (But there is no built-in photodetector as would be desirable for an SFPI.) More likely, the detector simply vanished. The unit consists of a HEAVY beautifully machined chrome-plated resonator cylinder about 22 inches in length and 1-3/8 inches in diameter installed in an insulated heater jacket with a 1.5K ohm thermistor temperature sensor for thermal regulation. The front mirror is mounted in an end-cap that has threads for fine tuning the cavity length and a ring to lock it in place. The back mirror and PZT is in an end-cap that screws on and seats against a fixed joint. It also has the 1K ohm thermistor and an SMA connector for the high voltage PZT drive. Both front and back have glass windows to prevent contamination. I don't know if it is intended to be hermetically sealed but it does come at least close with O-rings on all joints. And there is a small flat-head screw also with O-ring seal to perhaps allow for the pressure to be equalized - or something. ;-) The distance between the mirrors is just about 19.7 inches (0.5 m) for an FSR of 150 MHz. This entire assembly - which weighs in at around 8 pounds - is housed within the insulated heater in a light-weight aluminum cylinder about 23-1/2 inches in length and 2.5 inches in diameter. The total weight is about 11 pounds. There are separate cables for the high voltage PZT drive and the heater/thermistor.

    I acquired this unit on eBay. Well, parts of one plus a second outer cylinder with insulated heater and not much else. In fact, on eBay they were listed as "Burleigh Lasers". :) At first I was unable to identify the wavelength of the mirror set. It had the appearance of having high reflectance for the all too common (and useless to me) 1,5xx nm IR wavelength - blue/green in reflectance and pale pink/orange in transmission. However, these usually have at least some amount of reflection at 532 nm. This had very little - under 25 percent. But using an Ocean Optics USB2000+ with a tungsten lamp to get an idea of the transmission in the visible range, there was a very obvious stretch of low transmittance from around 475 to 510 nm. Indeed, at 473 nm, the transmission is around 1 percent and at 488 nm the transmission is only around 0.2 percent. Bingo! This should have a nice high finesse for the quite common 488 nm Coherent Sapphire OPSL and other modern replacements for the 488 nm argon ion laser. However, this does not appear to be one of the standard mirror sets, which have a wider wavelength range (450 to 550 nm would be the relevant one). The finesse may be more than 750 at 488 nm for a resolution of better than 200 kHz. It's still possible that the design wavelength is something other than 488 nm. It could even be an IR wavelength 3 times one somewhere in the range of the visible wavelengts (475 to 510 nm) or 1,425 to 1,530 nm, in which case its reflectance would be even closer to 1 and the finesse could be higher by as much as a factor of 10 or more. This in fact seems like a more realistic range except that none of the standard CFT-100 mirror sets have a wavelength range in the IR consistent with the measured VIS transmission.

    Unfortunately, one of the first things I found out was that the PZT with mirror attached had broken free of its glue and was dangling by a single wire. (Not surprizingly, there was a scribbled note on the outside of the outer cylinder: "Broken, please fix.". The seller didn't mention that or show it in the auction photos!) A chunk of the PZT material was also missing from the PZT cylinder, pieces bouncing around elsewhere. The good news is that neither mirror appears to have been damaged and the missing piece of the PZT probably won't make that much difference in performance. Re-attaching the PZT cylinder was a bit of a challenge though because it is recessed inside the end cap of the main SFPI cylinder and that doesn't come apart. I was afraid at first that it wouldn't be possible to assure decent mirror alignment. But at what is probably the original orientation, it appeared to seat square to the axis. Some careful application first of 5 minute Epoxy in 3 equidistant spots, then slow-cure stronger stuff in between seems to be satisfactory. The hole where the missing bits of PZT used to be came in handy though as the electrical connection to the outer surface of the PZT cylinder used to be via spring clips between it and the inner surface of the metal end-cap. (I have no explanation as to why a wire wasn't soldered to that like the inner surface.) There was no practical way of installing them ahead of time, but once the glue had cured, it was a simple matter to use a dental pick to carefully insert a clip on each side of the hole. (I never did find a 3rd clip, though I suspect there may have been 3 originally.) Part of the SMA socket had also broken off and was stuck in the cable plug. That was soldered back in place. It broke off again later, so a wire was simply soldered to the connector and attached via the screw that sealed the interior. The wires to the thermistor had been cut flush, so the remains of the cable had to be drilled out and the wires were then reattached directly. Aside from these minor problems, it was in perfect condition. :) But the final result should be very close to what it was when new. Burleigh High Resolution Scanning Fabry-Perot Interferometer (Tunable Etalon) shows the resonator cylinder, heater (foam insulating cylinder not shown), and outer aluminum cylinder, along with closeups of the front and back of the resonator and the interior of the rear end-cap housing the PZT and rear mirror. (My designations of front and rear are of course arbitrary.) The SMA connector is for the PZT HV, the twisted pair is for the temperature sensor, and the flat head screw provides access to the interior. The mirrors are pale orange/pink in transmission and blue in reflection, so lighting and angle determined how they appear in the photos.

    The rear mirror is glued into a holder which attaches to a flange glued to the end of the PZT cylinder via the 3 screws that are visible in the bottom middle photo. The other end of the PZT cylinder is simply glued to the end-cap, deep inside, adjacent to the window. Repair would have been somewhat easier if the window were removed, but that could not be done non-destructively. So, I had to use a long narrow stick to reach in and apply glue taking care to avoid getting it all over the window and inside of the PZT cylinder. This was mostly successful. ;-)

    My initial test with a JDSU 2201 488 nm air-cooled argon ion laser was inconclusive as the signal was extremely noisy with the PZT drive turned off, even after minimizing back-reflections, but maximum amplitude when the interferometer was perfectly aligned. The noise wasn't apparent with the photodiode directly in the beam. So, perhaps, it was actually frequency jitter due to high ripple in the ion laser power supply interacting with the highly selective interferometer. Yes, I know this is grasping at laser straws.... However, the expected spots did appear at the output indicating that alignment is probably acceptable. And the PZT makes nice ticking sounds when driven with my SP-476.

    Final testing had to await a suitable narrow line-width low noise 488 nm laser. A Coherent Sapphire 488-50 was the perfect candidate. It was first checked on a Spectra-Physics 470 SFPI head, which has an 8 GHz FSR, and was found to be pure single longitudinal mode from 12 to 50 mW (the range over which output power can be adjusted via RS232) both when stable and during the transitions. Then the Burleigh high resolution SFPI head was mounted on the same platform used to test the Ultra-High Resolution Scanning Fabry Perot\ Interferometer.

    The results so far are disappointing. The PZT works well, with ~200 V to span 1 FSR. There is no evidence of asymmetry from the missing chunk of PZT. But the maximum finesse appears to only be around 100, not the ~750 or more I was expecting. No amount of alignment was able to reduce it significantly. I'm quite sure the mirrors are clean and aligned well enough (parallel to each-other and centered). I did try several random focusing lenses at the input with no improvement. Proper mode matching at the input could still be the key though. But now I wonder about the line-width of the Sapphire laser. Could it be that the SFPI resolvance is actually much better than the laser and what's being displayed is really the laser line profile? There are no published specifications for its line-width. For that matter, Sapphire lasers are not even guaranteed to be SLM, just that many like this one are SLM. The Sapphire is a doubled OPSL - Optically Pumped Semiconductor Laser. But will its line-width be closer to that of a diode or a solid state laser? My inclination is that the wide peaks are still likely due to an SFPI issue but cannot be certain at this point.

    Tropel 2440 Scanning Fabry-Perot Interferometer

    This is a very unusual compact SFPI head best described as "cute". :) See Tropel Model 2440 Scanning Fabry-Perot Interferometer. It is about one half the diameter and length of most of the other common confocal SFPI heads such as the Coherent/Tropel 240 and SP-470. Note the AA cell for comparison. (No, unfortunately, it doesn't run on batteries!) If one were to use a miniaturizer on a more common SFPI head, the Tropel 2440 might be the result. :-) It is small enough to fit in some one inch adjustable mirror mounts, though minor modifications to the mount may be necessary. But an adapter (not shown) enabled it to be installed in a standard two inch optical mount - which of course partially eliminates the benefits of being so compact! But there are at least two other notable characteristics associated with the 2440 beyond its size:

    The 2440 head is easily disassembled (reversibly) to its major components: The detector assembly unscrews from the body of the SFPI revealing the recessed lock ring securing the rear mirror. A shroud enclosing the front of the head is simply pressed onto the main body cover over an O-ring. (The cover is normally sealed with adhesive and not supposed to be removable minor details like that never got in the way of a nice dissection. But that's also reversible.) The shroud protects the front mirror and has a threaded hole in front for an aperture or possibly other optic (like a polarizer). Pulling off the shroud reveals the threaded cap securing the front mirror to the end of the PZT cylinder.

    From left to right at the bottom of the photo are: photodiode assembly, main body with PZT cylinder and rear (curved) mirror (hidden), front (planar) mirror, threaded ring to secure front mirror, main body cover, front shroud, and front aperture (to assist with alignment). The rear mirror is recessed but can also be removed using the appropriate spanner or other suitable tool to unscrew its lock ring. However, I don't know if the mirror sets were supposed to be replaceable by the user. It's not obvious how the precise mirror spacing is adjusted, although it may simply be that the rear mirror seats against a spring or O-ring and the lock ring is then able to move it enough to cover the required adjustment range. I haven't attempted to remove the rear mirror - yet.

    While the entire detector assembly is easily replaced, it's locked together with glue and this one had a bad an intermittent connection inside! So I had to rip it apart, which required the use, among other things, of a heat gun, files, pliers, wrenches, clamps, and a liberal sprinkling of 4 letter words. :( :) But after reassembly, it still looks decent. And although the PD itself was still good, it is now socketed just in case. ;-)

    This particular 2440 probably has a mirror set with an FSR of 7.5 GHz. That's somewhat of a guess based on eye-balling the mirror spacing and assuming the curved mirror is the same as one of those for the 2440's bigger brother, the Model 240. It's possible there is sufficient space to install a mirror for an FSR of 1.5 GHz, requiring a hemispherical confocal mirror spacing of around 25 mm, one half that of the standard confocal cavity. With a suitable spacer, a 50 mm RoC rear mirror could be installed about 1 cm further back, and the front planar mirror (which would be unchanged assuming the wavelength range remained the same) could be positioned forward to extend the cavity (though it may not fit inside the shroud). Or, mirrors for a non-confocal Mode-Degenerate Interferometer (MDI) spacing could be installed. This would enable the head to be set up for an intermediate FSR using a 1.5 GHz mirror set. Now that's a thought. ;-) For example, the 3rd order resonance would have an FSR of around 2.0 GHz with a mirror spacing of 1.25 cm, which should be easily accommodated by adding a spacer in front of the rear mirror. (See the section: Selectable FSR Mode-Degenerate Fabry-Perot Interferometers.)

    The wavelength range of this unit appears to include 1,064 nm. There is, of course, nothing on the label, only "Tropel 2440". The mirrors are bluish in reflection, slightly pink in transmission, which normally suggests either VIS blue/green, or IR at 1,5xx nm. But wavelengths from 457 to 544 nm (multi-line argon ion and green HeNe) pass right through, as do wavelengths near 1,319 nm (YAG) and 1,523 nm (IR HeNe). And the photodiode is silicon since its voltage drop is around 0.7 V and it has good sensitivity to 532 nm. An Si PD would be blind above around 1,100 nm. Finally, I dug up a 1,064 nm laser and bingo! The mirrors block its beam, so 1,064 nm is likely within the wavelength range. The beam from a 870 nm diode laser passes through without much attenuation so the wavelength range can't extend below around 900 nm. My guess would be that it has the same coating, and is similar or identical to the mirrors for the 1,000 to 1,100 nm range or possibly the 900 to 1,070 nm ranges available for the Tropel model 240. But to achieve similar finesse in the hemispherical cavity (which cuts finesse in half), it would need to have a higher reflectance (assuming the coatings on both the planar front mirror and curved rear mirror are the same). (See the section: Coherent/Tropel 240 Scanning Fabry-Perot Interferometers.) Thus the complexion of the mirror coatings was a false trail. Actual testing will have to await a low power reasonably well behaved (longitudinal mode-wise) 1,064 nm laser I can conveniently power where my SFPI test setup is located. ;-)

    The only reference I have to the Tropel 2440 is a 1971 advertisement which states: "The Model 2440 is the only spectrum analyzer that converts to Pz drive. Unmode-matched finesse is >150. Scan 1 free-spectral range in 25 V. And it's inexpensive.". I'm not sure what "Pz drive" means :) but everything else is consistent with a hemispherical confocal cavity SFPI. The complete print Ad which also lists other Tropel SFPIs may be found at Tropel Laser Spectrum Analyzers (1971).

    I'm tempted to replace the mirrors ones for 633 nm. The rear one would be 43 mm RoC/1.7 GHz mirror as in my home-built SFPIs, and the front one would be a planar HR or near-HR. Even if the cavity length can't be extended to the required 21.5 mm for the hemispherical confocal cavity, a higher order (shorter) setting could be used, with the benefit of being able to be set for a larger FSR as noted above.

    The miniature RF connectors it uses are the same as those on some other Burleigh SFPI heads, but so far, I've been unsuccessful in identifying them. If anyone knows what they are called and/or has mating cables available, or more information on or other examples of this type of SFPI, please contact me via the Sci.Electronics.Repair FAQ Email Links Page.

    Burleigh LD-1500S Scanning Fabry-Perot Interferometer

    (Portions from: Bob Arkin.)

    The Burleigh LD-1500S is a complete plano-plano SFPI (optical spectrum analyzer) in a self-contained package except for the display (oscilloscope). The 1500S has a range of 1,280 to 1,580 nm from an FC fiber input which goes to a set of flat etalon mirrors in a TL series piezo. The fine alignment of all three axes was adjustable from the front panel. (Unlike confocal cavity SFPIs, alignment is much more critical for plano-plano interferometers.) It should be possible to install mirror sets for other wavelength ranges that will work with the IR photo detector. (It was possible to order a LD-800S which covered 760 to 860 nm.) Or, by also changing the sensor, to almost anything. :)


    Here are some photos, also courtesy of Bob:

    Melles Griot Scanning Fabry-Perot Interferometer

    This is a classic SFPI, though Melles Griot called it an "Optical Spectrum Analyzer", which is normally considered to be a totally different beast. :) (An alternative designation is "Laser Spectrum Analyzer".) See Melles Griot Scanning Fabry-Perot Interferometer. The Melles Griot SFPI was probably not very popular as it has taken 20 years to come in contact with one - and that was sent to me for repair from a former high level employee of Melles Griot! Note the inventory label. :-)

    The controller (ramp driver/photodiode preamp) is fairly basic but adequate with controls for ramp SHIFT (offset), RANGE, and RATE, and PD OFFSET and GAIN.

    Melles Griot SFPI heads

    There have been at least two styles of SFPI head. The ones I've seen I'm calling "Type 1" and "Type 2". They both have a 2 GHz FSR and a wavelength range from 540 to 640 nm, so should be more or less interchangeable, but not quite. ;-)

    If either of the photodiode cables disappears, hookup is at the very least, annoying. Using strange cables was probably intended to minimize the chance of accidentally plugging the HV into the PD or scope. For the Type 2 head, a normal BNC screw-type panel mount connector can be retrofitted relatively easily.

    Melles Griot SFPI controller

    As noted, the controller is bare-bones but adequate. The minimal set of controls means it's straightforward to adjust (and more difficult to screw up).

    It was necessary to repair two of these controllers. The major problem were failed high voltage capacitors and the subsequent blowing of the line fuses. The HV for the ramp driver is provided by a 420 VRMS winding on the line transformer feeding a voltage double for a nominal output of around 1,200 VDC. The doubler circuit consists of four 1N407 rectifiers, four 100 µF 450 V electrolytic capacitors, and 4 220K or 330K 1/2 W equalizing resistors. Those resistors were marginal in terms of power dissipation, so over time their value drifted and eventually caused enough of a voltage imbalance on the caps for some of them to fail, thus blowing the line fuse. One was even a dead short. Replacing the capacitors and equalizing resistors (with 345K, 1%, HUGE mil-spec resistors which were laying around), and the fuses restored operation. There was clear evidence of prior incomplete repairs - several caps were replaced, but the underlying cause of the failures - the equalizing resistors that drifted in value - had not been replaced. Further, whenever any capacitor in a series setup like this is replaced, they should all be replaced because the leakage currents will not be even close to equal, especially bewteen new and old caps, and those that may have been overstressed but didn't fail totally.

    WARNING: When the covers are removed, there are MANY exposed component leads and PCB traces, as well as the heatsink for the high voltage ramp driver transistor, that may have up to ~1,200 V on them. The energy stored in 400 µF 450 V capacitors is potentially lethal. Even after powering down, while the equalizing resistors also act as bleeder resistors, the time constant is typically 20 seconds or longer depending on their value.

    The low voltage circuitry is very simple and straightforward to troubleshoot consisting of an ICL8038 timing generator IC (sort of a 555 on steroids, which can generate square, triangle, and sine waves directly, which is unfortunately obsolete - it IS a cool chip), an LM348 quad op-amp for driving the HV TO3 NPN transistor with feedback to maintain a linear output, and an OPA op-amp for the photodiode. There is a ±12 VDC power supply for the LV circuitry using 7812 and 7912 three-terminal regulators. The PCB is single-sided and wide open. And while a schematic is not available and even people who worked at Melles Griot have no idea if one exists, a suitable facsimile could be traced out in an hour or so should the need arise.

    The power indicators used telephone "slide base" incandescent lamps were also burnt out. They run on ~18 VRMS AC from the power transformer and were replaced with a pair of ultra-high brightness 3 mm green LEDs wired in parallel with reverse polarity and soldered to the sides of the original slide base lamp assembly. A 10K ohm resistor added external to the indicator assembly provided current limiting. These LEDs should outlast the Universe. ;-)

    Overall, the performance of the system is good, though resolvance would probably benefit from a smaller beam diameter than the 05-LHR-911 provides. The Type 1 head may be slightly better in this regard than the Type 2 head but that's not certain.

  • Back to Laser Instruments and Applications Sub-Table of Contents.


    Basic Description

    While there are many ways of determining the wavelengths produced by a laser or other light source, the simplest one beyond the use of calibrated eyeballs is probably a monochromator. It's possible to construct one from inexpensive parts but they also show up surplus by themselves or as part of other optical devices like spectrophotometers, DNA analyzers, fluorescence spectrometers, and other lab equipment with even more obscure names.

    A monochromator is an instrument which accepts a light source as its input and can select not quite a single wavelength, but a narrow band of wavelengths. A monochromator is probably the simplest device for determining wavelength of a laser or other light source where standard calibrated eyeballs aren't sufficient. (The longer harder to pronounce term "monochronometer" is also commonly used but it refers to the same type of device.)

    There are many types of monochromators but here we only describe one of the simplest, consisting of the following:

    The arrangement above using a planar diffraction grating is acceptable if the input light source is well collimated and aligned with the monochromator's optical axis. But, making the diffraction grating slightly concave with its two focal points at the slits makes the system less sensitive to the orientation or divergence of the input beam and provides better selectivity since it is essentially imaging the light at the entrance slit into the exit slit. Alternatively, spherical lenses or mirrors can be used with a planar diffraction grating to achieve the same effect.

    Due to the way a diffraction grating selects wavelength, if the linear travel of the lead screw is converted into rotation of the diffraction grating by a lever of the proper length, the result is a linear relationship between the "nut" location on the lead screw and wavelength. As long as the pitch (lines/mm) of the diffraction grating is known accurately, the relationship will be exact. Thus, a simple multiturn precision dial can be used to read off wavelength. Or, for an automated instrument, a stepper motor will have a constant nm/step size.

    A fully optical monochromator - with no electronic detector - is perfectly adequate and may actually be preferred for measuring laser wavelengths in the visible range at least since almost any laser is powerful enough to result in a beam at the output of the monochromator to be easily seen. However, for measuring the spectrum of something like a glow discharge (as in the bore of a HeNe laser tube) or UV or IR lasers, a high sensitivity detector is essential. Spectra for varioue elements and compounds can be easily found by searching the Web. The NIST Atomic Spectra Database has an applet which will generate a table or plot of more spectral lines than you could ever want.

    CAUTION: When exploring the interior of a monochromator, DO NOT touch the surface of the diffraction grating. Cleaning a grating without damage is difficult at best, and may be impossible for some types of gratings

    The following sections describe some typical monochromators.

    Intruments SA Model H2O 1200 Monochromator

    An example of a simple monochromator is the Instruments SA H20 1200VIS, designed to span the visible spectrum with either manual or motorized control. A slightly battle weary sample of this unit is shown in Instruments SA H20 1200VIS Monochromator. A diagram of its organization is shown in Basic Monochromator Opto-Mechanical Layout and the actual underside mechanism in Instruments SA H20 1200VIS Monochromator Lead Screw Lever System.

    The two mirrors allow the input and output beams to be co-linear but have no other effect. A variety of input and exit slits permit the resolution and sensitivity to be easily changed. All inner surfaces of the monochromator as well as some additional "Absorbers" are coated with a super flat black material to absorb as much stray light as possible. This includes both unavoidable scatter as well as the zeroth and higher order diffracted beams (not shown) from the grating.

    In the diagram, a hypothetical light source consisting of red and green lines is shown, perhaps from a very strange Ar/Kr ion laser. The green beam is diffracted less than the red beam and is thus not passed through the exit slit.

    Assuming the mechanical design of the monochromator is correct, only two adjustments are needed for calibration: The angle of the diffraction grating with respect to the lever, and the dial setting with respect to the lead screw.

    When I found the 1200VIS, both of these were far off. Simply adjusting the dial to coincide with a 632.8 nm red HeNe laser resulted in a green 532 nm DPSS laser pointer reading 529 nm. This indicated that the lead screw wasn't moving the lever enough over that range. To remedy this, I slightly loosened the set screw locking the lever to the diffraction grating shaft, but still tight enough so that normal dial twiddling wouldn't affect the relationship. Then, using the 632.8 nm and 532 nm lasers as references, the diffraction grating was rotated incrementally with respect to the lever until the difference between the readings was exactly 100.8 nm. It would have been even better to use more extreme wavelengths like a 457 nm DPSS or argon ion laser and 647 krypton ion laser, but these will have to do for now. Checking some other known wavelengths including 543.5, 594.1 and 611.9 nm (green, yellow, and orange HeNe lasers), as well as a 640 nm (errant laser line being produced by the red HeNe laser) showed them to be quite accurate.

    For determining laser wavelengths, a simple white card is all that is needed on the output. However, in its original application of detecting spectral signatures of plasma flames and such, a sensitive photodiode or PhotoMultiplier Tube (PMT) detector would be mounted beyond the exit slit.

    The overall performance (including wavelength precision and repeatability) using the dial, is now better than 0.5 nm, limited by very noticeable backlash in the multiturn dial mechanism. Originally, the 1200VIS also had a stepper motor (which I removed), PMT detector, and controller with data acquisition system (all whereabouts unknown). That was probably much more accurate but for my intended uses, this will be fine.

    Verity Instruments Model EP200 Monochromator/Detector

    These devices have been turning up on eBay lately in a variety of flavors, both manual (EP200Mmd) and motorized (EP200Msd). The latter is really only more useful if one of the mating controllers is also acquired - or one is built. Contructing one would not be that difficult with a microcontroller and stepper motor driver. Both versions include a micrometer adjustable monochromator, PhotoMultiplier Tube (PMT) detector, its high voltage power supply, and preamp, all built into a case about 7.5" x 7.2" x 2.6" inches. It runs on +/-15 VDC.

    As its name implies, the optical output of the EP200's monochromator is sent directly to a detector inside the unit which produces a voltage between 0 and +10 V proportional to light intensity at the selected wavelength. There is no exit port for light. The case is very well sealed against stray light - the light goes in but it never comes out. :)

    An annotated photo can be found in Verity EP200 Monochromator/Detector Organization. In addition to the parts being labeled, the beam paths for the a sample input, zeroth order (reflected input), and first order (useful) spectral lines are shown. The input here may be from the bore discharge of a HeNe laser tube. The wavelength micrometer is set for 587.6 nm in the photo thus selecting the intense yellow (helium) line which passes through the exit slit to the PMT. Of all the other lines shown (there are many more in the spectra of He and Ne), only the green one even makes it to the vicinity of the exit slit, but it's off to one side. Note that only the central ray for the incoming beam and each of its spectral components has been drawn on the photo. However, for a diffuse source like a glow discharge (as in this example), light bulb, or even an LED, the internal beam will expand to fill the large concave holographic diffraction grating (which provides high light gathering power). The bounce mirrors must also be larger than what might be expected due to the size of the internal beams. Only with a collimated laser, would the actual beam paths closely resemble the narrow ones shown.

    This model uses an actual machine shop type micrometer assembly to select wavelength. While not as convenient as a direct reading multidigit dial, rotation selectivity is better since there are only 25 nm per revolution compared to 100 nm for the typical dial. And, there is no backlash in the readout so the precision is better. However, when using narrow slits, the wobble in the micrometer becomes significant. The motorized version with its long shaft may be even better but it's useless without the matching controller because there is no readout.

    The sensitivity using the PMT is truly amazing. With the HV nearly as low (close to 0 V) as it can be, -200 VDC, and the gain turned nearly all the way down (10 percent), simply placing a small neon lamp power indicator near the entrance slit overloaded the preamp. Numerous lines in the neon lamp spectrum could be easily found. Since the PMT provides most of the amplification, the preamp really isn't that sensitive in the grand scheme of things. I measured 10 V out for 3.75 uA in at the 100 percent gain setting. Thus, replacing the PMT with a photodiode would only be useful for high intensity sources or low power lasers aimed directly into the entrance slit. But that would be such a waste of the EP200 since one of its main benefits is having the super high sensitivity detector built-in. In fact, when using the EP200 with coherent sources like lasers, there may be a small amount of ripple in the peak response versus wavelength and orientation of the instrument, presumably due to interference effects similar to visible speckle. This is not present with gas discharge sources. The EP200 is also polarization sensitive with a difference in response of about 2:1 for s and p polarized light. Thus, using the EP200's output to monitor the mode sweep of a random polarized HeNe laser may result in excessive amplitude fluctuations. But why would anyone want to do that with a monochromator?!

    Detailed information on this series of instruments can be found at Verity Instruments Monochromators. A description and photo of the interior is there as well as connector pinouts. One thing I did determine that isn't on the Web site is that there is a small slide switch on the PMT HV PCB inside the unit to select internal (adjustment pot) or external PMT high voltage control on pin 1 of the DB9 (+2 to +10 VDC for -200 to -1,000 VDC). The Web site simply mentions that internal or external HV control is selected at the time the order is placed. Some units (don't know if they would be newer or older than the ones I've seen) may only have a jumper. The lid can be removed without much risk of contamination as the box is not sealed. Just don't touch the diffraction grating as it cannot be cleaned.

    A sticker under the micrometer cover as well as another one inside the unit details the function of the 4 position DIP switch, which is to control the PMT preamp bandwidth as follows (0=Off):

         Position     Preamp
           4321      Bandwidth
           0000      530    Hz
           0001        5.3  Hz
           0010        0.53 Hz
           0100        0.24 Hz
           0110        0.17 Hz
           1000        0.11 Hz
           1010        0.09 Hz
           1100        0.08 Hz
           1110        0.07 Hz

    For manual control, only the first or second setting would probably be useful. Otherwise, the response speed is so slow that spectral features would be missed.

    The pinouts for the DB9 connector are available on the Verity Web site but here is a summary with the voltage polarities explicitly noted:

      Pin  Function
       1   HV Programming (if enabled, +2 to +10 V)
       2   -15 VDC (225 mA, polarity protected)
       3   +15 VDC (50 mA, polarity protected)
       4   Remote HV Monitor (-2 to -10 VDC)
       5   Signal Output (0 to +10 V)
       6   DC Offset (Zero voltage)
       7   Circuit Ground (Power and signal return)
       8   NC
       9   Circuit Ground (Power and signal return)

    I think the HV will actually go down (up?) to 0 V but probably isn't very useful much closer to 0 V than about -200 VDC. Note: Some documentation I've seen shows the HV Programming input being -2 to -10 V but all the units I have work fine with +2 to +10 V. I do not know for sure what the DC offset pin is used for. It is connected to the wiper of the ZERO pot and produces a small DC voltage (less tha 1 VDC) that varies with the pot. But, it may also be an input to allow the controller to set the DC offset remotely.

    The next test was to look at the spectral lines of the discharge in the bore of a 1 mW HeNe laser tube. Placing the bare tube next to the EP200 entrance slit and approximately level with it resulted in a large response. (HV of -300 VDC and gain of 50 percent.) However, this particular EP200 came with 500 um slits, too wide for my taste. :) While the strong lines could be seen, weaker ones adjacent to them were buried. :) So I modified the slits by using 5 minute Epoxy to glue a piece of a single edge razor blade to each, positioned to reduce the slit width to between 100 and 150 um - about as narrow as could be done by eye. (See below for more details on modifying the slit width.) This worked great in improving the resolution and allowed weak lines adjacent to strong ones separated by well under 1 nm to be resolved. However, since each slit was narrowed from one side only (that was enough of a pain in itself!), the wavelength calibration shifted by about 1 nm. (I must have guessed wrong since if they had been narrowed from the proper side relative to each-other, the wavelength calibration shouldn't have changed.) To remedy this, the micrometer mounting plate screws were loosened just enough to allow the micrometer to be nudged by about 1/1000th of an inch using the 585.25 nm and 587.56 nm yellow lines of the HeNe laser tube bore discharge as references, and confirmed with the 632.8 nm red lasing line.

