Help me out with opamp problem

[dipak.rf]

It is a folded cascode opamp and all the baises are fixed.

My problem is when i change bias3 (see in attached figure) by lets say 0.01%(assuming any variation for bias change) the output is changing drastically to half or more, but when i change bias4 (by same amount) the output doesnt change much (very less amount ) even though impedences for both the transistor are same and gm also comparable

Ques1. why much change in output for bias3 variation as compare to bias4?
Ques2. Is it good have this much variation (approximately 50%) for small amount of bias variation?
Ques3. If it is not ok then suggest me a solution?

Thanks in Advance

 

[saro_k_82]

1. The gain from bias3 to OUT will be higher than it is from bias2. From bias2., it is a source degenerated CS amplifier and the degeneration is so high that OUT is insensitive to bias2. But bias 3 has two parallel impedance contributions to it's source degeneration. So if all the conditions for all the transistors being same, you could expect 6dB gain difference. If that is more it tells that the impedance is dominated by the input transistors' channel length modulation. Check M13 and M14
But before all this first check whether all transistors are properly biased and have sufficient margins above vdsat. So much of sensitivity can only come from poor bias margins.

2. No good design can have so much of dependence on bias

 

[drDOC]

It also seems to me that you have an fully differential opamp without common-mode feedback, and you are simulating it really open-loop. My guess is that with proper feedback you might be able to get a correct biasing.

 

[Warlike]

M0 and M1 work in saturation?