Hi all:
Now I am designing a fully differential amplifier with unity-gain buffer connection. The fully differential amplifier is a two stage miller compensation structure. When I simulated AC response, the bode plot was ok and the phase margin was above 75 degree. Also, when I simulate the step response with small signal, it worked well and did not oscillate.
But when I simulate the step response with large signal, it oscillated.
So could you tell me how to solve the problem? I appreciate your help.
You might look for transient saturation (if this is a bipolar
design) or "windup" of stages under high overdrive conditions.
These will give you large signal overshoots that small signal
analysis won't indicate.
Sometimes you will see features like cross-clamping of diff pair
outputs, to address this sort of thing.
Yes, I have checked the stability of your common-mode feedback loop, the AC response is OK. I use four equal resistors to form the unity-gain buffer. Two worked as input resistor and two worked as feedback resistor. Thanks!
Hello, I also have this problem, the common mode loop is almost ideal (DC AC are all fine) but it is oscillating in transient simulation. I don't know why? anyone have a clue?
I am quite sure it is due to the simulator, because the oscillating frequency is far beyond the unity gain frequency of the CM and DF loop. But what setting should i change in the transient sim?:?:
Hello, I also have this problem, the common mode loop is almost ideal (DC AC are all fine) but it is oscillating in transient simulation. I don't know why? anyone have a clue?
I am quite sure it is due to the simulator, because the oscillating frequency is far beyond the unity gain frequency of the CM and DF loop. But what setting should i change in the transient sim?:?:
To check to see if it is a simulation artifact, reduce the minimum timestep and see if the apparent oscillation changes. If it is a simulation artifact you will be seeing a triangular "oscillation" where the time between points is the minimum timestep. Another option is to change to "gear" integration rather than trapezoidal, although that is not usually the preferred solution.