    Those 585.25 nm and 587.56 nm lines are significant in that they are from neon and helium, respectively, and if reasonably similar in amplitude, the ratio of helium to neon is correct inside the tube. On the tube I tested, the intensity of the He line was about double the Ne line, indicating that the He:Ne ratio was high. That's better than the other way around. :) This thing makes such determination so easy. :)

    I've built a DC power supply and detector meter box (from junk parts of course!) to drive the EP200 heads conveniently. By default, its analog meter shows the detector output. However, by pressing a button, it will show the HV, which is adjusted via a 10-turn lockable knob if the EP200 is set up for external HV programming.

    I also constructed my own fiberoptic cable adapter (these are also available from Verity) which positions the tip of an SMA fiber connector at the entrance slit. The other end of the fiber would then have a focusing lens that can be positioned conveniently near the spectral source. Even though the amount of light coupled through the fiber into the monochronometer is generally quite small for anything but a laser, with the high sensitivity of the EP200, there is easily enough to take readings.

    A second EP200 I acquired had a bad photomultiplier tube but I replaced the original (Hamamatsu R928HA Hamamatsu R928 Datasheet) with an RCA 931A I had laying around. That seems to work OK, certainly well enough for my needs - it's way too sensitive! There are probably many other compatible PMTs. As long as the PMT has the same side-input, pinout, and fits the socket and housing, that's probably good enough for non-critical uses of the EP200.

    On a third EP200, the grating had fallen out of its mount. How the set screw loosened up will probably remain a mystery. Although there is a long scratch on the grating (possibly from the trauma, possibly from bouncing around during shipping, or possibly it was there even when new). Since the scratch just happens to be perpendicular to the rulings, it really doesn't cause any degradation in performance of any consequence. This unit worked fine after reinstalling the grating and calibration.

    On a forth EP200, the PMT had actually cracked - the main cover had a major ding in exactly the wrong place. Although no internal damage was visible, this must have whacked the PMT. It's otherwise in good condition awaiting a transplant.

    Yet another one had a bad PMT (very high dark current or something) AND a bad PMT transconductance preamp op-amp. The main problem replacing it was that the space is very limited. It's also the only IC that's not socketed. An LM741H in a TO-99 can is pin compatible, but with some creative lead bending, a jelly bean LM741CN DIP (which is much less expensive) can be made to fit. The original very high quality OPA111AH op-amp from Burr Brown (now TI) is quite expensive, but the 741 should work well enough for measuring wavelengths since long term stability is not required. Almost any other common op-amp could probably be shoe-horned in its place. Pin 2 is the input and pin 3 comes from the Zero Offset trim-pot on the front panel. The op-amp offset trim pins are not used.

    CAUTION: Do NOT attempt to measure the output of even a very low power laser by aiming it into the entrance slit. That will completely overload the system and may damage the photomultiplier tube. For a typical 1 mW laser, just arranging the beam to hit a white card positioned at 45 degrees near the entrance slit will provide more than enough signal with the PMT HV near the lower end of its useful range (say -250 to -300 VDC).

    CAUTION: Do NOT attempt to adjust calibration unless you have a source of a known unambiguous wavelength to use as a reference! It's way too easy to shift it too far to get back easily without one. A low power HeNe laser or green DPSS module shining on a white card would be suitable, but NOT a red pointer or other diode laser unless its peak wavelength is known exactly.

    If anyone has proper narrow slits or anything else related to these monochromators that they don't need, please contact me via the Sci.Electronics.Repair FAQ Email Links Page.

    Initial Testing and Adjustment of an EP200

    Checking one of these units for basic functionality is quite easy, requiring only a spectral test source like a neon lamp, HeNe laser tube, or other gas discharge lamp. Given the extremely high sensitivity of the EP200, using the light shining through the window from an outdoor high intensity sodium or mercury vapor street lamp may even be possible.

    Additional equipment that will be needed are regulated +15 VDC and -15 VDC power supplies (50 and 225 mA maximum, respectively), a 10K ohm pot and 5K ohm resistor to build a circuit for setting the high voltage (HV) if using external HV Programming, and a DC voltmeter capable of reading 10 V full scale. A flashlight may be useful as a quick test for confirming that the photomultiplier tube (PMT) detector is responding to light.

    Referring to the pinout for the DB9 interface connector, above, wire up the 15 VDC power supplies, Signal output, and HV programming (just in case). Circuit Ground is the common for everything including all voltage measurements. The COM test point is connected to Circuit Ground. The 5K resistor goes to +15 VDC and then to the top (clockwise end) of the pot, the wiper goes to the HV Programming input, and the bottom (counterclockwise end) goes to Circuit Ground. The value of the resistor and pot aren't critical as long as their ratio is 1:2 so that the HV Programming input is 0 to +10 VDC. Anything from 1K to 50K should be fine for the pot. For a permanent setup, a 10 turn pot may be desirable as the gain is quite sensitive to HV.

    There is no need to remove the large cover which encloses the optics and electronics of the EP200 except to flip the PMT HV selector switch if needed, or a major problem is found. For all tests, this cover should be in place with the screws tightly secured. Ample details on what's in there can be found on the Verity Web site or from the photo Verity EP200 Monochromator/Detector Organization.

    However, we all know that curiosity will get the better of you, so as long as it's open, check that nothing has fallen out, that the diffraction grating is secure in its clamped mount, it rotates freely against spring tension, and that the metal shroud surrounding the grating glass itself is pressed in as far as it will go. DO NOT touch the surface of the grating as it can't be cleaned without degrading its performance!!! DO NOT attempt to remove or adjust the grating in its mount - that affects focus and precise wavelength calibration (beyond what is discussed below). It's not supposed to be all the way in. If for some obscure reason it must be removed, use a depth gauge or other instrument to determine exactly how far in it should go and note which side is up (so that the blaze angle is correct when reinstalled). If there are any serious dents in the cover, confirm that there is no corresponding internal damage.

    CAUTION: On all the EP200s I've tested, simply removing and replacing the cover may alter wavelength calibration by a fraction of a nm. I assume that any slight change in stress on the baseplate deforms it enough to shift the wavelength peak. So, expect to have to do the basic wavelength calibration described below if you do go inside.

    Now that that's out of the way, block the entrance slit with a piece of black tape and double check your wiring before proceeding. It doesn't matter if the micrometer compartment cover plate (two thumbscrews) is installed for any of these tests.

    1. With your HV Programming pot fully counterclockwise (0 V), apply power and confirm that +15 VDC and -15 VDC are present at the proper DB9 connector pins.

    2. Connect your DC voltmeter or a multimeter between the HV monitor test point or pin 4 on the DB9 and Circuit Ground or COM.

      • If there is between -2 VDC and -10 VDC present, the internal HV pot is in control. Set it for -3.0 VDC corresponding to -300 VDC, which is close to the minimum useful HV.

      • If there is no voltage present, adjust your HV Programming pot to confirm that it is in control. If so, set it for about -3.0 VDC output on the HV test point (-300 VDC to the PMT). Note: According to some documentation I've seen, the HV Programming input should be negative. However, all the units I have work fine with positive HV Programming input. But if for some reason yours doesn't, flip the polarity and try again.

      If it is not possible to obtain any high voltage or the full -10 VDC (-1000 VDC to the PMT) on the HV test point with either the EP200 pot or external HV control, the HV power supply (a potted brick, an EMCO model 6858) or associated circuitry may be defective, or there may be a short in the PMT or its wiring.

      To change from internal to external HV programming or vice-versa, remove the main cover on the EP200 by taking out all the screws around its periphery and locate the HV select slide switch near the edge of the electronics PCB between the PMT housing and slit mounting plate. Flip it to the opposite position (toward the control/connector panel to select the EP200 HV pot).

      However, I highly recommend using external HV control as it is more flexible and convenient allowing for very quick and easy control of PMT gain, and won't wear out the internal pot! This is very likely the default for most of these units.

    3. Once the presence of HV has been confirmed and it is set for about -300 VDC, connect the Signal test point or Signal Output (pin 5 on the DB9) to your voltmeter. The reading could be anywhere from negative to off scale at this point.

    4. Set the 4 DIP switches to the OFF position to select maximum signal bandwidth.

    5. Press and hold the "GAIN" pushbutton and adjust the GAIN pot for approximately +5 VDC (on the Signal test point or pin 5 of the DB9) which corresponds to 50 percent gain. (The useful range is from 0.1 to 10.)

    6. With no light input, it should be possible to set the ZERO pot to produce near 0 VDC on the meter, though it may be quite touchy and like to go way negative. Keep it just positive if it won't cooperate by going to exactly 0 VDC. If the Signal Output remains pegged beyond +10 V and won't settle down in positive territory, try turning down the HV to 0 VDC. If the meter can now be zeroed but pegs to over +10 V with almost any HV, the PMT is probably defective. Else, it may be a circuit problem. (Some dark current which varies with HV may be normal but with -200 to -300 VDC, it should be negligible.)

    7. Set the wavelength micrometer for approximately 600 nm. (If the micrometer doesn't turn easily, check the black locking ring near its base, counterclockwise a fraction of a turn to unlock.) Each turn of the micrometer corresponds to 25 nm with each division being 1 nm. Multiply the direct reading by 100 to obtain the actual wavelength.

    8. Shine a flashlight (if available) head-on into the entrance slit. With a broadband source, there should be a response on the meter regardless of wavelength. However, some LED flashlights may not he broadband. So, if using an LED flashlight, rotate the micrometer through the visible range and look for a response. If this test is successful, the EP200 is likely fully functional but may need to be calibrated for wavelength.

    Before proceeding with wavelength checks or calibration, the EP200 main cover should be securely screwed down including the locking screw of the input slit since a slight shift in wavelength may occur when this is done. The EP200 should be placed on a solid surface so it can't wobble as any position or orientation change may result in a variation in the signal output and confusion as to where a peak is located.

    1. Set up your spectral test source in front of the entrance slit, or aim the monochromator to the source. the f/3.5 acceptance of the system is low enough that precise alignment isn't needed but arranging it to be fairly close to in-line with the slit will be best.

      CAUTION: DO NOT shine a laser beam directly into the entrance slit. Almost any laser - even a half dead laser pointer - is way too powerful to be used directly. However, it is safe to shine a low power laser on a white card placed near the entrance slit. More on this below.

    2. Refer to the appropriate spectral charts to determine what wavelength emission lines should be present with your source. For example, three relevant lines for the bore discharge of a HeNe laser are: 585.25 nm (neon yellow), 587.56 nm (helium yellow), and 640.2 nm (neon red). The normal HeNe red 632.8 nm is also present, but quite weak. There are dozens of other lines present but the three cited are particularly strong. If these can be located exactly where they should be, you're done. If they are there but not exactly where they should be, you're almost done. :) When turning the micrometer, avoid applying side-pressure as this will result in a wavelength shift and hysteresis.

      Note that with the typical 500 um slits, it is difficult to resolve the 585.25 nm and 587.56 nm lines as their spacing is less than the spec'd resolution of the EP200. But it's no problem with 200 um or narrower slits.

    Other lines that can easily be checked in a HeNe laser tube bore discharge include: 447.1 nm, 471.3 nm, 492.2 nm, 501.6 nm, 587.6 nm, and 667.8 nm from helium, and 540.1 nm, 585.2 nm, 588.2 nm, 603.0 nm, 607.4 nm, 616.4 nm, 621.7 nm, 626.6 nm, 633.4 nm, 638.3 nm, 640.2 nm, 650.6 nm, 659.9 nm, 692.9 nm, and 703.2 nm from neon. Note that the lasing wavelength of 632.8 nm is not among these medium to strong lines.

    If you're not using a HeNe laser tube but a source like a gas discharge spectral lamp containing some other gas(es), spectra for varioue elements and compounds can be easily found by searching the Web. The NIST Atomic Spectra Database has an applet which will generate a table or plot of more spectral lines than you could ever want.

    If the spectral lines located above are at the proper locations on the micrometer or within the uncertainly due to what minimal backlash there is, you're done. Otherwise, adjustment of the micrometer assembly will be required. This is best done with a HeNe laser or other source with a single spectral line. Otherwise, it may be difficult to know which line you're seeing. The use of such a laser is assumed below. Note that a red diode laser pointer is NOT suitable as its wavelength is not known precisely, nor is the line very narrow. However, a green DPSS laser pointer at 532 nm is perfectly fine.

    1. Set up a low power HeNe laser with its beam shining on a white card placed at a 45 degree angle next to the entrance slit, approximately centered and level with it. CAUTION: Do NOT shine the laser into the entrance slit as it will be way too powerful. This would grossly overload the system and may damage the PMT.

    2. Attempt to locate the 632.8 nm wavelength (assuming a red HeNe laser) by adjusting the micrometer. Unless someone has totally disassembled the EP200 at some point, it should appear somewhere within the adjustment range of the micrometer, probably quite close to where it should be. A HV setting of -300 V and gain setting of 5 V (50 percent) should result in adequate response for a 1 or 2 mW laser. You can try increasing the HV slightly if the response is very weak so that the reading reaches at least half scale.

    3. If the wavelength location is too low, the micrometer must be moved toward the entrance slit side of the unit. If too high, it needs to move away. The return spring on the grating tends to push it away but with the cover in place and pressing on the seals, some persuasion may be needed.

    The following three steps can probably be skipped if the wavelength is within a few nm of the correct location.

    1. Use a right-angle hex wrench to loosen both of the micrometer plate mounting screws inside the micrometer compartment just so they are barely touching the plate. The return spring will move the micrometer assembly away from the entrance slit side of the EP200 to its stop.

    2. Set the micrometer to 632.8 nm exactly.

    3. Use a narrow screwdriver or other suitable tool as a lever vertically between the micrometer mounting plate and EP200 main cover. Slowly move the micrometer assembly toward the entrance slit side of the unit (pull the tool toward the control/connector panel) while monitoring the Signal Output test point or connector pin. When the position coincides with 632.8 nm, the reading will jump. With care, it should be possible to lever the plate back and forth very slowly around this point stopping exactly at peak output so the mounting screws can be tightened. 1/1000th of an inch movement corresponds to 1 nm. With care, a precision of 1/10,000th of an inch is possible. This sounds more daunting than it really is.

    1. Check the peak setting. If the wavelength is only slightly off, it may be possible to loosen the micrometer mounting screws so they are just snug enough to hold the plate in position. Then, gently nudge the plate in the appropriate direction to center the peak at the 632.8 nm setting.

    2. If possible, check some of the lines in the HeNe tube discharge spectrum to confirm calibration. If fine adjustment is needed, it's better to use one of those because there are negligible interference effects to confuse the peak location, as there are with the coherent laser output, which may be somewhat touchy and dependent on input position and orientation.

    3. If another laser is available with a much different wavelength (like 532 nm), it would be worth checking that it too is now lined up correctly with the correct micrometer setting. Should this for some reason be off by more than a fraction of a nm, the relation of the grating shaft to its lever arm has changed. Adjustment of that is a more involved process reserved for the advanced course. :) But, no one should have touched and unless the unit was terribly abused, it shouldn't have changed on its own.

    This description probably makes the procedure sound like it will take all day. Wavelength calibration should require only a few minutes unless you are an absolute perfectionist, in which case it will take forever. :)

    Narrowing the Slits in an EP200

    For end-point detection in whatever processes these instruments normally monitor, the most common 500 um slit is perfectly adequate. But for looking at closely spaced spectral lines, a narrower slit is almost essential. Although the slits width isn't adjustable on the EP200, modifying the slits to be 100 to 200 um is relatively easy. The first procedure is reversible:

    1. Remove both slit assemblies (slit plate in plastic holder). The entrance slit is locked into position by the cover screw next to it. The slit assembly can then be pulled out. The exit slit is under the circular cover plate held in place by 4 small screws on the bottom of the EP200. Once the cover plate is removed, the slit assembly can be pulled out. The entrance and exit slit assemblies are identical.

    2. Gently confirm that the thin metal slit plate itself is secure in each plastic holder. On some units, I've found that the glue has weakened and almost any pressure results in the slit plate popping out. If loose, fallen out, or in doubt, use a tiny dab of 5 minute Epoxy to secure it. Make sure it is parallel to the edge with the shiny side should be down

    3. Prepare a sacrificial single edge razor blade. It doesn't have to be new but the edge should be free of any nicks or dings. Use a pair of pliers to break off pieces that will fit in the hole in the plastic behind the slit.

    4. Narrowing the slits from both sides is not essential but is desirable to avoid a wavelength shift depending on where an small diameter input source is located - above, below, or centered on the slit. This is because the blades on each side will be at the different distances from the diffraction grating. This may result in an error of a fraction of a nm depending on source location, though resolution will not be affected significantly. But, to prevent a fixed wavelength calibration shift if doing only one-side surgery, the entrance slit should be narrowed from the opposite side as the exit slit.

    5. Put a tiny dab of 5 minute Epoxy in the recess near the edge on the appropriate side and use tweezers to place the bit of razor blade in position. I just eyeballed the slit width but using something more sophisticated to measure it is permissible. :)

    6. Check the status of each slit as the adhesive cures to make sure it hasn't shifted position.

    7. Reinstall the slits and check calibration. Since the slits are much narrower than before, increasing the HV and/or gain may be necessary. If the calibration did change significantly, use the procedure in the previous section to correct it. With the narrower slits, calibration error will be more noticeable.

    Where it is known that going back to the wider slit will never be desired, then the following may result in better performance:

    1. Carefully pop the original metal slit insert off of the plastic holder.

    2. Use a sharp pair of scissors to split the slit into two parts without harming the slit edges.

    3. File each side to the two halves can be mounted at the appropriate closer distance.

    4. Use 5 minute Epoxy to reattach them to the plastic holder confirming the correct spacing and adjusting as needed before the adhesive sets.

    The focus is more critical with narrower slits. If the diffraction grating's position in its mounting clamp isn't exactly right, resolution will suffer. If it's never been touched, then don't touch it now as it's unlikely to have moved on its own, and wavelength calibration may be affected by position. But, if you've been fiddling with the grating, now is the time to adjust focus using a diverging beam (laser or LED) as the input. Since this has to be done with the cover removed, the source will need to be bright with gain set low enough so the ambient light doesn't overwhelm the system.

    Monolight Model 6100 Scanning Monochromator

    These are based on a "head" unit which is a compact Czerny-Turner monochromator where the diffraction grating rotates continuously on a motor shaft. With a suitable controller, an optical spectrum over a wide range (e.g., from 300 to 1,100 nm) can be acquired in about 85 ms. Thus one application is in a fast, though not particularly high resolution, optical spectrum analyzer. The resolution is about 1.4 nm for my particular unit.

    Some general info used to be found at Macam Photonics: Monolight Monochromators. But now it simply says to contact the company. And it may be that Gamma Scientific and/or EG&G (now part of EXCELITAS), sold these systems under their name.

    A montage of the head unit is shown in Monolight 6100 Scanning Monochromator. It includes the motor speed control and a transimpedance amplifier (but not the photodiode) with BNC input and output. A trigger output signal is provided on another BNC for syncing the scope. There may be several other signals as yet unidentified on a 7 pin DIN connector and a separate jack for 15 VAC power. There are also 4 trim-pots, associated with the motor driver, as well as a gain adjust for the preamp.

    Simply applying 15 VAC to the power jack will make the motor spin, though the stability on the unit I have isn't that great. Although motor speed is regulated based on a voltage-to-frequency function from the 36 pulse/rotation optical encoder disc, it's likely intended to be more precisely phase locked to a crystal reference by the controller. Or, perhaps as someone suggested, the motor is high mileage with worn bearings or dirty brushes or something. While the 6100 has a built-in trans-impedance (current to voltage) preamp for a photodiode, the detector itself must be mounted externally. If displaying the spectrum of a laser, the photodiode can be almost anything as long as it's relatively small area (low capacitance). I used one from a barcode scanner for testing, just positioning it near the output slit. The 6100 provides a trigger signal that can sync an oscilloscope which can then be used to display the spectrum in leu of the controller and data acquisition system. Although a digital storage scope is desirable, my Tek 465B worked just fine in showing the 3 lines of my funny yellow-orange-orange PMS/REO LHYR-0100M HeNe laser head. However, the resolution is orders of magnitude poorer than would be required to view the individual longitudinal modes of any HeNe laser, as wtih a Scanning Fabry-Perot Interferometer (SFPI). The main problem was excessive sensitivity. My photodiode detector had no gain control, being, well, just a photodiode. Perhaps the intended detectors have an adjustment. I had to use neutral density filters to reduce the intensity to a level that didn't saturate the preamp.

    If anyone has more specific information including schematics for the 6105 head unit, or has Monolight hardware they'd be willing to contribute to the cause, please contact me via the Sci.Electronics.Repair FAQ Email Links Page. I had to add a pot to adjust the pullup resistance on the optical encoder on my sample as the signal was about half the amplitude it should have been leading to some peculiar behavior. Although it appears as though this pullup is a "select on final test" resistor, being in a two terminal header rather than being soldered into the PCB, I assume that somehow the output has gone down due to a weak LED or other problem in the opto-detector reading the disc pulses by reflection, similar to the reel rotation sensors in some VCRs. It is probably a standard 3 terminal device (LED and photodiode) which could be replaced but I can't read the part number without disassembling the unit. And I'm not THAT curious. :)

    Rees 200 Series Laser Spectrum Analyzers

    These may also be labeled "ist Laser Spectrum Analyzer" and possibly sold through Heraeus Nobelight Analytics, Ltd., but Rees is apparently the original manufacturer. Like the Monolight unit above, these are based on compact a Czerny-Turner monochromator using a DC motor driven spinning grating. However, the ist/Rees optical head is about half the size of the Monolight unit, and includes the detector whereas the Monolight housed a preamp with BNC input and output connectors, but not the actual detector, only a slit with mount for a detector. On the other hand, there are no other electronics in the optical head, so motor speed regulation is done externally.

    The input can be an optical fiber (FC/PC) but a free-space beam can simply be directed into the fiber connector. The input and output slits appear to be more like pinholes, so there is more loss but they work well enough. Focusing the beam with a 0.5 to 1.0 inch focal length positive lens will greatly increase the power that gets through the pinhole. A separate control box provides power and includes an A/D with digital storage for for the spectrum data, and a D/A to drive a normal oscilloscope (analog or digital). There are three BNC outputs which are "Signal" (analog spectrum), Markers (ticks at 0.2, 1, and 10 nm, as well as a blinking up and down tick that may be positioned via front panel buttons on any integer nm location within the scan range), and Trigger to sync the scope. Two basic versions were available. The model 201 for VIS (350 to 1,100 nm) and model 202 for NIR (750 to 1,650 nm). A Web search on "Rees E200 Laser Spectrum Analyzer" will return a spec sheet with more info.

    A typical system is shown in Rees/ist Series 200 Laser Spectrum Analyzer System. I had to construct the colorful cable since it was missing from the unit I acquired - perhaps "liberated" for some other purpose - and I figured rainbow spectral colored wire would be appropriate and the wire was available. ;-) The two buttons on the left set the Marker wavelength in increments of 1 nm. The three buttons on the right are for HI GAIN (or low gain for the spectrum input signal), HOLD (freezes the display like a DSO), and PRE TRIG which offsets the Marker position relative to the scope trigger. The GAIN knob adjusts signal amplitude while the SPAN knob adjusts the wavelength range on the display. However, its range isn't that large so a combination of the SPAN control along with the scope's horizontal sweep rate are required to zoom in and select any given portion of the spectrum.

    The inside of the head unit is shown in Interior of Rees/ist Series 200 Laser Spectrum Analyzer Optical Head. The grating and 36 slot encoder wheel are clearly visible. One of the cutouts in the encoder wheel is smaller and used as the reference. The DC motor has very good bearings and the rotating assembly has been dynamically balanced so there virtually no vibration. :) The PCB under the grating/encoder wheel is for the signal preamp which takes in +/-12 VDC and outputs the spectrum signal. It is not known whether there is a gain control line, if two different amplitude output signals are provided, or gain changing is done in the control box. Connections to the motor and photo-interrupter are routed from the 8 pin Mini-DIN connector via the PCB but do not use any of its circuitry.

    A reverse-engineered schematic of the head has been provided by Oscar Spierings and may be found at Rees Model 200 Optical Head Schematic.

    For reference, the laser head to controller cable has a male 8 pin Mini-Din at both ends with all pins wires 1:1 as well as the shield (which appears to carry the motor current). CAUTION: Cables labeled "Apple Printer" or similar will NOT work and may cause bad things to happen as they have some pins swapped - NOT wired 1:1. Specifically, pins 3 and 5 are swapped and 6 and 8 are swapped. The major issue is that +/-12 VDC power to the head on pins 6 and 8 would have reverse polarity, which can't be good! The preamp signals on pins 3 and/or 5 would also be swapped.

    The only thing I had to do to tune up the optical head was to slightly reposition the pin-hole to line up with the FC fiber.

    The first test was with a low power fiber-coupled red (633 nm) HeNe and that came right up at 632 nm. Fiddling with the "Calibrate" switches at the rear moved it to 633 nm, where it belongs! It's possible some additional adjustment of the switches will be required eventually so wavelengths line up at both ends of the 350 to 1,100 nm range.

    Although marketed for use with ultra-fast lasers, this would seem perfect for multi-line HeNes. And what better laser to use for testing than one of the strange PMS/REO HeNe lasers. See specifically The PMS/REO External Resonator Particle Counter HeNe Laser. Up until now, only 7 (!!) lines were seen from this laser, but by using the Rees LSA, an 8th has appeared:

    Nowadays (2015), a fiber-optic spectrometer like the USB-2000 from Ocean Optics may serve a similar function and has the advantage of being a more compact overall system sending data directly to a computer. The Rees/ist may still have an edge in resolution since its monochromator generates an analog signal that can be digitized at a high rate, not limited by CCD pitch. So subtle changes in position and/or spectral line width may be easier to see. But it may have simply come along at the wrong time to really become popular.

  • Back to Laser Instruments and Applications Sub-Table of Contents.

    Optical Wavelength Meters

    Principles of Operation

    While instruments like monochromators and optical spectrum analyzers are capable of determining the wavelength of light sources from light bulbs to lasers, their accuracy depends on the precision of multiple mechanical parts and the quality of the initial calibration. This is because they use what might be termed an indirect method of analysis - typically a diffraction grating moved by a precision mechanism. If there is any real-time reference, it is likely only at a single wavelength so there could be significant error at wavelengths not close to it.

    Where the source to be measured is broadband or has multiple spectral lines, such techniques are generally the easiest and fastest (but see below). However, where a single wavelength CW laser's output needs to be determined very precisely, alternative methods are generally used. (Wavelength meters capable of reading pulsed lasers also exist. See the section: Pulsed Wavemeters.) Instruments of this type may have an accuracy and resolution that is orders of magnitude higher than a monochromator or optical spectrum analyzer.

    The following description is for one common approach and the one used in the Burleigh WA-20 a typical older model dating to the early 1980s. For modern wavemeters using a different approach, see the section: Pulsed Wavemeters.

    The basic wavelength meter or "wavemeter" compares the unknown input with a reference laser by counting fringes for both sources simultaneously in a Michelson interferometer where the path length difference is varied periodically by a motor-driven mechanism. The unknown wavelength (or frequency) is then related to the reference by the ratio of the number of fringes for each during a fixed period chosen to be near where the path length difference is small (to minimize the effects of the coherence length of the unknown laser). Where the desired display is in wavelength (e.g., nm), the reference wavelength is divided by N/No (where N is the number of fringes for the unknown and No is the number of fringes for the reference). Where the desired display is in frequency (cm-1), the reference frequency is multiplied by N/No. (Frequency here is the meaning used by spectroscopy-types: 1 cm/wavelength, or wave number.) The division and multiplication can be easily accomplished with digital counters and simple control logic, similar to that in any vanilla-flavored electronic counter/timer. Phase-Locked Loops (PLLs) will generally be used for both the reference and unknown detector signals to multiply the fringe counts and thus the resolution.

    A red HeNe laser is generally used for the reference laser. For the Burleigh WA-20, an Aerotech OEM1P laser head with an Aerotech brick power supply was found in one unit I checked. (I don't know if all WA-20s used the same laser.) The brick ran from 12 VDC (it's labeled 10-14 VDC) and had a fixed output of 4 mA with a compliance range of 1,200-2,000 V. The specs of the OEM1P laser head (from an old Aerotech brochure) are:

    I doubt that any of these specifications are really critical. A rated output power between 0.5 and 2 mW should be acceptable. Where a replacement laser head is not electrically compatible, another HeNe laser power supply can be easily substituted. Even 0.5 mW is much more than actually required. The only "adjustment" is a linear polarizer that can be rotated to reduce REF power. Even the fact that the OEM1P is linearly polarized probably doesn't make a lot of difference, though a random polarized laser might result in a larger variation of REF signal amplitude during mode sweep and the polarizer/attenuator won't work so an variable ND filter would need to be substituted. To avoid this, a linear polarizer can be added at 45 degrees to the polarization axes of the laser resulting in a beam indistinguishable from that of a random polarized laser at the expense of 50 percent of the output power. (More will be lost with a sheet polarizer, but there will still be 30 or 40 percent of the original power.) With so little power actually needed, this should not be a problem.

    However, one thing that would make a very slight difference is the isotope ratio of neon in the gas-fill, which can shift the peak of the neon gain curve by almost 1 GHz or 1.4 pm, more than 1 count in the LSD of the WA-20 display. This in addition to the contribution of REF frequency change due to mode sweep. The hardest part may be in mounting a non-exact replacement laser - there are a pair of narrowed sections on the Aerotech laser head which mate with brackets in the WA-20! (Melles Griot may have a suitable laser head that would be a drop in replacement.)

    Even if not stabilized (but with a known gas-fill), its wavelength will be accurate to better than 1 part in 106. A stabilized HeNe laser locked to the gain curve can be a couple orders of magnitude better and an iodine line stabilized HeNe laser, even better. (Some later Burleigh/EXFO wavemeters do incorporate stabilized HeNe lasers for the reference.) Another source of error is the change in the refractive index of air over the typical wavelength range of at least 400 to 1,000 nm. For this reason, for better accuracy, some wavelength meters put the interferometer optics in a vacuum chamber (less than 10 Torr). However, simply providing a lookup table for wavelength correction would be nearly as effective and much less of an implementation issue, though the actual pressure and temperature have to be taken into consideration.

