Hi Guys,

I have a testbench for stability of a regulator, which is shown in the following picture. The loop response is weird if the loop is broken in the feedback path. The loop phase started from 0 degree, increases 90 degree below 1 kHz. I get correct loop response if the loop is broken at the gate of MOS transistor. I know I have not broken all loops in the testbench if the loop is broken in the feedback path. However, I want to know if anybody can explain the weird behavior.

Thanks
Yawei

Hard to say without see the loop response or the schematic of your buffer. One guess is that you have two loops, that can give this 0 phase at low freqs and the increasing before it rolls of.

I do not understand the capacitor either. In fact, I do not understand the discussion about breaking the loop at different locations. The stb will give you the same respons independet where the probe is inserted, as long as it is in the feedback loop.

I agree that there is no second loop in the drawn schematic. But since the amplifier is jut one block it is hard to know how it looks like inside. What I mean with two loops is that stb can give you this respons in there is two nested loops in the circuit. I had that problem in a CMFB loop once. The phase started at 0 degrees and then increased before rolling of.

Hi Raj and rfidea,

I am sorry that I have not provide enough details like loop response. I did stb analysis. And the following are loop response. The first is wrong loop response when I insert iprobe in the feedback path.

Yawei

Strange indeed! What happens if you remove the capacitor and replace the inductor with a short. At DC those changes should not change your circuit, loop gain and phase of 0 or 180 degrees should be the same.

Do you even then get "wrong" response with the probe in the feedback and "correct" response with the probe in series with the gate?

It did not mean that you should remove the capacitor, or the inductor, from the design. It was just a test to see if and how the loop response at DC changes in such case.

Ok, well done. Two loops can give you strange phenomena. Even your "wrong" response is true in some means. There is usually no danger to have 0 degree phase at low frequency as long as it increases to +90 before it rools off. But it is always good to know why it behaves as it does.

It was probably a simplification from my side. If the open loop gain and phase behaves strange, which it was in this case and also in the one you suggest, I think one should check the stability with the complete Nyquist criteria. In that you should count the number of encircles of -1 but also the number of poles in the right half plane of the open loop transfer function.

If you do this for the "wrong" respose in this example you see that the transition from 0 degree to +90 degree is a pole in the RHPL for the open loop. But you will also find an encircle of -1. The signs are opposite so the number of poles in the RHPL for the closed loop is zero and the closed loop is then stable.