[sdiy] Temperature compensation results

[René Schmitz]

Hi Ian and List,

> So you assume the same model equation for both transistors. I don't see 
> the justification for this -- small offsets could be caused by small 
> deviations from the equation.

Say you would extract the parameters for a SPICE simulation, you could 
fit XTI and EG to match with observed data. Because if you try to treat 
the transistor numerically, you have to use a model of some sort. Thats 
physics essentially, modelling the world arround us.

> I don't follow you here.  IS(T0) looks like just  a number that is 
> factored out. The rest of the expression collapses to unity at T=T0.

What I mean is that a perfect match could occur *by* that deviations 
from the model equation as well.

> I believe EG is just the Si bandgap, which is a fundamental number.  Why 
> do you think it would change?

I don't, thus I think that the quotient of Is1 and Is2 is a constant. 
Also I think XTI is a physical property of semiconductors and should not 
change as well.

If the parameters are the same for both transistors are the same, then 
as you say, the equations collapse, and become independant of T, thats 
what I was aiming at.

But supposed Is1/Is2 had a temperature dependance, it would be pretty 
mild, its 1st order term would generate a tune drift, which could be 
cancelled by applying a PTAT voltage or your trick, only the higher 
order terms could be annoying. But I guess for an "almost matched"
pair this shouldn't be too strong.

> I think the crucial question is whether ideal model behavior can be 
> assumed when you are looking for small effects that could cause mismatch..

I see it this way: If I randomly pick two 2N3904s they will have 
mismatch. I would also be able to find another pair to make two matched 
pairs. But this doesn't mean there is anything wrong with the 
transistors, just because I was looking at the wrong combinations of 
transistors.
(Unless one of them is off by miles in the Vbe figure.)

Also we should not forget that what we consider to be matched pairs, 
aren't really matched, they still have an offset. Its just smaller for a 
MAT02 than for a CA3046. So you would have to apply your argument for 
the MAT02, LM394 and so on as well. To me there seems to be no 
fundamental difference between a pair that is matched to 0.5mV to one 
that is matched to 50uV. At least there is none that would be predicted 
by the Gummel Poon model.

> Right.  But I think you can take care of this simply by putting the 
> circuit in a small enclosure, like an Al or plastic case.

Yes, that sounds like a good idea, and also takes care of most of the 
drafts.

> BTW, do you know that all the mail to your address uzs159 at uni-bonn.de is 
> bouncing?

I just checked, but nothing, but Martins adress is sending an autoreply.

Cheers,
  René

-- 
uzs159 at uni-bonn.de
http://www.uni-bonn.de/~uzs159