    The performance of this fundamentally simple device is quite amazing. The resolution and accuracy of the Burleigh WA-20, which is one of the earliest commercial wavemeters, is better than 1 part in 106 (less than 0.001 nm or 1 picometer over the measurement range of 400 to 1,000 nm!). No routine calibration is required. While degradation in alignment is possible, the effect will be to increase the power level needed to take a reading but will not noticeably effect the resolution and accuracy. As long as the instrument is happy with the signal levels, the resulting display should be accurate. The most common problem may be a bad belt between the motor and interferometer drive! And an elastic band or tape player belt will work just fine, thank you. :-) (But really old belts may decay into a gooey black mess.)

    The successors to the WA-20 are the Burleigh/EXFO WA-1000 and WA-1500. They are virtually identical mechanically to the WA-20 but keep the interferometer at atmospheric pressure instead of in a vacuum. Thus their weight is much lower and similar that of a WA-10. Software correction for the non-linearity of the index of refraction of air is then used with inputs from pressure and temperature sensors. The WA-1000 uses the same reference HeNe laser as the WA-10 and WA-20 and has similar accuracy. The WA-1500 uses a stabilized HeNe laser providing for better accuracy - +/-0.2 picometer compared to +/-1 pm for the WA-20. Both are microcprocessor-controlled resulting in a There is more information on the WA-1000 and WA-1500 in subsequent sections.

    A note to those out there who believe in running wavemeters continuously because they believe this results in better performance: There is no need. The performance of wavemeters like the WA-20 or WA-1000 is essentially the same as soon as the reference laser turns on as if run for a year. There is nothing to warm up that matters. So, save the reference laser and mechanics (where appropriate) and turn your wavemeters off when not being used! This also applies to those like the WA-1500 that incorporate stabilized HeNe lasers for the reference except that a warmup period of a few minutes is required for the laser to lock. But after that, the wavelength drifts at most by much less than 1 count in the last digit (0.1 picometer).

    It should be noted that this implementation of a wavemeter is a subset of a more general technique called Fourier Transformer Spectroscopy which is capable of dealing with arbitrary spectra. (See, for example: World of Physics: Fourier Transform Spectrometer.) Rather than simply counting fringes, the Fourier transform is taken of the fringe waveform during one or more scans of the path length difference. For a single spectral peak as with a CW single frequency laser, the FT is a single peak. For a source with multiple peaks, the fringe pattern becomes visually complex, but the Fourier Transform will be the desired spectrum. This approach is also used in some wavemeters that can deal with multi-line laser input. For example, the WA-650 is an add-on that converts the WA-1500 or WA-1000 into an optical spectrum analyzer by Fourier Transform processing of the fringe pattern. And later models called "multimavelength meters" incorporate FFTs as standard equipment. :)

    In fact, it should be possible to process signals from the back of almost any wavemeter which has a built-in reference laser to use it as an optical spectrum analyzer. The interference signal for the unknown source, the interference signal for the reference laser, and a scan position sync pulse are required. This would be very simple if the scan was linear. But with wavemeters using a motor-driven scan like the WA-20 or WA-1500, the speed and thus fringe frequency isn't perfectly constant and this would totally mess up the FFT. It should be possible to correct it as long as a reference laser signal is available. The details are left as an exercise for the student. In fact, this would make a nice term project in DSP course. :)

    But, the beauty of the basic single wavelength wavemeter is at least in part due to the simplicity in terms of its principles of operation, mechanical construction, and electronics.

    A Scanning Fabry-Perot Interferometer (SFPI) may have better resolution, it typically doesn't have very good accuracy or stability with respect to absolute wavelength or frequency unless additional techniques are used, adding to complexity.

    While minor enhancements like the use of a voice coil magnetic drive instead of a motor can improve the speed and reduce the size of the Michelson interferometer-based wavemeter, higher performance instruments may use something called a Fitzeau interferometer with no moving parts. Multiple wedged etalons generate fringe patterns which depend on the source wavelength. These are captured via CCD arrays and analyzed in software. These instruments can deal with pulsed lasers and have much faster dipslay rates (100s of Hz or more compared to a few Hz for motor driven interferometers) and even more immune to alignment problems.

    Tuneup of a Burleigh WA-20 Wavemeter

    The Burleigh WA-20 is a typical older wavelength meter that uses a motor-driven moving interferometer mirror and fringe counting to determine wavelength (in um) or inverse frequency (in cm-1) of CW lasers between 0.4 and 1.0 um for the visible, which may be extended to 4.0 um with the IR option (which substitutes a different beamsplitter and detector). This model dates from the early 1980s, though the specific unit I'm working on has a manufacturing date of 1995. There was also a WA-10, with the only difference being that while the WA-20 maintains the entire interferometer inside a chamber that can be evacuated to below 10 Torr to for better accuracy, the WA-10 simply has a dust cover. The reason that a vacuum is beneficial is that there is a small non-linear depedence of the index of refraction of air on wavelength so a measurement of a laser with a wavelength far away from the 633 nm reference might see an error of as much as 3 parts in 106 in air. If the temperature, pressure, and humidity are known, a lookup table can be used to eliminate the error, but that's probably more trouble than it is worth. However, the nice thing about the WA-10 is that it's a lot easier to work on it when doing alignment not having to deal with the vacuum chamber and vacuum-tight covers over the mirrors! And it weighs less. :)

    A red HeNe laser (polarized but not stabilized) is used as both the wavelength reference, and to provide a "tracer" beam to facilitate alignment of the unknown laser to the input of the WA-20. It passes through the same interferometer optics, but more-or-less in reverse. Thus, alignment of the reference laser beam is sufficient to guarantee alignment of the entire system.

    Neither the WA-10 or WA-20 are manufactured or supported now, but the modern replacements, the WA-1500 (with stabilized HeNe laser reference) and WA-1000 (without) are substantially similar in design, though they both operate without a vacuum, but have pressure and temperature sensors using software correction for the non-linear index of refraction of air. With its stabilized reference laser, the WA-1500 has somewhat better accuracy than the WA-20 while the WA-1000 is similar. They both have better sensitivity (20 uW instead of 100 uW). For more information, see the section: Burleigh WA-1000 and WA-1500 Wavemeters

    The WA-20 I had on loan for testing and adjustment was so misligned when I received it that the tracer beam was partially cut off and less than 1/10th the intensity it should have been, and the "Ref Error" light was flashing due to low signal level. Yet, despite these problems, it was still able to measure the wavelength of an external red HeNe laser to the expected accuracy, though higher than spec'd power was required. That is, after the drive belt which had fallen off was put back in place. :) Consider yourself lucky if the belt simply falls off; on many of these, the rubber has decomposed and turned into a gooey black mess. A common elastic band will suffice until a proper drive belt can be obtained.

    There is an alignment procedure in the user manual that is strightforeward, if somewhat tedious. It uses the reference laser entirely to align the mirrors and beamsplitter in relation to the input aperture. Once this is done, the unknown laser input is also automagically aligned since it uses the same optics. Once this procedure was complete, the system was able to read the red HeNe laser as well as a highly attenuated C315M green (532 nm) laser at power levels below the spec'd minimum of 100 uW.

    Additional items that still require attention are obtaining a replacement drive belt and replacing the O-ring in the motor vacuum seal feed-through since it's leaking at too high a rate. However, except for being run at 1 atm instead of a vacuum and accepting the slight reduction in accuracy, it's now in good shape.

    And, the Power indicator uses a strange 60 V, 20 mA incandescent lamp, and of course is likely to be burnt out on a well-used WA-20. I replaced it with a high brightness LED, soldered to the slide contacts of the original lamp along with a 1N4148 across it for reverse polarity protection. An additional 3.9K, 2 W resistor and 1N4007 diode were added in series with the original 3.9K resistor in series with the one already there. An alternative that wastes less power would be to tap off one of the 5 VDC or 12 VDC supplies but this makes it easy to restore the original arrangement if desired.

    Burleigh WA-1000 and WA-1500 Wavemeters

    The WA-1000 and WA-1500 are more modern Michelson interferometer-based instruments. They are both smaller and lighter than the WA-20, and are microprocessor-controlled. The use of environmental sensors for wavelength compensation eliminates the need for the massive vacuum chamber of the WA-20. The primary functional difference between the two models is that the WA-1000 uses a normal 1 mW polarized HeNe laser head for the reference laser while the WA-1500 uses a frequency stabilized HeNe laser. The use of the stabilized reference laser allows for approximately a 5X increase accuracy and a 10X increase in resolution. But one cannot simply install a stabilized HeNe in the WA-1100 and obtain all the features of a WA-1600 as there are other differences. However, the result would be less variation in readout value even if the number of digits of the display doesn't change.

    The WA-1000/1500 come in three wavelength ranges that are documented: VIS (400 to 1100 nm), NIR (600 to 1800 nm, and IR (1500 to 4000 nm). There is also a UV version but its specifications are unknown. It's possible to convert from one version to another in the field by at most changing the beam-splitter (the only transmission optic in the interferometer beam paths) and signal photodiode. Except for the IR version, all accept either a free-space beam fed in from the side (similar to the WA-10/20) or a fiber-coupled input fed in via an FC/PC connector on the front panel. (There's really no reason why an appropriate optical fiber could not be used with the IR version, so I don't know why that is not spec'd. The WA-1000-IR I tested had the fiber port, but that may have been a conversion from VIS or NIR.) The physical layout of the interferometer is similar to that of the WA-10 but thankfully, a higher quality motor and belt are used to drive the moving retroreflectors, so belt replacement and belt goo cleanup are probably no longer a part of regular maintenance. :) The motor-driven attenuator is internal and can be set for automatic or manual (using front panel buttons). Complete specifications for both instruments are in the operation manual, easily found on-line by searching for "Burleigh WA-1000 manual" or the later "EXFO WA-1000 manual" (which for the most part is identical except for some formatting enhancements).

    Burleigh WA-1100 and WA-1600 Wavemeters

    The WA-1100 and WA-1600 are more modern Michelson interferometer-based instruments. They are both smaller (shorter) and lighter, and like the WA-1000/1500, are microprocessor-controlled along with environmental sensors providing for wavelength compensation and eliminating the need for the massive vacuum chamber of the WA-20. The primary functional difference between the two models is that the WA-1100 uses a normal 1 mW polarized HeNe laser head for the reference laser while the WA-1600 uses a frequency stabilized HeNe laser. The use of the stabilized reference laser allows for approximately a 5X increase accuracy and a 10X increase in resolution. However, one cannot simply install a stabilized HeNe in the WA-1100 to convert it to a WA-1600 as there are other differences including the firmware and possibly the actual interferometer as the update rate on the WA-1100 is 10/second compared to only 1/second for the WA-1600. The WA-1100 and WA-1600 appear in some ways to be cost/size/weight-reduced versions of the WA-1000 and WA-1500. The input to both of these instruments is fiber-coupled, with an FC/APC as the default connector. (There is no free-space option.) Complete specifications for both instruments are in the operation manual, easily found on-line by searching for "Burleigh WA-1100 manual".

    There are several subsystems inside the unit as shown in Burleigh WA1100 Wavemeter Interior View: DC power supply (upper right), reference laser with its HeNe laser power supply brick (right), interferometer optics with processing PCB on top (covered, bottom), microcontroller PCB (upper left), display PCB (attached to front panel, hidden), and a little motor driver PCB (left).

    The interferometer in the WA-1100 has been greatly reduced in size and complexity. (I have not seen the internal organs of a WA-1600.) It uses a single retro-reflector (cube-corner) instead of a dual-sided one as in the WA-10 and WA-20. It's on a much much smaller motor driven linear slide, and the remaining optics are installed in a compact single-piece precision milled aluminum structure. There is still a rubber belt but hopefully, it won't decay and require replacement like those in the WA-10 and WA-20 as gaining access to the belt appears to be more involved, buried beneath the optics. A single PCB (the Sensor Board) mounts above the interferometer as shown in Burleigh WA1100 Wavemeter Sensor Board. There are 3 photodiodes (PDs) facing downward into it, one each for input laser power, reference fringe detector, and signal fringe detector. Repair would be more difficult as (1) most parts on the PCB are SMT and (2) it's not possible to get to the optics with the PCB (and PDs) in place. However, like the WavemeterJr, there really is only a single mirror that requires alignment and those adjustments are accessible without disassembly. That is accessible from the right side of the interferometer without any further disassembly and consists of 2 kinematic adjustments screws and locking set-screws near them. Peaking the reference fringe amplitude should also result in optimal signal alignment. Everything else is glued in place.

    The reference lasers are typically OEM versions of the Melles Griot 05-LHP-491 for the WA-1100 and 05-STP-910 for the WA-1600, fiber-coupled directly to the interferometer. The fibers are not connectorized, only terminated and glued. The input fiber needs to be able to pass wavelengths up to 1,700 nm with minimal losses, so it is assumed to have a rather large core. However, the reference fiber may be the same and thus multi-mode at 633 nm, and it is sensitive to power changes with even modest bending radii or changes in routing. There is a beam sampler and silicon photodiode that monitors laser power before fiber coupling, which may be read from the front panel, so it won't detect problems with the fiber. There is nothing special about these particular model reference lasers, but if the beam diameter and divergence of a replacement are not close to the original, the coupling efficiency may be reduced unless the focus position of the fiber (which is glued) at the laser end is changed. The WA1100 in the photo actually has a JDSU 1107P installed as the replacement reference laser. And it indeed appears to have significant power coupled into the fiber cladding for this reason. Attempting to break the glue bond was deemed to risky.

    The specifications list a wavelength range of 700 to 1,700 nm. And indeed, the lower bound seems to be quite strict. There is even a VIS-blocking filter in the optical path, with the reflection from it going to the laser power PD. It's not clear why the wavelength range doesn't extend further into the visible. Certainly the IR PD still has some sensitivity at 633 nm so it should be possible to measure the wavelength of a common HeNe like its own reference! And for that matter, why not include a second signal PD so that the measurement range could be extended down to 400 nm or beyond as is done with the WavemeterJr? With minor modifications to the firmware, it should not require much more than a dichroic (or even broad-band) beam-splitter and $2 silicon PD. Darn Marketing! Cost reduced, strip out useful features to promote sales of the higher priced spread. ;-)

    Tuneup of a Burleigh WA-2500 Wavemeter

    The WA-2500 is appropriate named "WavemeterJr" as it is a much smaller, lighter, and simpler instrument than the WA-20 which may be used to measure wavelengths from 400 to 1,800 nm using separate detectors for VIS (400 to 1,100 nm) and IR (up to 1,800 nm). The precision is lower (5 digits instead of 6 digits) and it lacks the built-in reference HeNe laser, so an external HeNe laser must be used for calibration. However, this can be done regardless of whether the instrument is set for VIS or IR as the IR photodiode still has decent sensitivity at 633 nm. Given the lower resolution, this isn't nearly as important as with the WA-20. But the WA-2500 also lacks the tracer beam, and in fact, only allows for fiber coupled input via an FC connector on the rear panel. It is microprocessor-controlled and includes an RS232 port for data collection. A Web search for "Burleigh WA-2500" will locate an operation manual with description, specifications, and alignment information.

    Like the WA-20, the WA-2500 is based on a Michelson interferometer, but the optical setup is much simpler. The main component is a single Cube-Corner (CC) on frictionless dual flexure mounts with a electromagnet ("voice coil") to "excite" it at its mechanical resonance of about 10 Hz, providing 10 readings/second (or 1/s if averaging is turned on). An automatic locking mechanism keeps it from bouncing around when power is off, but makes an annoying loud clunk in doing so. :) The other interferometer optics include a 45 degree beam-splitter mirror, and two 0 degree mirrors. (A detailed layout is in the manual.) The alignment of one of these mirrors is critical, with adjustments accessible from the back panel. I just wish they had used a higher quality mount with finer pitch screws! There is also a fixed (45 degree) fold mirror which simply directs the internal beam to the adjustable one on the backpanel to make the optical layout work out in the available space.

    The only time there's a need to go inside is to flip the detector board to select VIS or IR. (One screw and one jumper.) A photo is shown in Photodiode Preamp PCB from Burleigh WA-2500. On this WA-2500, the final op-amp (LF347) was found to be blown along with a toasted (but apparently still functional) resistor. Someone may have plugged the cable in incorrectly, though this would seem to be difficult. :) Or it may be that the "Monitor" BNC on the rear panel is in parallel with this output, so perhaps someone plugged something in there that shouldn't have been plugged in there. :) So, the signal level was very low almost never showing up on the bar-graph display and only a few hundred mV at most from the Monitor BNC, with the machine either producing Lo Signal or Alignment Error. There was also no Window signal (middle BNC) at all which initially led me to suspect there might have been logic problems. But apparently, that isn't generated until some minimum signal level is detected. Once the op-amp was replaced, it instantly sprang to life. However, for optimum performance, using the identical op-amp (or at least one with adequate bandwidth) for that stage at least seems to be critical. Substituting an OP27 in place of the LF347 (which was the only single op-amp I had available at the time) resulted in a non-uniform signal envelope and incorrect calibration with respect to Hi Sig and the bargraph as well as limiting the range of the bargraph. With correct op-amp, the trim-pot can be adjusted for a reasonably flat signal envelope.

    The tuneup then consisted simply of peaking the signal level using the mirror alignment screws.

    The WA-2500 works very well with single longitudinal mode (single frequency) and multimode lasers where the modes stable and closely spaced, as they are in all HeNe lasers. Ifthe bandwidth of the lasing modes is larger than the coherence period over which the Mechelson interferometer samples, the WA-2500 will reduce the resolution so that a meaningful measurement can still be made. This is done by looking at the envelope of the fringe signal and only sampling during a segment between where it goes to zero. However, (not surprisingly) there can be real problems with multimode lasers where the modes may be jumping around. I could not get reliable measurements using a green laser pointer or crappy green DPSS laser module (which is probably similar). But it works flawlessly with any HeNe and high quality DPSS lasers like the Coherent C315M. There can also be some loss of resolution if using a multi (spatial) mode fiber for the input though this is usually minor, may be worth the greatly reduced hassle in coupling to the large fiber core.

    Performance is somewhat better in terms of sensitivity and consistency with a single mode fiber having a core size appropriate for the laser wavelength, but a multimode fiber can be used without too much difficulty. Even the one Burleigh provided can only be truly single mode for the longer wavelength range. A 9/125 telecom fiber will not be single mode for a 532 nm green laser, which requires a 3 or 4 µm core to be single mode inside the fiber.

    Most of the above also applies to the WA-2200, of which there appears to be no record on-line. The sample I tested did not allow for swapping between VIS and IR, and the detector PCB was slightly larger. But otherwise, the optical components and layout are identical to that of the WA-2500. My unit had a broken glass Moire plate, probably from being dropped hard. There are two glass plates with patterns of fine lines in close proximity to generate the reference signal in lieu of the HeNe reference laser. But the mass of the moving part of the optics could indeed cause them to contact if the shock is severe enough, even when locked. I remounted the remaining good piece as best I could - the other piece was no where to be found, presumably lost by the previous owner. I believe there is enough remaining to provide the necessary signal. I finally found the PLL test-point. Here are all the test-points with their function:

       ID    Function
      TP1    +5 VDC
      TP2   +12 VDC
      TP3   -12 VDC
      TP4    +5 VDC
      TP5   PLL Signal
      TP6   Common/Gnd

    I have no idea why there are two test-points that appear to have +5 VDC on them, but perhaps one is a reference voltage or digital and analog, or something. :) TP1 through TP5 are near the power supply; TP6 is on the other side of the PCB near the front panel. There are 4 trimpots on the main PCB:

       ID   Function
       R3   PLL Gain (set fully CW)
      R36   Bargraph Sensitivity?
      R49   ??? but critical setting, else PLL and other errors
      R53   ??? no obvious effect (set fully CCW)

    Adjusting alignment of the two Moire plate varies the PLL signal amplitude as expected. But the separation appears to have to be at the limit the mounting screws will permit to get a clean signal. With some fiddling, it was possible to get it up to almost 1.5 V p-p, a bit below the 2.0 V p-p the manual wants. But behavior is essentiaally identical down to less than 0.5 V p-p so that's probably OK. "PLL Err" is displayed on power-up a couple times but it then disappears, and that may be totally unrelated to signal level.

    I also found that the pot on the photodiode preamp board changes its bandwidth and setting it lower produces a flatter signal envelope at the expense of some gain. I'm not sure that this WA-2200 is quite as stable as the WA-2500 but it isn't too bad now. However, there is an annoying intermittent problem: Occasionally, usually a few minutes after being powered on, it will get into some state where it's not happy with any signal level producing "Lo Sig" or "Hi Sig" or possibly other errors continuously. This may continue for a few minutes and then totally disappear. There's a slight possibility it has to do with back-reflections destabilizing the test laser (Melles Griot 05-LHR-911) which is directly fiber-coupled to the wavemeter without an optical isolator, but that's not too likely.

    If anyone has service information on these (or other) wavemeters, please contact me via the Sci.Electronics.Repair FAQ Email Links Page.

    Burleigh WA-7100 and WA-7600 Multi-Line Wavemeters

    The WA-7100 and WA7600 appear are geared to the telecom industry providing spectral analysis capabilities using the same Michelson interferometer technology as the WA-1100 and WA-1600. Rather than simply displaying the precise wavelength of a single laser line, these instruments provide a spectrum analyzer type display using the Fourier transform of the fringe signal rather than simply counting fringes and comparing the count to a reference laser or grating. In other fields, this would be called a Fourier Transform Infra-Red Spectrometer or FTIR. ;-)

    These are larger and heavier than the WA-1100 and WA-1600, probably primarily to provide space for the larger LCD screen on the front panel. But the interferometer "engine" appears to be similar or identical. The microprocessor-based controller is physically similar but provides much more sophisticated measurement and display capabilities. Similar environmental sensors provide for wavelength compensation. As above, the primary difference between the two models is that the WA-7100 uses a normal 1 mW polarized HeNe laser head for the reference laser while the WA-7600 uses a frequency stabilized HeNe laser. The use of the stabilized reference laser allows for approximately a 5X increase accuracy and a 10X increase in resolution. And, a stabilized laser cannot simply be installed to increase resolution. The input to both of these instruments is fiber-coupled, with an FC/APC as the default connector. (There is no free-space option.) Complete specifications for both instruments are in the operation manual, easily found on-line by searching for "Burleigh WA-7600 manual".

    There are several subsystems inside the unit as shown in Burleigh WA7600 Multi-Line Wavemeter Interior View: DC power supply (upper right), fiber-coupled Melles Griot 05-STP-910 stabilized reference laser with its HeNe laser power supply brick (right), interferometer optics with processing PCB on top (covered, bottom), microcontroller PCB (upper left), display PCB (attached to front panel, hidden), and a little auxiliary PCB (assumed to be for the motor driver as with the WA-1100, left).

    As noted, the interferometers in the WA-7100 and WA-7600 are probably similar or identical to the ones in the WA-1100 and WA-1600, respectively, though I have not removed the cover to inspect it. The reference lasers are the same as well. See the previous section for more info.

    The VERE WM4100 Wavelength Meter

    While one of the systems described above is named "Wavemeter Junior", this one should perhaps be called "Wavementer Lite". And it's not really a wavemeter since it doesn't count fringes in any way, shape, or form using either an actual laser or mechanical reference. And there's no interferometer, so it really doesn't compute wavelength directly. Rather, a diffraction grating directs the incoming light from a 600 µm fiber onto a Position Sensitive Detector (PSD). Analog circuitry generates a voltage proportional to the wavelength based on the centroid of the spot position. An A/D converts that to a display of wavelength. The spec'd default wavelength range is 500 to 1,000 nm. The manual including specifications may be found at WM4100 Wavelength Meter Operation Manual. (A Web search will also find this manual.) Hamamatsu, one supplier of PSDs, has an on-line "optics handbook" extensive technical info on PSDs in Chapter 02: Silicon Photodiodes. Or, for a quick intro, see Wikipedia Position Sensitive Device. The 1-D version would only require the equation for X:

            x = Kx * -------

    where Ia and Ib are photocurrents from the ends of the PSD and Kx is a scaling factor. The analog circuitry of the WM4100 implements this equation directly using AD706 dual op-amps, PMI AMP03 differential amplifiers, and an Analog Devices AD632 divider IC. However, since for the diffraction grating, angle is a function of the arcsin of the wavelength, and the spot is projected onto a flat surface, additional calculations are required to correct for these non-linearities. Thus there is an AD538 "Real-Time Analog Computational Unit (ACU)" involved somehow. More on this when I figure it out. :)

    Where incredible accuracy is not of key importance, this approach works well. Any source with a single peak wavelength that can be coupled into the input fiber will produce a reading and the response is a fraction of a second - essentially the update rate of the A/D. So, multi-longitudinal mode laser diodes as well as high brightness LEDs will be acceptable. If you can eyeball a peak wavelength using a spectrometer, this thing will find it. More or less. :)

    And it is relatively small and light weight!!!

    I now have reverse engineered circuit diagrams for most of the system.

    The unit I acquired appeared to work fine in general and displayed around 633 nm for a red HeNe. But a green (532 nm) laser pointer known to have an effective IR-blocking filter produced 1,070 nm - around twice what it should be. It turned out that the 2nd order green spot appeared at the high end of the sensor rather than being blocked (or so I assumed) and the 1st order spot was blocked, and that obviously some calibration would be required. Someone *had* been inside as the internal ST cable at the diffraction grating-end was only half pushed into the connector and some mounting screws were loose. By slightly adjusting the angle of the diffraction grating (which secured tightly), the green spot was easily positioned in the Hopefully the 3 user adjustments will now suffice to tune it up and they didn't also twiddle the 3 unmarked trim-pots!

    However, upon further testing - after attempting to adjust the grating and all of the pots (including those that shouldn't be twiddled), I came to the conclusion that this instrument must have been set up for something like 600 to 1,100 nm. There was no electrical or optical adjustment that would ever result in a reading much below 600 on the display. Indeed. the full model number is WM4100C. So, perhaps the "C" means something. While an electronic failure could have resulted in a bogus 100 nm offset messing up the readings, that wouldn't explain how the 2nd order 532 nm spot hit the PSD. The grating position didn't appear to have been changed - at least its screw was very tight. No other optical adjustment could have had such a large effect. The screws securing the input lens were just snug, but that optic has very limited adjustment range.

    At present it is awaiting final calibration, but does read down to the HeNe 594.1 nm (yellow) line - barely.

    Here are a couple of photos and more information on calibration adjustments.

    Pulsed Wavemeters

    Most of the previous descriptions are for older wavemeters that only can be used with CW lasers since they are based on Michelson interferometer with motor driven mirrors and require significant time to take a reading. While these type of systems are still manufactured, modern high performance wavemeters that can be used with both CW and pulsed lasers are now fully "solid state" with no moving parts. They typically employ one or more Fizeau interferometers and linear diode (e.g., CCD) detectors. The interferograms (fancy name for fringe pattern images) are then transferred via to a PC for analysis and display. For more info, check out companies like Bristol Instruments and High Finesse GmbH.

    (From: Sébastien Bourdeauducq.)

    I think they're not that complicated and no less "natural" than Michelson interferometers :)

    You can think of each etalon as something like a Fabry-Perot cavity but wedged, i.e., the length of the cavity varies along one axis.

    The first coarse wavelength determination is done naively, i.e., by measuring the distance between the peaks in the interference pattern. The system should be engineered so that this coarse wavelength determination is sufficient to determine the order of the interference without ambiguity (the cavity length must be known in advance and stable to sufficient precision). Then the phase of the interference pattern is used to determine the wavelength with greater precision. The error on the phase measurement is divided by the interference order, which is where the greater accuracy comes from.

    This process is then repeated by cascading another (or several) similar etalons with increasing cavity spacings (lower FSRs, higher interference orders), each using the wavelength measurement from the previous etalon to determine the interference order and produce a new wavelength measurement with greater accuracy.

    I think the first publication of this idea was in this 1978 paper by Snyder (in Russian): Small-sized wavemeter for pulsed and continuous laser radiation, Quantum Electronics, 1978, Volume 5, Number 8, Pages 1682–1684 (Mi qe10622). Click on: "Full text: PDF file (433 kB)". blocks/redirects direct links to the PDFs.

    This is also an interesting read: A Laser Spectrometer and Wavemeter for Pulsed Lasers (NASA 1989.

    Michelson wavemeters require less advanced techniques to build (no need for tightly controlled and stable dimensions and well-polished optical surfaces), but there are moving parts that wear off and create vibrations, the measurement time is much higher, and they cannot measure laser pulses. And alignment is a pain.

  • Back to Laser Instruments and Applications Sub-Table of Contents.

    Ring Laser Gyros

    Basic Description and Operation

    Mechanical gyroscopes measure rotation by measuring forces on a rotating mass which has been machined and spun at high precision. Having moving parts they are often clunky, bulky and distinctly 'low-tech'. They often take a long time to 'spin up' and stabilize.

    In principle the ring laser gyroscope can replace these with a fully optical system using counter-rotating laser beams, photodetectors, and digital electronics with no moving parts larger than photons and electrons.

    In practice, it isn't so easy.

    In its simplest form, the ring laser gyro (RLG) consists of a solid triangular block of glass with a hole drilled out parallel to each edge. Mirrors are added at each corner, a laser gain media such as a mixture of Helium-Neon gases is added and a stable laser cavity is formed by a suitable set of 3 mirrors. The gain media is chosen to allow two counter-rotating laser beams to be established - one clockwise (CW) and the other counter-clockwise (CCW).

