Hi, there:
    I am designing a constant gm biasing block. stable issue is analyzed using Spectre's STB analysis and a voltage source is emploied as the probe component. Below are the schematic and simulation result.
   Now here are my questions
   1.  Why the phase of the loop gain at low frequencies is 180 deg instead of -180 deg? Is this because T(return ratio) = -Vo/Vx?
   2.  If I want to improve the stability, what should I do? adding Miller cap? then which branch should I choose to hook up the comcap?
   3.  Any analytical papers with regard to the stability issue of this circuit?

Thanks In Advance

Hi Julian,

Please be aware of the fact that you have two loops in this circuit. One of those is a positive feedback loop, while the other one is a negative feedback one. In order to have a stable system the negative feedback loop has to have a larger gain than the positive one. It is ok to analyze the stability of both loops at the same time by "breaking" the loop where you are placing the probe, since both loops are sharing such node.
The opamp BW should be larger than the pole of the bandgap circuit "without" the opamp in order to make it stable. So one option is to increase as much as possible the GBW of your opamp (not an easy task).

Hope this helps
Tosei