temperature stable photo detector (and emitter) for power measurement

[SArepairman]

What are the most temperature stable photo detector? I simply need to detect the magnitude of visible light. The temperature of the detector and emitter will change by ~20c. I might be worrying about nothing, but I would like to  make wise choices, in case this proves a hurdle towards calibration.

http://en.wikipedia.org/wiki/Photodetector

Wiki has a bunch of them.

At first thought I would think to use two LEDs.


As for the emitter,
 I don't need much light, a single dim LED is enough.

Making a temperature stable current source to put energy into something is easy enough, but what should that something be?

some kind of light bulb? LED? Perhaps a particular kind of LED?

This is for detecting a change in light conductance through a sample (like a photo spectrometer)



Is there something like two LED's on a common substrate? I would imagine the detection capability of one LED would scale with the emission capability of the other LED over temperature, ? I could use fiber optics to route light in this case. (or its still non linear even if they were made at the same time on the same substrate? )


I guess i can over nize it if no near solution exists, but that is pretty ugly.

*wave length is not critical. So long it is visible light.

[Paul Price]

if you can work with a simple and very cheap chip that can do some think, you could easily  program it to use it to calibrate your  light sensing circuit to compensate for the temperature changes, you just add a temperature sensing chip or use a cheap thermistor and calibrate it with a thermometer.

 You might find that adding intelligence to your hardware design would be a very nice wayl to make  your project  do something with precision and  versatility.  An easy breadboard-able solution would be the Tiny Arduino or even the so very much more cheap and easy to use, Chip on a DIP 44 pin breakout board, the ARM-Cortex 32-bit CY8CKit-049-42xx which costs only $4 US that programs directly from a USB cable to your desktop/laptop, and yet can be easily connected to a wire-less breadboard.

[miguelvp]

This is for detecting a change in light conductance through a sample (like a photo spectrometer)

...

*wave length is not critical. So long it is visible light.

If you don't care about the wavelengths then you don't need a spectrometer, just a simple candlelight meter might suit your needs?

Like a photography light meter, I would hope they are still accurate at different temperatures.

Maybe I'm not understanding what you are asking and i'm oversimplifying it.

[miguelvp]

BTW I think Paul meant the CY8CKit-049-42xx prototype board.

Edit: For example these two project will measure light on their pioneer kit, the project should be fairly easy to port to their $4 prototype kit, only difference is that you have to make it bootloadable. The Pioneer kit goes for $25 otherwise.

http://www.element14.com/community/message/76059/l/psoc-4-pioneer-kit-community-project06-danger-shield-with-light-sensor-control

http://www.element14.com/community/message/89328/l/psoc-4-pioneer-kit-community-project096-light-sensor-project

Edit again: but it seems the sensors they are using can also be used on an arduino or anything that has a decent ADC.

[SArepairman]

Well I am thinking about what I am building and I guess a thermometer makes sense, as fiber optics will be necessary. In that case I can use two LEDs, a thermistor and some epoxy, there is a ADC to read the output of the photo detector so it can be multiplexed to the thermistor. I don't need the oven.

This seems like a good medium stability solution. It will probably be enough, but I would welcome additional discussion.

[SArepairman]

actually, that has the damn problem of requiring a temperature sweep for calibration. Component variation 
very production unfriendly.

ovenizing cures this some what, but thats way uglier.

I would like to avoid extremely time consuming calibration procedures and reliance on buying out a stock of something. I guess maybe a shit load of LED's from the same batch might do it but I don't really like it.

[David Hess]

The usual solution for this is to use two detectors and one source.  Photodiodes are pretty accurate over temperature (500ppm/C) compared to an LED source so one is used to measure the output of the source directly and the other is used to measure the output through the sample which helps cancel out their already small error over temperature.  The sensing photodiode may be used as part of a feedback loop to drive the source to constant output.

[SArepairman]

ahhh that makes sense.

And here I am thinking about how to hack one of those tiny chip-scale 7-segment displays (common substrate multiple LED).