NTC self-heating things, I am thinking whether thermal runaway will really affect the measurement. Therefore, it is not an offset, but actually a drift of resistance.

So I try to think about what will happen to such a circuit:

Due to the maximum power transfer theorem, the maximum power absorbed by the thermistor is when having the same resistance as Rref. Making this graph of power absorption versus temperature (ie resistance) and opposition, I get something like this:

Of course, in At 25 ° C, both resistors are at the same value. Now, for temperatures above 25 ° C, I can see that some negative feedback can be useful to prevent sensor drift: as the temperature increases, the power consumption decreases, and the temperature drops again. I expect it to remain at an equilibrium point low There is no such negative feedback offset.

On the other hand, for temperatures below 25 ° C, when the sensor raises its temperature, it will increase the dissipated power and further increase until it reaches Stop point at 25 ° C.

Now, if these assumptions are correct, how to consider this in planning? Will this effect be thrown away in some way?

I think such a problem also applies to certain "fixed" resistors, but it has obvious parasitic thermal dependence.