It really sounds like you are trying to follow g
m, not V
t since you are trying to compensate for mobility. There is a very simple constant g
m circuit that is found in a number of text books. One that I think of off the top of my head is the Johns and Martin book, "Analog Integrated Circuit Design".
I ran a quick search on google, and found a paper that describes the technique.
www.eecg.toronto.edu/~kphang/papers/iscas2004_nicolson_opamp.pdf. I have not used the additional techniques presented by the paper, but it does give a nice overview of the traditional way to bias amplifiers. One difference is that I always implement the resistor on chip, so I don't ever see the stability problems that they were talking about.
I realize that everything that I have talked about is generated currents that will produce a constant g
m, and you asked for a voltage. Do you really need a voltage? If so, and if you have a low tempco resistor available, you can just drive the current into a resistor.