0

Variable Wavelength Emitter?
Moderators: adafruit_support_bill, adafruit

Please be positive and constructive with your questions and comments.

Variable Wavelength Emitter?

by gtaubman on Sun Aug 23, 2009 7:11 pm

Hi Everyone,
My parents are both wetplate photographers (http://en.wikipedia.org/wiki/Collodion_process). Apparently the plate is mostly sensitive to UV, so normal light meters don't work at all. I'd like to try building them a light meter. I'm assuming that calibration is going to be the hardest part. I was thinking it'd be useful to be able to emit light at a fixed intensity over a range of wavelengths, and then I could measure the plate's sensitivity over a spectrum. Then the next step is getting some photo sensors and knowing their sensitivity over the same wavelength range, voila, a light meter.

So, I was wondering if anyone knows where to get such an emitter? I've looked around online, but I don't think I know the right terminology to get what I want. I investigated RGB LEDs but they probably don't have a uniform intensity over the range, and I need UV as well.

Does anyone have advice? I don't expect this to be a quick project, but this seems like a good place to start.
Thanks!
Gabe
gtaubman
 
Posts: 3
Joined: Sun Aug 23, 2009 7:02 pm

Re: Variable Wavelength Emitter?

by John_NY on Sun Aug 23, 2009 9:03 pm

That's a non-trivial problem. I might start with Oriel (which might send you the "Oriel Light Catalog" or "Oriel Light Reference"), which supplies some good test equipment to major sensor & camera manufacturers. Newport, Melles-Griot, Thor Labs, PI (Photonics Instruments?) are other good names in the business.
To me, the question is "What is the minimum requirements of your instrument". It sounds like you are looking to build a variable-wavelength detector, and are asking about building a varialble wavelength emitter to calibrate.
If you did not emphasize blue and ultraviolet, I recomend a spectrometer/spectral measurement device. You probably want to base this on a grating and a bright incandescent light source, but it will be weak in the UV unless you have a very hot 200 to 400 watt bulb. These bulbs will have short life, and are much brighter in the red than in the UV, so accuracy in positioning the grating is important.
If you know the approximate spectral characteristics of the photographic plate, you may want to consider using LEDs of known brightness and known spectral characteristics. You'll want to get one that is stable over time and whose radiance over wavelength is well-described by the manufacturer. You'll also want to hook it up on a circuit with a stable power supply and the recommended operating voltage. I'd guess if you were cautious and meticulous you could get accuracy of 10% to 20% of the actual value at each wavelength. LED calibration is not something I've done before, so you might want to read up on it (search "led calibration source").
See if you can get the "Oriel Light Reference" and search on alternate "LED Calibration". If you want good accuracy you will need to track response over wavelength for source and filter (or grating), and try to find out what type of radiance changes could result from changes in power supply voltage and current.
You might alternately consider sending away for a NIST calibration of your sensor and an average plate. I don't know what it costs, but if you want accuracy under 5%, you might want to price it rather than buy the calibration equipment yourself!
Let me know please how low an accuracy you can settle for.
I have some time to think about this, so I'd be happy to figure out how to scale it down to your budget and requirements once I am near my reference bookshelf.
Also, what is the specific wavelenth range of the spectrum you want to measure?
-John
Update: I am definitely overthinking this. The purpose of this light meter is to make sure your plate doesn't saturate, right? You want to measure the spectral response of the plate first, and produce a curve over wavelength. If you are using solar-illuminated scenes (if your source is the sun) you want to weight your plate by the solar spectrum (5800K blackbody?) by multiplication. Then, take your high quality photosensor and put it in a good and stable circuit (or buy one), and put a filter in front of it. Multiply the photosensor quantum efficiency (QE) by the filter transmission and by the solar spectrum. Integrate each curve and take the ratio. That should be the ratio between meter radiance and plate response, right?
John_NY
 
Posts: 55
Joined: Sun Jul 05, 2009 2:54 am
Location: SoCal

Re: Variable Wavelength Emitter?

by gtaubman on Sun Aug 23, 2009 9:41 pm

Hi John,
Thanks for the detailed response! You're right, I wasn't very clear in exactly what I'm trying to do. Let me give it another go:

The long term goal is to build a light meter that is calibrated for taking wet plate collodion pictures. These pictures are all shot via sunlight. I was thinking that the first step in this project would be to understand the frequency response of a wet plate picture. I've looked around online and I haven't been able to find that information. This is the step that I was guessing I'd need a fixed intensity variable wavelength light source for. I could do a test shot where I fired it at a plate using a different wavelength as I moved the emitter, and then I'd be able to measure the frequency response.

