A overshoot problem which costs me over 100 hours!

[xiexi]

The op is connected as the figure. When the capacitor is charged and discharged there is an overshoot at the capacitor.
The op's phase magin is about 90 and when I only connect the op as unity gain buffer with the same cap load and input the step as simulus at the Vin+, there is no overshoot at the step response. So the op is stable.
Now I am confused, the op is stable why does the overshoot comes out in this case. I guess the cap is charged very quickly and faster than the op respone time, so the overshoot comes out. If it is so, how can we solve it. Please give me intruction, thank you all.

 

[echo47]

Here's a smaller image to help folks respond to your question.

 

[steer]

You did not show where is the output in your op-amp. Depending on your answer, the op-amp either 2 or 2.5 stage. Switched-cap circuits usually based on a single stage amplifier. Correctly designed single stage amp has no overshoots. On the other hand, 2-3 stage op-amps do have it, unless super-carefuly designed.
Why dont you go the usual path and use a single-stage amp?

 

[xiexi]

I am sorry. Vout is at the follower's source of the 3rd stage. Now I am very worried is the stability. Is the overshoot caused by the stability?

 

[steer]

Yes, it's stability problem. Your op-amp is stable, but I doubt you can completely eliminate overshoot in such an architecture. At least not across process and temperature variations.
I strongly recommend you to use a single-stage op-amp, such as casode or folded cascode, depending on your voltage swing range.

 

[xiexi]

But the one stage's op' gain is very low and the offset is large, can we do anything to improve them?

 

[xiexi]

If the one stage output connecting small res, the amp's gain is very low. How can I solve the problem?

 

[steer]

OK, I see what you are saying. Then 1-stage amp does not fit.
If you are forced to use 2-stage amp, I'd recommend to put a small resistor in series with your switch Q.
This should reduce the overshoot and help to improve the amplifier stability.

 

[sutapanaki]

here is one speculation about the overshoot. Initially the opamp is disconnected from the 3p cap, which is charged to 0.6v. Then you connect the cap to the output of the opamp, but at that initial moment the input difference to the opamp is 0.4v and the amplifier may be going into slew rate limitation at the output or at least the diff-pair is strongly disbalanced. The current of the first stage is going through the compensation network flowing into the NMOS of the diff pair. This current creates a voltage drop across the resistor with a polarity that will make the output of the second stage spike up and also the output of the opamp.
Try to decrease the resistor (or remove it) and tell us if this helps with the overshoot. If you really have 90 deg of PM you can play with the resistor value till you get somewhere 70-76 degree.

 

[xiexi]

To sutapanaki, reducing the compensate res doesn't help. And steer's method is useful. Will steer explain the reason. My explain is that the charging speed is faster than the op. When a res in series with the cap, the charging speed is reduced which matches the op's speed well. Is it right?