[sdiy] Peltier Elements

[Ian Fritz]

At 05:33 AM 5/13/2003, Czech Martin wrote:
>The question is also how low the temperature must be.
>If the curves to be observed would be linear with little
>deviation, it would be enough to mesure at 25C and 55C
>to get a good aproximation of what's going on.

Yes, a quick-and-dirty look at 25C and 50C is often useful, especially if 
you are trying to dial in a correction.

>If serval factors come into play, some with positive, others
>with negative sign, I could think that the curves are at least
>of third degree (polynomial aproximation), so extrapolation
>of only two points could be misleading.

Yes, and even a standard tempco resistor has a small quadratic part.  Also, 
temperature differences within chips/compensators can give nonlinear 
temperature curves.

>In this case I think that one has to go lower then the
>expected operating temperature range of the circuit
>to get a better understanding of the first and second derivative.

Yes.  Another point is that some leakage effects come in at high 
temperature (FET input current) so you want to limit the high-T 
extreme.  Going cold then gives you a larger T range to help fittting 
non-linear coefficients.

>I think 10C would be sufficient even in this case.

Agreed.

   Ian