    One of the mirrors is very slightly transmitting so that the CCW and CW beams can exit the laser cavity. They are then mixed together and (due to the slightly different optical frequencies), the result will be a beat in the audio band called the "Sagnac frequency". This can be measured to very high accuracy either by simply detecting the fringes when the CCW and CW beams are directed to overlap at a slight angle, or when combined with optics similar to those used for homodyne metrology applications. One such setup adapted for the RLG is Sagnac Frequency Detector 1. In this design, the CCW beam is reflected and rotated 90 degrees by a Half WavePlate (HWP) so that its polarization is orthogonal to the CW beam. They are then merged by a Polarizing Beam-Splitter (PBS) cube. A Quarter WavePlate (QWP) converts the two now orthogonally polarized CCW and CW to left and right circularly polarized components. Another PBS and a pair of photodiodes produces the quadrature sin and cos (or A and B channel) electrical signals. There are variations on the way the final quadrature signals are generated from the combined CCW/CW beam as shown in Basic Homodyne Laser Interferometer Quadrature Decoders but these all produce similar results. However, Types 1 and 2 generate up to twice the signal amplitude for the same laser power and provide more flexibility in setting the precise phase shift by adjusting the relative orientation of the linear polarizers compared to Type 3. For more info, see the section: Quadrature Decoders for Homodyine Interferometers.

    Now for some not particularly hairy math. ;-)

    The basic equation for the Sagnac frequency as a function of the angular velocity of the RLG is given by:

       δf = ---- Ωn


    Thus the Sagnac frequency is proportional to the rotational velocity, operating wavelength and ratio of cavity area to perimeter. Or for the same shape "ring", simply to the diameter or perimeter.

    Strictly speaking, this equation isn't totally accurate. Among the issues that crop up when engineering a practical system, at very low rotation rates, cross-coupling of the CCW and CW beams produces a condition called "lock-in" where they have identical optical frequencies resulting in no signal. And at high rotation rates, mode pulling causes the CCW and CW lasing lines to be slightly offset toward one-another and not based solely on the Doppler shifts, thus reducing the Sagnac frequency compared to what it would be if the response were perfectly linear. The latter is a very small and mostly correctable non-linearity. However, a great deal of effort has gone into solving the lock-in problem without achieving an elegant yet practical solution. The vast majority of commercial RLGs use mechanical "dithering" to rotate the laser block back and forth by a small angle (typically hundreds of arcsec corresponding to hundreds of counts) at a few hundred Hz. It passes through the lock-in region only for a very short time at the extremes, mostly accumulating counts in the largely linear region, so the errors due to lock-in average out, there is no longer a dead zone, and bias/drift are reduced to an acceptable level. In fact, this simple, if clunky method, can enable an RLG that would otherwise require a substantial fraction of a degree per second to unlock, to accurately measure the Earth's rotation of around 4.166x10-3 degrees/second.

    As an example of a commercial RLG from the mid-1990s (but a version of which may still be in production), consider the unit described in the section Honeywell GG1342 Ring Laser Gyro, which is likely a version of the Honeywell GG1342. Although based on the "42" in the model number it should have equal leg lengths of 4.2 inches or 10.6 cm, the triangular ring has sides of approximately 12.5 cm, 10.6 cm, and 12.5 cm, so only one matches, and it's isosceles instead of equilateral, but that's close enough for government work. ;-) The GG1342 was apparently designed that way to get a bit more path length and enclosed area into a pre-existing limited space. The perimeter is thus 35.6 cm or 0.356 m with an area is 60 cm2 or 6.0x10-3 m2. For testing, n is equal to 1, rotating flat on the table. Calculating for 1 degree/second as an example, the result would be given by:

               4 * 6.0x10-3       1
       δf = ----------------- * ----- = 1.86 kHz
            6.33x10-7 * 0.356    57.3 

    Where the last term is the conversion from degrees to radians. This behavior is analogous to that of an optical rotary encoder with 1,860 cycles/degree.

    The Honeywell GG1320AN, which was originally developed around 1994 but an updated version is still state-of-the-art as of 2018, is just over 1/2 the diameter of the one above, housed in a case less than 3.5 inches across weighing in at around 1 pound. The discharge leg length is 2.0 inches based on the "20" in the model number. It is spec'd at 1,164,352 +/-18 pulses per full rotation or ~3,234 pulses per degree. This much higher value compared to the one above is the result of the quad sin/cos decoder generating 4 pulses per fringe cycle. For the same shape "ring", the number of pulses per angle change is proportional to the diameter or perimeter. Without the 4X multiplier, it would be ~808.5 pulses per degree. See Honeywell GG1320 Digital Ring Laser Gyroscope if you'd like to display your heading accurate to a few arcsec using an Arduino in your sailboat. :) I've archived the Brochure and Manual if the links don't work.

    One of the smallest commercial RLGs was probably the Honeywell GG1308, developed in the early 1990s for use in low cost inertial platforms for drones and other remotely piloted vehicles, helicopters, and the like. It had discharge legs of only 0.8 inches in length. (!!) The Sagnac frequency for the GG1308 was 116,000 +/-200 pulses per full rotation or ~322.2 pulses per degree (not multiplied by 4). There is some interesting detailed information on the GG1308 in Performance Evaluation of the Honeywell GG1308 Miniature Ring Laser Gyroscope - DTIC.

    And while we're at it, the largest RLG known to have been constructed was the Canterbury University UG-2 Ring Laser. This definitely NOT portable monster had ring dimensions of 39.7 x 21 m for a perimeter of 121.4 m and an enclosed area of 834.34 m. The Sagnac frequency was ~757.8 kHz for a 1 degree rotation! ;-) Since it couldn't be picked up and mounted on a plane or even an aircraft carrier :), about all it could measure was the Earth's rotation with an average Sagnac frequency of 2.177 kHz. However, it was able to detect polar wobble due to the Moon's gravity. :)

    A complete 3-axis inertial platform would require 3 RLGs mounted at 90 degrees to each-other. The entire affair can be fabricated inside a solid block - at least up to a point! However, it's generally a lot less costly to simply replicate a single axis gyro. :)

    However, as noted, there are problems with this simplistic implementation:

    For the most part, these difficulties have been overcome to a degree sufficient to allow for navigation of aircraft, spacecraft, ships, and submarines to very high precision. In such applications RLGs are increasingly being used in place of mechanical gyroscopes.

    A much more complete commercial RLG assembly including photos, analysis, and testing of the Sagnac frequency output is described in detail in the section Honeywell GG1342 Ring Laser Gyro, below. This includes in depth practical operational information on an actual working RLG, something that I have not been able to find anywhere else.

    A more extensive (than you probably desire) introduction to RLGs can be found at Leybold Laser Gyro Kit Instruction Manual. This is associated with Leybold's complete (and $$$) ring laser gyro kit. And a Web search will return a variety of publically-accessible research papers on experimental RLGs (among others) as well as many RLG patents.

    (From: A. E. Siegman)

    If you go to the Laser Gyros Directory, you'll find photographs of an early square He-Ne ring laser gyro built by Sperry and some early designs for Honeywell's monolithic ring laser gyros. (This link is long dead but while unlikely, it might have been preserved by the Internet Wayback Machine, and I've let it remain here in memoriam.) Sperry Experimental Ring Laser Gyro is probably similar.

    The Sperry gyro couldn't actually be rotated in the lab - kind of hard to spin a one ton or thereabouts optical table. So they relied on the Earth's rotation, or at least the vector component of it perpendicular to the table at Sperry's latitude, to test their system.

    I recall a conference talk on their work in which the speaker noted that, given the backscattering and lock-up problems associated with a ring laser at this low rotation rate, their primary conclusion was that as best they could tell the Earth was still rotating, but at a highly uncertain rate. [That could be annoying. ;-) --- Sam.]

    (Portions from: Flavio Spedalieri.)

    Photos of a partial RLG laser assembly from a commercial inertial platform can be found at Flavio's Ring Laser Gyro Page. This is a triangular ring cavity in a solid block. But the dual discharge HeNe laser plasma tube is external to the block, which is really quite strange. :) At least one of the mirror mounts has a coil, presumably to do the dithering using the Kerr-effect to overcome "lock-in" rather than the mechanical dither motors used in other systems that literally vibrate the entire block at an audible frequency (typically between 400 Hz and 1 kHz) with an amplitude above the lock-in threshold. The Kerr method is said to be advantageous as it does not introduce errors in accelerometers and other axes on the inertial platform. [And it is completely silent so there is no obnoxious sound doesn't take the hair off your teeth. ;-) --- Sam.] The detector optics can also be seen. Apparently, Sperry made two sizes of their RLG used in the SLIC-7 and SLIC-15 IMU strap down assemblies. The laser wavelength is actually 1,1503 nm. I have additional diagrams which apparently are out of the service manuals back in the 70s including an optical schematic (but it does not show the etalon which I have on mine). Also there is the diagram of the output mirror and combination prism and paths. Would love to try and power up the other unit that I have.

    (From: Sam.)

    Yeah except that testing at 1,153 nm would be interesting....

    Increasing the Sensitivity of Ring Laser Gyros

    The RLG generally uses a cavity configuration which is usually either a triangle with 3 mirrors or a rectangle with 4 mirrors. The effective cavity length is the perimeter of the "ring" (c/FSR) and based on the Sagnac equation, sensitivity is proportional to perimeter. So increasing the sensitivity requires a larger perimeter. But this doesn't necessarily require a physically larger ring. Various techniques can be used to force the effective cavity length to be a multiple of the perimeter. This can be accomplished physically or optically. Here are three possibilities:

    1. Mirror configuration: Add additional mirrors so that the ring doesn't close on itself but wraps around 2 or more times.

      These schemes are complex in terms of both topology and the need for increased precision and stability as additional mirror mounts are added. I am not aware of anything like this being used even in research RLGs. But feel free to correct me and send photos of your 17 mirror RLG. :)

    2. Faraday rotation: Insert a Faraday Rotator (FR) in the beam path that rotates the polarization by a sub-multiple k of 180 degrees. It then requires k circuits of the ring for the polarization to align. For example, with am FR of 45 degrees, the light will go around the ring 4 times before it again aligns with itself. This requires either a gain tube with perpendicular windows or an integrated cavity or cavity block.

      Implementation could be rather straightforward if the added losses of a Faraday rotator can be tolerated. Common optical isolators general implement a 45 degree rotation internally. Since these are designed to minimize back-reflections, their AR coatings should be very good. So utilizing just the rod and magnet from one of these with a means of adjusting how far into the magnet the rod is placed could permit this concept to be tested.

    3. Half WavePlate (HWP) rotation: Insert a HWP in the beam path that rotates the polarization by a sub-multiple of 180 degrees. It then requires multiple circuits of the ring for the polarization to align. This also requires either a gain tube with perpendicular windows or an integrated cavity or cavity block. Since the HWP cannot be placed at the Brewster, it must be AR-coated to minimize losses.

      It seems like a doubling should be straightforward. But going beyond that may be tricky or impossible since a HWP doesn't function like a Faraday rotator independent of input angle.

    Whether this can work will depend on other factors. For example, even random polarized HeNe lasers normally exhibit a polarization preference with adjacent longitudinal modes being orthogonally polarized with an orientation locked to the physical laser tube. This is also true of ring lasers where the asymmetry of non-normal reflections from the mirrors introduces a polarization preference. This preference could override and/or compete with the action of the FR resulting in an unstable or chaotic polarization state.

    The Challenge of Unlocking the RLG

    One of the fundamental limitations of all ring laser gyros is mode locking at low rotation rates. Here "mode locking" refers to the CCW and CW lasing modes becoming identical in frequency and locked in phase thus losing any rotation information when the RLG is nearly stationary. For a typical small commercial RLG, the "lock-in" threshold of the laser block itself is a fraction of a degree and is greater by orders of magnitued than the few arcsec or less of its native resolution. Lock-in occurs due to unavoidable cross-coupling between the beams as a result of imperfections and scatter inside the ring cavity, mostly from the mirrors. No matter how well the substrates are polished and coated, there will always be some unavoidable imperfections. Various schemes have been used to compensate for lock-in in small RLGs. Here are a few:

    More complex non-mechanical approaches have also been implemented, though it's not clear if they are significantly better and to what extent they have been used commercially. Simply fabricating the normal laser block is challenging enough without adding additional intra-cavity optical components, magnetic fields, strange non-planar geometries, etc.

    The totally non-mechanical lock-in schemes appear at first to be quite attractive but the devil is in the details. Non-linearities, sensitivity to temperature and magnetic fields, and managing the issues associated with more complex lasing (i.e., beyond gain and mirrors) make these very complex to implement in a consistent reproducible manner. Thus, mechanical dither continues to dominate the industry.

    The "Lock Threshold" or "Lock-in Threshold" does go down as the enclosed area of the RLG increases, so football field-size RLGs may be sensitive and accurate enough without resorting to these schemes. :)

    So let's explore a couple options for not just minimizing the locking problem but eliminating it entirely. Before you get your hopes up, there is no magic solution that will satisfy all criteria, at least none are on today's agenda. ;-). There have been hundreds of research papers over the years representing countless thousands of person-hours of effort with no perfect solution. However, if either of those described below has not been proposed before, I'm claiming them as of January 2019. ;-) The first might be considered clunky or kludgey and definitely not elegant, but is technically dirt simple and guaranteed to work. The other is more speculative and may or may not even be possible or practical.

    1. Adding rotational bias: Since lock-in only occurs at low angular velocities - typically well under 1 degree per second - it could be totally eliminated by continuously rotating the RLG block or entire RLG at a precise angular velocity enough above the maximum spec'd value so that at the extreme that reduces it, the value is still well above the lock-in threshold. For example, if the spec is +/-450 degrees/second, a continuous rotation rate of +500 degrees per second would guarantee that the RLG would never go below 50 degrees per second. This is still only 83-1/3 revolutions per minute (rpm), just above the 78 rpm of pre-Jurassic audio recordings. :) So it's not like a mechanical gyro spinning at thousands of rpm. The rotation seen by the RLG would now range from +50 to +950 degrees per second, which might mean that a parameter in the setup file needs to be changed for it to meet spec, but so be it. ;-)

      From an engineering standpoint, the mechanical and electrical implementation should be quite straightforward. A motor would rotate the RLG assembly at a fixed crystal-controlled rate (83-1/3 rpm for the example, above). At such a low rotation rate, all that's required is for the rotating assembly to be fairly well balanced. There are no issues with angular momentum as would be present with a high speed spinning mass on a fixed mount. AC power can be coupled to the rotating assembly inductively via a rotary transformer. It would be converted to the required DC voltages internally. Only two electrical signals are required and can both be in a serial data format like RS232/RS422. One is for transmitting commends to the RLG and the other is for receiving angle data and status from the RLG. These can be transferred using rotary transformers, optically, or via an RF link. Slip rings could also be used for both power and signals but are subject to electrical noise and wear.

      However, several sources of error may be introduced. The inertia of the rotating assembly will resist sudden changes in angular velocity. Even with a stiff feedback loop for the servo motor, a rapid maneuver could result in its orientation lagging behind and/or there could be oscillations. The result would be errors in angle, though they would disappear as the system settled down. Small transient inaccuracy may not be of major significance though. Any variation in the actual angular velocity of the RLG due to bearing friction or servo motor cogging (periodic offsets due to the motor's discrete pole pieces) will show up as a corresponding oscillation of the angle even if the inertial platform is stationary. Temperature changes that affect the cavity length (perimeter) will alter the Sagnac frequency. Further, the null point where the angle output should read zero is based on the difference between two large numbers typically more than 106 Hz - the Sagnac output from the RLG and a frequency synthesizer-generated value derived from the crystal reference for the slow rotation. The net result may be random changes, offset, or drift of the angle readout. And there are no doubt others. However, it could be argued that these are not fundamental like lock-in but are engineering issues and can be reduced to an arbitrarily low level through careful design and implementation.

      I promised that this scheme was not exactly elegant. Now you know beyond a shadow of a doubt. ;-)

      As a specific example, consider retrofitting a Honeywell GG1320, which is probably the "gold standard" of commercial RLGs in 2019. It's compact and light-weight with full digital processing done internally. However, the GG1320 (and other modern commercial RLGs, most using dither) are already darn good, with sensitivity way higher than needed to be able to cleanly display the Earth's rotation, low jitter, noise, bias error, drift, low nonlinearity, and overall accuracy sufficient for most navigation applications. So the following would likely end up being just an academic exercise.

      The main modification would be to disable the dither motor and associated processing and error checking. How to accomplish this is not clearly documented in the publicly-available operation manual, but is almost certainly possible, for testing if nothing else. Then at most, only a simple adapter board would be required to deal with the handful of status and control bits. (See the section: Honeywell GG1320 Ring Laser Gyro for info on construction and operation.)

      Power requirements for the GG1320 are modest: +15 VDC at 150 mA continuous (300 mA max), and +5 VDC at 150 mA. However, with no dither motor drive required, the +15 VDC current would be somewhat lower. AC power could be inductively transferred across the rotary joint and converted to regulated 15 VDC, and that could then be stepped down via a DC-DC converter to 5 VDC.

      No signals need to be sent to the GG1320 to implement basic operations including angle readout except a Sample Data Request clock for each data packet - and that can be generated by a local 555 timer with Clear to Send tied high. Initialization can be configured to be automatic at power-on, though being able to send a pulse to reset the system should something unfortunate occur may be desirable. That could be coupled optically via a stationary LED and photodiode on the rotating assembly. Sending other control and configuration information should be possible using the normal Serial Data In stream, piggy-backing additional control functions on it by taking advantage of unused codes. Angle data and status would be returned using the normal Serial Data Out stream. The small non-linearity resulting from operating well away from 0 angular velocity and above the original spec can be corrected via calculation or a lookup table. Both input and output data can most easily be transferred via an RF link.

      With its small size and weight, the mechanical design is straightforward. A "pancake" servo motor would rotate the GG1320 at a fixed rate based on a crystal reference. Possibilities include a brushless DC motor with a high resolution optical angular encoder for feedback (closed loop) or an AC motor or microstepper with programmable multiphase drive (open loop). Common ball bearings would suffice for support and should last essentially forever at the modest load and slow rotation rate. For power, a pair of pancake coils with an appropriate turns-ratio in close proximity may suffice while the rotary head assembly from a bygone era VCR could perhaps be modified to handle the data signals if inductive coupling was preferred over RF.

      So it's possible that with at most a very small amount of additional information on disabling dither, this scheme could be tested. Simply disconnecting the dither motor would result in an error and the "Gyro OK" bit would get turned off. However, it's possible the GG1320 would continue to send valid angle data even without dither, as it's done in tests where the GG1320 was not bolted down and the behavior wasn't what its microbrain expected.

      And as a bonus, in addition to potentially improving the performance by an order of magnitude or more, the GG1320 would then no longer make the fingernails-on-chalkboard squeal produced by its dither motor when in operation. ;-)

    2. Decoupling the CCW and CW beams: The most obvious approach to unlocking is to separate the CCW and CW beams entirely. If they are nowhere near each-other there would be no chance for them to lock together. So, iInstead of two counter-rotating beams in one laser, why not implement the RLG cavity as two separate unidirectional ring lasers with their outputs combined externally? The Sagnac equations would be identical. Sounds simple, right? Well not exactly.

      Note that the issues here are NOT the same as with a "Fiber Optic Gyro" or FOG, where an external laser is split and injected as counter-rotating beams inside a spool of fiber to then be combined to interfere and generate a a signal. The primary difference is that the laser is part of the RLG and rotation changes the optical frequencies of the counter-rotating beams with a difference frequency proportional to rotation rate, while the FOG is a passive interferometer where rotation changes only the effective path lengths and thus the phase between the two beams. Since they originate externally, their frequency cannot be changed by any simple interaction. See the section: Difference between Ring Laser Gyro and Fiber-Optic Gyro.

      Building unidirectional ring lasers is simple in principle: An "optical diode" is installed inside the ring cavity which favors transmission in one direction over the other. For the RLG laser, optical diodes of opposite polarity are incorporated into each of the two beam paths, fabricated totally separately in the cavity block. An optical diode is normally implemented using the Faraday effect: By applying an axial magnetic field to a "gyro-magnetic" material, the orientation of a linearly polarized beam passing through it will be rotated by an angle proportional to the magnetic field and path traversed in the material. This is a non-reciprocal effect. Non-reciprocal means that a beam passing through the material in the opposite direction will experience the opposite rotation with respect to its frame of reference. So, for example, if the polarization of a beam in the forward direction is rotated by +15 degrees with respect to fixed coordinates, the same beam reflected in the backward direction will be rotated by another +15 degrees, and will end up at +30 degrees compared to the original beam, rather than retracing its steps. Then if a Half WavePlate (HWP, which is reciprocal) is introduced in the beam path to precisely compensate for the orientation change, transmission in one direction will be favored over the other as the polarization around the ring will not "align" in the backward direction so there will be higher losses. This technique is used to implement Faraday isolators (where the directionality is near perfect) and is commonly used in the cavities of solid state ring lasers which despite their relatively high gain, require only a small loss in one direction to force unidirectional operation in the other.

      Doing this for a low gain HeNe laser is possible in principle, though there are definitely major challenges in implementation to keep intracavity losses low enough. See Experimental HeNe Clockwise Unidirection Ring Laser. A rod to provide the medium for the Faraday effect and Half WavePlate (HWP) represent 4 optical surfaces added to the intra-cavitybeam path. However, if the rod has a Brewster-angle polished surface at one end and the HWP were oriented at the Brewster angle, losses would be minimized and should be acceptably low for HeNe operation especially since high output power is not a major consideration for an RLG. Iodine stabilized HeNe lasers typically have 8 Brewster-angle intra-cavity surfaces. For an experimental configuration with a two-Brewster plasma tube, the number of surfaces would be the same, though without a convoluted non-planar arrangement, one end of the rod would have to be AR-coated rather than Brewster angle. Wrapping one's brain around a non-planar RLG cavity could prove interesting but would be doable. :) If an etalon needs to be included to force single frequency operation, that's another 2 optical surfaces, though they should "disappear" at resonance.

      So, implementing a pair of independent unidirectional ring lasers is definitely a challenge, but one that can be overcome with enough care and determination. Think of a pair of RLG cavities one above the other differing only in the orientation of the HWP to be rotated by the same amount in the opposite direction. They could even be fabricated inside one monolithic block, possibly even sharing the same magnet.

      But ironically, constructing the unidirectional lasers is not the real problem. The real problem is, well, locking them together. :) No matter how precisely the cavities are machined, there will always be differences in dimensions of the paths around the two rings. And a difference (or more precisely, relative stability) of 1 picometer is signficant as will be shown below.

      The Free Spectral Range (FSR) of a ring cavity is given by c/L (twice that of a Fabry-Perot cavity of length L). This is also how much the optical frequency of a laser's longitudinal modes will change with a change in the cavity length of one wavelength (assuming no mode hopping). As an example, for a ring with a 1 meter total path length (circumference), the FSR is approximately 300 MHz. If the path length changes by 1/10 wavelength at 633 nm (63.3 nm), the lasing wavelength will change by 30 MHz. For 1/1000th of a wavelength (0.316 nm) it will change by 300 kHz. Open loop (without optical feedback) techniques using temperature control alone can stabilize a laser to a few MHz. And using optical feedback, to much less. But then any close-loop optical control would be fighting the precise effects that the RLG is supposed to be measuring.

      What is needed is to stabilize the path length difference between the two rings so that 0 Hz (or a constant and known difference) is maintained. ;) But there's no way to guarantee the CCW and CW beams have the same optical frequency - or even close - by temperature control or other technique that doesn't use the wavelength itself to lock. And now we're going around in circles in more ways than one. :)

      When both the CCW and CW beams with the same spatial mode are using the same cavity, they will be exactly the same if the RLG is stationary since the CCW and CW path lengths are identical. If the relative path lengths could be maintained within 0.000001 nm (1x10--15 m) of each-other, about the diameter of a proton :), the optical frequency difference would be constant to within 1 Hz. :) Even if machined next to each-other inside a solid Zerodur block, and one of them were maintained constant by any means possible, the other would still drift so the relationship between the two path lengths could not be maintained to that precision. While they may be close enough for a finite time, the result would not be digital precision, but more akin to drift in mechanical gyros, which kind of defeats the entire purpose.

      So, is there some way to lock together the cavities themselves without affecting the LRG beams? One way to do this in principle would be to have them operate as dual-wavelength lasers. After 633 nm, 640 nm has the highest gain in the visible (almost half that of 633 nm) and might make a suitable secondary wavelength. In fact, where a HeNe laser manufacturer doesn't implement the mirror recipe quite correctly, some coherent 640 nm may be present in the output. One of the cavities would operate with both 633 and 640 nm in the same direction, say CCW. For this, the HWP would need to be designed such that it provides a similar rotation for both 633 and 640 nm. This HWP should be relatively close to a standard part since the wavelengths differ by less than 1 percent, though it would need to operate at the Brewster angle. The other ring would need to operate with a CW beam at 633 nm and a CCW beam at 640 nm, which would require the HWP to provide the opposite rotation angle. Could this be done? This is yet another 64 thousand photon question. :) Again, in principle. But it won't be found in the Newport catalog. However, assuming that's it's possible, there would then be two CCW beams at 640 nm whose difference frequency would not be affected by rotation.

      The secondary wavelength doesn't absolutely need to be 640 nm. It could be in the near-IR. For example, 1,152 nm might be better if it's gain were high enough, but that could also add additional complications in terms of dual-wavelength mirrors and HWPs. Or being further apart might even be advantageous.

      Either way, if the lasers could be locked together using the secondary wavelength, the cavity lengths would be identical and the 633 nm beams would still be independent. But it may be more likely to catch sight of some flying pigs than to make this work. ;-) Even if suitable optical parts (HWPs, Faraday rod and magnet) could be procured, implementing a dual-wavelength HeNe laser in an open cavity, whether using a two-Brewster plasma tube or inside a solid block, could prove impossible.

    To be continued.....

    Difference between Ring Laser Gyro and Fiber-Optic Gyro

    Unlike an RLG, the "Fiber Optic Gyro" or FOG uses an external single frequency laser (generally a telecom DFB diode laser at 1,5xx nm). Its beam is split into two parts and injected in opposite directions into a spool of single mode fiber. The beams that emerge at the opposite ends of the fiber are then combined with interferometer optics.

    At first glance, it might appear that the Ring Laser Gyro (RLG) and Fiber-Optic Gyro (FOG) should be very similar. They both use lasers, are shaped like rings of sorts, have counter-rotating beams, and generate a Sagnac signal based on interference between them. However, aside from using coherent light, there really isn't that much in common. And their characteristics, well, differ by the speed of light. :-) The primary physical difference is that the laser itself is an integral part of the RLG so that a rotation changes the optical frequency of the counter-rotating beams. This is the direct result of the effective change in path lengths around the ring seen by the CCW and CW beams, which causes the lasing modes to shift up and down in optical frequency. The result is that the RLG produces a Sagnac frequency (rate of change of phase) that is proportional to the angular velocity. However, the FOG uses a passive optical fiber whose rotation changes only the effective path lengths seen by the CCW and CW beams. Thus, the FOG produces a Sagnac phase signal that is a function of the instantaneous angular velocity. And while a typical small RLG generates hundreds to thousands of complete fringe cycles (each 2π radians) per degree, a FOG with 1,000 meters of fiber wound on a 10 cm form will only produce a phase difference between the two beams of around 0.012 radian for an angular velocity of 1 degree/second! When the widely used Honeywell GG1320 RLG is compared to this example of an FOG, the ratio in sensitivity is over 423,000:1. However, the RLG is not overwealmingly better. That would be like comparing apples to aardvarks. The RLG has its own issues with respect to non-linear response, complexity, and cost.

    Instead of going through the complete equations for both types, for the following thought experiment, assume that the light path is circular for both the RLG and FOG and both run at 633 nm. Yes, I know, building a circular HeNe could prove challenging but this IS a thought experiment. ;-) (Actually, hollow-fiber HeNes have been proposed, though I don't know if any were ever built successfully.) However, a circular perimeter allows the RLG and FOG to be treated in the same way without worrying about enclosed areas and such which simplifies the math. Also assume the axis of rotation is perpendicular to the plane of the ring and ignore relativity since these velocities are very low compared to the speed of light (c).

    The wavelength for both is assumed to be 633 nm for an optical frequency of ~474 THz.

    Then for a ring of radius r, the difference in linear velocity around the perimeter of the ring for the CCW and CW beams relative to the speed of light will be equal to:

       Vl = Ω * ----

    For this thought experiment RLG, the Sagnac frequency is:

          δf = Vl * 474 THz

    The change in angle is then the integral over time of the Sagnac frequency times 633 nm/(2r) or equivalently, the accumulated phase.

    Now for the thought experiment FOG, the Sagnac phase shift is equal to:

              2πrNVl     2πrNVl * 474 THz     2πrN
        Φ  = -------- = ----------------- = (-----) * (Vl * 474 THz)
              633 nm            c              c

    Where N = the number of turns of the fiber coil.

    So what do these equations indicate in terms of the comparison of the RLG to FOG??

    1. The first difference between the RLG and FOG is fundamental: The signal from the RLG is a frequency proportional to angular velocity. This is analogous to the output of a rotary encoder. But the output from an FOG is a phase shift proportional to angular velocity.

      The Sagnac frequency δf from the RLG can be counted or accumulated to determine the precise change in orientation. It is essentially a digital output of angular change in 633 nm increments around the perimeter - typically less than 1/1,000th of a degree of rotation per count for even a small RLG - with interpolation only required for finer precision.

      However, the Sagnac phase Φ from the FOG is simply an analog signal that is related to the instantaneous angular velocity. It's not even linear, but sinusoidal since it is the result of interference. (Though as a result of (2) below, it's only the first quadrant of the sinusoid that typically ever matters.)

    2. The second difference is in sensitivity. While the wavelength of 633 nm in the denominator of the first instance of the equation for Φ (left) is a very small number, the movement of the perimeter of the ring for realistic angular velocities relative to the speed of light - even when multiplied by a large number of turns - is a much much smaller number and c totally dominates regardless of the other factors. This is most dramatically evident when the equation for Φ of the FOG is rearranged (far right) as the product of a constant term with c in the denominator and the Sagnac frequency δf for the RLG.

