PDA

View Full Version : optical bench



Evan
12-17-2009, 08:13 PM
I finally got around to setting up my optical bench so I can do some experiments in calibrating measuring instruments from primary standards that don't rely on traceability to a source. In particular, known wavelengths of light are as good as direct NIST traceability for accurate and quantitative calibration of distance measurements. It is relatively easy to produce precisely known wavelengths of light using readily available light sources such as low pressure sodium lamps or 630 nm laser diodes.

I still have to make some stands and adjustable holders for some of the optical devices. This collection doesn't include my lenses, prisms or filters with a couple of exceptions only. This is mainly just the mirrors that I have. These are all first surface including the two Perkin Elmer master diffraction gratings which are worth around $1000 each.

http://ixian.ca/pics7/optics1.jpg


http://ixian.ca/pics7/optics2.jpg

Here is the obligatory laser path demonstration that actually is totally meaningless but looks cool.

http://ixian.ca/pics7/optics3.jpg

Doc Nickel
12-17-2009, 08:37 PM
Did you have to blow some smoke in there or was that just a long exposure?

Doc.

Evan
12-17-2009, 08:45 PM
Smoke. I don't usually smoke in that room. It has two HEPA filter units that keep the air clean. It's where I store all my optical parts. The laser doesn't show at all without some smoke.

mechanicalmagic
12-17-2009, 08:45 PM
A question about those 630 nm laser diodes. I thought laser diodes had a very broad spectral bandwidth, (as compared to gas lasers/lamps). It would seem the coherence length would not allow any useful distance change measurement. Also, I thought they were mildly temperature/wavelength interdependent.

If you have a link to a good diode, I would appreciate you posting it.

Evan
12-17-2009, 09:53 PM
Even the relatively cheap 630 nm diodes are very good now if you don't over drive them. They are temperature compensated and very stable with very little drift. The diode above is from Deal Extreme and costs about $15. It's not a laser pointer, but a good quality diode with decent packaging and adjustable regulator and collimator. It is speced to run at 4.5 volts and is very bright but runs in mixed mode at that level. If you turn down the input voltage at 3.2 volts it switches to Tem00 mode and becomes super stable. It's still plenty bright enough to use for diffraction and interference measurements.

I plan on using automated fringe counting to tabulate distances for calibration purposes. The room that it is in has very stable temperature conditions that will not vary more than a degree or so over 24 hours. It's the same room that I have my seismograph in so I have years of data on just how stable the environment is. The seismograph is extremely sensitive to both tilt and temperature variations.

Greg Q
12-17-2009, 11:22 PM
I have an optical flat that came with a set of gauge blocks. Is one of those diode lasers the best way to get a suitable light for things like checking micrometer anvils and ga. blocks?

Greg

Black_Moons
12-18-2009, 02:07 AM
So does this mean we can send our messurement tools/gage blocks etc to be E.v.a.n. tracable calibration/certification? :P

Evan
12-18-2009, 05:19 AM
Greg,

It's one way. You don't need an extremely narrow band light source to easily see interference fringes. It can be pretty wide band relatively speaking. An ordinary super bright "pure green" LED is all you need.

oldtiffie
12-18-2009, 05:27 AM
I have an optical flat that came with a set of guage blocks. Is one of those diode lasers the best way to get a suitable light for things like checking micrometer anvils and ga. blocks?

Greg

Greg.

The theory of optical flats and their uses is at pages 196 and 197 of Culley/RMIT "Fitting and Machining" (copies posted here) available at:
https://www.machineryhouse.com.au/Products?stockCode=L341

If you don't have it, I can thoroughly recommend it.

http://i200.photobucket.com/albums/aa294/oldtiffie/RMIT_book_details/RMIT-Opticalflats1.jpg

http://i200.photobucket.com/albums/aa294/oldtiffie/RMIT_book_details/RMIT-Opticalflats2.jpg

See also:
http://en.wikipedia.org/wiki/File:Visible_EM_modes.png

From:
http://en.wikipedia.org/wiki/Electromagnetic_radiation

and:
http://bbs.homeshopmachinist.net/showpost.php?p=492764&postcount=66

So, if you have monochromatic light (source) and know the frequency it is not hard to get the wave-length.

The dark lines in an optical flat are one wave-length apart.

The optical flat MUST rest on a wedge of air. It will NOT work if it is flat down on the work-piece.

In my younger days I used optical flats in the Tool Room and Instrument Shops (we did a lot of our own small optical stuff). I have also used an optical flat that had a square grid etched into it as a reticule/graticule that was used on scraped surfaces to get the "closeness" and "spread" - ie the "count" - of a scraped surface. It was used with spirits and Prussian Blue - but the "flat" was pressed hard down(no air wedge) for that purpose.

