Phase Margin for fully differential OpAmp

[Fabien]

Hi everyone,

I'm trying to size the miller cap for a fully differential OpAmp. Without the miller cap, the phase is around -230° at 0dB (open loop, differential output). I guess, the OpAmp is not stable, but I try to run some simulations and I'm not able to see it oscillating!
Anyway, I'm just wondering what should be the phase margin? 12° or 45°? (it will result in a lower gain, for sure).
Should I check the phase margin for each independant output (+ and -) or for the differential output (Vout+ - Vout-)??

Thank you for your help.

 

[nitishn5]

The Phase margin should not affect the DC gain at all.

The better the phase margin, you would see less ringing at the output with step inputs. 45° is a good number.

I used to check for the differential input to differential output.

 

[LvW]

Without the miller cap, the phase is around -230° at 0dB (open loop, differential output). I guess, the OpAmp is not stable, but I try to run some simulations and I'm not able to see it oscillating!
.

The phase margin needs to be determined from the loop gain - not from "open-loop" gain of the amplifier only.
What do you mean with "open loop"?

 

[Fabien]

What do you mean with "open loop"?

By "open loop", I mean nothing is connected between the output to the inputs.

Here is my test bench and the gain and phase results (without Miller caps).

nitishn5: for sure it doen't affect the DC gain, but it does for the GBW, so it does for the specifications. GBW should be 5 times the switching frequency.
I tried some pulse waves at the inputs, but no oscillations at the outputs!?

 

[nitishn5]

You need to put a feedback and close the loop before applying the AC signal.
There is a method using ideal inductors of huge values in the feedback path to set the DC operating point and using an ac source to get the loop transfer function plots. I don't remember it exactly but I have seen it in this forum.

 

[kenambo]

phase margin should be between 45 to 60 degrees... to achieve good stability in closed loop configuration....

thanks

 

[Fabien]

I agree but... if the OpAmp is stable in open loop, it should be in closed loop. As it should be used for SC cap, the OpAmp will work (in a little laps of time) in open loop, and the closed loop also depends on the caps I'll use.
Anyway, I'll check for the inductor method, but I still don't know how to check the OpAmp oscillations regarding to the phase margin...

 

[LvW]

I agree but... if the OpAmp is stable in open loop, it should be in closed loop. As it should be used for SC cap, the OpAmp will work (in a little laps of time) in open loop, and the closed loop also depends on the caps I'll use.
Anyway, I'll check for the inductor method, but I still don't know how to check the OpAmp oscillations regarding to the phase margin...

Fabien, I have the feeling you are not sufficiently familiar with the term "phase margin". Therefore, some basics:

*The phase margin applies to systems with feedback only. It is a measure for "safety" against oscillations in case of a closed loop - but it will be measured/simulated in OPEN loop conditions.
* Thus, you have to open the loop for applying a test signal ac source.
*But the problem is: In many cases, there will be no stable DC operating point under open loop conditions.
*Therefore: Open the loop and place a very large inductor (1 H or more) between both nodes (left and right). Thus, the loop is closed for dc but open for frequencies of interest.
* Then, inject a test voltage into the feedback path using a very large capacitor (1 F or more at one end of the inductor) and measure/simulate the voltage at the other end of the inductor.
* The voltage ratio gives you the loop gain response, which can be evaluated (magnitude and phase) to see the phase margin, which is the phase difference to -360 deg at the point where the magnitude is 0 dB.

 

[Fabien]

*The phase margin applies to systems with feedback only. It is a measure for "safety" against oscillations in case of a closed loop - but it will be measured/simulated in OPEN loop conditions.

Sure, that's the reason why I plot the Gain and Phase in open loop. In my simulation, the DC operating point seems to be stable, so if I understand, the inductor is not useful in this case.

The thing is: in OPEN loop, with no miller cap, the phase is far below -180° at 0dB. It seems the OpAmp is not stable! That's what I understand and what I can read in the books. If it's not stable, if I close the loop, the OpAmp may oscillate. So I put a pulse wave in this condition at the inputs, and nothing happen.
So the question is: how to see the system unstable ? Because with a phase margin of 12° (what I set with miller cap) the OpAmp should be stable but may have decreasing oscillations in closed loop. What I can't see.

May be I'm wrong...

 

[LvW]

So I put a pulse wave in this condition at the inputs, and nothing happen.
So the question is: how to see the system unstable ? Because with a phase margin of 12° (what I set with miller cap) the OpAmp should be stable but may have decreasing oscillations in closed loop. What I can't see.

What means "...and nothing happen"? Is the opamp dead? No change of the ouput voltage?
It would be best to show the circuit - otherwise, it is problematic to comment on a circuit that can`t be inspected.

 

[kenambo]

i think, the op-amp's output starts oscillating and ends with decreasing oscillation and finally becomes DC...

Am i correct..? fabien..

 

[Fabien]

ithe op-amp's output starts oscillating and ends with decreasing oscillation and finally becomes DC...

It should act like this, that's what I wanted to see but not able to.

What means "...and nothing happen"?

It means nothing is oscillating at the output.

Anyway, guess it's OK, I re run a simulation with the inductor from input to output, and I got the oscillations. Here are the new snapshots.
Now, I think to define the phase margin, I should know what capacitors should be connected in the close loop, right? An then I can define 45°? Am I right?

But, nevertheless, the OpAmp should be unstable, so I expected to see infinite oscillation, not becoming DC!

 

[LvW]

Anyway, guess it's OK, I re run a simulation with the inductor from input to output, and I got the oscillations. Here are the new snapshots.

No - that´s not correct. You have performed a TRAN analysis using the inductors? WRONG!

The inductors are used ONLY for opening the loop ac-wise for the purpose of a loop gain simulation (AC analysis).
More than that, since the ac testsignal to be injected via a large capacitor must be the ONLY signal in the circuit, the "normal" input must be grounded during this simulation.