      While a typical small RLG (less than 0.5 meter perimeter) generates thousands of fringe counts per degree, an FOG with 1,000 meters of fiber wound on a 10 cm form will only produce a phase difference change between the two beams of less than 1 degree) for an angular velocity of 1 degree/second.

    3. The third difference is that RLGs provide an absolute angular increment. The use of the HeNe laser as the basis for most RLGs means that the lasing optical frequency is an intrinsic property tied directly to the neon gain curve. Even without stabilization, this will result in an absolute accuracy of around +/-1 ppm. While there will be variations due to changes in temperature and other physical parameters, these are manageable. FOGs on the other hand generate an output that is a function of angular velocity. While the expected value can be computed precisely from a known rotation rate, when going the other way, the actual output is from (2) above, a very low level analog signal. So mechanical and electrical noise, offset and drift in the electronics, and thermal and other environmental changes can make the accurate computation of rotation and tracking over time challenging.

    If it weren't for the lock-in effect inherent in the RLG, little more than a simple up/down counter could be used to provide orientation information to a few arcsec precision that is stable over an arbitrarily long time scale. And going down to a small fraction of an arcsec requires only minimal additional electronics and simple computations. However, the FOG requires sophisticated baseband processing of the non-linear phase signal with very low and possibly noisy voltages, especially for typical low rotation rates. Further, orientation is the integral over time of an analog quantity meaning that this is not an absolute measurement but will be subject to offsets and drift in the electronics, similar to those of a mechanical gyro. Yes, there are no moving parts, but making the FOG live up to its potential takes some serious engineering.

    Thus the RLG and FOG have benefits and deficiencies. Through careful design and manufacturing, both can be implemented with acceptable performance useful in numerous applications.

    Comparison of Honeywell and Research Ring Laser Gyros

    Honeywell Aerospace is the largest manufacturer of commercial ring laser gyros. Here are some physical and functional specifications for 6 Honeywell RLGs. The GG1342 and GG1320 were manufactured in large quantities. It's not clear if this is also true for the GG1308 which was a low cost RLG intended for drones and other unmanned vehicles, and helicopters. The GG1328 was apparently intended for missile guidance. Since both the GG1308 and GG1328 appear to have been intended primarily for military applications, perhaps there isn't a lot of unclassified information available. There is also a GG1389 intended for high precision and marine navigation. Honeywell defines the GG part number where the last two digits are the leg length in inches x 10. Except, that is, for the GG1300, which was the original prototype before anything other than the "GG13" had been selected. :-) It should be designated GG1357 but was never updated. And also except for the GG1342, all 3 legs are equal. :) Values that can be computed based on dimensions are present for all 6 RLGs. Somewhat detailed specifications are only available for the GG1308 and GG1320, thus the gaps for the others. Slide #46 in Introduction to Navigation Systems shows the cavity blocks for all of these side-by-side glowing from lighting or Photoshop, not an electrical discharge. There is also one labeled just "Rate Sensor" without a model number that looks like the second smallest after the GG1308. That slide from the 2015 presentation indicates that the GG1342 has by far the highest production volume compared to all the others combined for which data is shown: over 50,000 units. However, the date of the slide itself is not known. And the photo is from a print Ad from October, 1986. Information from other sources suggests that the actual production numbers for Honeywell RLGs are much higher.

    In the chart below for Honeywell RLGs, all computed values are based on a wavelength of 632.8 nm and the dimensions taken from the model number (or in the case of the GG1342, measurements of an actual unit as the leg lengths are not equal). These may differ slightly from the values in the published specifications, if available.

     Honeywell Model           GG1308    GG1320     GG1328     GG1342  GG1300/GG1357  GG1389
     Leg 1 Length (in)          0.80      2.00       2.80       5.00        5.70       8.90
     Leg 2 Length (in)          0.80      2.00       2.80       4.20        5.70       8.90
     Leg 3 Length (in)          0.80      2.00       2.80       5.00        5.70       8.90
     Perimeter (in)             2.40      6.00       7.40      14.20       17.10      26.70
     Enclosed Area (sq in)      0.277     1.732      6.253      9.529      14.069     34.29
     Cycles/Full Rotation      116,487   291,217    553,313    676,966    829,969   1,295,916
     Counts/Full Rotation X4   465,947  1,164,869  2,213,251  2,707,863  3,319,876  5,183,665
     Cycles/Degree              324.0     808.9     1537.0     1880.5      2305.5     3599.8
     Counts/Degree X4          1,294.3   3,235.8    6147.9     7521.8      9221.9    14399.1
     Resolution (arcsec)        11.13    4.4503     2.3423     1.9144      1.5615     1.0001
     Resolution X4 (arcsec)      2.78    1.1126     0.5856     0.4786      0.3904     0.2500
     Linearity                 50 ppm     5 ppm
     Bias (deg/hr)                1       0.005      0.01      0.001       <0.001     0.0001
     Dither Frequency (Hz)    900-1,550  485-669
     Earth's Rotation δf (Hz)    1.35     3.38       6.42       7.85        9.63       15.0
    The Sagnac frequency δf for the Earth's rotation assumes the normal n of the RLG is aligned with the Earth's axis.

    For a comparison with several research RLGs that have been constructed to date, as well as one I'm building :), see Key Parameters of Various Ring Laser Gyros. Note the units change: Honeywell RLGs (and mine) are spec'd in inches while the research RLGs are spec'd in meters. :) At first glance the order of magnitude of the leg lengths appear similar but the research RLGs are ~40 times larger as a result of the inch to meter conversion! And UG-1/UG-3 and UG-2 can only be described as gargantuan or monstrous. But even most of the other research RLGs are too large and heavy to move, so performance measurements are based on the Sagnac frequency from the Earth's rotation. And as above, the values assume the RLG normal vector and Earth's axis are aligned. (The actual values for their location and orientationare given in the papers.) More on mine may be found in the section: Sam's Home-Built Ring Laser Gyro 1, below.

    Honeywell GG1342 Ring Laser Gyro

    (The following includes contributions from Dr. Walter Luhs of Photonik Ingenieurburo who offers an RLG as well as other laser/optics/photonics educational kits, and developed the original experiments for MEOS and Leybold.)

    The description and tests are based on access to an entirely intact and completely functional commercial HeNe RLG assembly from the 1990s for an aircraft inertial navigation system. Athough there were no identifying labels because it was acquired with the outer cover already removed, there is no doubt it is a Honeywell GG1342 based on the general style of construction and dimensions, and information found in Blazing Gyro - The Evolution of Strapdown Inertial Navigation Technology for Aircraft. (That article also includes extensive historical information on the development of the first Honeywell RLG presented with a wee bit of humor thrown in.) The photo on page 37 matches the design of this unit quite well. Even the small circuit boards appear to be identical to those in the photo. And the photo on page 36 of the complete IMU shows the RLGs with covers that have the same shape and style. This was confirmed a few years later when a pristine GG1342 with outer cover was obtained and it is virtually identical down to the color of the paint job.

    The GG1342 was the first Honeywell RLG to be fully commercialized and shares most basic design features with the smaller GG1320, a version of which is still state-of-the-art and in production as of 2018. Although the GG1342 was originally developed in the late 1970s, a version it also continued to be produced at least as late as 2015. And even if not now available from Honeywell, new/NOS GG1342s can be found on eBay. ;-)

    Honeywell GG1342 Ring Laser Gyro Assembly Powered shows it in all its glory. An annotated photo is shown in Honeywell GG1342 Ring Laser Gyro with Major Parts Labeled. The heart of the RLG assembly is a solid three-sided block fabricated from Cervit, a glass-ceramic material which has a near-zero coefficient of thermal expansion. The block includes the HeNe laser discharge paths, mirrors at the corners with PZTs *behind* two of them (lower left and right), and the detector assembly at the third (top). The actual path lengths of the laser cavity are approximately 5.0" (12.5 cm), 4.2" (10.6 cm), and 5.0" (12.5 cm). Based on the "42" in the model designation of GG1342, they should all be 4.2", but this was designed to fit an existing case and making it slightly larger would result in better performance. Or so the story goes. :) The block is mounted on a PZT-driven dither motor in the center which rotates the entire block back and forth by a minute amount (typically 250 arcsec) to compensate for mode-locking of the two counter-rotating beams at low angular velocities. 250 arcsec is several hundred (X4) counts but by filtering and averaging, the result is insensitive to low rotation rates or even being completely stationary and very stable. There are many diagrams of RLG structure on the Web but they differ in some details compared to this model. The labeled photo should provide better information.

    In the glowing photo :), the laser is running stably at a current of around 0.95 mA (!!) for each of the two discharge paths. The voltage of the discharges is around 650 V with around total 1 kV before the ~370K ohm (also !!) ballast resistors. The stable current could be much lower but 1.9 mA was the limit of the instrumented power supply used for these initial tests. (In fact as will be seen below, the discharge remains stable below 0.25 mA/discharge!) 0.95 mA is already by far the lowest stable HeNe current I've ever encountered, less than 1/2 of of the dropout current even for a very short tube. The ballast resistors being around 370K ohms also help, as do the relatively wide bores - which are much much wider than the beam, which is less than 0.5 mm at the output mirror. So the mode diameter must be limited at some point around the ring, or by the mode volume determined the mirror RoCs.

    And it does lase. There is a tiny red dot barely visible near the top of the photo, which is scatter off the inside of the output mirror. The beam can also be intercepted between the output mirror and photodetector assembly. It's not going to burn anything but is still a decent amount of power for leakage from a super high quality high reflector mirror with a short discharge path HeNe and super low current. Well, only perhaps 6 photons per second. Actually a few µW, But that's enough for the Detector PCB to do its thing. :)

    The getter appears to be nearly perfect with just the slightest feathering at the edge, which is a testament to the fabrication to be like new after 25 year. Various methods have been used to seal RLGs including special low temperature metals and glass frit. It's difficult to tell by inspection which of them - or multiple types - are used here. The internal volume is quite small with little gas reserve space, so the cathode may be very special with its limited surface area.

    Many more photos of this assembly can be found in the Laser Equipment Gallery (Version 4.90 or higher) under "Helium-Neon Ring Laser Gyros".

    All Honeywell RLGs are triangular and the basic appearance of the laser blocks is very similar. Some other manufacturers opted for a square geometry since that has a larger enclosed area (and thus higher sensitivity) for the same overall diameter. But those advantages are offset by the added complexity and mass of the block, and the requirement for a 4th mirror and its scatter, losses, alignment, and sealing.

    Overall description and testing

    See Honeywell GG1342 Ring Laser Gyro with Parts Labeled. It is recommended that this photo be open in another window while referring to the description below.

    While the removal of the cover on the first sample involved a moderate amount of abuse (and likely some selected 4 letter words), I was not involved. :) But knowing was was inside made getting inside the second one much easier. Note that since the laser itself is not being attached, no expensive chants and incantations are required. ;-) The GG1342 consists of a thin steel outer cover in two pieces: One is mostly flat covering the bottom while the other is form-fitting for the top. They are attached via really tough adhesive. The top one must be removed or at least the seam around the inner cover must be exposed to get inside. It might be possible to use a heat gun or welding torch :) to soften the glue, but that could be risky. And while any labels will be distressed at the very least, the simplest approach is to peel it off all around to expose the seam. The, with the entire RLG clamped in a vice, a hacksaw or rotary ("Dremel&trade") tool can be used to create a slot the end of the narrow section. A flat blade screwdriver is then able to pry the top and bottom apart without damage to anything inside. Just limit the incursion to no more than 1/8 inch.

    Popping the inner cover reveals everything in all its glory, that is, until it is powered. ;-) Besides the resonator block, there is a circuit board for the photodetectors as well as another one that is associated with powering and controlling the HeNe laser discharge, and a potted high voltage module directly connected to the ballast resistors. The HV to the HeNe discharge paths in particular is somewhat peculiar, though the actual HV power supply is NOT part of this assembly. More below. There is also the interconnect wiring (mostly via a Kapton ribbon cable) and the main connector that goes to the outside of the case attached to a small board with a few resistors and a connector for all the PZTs and a lone sensor (probably for temperature).

    If having viewed the photos and read the description, you have something to correct or add, or can come up with a manufacturer and even better, a model, technical specifications, complete schematics, and an operation and service manual ;-), please contact me via the Sci.Electronics.Repair FAQ Email Links Page.

    Triple GG1342 HeNe laser power supply PCB

    This board came with the second GG1342, probably suspected as being faulty. Since there are 3 pairs of high voltage cables with connectors that match those on the GG1342, it is assumed to power the HeNe laser discharges in all 3 RLGs in the Honeywell Inertial Measurement Unit (IMU). See: Triple Output HeNe Laser Power Supply for IMU using the Honeywell GG1342. However, little is known about this board other than that the connectors mate with those on the GG1342. ;-) Due to the lack of any power transistors to drive an inverter, the input is probably something like 28 V at 400 Hz. The 3 sets of 2N5416 transistors at the lower right are probably the linear regulators for the total current to each pair of discharges in the 3 RLGs. The current balancing is then done by the Laser Regulator PCB in each RLG. Additional photos of this board can be found in the Laser Equipment Gallery (Version 6.02 or higher) under "Helium-Neon Ring Laser Gyros".

    Custom power supply box

    A stand-alone power supply for the laser was built into a small lab-style HeNe laser power supply case. See Honeywell GG1342 Ring Laser Gyro with Custom Power Supply Box. The operating current is provided by a pair of defective bricks from HP-5517 lasers that have lost their regulation. They use a self-oscillating inverter and the operating current is now a function only of input down to a very low voltage. (That may also be true of undamaged bricks of this particular model up to the point where regulation kicks in, but without regulation these were useless for their intended application.) The RLG HeNe runs down to below 0.5 mA total current (0.25 mA/discharge) and there is no commercial supply that goes that low. The input to the box is 12 VDC and an inexpensive DC-DC buck converter provides an adjustable voltage from 0 to 7 V to the brick. This results in a maximum total laser current of approximately 3 mA. But the two bricks do not behave quite identically, expecially at these low currents, so an additional ballast resistor had to be added for one supply and a diode and resistor in series with the input to the other to balance the currents. Running both discharge paths from a single brick would have avoided this, but since the brick already provides a starting voltage, one of the two discharges would always strike. However, starting the other one at low current became tricky, requiring a starting voltage to go in via the long HV connector. That was actually my original intent with a momentary toggle switch to apply the starting voltage, but starting was not consistent, cause unknown. And a disadvantage was that as designed, it required operator attention. So on to Plan B with the dual supplies. ;-) A digital panel meter monitors discharge current. And there are LEDs for DC Power and one below the meter whose brightness is proportional to the adjustable voltage from the DC-DC converter to the bricks for the operating current. And so as to not waste a perfectly good hole :), the momentary toggle originally intended for starting can apply 8 VDC from a 3-terminal regulator to the supplies just as check. This results in around 4 mA total laser current.

    Now with a proper power supply :), laser output power versus current was measured by removing the Detector Assembly (which was a bit of a pain to realign after replacement). Here are the data:

      Total Current    CW Beam    CCW Beam  CCW Sample
         0.5 mA         1.9 µW     2.0 µW     1.0 µW
         1.0 mA         4.2 µW     4.1 µW     2.1 µW
         1.5 mA         6.5 µW     6.3 µW     3.0 µW
         2.0 mA         8.0 µW     7.5 µW     3.6 µW
         2.5 mA         9.0 µW     8.9 µW     4.2 µW
         3.0 mA        10.8 µW    10.2 µW     5.0 µW

    The dropout current is actually below 0.25 mA for each discharge (WOW!) and it still lases at minimum current. If additional ballast resistance were added, it probably could go much lower. The starting voltage is around 2.5 kV.

    "CCW Sample" is split off of the CCW beam and used to monitor laser power. There's also a "CW Sample", but that beam is blocked by a patch of white paint and not used. Interestingly, in this regime of very low current, the output power is approximately proportional to current, between 5 and 6 µW/mA/discharge. The discrepencies between the entries are almost certainly mostly due to measurement error at this very low power. Thus at the maximum current setting of the power supply here limited to 3 mA, the actual total output power is over 30 µW (although part of that is blocked). I take it back. Perhaps it could burn a minutely tiny something. ;-)

    Testing the Detector PCB

    After wiring a connector to fit the header H1 and double checking power and ground connections to the LM119H the board was reinstalled on the RLG assembly and reconnected to the Detector Assembly. Using 12 VDC from the power supply box and a dual channel scope, the thing actually works, sort of. ;-) Just pushing on the very solid workbench results in a response. With the RLG assembly on a ball-bearing turntable, pulse trains can be readily detected with even the slowest manual rotation. However, the two counter-rotating beams are probably locking together since the Earth on which it sits is also rotating, very slowly and there should be fringes changing at a few Hz without touching anything. Of course the dither motor and mirror PZTs are not being driven so this isn't totally unexpected. What is a bit strange is that both outputs normally sit at a low level and only pulse high with a detected change in rotation. This suggests that something on the Detector PCB is AC-coupled, perhaps expecting the fringes to be dithering. The outputs are from the LM119H dual comparator so they are digital signals. But they have a huge tendancy to oscillate on their own, made much worse if an additional pullup resistor is added. There is also a header pin that goes to the output of the Harris chip connected to one of the sensors. This does show a fringe waveform - but only if the rotation is fast enough. The overall detection scheme is so sensitive that without covering the area of the sensors, the light from the overhead fixture results in pulses at the output, so the area of the detector assembly was covered with black tape. This is probably working correctly.

    And there is a response with a laser current below 0.5 mA total (!!), though the behavior is more stable at around 1 mA and is generally similar up to at least 2 mA. However, there may be an optimal current because there are subtle differences in the waveforms and consistency between the two outputs. And this changes over time. So something may be going on with respect to laser output power as well as single longitudinal mode operation. Unfortunately, it would be rather challenging to actually monitor the laser beams themselves with a Scanning Fabry-Perot Interferometer (SFPI). Aside from having to remove the Detector Assembly again and fight with realignemnt after replacing it again, 1 or 2 µW of beam power is somewhat marginal for a decent response. A combination of feedback control of the laser current and voltage to the mirror PZTs may also be required to maintain single mode operation. And they are generally called PLC (Path Length Control) PZTs. I'm definitely not going to attempt driving all the PZTs at the same time. :( Not enough hands. :)

    It turned out that 12 VDC was too high for the power to the Detector PCB resulting in the excessive noise on the waveforms, so an adjustable regulator was added along with pullups on the two comparator outputs. Around 5 or 6 VDC results in much cleaner transitions and 100K ohm pullups provide a flat response to more than 100 kHz. But for whatever reason, the default output state is now high instead of low probably due to changes between the sensor outputs and comparator thresholds. The supply voltage also appears to be rather critical for proper response from both channels, so it may have also been feedback regulated based on laser power or behavior.

    Honeywell GG1342 Ring Laser Gyro Fringe Detector Outputs A & B shows the outputs for CW and CCW rotation. For CW rotation A leads B while for CCW rotation A trails B. A video might have been better but attempting to make something meaningful without a programmable rotation stage proved impossible.

    At this point, the A and B signals could have been sent to a measurement display like µMD1 to track the rotation (at least at modest angular velocity). (µMD0 and µMD2 did not exist at the time of these tests.) The behavior with no dither would have been similar to that of a home-built RLG. See the section: Sam's Home-Built Ring Laser Gyro 1.

    The Detector PCB doesn't appear to implement anything very special. If the sensors were simply a pair photodiodes, basic high gain trans-impedance amplifiers could be substituted for the unidentified Harris parts (which are probably just op-amps). But the sensors are still a slight mystery since their forward voltage is slightly high for silicon diodes. However, assuming them to be peculiar photodiodes would probably not be that far off.

    More not so hairy math

    For the GG1342, the triangular ring has sides of approximately 12.5 cm, 10.6 cm, and 12.5 cm for a perimeter of 35.6 cm or 0.356 m. Its area is 60 cm2 or 6.0x10-3 m2. For testing, n is equal to 1. Calculating for 1 degree/second as an example, the result would be given by:

               4 * 6.0x10-3       1
       δf = ----------------- * ----- = 1.86 kHz
            6.33x10-7 * 0.356    57.3 

    Where the last term is the conversion from degrees to radians. Although I currently have no way to measure this precisely, moving the turntable by hand results in waveforms that appear to be consistent with these calculations.

    The Sagnac frequency for this RLG sitting flat on the table with it's axis vertical due to the rotation of the Earth is given by:

      δf = ---- ΩEsin(φ)


    If the RLG was at the North pole :), δf would be around 7.74 Hz. With the RLG at the latitude here (39.95°, the expected δf drops to 4.97 Hz. But this is not being detected, most likely due to lock-in without dither. 0However, applying up to 20 V p-p from a function generator to the dither motor or mirror PZTs over a range of frequencies from 1 Hz to 1 kHz resulted in the dither frequency showing up as pulses, but there was no correlation with the expected 4.97 Hz due to the Earth's rotation. With up to 140 VAC at 60 Hz applied to the dither motor (through a resistor for protection!), there was just the barest hint of unlocking, but still nothing unequivocally correlated with the Earth's rotation. Now having experienced the dither of the GG1320 (see the next section), the amplitude needs to be much higher on the order of 250 arcsec for more than 100 (X4) fringe counts to minimize the effects of the dead zone. To achieve this requires running near the angular resonant frequency of the block with respect to the spring constant of the dither motor core.

    While the lock-in threshold (ωLock) cannot be calculated without knowing a parameter which is unknowable for the intact laser :), we can make an estimate. The simplified equation for ωLock is given by:

      ωLock = -----


    The scattering coefficient is basically a measure of mirror quality. and can only be estimated at best without direct access to the mirrors, and only with fancy instruments even with direct access to the mirrors. ;-). (Calling it a wild guess might be more accurate.) For common HeNe laser high reflector mirrors, 10-4 is a typical value. But a high quality RLG like this, the mirrors will have super-polished substrates and special coatings having an r value that's likely one or two orders magnitude lower. If r is 10-5, ωLock for this RLG would be 0.72 degrees/s. This could be credible based on behavior. Ultimately, confirmation may require a turntable with programmable rotation. ;-) However an ad-hoc test was done rotating it by hand. The diameter of the turntable is 12" (30.48 cm) for a perimeter of 37.7" (95.8 cm) so 1 degree of rotation is 0.104 inches (2.65 mm). A movement of 1 degree/second is picked up easily, and possibly somewhat lower, though probably not an order of magnitude lower. When the RLG is in a good mood, it simply is not possible to touch it without evoking a response. And pulses may continue to show up for several seconds after just tapping on the workbench. But even if ωLock was only 0.072 degrees/s, that's still more than an order of mangitude larger than the Earth's angular velocity of around 0.004 degrees/s, so it's not surprising that the RLG is blind to it. To get a lower ωLock requires even better mirrors or a larger enclosed area or both. Experimental RLGs are generally huge compared to this one - typically 10X larger at 1 to 2 meters on a side. But some truly gargantuan experimental RLGs have been built, at least one with dimensions of 39.7 x 21 m on a side. ;-) For that one, the Sagnac frequency for the Earth's rotation is 2.18 kHz. Bigger is better as far as both increasing δf and decreasing ωLock. Though using one for general navigation would be somewhat limited, at least for anything with dimensions much smaller than that of a football field. :)

    Honeywell GG1320 Ring Laser Gyro

    The GG1320 is the only RLG model currently on Honeywell's Web site in 2018 though others like the GG1342 and GG1308 may be part of inertial platforms and not offered separately. However, those I could locate either don't specify a model at all or list the GG1320. See Honeywell Digital Ring Laser Gyroscope. Also there under "Related Downloads" is a separate brochure with specifications as well as a user manual with power requirements, connections, and communications protocols to acquire orientation data. "Digital" here means that unlike the GG1342 which provides only fringe data, the actual angles are returned via RS422 so all processing is internal to the GG1320 enclosure. The entire device is only 3.45 inches overall which is less than half the diameter of the GG1342, and weighs only one pound including all the electronics. So you could put a GG1320 in your car to keep the GPS honest or three of them along with accelerometers if you're a true inertial navigation geek. ;-) Or on your bike or scooter. One such unit is shown in Inertial Platform with Triple Honeywell GG1320 Ring Laser Gyro. This unit had no make or model but is typical, around 6x8x8 inches overall and 5 pounds. So it would fit, mostly. However, the more compact HG1700 IMU would be easier to mount. ;-) See the info below.

    Overall description

    Honeywell GG1320 Ring Laser Gyro shows the intact unit next to an inch ruler. It has a diameter of only 3.45" and a height of only 1.77"and weighs in at 1 pound. The leg lengths in the laser block are 2 inches (thus the "20" in GG1320). The specifications state that there are 1,164,352 (X4) counts in 360 degrees or 1.113065 arcsec/count. (There is a discrepency of around 1 part in 400 compared to the calculated values for the 2 inch leg length equilateral triangular laser cavity assuming a wavelength of 632.8 nm. This remains a mystery. Perhaps one or all of the legs are not exactly 2 inches.) The only disassembly via fasteners are the sheet metal covers on top and bottom which are secured with single tiny flathead screws. They do not expose anything inside and only serve as a magnetic shield. The guts of the RLG are housed in precision milled aluminum castings, apparently press-fit together and thankfully not laser-welded. :) I feared this would be more difficult to enter than the GG1342, above, where the top cover was formed 16 gauge aluminum and some serious grinding with a cutoff wheel was required to remove it. Or that potting compound would be involved. However, an improvised spreader tool did the trick with little fuss as there were convenient recessed slots in which it could be inserted. At most, there was perhaps some miniscule amount of adhesive that was doing most of the work since putting it back together required almost no force. And a single-edge razor blade was then sufficient to separate it again. This would probably have been similar for the GG1342, but it did not have any convenient recesses for a spreader tool to grab. See Honeywell GG1320 Ring Laser Gyro First Level Interior. The PCBs do differ from the photo on Honeywell's Web site, possibly being an earlier or later version. Next, removing the 6 screws securing the black plate and the 3 screws in the center (which actually go all the way through the RLG dither motor core to the bottom of the case) frees up the (drum roll please) Honeywell GG1320 Ring Laser Gyro Laser Block with its pristine shiny getter.

    Later, I found from the manual that it is supposed to be in an "environmentally sealed case filled with a dry nitrogen gas mix". Just as well that wasn't known before this stunt. ;-) But I'm somewhat suspect that the seal was all that good. Perhaps "environmentally sealed" means "more or less sealed well enough to pass QA". :)

    Many more fabulous photos of this unit can be found in the Laser Equipment Gallery (Version 4.95 or higher) under "Helium-Neon Ring Laser Gyros".

    For the following, please see Honeywell GG1320 Ring Laser Gyro with Parts Labeled. It is recommended that this photo be open in another window while referring to the detailed description below.

    There are no active devices associated with the block unless one includes the laser itself. Note the inch ruler: The block is only about 3 inches across for the 2 inch leg lengths. It is not actually attached directly to the black separator plate underneath it, but is suspended from the dither motor core, which is normally secured with the 3 screws through to the base. There are a total of 18 power lines and signals to it via the two long white connectors and that's all. The wires are extremely flexible so that they won't hamper the motion imparted by the dither motor, or break due to metal fatique. Unplugging the connectors means the block is completely detached and could in principle be a field replaceable part (though of course that would be insane). Here are the PCB to Block connector pinouts. There are no identifying marks for pin 1, so pins are labeled from top to bottom where "top" is defined where the connectors are closer together near the Detector Assembly. And only pins that are present are labeled without regard to empty spaces. BC1 is on the left with the block down toward the separator plate:

       Pin    Color   Description
      BC1-1   White   Fringe detector pin 1 cathode 1
      BC1-2   Black   Fringe detector pin 2 anode
      BC1-3    Red    Fringe detector pin 3 cathode 2
      BC1-4   White   Laser Power Monitor pin 1 anode
      BC1-5   Black   Laser Power Monitor pin 2 cathode
      BC1-6   White   Dither motor drive
      BC1-7   Clear   HeNe laser anode 2 (red tip-off)
      BC1-8   White   Mirror 2 PZT front left
      BC1-9    Red    Mirror 2 PZT Rear
      BC1-10  Black   Mirror 2 PZT Front right
      BC2-1   White   AD590 pin 1
      BC2-2   Black   AD590 pin 2
      BC2-3   Black   Dither motor return
      BC2-4   Clear   HeNe cathode
      BC2-5   Clear   HeNe anode 1
      BC2-6   White   Mirror 1 PZT front left
      BC2-7    Red    Mirror 1 PZT Rear
      BC2-8   Black   Mirror 1 PZT Front right

    The connections to the cathode and both anodes are direct - 0 ohms. A pair of 38K ohm SMT ballast resistors on the PCB next to the white connectors go directly to the anode pins. There appears to be additional ballast near the HV power supply - several 500K ohm resistors, possibly as much as 2M ohms for each side! But tracing the wiring through to them from the block is not something I intend to do. More below.

    For the most part, the GG1320 laser block is a miniature version of the one in the GG1342 with the dual anodes, two mirrors with PZTs behind them, the third mirror with the fringe detector and laser power monitor (though in this version, glued and globbed in place with a wedged mirror substrate and no corner prism), dither motor, and temperature sensor.

    The electronics consists of two PCBs. The "Processor" PCB is visible immediately after opening the case while the "Power" PCB is underneath it as shown in Honeywell GG1320 PCBs. The HeNe laser power supply is almost certainly on the Power PCB. It probably involves the power transistor and large Precision Inc. part. Precision Inc. is a manufacturer of magnetic components like transformers. :) That part may also be for driving the PZT(s). Both PCBs are loaded with discrete components and a few small ICs. There are 2 LSI parts (68 and 100 pins) with Honeywell house numbers on the bottom of the Processor PCB. Everything is surface mount. See the closeup photos of the PCBs in the Laser Equipment Gallery.

    Examining the GG1320 photo on Honeywell's Web site, it looks like they have the power PCB directly above the separator plate (rather than behind the Processor PCB in the lid) with different connectors to the laser block. And the "hole" in the PCB is not in the same location as the one in the separator plate on mine suggesting that the laser block itself may also not be quite the same. See below under "Initial system tests".