It is hard to find more soul-destroying work than lots of scraping and lapping (all by hand) - especially on Mondays!!!.

Evan
12-18-2009, 05:52 AM
Here is the measured spectrum of a "pure green" LED. Because of the particular construction of pure green LEDs they have a very narrow band emission with most of the power concentrated within only a small bandwidth. In particular they have no emission in other bands.

http://ixian.ca/pics7/puregreen.jpg

Greg Q
12-18-2009, 06:34 AM
Tiffie: Thanks for the link. I have three of the RMIT texts, but have not read them all yet. I have read "Fundamentals of Dimensional Metrology" and others. I don't know much about machining, rather more about machine rebuilding including scraping. You're right about the boredom, for which there is the Biax power scraper. Also Zen, and ABC radio.

I have been sidetracked by the metrology aspects of metal work: I find it fascinating to the point that I'll be hoping for a copy of "The foundations of mechanical accuracy" under the tree. Geekoid.


Evan: thanks for the info on the green LED. That looks like a great solution for an affordable monochromatic source. As an aside, where are you with your LED worklight projects? Is there are particular part number that is optimum for a white worklight since I'll be ordering anyway?

Thanks
Greg

Evan
12-18-2009, 06:44 AM
My LED work light projects are on hold. I was building an LED flood light using my last four 5 watt emitters and accidentally connected them to 110 ac with no current limiting due to a faulty capacitor. They emitted a very bright flash for exactly 16.5 milliseconds and then turned into DEDs. (Dark Emittiing Diodes). I have to order some more.

As for recommendations, that depends on what is actually in stock right now as well as what is the best price per lumen per watt. That varies almost daily as new products come out.

Greg Q
12-18-2009, 07:05 AM
Bummer on the supernova. Thanks of the info. Next I have to find a copy of electronics for dummies.

Greg

lazlo
12-18-2009, 10:00 AM
Culley/RMIT "Fitting and Machining" (copies posted here) available at:
https://www.machineryhouse.com.au/Products?stockCode=L341

If you don't have it, I can thoroughly recommend it.

I'd love to order that book Tiff, but Machinery House wants more to ship it to the US than the cost of the book :(

I've looked for a used copy, but apparently that's the VoEd/Apprentice text in Australia, so it's not available anywhere else (including England, where the Royal Airmail overseas is cheap).

whitis
12-18-2009, 10:15 AM
I finally got around to setting up my optical bench so I can do some experiments in calibrating measuring instruments from primary standards that don't rely on traceability to a source. In particular, known wavelengths of light are as good as direct NIST traceability for accurate and quantitative calibration of distance measurements. It is relatively easy to produce precisely known wavelengths of light using readily available light sources such as low pressure sodium lamps or 630 nm laser diodes.


In order to have NIST traceability equivalent length measurements, your wavelength standards and the measurement instrument it is used in must be of very high quality.

traceability must go through the manufacturer of the laser measuring system to nist.
The manufacturer needs to use a iodine stabilized hene laser from nist and spend a lot of time to calibrate their laser and characterize drift
If the measurement is not done in a vacuum, you must measure and compensate for temperature, humidity, air pressure, and CO2 content which when done with state of the art sensors and algorithms still gives results that are several orders of magnitude worse than in a vacuum.
Calibration of the laser itself is not adequate - the entire system needs to be verified
Unlike gage blocks, lasers are notoriously unstable
the laser needs to be stabilized

Laser interferometry length measurement systems are so dicey that you can not just send one to NIST to have it calibrated.

http://emtoolbox.nist.gov/Publications/NBSTechnicalNote1248.pdf

Now, lets say your 635nm diode is accurate to +/-1nm. This is a very generous assumption. Many manufacturers of laser diodes don't even give a wavelength tolerance in their data sheets. Papers have been written about stabilizing laser diodes to +/-2nm accuracy. Now, lets say you measure 1". That is 40000 wavelengths. Which means your accuracy is now +/-40000nm or +/-1.57mils. My $16 chinese digital calipers are more accurate. Ondax makes special wavelength stabilized laser diodes such as the TO-640-PLR09. The accuracy? 640nm +/-1nm. By comparison, a sony SLD1332V is 670nm+/-10nm at 25degrees C and a sanyo DL-3148-023 is 635nm+5/-??? at 25 degrees C.

The wavelength of a typical Fabry-Perot laser diode changes by +/-0.3nm for every kelvin degree of temperature change of the diode (not ambient air, not case temperature). That is 472 ppm per Celsius degree or about 28 times worse than the tempco of stainless steel used to make calipers. Worse, it is in a power dissipating element that is hard to stabilize.

