kell said:

Some references describe diode tempco as -2.1 mV per degree C and also

as 3300 parts per million.

If parts per million, then the tempco would vary with the forward

voltage of the diode, for example, with a very low forward current, Vf

could be half a volt; at 3300 parts per million that comes to about

-1.7 mV. So is it -1.7 or is it -2.1; in other words is tempco a

constant or is it a proportion of Vf?

The voltage across an ideal diode is the natural log of a ratio of

bunch of terms that involve doping densities and thicknesses and

widths multiplied by the diode current. And that log is multiplied by

k*T/Q.

The derivative of all that with respect to T is the same thing with

the T replaced by 1. So as long as the current is constant (and all

those other constants do not vary with temperature), the slope is

constant, or a fixed increment per degree, not a fixed fraction of Vf

per degree.