actually I had already add a resistor in series with Cc, but it did't improve the PM so much, only a few degree. I find that the PM between second stage and first stage is enough, but after the second stage and before the source follower, it is not enough. The CL is around 33f. Do I need to add a Cc between the second stage and source follower? If so, it will affect the BW so much and how to solve the problem?
I am sorry that I can not remove the second stage, I don't know why, maybe for a large swing. This circuit is from a paper published by MIT. It is used for EEG signal acquisition