Now, if you have a NIST calibrated iodine stabilized HeNe laser, the accuracy of the wavelength is 2.3 parts per 10^10 (230 parts per trillion).
DealExtreme laser diode modules need not apply. Your DX laser is likely worse than a million times less accurate than the ones the big boys use.

Now spectral lamps, such as the common Pen-ray, designed for calibration can have better accuracies of a few ppm. But this accuracy depends on conditions that are not true for street lights and even these widely used calibration lamps degrade the accuracy by a factor of 20 or so by having the gas under less than ideal conditions. Still have to deal with air vs. vacuum, of course and filter out a single line. Used in a vacuum, if you did everything right, you would end up with a measurement accuracy perhaps equivalent to a Grade 2 gage block.
http://pas.ce.wsu.edu/CE415/PenRay_lamp_spectra.pdf

These low budget light sources are find for some applications; length metrology isn't one of them.

Evan
12-18-2009, 10:25 AM
Here you go Robert, all three volumes for $68 USD.

http://www.tomfolio.com/bookdetailsmem.asp?book=38651&mem=292

lazlo
12-18-2009, 10:32 AM
Here you go Robert, all three volumes for $68 USD.

http://www.tomfolio.com/bookdetailsmem.asp?book=38651&mem=292

Thanks Evan, but that's an ancient printing (1982), when it was three volumes. The current printing is 2008, and is two volumes.
The other issue is that bookseller is in Australia, and they don't quote shipping, so the shipping will be probably be similar to what MachineryHouse is charging.

In any event, sorry for the hijack :o

Evan
12-18-2009, 10:45 AM
These low budget light sources are find for some applications; length metrology isn't one of them.

I think that you possibly have even less sense of humour than I do.

However, would you care to explain why a laser diode from DX is less accurate than others?


Now spectral lamps, such as the common Pen-ray, designed for calibration can have better accuracies of a few ppm

Yes, I have one. I can also use the diffraction gratings blaze frequency as a standard to calibrate the laser by interferometry.

I also have a 1 Ghz and 10 MHz crystal oven timebases calibrated locally by a triple caesium clock stack to an accuracy of +- one count.

whitis
12-18-2009, 03:14 PM
Here is the measured spectrum of a "pure green" LED. Because of the particular construction of pure green LEDs they have a very narrow band emission with most of the power concentrated within only a small bandwidth. In particular they have no emission in other bands.

http://ixian.ca/pics7/puregreen.jpg

Very narrow? About 35nm full width half maximum. Or roughly 17,500 times wider than the line from a helium neon laser. That is as wide as a barn. There are theater gels that are about 100nm full width half maxium, such as roscolux 4490.
http://en.wikipedia.org/wiki/File:Helium_neon_laser_spectrum.png

If you are measuring over a significant distance, then the broad width can do some weird stuff. For example, the half power points are at approximately 509nm and 544nm. At a distance of approximately 0.0109 inches (and every multiple thereof), the extra fringes from those half power wavelengths constructively interfere to make a full power fringe. This will cause fringe counting systems to malfunction. And while the plot shows the spectra as a smooth curve, it is probably really a Fabry-Perot comb pattern, which is like having a bunch of narrower spectral lines at intensities that correspond to the height on the plot.

However, there is an easy solution - at least for the light source part of the problem. Set your diodes aside for other uses. The wavelength accuracy of a basic helium neon is reportedly around +/- 1ppm. If you actually get 1ppm, you are around grade 0.5 gage block accuracy. Not provable or NIST traceable. But since you can get a HeNe laser for under $100, not to bad. However, the variability of the refractive index of air with temperature, pressure, humidity, and CO2 content (greenhouse gas emissions long term, people breathing short term) is still a very big issue at this level of accuracy. NIST gets the refractive index error down to about 0.1ppm or 0.01ppm after very precise compensation for these variables. However, CO2 is a small effect, about 3.75ppb per person in the laboratory.
A 0.1ppm error appears to result from a 0.1degC change in temperature, 0.3mm change in air pressure, a 12% change in relative humidity, or a 670ppm change in CO2 levels. Thermal expansion of the components of the interferometer can also be significant. And bear in mind, these are absolute accuracies measuring these variables; relative accuracies aren't good enough. A $500 thermocouple temperature meter only gets you to within 1degC, not traceable. An omega HH804U $149 RTD meter (NIST traceable) and a P-M-1/10-1/8-6-0-P-3 $105 probe (1/10 DIN accuracy) gets you to 0.03degC to 0.08degC from 0 to 100degC. A $2,415.00 Omega DPI740-KIT barometer (NIST traceable), gets you to 0.02% on air pressure with a drift of 0.01% per year. An Omega RH820U $199, gets you humidity measurements within +/-2.5% RH, NIST certificate. The above instruments have USB or RS-232 ports. So, for a mere $2904 you have the necessary instruments to compensate for the refractive index of air without blowing your error budget.

