I had made a corner simulation of this circuit and I found out that it failed on the SS corner.
I can't explain why.
Do I need a start-up circuit for this?
Below is the circuit diagram.
Below also are the pertinent node voltages.
The DC output voltages of OPA in all corners are the same.
I just used the VB2 from the bias (VB2>VB1).In fact, I decreased the size of Mn2 (by 10 times) so that most the drop of Mn2 is much more higher than that of the Mn3. As much as possible I want a very low voltage of MN3 which also happens to be the node voltage of N4.
From this experimentation, I found out that the + terminal of the OP copied from the - terminal. I always believe before, that - terminal is the one who do copying.