Designer’s Guide Community

Cri Azzolini

Community Member

Offline

Posts: 48
University of Parma, Italy

Strange problem in AC/PAC analysis with Spectre
Dec 2^nd, 2005, 1:09am

Hi all!

I have got a problem during my AC analysis of op-amps.
When I try to perform the analysis of the open-loop gain of the amplifier, the resulting Bode diagrams show very strange shape. In particular, both the magnitude and the phase of the open-loop gain show the typical low-pass shape but at medium-to-high frequency appear saw-toothed curves on the graphs. This saw-tooth profile is very sharp since it may cause the magnitude to change (randomly) of 10-20 dB in just few hertz. For this reason I do not think that is the real behaviour of the op-amp.
These sharp sawtooth profiles lay upon the familiar low-pass open-loop curve but are very tiresome since, usually, they start around the unity-gain frequency where the phase-margin should be calculated.
I tried to change the number of frequency points for the simulation but the result is the same.
Probably it depends on the system I used for breaking the loop: indeed using ideal switches (AC position = open) the problem never appears. On the contrary, using AHDL or VerilogA switches, the saw-tooth usually appears.

The same problem results during PAC analysis of switched-capacitor circuits. Unfortunately the ideal switches in "analogLib" library do not work for PAC analysis so I have to use AHDL/VerilogA switches for PAC analysis.

Should I change some simulator options?
Could anybody of you help me?
Thanks in advance!

Back to top

« Last Edit: Dec 02^nd, 2005, 5:25am by Cri Azzolini »

IP Logged

Ken Kundert Global Moderator Offline Posts: 2386 Silicon Valley	Re: Strange problem in AC/PAC analysis with Spectr Reply #1 - Dec 3^rd, 2005, 8:14am In general breaking the loop is a bad idea. You can find alternative approaches in "The Designer's Guide to Spice and Spectre", by going to Prof. Middlebrook's website (ardem.com), or http://www.thekunderts.net/ken/docs/c%26d2001-01.pdf. -Ken
Back to top	IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: Strange problem in AC/PAC analysis with Spectr
Reply #2 - Dec 3^rd, 2005, 1:08pm

There is also the "stb" analysis in spectre (which is a means
of analysing stability with the loop kept closed). There's also
a "pstb" analysis which is currently present as an engineering option in spectre for doing the same kind of thing with periodic circuits.

Ah, I see the paper that Ken was referencing was for the "stb" analysis in spectre, or at least the background to it.

Andrew.

Back to top

IP Logged

Cri Azzolini

Community Member

Offline

Posts: 48
University of Parma, Italy

Re: Strange problem in AC/PAC analysis with Spectr
Reply #3 - Dec 4^th, 2005, 7:48am

Hi Ken, hi Andrew,

thanks for your replies.
I read Ken's book about Spectre and Spice and I know that the Middlebrook method is the best one for open-loop gain evaluation but I was not sure how to apply to circuits with a periodic steady-state.

I am going to try the "pstab" analysys as suggested by Andrew.

Bye,
Cri

Back to top

IP Logged

Cri Azzolini Community Member Offline Posts: 48 University of Parma, Italy	Re: Strange problem in AC/PAC analysis with Spectr Reply #4 - Dec 9^th, 2005, 1:11am Andrew, in your last reply to my post, you talked about a periodical stability analysis (pstb). Actually I am not able to find it in Spectre5.033: is it an analysis added in the 6.0 version? What do you mean with "engineering option"? Thanks, Cri
Back to top	IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: Strange problem in AC/PAC analysis with Spectr
Reply #5 - Dec 12^th, 2005, 12:26pm

My mistake; I thought it had made it into MMSIM60 USR2 - but currently it's only available to selected customers rather than it being generally released. We tend to do this approach to allow something to be selectively introduced rather than releasing something before it's been properly validated.

Andrew.

Back to top

IP Logged

Frank Wiedmann Community Fellow Offline Posts: 678 Munich, Germany	Re: Strange problem in AC/PAC analysis with Spectr Reply #6 - Dec 15^th, 2005, 1:38pm If in the meantime you would like to set up some kind of pstb analysis by hand, you might want to take a look at the thread starting at http://www.designers-guide.org/Forum/?board=power-electronics;action=display;num... (please note that the thread spans 3 pages).
Back to top	IP Logged

Ken Kundert Global Moderator Offline Posts: 2386 Silicon Valley	Re: Strange problem in AC/PAC analysis with Spectr Reply #7 - Nov 2^nd, 2006, 7:57pm The paper I cited in my earlier post has moved. You can now find it at http://www.kenkundert.com/docs/cd2001-01.pdf. -Ken
Back to top	IP Logged

Andrew Beckett Senior Fellow Offline Life, don't talk to me about Life... Posts: 1742 Bracknell, UK	Re: Strange problem in AC/PAC analysis with Spectr Reply #8 - Dec 12^th, 2006, 2:57pm pstb is in MMSIM61, by the way. Regards, Andrew.
Back to top	IP Logged