I took a quick look at Oriel, and they seem to have a 150W "Low Cost" solar simulator for *only* $7k :D. Perhaps I could just get one LED at every frequency I can find, and use those to make a bootleg emitter for calibration?

Then the next step would be finding detectors that covered the appropriate wavelengths, and compensating for their frequency response. I think that with all of that information, you could build a moderately successful light meter unless I'm forgetting a huge chunk :).

Ahh, I just saw you updated your post. Yeah, purpose is to make sure it doesn't over or under saturate. The exposures are usually between 1/15th of a second and at most 5 minutes, so it doesn't need to have accuracy for high shutter speeds.

You lost me a little bit. I was thinking you'd take a reading with the meter for a fixed duration, you'd see what each of your fixed wavelength detectors said, you'd unweight them by their frequency responses to get what you believed to be how much light is actually coming in, then you'd multiply by the response of the plate at each of those frequencies, decide that you want the average to be something (18% grey?) and then you have your exposure time for that duration. I'm assuming that's a pretty huge generalization though.

Thanks again!
Gabe
gtaubman
 
Posts: 3
Joined: Sun Aug 23, 2009 7:02 pm

Re: Variable Wavelength Emitter?

by John_NY on Tue Aug 25, 2009 9:10 pm

Here are a few thoughts -- I haven't done all the research I hoped to just yet, but it's still planned.
I found, as I thought I remembered, that photographic response is logarithmic rather than linear. I don't have a physical feel for what that means until I look back at my old astronomy textbook to see how photographs can be used for photometry. The implication of this is that if you want to measure the spectral response of a photographic plate, you'll need to factor in the normal photographic response of the plate to changes in radiance before you know how much less sensitive it is to wavelength X compared to wavelength Y. Since the only two worries are whether you're getting too little or too much light, I don't know if that'll be important. This concern would come into play once you try to determine the spectral response of your plate (which I think may end up being a necessary step).
Measuring the response to different LEDs (e.g. from 400 to 800 nanometers Center Wavelength) might be a good way to calibrate a low-cost LED spectral measurement device.

The below steps describe the calibration of a spectral measurement assembly using multiple LED sources:

Since ordinary CCD cameras (or maybe only high-end ones) might have Quantum Efficiency or spectral response published in their technical documentation (or some enterprising soul may have done his or her own measurements and published them online), they could be used to calibrate an array of LEDs of different spectral wavelengths.
<oops: I forgot that ordinary CCD cameras are typically R/G/B! see "camera selection notes", below>
I'll call this well-documented camera the "Reference Camera", and it could be a digital SLR, or maybe even a slim consumer digital camera. You can use this to determine the relative radiance of your 5 or more different LEDs (especially those in the peak response region of the Reference Camera). Take a CCD image of the 5 LEDs using the same shutter setting and exposure time, and make sure that none of them are saturated. I recommend building either a separate circuit for each LED, or designing your circuit to run each LED at its proper voltage and current setting. You won't change the circuit from this point on, since you'll ruin your calibration by changing the circuit! I know from experience with the 3 color LEDs (R,G,B) in the adafruit starter kit that hooking the LEDs in parallel will reveal that some draw more current than others (and also that I should have put some current-limiting resistors in my circuit). If you change your circuit (or change LED bulbs, or go a few weeks without recalibrating the LEDs) recalibrate with the reference camera, and discard your previous calibration.
To calibrate, take an exposure with the reference camera, divide out the relative spectral response of that camera (either at the center wavelength only, or if you feel adventurous, try weighting it to the LED spectral curve), and normalize your resulting values to the brightest LED. That curve is approximately the true relative radiance of each LED compared to the radiance of the brightest LED. You may find some surprises at this stage (as to which LED is brightest).
You won't get the accuracy of a laboratory calibration setup, but this way is cheap and might get your curve within 5% of the true relative spectral response curve of your LED spectral calibration reference.

