The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> analysis with two Strongly Large Signal Tones
https://designers-guide.org/forum/YaBB.pl?num=1402554609

Message started by saeedp on Jun 11^th, 2014, 11:30pm

Title: analysis with two Strongly Large Signal Tones
Post by saeedp on Jun 11^th, 2014, 11:30pm

Hello,

Is that possible to run QPSS/QPAC/QPnoise with two strongly large signal tones (clocks) in the circuit at two different frequencies (for example 1G and 1.1G)?

say for example a circuit with two switched capacitor filters each running from two different clocks.

Thanks,
Saeed

Title: Re: analysis with two Strongly Large Signal Tones
Post by RFICDUDE on Jun 13^th, 2014, 8:19pm

Yes in principle it should be possible, but the two frequencies must be divisible by a common frequency (i.e. 100MHz in your example). The fundamental associated for the QPSS large signal is 100MHz. You need to include enough harmonics of the large signal fundamental to capture the harmonics of the high frequency signals, so this may create computer memory issues as the number of harmonics increase.

Title: Re: analysis with two Strongly Large Signal Tones
Post by saeedp on Jun 15^th, 2014, 6:26pm

Thank you.

but if we are going to take 100MHz as the fundamental, why not run PSS instead with fund=100MHz and "number of harmonics" of 55 to cover both 1G and 1.1G to their 5th harmonic.

Title: Re: analysis with two Strongly Large Signal Tones
Post by RFICDUDE on Jun 16^th, 2014, 4:44am

Yes, in this case there wouldn't be much of any difference between PSS and QPSS. I did not understand that you were asking if QPSS will work if one of the moderate tones is actually a large switching signal.

If this is your question then the answer is most likely no because the QPSS is designed such that there is only one fundamental frequency (with harmonics) that can be considered a large signal where the answer has many harmonics (like a square wave).

If you need to find a steady-state solution with two non-harmonic switching signals then you're stuck with PSS. However there are multiple approaches you can try if it is applicable to your specific problem.

PSS with shooting method will work, but it takes a long time to solve depending on how low the beat frequency is between the two switching frequencies.

PSS with harmonic balance (flexbalance) may give a quicker answer, but you have to include many harmonics of the two fundamentals to accurately represent the square wave switching signals. Harmonic balance is also tricky to debug and verify that you've got a good solution. I usually compare with transient simulation first to make sure the answer is reasonable.

Transient noise analysis can be useful for looking at noise response of switched circuits, but it also takes some time to get it set up properly and make sure you know how to analyze the result compared to your expectation.

Title: Re: analysis with two Strongly Large Signal Tones
Post by saeedp on Jun 19^th, 2014, 11:08pm

Thanks RFICDUDE,

I took the first step with your help and now I think the PSS+PAC sim shows correct results.

However, I see some discrepancy in Phase noise sim.

Here is the issue:
assume I have only one 1GHz large signal/clock (I removed the other one to simplify the issue) in the circuit and I would like to measure the phase noise on that.

two methods that I tried and observed different results are:

1) PSS, beat freq=1G, output harmonics=3
Pnoise, sweeptype=relative, relative harmonic=1

2) PSS, beat freq=0.1G, output harmonics=30
Pnoise, sweeptype=relative, relative harmonic=10

the phase noise at 100MHz offset is low for method #2 around-171dBc
but I think method #1 shows correct phase noise.

what do I miss in method #2 with 10X lower beat frequency that results in improper phase noise?

Thanks,
Saeed

Title: Re: analysis with two Strongly Large Signal Tones
Post by RFICDUDE on Jun 21^st, 2014, 8:44am

They should give the same answer.

One possible problem is that limiting the analysis to the third harmonic of the clock may cause some error in the calculated noise.

If determine if this is an issue or not, rerun the 1GHz fundamental frequency simulation and increase the number of harmonics to 5, 7, 9 etc. until the phase noise result doesn't change. If the answers are no different than the harmonics=3 then this is not an issue.

If it is an issue, repeat your experiment including harmonics where the result is not changing with higher harmonics.

Another issue might be the maxacfreq parameter, but I'm not 100% clear on how this parameter works for the 0.1GHz fundamental with harmonics=30 case.