[sdiy] Temperature compensation results

[René Schmitz]

Ian,

> I would say that what you are describing in engineering, not physics. 
> You pick some ideal model and blindly assume it is correct.  

I didn't mean to imply that.

> Again, if the mismatch is caused by defects, edge leakage, etc., then 
> you have to use model equations that include these effects.  *THAT'S* 
> what physics is about.

I agree with that.

A refined model would still be just a model. You end up with some sort 
of equation. I'm not saying that the models used in spice are the last 
word. (Hence there is VBIC.)

> That's circular logic. You assume an ideal model, then conclude that the 
> behavior must be ideal.

No, I merely state that this model doesn't predict the effect you are 
suspecting.

I would want to know what makes you think that these nonideal effects 
aren't present in the transistors which are considered "matched"?
Or put differently, a mismatch of 50uV would be caused by a difference 
in saturation current and a 1mV mismatch won't?  Where would the 
boundary be?

Cheers,
  René

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