I'm engineering a system that is temperature-sensitive, between ~-10-30C. I've moved to using cermet potentiometers because of their sub-200PPM/C temperature coefficients.
If I wire such a pot with the extremities at gnd and Vcc assumed to be temperature-invariant, and take the wiper as a Vref, then can I assume that the "lower" and "upper" parts of the pot on either side of the wiper have equal temperature variance and so Vref's temperature variance will be nearly zero?
It's a reasonable assumption that the tempco will much be better than using it as a rheostat, and your statement should be reliable for some value of "nearly".
A wild guess would be that you can expect 10-20:1 improvement if you draw negligible current from the wiper.
Rather than using GND and Vcc, it's often useful to shunt a cermet pot element with a low tempco resistor.
Of course the best approach is to first minimize the adjustment range, then the effects of pot position, CRV and tempco are minimized, then play with the other factors.
To a first approximation, yes.
If you want to get really, really picky, you could consider that, for unequal resistances in the two sections (pot not at 50%) the two sections will dissipate different amounts of heat. It's true that the section dissipating more heat is larger, so the temperature rise should be about same, and the two sections are thermally coupled so they ought to be about the same, but there still might be some tiny effect. This would be accentuated if the wiper is not driving a very high-impedance load, since then the two sections would have different currents and different heat dissipation.
But as a practical matter, yeah, you can usually consider the voltage ratio (input across pot to wiper voltage) to have a zero TC.
