Constant Gm-R topology comparison

topology comparison

Dear all,
I need you to tell me what is the difference between the 3 topologies?
What is the advantages and disadvantages of each one of them?
regards.

constant gm bias

A- This is the typical "constant Gm" bias generator. This circuit works pretty good, and produces an easily derived current level. In order to produce a significant voltage on the resistor, the current densities in the two NMOS are typically VERY different (either by having different current levels, or different aspect ratios). What this means is that there probably is not a true constant Gm point.

*Side note: I believe that the "Constant Gm" terminology comes from the use of this topology with bipolar transistors, where the transconductance vs current density that is very well behaved over a very large range. (Current in a bipolar circuit would be I=Ut*ln(Id1/Id2)/R, and Gm = I/Ut = Ut*ln(Id1/Id2)/R/Ut, so Gm = ln(Id1/Id2)/R, (where Id1 and Id2 are the current densities in the bipolar devices and R is the resistance). Unfortunately, MOS devices do not follow such a nice Gm vs current rule like bipolars, so the circuit does not produce a true "constant Gm" with MOS.

B- This circuit is similar to A, but has an interesting property: If the circuit can be designed with R approximately equal to 1/Gm of the left NMOS, there is less sensitivity to matching of the PMOS transistors.

C- This is not a "constant Gm" or Gm/R type bias generator, but is rather will be equal to Vt/R circuit. Because the voltage across the resistor is larger than in the other circuits, it will need a larger resistor for the same current level.