http://documents.exfo.com/appnotes/anote154-ang.pdf

Oh, helium neon lasers can be calibrated for about CDN$2075 (still limited by the stability of the laser, of course):
http://www.nrc-cnrc.gc.ca/eng/services/inms/calibration-services/optical-frequency.html

http://ts.nist.gov/MeasurementServices/Calibrations/upload/4998.pdf

So, sure, you can save yourself the few hundred to few thousand dollars for a decent set of gage blocks. But you might need to spend as much on a bunch of other NIST traceable standards and measuring equipment to get there, plus your interferometer and laser, before you can derive your measurements from first principles and bypass NIST standards.

A mere 0.1" of glass or 0.3" of borosilicate glass between the reflective surface of you interferometer and the part being measured is enough to blow your entire error budget for Grade 0.5 accuracy with just 1degC temperature change. Fused quartz or zerodur will be better. Not to mention the material between there and the back anvil on the opposite surface. It won't take much to knock you down to Grade 3 or economy grade.

Metrology standards don't just easily come from scratch using first principles. Every accurate standard out there has relied on other accurate standards to get there. I.E. each generation precision standard has relied on the previous generation standards for measuring other quantities as well as the previous generation standard of the same unit. Sure, you can use an interferometer for spiffy length measurements but you become dependent on slightly less spiffy or as spiffy standards for time, temperature, pressure, humidity, voltage, etc. It took over a century of successive refinement and labs full of expensive equipment to get to the accuracy levels available today. This isn't to say some creative homebrew solutions aren't possible but it takes a lot more than $15 laser diodes or street lamps to get there.

Evan
12-18-2009, 04:16 PM
So, for a mere $2904 you have the necessary instruments to compensate for the refractive index of air without blowing your error budget.



Or, I can build a box of Lexan that I get for free from a local supplier and then walk out to my garage and pick up a tank of helium and fill the box with that after everything is aligned. That reduces the refractive index by an order of magnitude, for a few dollars. Then I can use a dollar store thermometer. Of course instead of that I would wind up a thermocouple from pure platinum and aluminum or something similar to give orders of magnitude better accuracy characterizing the system temperature.

Incidentally, it's extremely easy to calibrate a thermometer to an absolute standard. That standard is the triple point of water.

I'm not going to explain why but my work area is in a room that is insulated heavily on all six sides. As I already said, it is a very stable environment.


As for the pure green diode, it isn't the width of the spectrum that determines the accuracy. All that controls is how easy it is to see the fringes. The accuracy is determined by the actual wavelength of the peak power output and that is a function of the bandgap energy of the particular alloy and dopants used to make the junction. Pure green LEDs are special in that respect which is why they are called "Pure Green". The centre point of the spectral energy of a pure green LED doesn't vary more than a couple of nanometres.

Evan
12-18-2009, 04:23 PM
Every accurate standard out there has relied on other accurate standards to get there. I.E. each generation precision standard has relied on the previous generation standards for measuring other quantities as well as the previous generation standard of the same unit.

That isn't true. Temperature has been arbitrarily tied to the triple point of water consisting of an arbitrary mix of isotopes. It doesn't depend on previous standards. Length has been arbitrarily defined in terms of the velocity of light and owes nothing to a bar of platinum sitting under a bell jar in a lab somewhere.

davidfe
12-18-2009, 09:26 PM
Thanks Evan, but that's an ancient printing (1982), when it was three volumes. The current printing is 2008, and is two volumes.
The other issue is that bookseller is in Australia, and they don't quote shipping, so the shipping will be probably be similar to what MachineryHouse is charging.

In any event, sorry for the hijack :o

$ 85.30 including shipping from RMIT.edu.au

RMIT Publishing on 14 DEC 2009

HTH

David

mechanicalmagic
12-18-2009, 10:37 PM
The accuracy is determined by the actual wavelength of the peak power output and that is a function of the bandgap energy of the particular alloy and dopants used to make the junction. Pure green LEDs are special in that respect which is why they are called "Pure Green". The centre point of the spectral energy of a pure green LED doesn't vary more than a couple of nanometres.

You have no control over the actual frequency of the laser diode. If it mode hops from 510nm-550nm a graph like you posted is quite likely. It might spend more time at 530nm.

If you have an equal length leg interferometer this is no problem. However, if you are changing one leg, and the modes (wavelength) is changing, the fringes will die out in a very short distance. Mode hopping happens very quickly.