    Only a few parts have labeling that may be date codes. One type is the AD822 dual op-amp, one of which is in the lower right of the first PCB closeup - 0046. This could be week 46 of the year 2000. Another one has 0043. A part on the assembly from which the GG1320 was removed has a date code 2401 which could be week 24 of the year 2001. So probably this is in the 2001 time frame. Darn, and I thought it would still have that new RLG aroma. ;-) Now if I had only read the manual first: "3.6.1 Labeling - The laser gyroscope serial number is comprised of a four-digit date code followed by the six-digit laser block serial number. The date code consists of the last two digits of the year followed by the week of manufacture." For this unit, the RLG SN is: 0112182119. So, it's week 12 of 2001.

    Laser tests

    Powering the laser block turned out to be a bit strange through no fault of its own. I expected this to be even easier than for the GG1342 since the discharge voltage should be lower. But the behavior was very peculiar using my custom RLG power supply box which runs the GG1342 perfectly. (See the info on the GG1342, above) with 470K ohm ballast resistors resulted in only one side lighting most of the time even though they are powered from separate bricks. Increasing the power only resulted in increased current to that side and the opposite side flashing. Although the same side generally started and ran, it would sometimes swap with similar behavior. And perhaps 1 in 10 starts both discharges would light. The current could then be reduced to around 1 mA total and be stable. The ballast resistors on the PCB near the white connectors are only 38K ohms, that's is just the ballast close to the laser; there could be additional resistance near the power supply itself but tracing that would be quite convoluted running on over and through 2 surface mount multilayer PCBs. :( :) It did try powering it with up to 1M ohms total with little change. Go to Plan B.

    So testing was repeated using a pair of Hughes 3595H bricks. These are self oscillating inverters similar to the broken HP power supplies used in the custom box, but driven from a variable linear power supply. Each ballast was a 47K ohm resistor directly at the anode in series with a 1M ohm resistor (in addition to whatever is inside the bricks). This setup is much better behaved. :) The starting voltage is close to 2 kV and the operating voltage is around 400 V (at 1 mA per discharge) (monitored with a vintage Simpson 260 on its 5kV range at the junction of the 47K and 1M ohm resistors). The dropout current is below 0.15 mA (!!) for each discharge. And it's possible to make out tiny red dots on the inside surfaces of the mirrors at a current of below 0.25 mA, though I cannot tell if they are there down to 0.15 mA because of the glow of the discharge. There is no convenient spot to poke a white card in the detector assembly since it's globbed over with gray stuff. That's technical term. ;-) But it appears as though the GG1320 laser is healthy.

    The peculiar behavior was probably an artifact of the way the power supply box works with the very different characteristics of these tiny lasers. But an explanation is not forthcoming. And no, it didn't somehow go bad. It still works perfectly with the GG1342.

    Initial system tests

    Since the GG1320 is fully documented on the Honeywell Web site including the operation manual - which is truly remarkable - interfacing with it should be straightforward. Power is +15 VDC (150 mA continuous, 300 mA max, probably for starting the laser) and 5 VDC (150 mA). It only needs a reset pulse and then a Sample Request for packets of angle data. A dual power supply, push button, and LED should be sufficient to determine if it is alive.

    The GG1320 Ring Laser Gyro main Connector is a Cinch Micro-D part number 0046 10154909-101. At first, it looked like locating an affordable mating connector would be a problem even if it didn't have to be deep-space-qualified. ;-) But eBay came to the rescue with a Glenair MWDM2L-25P-4E7-48K, which based on the 333 page Glenair databook :) is the perfect mate. And it even has wires already attached. Not cheap even on eBay but given the amount of work that would save is more than worth it. Unfortunately, Glenair apparently is not familiar with the resistor color code. While the color sequence is correct, it begins with black for pin 1. :( So, it's essential to quadruple check all connections.

    Here is the connector pinout for the GG1320 from the Honeywell manual with the minimum required hookup to reset it and request data:

       Pin   Description           Connection
        1    CGND                  Ground    
        2    Spare                 x
        3    Master Reset Loop     Connect to pin 16
        4    +5 VDC                +5 VDC Power, 150 mA max
        5    Spare                 x
        6    Spare                 x
        7    LGND                  Ground
        8    User Sample Request   Sample clock
        9    Spare                 x
       10    Spare                 x
       11    Factory Mode Enable   Do not connect
       12    VPP (Program Enable)  Do not connect
       13    CGND                  Ground
       14    Serial Data Out       Serial Data out
       15    +15 VDC               +15 VDC Power, 300 mA max
       16    Master Reset In       Connect to pin 3
       17    AGND                  Ground
       18    Dither Square         Do not connect
       19    Spare                 x
       20    Clear to Receive      Do not connect
       21    Clear to Send         Tie to logic high (+5 VDC)
       22    Gyro OK Monitor       Low indicates failure
       23    Spare                 x
       24    Spare                 x
       25    Serial Data Input     Do not connect

    As can be seen, very few signals are actually required to make this thing sing. (How prophetic that phrase is will be seen below.) It is believed that the +5 VDC is for the logic and the +15 VDC is for the HeNe laser power supply and probably the dither motor. The only required input signal is User Sample Request which triggers the generation of a data packet on its rising else. Jumpering the two RESET pins together will enable a power-on reset. Clear to Send should be tied high. The outputs are Gyro OK Monitor and Serial Data Out, 1M baud RS422 (TTL levels from GG1320). Just to determine it it's healthy only requires power, the reset loop jumpered, and Gyro OK status. Here is the list of only the required connections:

          Pins      Description    Connection
        1,7,13,17   GNDs           Power supply return and signal grounds
          3,16      Reset          Jumper pins 3 to 16
            4       +5 VDC         +5 VDC Power, 150 mA max
            8       Clock          Sample request, TTL rising edge
           14       Data           Serial Data Out
           15       +15 VDC        +15 VDC Power, 300 mA max
           21       CTS            Tie to logic high (+5 VDC)
           22       Status         Gyro OK if high.

    The basic Interface board was constructed first with a 9 pin connector to the RLG for power/control/status, and a 3 pin connector to the RLG for the communications. Here is the pinout for the RLG cable (for my own reference):

                      RLG   Wire
      Pin   Function  Pin   Color    Description/Comments
      P-1   +5 VDC     4    Orange   LM7805 regulator from +15 VDC
      P-2   +15 VDC    15   Yellow   15 VDC 0.9 A from wall adapter
      P-3   Ground     1    Black    CGND
      P-4   Ground     7    Blue     LGND
      P-5   Ground     13   Red      CGND
      P-6   Ground     17   Blue     AGND
      P-7   Reset-     3    Red      Master Reset Loop
      P-8   Reset-     16   Green    Master Reset In pushbutton
      P-9   Status     22   Brown    Gyro OK Monitor green LED
      C-1   CTS        21   Black    Clear To Send
      C-2   Clock      8    Violet   User Sample Request
      C-3   Data       14   Orange   Serial Data Out

    And as noted above, the wire color versus pin number of the pre-made cable is screwed up. :( :) Did you hear that, Glenair? ;-) With pins 7 and 8 tied together through a 1K ohm resistor, Reset is automatic at power-on, but a pushbutton was provided to force reset at any time. That turned out to be useful. The Status signal drives a super bright green LED through a 2.2K ohm resistor (just to be safe - while the input and outputs are TTL levels, there is no specification for logic drive capability). At less than 2 mA it's still plenty bright.

    To be able to observe the exciting action during initial testing, the laser block was plugged into the Processor PCB and secured to the separator plate with nuts and bolts, but not to the bottom of the case. First light on internal power is shown in Honeywell GG1320 Digital Ring Laser Gyro Testing. The laser comes on with an estimated current between 0.25 and 0.50 mA/side based on the appearance of the discharge. The Gyro OK status LED lights for a few seconds but then goes out. During that time, there is an audible whine from the dither motor, struggling to accomplish something - whining and moaning. So perhaps it's not happy with the low mass of the plate on which the block is bolted. Time to button it up and hide the beautiful glow forever. :)

    And in fact, once reassembled, the green LED remains on. But the whine from the dither motor could take the hair off one's teeth. Think constant multiple fingernails on a chalkboard. Geez. An inertial platform would have 3 of these in a chorus at slightly different dither frequencies. :) Apparently, silence is not one of the RLG's attributes.

    According to the manual, the dither "Zero to Peak Amplitude" is spec'd as 250 arcsec. The p-p excursion would then 500 arcsec or about 0.129 degrees. 500 arcsec is about 450 (X4) counts. A rough calculation shows 500 arcsec to be a p-p excursion of around 0.005 inches at the periphery of the laser block. So that could explain the annoying sound and vibration. The manual also states that the GG1320 needs to be rigidly mounted to a structure with at least 100 times its moment of inertia. So, just sitting on the table may not be sufficient (though it wasn't complaining if left alone).

    And something interesting: When the laser starts, there is a momentary flash of gorgeous blue light inside the block near the area of the detector, probably a blue LED around 450 nm. Not even a microscopic part is visible on the top or bottom of the block, so the LED must be in the detector assembly itself shining through Mirror 2, powered somehow from two of the photodiode pins. And no, it isn't arcing. :( :) Providing light to aid in HeNe starting is a well known technique and is mentioned in at least one of the RLG patents and the GG1308 information, below. Since this unit is from 2001, that would be not so long after the development of the first practical blue LED in the 1990s. Red or green LEDs (or even incandescent lamps before LEDs) also work for this purpose so it's not clear why they used blue except that theoretically, the shorter wavelength may be superior. There is also a ~1/4 inch hole in the separator plate directly above the area of the LED. It's not clear what purpose that serves unless to make the LED's operation visible when the GG1320 is split open with jumper cables for testing.

    Data testing

    Now that it has been determined that the laser comes on and the Gyro OK bit stays high, time for the basic data interface. A 555 timer to generate the User Sample Request pulses was added to the Interface board. A scope was used to check that data is generated. Here is the pinout for the Data cable (again for my own reference):

      Pin   Function   Description/Comments
      D-1   +5 VDC     Power
      D-2   GND        Return
      D-3   Reset-     Master Reset In-
      D-4   Status     Gyro OK
      D-5   CTS        Clear to Send
      D-6   Clock      User Sample Request
      D-7   Data       Serial Data Out

    The manual provides a detailed description of the Serial Data Out format. The rising edge of Sample Request initiates the return of a 6 byte packet:

      Byte   Description   Comments
       1     Gyro Status   Errors and other indicators of gyro health
       2     Tag ID        Specifies which variable appears in Tag Data
       3     Theta LSBs    LSBs of cumulative (X4) counts since last Reset
       4     Theta MSBs    MSBs of cumulative (X4) counts since last Reset
       5     Tag Data      Value of the tagged variable specified by byte 2
       6     Checksum      1's complement sum of bytes 1 through 5

    The Status byte includes a 2 bit field for Normal, Caution, Warning, and No Computed Data, along with some flag bits. The Tag Date byte returns non-real-time information like serial number, total operating time, operating time since last Reset, temperature, etc. determined by the Tag ID value which increments after each packet is sent. The actual angle (Theta) is a 16 bit number spanning Bytes 3 and 4 (20 bit first) representing the cumulative (X4) counts from the fringe decoder since the last Reset. The Checksum byte can be used to confirm reliable data transmission and as for framing the six bytes as they come in.

    It's possible to get an idea of whether the angle (Theta) is changing in a predictable manner from the scope display alone by just observing Bytes 3 and 4. The bits are arranged with the LSBs on the left so as Theta increases or decreases, the appearance in a mirror :) would be that of a 16 bit binary number incrementing or decrementing, but with a gap between the two bytes.

    And it's alive! With my trusty vintage Tek 465B set on delayed sweep, bytes 3 and 4 can be centered and expanded to fill the screen. However, not entirely unexpectedly after reading about the required moment of inertia of the support structure in the manual, the GG1320 had to be screwed down. Otherwise, Gyro OK might eventually go off and stay off if the GG1320 was touched or moved, though data continues to be sent regardless. If my eyeballs were fast enough, the fault could be determined from the Status byte. ;-) With that resolved, the show can continue.

    Scope Display of GG1320 Data Packet Bytes 3 and 4 with Earth and RLG Rotation is a mediocre video of the scope screen showing all the action. From left to right on the screen there is a Low for the start bit of byte 3, bit values 20 through 27, a High for the stop bit of byte 3 followed by a Low for the start bit of byte 4, bits values 28 through 215, and finally a High for the stop bit of byte 4. Bits beyond 215 are not sent - software must take care of carry/borrow to extend the number beyond 16 bits. The times are approximate:

    As can be seen if reading binary numbers backwards or in a mirror is something at which you are proficient, Theta increments and decrements nicely when the GG1320 is rotated. And when the GG1320 is stationary, the rotation of the Earth is clearly visible as Theta increments in the LSBs near the left of the screen. Based on a careful analysis of the scope screen (with only the occasional use of a loaded die), the results are in:

           |<---- Scope ----->| |<---- Binary ---->|
      Time . Byte 3 .. Byte 4 . C/B MSB   Word   LSB  Decimal  Change     Angle
       0 s _-__------_---__--_-  0  0110011111111001  26,617        0    0.000 deg
       5 s _-----___-____-_--_-  0  0110100000011111  26,655  +    48  + 0.015 deg
       9 s _--_-_-__-___---_-_-  1  0101110000101011  89,131  +62,466  +19.318 deg
      10 s _--___--_-___---_-_-  1  0101110001100011  89,187  +    56  +19.335 deg
      15 s _--__---_-_-___---_-  0  0111000101110011  29,043  -60,144  + 0.749 deg
      18 s _----___--_-___---_-  0  0111000110001111  29,071  +    28  + 0.758 deg

    The "." are the start (0) and stop (1) bits and C/B is the carry/borrow from the MSB. Thus the actual value consists of bytes 3 and 4 (flipped) along with C/B to form a 2's complement number. It just happens that for this test, the angles are always positive. With a maximum of 65,536 unique values sent, the data wraps almost 17.8 times in 360 degrees. So C/B would extend it to handle however many full barrel rolls are anticipated. :-)

    If the GG1320 were sitting on a table at the North pole, the Sagnac frequency δf due to the Earth's rotation would be around 3.38 Hz. At the latitude here of 39.95°, δf drops to 2.17 Hz. Multiplying by 4 for the GG1320 X4 resolution results in 8.68 counts/second, which is consistent with the observed count rate. It's comforting to know that the Earth is still spinning correctly as advertised. ;-) The larger than expected change from 9 to 10 seconds when the GG1320 is supposed to be stationary is almost certainly because I was touching it. And as further confirmation that the slow counting is indeed due to the Earth's rotation, with the GG1320 tipped at around 40 degrees away from true North, the counting slows to a virtual stand-still with just some randomness in the LSB due to vibrations, and reverses direction when tipped beyond 40 degrees. When tipped the other way up to 50 degrees towards North, the counting is most rapid. Isn't Physics wonderful? ;-)

    Display interface

    At first, I was considering constructing an Arduino interface to a PC or one with its own color LCD display. It would need to be able to read RS422 or serial TTL at 1 Mbaud, format the data, and either send it to the PC or display it on the LCD. (Or both.) The $2 Atmega 328P Nano 3.0 would probably suffice since this is not exactly compute-intensive except for the loop that acquires the 6 bytes of data. Updating at modest rate like 10 Hz should suffice. However, there are many newer boards which are much more capable and include a color LCD. But I fear that the motivation quotient for this project will be lacking. Putting RLG technology on my road bike wouldn't be very practical. I don't own a boat or plane (or ever intend to) and GPS is fine for the car. And there isn't a huge market for DIY RLG-based inertial navigation system displays. :)

    So a simpler option would be to interface directly to a PC via a serial USB-to-TTL adapter. Then it would just be a matter of a basic GUI to format and display the data. The FT232RL USB-to-TTL adapter should be capable of up to 3 M baud and is less than $2 on eBay. It could run at a fixed update rate (determined by the 555 timer) or could use CTS as the User Sample Request. Initializing to 0 angle would be via software. The hardware Reset button would be needed only if the GG1320 screws up. :) This solution is much more likely to be implemented.

    (From: Zack Dougherty.)

    I came across a GG1320 laser and was inspired to see if I could measure the earths rotation with it. After googling the gyro to see what information was available, I found your site. As an uneducated reverse engineer this Web-page, was instrumental in helping me take on the personal challenge of getting it operational.

    I am using an Arduino to try to read the serial and decode the bytes into delta theta. This is my first project that has introduced me into binary, and between your web-site and the manual for the GG1320 I am struggling to understand how the device (or how to know in general) when overflow / underflow is flagged. Your test didn't go negative but it did subtract back to around 0 degrees. I am not understanding how the C/B is marked 1 or 0 based on your diagram or How the gyro defines underflow or overflow for counts above 65,536.

    With my known unknowns, I understand that this may be a question that has a large answer and do not wish to burden you with that as it might be very foundational knowledge. In any case I would greatly appreciate any information such as a link to point to relevant information on how to learn about it or anything you would be willing to answer.

    Here is some info on the system in this photo: Inertial Platform with Triple Honeywell GG1320 Ring Laser Gyro. A nearly identical version is found inside a Honeywell Mission Computer. The one I have was from a AW101 CSAR helicopter and is marked V (instead of 3 in the photo above), W,R,U,T,S. The dither rates on the set of LRGs are 523 Hz for U, 589 Hz for V and 625 Hz for W. The accelerometers inside are Honeywell Q-Flex QA2000s.

    Honeywell GG1308 Ring Laser Gyro

    The GG1308 may have been the smallest RLG ever built, or at least ever developed commercially with a leg length of only 0.8 inches, a block width of 1.5 inches, and a block weight of 2.3 ounces! :) The bulk of publicly-accessible information I've been able to locate may be found in Performance Evaluation of the Honeywell GG1308 Miniature Ring Laser Gyroscope, though a few research papers in peer-reviewed publications do cover it. The GG1308 was intended for "low performance" and "low cost" applications in drones and other remotely-piloted vehicles, and helicopters. Compared to its larger cousins, the GG1308 had lower accuracy, stability, and shorter life - and much lower cost. The G1308 is used in the Honeywell HG1700 Inertial Measurement Unit and Honeywell HG5700 Inertial Measurement Unit. The HG5700 is an enhanced performance version of the HG1700. Although the basic structure of the GG1308 is similar to other Honeywell RLGs, the laser block is made of common BK-7 optical glass rather than the ultra-low expansion Cervit, and other design and manufacturing techniques have been used to reduce the target cost to under $1,000 (in 1990s dollars). And while "repairable life" is quoted as 2,000 hours, tests suggested that laser failure after 200 or 300 hours might be more likely, due in part to the extremely small gas reserve in the block, though lifetime may have been improved since this report was written. The total discharge path length is also very small, order of 0.25 inches, making the gain and reserve power low to begin with. And the bore diameter is probably much larger than optimal for higest gain.

    Honeywell HG1700 IMU with Triple GG1308 Ring Laser Gyro

    The HG1700 IMU (Inertial Measurement Unit) is totally self-contained including 3 GG1308 ring laser gyros, 3 vibrating quartz accelerometers, and all the control electronics to provide attitude information in digital form. The entire unit is around 3 inches high by 4 inches in diameter and weighs in at under two pounds. It is intended for applications requiring less precision and shorter life than those like aircraft navigation. These include UAVs and UUVs (drones), as well as missile guidance. A Web search for "Honeywell HG1700" will turn up a variety of general information on the HG1700 including brochures and minimal specifications.

    I now have an HG1700 to analyze and document. The first challenge was to open it more or less non-destructively. There is a lock ring having 12 holes for a special spanner wrench which secures the resilient mount to the body of the HG1700, and the top and bottom of the unit as well. Of course that wrench is not exactly something one would find at Home Depot, and unlikely to be available from Honeywell. Even if it was, the cost would probably be typical aerospace stratospheric. :( :) Some careful pushing and prodding as well as modest tapping with the main housing clamped in a vice so it wouldn't get upset too much did nothing. So some destruction was necessary. My first action was to drill holes in the lock ring in an attempt to separate it into multiple pieces which could be more easily moved. But this didn't do much beyond resulting in ugly dents and holes in the lock ring, mount, and some on the edge of the main body, which can be seen in some of the photos. So on to Plan B which was much more successful. Plan B involved simply cutting the resilient mount ring at two places 180 degrees apart just shy of damaging the main housing and then prying it apart with a large screwdriver.

    This revealed the seam in the main body where the top and bottom were clamped together by the resilient mount, sealed with an O-ring. A single-edge razor blade carefully inserted in the joint and worked around the perimeter allowed for full separation in about 10 seconds. However, the portion with the sensors (RLGs and accelerometers) was still stuck to the section with the controller PCB via 5 ribbon cables which had to be carefully detached. And then, voila, we are inside. ;-) The RLG assembly is secured to the bottom of the case with 6 screws and the controller PCB is secured to the top of the case with 4 screws and one plug that goes directly to the external 30 pin connector.

    Numerous photos can be found in the Laser Equipment Gallery (Version 5.20 or higher) under "Helium-Neon Ring Laser Gyros" including the RLG blocks lit up in all their glory. ;-).

    There are three assemblies:

    1. Triple GG1308 ring laser gyro: This includes the three super small RLG blocks mounted precisely with their principle axes orthogonal, each with one anode and two cathodes, power sensor assembly (one photodiode) at one corner, small PZT behind mirror for cavity length tuning at the second corner, and the main detector assembly at the third corner. See Triple GG1308 RLG Assembly in Honeywell HG1700 Inertial Measurement Unit.

      The three RLG laser blocks are identical and are "machined" from common BK7 optical glass (not Cervit or Zerodur as used for the larger RLGs) and frit seals are used for all optical and electrical attachments. (Actually, in this unit, they aren't quite identical: One is fine ground on the outside and the others are polished, but I assume this has no cosmic significance, just that they came from different manufacturers or manufacturing lots.) Each has its own ribbon cable to a connector that plugs into the controller PCB. The whitish tabs with barcodes visible in the photo are the undersides of the RLG connectors. Disignations like L_01440713 are probably serial numbers but the meaning of ARW .06, ARW .07, and ARW .09 are a mystery. The anti-lock PZT dither motor is implemented using multiple crystals glued directly to the frame. Several of them are visible in this photo. A black potted module appears to control the HeNe laser current for all 6 discharge paths (two legs for each RLG block) via the cathodes, and also provides the drive for the main dither motor. It has a VMI part number. More on this below.

      • Anode and cathodes: These are just miniature versions of those in the other larger RLGs. The anode is simply a flat disk with the oversize tip-off at its center with a direct connection to the RLG cable; the cathodes are cups with the connections to the VMI module soldered directly to them.

      • Mirror 1 on PZT: The two solder pads go directly to the RLG ribbon cable and connector.

      • Mirror 2 with detector assembly: The detector uses the classic angled mirror to combine the CW and CCW beams to form a fringe pattern at the two photodiodes.

      • Mirror 3 with power sensor assembly: This tiny PCB is glued onto what looks like a conventional (though small) HeNe laser mirror. There are four solder bumps which go to its ribbon cable but only one pair shows any continuity and tests as a silicon diode. The function of the other pair is not known.

      • Discharge paths: With the block being under an inch on a side, the actual usable discharge path length (or "bore" of a conventional HeNe laser tube) is only about 0.25 inches total for each of the two cathodes.

    2. Triple vibrating quartz beam accelerometers: These are secured to the bottom of the case and attached to the controller PCB by one large ribbon cable that plugs directly into the controller PCB. They look like tall TO5 transistors with 6 pins. The markings on the case are: "ACCELEREX P/N 976-0513-004 CAGE CODE 0YFP0 Honeywell". A Web search doesn't turn up anything useful for this part number, but the Accelerex RBA500 appears to be similar and there is some info on those.

    3. Controller PCB: For now I will just say it has lots of SMT stuff including the a multiple winding transformer for the HV power supplies. ;-) There is only one obvious set of external connections identified which appear to be for 5 VDC power. However, according to a thesis dealing with tests of the HG1700 IMU, ±15 VDC is also required.

    Laser tests

    Although I originally intended to power the entire unit with the bottom section and aaccelerometers removed so the cheery glow of the lasers would be visible, this turned out to more involved than first expected. A thesis dealing with the HG1700 described the setup and power supplies: +5 VDC, and ±15 VDC. Locating the 5 V input appeared to be straightforward based on a large tantalum capacitor connected directly to two external pins. But nothing similar is present for the 15 V inputs. None of the other large tantalum capacitors are across two external pins Tracing the connections to parts likely using 15 V would be required. Since having the laser blocks tethered to the control PCB would limit the photographic options anyhow, it was decided to power them separately.

    Each GG1308 RLG laser block is independent with its own ribbon cable and connector except for the two cathodes that go to the potted VMI module. The anodes were easily traced to one pin of the connector but determining how the cathodes were wired through the VMI module was less obvious. Resistance tests didn't reveal anything - they all tested open indicating that there must be some semiconductors inside. However, applying +5 VDC through a current limiting resistor did show current flowing from each cathode to a pin on the VMI 10 pin connector. The amount of current was consistent with an internal resistor of between 200K and 300K ohms.

    Initilaly, I did manage to light up one of the RLGs by applying variable HV+ to the anode pin on its connector and HV- via separate 3.3M ohm resistors to the cathodes. That is, after possibly zapping the cavity length PZT because it appeared as though the pins near the anode pin on the connector were for the cathodes but they were not. :( It of course runs stable with way less than 1 mA but the two legs do not quite have have identical characteristics. While it's possible to get them to both light at high enough current, one will go out way before the other as the current is reduced. I can only assume they thing is lasing as it's impossible to see inside clearly enough to determine much of anything.

    Based on these measurements, a jig was built with nine (9) 3.3M ohm ballast resistors - one for each anode and each cathode. The value of 3.3M was chosen based on (1) the expectation that it would be large enough so that any discharge being lit would not greatly impact the input voltage and (2) that they were available. :) All the anode ballasts were tied together and connected to the +HV output of a Laser Drive 103-23 HeNe laser power supply brick. All the cathode ballasts were tied together and connected to the -HV output via a 10K ohm resistor to be used for monitoring the total current by measuring the voltage across it. The 103-23 brick was connected to a variable DC power supply since it was assumed that only a relatively low high voltage would be required and varying the input voltage would also allow the current to be adjusted.

    Honeywell GG1308 Triple Ring Laser Gyro Laser Test Circuit. The total high voltage was monitored using an ancient Simpson 260 VOM which has a 5 kV range. And the current was monitored using a DMM across the 10K sense resistor (RIS) so the calibration was 0.1 mA/V. RTA, RTC, and RTV provide isolation so that measurements will not significanly affect the stability of the discharge. Only the 3.3M anode and cathode ballast resistors are duplicated for the other two RLG laser blocks.

    Here are the approximate data for each RLG "laser":

    This rig worked out quite well. It was generally possibly to get all six discharges to light, though as noted, sometimes ramping the HV up and down a few times would be required, particularly from a cold start. Once lit, the discharges were stable down to less than 0.067 mA for each one, which was at around a total voltage of 1 kV. Starting required 1.5 to 2 kV. And that's how the glowing photos were taken. ;-)

    Stay tuned for more exciting developments.

    If anyone has information including operation, interface, and service manuals, please contact me via the Sci.Electronics.Repair FAQ Email Links Page. While I think it rather unlikely that actually doing anything more than possibly figuring out how to apply DC power without smoke, anything would be useful. For one thing, none of the connectors mates to anything in my junk box or anything I've ever seen before. :) The main connector on the outside of the case looks like a three-row DB31 but the (1) there is no such thing as DB31 and (2) the pin spacing is slightly larger than for the standard DB connectors. The internal connectors are also not exactly common (at least not outside the aerospace industry perhaps).

    Home-Built Ring Laser Gyro?

    So you want to build one? Good luck! :-) Or at least, be prepared for some challenges.

    Kits do exist, but if you have to ask, then they are definitely out of your price range. Think $5-figures (though the hardware can be replicated for much less). But the arm and two legs gets you complete instructions, supporting instrumentation, and a known working setup. For an educational institution or as a means to jump-start a research project on RLGs, the cost might be justifiable. Check out: Photonik Ingenieurburo Laser Gyroscope and Leybold P5885 Laser Gyroscope. The instruction manual for this one may be found at Keybold Laser Gyro Kit Instruction Manual.

    Building and aligning a HeNe ring laser is a challenge in itself, especially for someone who hasn't already even worked with an open cavity Fabry-Perot (linear) HeNe setup. Typically the ring will require a two-Brewster window HeNe laser plasma tube, mirrors that have the necessary Radius of Curvature (RoC) and reflectance at an angle of incidence of between 30 and 45 degrees, and a rigid but adjustable (and rotatable) mounting scheme for the tube, mirrors, and fringe detector. Once it's lasing, it must be forced to single frequency operation either by current-starving the discharge (which probably is only possible in custom RLG blocks with short discharge paths), mirror misalignment (which isn't very stable), the use of a frequency selective intra-cavity component like an etalon, and possibly required active feedback to maintain it. Then the interferometer or other detection optics and electronics must be implemented. And that's assuming you haven't moved onto another project and relegated this one to a box in your attic alongside the unfinished telescope mirror. :) But it is straightforward given availability to the key parts, at least to build something to demonstrate RLG principles. Much of what enable a commercial RLG to have such high performance is in the details, including that solid thermally stable block for the resonator structure, super polished mirrors, and the techniques used to avoid mode locking between the two counter-rotating beams.

    Having said all that, if you are still really determined to build an RLG, don't be discouraged. But do be prepared for what to expect. ;-) I would not recommend such a project for someone with little or no experience in lasers though. Simply aligning a ring cavity is much more difficult than a linear cavity. It may require incorporating a laser spectrum analyzer into the ring itself by putting one of the mirrors on a PZT. And assume whatever you come up with probably won't be useful for steering your Lazer sailboat. ;-)

    As a challenge, I have built an RLG similar to the one in the kits. It uses a Melles Griot 05-LHB-290 2-B tube with a 60 cm RoC HR and two planar HRs in a triangular configuration with leg lengths of averaging 10.9 inches. Optics similar to those used for homodyne interferometers are used combine the CCW and CW beams to generate the output signals. My design could be replicated, though there is nothing particularly special about it, being partially based on parts that were available and keeping it compact enough to be portable, sort of.

