🚀 go-pugleaf

optoeng@pioneernet.net (Paul Mathews) wrote in message news:<ed443a10.0204240648.678bd1c5@posting.google.com>...
> aseychell@vet.com.au (Adam Seychell) wrote in message news:<21a524b0.0204231906.329565f6@posting.google.com>...
> > I am trying build a photometer that will measure absorption of light
> > from an high intensity red LED through a substance. Knowing that LEDs
> > and photodiodes are quite temperature sensitive devices (in terms of
> > luminescence vs temperature for fixed current) I designed a
> > compensating photodiode circuit so the LED puts out constant
> > brightness. So the arrangement is a LED light source that shines some
> > of its light off a mirror to a compensating photodiode and some of its
> > light to another photodiode which does actual measurements. My circuit
> > works well and everything is mechanically stable, but the LED still
> > remains temperature sensitive. The only conclusion I can come up with
> > is that the distortion pattern emitted from the LED is temperature
> > dependent. Intensity change alone is canceled out with the
> > compensating photodiode so it must be some other effect. Does anyone
> > have information about this effect with LEDs ?
> > 
> > So I'm wondering if my idea of using a LED for a photometer is such a
> > good idea after all. I don't really want to do anything fancy like
> > calibrated microprocessor temperature compensation and ADCs, including
> > pulse sampling techniques to avoid LED self heating effects.
> 
> 
> It's actually rather difficult to insure that the reference optical
> path has constant coupling over temperature, and, of course, you also
> need to temperature stabilize your feedback photodiode circuit.  Here
> are a couple of alternatives to consider:
> 
> 1. Use a laser diode with internal feedback photodiode instead of your
> own LED/photodiode assembly.
> 
> 2. Drive your LED with constant current.  Add a temperature sensor and
> a microcontroller with A/D.  Calibrate the entire setup, so that the
> microcontroller corrects readings according to temperature.
> 
> 3. Scan your optics between 2 paths, the sample and a reference.  You
> can use an oscillating mirror, shutters and a beamsplitter, or other
> scanning method.  Use the reference channel to normalize the sample
> channel.
> 

Paul,

The problem lies in a mismatch in temperature sensitivities between 
the two optical paths. When I touch one leg of the LED with a 
soldering iron tip (1 second) the effect is easily produced. I can
also see
the output takes several seconds to settle after power is applied, as 
if the LED is heating up and stabilizing. I suspect the LED is 
changing its radiation pattern with temperature, thus altering
relative
amounts of light to the feedback and measuring photodiode. 
Currently a mirror is used to reflect part of the light to the
feedback photodiode. The two photodiodes therefore receive light from
two different
spots of the LED. Your idea of using a beamsplitter should get around
this
problem because the same light path will be directed to both 
photodiodes.

As for the circuit I'm using all 100ppm/K components and photodiodes
are run in shortcircuit mode (i.e across the inputs of a high
impeadance op-amp).