Evan
12-18-2009, 10:48 PM
I was talking about a light emitting diode used to illuminate an optical flat, not a laser.

mechanicalmagic
12-18-2009, 11:14 PM
I was talking about a light emitting diode used to illuminate an optical flat, not a laser.

Your original post was about an optical bench. To be used for some sort of callibration. I have seen one reference to optical flats, posted by oldtiffie.

A LED has a very broad bandwidth, a very short coherence length. Your fringes will die in a short length.

Evan
12-18-2009, 11:19 PM
Greg asked this question:

I have an optical flat that came with a set of gauge blocks. Is one of those diode lasers the best way to get a suitable light for things like checking micrometer anvils and ga. blocks?

Greg


If you actually read the thread you will see what the context of my replies is.

nheng
12-18-2009, 11:45 PM
As Whitis mentioned, laser diodes will have a spec on their central wavelength as well as a temperature coefficient related to the semiconductors involved.

One way around this is to use a TEC device. In these, the laser die is mounted on a thermoelectric element that can heat or cool. I use several different communications wavelengths such as 1625nm (+/- 10nm) that are TEC packaged (DIP or butterfly packages).

These give you 3 main benefits. If your central wavelength accuracy requirement is met by the device at its room temp spec., simply hold 25 degrees C with the TEC. If you need to coax the wavelength a bit, you can run it at either an elevated or reduced temperature, gaining the approximately 0.3 to 0.4nm per degree C adjustment coefficient.

The 3rd advantage of a TEC device is that the output power is stabilized by the controlled temperature. I've measured shifts in the 0.01dB range over 8 hours.

The TEC lasers I've used were in standard telecommunications windows but there's no reason you can't roll your own at other wavelengths. You may be hard pressed to get the very short time constants that a packaged device gives but if you provide tight thermal coupling between your laser package and controlled mass, your results can be very good.

A simple power transistor mounted to a heatsink with a thermal sensor can get you started. The transistor is driven on to dissipate power, the sensor and a control loop (as simple as a single opamp) control the heatsink to some point that is a little hotter than your hottest ambient temperature.

Den

Evan
12-19-2009, 01:27 AM
This green laser seems to be very stable and it is output compensated for die temperature. I have it mounted in an electrical housing in contact with the brass heat sink of the laser so that helps to keep it at a constant temperature.

Here is a diffraction pattern produced by the point of a pin. It is absolutely stable for as long as the voltage doesn't drop below 2.9 volts and stays under 3.2 vdc.

http://ixian.ca/pics7/laserdiff.jpg

whitis
12-19-2009, 10:55 AM
I think that you possibly have even less sense of humour than I do.

There was no attempt at humor in the line you quoted.



However, would you care to explain why a laser diode from DX is less accurate than others?

In the post you were replying to, I gave specific examples of diodes available from industrial suppliers that varied by an order of magnitude in accuracy specifications and even the good ones weren't good enough. And the ones that might be close to suitable are expensive and exotic. Which do you honestly think DX is selling at cheap prices for making laser pointers, laser shows, and other applications which don't require wavelength accuracy?

Now, I plan to get some laser diodes/modules from DX myself but have been slow placing the order. They do have their uses.



Yes, I have one. I can also use the diffraction gratings blaze frequency as a standard to calibrate the laser by interferometry.

You mean the ones which say 25 and 100 lines per mm? Not 25.0 or 100.0, let alone 25.0000 or 100.0000? Hope you have some better specs than that.
But you said you were going to get NIST traceable accuracy from first principles. If you use the grating to calibrate the laser and the NIST traceability isn't there, your NIST level accuracy goes out the window and if it is there then your deriving from first principles without relying on standards goes out the window. But, ok, you said known wavelengths and if you have specs for the diffraction gratings and allow for the error in your experimental setup and the drift of the diodes you can wind up with known wavelengths of some level of accuracy.

The patent mentioned on the grating containers doesn't give accuracy and an article (http://www.optometrics.com/products_gratings_gb.html) which refers to that 1947 patent gives typical dimensional accuracy specs of +/- 0.5mm. I.E. the rulings may be very precise but not necessarily very accurate. You need precision in a diffraction grating but accuracy is imparted when you calibrate the spectrometer it is installed in using a calibration lamp.

Using the gratings to calibrate is also only as accurate as your angular position or measurement of those gratings. The linked article suggests that the ruling angle vs edge of the grating is only good to 0.5degrees which translates to a length error of about 8727ppm if you rely on the edge.



I also have a 1 Ghz and 10 MHz crystal oven timebases calibrated locally by a triple caesium clock stack to an accuracy of +- one count.