    Much more in the next section.

    Here are some contradictory opinions on DIY RLG feasibility and implementation:

    (From: Dr. Walter Luhs (

    You can build an RLG on a wooden plate (20 mm thickness) and rotating it using a stepper motor (micro steps) is sufficient, with a speed of a few degrees per 10 seconds. You do not need super polished mirrors.

    (From: Sam.)

    If I had realized the bit about wood when building mine (see the next section), I could have saved some money on the plate. ;-) The stability of even soft woods is actually much better than aluminum or most other common metals with a coefficient of thermal expansion that is 4 or 5 times lower. I would recommend cabinet-quality plywood (NOT flake-board) an inch or more in total thickness, which could be made from two pieces surface glued together. It might be worthwhile using small metal plates under the optics mounts to minimize deformation of the wood surface when they are tightened down. And staining and varnishing would provide that authentic period look. ;-)

    (From: Douglas P. McNutt (

    The mechanical precision is the hard part and that's what makes it virtually impossible for an amateur to construct a ring laser gyro. The two opposite traveling waves have to have extremely high spectral purity which translates to high quality, high reflectance flats at the corners. Not a home job.

    It might be easier to build a fiber gyro in which the light passes many times around an effective ring through a wound fiber.

    (From: Chris C.)

    The mechanical part is horrendous. We have an open cavity HeNe at my school's lab, and it is a challenge to keep lasing on a heavy damped breadboard with the mirrors mounted on a thick dovetail rail, bolted to the breadboard.

    Then you complicate that by going from a straight, two-mirror cavity to a three or four mirror cavity ring configuration, and then spin it real fast. Can you say "centrifugal force?"

    A fiber loop isn't quite the same as a ring laser, because the ring laser actually has the laser gain medium in the ring. As opposed to having the beam directed into a ring. The gain medium in the ring cavity ensures a standing wave is set up in the cavity, which would not be so for the fiber loop.

    Of interest for the future of laser gyros are the new photorefractive polymer devices that exhibit the property of two-beam coupling. This device allows coherent transfer of energy from one beam to another, when the beams are intersected in the material. This can be used to assemble a ring resonant cavity, pumped from the outside by a laser. This can be done with a small diode laser resulting in an assembly much smaller and easier to keep still while spinning than a gas laser ring cavity.

    Photorefractive oscillators using inorganic PR crystals have been studied for some time. The first announcement of a resonant cavity using a PR polymer has just occurred in the past few weeks (March, 1998).

    (From: Douglas Dwyer (

    If you are trying to make a laser gyro as a home project you've got a lifetime project.

    The ring laser is often carved out of a solid block for stability , a major problem with both ring lasers and fibre gyros is locking of the two phases - when rotated the phase relationship between the two paths sticks until a certain rotation rate is reached at which point the two paths unlock and it starts to work properly The solution to this could be to deliberately modulate the phase of the light with pseudo random noise and demodulate at the phase detector. Also as stated the fibre gyro is less attractive because of the inherent greater spectral width of the laser.

    I wonder if one could make a Mossbauer gyro. I once saw turntable rotation detected by the relativistic effects on the gamma radiation and absorption. That could be easier.

    Sam's Home-Built Ring Laser Gyro 1

    Since some people considered a DIY ring laser gyro an impossibility - myself until recently - that only meant it was something I'd have to attempt. Of course, others thought it was no big deal. It turned out to be somewhere in between. This is not the sort of project that could be completed successfully in a reasonable time by someone with no experience in constructing or testing lasers, and in particular open cavity HeNe lasers. But it's also not rocket science or even really laser science. However, it does call upon a variety of skills that can only be mastered by experience. There are simpler laser projects that can be done to acquire those skills and would have a better chance of a successful outcome leading up to construction of the DIY RLG.

    The design was largely based on (1) keeping it small enough to be portable (or at least storable without disassembly), (2) using a short (9.5 inch) two-Brewster HeNe laser plasma tube ("gain tube"), and (3) using as many other parts from my junk drawers as possible. ;-) It is a triangular ring with leg lengths averaging around 10.9" using on a 12"x14"x1/2" aluminum plate which can be placed on a turntable ("Lazy Susan") for tests. A diagram is shown in Opto-Mechanical Layout of Sam's Ring Laser Gyro 1.

    The final outcome was actually better then expected as an excellent means of demonstrating the concept of the RLG (among other things). Since it is built on an aluminum slab with aluminum optics mounts, the thermal stability isn't great, and occasional adjustments and optics cleaning are required to keep it tuned and responsive. A great deal of additional work would be required to make the thing useful for navigation. :) But its behavior should be generally similar to that of the educational RLG kits from the companies listed above. And aside from the baseplate, it was built totally from available parts. And if I had realized at the time that a wood base could actually be more stable than metal, even that cost could have been avoided. Nearly everything except the mirrors originally came from eBay in one form or another for other projects. The mirrors are from my stock of new/NOS surplus Melles Griot optics like those I currently have for sale on eBay. Replicating this RLG or something similar could probably be done for less then $1,000 if everything had to be purchased (used or surplus), with most of the cost being for the two-Brewster HeNe laser tube. The cost of the other parts could be much lower dependent on one's scrounging/fabricating ability.

    Sam-1 RLG approximate specifications

    Note: All calculated values above are based on the spec'd physical leg lengths. The effective optical perimeter will be slightly longer when the etalon is installed and may also differ due to assembly tolerances and adjustments.


    I usually don't put dates on these descriptions but for this project, the chronology seemed to be worthwhile. However, as a concession, edits may have been down to previous segments for corrections or updates. ;-)

    (16-Nov-2018) After a request from someone who also wanted to build an RLG, I tested samples of the available new/NOS 633 nm HeNe mirrors at a 30 degree Angle of Incidence (AoI) to determine the effects of non-normal incidence on reflectance. Unless a mirror is spec'd for non-normal incidence, one cannot assume that its reflectance will not change compared to normal incidence - and possibly quite dramatically. All the HR mirrors tested remained very close to HR at least to a 30 degree AoI. This may be due to HeNe laser mirrors generally NOT being designed to have their reflectance peak at the spec'd wavelength, but to be somewhat flat with a "cliff" on one side at the expense of extending somewhat on the other. This is done in part to suppress unwanted rogue wavelengths, especially in high power HeNe laser tubes. The matching OC would have the cliff on the opposite side. Only the overlap has the desired reflectance which drops quickly on one side or the other and kills lasing at unwanted wavelengths. So there's a 50 percent chance that the wavelength shifts in the direction at the 30 degree AoI such that the reflectivity doesn't change by much and is still HR. Murphy must have taken the day off with respect to these mirrors since as noted, they were all well behaved. ;-)

    The final decision on mirror selection was to use two Melles Griot LO1003 planar HRs and one 60 cm RoC HR intended for the Melles Griot STP-901/SP-117 stabilized laser tube. The decision to use a single curved mirror was to minimize the astigmatism that results from a non-normal AoI. The 60 cm RoC was chosen based on resonator stability and to minimize the intra-cavity mode diameter in the gain tube, which has a narrow bore. The STP-901/SP-117 HR has a slightly lower reflectance compared to a standrad HR to provide enough power for the waste beam to be used for stabilization feedback. For the RLG, this will result in the available power being a bit higher for the detector. The outer surface is AR-coated. A possible alternative was an LO1130 60 cm RoC HR mirror. This is the internal HR used in Melles Griot's high-Q one-Brewster tubes like the 05-LHB-568, and the recommended external HR for use with it capable of 10s of WATTS of intra-cavity circulating power. The LO1130 would probably have lower scatter, but also lower transmission even at a 30 degree AoI. Other options included the LO1013 or LO1118, which are the 60 cm RoC HR mirrors used in some other Melles Griot one-Brewster and internal mirror HeNe laser tubes.

    (1-Dec-2018) It was also decided to use genuine Thorlabs and Newport mirror mounts rather than to construct them from scratch. Taking advantage of commercial parts like mirror mounts may be considered cheating, but there were going to be enough other challenges to deal with. And the mounts were available begging to be gainfully employed. ;-)

    (12-Dec-2018) The aluminum slab arrived and was then lightly sanded and cleaned. The sub-plate with dual 4-screw adjustment rings for the gain tube was constructed. With a dummy tube installed in the rings, it, along with the mirror mounts with 8 mm to 1 inch mirror adapters and "sacrificial" mirrors were positioned to form the near-equilateral triangular ring with enough space reserved beyond mirror 2 so that the detector assembly sub-plate could be attached later. (The use of a PZT for mirror 3 was not anticipated at this time.) Hole positions were marked and drilling and tapping commenced.

    (15-Dec-2018) A photo of the overall RLG assembly early on is shown in Sam's Ring Laser Gyro 1 Under Construction. An internal mirror HeNe laser tube (05-LHR-120) and expendable mirrors are substituting for the actual 2-B tube and high quality new/NOS mirrors that will eventually be installed. All the holes that could possibly be needed have been drilled and tapped. The open ones are for the detector assembly sub-plate that will attach beyond the far mirror and along the leg paths that might be for an etalon, waveplate, or other intra-cavity optics, which would also mount on their own sub-plate(s) so that no shop work will need to be done to the main RLG assembly once it's aligned and lasing. Although the RLG is sitting flat on the table, for the purposes of what follows, the RLG components are arranged with the three mirrors numbered 1, 2, and 3 clockwise starting from the one at the lower left next to the anode-end of the gain tube at the bottom matching, the diagram, above. The time lapse to the next installment resulted from a diversion playing with the Honeywell GG1320 RLG. ;-)

    (27-Dec-2018) The first serious alignment was attempted without the Melles Griot 05-LHB-290 two-Brewster "gain tube" using a high quality green laser pointer modified to run on a DC power supply. The mirrors pass 532 nm quite well, too well in fact. Since all 3 mirrors are HR at 633 nm, not much would get through if a red laser was used. However, the beam diameter of the pointer was too large and the beam quality could not compare to that of a HeNe. So a green (544 nm) HeNe on a mount with a pair of 3-screw alignment rings was then used. The difference between 532 and 544 nm was not significant but the beam was narrower and very high quality. The only other option besides red was yellow at 594 nm and the mirrors have similar reflectivity there. A wavelength in between would have been better but unfortunately, there is no 570 nm HeNe :) and nothing else was available.

    By aiming the alignment laser through mirror 1 to mirror 2 so that it was centered on both of them, and then adjusting mirrors 3, 1, and 2 in turn to center the beam on all the mirrors, alignment appeared fairly good. But with the 60 cm RoC of mirror 2, the spot became smeared in the plane of the ring after 1 or 2 circuits making the cavity stability in the ring-plane axis suspect.

    Nonetheless, the gain tube was installed and removed multiple times with alignment in between in an attempt to achieve some lasing - any lasing - even a dozen coherent photons. :) But with its narrow bore, it was impossible to align it to the intracavity beam in a way that was indisputably centered. And there was no lasing or even a flash to go along with jiggling it or systematic searching using the thumbscrews.

    So suspecting that the smeared spot did indicate a real issue, a 120 cm RoC mirror was substituted for the 60 cm mirror. The same procedure did result in a much nicer round spot likely after several circuits of the ring, and possibly even some dancing interference effects indicating that alignment was good, but still no lasing when the gain tube was installed, being even more difficult to center with the larger beam diameter resulting from the 120 cm RoC mirror.

    (29-Dec-2018) So on to Plan B. This consisted of two parts: First was to use a medium power (12 mW Melles Griot 05-LHR-991) red HeNe for alignment. As noted, almost nothing gets through the HR mirrors but whatever does get through would make many circuits of the ring due to the near perfect reflectance of all the mirrors at 633 nm. So the second part was to incorporate a Scanning Ring Cavity Interferometer (SRCI) into the RLG itself. (See the section: Scanning Ring Cavity Interferometer.) This is a fancy way of saying that a PZT was added behind mirror 3 with a Thorlabs PDA-55 amplified photodiode monitoring leakage from mirror 2. After aligning the mirrors WITHOUT the gain tube so that the spots converged, some additional fiddling resulted in a signal being detected. And with careful mirror tweaking, the display was quite respectable. The finesse was at least 100 and probably a lot more since all three mirrors have reflectances near 100 percent. But the display was very noisy and scrambled since there was no spatial mode confinement nor could the ring cavity be made mode-degenerate. So the 05-LHR-290 gain tube was added and carefully aligned to obtain an SRCI display. With it in the ring (not powered), the finesse went way down likely because its bore diameter is smaller than the intra-cavity mode diameter. More about that later. But it did serve to force single spatial mode operation so the SRCI peaks were no longer messed up and resulted in a very nice near-textbook display of the longitudinal modes of the 05-LHR-991 marching along just as they would when viewed with an SFPI. However, that in itself appeared somewhat strange.

    There are two thermal effects going on: First, when the 05-LHR-991 alignment laser is initially powered on, its tube starts expanding and its modes move through the neon gain curve at a fairly fast rate. But once it has reached thermal equilibrium, they become relatively stationary. Second, the base of the RLG is expanding due to heat from the two-Brewster tube and its ballast being powered. So the SRCI is drifting with respect to the HeNe as a result of its mirror spacing increasing.

    A decipherable mode display was unexpected for the SRCI due to its small FSR. The FSR for a ring cavity with perimeter L is c/L, twice that of a Fabry-Perot (linear) cavity of length L. (FSR is really just c divided by the round trip length regardless of the cavity geometry.) For this RLG cavity, L is 32.6 inch or 0.828 m. The FSR should then be 362 MHz. Fine so far. But the 4 or 5 modes of the LHR-991 are spaced at 341 MHz with a maximum total width of around 1.5 GHz. They show up perfectly moving slowly through the display just as though it had an FSR of many GHz. The $1 beeper PZT that usually scans several FSRs in an SFPIusing the Wavetek function generator to drive it could only achive a fraction of one FSR, so the clump of beautiful 05-LHR-991 modes would drift through the display and then move off-screen only to eventually reappear on the other side. (This was due to an FSR in a ring cavity requiring a full wavelength of movement to scan rather a half wavelength, the PZT being attached to a 1 inch ring to fit the Thorlabs mirror mount which is smaller than the normal mounting location at its edge, and the hole in the PZT having been made as large as possible to be able to pass the ~30 degree angled transmitted beams during testing.) To rectify this, the PZT was connected to a (high voltage) SP-476 SFPI driver with a 3:1 voltage divider, which then enabled more than one instance of the 05-LHR-991 modes to be displayed as would normally be expected with an SPFI.

    The explanation for this weirdness ends up being pretty mundane: The modes are being displayed in a normal manner as a result of aliasing between the 05-LHR-991 mode spacing of 341 MHz and the SRCI FSR of 362 MHz. Imagine two combs with slightly different tooth spacings (341 and 362) moving with respect to each other. Whenever a two teeth intersect, there is a peak in the display. There can be benefits to violating Nyquist sampling. ;-)

    Thus the effective FSR of the SRCI is actually the FSR of the difference in the lengths for a round trip of the two cavities. The round trip cavity lengths are 87.9 cm for 05-LHR-991 and 82.8 cm for the ring cavity. The difference in the round trip path lengths is then 5.1 cm and this becomes an effective FSR of 5.88 GHz. The mode spacing and FSR being close together was fortuitous. If a shorter alignment laser had been used, the mode display might have been hopeless jumbled. Not that the actual mode display really mattered (only the finesse), but it did make for an interesting exercise. :)

    At this point, there was still no lasing with the 05-LHB-290 2-B tube even though the SRCI showed the alignment to be perfect with the tube installed. However, it was providing definite gain as the amplitude of the SRCI peaks increased by a factor of 6 or more when the tube was powered. But that was just not enough to exceed threshold. The hypothesis at this point was that the limiting factor was the narrow 05-LHB-290 bore since the intracavity mode diameter was larger with the 120 cm RoC mirror. The bore size cannot be changed :) but the RoC of the curved mirror could be reduced, which in turn would reduce the intracavity mode diameter in the area of the tube. The 120 cm RoC mirror would probably work well if the bore were wider. As noted, originally a 60 cm mirror was installed but the cavity didn't appear stable with the green alignment laser. However, based on theory, 60 cm should be stable even accounting for the astigmatism of the off-axis beams.

    The effective RoC for an off-axis beam in the Y direction (parallel to the ring's axis) doesn't change, but for X it is given by:

        RoCeff = 2RoC(1 - ------)

    where θ is the Angle of Incidence (AoI) relative to the normal to the mirror. At the ~30 degree AoI of this RLG, RoCeff for the 60 cm RoC mirror will be 51.7 cm. The RoC for a stable cavity must be at least 41.2 cm (1/2 the ring perimeter), though a comforatable value would be slightly greater.

    So perhaps the observation was a red herring and not correct.

    The gain tube was located opposite the curved mirror in order to minimize the mode diameter in its bore, which would in effect be equivalent to placing it near the planar mirror-end of a near hemispherical linear cavity. Calculations show that the intracavity mode diameter should be less than 0.5 mm within the bore. This isn't ideal - even narrower would be better to reduce diffrection losses - but it should be acceptable. And the narrow bore should still help with spatial mode confinement.

    (30-Dec-2018) So, with the rest of the cavity aligned, a 60 cm RoC mirror was installed. 80 and 100 cm RoC mirrors were also available as backup. (Assuming lasing with decent power was possible using a larger RoC, it would be more likely to be single spatial mode and thus beneficial, but with a sacrifice of output power. But it's not clear what the additional scatter from the walls would do in so far as the lock-in threshold is concerned.)

    And we can now report First Light for this RLG laser as shown in Sam's Ring Laser Gyro 1 Laser Lasing!. :-) Less than 1 minute was required to realign the mount with the 60 cm RoC mirror before it was obvious that something beyond the SRCI display of the alignment laser's modes was appearing on the scope - the detector saturated and lost correlation with them. That was evident even before noticing the multiplicity of bright red reflections from various locations on the RLG assembly as well as the 6 output beams which weren't present with the red alignment laser alone. And having now replaced mirrors several times, this approach to alignment has always been required. Even knowing where the beam should be, getting it through the gain tube bore, and superimposing spots proved futile. But monitoring the alignment using the SRCI resulted in success quickly because the output is a linear function of alignment, while lasing has a threshold. So very slight improvements are immediately obvious even if the SRCI blips are buried in noise.

    And wow! While lasing is solid, the power from the CW and CCW beams is extremely unstable. :( :) And not unexpectadly, it is often multi-spatial mode, typically diagonal TEM01 with the CW and CCW beams being different modes, though each is always TEM00. The photo above just happened to catch the CW and CCW spots in the upper right corner being more or less equal intensity.

    As a side note, on this tube (originally from a NIST iodine stabilized HeNe), the Brewster windows aren't quite perfectly aligned. I should complain to Melles Griot as the performance could be better. :-) Oh, never mind, the morons there don't do HeNes any longer. :(

    The total power being spewed forth from all the ring cavity's orifices is between 0.75 mW and 1.0 mW in the 6 beams leaking through the 3 mirrors (3 CW and 3 CCW) and the 4 beams from the Brewsters (2 from the 2 surfaces on each window). The beams from the mirrors have an output power between 50 and 75 µW. That's quite good considering that all the mirrors are HR at 633 nm, and the reflectivity is still nearly as high at the 30 degree AoI. However, the available power may go down once a means is added to force single mode operation, a few µW is all that's needed for the detector. Since most optical surfaces are vertically oriented, it can sit out in my not exactly cleanroom conditions and still lase after sitting overnight or longer, and then be peaked again with minor dusting of the Brewsters.

    Initially, the output power balance tended to vary at random between pure CW or CCW, though more often than not a mixture of the two, and may oscillate between them with a frequency up to 10s of kHz or totally chaotically. Which happens is correlated somewhat with cavity length and almost everything else. Sometimes it even appeared that the output power contained a signal that looks suspiciously like a Sagnac frequency. That may be just an illusion but the unavoidable inter-beam coupling due to scatter can result in power modulation.

    (31-Dec-2018) Adding an etalon inside the ring cavity resulted in much better stability but confirmation of the resulting mode structure would require a test with an external SFPI. The etalon is shown in Sam's Ring Laser Gyro 1 Etalon It is just a glass block with two polished faces I had gotten a pile of friebies from somewhere called a "compensation plate", whatever that means. The thickness is about 1 cm for an effective optical path length of around 1.5 cm or an FSR of 10 GHz, which is a bit wide. The faces are parallel and uncoated resulting in a relatively smooth variation in transmitted power of around 14 percent over the FSR. Perhaps two of them glued together with optical cement would be better. ;-) But the single block appears to get most of the way there by singificantly quieting down the erratic behavior, which is actually rather surprising given the large FSR and small reflectance variation, so perhaps there is something else going on. The power balance is fairly constant and as will be seen, there's a range of adjustment of the etalon mount (pan/tilt) which results in clean signals.

    (1-Jan-2019) Rather than doing something logical like testing the mode structure :), the beam combining optics of the detector were added. (Inputting one of the beams to an SFPI would have required raising the RLG or improvising a periscope to match the beam height. Definitly too much work!) These are shown in Sam's Ring Laser Gyro 1 Basic Detector It includes a planar HR as a bounce mirror along with a HWP and PBS to merge the CCW and CW beams, and a photodiode behind a polarizer at 45 degrees. The black tape is securing a sliver of HWP in the CCW beam. This is now everything in the layout diagram, above, except the final parts of the quadrature decoder (QWP, NPBS, polarizers, PDs) in Sagnac Frequency Detector 1. That will be built as a module and then replace the single polarizer and PD. And a dual preamp will be added. The single channel is enough to actually see Sagnac frequency activity, but not determine the direction.

    The current version is shown in Sam's Ring Laser Gyro 1 Nearly Complete And the darn thing actually works! ;-) Once it was put on my high tech Lazy Susan $1 garage sale turntable, the Sagnac behavior became obvious. Unlocking requires a fraction of a degree per second rotation, which isn't all that much different than for the Honeywell GG1342. It must usually be mostly single longitudinal mode with the etalon because there is very little instability. Or at least that one mode is sufficiently larger than any others to dominate. (That's the "mostly" part) However, the etalon does need to be adjusted from time-to-time due to thermal changes both to peak power and response. It's obvious when it's on the edge because the waveforms become chaotic. The etalon and everything else should really be temperature controlled. But that's not going to happen. :) Although with the etalon, it may not always be SLM, it must be the majority of the time because the response is then as expected from rotation. Or at least that there are both a CCW and CW lasing mode that are the same cavity mode. If there are other modes far away mutually distant from one-another, they would be ignored in the detector except to the extent of reducing the usable power.

    (3-Jan-2019) A dual channel photodiode preamp PCB was repurposed and populated to provide higher amplitude and better signal quality. It provides a variable trans-impedance gain of up to 10 V/µA, along with offset adjustments for both channels to compensate for the DC level due to imperfect alignment and power balance. It is in two stages using TL072 op-amps, which provide adequate bandwidth for the anticipated maximum rotation rate of the RLG. After the PCB was tested (following the correction of some stupid rewiring errors), the final parts of the Sagmac frequency detector - the quad-A-B optics and photodiodes were added. Most of these are actually the waste beam sampler from an Axsys 150 laser normally used to generate the feedback signals for dual mode stabilization. But with the addition of a QWP in front of one of the photodiodes, it is serving a new noble purpose. ;-) And as a concession to simplicity, the beam sampler was aligned in the combined beam from the PBS cube for maximum signal and glued to a sheet metal 45 degree bracket. (It has to be at 45 degrees because the polarizers already in front of the photodiodes are aligned at 0 or 90 degrees. Therefore the orientation of the QWP and LPs in the detector do not agree with the layout diagram, above.) A miniature 4 axis positioner would have been just a wee bit of overkill and was not available. The only thing left to do now is to slip a piece of QWP into position at 45 degrees relative to its polarizer.

    (4-Jan-2019) A sliver of QWP (similar to those from HP/Agilent lasers) was installed at a 45 degree orientation (for the same reasons as described above) in front of one of the photodiodes. The outputs are now shifted in phase, though whether it's a full 90 degrees is hard to say. The original QWP was not labeled as to optical axis, so its orientation was estimated by aligning the QWP between crossed polarizers to go along with taped-in-place HWP. In addition, these QWPs are made from optical grade mica and their precise retardation is fine tuned by tilt. This snippet lying is lying flat, so its retardation may not be precisely 90 degrees.

    However, the waveforms are generally not sinusoidal, possibly due to a combination of multiple longitudinal mode operation and/or unlocking distortion until the rotation rate is quite high, and it could also be impacted by the frequency response of the preamp. This general behavior was similar when using a back-biased photodiode with only a resistor for a load except that it was much noisier and the frequency response was quite poor.

    And the sounds the signals make when connected to an audio amp are quite cool including various chirps and tweets as the RLG is rotated back and forth depending on velocity. Most of it tends to be above the normal range of human hearing though - the ultrasonic glissandos and such would probably drive bats and dolphins quite nuts. :) See and hear Signals from Sam's Ring Laser Gyro 1 while Rotating Back and Forth 1 for a pathetic clip of the two channels from the Sagnac detector on the scope and audio via a stereo amplifier while manually moving the RLG. You might want to put cotton balls or more serious plugs in your ears though. :) Each obnoxious brain tingling sqeek means that the direction has changed. The baseplate also acts as a decent microphone even if within lock-in, presumably due to changes in alignment and laser power, and results in acoustic feedback if the volume is set too high. It is possible to make out the lead or lag of the two waveforms depending on direction and study it in Snapshot from Sam's Ring Laser Gyro Video. The phase shift may not be quite a full 90 degrees but so be it. :) The scope is set at 50 µs per division so 2 cycles on the screen epresents a period of 250 µs or a frequency of 4 kHz. The variation in amplitude and shape of the waveforms of the two signals is due both to which modes are lasing, how close the Sagnac frequency is to lock-in, and the high frequency rolloff of the photodiode preamp at around 50 kHz.

    I did finally check mode structure using an SFPI. With the etalon installed, it turns out to be a 3 longitudinal mode laser which is usually pure SLM over a narrow range when one of the modes is centered on the neon gain curve. The etalon seems capable of forcing SLM but to maintain it for any length of time would require careful temperature control of both the etalon and overall RLG structure. A more selective etalon would force pure SLM but might require feedback control of the mode structure itself to maintain it in addition to temperature control.

    Based on behavior, it is believed that with the etalon, the spatial mode is TEM00 for both the combined CW and CCW beams. It's not currently convenient to use the SFPI to display the modes of the combined beam.

    While this setup is small enough to be portable, I wouldn't want to have to store it in a box or move it more than across the room. Lasing would cease if alignment were lost on any of the 12 precision adjustments (two tube ring mounts with 2 axes each and 3 mirror mounts with 2 axes each), as well as 2 axes of the etalon (though it may be possible to simply remove that without significantly affecting alignment). 1/10th turn goes from peak to nothing on any of the adjustable mounts. Unless the specific adjustment that was affected is known, getting it back could require starting at the beginning with an external alignment laser and the SRCI. So once it's decided that nothing more will be done with it for awhile and its put away, my assumption is that it will require alignment if ever resurrected. Just seriously bumping a mount while putting it in a box could shift it position even though they are all really well tightened down. To minimize the chance of an adjustment screw being accidentally turned without realizing it, the knobs on the Thorlabs mounts will eventually be removed. But even then there is the issue of contamination on the Brewster windows, mirrors, and etalon surfaces. A single speck of dust anywhere can kill lasing and make it difficult to determine if that is due to dirt or misalignment.

    (11-Dec-2019) Just as a wild test, the Wavetek function generator was again connected to the PZT behind mirror 2 to see what effect various frequencies and amplitudes would have on RLG operation. The results were .... interesting, though probably not terribly useful. Generally, since a variation in the cavity length (perimeter) moves the lasing modes, there would be a correlated change in output power in the detected signals. If the amplitude was large enough, mode hoping would occur. However, under certain not entirely well defined (or repeatable) conditions, the response to very small rotational velocities appeared to be enhanced resulting in a lower lock-in threshold or a significant phase shift if unlocking didn't occur.

    (26-Dec-2019) Construction has commenced on a rudimentary angle readout using µMD1 (Micro Measurement Display 1) developed for and used with laser interferometers in metrology applications. µMD1 interfaces to the interferometer hardware via USB to a chipKit DP32 development board with a MIPS microprocessor. Its Windows Graphical User Interface (GUI) displays data in a variety of formats tied to the change in path length difference of the two arms of an interferometer. In the case of an RLG, the two arms are the CCW and CW laser beams. And the change in path length difference is equivalent to the Sagnac frequency.

    The original plan was to construct a hardware quad-A-B to pulse converter using discrete TTL, or a PLD, or FPGA. It would run at over 20M counts/second, but that's way overkill since the RLG is tethered to its power supplies and detector electronics and can only rotate a total of less than 90 degrees and slowly. So it was decided to use my favorite Arduino-compatible $2 microprocessor board, an Atmega 328 Nano 3.0. With the quad pulse code running in a tight polling loop using Arduino digitalread and digitalwrite instructions, it is capable of a throughput of well over 40K cycles/second for a single quad-A-B channel. By replacing the those system calls with direct PORT reads and writes, order of 1 million (4X) counts/second. That's well beyond what is reasonably possible. :)

    However, if you're intending to spin the RLG in a centrifuge and really want to stretch the envelope, Aynchronous Quadrature to Pulse Converter should run at 5 MHz or more. But there are a pair of discrete delays and the relative propogation delays of the 74LS04s and 74F153 are critical which may be just as bad as using monostables (which would earn you an "F" in my logic design course).

    With some tweaks, the maximum speed can be greatly increased and this would fit into a small PLD or FPGA.

