I am trying to get the transmittance through various liquid samples using specific wavelength LEDs. Originally I was experimenting with photodiodes, but do not seem to be ideal.

The target color range is near UV / Violet and Blue (~400-550nm). I am attempting to source a CMOS image chip that I can experiment with and am not sure as to which requirements I am looking at.

The basic requirements would be: Drivable with an atmel (Xmega, Atmega, etc). I think ARMs are more standard, or FPGAs but in terms of cost and programming complexity, would prefer to stay within microproc. territory.

Essentially I intend to sample the light from the LEDs with the intent of processing with an FFT algorithm in order to detect frequency components. A standard supported protocol like i2c or similar would be nice. Working with an AVR dragon.

Essentially help to find out what the important criteria in CMOS selection are.

  • 1
    \$\begingroup\$ What are you hoping to get from an FFT? An FFT of the light intensity won't tell you what frequencies of light are passing through the medium. \$\endgroup\$ – JRE Feb 1 '16 at 16:47
  • \$\begingroup\$ I'm not sure what league / budget you are in but the Public Lab: Smartphone Spectrometer may be worth a look to give you a very low-cost way of testing standard camera chips. It that works it may point you towards a solution - or maybe away from one. \$\endgroup\$ – Transistor Feb 1 '16 at 18:03

You won't be able to separate the light frequencies with just an optical sensor no matter how much processing you do on the output.

From your question it appears you want to build an optical spectrometer. The below is one way:

enter image description here

There are also methods involving micromirror (DLP) technology.

The idea is to optically spread out the light frequencies into a spatial distribution (as with a prism or diffraction grating) and then sense the light intensity at different points.

| improve this answer | |
  • \$\begingroup\$ Maybe I am missing a key point about fourier series? Isn't the light analog signal subject to that? IE, a DFT will show the spectral distribution? Is mechanical / optical filter necessary? \$\endgroup\$ – Subpar Greg Feb 1 '16 at 16:54
  • 1
    \$\begingroup\$ @JeremySmith It will show the spectral distribution of variations in the light intensity but the bandwidth will be MANY orders of magnitude too low to pick up the ~500THz light frequency. Nothing exists on earth that can do that directly in the electronic realm. \$\endgroup\$ – Spehro Pefhany Feb 1 '16 at 17:05
  • \$\begingroup\$ Thank you very much for the clarification. I was assuming I could view a dip in intensity at a certain spectral distribution, but from what you're saying, some sort of optical filter to get it into a single (or close) frequency would be required? Only looking for a dip in intensity of a fairly narrow region (~10nm). \$\endgroup\$ – Subpar Greg Feb 1 '16 at 17:21
  • 1
    \$\begingroup\$ Yes, a diffraction grating would be normally used. That way you can look at the light intensity, say along a line sensor and see the spectrum from the variations along the sensor length. You may or may not want another sensor that measures light that doesn't go through the sample. \$\endgroup\$ – Spehro Pefhany Feb 1 '16 at 17:24

Using the Fourier Transform to analyze wavelength in an optical system is possible, and even widespread, but it is much more complicated than you think.

Please keep in mind that using an FFT/DFT for audio analysis requires sampling the audio faster than the twice the highest frequency to be analyzed (Nyquist Theorem, right?). Unfortunately, to apply the same concept to light requires exactly the same constraints, and since visible light has frequencies in the (roughly) 500 THz range, a simple approach requires a 1000 THz A/D converter, and you won't find one of those very easily.

But (I hear you say) you said FFT/DFT is widely used - how can that be?

Well, the idea is called Fourier Transform Infrared Spectroscopy (FTIR) , and it uses an interferometer with one arm which is precisely movable. Moving the arm causes different light wavelengths to interfere and beat at different rates. With the arm moving at the correct rate and the proper A/D converter looking at the resulting optical signal, an FFT will identify the beat frequencies and their amplitudes, which allows the wavelengths of the components to be identified. Also see here, for example.

This scanning analysis is most useful for IR, but can be used to look at visible and near UV. A homebrew FTIR is certainly possible, but be warned that it's not for beginners and will cost you. Precision motion control, machining and optics are required.

| improve this answer | |
  • \$\begingroup\$ Thank you very much for the reply. I was confusing FT spectroscopy and regular spectroscopy. I see that the use of something like the Michelson IF is needed to get the beating. I am not a beginner, but also do not work for a spectroscopy manufacturer. (EE Degree, and work as one). I have a budget and access to quality 3-d printers and mech engineers for design. Ball park curiosity, what sort of cost are you imagining? Like 500? or 5000? And is CMOS detector pretty much irrelevent? \$\endgroup\$ – Subpar Greg Feb 2 '16 at 16:17
  • \$\begingroup\$ @JeremySmith - Depending on your ingenuity, a "working" FT spectroscope ought to be well under 1k. In principle, all you need is a collimator, a beam splitter, 2 mirrors, and a photodetector/ADC combination, with some way to linearly and smoothly move one of the mirrors. eBay would seem a good source for the optical components, but I'm not sure about moving the mirror. A saving grace is that you really don't need (relatively speaking) very high wavelength resolution, which eases your movement precision requirements. The Devil is in the details, of course. \$\endgroup\$ – WhatRoughBeast Feb 2 '16 at 17:28
  • \$\begingroup\$ Really good information. Thank you. Also, if I were just looking at eV delta (absorbtion) of a narrow frequency range, could I just use something like a laser, or narrow band LED through an optical filter? Specifically just looking at ~420nm (+-10nm)? \$\endgroup\$ – Subpar Greg Feb 3 '16 at 23:50
  • \$\begingroup\$ If you're just interested in a single waveband, you can indeed use a filter and a detector. Be aware, though, that narrow filters will cost you. Here search.newport.com/?q=*&x2=sku&q2=10BPF10-420 for example is a commercially-available 420 nm filter with 10 nm bandpass. Of course, it's $160. Of course, you can skip the filter, and just illuminate with a laser or LED, but your 420 nm spec is hard. There simply don't seem to be a lot of choices at that wavelength. If you're willing to accept 445-450 nm, there are plenty of lasers on eBay fairly cheap. \$\endgroup\$ – WhatRoughBeast Feb 4 '16 at 0:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.