Frequency is one area where you can get decent accuracy for peanuts. $31 will buy a 20MHz TCXO with +/-1ppm initial accuracy plus +/-0.025 over the 0-70C temperature range, 1ppm hit from soldering to board and 1ppm over the first year for a total of about +/-4.6ppm. Some GPS receivers have a time sync pulse output with good accuracy. Basic specs are comparable to some OCXO (oven controlled crystal oscillators). Some other OCXOs can do 5ppb over 0-50C, +/-20ppb over supply voltage, and 100ppb per year (500ppb over 10 years) for $121 qty 10. And with fancier stuff, much higher accuracies are available, as you know.

If you have an accurate wavelength standard, you can heterodyne that with an unknown wavelength, run it into a photodetector and get a frequency out which can then be measured. I have used that. Well, we had a multichannel spectrum analyzer on it since the frequency out wasn't all that clean, but the source was on another planet.

Or there is a grating doppler shift effect you can exploit using the timebase and the grating, but the results will only be as accurate as the lesser of the two.

You can get nice precision with a well built interferometer and that is a worthy accomplishment in itself. For $90, you can have a single 1" gage block calibrated to about 1.1ppm by NIST or for $5 to 2.3ppm by mitutoyo and can then hopefully convert that precision to accuracy approaching that of the gage block. You aren't bypassing NIST and doing it all from first principles then but you have a very practical, and likely cheaper, solution. Within the limits of stability/repeatability, of course. But if you calibrate your interferometer to the gage block within close time proximity of your measurement of an unknown, you don't need stability over a very long time. And all the stuff that throws the refractive index out of whack can be calibrated out with a cheap gage block instead of expensive test equipment as long as it changes slowly enough. No doors opened, Heat/AC turned off during measurement, everything warmed up and stable before use, don't breath on the equipment, etc. Instead of bypassing NIST, you have just leveraged them highly.

Of course, if you want to calibrate your laser using a grating, etc, enjoy.

I wish you luck on the interferometer, seriously, and will be interested to see what you come up with.

OT:18" of snow on the ground (Charlottesville, VA), and still coming down.

lazlo
12-19-2009, 11:52 AM
$ 85.30 including shipping from RMIT.edu.au

RMIT Publishing on 14 DEC 2009

David, don't want to hijack Evan's thread: I can't find any place to order at that RMIT university web page. PM sent...

Evan
12-19-2009, 03:05 PM
You mean the ones which say 25 and 100 lines per mm? Not 25.0 or 100.0, let alone 25.0000 or 100.0000? Hope you have some better specs than that.


The lines per millimetre do not define the accuracy of the blaze frequency of the grating. It is the angle of the triangular cross section of the lines that defines the accuracy of the blaze frequency. The number of lines per millimetre only determines the efficiency of the grating at a particular wavelength, not the wavelength of the grating.

With the grating I ruled recently the changes in colour resulted from changes in the angle of the ruling caused by changes in the spacing interfering to varying degrees with the previously ruled line.

You should know that. That you do not calls into question your other assertions. Credibility is so easily lost, isn't it?

Optics Curmudgeon
12-19-2009, 04:37 PM
Evan, you have quite a bit to learn about diffraction gratings, and quite a bit to unlearn first. You toss out the term "blaze frequency" as though it's some sort of absolute point of reference. Blaze frequency is just the reciprocal of the blaze wavelength. As you say, the blaze wavelength is determined by the angle of the ruling faces, but it isn't some magical distinct point, it is a positioning of the peak of the efficiency curve, which is broad and complex. Blaze frequency is rarely specified by grating manufacturers, normally the blaze angle and wavelength are specified. In the case of your P-E gratings, the are extremely coarse, intended for long wave IR, where frequency (wavenumber) is commonly used. Also, the variation in color you mentioned is the result of variations in groove spacing, not angle. It's the most common fault in gratings and the target or the greatest effort in ruling engine development. The papers by Babcock describing the efforts at Mount Wilson are interesting to read. The driving force behind using interferometers to control ruling engines wasn't to produce absolute standards of wavelength, but to eliminate ghosts and satellites. Unless you find a grating made with a substrate that has truly zero TCE, or can maintain it at exactly the temperature it was ruled at, you have a poor standard of wavelength. There are a number of references out there, like Newport's Diffraction Grating Handbook, Babcock's papers and others, that can fill hours of time, productive or otherwise. But who cares about productive, this is supposed to be fun.
Speaking of fun, why do I suddenly pop up here and comment? Because I usually watch these sorts of threads with amusement. They always go the same way, Evan proposes something, sometimes provocative. Someone replies, sometimes they know what they're talking about, sometimes not. When the reply is particularly pointed, and / or accurate, there occurs what I think of as "zig-zag and make smoke", where some irrelevant thing is tossed out, maybe to shake the pursuer. The post with the gratings and oscillators is an example. WTF was that? Two items totally irrelevant to the original post's point, which had to do with using interferometry to do precision measurement without reliance on external standards. So, I've finally reached a point where I had to comment, where it doesn't seem funny any more. Here's the deal, If you want to use LED's, cheap laser diodes, diffraction gratings or TXCOs for absolute references you'll get what you get, and if that's good enough for you than so be it. But don't say you've equalled what NIST or any other standards organization get, because you won't. Argue or make smoke all you want, but it won't change that. People that know what they are talking about have tried to help, and gotten dissed, talk about loss of credibility. By the way, if looking at the diffraction pattern of the head of a pin is the best diagnostic of laser performance you've got, you really are in trouble. Sometimes the biggest pitfall is not being aware of what you don't know.