    The quad-A-B signals from the RLG are converted to TTL using two sections of a 74HC14 hex Schmitt Trigger as voltage comparators and applied to digital inputs D2 and D3 of the Atmega programmed to generate up/down pulses on D4 and D5 for µMD1 running in Homodyne Mode. Without modifications to the Windows GUI, the label for the units won't be correct since it would have to use normal Displacement mode which displays a value proportional to the accumulated counts. (While µMD1 has an angle mode, due to the way the normal angular interferometer works, the calculation is non-linear.) But it should be possible to adjust the scale factor so that the correct value is displayed in arcsec, arcmin, or degrees.

    (27-Jan-2019) The converter with Atmega board has been mostly completed and tested stand-alone. It was first built on a solderless breadboard. But while that worked, it proved too unreliable. So the circuit was soldered on a Perf. board with the Atmega board plugged into a socket.

    (29-Jan-2019) Everything is now working together with µMD1 on an old Window 7 notebook as shown in Sam's RLG 1 Complete Setup and a closeup of the circuit boards in Sam's RLG 1 Processing Electronics. From front to back are the dual photodiode (detector) preamp, Atmega quad-A-B to up/down pulse converter, and µMD1 chipKit DP32. The optimized Atmega code was necessary to avoid excessive count rate errors due to the non-uniform manual rotation even at low rates. An "Encoder" checkbox has been added to the µMD1 GUI configuration so it reads linearly with the correct units, "Encoder Angle" since the RLG is similar to a rotary encoder. And will also make the GUI compatible with robotic arm and construction crane rotary encoders. ;-)

    And RLG works reasonably well. See Screen Shot of Sam's RLG Angle Readout using µMD1 shows manual rotation with approximately correct angle calibration.

    Stay tuned for more exciting developments, though I imagine they may be anticlimactic. It will be hard to beat First Light for a ring laser and First Sagmac Frequency Activity for a DIY RLG, and a tooth shattering visual-auditory experience. ;-) In fact, at this point the effort is likely to concentrate on RLG Sam-2.

    Summary of major parts

    Consider this just a serving suggestion. ;-) Most of the details are not critical. The list does not include screws and other hardware. All optics are spec'd at 633 nm.

    Although commercial mounts are specified, there is no reason this cannot also use DIY mounts, which can be just as good, if not quite as esthetically pleasing.

    Sam's Home-Built Ring Laser Gyro 2

    The objective here was to build an RLG that is around 50 percent larger than the one above using a Melles Griot 05-WHR-252 (same as 05-WPR-252) for the gain tube, which has two perpendicular windows and is about 14 inch in length. By using perpendicular rather than Brewster-angle windows, it will permit both S and P polarization for the intra-cavity beams, which could be beneficial in doing experiments with reducing lock-in or increasing resolution. Most of the design will be virtually identical to the one above except for size, which has no mystical significance and was only suggested by the the ability to accommodate the longer gain tube and the availability of a suitable sturdy, flat, and attractive wooden base, cast off from an unknown someone's remodeling project. :) Nearly all other parts are on hand.

    The objective is stated in the past tense because as will become clear, some basic issues arose that will probably make the use of a perpendicular window tube impossible. So what's below is how it stands now but that tube will likely end up being replaced with two Brewster tube.

    A diagram is shown in Opto-Mechanical Layout of Sam's Ring Laser Gyro 2. The major differences compared to Sam-1 are (1) the leg lengths, (2) gain tube, and (3) location of QWP in the detector. And yes, the leg lengths as depicted in the diagram aren't quite to scale, but it wasn't worth redrawing almost the entire thing because the angles differ by a few degrees. :-) The bottom leg is also just long enough to permit a LASOS LGR-7627-BF two-Brewster tube (~15.5 inches) to be substituted for the 2-W tube if desired.

    Sam-2 RLG approximate specifications

    Note: All calculated values above are based on the spec'd physical leg lengths. The effective optical perimeter will be slightly longer when the etalon is installed and may also differ due to assembly tolerances and adjustments.


    As with the Sam-1 RLG, this will also have dates for all the exciting events and milestones.

    (19-Jan-2019) A photo of the overall RLG assembly early on is shown in Sam's Ring Laser Gyro 2 Under Construction. As with the first one, an internal mirror HeNe laser tube (05-LHR-150 in this case) and expendable mirrors are substituting for the actual 2-B tube and high quality new/NOS mirrors that will eventually be installed. Since there's no problem adding holes for additional parts, there is no real need to pre-drill and tap the wooden base, which will be stained antique walnut and varnished with rounded over routed edges. ;-).

    (20-Jan-2019) Sam's Ring Laser Gyro 2 Baseplate Ready for Permanent Parts Installation shows the final result. Or almost. It still needs some sanding, touch up, and another coat of polyurethane. Since trimming wood is much easier than trimming aluminum, it was shaped to be more in keeping with the triangular ring. And yes, there is one extra hole. ;-) Another wooden plate cut to be slightly smaller will be secured to this one to increase stiffnes.

    (21-Jan-2019) Applied second coat of polyurethane and cut and finished stiffener plate including countersunk holes for wood screws to secure to main plate along with wood glue.

    (22-Jan-2019) The main panel was hand-rubbed with fine steel wool and then a soft cloth for a semi-gloss table-top appearance. Lemon scented furniture polish optional. The bottom was spray-painted black to reduce visibility. Never underestimate the importance of esthetics! ;-) This was also done to minimize moisture absorption. (Though these panels have been sitting 20+ years with no signs of warping.) It's now ready for final assembly.

    Next, the mirror mounts with new/NOS mirrors will be installed permanently, along with the internal mirror 05-LHR-150 laser tube for alignment, and then the 05-WHR-252 will replace it. With its wide bore, alignment will hopefully not be as much of a saga as with Sam-1. Passing the alignment beam cleanly through its bore should be possible. However, the wide may create additional challenges like achieving robust single spatial mode, which may require the addition of an intra-cavity aperture or selecting an RoC for the curved mirror to create a near-hemispherical cavity.

    (6-Feb-2019) After a short hiatus to get other unimportant stuff done :), the RLG was reassembled on the stained polyurethaned hand-rubbed wooden base secured to a stiffener board and turntable. ;-) While alignment still required the use of the built-in SRCI, knowing the drill, it was not as much of a saga as with RLG Sam-1. With the ~50 percent larger cavity, alignment is somewhat more critical, though not unmanageable even on the wooden base. So, we can now report First Light for RLG Sam-2 using the 05-WHR-252 two window gain tube as shown in Basic Testing Setup for Sam's Ring Laser Gyro 2 with the RLG and silver alignment HeNe propped up on blocks. The red alignment HeNe has an output power of around 6 mW of which only ~0.5 µW makes it through the mirror. But WOW, if you thought the Sam-1 cavity was unstable, this is infinitely worse when tested with instruments, though unlike RLG Sam-1, there is no evidence of gross noise or mode switching by eye - all beams appear fairly stable. But the SRCI display of the lasing modes looks exactly like random noise with no sign of anything correlated with the scan ramp, and turning the driver on and off makes no difference. More on this below.

    The sum of the CCW and CW beams (using scope A+B) from the same mirror has *only* about 20 percent p-p noise when both the CW or CCW beams are nearly 100 percent noise. This is still much more noise than would be present from a common HeNe laser. The total power of 6 beams is only slightly greater than that from RLG Sam-1, possibly due to bore clipping due to the larger intra-cavity mode diameter. In a linear cavity with optimal mirrors and geometry, the 05-WHR-252 is capable of around three times the output power of the 05-LHB-290.

    With the compensator plate etalon block, which totally cleaned up the laser in RLG Sam-1, it may quiet down somewhat, but is still strange. At this point, it wasn't clear how much of the instability is the likely non-TEM00 spatial mode, or the dual window tube, or the cavity geometry with the 120 cm RoC Mirror 2.

    (7-Feb-2019) An 80 cm RoC mirror was substituted for Mirror 2. Alignment appeared much more critical though I would have expected it to be simpler. (The cavity turned out to be on the border of stability in the X axis.) And it was not possible to obtain any SRCI display with the 2-W tube in place, even after it was lasing and aligned for peak output power. Strange. The noise is still present and remains in some form even with an aperture, etalon, or Brewster plate (though they weren't tested at the same time). And under certain conditions, the character of the noise could be affected by a magnet near the tube. (??) The working hypothesis is that the noise is mostly an artifact of the reflections from the AR-coated surfaces of the 2-W tube. With the high intra-cavity circulating power (estimated to be around 1 W), the reflections from the 2 surfaces of each of the 2 windows is substantial even though they have excellent AR coatings. And unlike Brewster windows where except for scatter, the reflections are directed away from the cavity and tube, the reflections (as well as the scatter) from perpendicular windows are nearly aligned with both. So the 2-W tube may not be a viable option after all. To confirm, an 05-LHB-290 2-B tube like the one in RLG Sam-1 will be substituted for the 05-WHR-252 2-W to see if the noise goes away. (At least this totally random noise.)

    And it looks like Murphy has returned to work. :) The SRCI display of the alignment laser's longitudinal modes is quite scrambled even with a pinhole to force single spatial mode. This is caused by the aliasing of the alignment laser's longitudinal modes with respect to the ring cavity FSR.

    (8-Feb-2019) So we have a problem......

    It won't lase with the same model 2-B tube as used with RLG Sam-1 (Melles Griot 05-LHB-290). Succeeding with that tube was probably just luck. The hypothesis is that the bore diameter is just on the edge for the smaller ring with a narrower intra-cavity mode diameter. With the two planar and one 60 cm RoC mirrors, the mode diameter at the center of the gain tube in Sam-1 is around 0.47 mm, increasing to somewhere around 0.50 mm at the ends of the bore. For the larger ring of Sam-2, it is 0.52 mm at the center of the gain tube and increases to around 0.57 mm at the ends. It was an accident that the slightly narrower mode diameter worked in Sam-1 and that may also explain why the spatial mode seems to be fairly rebust TEM00 there. All attempts to even get an SRCI display with the 05-LHB-290 tube in place in Sam-2 have only resulted in a mediocre finesse at best. With the 05-WHR-252, the finesse is decent, though not as high as with the tube removed. During alignment, while the intra-cavity beam from the test laser used with the SRCI can be passed through the bore of the 2-B tube, it is cut off by the bore wall and gets attenuated significantly even with one pass. The only way to reduce the mode diameter would be with a shorter RoC for Mirror 2. However, there isn't much wiggle room to reduce the RoC. Due to the off-axis reflections for Mirror 2, its effective RoC in X is only 0.819 times the spec'd RoC. The minimum is one half the ring perimeter of 128.46 cm or 64.23 cm and that limit is reached with a spec'd RoC of around 78.5 cm. Thus there is really no wiggle room at all and running even that close to the limit as it is now would be a challenge. So there are only two options: find a 2-B tube with a wider bore or reduce the perimeter of the ring.

    To round out the initial testing, 100 cm and 120 cm RoC mirrors were also installed. Without the 05-LHB-290 in place, the SRCI display was excellent, but with the tube, the finesse again went way down - probably to less than 10. And while gain can be detected when the tube is on, it's not even as much as was present with the same model tube in RLG Sam-1 and the 120 cm RoC mirror. As you may recall, lasing was not achieved until Mirror 2 had a smaller RoC of 60 cm. This somewhat confirms the 'too narrow bore' hypothesis. Unfortunately, the bore diameter it not easily adjustable. :( :)

    (11-Feb-2019) Out of curiosity, the 05-WHR-252 tube was reinstalled in an attempt to characterize the behavior in more detail. As noted above, there is no evidence of noise visually as there was with RLG Sam-1. But on a scope and fed into an audio amp, check out the fabulous Video of Optical Noise with Two-Window HeNe Laser Tube in Sam's RLG 2. The spectrum exhibits a significant content in the audio range, though the spectrum extends from DC to around 100 kHz. This was confirmed with a Thorlabs DET-210 back-biased 1 GHz photodiode feeding into a 50 ohm load. The bandwidth may be limited at the high end by the cavity Q based on the 6 beams. Mirrors 1 and 3 have reflectivities of approximately 99.99% at 633 nm. Mirror 2 has a reflectivity of approximately 99.95% at 633 nm. The (perimeter or cavity length) is 128.46 cm.

      Q = -------


    For the aggregate loss through all the mirrors of around 0.14 percent, cavity Q is approximately 7.94x109, though this does not account for any losses due to bore clipping. It's a large number but doesn't really tell us much except that changing either the frequency or amplitude of the intra-cavity photons will be slow. Another way to look at it is using the cavity ring-down time:

      τ0 = --------

    Assuming the value of X is 0, the result is approximately 2.67 µs, in the ball park compared to the 10 µs corresponding to the observed 100 kHz, but too far away to be comfortable that the explanation is sound. And this doesn't take into account the additional losses from bore clipping and reflections and scatter from the 2-W windows, which would further decrease the ringdown time.

    The two levels of noise switching at around a one second rate during part of the video correspond to whether the path from the alignment laser laser is blocked or not, and the alignment laser is well aligned with the ring cavity such that is are strong SRCI peaks. Originally, it appeared as though the ~0.5 µW entering the cavity was required to change the character of the noise. However, further tests revealed that a similar change in noise level occurs even if the alignment laser is not turned on as long as the reflection of the CCW beam from its output mirror returns directly back to the cavity. When the path is clear, there is a higher average power level with lower noise. When blocked, there is a lower average power level and the noise amplitude is nearly 100 percent with zero power at the bottom of the display. The power of the back-reflection re-entering the cavity is around 5 picoWatts (5x10-12 W). That minute power is enough to significantly alter the noise characteristics because it is not only aligned with the cavity but coherent with respect to the intra-cavity beam. This really demonstrates in a dramatic way how sensitive a laser can be to back-reflections.

    (12-Feb-2019) With some more careful measurements, the average power in each of the 4 external beams from mirrors 1 and 3 is approximately 64 µW with the alignment laser perfectly aligned with the cavity. This may not even require that the alignment beam be present, only that there are back-reflections from its output mirror. With the path between the alignment laser and mirror 1 blocked, it's only 52 µW. The average power in the each of the 2 external beams from mirror 2 is approximately 310 µW with the alignment laser beam entering and aligned with the cavity and aligned with the cavity and 260 µW with it blocked. Thus the total power is between 728 and 876 µW in the external beams. In addition there are the reflections and scatter off the 05-WHR-252 windows, which is at least as much from each end but trapped inside the cavity and not possible to measure. All 6 beams appear similar and reasonably smooth, but vary from nearly circular to quite elongated depending on alignment. They are probably nearly always multi-spatial mode (e.g., TEM01) and the axis tends to be at an angle which varies with precise alignment, and this may differ for the CCW and CW beams. When elliptical, the shape remains similar with adjustments including tube orientation unless the alignment is way off (which forces symmetric TEM00 but at very low power). However, with sufficient fiddling, it is possible to come close to circular and possibly TEM00 for the CCW beam with the CW beam being elliptical and TEM01 at -45 degrees (though the individual spots merge so the TEM01 is just an assumption). Only then do fine adjustments have an obvious impact on the profile of the circular spot. Presumably, it would also be possible to swap the CCW and CW beam profiles. But there is little to no correlation between shape and noise level. The divergence of the narrow dimension is close to 1.00 mR for all external beams.

    The "output" beams are polarized roughly 3:1 horizontal:vertical. This is likely a result of the relative mirror transmission based on normal (vertical) or ~30 degree (horizontal) AoI. Adding a pair of ceramic magnets of 100 to 200 Gauss on the side of the tube (field horizontal) reduces the output power by a up to 50 percent, mostly stealing from the horizontal polarization. But the magnetic fields reduce the noise by even more, down to 25 percent or less p-p. However, there is significant reduction in output power and noise regardless of the magnetic field orientation, or even different orientations for the pair of magnets. Occasionally the noise will disappear for a moment (probably depending on longitudinal mode position).

    Experimenting more with alignment, there are settings where the average power is similar to the maximum seen, but the noise is much lower, between 0 and 5 percent of the total, depending on the longitudinal mode locations as selected by the PZT driver. Nothing else was done beyond fiddling with the mirror and tube alignment screws. But the magic settings have been lost with subsequent alignment attempts to achieve maximum power and most symmetric beam, presumably with the bore of the gain tube very well centered with respect to the natural cavity spatial mode. The output is again nearly 100 percent noise. Exactly what this means in the grand scheme of things or any implications for the 2-W tube to be usable in the RLG is not entirely clear as there is currently no predictable way to adjust alignment or anything else to get it to quiet down.

    (21-Feb-2019) Just some more fiddling with adjustments with no intracavity components other than the gain tube. It seems that under some conditions, the transverse modes and proportion of CW to CCW power can be affected by alignment in a repeatable way and be stable. Why this should be is not clear since there are no direction-determining components.

    Stay tuned for more exciting developments (but it may be awhile before the actual RLG is completed). However, lack of a suitable two Brewster gain tube is no longer an excuse as I've acquired a Pacific Lasertec (formerly Melles Griot) 05-LHB-294 2-B tube. This is about 8 inches in length but with a wide bore and capable of at least 1 mW. It would appear to be the ideal gain tube to replace the long and noisy 05-WHR-252. Now all that's needed is a revival of my motivation to complete this RLG. ;-)

    Summary of major parts

    Consider this just a serving suggestion. ;-) Most of the details are not critical. The list does not include screws and other hardware. All optics are spec'd at 633 nm.

    Although commercial mounts are specified, there is no reason this cannot also use DIY mounts, which can be just as good, if not quite as esthetically pleasing.

    Sam's Home-Built Ring Laser Gyro 3

    The only reason this is here is that a suitable piece of high quality laminated wood is available from which a base can be constructed. I have absolutely no current plans to actually build it. ;-)

    Sam-3 RLG approximate specifications

    HeNe Ring Cavity and/or Laser Alignment

    Although written specifically for exotic HeNe lasers with planar ring cavities, the same general approach can be used for other types of ring cavity lasers, and with obvious modifications to "bow-tie" and more exotic geometries which may not strictly speaking be called rings.

    HeNe ring cavities are used only in a few rather specialized applications, the most notable being the HeNe Ring Laser Gyro (RLG). (See the section: Ring Laser Gyros.) Most of the others are probably only found in the photonics teaching labs of sadistic instructors. :-) RLGs typically have 3 or 4 mirrors in a planar geometry with equal or nearly equal path-lengths between mirrors. At least one of the mirrors must be concave to form a stable resonator. If built with discrete parts on an optical breadboard (for example), a two-Brewster or two-window HeNe laser plasma tube between one of the pairs of mirrors typically provides the gain. (Multiple tubes can also be used where higher gain is required.) If integrated into a sealed enclosure filled with the HeNe gas mixture, then a section of the cavity path is excited either with RF or a DC discharge to provide gain. This method is used in large research RLGs.

    As an example of the small discrete variety, please refer to Opto-Mechanical Layout of Sam's Ring Laser Gyro 1. For this specific case, the components were secured to a 1/2 inch aluminum plate in leu of an optical breadboard. More on this above.

    Note that curved mirrors with non-normal Angle of Incidence (AoI) introduce astigmatism into the beam which must be taken into account in assuring that the ring cavity is stable. And even if stable for lasing, the asymmetry may do funny things to externally introduced intra-cavity beams, as will be required for alignment.

    Alignment of a cavity with greater than two mirrors is more complex than for a normal laser tube or Fabry-Perot etalon with a linear cavity but the general procedure is similar. The Horizontal (H) and Vertical (V) alignment will be done more or less independently.

    And to state this up front: Alignment of a ring cavity cannot be done by trial and error in a time period significantly less than the order of the lifespan of the Universe. Forget it. ;-) At the very least, initial alignment using either an external laser and alignment cards, or an internal mirror tube substituting for the gain tube will be required, and depending on the specific circumstances, additional aids may also make it a lot easier. The general arrangement is shown in Setup for Alignment of HeNe Ring Cavity Laser. These methods are based directly on experience in aligning home-built RLG Sam-1 and Sam-2, above.

    Initial alignment: This step is a quick way of getting all the mirror mounts set up and rough-aligned. With care under some conditions (like if the stars all align), the result will be close enough to drop the plasma tube in and get lasing with minimal additional fuss. But if the gain tube is long with a narrow bore, this will probably not be sufficient on its own.

    Substitute a normal internal mirror HeNe laser tube with the same diameter for the gain tube with its output beam toward mirror 1. (This is arbitrary but agrees with the following description.) The best choice would be a tube that is a similar model such as an 05-LHR-120 or 05-LHP-120 standing in for an 05-LHB-290. Adjust the mirrors and tube position to center the spots on all the mirrors. With the power available, they should be visible as scatter on all the mirrors. The output beam will serve to adjust mirrors 1 and 2 while the waste beam (with the output beam blocked) will serve to adjust mirror 3 and confirm alignment of mirror 2. The reflections off the tube mirrors should also align, but don't be confused by reflections off the outer (non-coated or AR-coated) surfaces of the mirror glass, which may be curved or at a wedge angle. Look for dancing interference patterns as perfect alignment is approached.

    With care, installing the gain tube will result in lasing with minimal further adjustments, especially if it has a wide bore. If it doesn't lase immediately, loosen the ring at one end and jiggle (technical term!) the tube looking for a flash of coherent light. Then try the other end. If this doesn't succeed after a short amount time not to exceed several hours, it's time to perform the full alignment procedure. :( :)

    Required or recommended equipment

    The last two helpers are not really needed if the beam inside the cavity from the alignment laser is strong enough to see scatter off the mirrors. However, the types of mirrors used in these applications are typically of such high quality that this may be difficult especially if the alignment laser and ring laser are the same wavelength..

    For the purposes of these procedures, the mirrors will be designated as in the diagram, above, with the alignment laser pointing into mirror 1. To ease alignment, transmission at the alignment laser's wavelength should be highest. However, there may be other considerations that prevent this from being desirable. The actual

    Also note that the ring cavity must have a stable geometry for lasing to be achieved. With a single curved mirror, this means that the effective RoC in both X and Y must be greater than one half the perimeter. The effective RoC is reduced by a non-normal Angle of Incidence (AoI) as follows:

      eRoC = 2*RoC*(1 - -------)


    Also from experience, even with a cavity that is stable for lasing but with a mirror RoC that is not much greater than the 1/2 perimeter limit (hemispherical), the size of the alignment beam may not converge after multiple traversals of the ring.

    Method 1: Alignment without instruments except your remaining Mark I eyeball

    This method assumes that the transmission of the alignment laser beam through at least one of the ring cavity mirrors is strong enough to be able to see scatter when it hits the other mirrors. The actual alignment should be done with the room lights turned off since scatter from clean high quality mirrors is very small and difficult to detect especially at low intensity. A night vision scope may be handy to have available. :) The alignment cards can also be used to assist in setting the precise height of the beam at each mirror.

    1. Set up the alignment laser in front of mirror 1 and align its beam position using the mounting ring adjustment screws until the beam is precisely centered on the coated surface of mirror 1 AND reflected from the precise center of the coated surface of mirror 2.

    2. Adjust the pan-tilt alignment of mirror 2 until the beam is precisely centered on the mirror coating or mirror 3.

    3. Adjust the pan-tilt alignment of mirror 3 until the beam is precisely centered on the mirror coating or mirror 1.

    4. Adjust the pan-tilt alignment of mirror 1 until the beam is precisely centered on the mirror coating or mirror 2 and is coincident with the direct beam from the alignment laser.

    5. If the cavity is stable, it should now be possible to "walk" the mirrors so that as the beam goes around the ring cavity multiple times, all the spots merge. With highly reflective mirrors, the beam can make 10s or 100s of trips around the cavity and still be visible, so any misalignment will be instantly obvious. Adjustments should be done in X or Y (mirror pan and tilt) but not both at the same time. Concentrate first on the Y adjustments to get the spots to merge vertically but perhaps be spread out horizontally. Then do X in a similar manner. It may be necessary to adjust multiple mirrors. With care, there should then be variations in the intra-cavity beam intensity at random or due to minute vibrations as a result of interference effects. Since the three mirrors form an etalon, there will be resonances of the alignment laser's longitudinal modes as the perimeter path length changes due to alignment and/or temperature. If this method is not successful in obtaining lasing, the second one will take advantage of this.

    6. Block the path between mirrors 2 and 3. Place alignment cards in the beam path between mirrors 1 and 2 positioned near mirrors 1 and 2, and adjust their X-Y position so the beam from the alignment laser passes cleanly through both holes.

    7. Unblock the path between mirrors 2 and 3. Any error in alignment should result in a "ghost" beam from 1 trip around the ring appearing offset from the hole (at least) on the card closest to mirror 1. Carefully fine tune the mirrors (separately for pan and tilt) to merge it into both holes. Remove the cards.

    8. Install the gain tube with the Brewster windows facing side-ways or down to minimize contamination from airborn dust. Check their cleanliness by inspecting for scatter with a high power laser pointer and clean if necessary. DO NOT touch the mirror alignment until lasing has been achieved. Adjust the tube position using the mounting ring screws so that the alignment beam passes cleanly through its bore. Place the alignment card at the output end (with respect to the alignment beam) to check for the transmitted beam. Fine tune until it is clean and circular and remains stationary on the card with respect to fine adjustments of the gain tube position indicating that it isn't being reflected from the bore wall.

      Block the path between mirrors 2 and 3 and place an alignment card between mirrors 1 and 2 so the alighment beam passes cleanly though its hole.

      Unblock the path between mirrors 2 and 3. There should be at most a halo around the hole in the alignment card. If there is a ghost spot away from the hole, alignment is bad. Go back and try again. :( :)

    9. Assuming the cavity parameters will support lasing (stability, intra-cavity beam diameter with respect to the bore diameter of the gain tube, mirror reflectivity, etc.), it should now lase as soon as the tube is poared. If not, very carefully adjust the tube's X and Y position but do NOT touch mirror alignment!!!

    10. Once it's lasing, mirror alignment and tube X-Y position can be fine tuned to maximize output power and beam profile.

    Method 2: Alignment where the beam inside the cavity is extremely weak

    When the alignment laser and ring laser are of the same wavelength (usually 633 nm) and mirror 1 is HR at the AoI, almost no power from the alignment laser will be passed by mirror 1. For a typical HR mirror, it may be less than 1 µW even for an alignment laser with an output power of 10 mW. With such low power, it's almost impossible to visually detect the location of the spot via scatter on the mirror surface. Therefore, the best that can be done is to estimate the location of the beam on the mirrors by using an alignment card to view the beam spot. However, by adding a PZT behind one of the mirrors, the entire ring cavity can be used as a Scanning Ring Cavity Interferometer (SRCI), the ring analog of the Scanning Fabry Perot Interferometer (SFPI) for a linear cavity. This will require a ramp generator to drive the PZT and an amplified photodiode and oscilloscope to view the response. By using a beeper element PZT, the drive can be modest - 20 V p-p should be sufficient to tune several FSRs. The scope can be almost anything that's still alive. :)

    The benefits of building an SRCI into the ring laser are that it will be much more sensitive than a Mark I eyeball at optimizing alignment before installing the gain tube, and may also be used with the gain tube in place. With lasing, there will always be a threshold that requires alignment to be near perfect. But to detect cavity resonance, as with horseshoes and hand grenades, close is OK for the SRCI. ;-) Or at least it's a linear response.

    1. As with the simple procedure, set up the alignment laser in front of mirror 1 and align its beam position using the mounting ring adjustment screws until the beam is precisely centered on the coated surface of mirror 1 AND reflected from the precise center of the coated surface of mirror 2. However, since scatter will probably not be visible, place alignment cards near mirrors 1 and 2 to set the precise height. Set its approximate position horizontally with a card close to (but not touching!) the mirrors' coated surfaces.

    2. Adjust the pan-tilt alignment of mirror 2 until the beam is centered on the mirror coating of mirror 3, determined with a card as above.

    3. Adjust the pan-tilt alignment of mirror 3 until the beam is centered on the mirror coating of mirror 1, determined with a card as above.

    4. Block the path between mirrors 2 and 3. Place alignment cards in the beam path between mirrors 1 and 2 positioned near mirrors 1 and 2, and adjust their X-Y position so the beam from the alignment laser passes cleanly through both holes.

    5. Block the path between mirrors 2 and 3. Place the alignment card in the beam path between mirrors 1 and 2 as close to mirror 1 as possible and adjust its X-Y position so the beam from the alignment laser passes cleanly through the hole.

    6. Unblock the path between mirrors 2 and 3. Any error in alignment should result in a "ghost" beam from (at least) one trip around the ring appearing offset from the hole. Since it was possible to set the vertical position of the beam fairly precisely (using the cards), only fine adjustments should be required in Y. Adjust mirror 3-tilt until the beam is at the correct height. Then move the card close to mirror 2 and do the same with mirror 1-tilt. Go back and forth until they converge to be both correct. Then do the same with mirrors 2-pan and mirror 3-pan. Eventually the spots should converge to mostly pass through the hole with just a halo around its edge. Remove the alignment cards.

    7. Alignment at this point may be good enough to install the gain tube, but that would be pushing one's luck. ;-) Now is the time to use the SRCI. However, if you feel lucky, skip the following steps and install the gain tube.

    8. Set up the amplified detector behind mirror 2 or mirror 3 (if there is a hole in the PZT). Mirror 1 cannot be used because the reflection from the alignment laser will totally swamp anything from inside the cavity. The sensor will need to be shielded from ambient light and even then, working in the dark (if you don't already do this routinely!) may be necessary to prevent it from saturating. Since there will likely be no actual beam transmitted by the mirror at this point, positioning will have to be done by estimating and maximizing the DC signal from the alignment laser's bore discharge glow. Set the scope to trigger from the ramp generator with the vertical (signal) input all the way up.

    9. With a ramp applied, if the alignment is close, there should be distinct peaks on the display that correlate with the ramp. If anything is detected, first adjust the X-Y position of the detector to maximize it, and then fine tune mirror alignment to maximize it. DO NOT change more than one adjustment at a time and NEVER change anythi