🚀 go-pugleaf

aseychell@vet.com.au (Adam Seychell) wrote in message news:<21a524b0.0204251637.5d74991f@posting.google.com>...
> Tony Williams <tonyw@ledelec.demon.co.uk> wrote in message news:<4b2cc6c7e6tonyw@ledelec.demon.co.uk>...
> > In article <21a524b0.0204250322.19119138@posting.google.com>,
> >    Adam Seychell <aseychell@vet.com.au> wrote:

<snipped the good oil from Tony>
 
> >  The real answer to the problem is to have two
> >  identical light channels, one as a reference (with a
> >  known absorption), and the other being the measuring
> >  channel. The output is the ratio of the two channels.

Search on dual beam spectrometers ... I got lectures on the subject as
a chemistry undergraudate in 1962.
 
> I'm glad you mentioned this Tony. I thought I was doing something 
> horrible wrong, but I have now learned that stabilizing absorption 
> photometers is not a trivial task. Well you seemed to eliminate your 
> optics problem with a temperature stabilized housing, but there still 
> remains the stability of the medium you are trying to measure. I'm 
> not yet sure if this will be a problem in my case. I like your idea of a 
> two channel device. I'd rather build two simple devices than a single 
> complicated one. Was there a need for a stabilizing photodiode even 
> though temperature remained at a constant 35�C ?

Yes. The biggest single source of temerature variation in the system
is likely to be self-heating in the LED junction - a normal-ish
non-red LED running at 20mA is dissipating 40-50mW in the junction,
and IIRR, that corresponds to some ten or twenty degrees Kelvin of
self-heating, with a thermal time constant of a few minutes, which in
our system cuts the LED output by about 20% as it warms up.

At one point in the development of our system, I allowed myself to be
seduced by the relative cheapness of open-loop compensation, and we
stuck a thermistor into a hole drilled in the back of a 10mm diameter
LED, not too close to the junction, and arranged for the LED drive to
rise as it got hotter. This reduced the excursion on the LED output
over warm-up to about 1%, but stretched the settling time to about an
hour - a very bad idea.

Now we glue a bit of glass fibre onto the side of the LED package with
UV-setting optical glue, passively monitor the light output with a
second photodiode at the otehr end of the (plastic) fibre, digitise to
16-bits (an over-kill, but the A/D converter was cheap) and get the
microprocessor to computer the ratio. Seems to work.