whitis
12-19-2009, 06:50 PM
Or, I can build a box of Lexan that I get for free from a local supplier and then walk out to my garage and pick up a tank of helium and fill the box with that after everything is aligned. That reduces the refractive index by an order of magnitude, for a few dollars. Then I can use a dollar store thermometer.


It does not reduce the refractive index by an order of magnitude but it does knock the delta between the refractive index and that of vacuum by an order of magnitude, which does reduce the amount of the errors by an an order of magnitude if you purge all the air. Of course, doing a good purge could empty a 15 cubic foot tank of helium ($50). If you don't get a good purge and keep it purged, you might end up with more uncertainty than with air. Clever idea, though. A mere 1% air infiltration (or incomplete purge) alone, even at standard temperature and pressure, knocks your accuracy off by about 3ppm or down from grade 0.5 to grade 3. Dry nitrogen is very close to air in refractive index, thus the impact of air infiltration is less, though still undesirable.

Also hard to maintain thermal equilibrium while purging with gas. Expanding as it comes out of the cylinder really mucks that up. I ran dry nitrogen through a closed loop temperature control system before it reached the interferometer.



Of course instead of that I would wind up a thermocouple from pure platinum and aluminum or something similar to give orders of magnitude better accuracy characterizing the system temperature.

Incidentally, it's extremely easy to calibrate a thermometer to an absolute standard. That standard is the triple point of water.

One point at 273K is not much of a calibration.

Thermocouple probes with meter are typically only good to +/-1 Kelvin degree. Often worse. Platinum maybe 0.5 kelvin. With a commercial sensor that has been well characterized. With 3 point calibration 0.25 kelvin. Oh, and the triple point works better on paper than it does it the real world. Been there, calibrating temperature sensors on the thermal stabilization system for an interferometer. But don't take my word for it. Lake Shore Cryotronics, for example, says (http://www.lakeshore.com/pdf_files/Appendices/LSTC_appendixD_l.pdf) "beware of liquefied nitrogen and ice-point temperatures. They can vary 0.5 K. Use a calibrated standard sensor if possible."



I'm not going to explain why but my work area is in a room that is insulated heavily on all six sides. As I already said, it is a very stable environment.

That is great. Wish I had that here.



As for the pure green diode, it isn't the width of the spectrum that determines the accuracy. All that controls is how easy it is to see the fringes. The accuracy is determined by the actual wavelength of the peak power output and that is a function of the bandgap energy of the particular alloy and dopants used to make the junction. Pure green LEDs are special in that respect which is why they are called "Pure Green".

Actually, I did a simulation and what happens is that the first fringe is visible, the visibility decays dramatically over the next dozen or so fringes, and by the time you have traveled 3/4 of a mil, the fringes are invisible (down by four orders of magnitude). That is for a gaussian spectral distribution roughly similar to yours. The attachment button is missing, so I won't post a plot. But if you imagine the plot of the sound from plucking a very lossy tuning fork, you will have the general idea. At 20,000nm (787 microinches) round trip (half that in actual displacement), the brightness of the fringes is +/-0.007%. From 23696nm to infiinity, the brightness was "4000.00" at every point, below the noise floor on a 16 bit A/D, vs about 0-8000 on the first fringe. Simulation sampled both displacement and source wavelength every 1nm; interference for every wavelength emitted from the laser was summed at every displacement.


The centre point of the spectral energy of a pure green LED doesn't vary more than a couple of nanometres.

As I explained earlier, that is a huge error.

Evan
12-19-2009, 09:52 PM
It does not reduce the refractive index by an order of magnitude ...

Odd. The refractive index of helium is approximately (to six places) 1/8th that of air at STP. That's closer to an order of magnitude than not.


One point at 273K is not much of a calibration.

Tell that to NIST.


The kelvin, unit of thermodynamic temperature, is the fraction 1/273.16 of the thermodynamic temperature of the triple point of water.


