scanning michelson interferometer

[shrad]

Hi all!

I'm planning on building a small scanning michelson interferometer for optical spectrum analysis

What I'll do is basically acquire temporal interferences, apply a fast fourier transform, and then have a spectrum

With some known stable wavelengths (632.5,532,1064) I'd be able to scale the spectrum and have something readable

I have everything needed (reference lasers, optics, piezzo, photodiode with preamplifier) but there are some unknown variables

What would be the scanning frequency for the piezzo mirror? (I assume it would be the frequency at which there is maximum mirror displacement, if too small for a piezzo I'll use a speaker voice coil)

What would be the ideal optical path length? I assume it is not important and mirror displacement will be the major concern about resolution

I think I'll use a modified USB soundcard for data acquisition and piezzo drive signal generation, so if there are people (especially those involved with the soundcard DAC) who have experience about modifying or reusing the driver, it would be very nice if they could contact me, I'd develop a software solution for data acquisition for free

I'd be so happy to make a photospectrometer for less than $200 and to give people instructions on how to make one themselves... that would be awesome

[lasersbee]

Hi all!

I'm planning on building a small scanning michelson interferometer for optical spectrum analysis

What I'll do is basically acquire temporal interferences, apply a fast fourier transform, and then have a spectrum

With some known stable wavelengths (632.5,532,1064) I'd be able to scale the spectrum and have something readable

I have everything needed (reference lasers, optics, piezzo, photodiode with preamplifier) but there are some unknown variables

What would be the scanning frequency for the piezzo mirror? (I assume it would be the frequency at which there is maximum mirror displacement, if too small for a piezzo I'll use a speaker voice coil)

What would be the ideal optical path length? I assume it is not important and mirror displacement will be the major concern about resolution

I think I'll use a modified USB soundcard for data acquisition and piezzo drive signal generation, so if there are people (especially those involved with the soundcard DAC) who have experience about modifying or reusing the driver, it would be very nice if they could contact me, I'd develop a software solution for data acquisition for free

I'd be so happy to make a photospectrometer for less than $200 and to give people instructions on how to make one themselves... that would be awesome

I would be very interested to see a working unit when it is done....

BTW... I think you meant 632.8nm....


Jerry

[shrad]

you're correct about the wavelength

if anyone with the related knowledge has any tips, please tell me

I'll try to get a basic layout down as soon as possible to demonstrate the principle

[mixedgas]

Google model train wavemeter
find something by:
B. W. Petley1 and K. Morris1

or google:

A toy-train wavemeter

Dont laugh, its paid for, and works.

Steve

[shrad]

Hi Steve

You always have valuable hints, thank you!

I found that document when googling: http://tf.nist.gov/general/pdf/1296.pdf

I think it is a highly valuable resource accessible to the amateur scientists like me

I'll dig google deeper when I have time and try to find some corner cubes

Fringe counting seems to be a nice method, but I have to learn a bit more about it to see if it could lead to a multi-wavelength display as with any optical spectrum analyser (fourier might still be easier for that purpose, especially with a well designed software)

[RedlumX]

Well I was pondering to build something cheap in this direction, since both a grating based CCD spectrum analyzer and a scanning interferometer use special parts that can be expensive, are bulky or hard to adjust etc. I was having a simple thing in mind that anybody could build and use. I thought of using a parallel glass plate (like a microscope slide) and reflecting light from a slightly diverging beam off both sides, and throwing the interference fringes along a long linear CCF sensor (cheap scanner type). There are no moving parts and all is software: reading off the sinusoidal pattern from the CCD and making a fourier transform. And then it would be cool to display the spectum by a device similar to eg http://www.seeedstudio.com/depot/smt...kit-p-446.html

I didnt yet make any computation to see how feasible it would be and to what resolution could be achieved for a given setup. But at any rate, it primarily all depends on how many fringes can be recorded simultaneously, probably serveral hundred, perhaps 1000, may be the practical limit. The resolution should be essentially 1/(number of fringes). The application would not be in absolute wavelenght measurements but rather to check the longitudinal mode spectrum of laser diodes. For this purpuse the demands on accuracy and resolution are moderate, 1/1000 should be enough (0.6 nm resolution). It just should be simple to use so that one can do this concurrently when shooting holograms. The other methods mentioned above are too cumbersome to be used on a daily basis, for the amateur holographer.

I know how to do the math etc, but I am no expert in microprocessor programming so that's what kept me off trying it so far.

[shrad]

If a fourier transform is to be done with a microcontroller, you need at least a 32bit MIPS architecture to achieve decent framerates

I have a PIC32MX development kit which would do the trick, but I highly prefer to use an USB soundcard as fringe signal has the same bandwidth as the soundcard

Anyway, my first tests will be with an analog oscilloscope to view untransformed spectra, and then I'll develop the software and tune the soundcard circuitry to get the same reading in parallel... after that, the only needed computations will be apodization, FFT and scaling (and maybe close-loop the scanning mecanism and implement a motorized slit)

But these things are already too much, I must first manage to get a compact working design

[RedlumX]

BTW with a piezo you can scan only a minimal amount, in this way you can build a scanning interferometer which has a range of a few Ghz and resolution of a few Mhz (with the proper highly curved dielectric mirrors). This is useful for very fine spectrum measurements (a few Ghz). For scanning a wider range, what is usually done is to translate the mirror of an interferometer for a considerable distance, like 10cm, and count the fringes. This requires a very good mechanical setup... like a high quality translation stage with a stepper motor. Not without reason are the commercial ones (eg from Exfo/Burleigh) very expensive.

I think a CCD based grating spectrometer is easier to build (if you aim for a broad range with moderate resolution). For small ramge hi-res measurements a scanning interferometer the better choice. I have built both kinds (see my site), each for less than $200, but this was due to luck in obtainung the rare but critical components from ebay. And as I was writing, I am ponderng to build something simpler for crude measurements, with like 1 or 1/2 nm resolution.

[shrad]

Quite interesting, could you point the url of your website please? I'd like to have a look at it!

I'm looking for a quite large bandwidth (400nm to 1500nm) but I lack the math to calculate the needed translation... maybe you could suggest some simple equations with which I could approximate decent values?

The reason I wanted to go the michelson route is everyone could scavenge an alarm clock piezzo or a 8 ohm speaker voice coil, and it is easy to acquire/generate signals with a cheap USB soundcard

Now, I know it would be quite difficult to align incoming beams without proper corner cubes and the like, but as soon as an amateur has the ability to measure optical spectrums, I think alignment time is not very important

I have some linear CCD arrays too, but I have never found any datasheet or implementation examples to get mine working... otherwise I'd have built a small microcontroller based interface a long time ago, so again if you have some tips, it's welcome

[RedlumX]

Hi,

the site is here .

Well the equations depend on the precise setup you have, but at any rate, the resolution is roughly given by the inverse of the number of fringes you count.

And yes, CCD arrays need a microcontroller for readout, and I made use of some preprogrammed device (CLS553) from www.framos.eu. The timing requirements are very complex and it would be a nightmare to program this oneself.

Best,
W