<Camera selection notes>
Black and white security cameras or second-hand astronomy CCD cameras may be your best bet, since they do not attempt tri-color imaging, and they may have decent documentation on their spectral responsivity. Security cameras may have automatic shutter features and be difficult to control so that you get a good image without saturation of the LEDs. Astronomy CCD cameras are a bit pricey and might require familiarity with some new software tools (I think there must be an open-source camera control package). There are other options if you can get a machine vision camera or other B/W CMOS or CCD camera.
Alternately, you could isolate the RGB channels of your camera using some photo-editing software and use the blue channel or to try to determine the relative radiance of each CCD. Since CCD cameras typically have some "white balance" function, you can't trust the output of the red-green-blue channels to have the same response relative to each other for all pictures. However, the "blue" channel should always have the same relative response once you isolate it. If you use the blue channel of a color camera, you'll want to have all your LEDs in the same image so you do not have to worry about the "white balance" screwing things up.
One final alternative would be to buy some Black & White photographic film whose spectral characteristics are well-documented. This might be the cheapest option, and you could use old astronomy photometric techniques to determine the relative radiance of each LED. (then, you'd correct for the known spectral response of the film, normalize to the brightest LED and you'd have your LED relative radiance curve)
</selection notes>

That's all for now,
-John

ps -- Variable wavelength sources are probably out of the question due to the cost, fragility and difficulty of building such a thing. If you can build an interferometer, you could probably assemble a serviceable monochrometer for spectral measurement, and I don't know of any COTS (commercial off-the-shelf) solutions that are not prohibitively expensive.

ps2 -- The goal of these steps is to determine a ratio of light meter response to predicted blue/UV photographic plate response. I haven't thought through an end-to-end project, but I'll post more when I think of it.
Last edited by John_NY on Thu Aug 27, 2009 12:30 pm, edited 1 time in total.
John_NY
 
Posts: 55
Joined: Sun Jul 05, 2009 2:54 am
Location: SoCal

Re: Variable Wavelength Emitter?

by macegr on Wed Aug 26, 2009 5:28 pm

This appears circular...you're building a light meter, so you want to build a light emitter to calibrate it. How are you going to calibrate your light emitter? With the light meter you're trying to calibrate?
macetech LLC - http://www.macetech.com
macegr
 
Posts: 292
Joined: Fri Apr 04, 2008 4:46 pm

Re: Variable Wavelength Emitter?

by John_NY on Wed Aug 26, 2009 7:56 pm

macegr wrote:This appears circular...you're building a light meter, so you want to build a light emitter to calibrate it. How are you going to calibrate your light emitter? With the light meter you're trying to calibrate?

Yeah, it's circular. I tried to compensate a bit with the "reference camera" idea. It turns out I cannot approximate the solar spectrum as neatly as I thought, since I forgot to account for the atmosphere.

To get a solar blackbody curve as it would be measured at the top of Earth's atmosphere (i.e. not useful here), you can plot the following over a wavelength series (use a spreadsheet or Matlab equivalent);
wavelengths = [200,210,220,...,480,480,500,510,520] nanometers;
blackbody_radiance = (NORM)*(wavelengths)^(-5) * (exp((h*c)/(k*wavelengths*1E-9*5800)) - 1)^(-1);
where
h = 6.63E-34
c = 3.00E+8
k = 1.38E-23
NORM = 4.46E+15

At the bottom of atmosphere, things get tricky:
Image

I haven't found a good look-up table of an average normalized solar spectrum measured from sea-level (possibly because it changes slightly with temperature, atmospheric content, water vapor, and many other variables), and running MODTRAN is not an option.
Hmm... still thinking this over.

-John

Update: Hey, maybe I can run MODTRAN! Look at this:
http://geosci.uchicago.edu/~archer/cgim ... ation.html
Hooray for wikipedia and University of Chicago! There's an option to set the sensor at altitude 0 and looking up, so you could simulate the solar spectrum. I'll have to compare it to my blackbody curve to see whether it reproduces the top-of-atmosphere versus earth-surface plot shown above...
John_NY
 
Posts: 55
Joined: Sun Jul 05, 2009 2:54 am
Location: SoCal

Please be positive and constructive with your questions and comments.