Regardless, let's go back to my original post.

Quoting myself, "I finally got around to setting up my optical bench so I can do some experiments in calibrating measuring instruments from primary standards that don't rely on traceability to a source."

What exactly did you think I meant by "measuring instruments"? The best I have is a TESA dial indicator that is calibrated to 1 micron per division. That is approximately 2 times the width of the entire spectrum from red to blue. Why would I care about a deviation of a few nanometres by a cheap laser? I'm interested in values a thousand times coarser than that and even that would be (and has been) characterised by most of the members here as impossibly small to deal with (it isn't). You are arguing about factors and errors that have no meaning to my goal or the likelyhood of achieving it.


As I explained earlier, that is a huge error.

Not for my purposes it isn't.

If I throw up a fringe pattern from a green laser that is nominally 532nm and count fringes as I move a mirror attached to a caliper using a stepper, 1879 fringes later the caliper should have moved 1 millimetre. A consistent error in the wavelength of the laser of 5 nanometres (it's unlikely to be that large) will result in a cumulative calibration error of the caliper of about +/- 9 microns. In inches that is approximately 0.00035".

If I wish to check the accuracy of the gearing in the TESA all I need is one rev of the needle which is 100 microns. An error of 5nm per fringe will result in a cumulative error less than a micron. Furthermore, the error will be consistent which will serve easily to detect periodic errors in the movement of the indicator.

If I want to get fussier I can buy a 405 nm violet laser module from DX now for $25. While they don't have specs they are on the market because of the large surplus of 405nm diodes left over from the Blue Ray/HD wars. These usually have a variation in wavelength from nominal of no more than 1 or 2 nm at most.

beanbag
12-20-2009, 04:11 AM
Evan,

I am not sure what you are trying to achieve here. Can you please explain? What good is a calibration if it is only good to 1%? Perhaps counting fringes can tell you the amount of backlash in something, but what use is there if it tells you that your calipers or lead screw moved somewhere between 0.99 mm and 1.01mm? The measurement of periodic error makes sense, though, but are you sure your setup has the requisite stability to make repeatable measurements?

nheng
12-20-2009, 09:36 AM
Had to jump back in here. Playing with optics is fun ... I've been doing it for several weeks now.

But to be practical about it, I think you'd be better off with one or more good gauge blocks in the size range you want to "self certify :) at". Their accuracy is way beyond what a laser or diode is going to give you "off the shelf".

Laser wavelengths are all over the place, both initially and with drive current and temperature. If you had a HeNe laser you'd have a pretty good wavelength reference (ignoring phase and polarization issues here). With any laser diode, you'd have to measure the wavelength at a specific drive current and temperature and then duplicate those conditions whenever you want to compare. Reflections back into a laser diode cavity will also have a huge affect on output modes and hopping. For laser sources that we build (fiber optic test) we include an isolator at the laser output to prevent this.

In most wavelength measuring equipment, like optical spectrum analyzers, you will find a very stable and accurate reference because the grating (very fine ones indeed) are not taken as absolutes. What's interesting is that you can create a quite good reference with a piece of (I think) polycarbonate and an excitation source (neon lamp comes to mind). I can't find my notes or references to it at the moment but it is an option. The intensity of the resulting light is probably too low for other than use as an internal reference source in a more elaborate piece of test equipment.

If you'd like, I could measure a laser diode spectrum for you at specific temp and drive with NIST traceability. PM if you're interested but it wouldn't be for several weeks.

Den

Evan
12-20-2009, 11:26 AM
But to be practical about it, I think you'd be better off with one or more good gauge blocks in the size range you want to "self certify at". Their accuracy is way beyond what a laser or diode is going to give you "off the shelf".


That depends on what you use. While a DX laser may or may not be accurate since it doesn't have docs there are reasonably priced lasers available from places such as Edmond optics that are +/- 1 nm. with a bandwidth of 0.1 nm.

I was thinking about this during the night and it occurs to me that there is a pretty good and reasonably stable source readily available for calibration purposes. It doesn't get any closer to first principles either.

For green 532 lasers there happen to be a very distinctive set of iron lines in the solar spectrum from 5316 to 5328 angstroms that would serve nicely as a calibration of a spectroscope.

At any rate I am busy today with our 18th annual christmas party.

Evan
12-20-2009, 11:30 AM
but are you sure your setup has the requisite stability to make repeatable measurements?


Probably. The repeatability of my milling machine according to my Tesa DI is +/- 2 microns. To conduct these experiments I will need to build a much more accurate X/Y stage for which I already have some very nice Thompson linear rails.

Incidentally, there are other optical methods to measure well below a micron using simple methods. Think moire patterns from vernier screens.