The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Stability analysis for two stage CMOS Op Amp in Spectre
https://designers-guide.org/forum/YaBB.pl?num=1244840188

Message started by Mohammed Omar on Jun 12^th, 2009, 1:56pm

Title: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Mohammed Omar on Jun 12^th, 2009, 1:56pm

Hi All,

When designing a two stage CMOS Op Amp using Spectre, where is the best point to put the IPROBE when doing STB analysis? At the input of the 2nd stage or at the input of the first stage?

Regards,

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Frank Wiedmann on Jun 15^th, 2009, 1:51am

Do you see a difference in the results? Normally you shouldn't. If you do, please post your schematic.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Mohammed Omar on Jun 15^th, 2009, 5:45am

Yes, I notice a difference in both the phase and gain plots. What is most noticeable is that cutting at the input of the second stage I find the phase plot to start at (-180 degrees) and starts going "UP" to zero degrees.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Frank Wiedmann on Jun 15^th, 2009, 6:50am

Your circuit obviously has two loops, one from the output to the input of the first stage, and another one from the output to the input of the second stage. You must put the probe into a wire that breaks both loops when opened, so the position in your schematic is correct. For details, see the section "Applicability of Loop-Based and Device-Based Algorithms" of http://www.kenkundert.com/docs/cd2001-01.pdf, this is the article that describes the algorithm used by the stb analysis.

For a circuit with negative feedback, the phase of the loop gain is supposed to be -180 degrees at low frequencies in stb analysis. For details, see http://www.designers-guide.org/Forum/YaBB.pl?num=1124688329.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Mohammed Omar on Jun 15^th, 2009, 7:39am

Thanks a lot for your fast reply Frank.

The problem is not with the phase starting at (-180) or 180, but with the resulting phase plot; The STB analysis is not able to calculate the phase margin. Also the resulting phase plot shows the system to be unstable while transient simulations show that is stable with adequate phase margin.

Attached is the Bode plot for the behavioral model.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Frank Wiedmann on Jun 15^th, 2009, 10:57am

I don't think that the plot shows the system to be unstable. If you do a Nyquist plot (see http://en.wikipedia.org/wiki/Nyquist_stability_criterion) of the loop gain (the negative loop gain as calculated by the stb analysis), you will see that it encircles the point (-1 + j0) counterclockwise and not clockwise, so I don't think that there are any closed loop poles in the right half of the s-plane.

Always use the Nyquist stability criterion in case of doubt. You can find another example with slightly counterintuitive results at http://www.designers-guide.org/Forum/YaBB.pl?num=1182388268.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Mohammed Omar on Jun 15^th, 2009, 12:24pm

What makes me say its unstable is that in the Bode plot the gain at the phase cross-over point is about 30 dB. Isn't this enough to judge unstability? Also is this phase response normal or possible to be judged?

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Frank Wiedmann on Jun 15^th, 2009, 11:33pm

No, it is not enough to judge instability. Use the Nyquist stability criterion in such unusual cases. The unusual behavior might be due to the fact that you have two loops or due to the component values in your models.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by pancho_hideboo on Jun 16^th, 2009, 9:29am

Mohammed Omar wrote on Jun 15^th, 2009, 12:24pm:

What makes me say its unstable is that in the Bode plot the gain at the phase cross-over point is about 30 dB.
Isn't this enough to judge unstability? Also is this phase response normal or possible to be judged?

See the following although your case is absolutely stable.
http://www.designers-guide.org/Forum/YaBB.pl?num=1190272820

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by RobertZ on Jun 17^th, 2009, 11:45pm

Frank Wiedmann wrote on Jun 15^th, 2009, 10:57am:

Hi, wiedmann,

by saying the counter-clockwise encirclement of -1, do you mean the nyquist plot begins from real x-axis, and goes through the 1st, 2nd, 3rd quadrant and then torwards orgin? it never goes into the 4th quadrant?

Thanks,
Qubo

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Frank Wiedmann on Jun 18^th, 2009, 1:34am

If you carefully look at http://en.wikipedia.org/wiki/Nyquist_stability_criterion, you can see that the Nyquist plot starts at the frequency -∞, not 0, and is always symmetric to the real axis.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Mohammed Omar on Jun 18^th, 2009, 4:30am

Thanks Frank,

I'm reading some literature about Nyquist plots and stability right now. Thanks a lot for your help and comments, it was quite useful.

One more question; In case I'm given a phase margin specifications, this means I must reside to the transient response? Or how to obtain the phase/gain margin from Nyquist plots?

Regards,
Mohammed Omar

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Frank Wiedmann on Jun 18^th, 2009, 5:53am

Please see my remarks at http://www.designers-guide.org/Forum/YaBB.pl?num=1182388268/4#4.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by RobertZ on Jun 18^th, 2009, 11:30pm

Frank Wiedmann wrote on Jun 18^th, 2009, 1:34am:

If you carefully look at http://en.wikipedia.org/wiki/Nyquist_stability_criterion, you can see that the Nyquist plot starts at the frequency -∞, not 0, and is always symmetric to the real axis.

Hi, Frank,

You are right about the symmetry. After I sketched the Nyuist plot, I realized that there are two counter-clockwise encirclement of -1. To make Z = 0, p must be =2, which means the open loop has 2 poles in RHP. Is this right?

Thanks,
Robert

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Frank Wiedmann on Jun 18^th, 2009, 11:48pm

Yes, I think this is right.

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by RobertZ on Jun 19^th, 2009, 12:03am

Frank Wiedmann wrote on Jun 18^th, 2009, 11:48pm:

Yes, I think this is right.

but how can you tell there are two RHP poles in open loop gain from his bode plot?

Thanks,
Robert

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by RobertZ on Jun 19^th, 2009, 12:07am

Frank, btw, can you take a look at my another post:
http://www.designers-guide.org/Forum/YaBB.pl?num=1245305413

maybe you have some clue.

Thanks,
Robert

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by Frank Wiedmann on Jun 19^th, 2009, 4:37am

RobertZ wrote on Jun 19^th, 2009, 12:03am:

but how can you tell there are two RHP poles in open loop gain from his bode plot?

Thanks,
Robert

Assume that you have an open loop gain function T(s)=A/[(s-p1)*(s-p2)] with two RHP poles at p1=2π*10³ 1/s and p2=2π*10⁶ 1/s. Do a Bode plot of this function for s=j2πf. What do you get?

Title: Re: Stability analysis for two stage CMOS Op Amp in Spectre
Post by RobertZ on Jun 19^th, 2009, 9:55am

you are right. there are 2 poles in RHP, other thatn that, there is a LHP zero, which is close to the 2nd pole, and a RHP zero after unitiy gain bandwidth.