I am using the spectre doing the Fully differential OTA simulation. In the AC, the GBP is more than 200MHz. What puzzled me is that in the transient simulation, there always offset votlage almost 100mV at the beginning, and need more than 20us, offset can be reduced to nearly 0V. How to solve that? In the DC simulation, the DC offset voltage is nearly 0V.
If u mean that u have a problem with ur transient response for only 20us then the circuit operates well, so i think that this small error is just at the starting of the circuit and has not important to try to eliminate it, THIS IF THE CIRCUIT IS OPERATING WELL AFTER THIS PERIOD
do you have maybe some caps with "incorrect" initial conditions according to the initial conditions of your voltage(current) sources ...
you should be able to test it if you fiddle a little bit with your (signal) sources ...
anyway it could be as said before harmless if everything settles fina after 20us and you can afford the simulation overhead and get no further problems ...
one thumb of rule maybe help you:
the GBP of CMFB should be larger than ur differential GBP
that's equivalent to that common mode feedback should be established faster than the operation of ur differential amp.
This is because at the beginning, the output DC level is 0, and it has to be charge to the correct operating point. So, I guess in your problem, there must be a rising or falling envelop overlapped by the small signal, is that right?