Designer’s Guide Community

vivkr

Community Fellow

Offline

Posts: 780

simulting PSTB analysis using PAC
Sep 26^th, 2007, 4:29am

Hi All,

I don't have access to the latest Cadence version with PSTB. So I was thinking of implementing the double-injection
technique using PAC sources in my switched-capacitor network. There were a couple of points which bothered me and
I have not been able to figure these out. Help would be appreciated on these:

1. Normally, one needs to use an explicit S&H when using PAC along with switched-cap filters. Would one need
to do the same here while assessing stability? One would thus determine the return ratio in say Phase 1 with 1 sample-and-hold,
and the return ratio in Phase 2 with another sample-and-hold (ideal ones used). This gives 2 sets of return ratios.

2. What is unclear is whether determining the return ratios in the 2 phases is really enough.

3. Also unclear is whether the phase is determined correctly. Normally, sampling will change the phase
because we are going from s-domain to z-domain. How can I check stability then?

If anyone has successfully realized a PSTB simulation using PAC, then I would be glad to know of the method.

Regards
Vivek

Back to top

IP Logged

Frank Wiedmann Community Fellow Offline Posts: 678 Munich, Germany	Re: simulting PSTB analysis using PAC Reply #1 - Sep 26^th, 2007, 7:10am I don't believe you need any additional S&H circuits. For some more information, you might want to look at http://www.designers-guide.org/Forum/YaBB.pl?num=1171399171.
Back to top	IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: simulting PSTB analysis using PAC
Reply #2 - Sep 27^th, 2007, 1:01am

Hi Frank,

You mention in your other post that one would need to insert the probe in PAC analysis at a node where there is a continuous signal.
I agree on this point. However, what is not clear is what continuous means in this case. I have a switched-capacitor amplifier, and
if I look carefully, I don't believe there is any node within the node which fulfils this criterion.

I inserted my probes at the summing junction as a test but I get a loop transmission factor which is always < 1 in magnitude.
I think the choice of my nodes does not fulfil the required criterion. What else could I try? I will try and see if I can get better
results at the output node, but I am not optimistic about it either.

For completeness, I am using double-injection on 2 identical copies of the same circuit. The method is to inject a PAC=1 A current
in 1 setup and apply a series voltage of PAC=1 V in the 2 different setups. The method is described in the EE214 lectures from Stanford,
by Boris Murmann. This method allows an easier computation of loop transmission, as only 1 voltage and 1 current need to be probed.
Of course, I am using the cmdmprobe to do differential probing.

Regards
Vivek

Back to top

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: simulting PSTB analysis using PAC
Reply #3 - Sep 27^th, 2007, 4:40am

I don't believe that using a point with a continuous signal is mandatory. If you are using a point with switching action, you will probably have to use "Sweeptype relative" in the pac simulation setup.

I also don't think that you can use the cmdmprobe if you are using pac analysis with your own sources instead of pstb analysis. I would suggest that you either use the approach with the ideal baluns I suggested (see page 206 of Murmann's lecture notes) or the setup shown on page 204 of Murmann's lecture notes.

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: simulting PSTB analysis using PAC
Reply #4 - Sep 27^th, 2007, 5:15am

Hi Frank,

Thanks for the feedback. Please ignore the part about the cmdmprobe. I realized after going through the documentation that
this would only have been useful for PSTB or STB. I do use the approach with the baluns. I seem to get Ti correctly, but Tv is too low.

Maybe, it will be better after using this sweeptype option. Let me try it.

Thanks
Vivek

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: simulting PSTB analysis using PAC
Reply #5 - Sep 28^th, 2007, 1:47am

Hi Frank,

No luck so far with using PAC setups to mimic a PSTB analysis. I am using the differential setup with baluns and
measuring Ti and Tv. Both seem too low, especially Tv. I expect them to be in the range of around 100 dB at low-f,
but Ti is around 80 dB and Tv < 0 dB. Not surprisingly, overall T is small.

Now, I plan to try to see if there is some error in my setup. For this, I can build an equivalent model for a continuous-time
circuit and see if my setup there gives the correct value for return ratios when compared to the standard STB analysis (I do have
that one available). If not, then I can work on changing my setup.

Will post here if I can get it to work, but for now, I am relying on transient analysis as always. The only problem is
that one cannot get an accurate assessment of the stability margins, only very approximate measures.

Thanks all the same for your help and suggestions.

Regards
Vivek

Back to top

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: simulting PSTB analysis using PAC
Reply #6 - Sep 28^th, 2007, 3:04am

I'm sorry I could not help you more. I must admit that I have never actually done such a simulation. One final thing you might try is to use ideal S&H circuits to probe your signals like you originally proposed (probably together with a relative frequency sweep). The S&H circuits should only be used as probes and not interfere with the function of the circuit.

Back to top

IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2386
Silicon Valley

Re: simulting PSTB analysis using PAC
Reply #7 - Sep 28^th, 2007, 10:53am

Vivek,
You should be able to duplicate a PSTB analysis with PAC. But I believe the strobing and the issue with the two phases that you brought up are red herrings. Since you are measuring loop gain, the circuit itself will naturally perform the strobing or suppress one of the phases as appropriate.

Have you tried replacing performing the test on an idealized case to work out any bugs before trying it on your circuit.

-Ken

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: simulting PSTB analysis using PAC
Reply #8 - Sep 30^th, 2007, 11:08pm

Hi Ken,

Actually, I did try an idealized circuit example, but didn't get good results there either when comparing STB with AC.

It is a little hard for me to understand why because I used the standard textbook setup with baluns and trying to estimate
Ti and Tv to compute overall T in the differential setup. I have used similar setups for single-ended circuits with success.

Something must be wrong with the way in which I calculate the various Ts, or with my current setup. I will check this carefully.
Meanwhile, I estimated the stability margins from transient analysis (closed-loop of course) and found that there is a large enough
margin.

Thanks for all the feedback.

Regards
Vivek

Back to top

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: simulting PSTB analysis using PAC
Reply #9 - Oct 1^st, 2007, 1:23am

Vivek, you might want to try a single-ended example first. Like this, you can separate possible problems with your differential setup from problems with the pac analysis.

Ken, what might be difficult is to set up the simulation in a way so that the results are easy to interpret. For example, a low-frequency signal at the input might appear close to the fundamental at a switching node. Using an ideal S&H could possibly transfer the signal back to low frequencies (one would have to choose the correct sideband at the sampled node, of course).

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: simulting PSTB analysis using PAC
Reply #10 - Oct 1^st, 2007, 2:31am

Hi Frank,

I have used the single-ended setup in the past with success, with continuous-time circuits. There, I
was able to get good results. I even tried to correlate the results from a single-ended double-injection
analysis (AC) with a STB analysis to check my setup and got a perfect match.

I think that something goes wrong with my differential setup. I use baluns there to convert the differential
signals to their diff. mode and common mode levels, use the double-injection technique at these nodes (2 copies)
and then reconvert back to the differential levels with another balun to complete the loop.

Yet, the results don't match even for a simple continuous-time example. I need to check what is going wrong.

Regards
Vivek

Back to top

IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: simulting PSTB analysis using PAC
Reply #11 - Oct 1^st, 2007, 6:49pm

Hi Vivek,

Except for a wrong differential set-up (as you suggested might be the case) I only can think of a different bias point for the circuit when running each AC and STB in order to explain the different results you are getting from each analysis.
What I'm not sure about is how to check what the calculated bias point is for the STB analysis so that you could compare it with the AC counterpart.

Regards
Tosei

Back to top

Keep it simple

IP Logged