The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Discrepancy between oscillation frequency with PSS and transient in Spectre
https://designers-guide.org/forum/YaBB.pl?num=1461681706

Message started by megaknaller on Apr 26^th, 2016, 7:41am

Title: Discrepancy between oscillation frequency with PSS and transient in Spectre
Post by megaknaller on Apr 26^th, 2016, 7:41am

Hi everyone,

I am designing a mm-wave Colpitts VCO and am running into big discrepancies (around 7-8 GHz) between the calculated PSS Harmonic Frequency and the calculated frequency from the transient simulations. I use harmonic balance for performing my PSS simulations with an harmonic number of 10. The version of Cadence I am using is the 6.16 and MMSIM is 14.10.

My oscillator seems to start after approx. 4 ns and I therefore set the tstab value in the PSS to 10 or 15 ns when the oscillation has already stabilized. I also set the step and maxstep from both the PSS and the transient simulation value typically to 1 ps in order to capture the oscillation behaviour. If i don't do this, Spectre can't typically resolve out the oscillation and the transient simulation shows a damped out response.

I´ve saved the initial transients (tstab) from the PSS simulations as well. Looking at them, the calculated frequency/wave seems to agree very well with my transient simulation, but if I look at the time wave from the PSS results, it does not match the obtained graph from the initial transient/transient simulations. The difference in frequency is around 7-8 GHz being the PSS calculated frequency the higher one.

Until now I´ve just tried to reduce the maxstep in the PSS options as well as use the sigglobal accuracy option to check if that improves the results agreement, but no succes by now.

I would like to ask if anyone has run into such problems and how could I solve it or at least improve the matching between my results?

Many thanks in advance.

BR :)

Title: Re: Discrepancy between oscillation frequency with PSS and transient in Spectre
Post by iVenky on Apr 27^th, 2016, 11:34pm

I have faced this issue before. Have you checked that the waveform got stabilised on PSS? I mean is it settled or increasing?

Title: Re: Discrepancy between oscillation frequency with PSS and transient in Spectre
Post by alialialialiali on Apr 28^th, 2016, 3:44am

I have met the problem with ADS software and I set steady state time more than default. then it was better.
simulation of Colpitts oscillator has this problem.

Title: Re: Discrepancy between oscillation frequency with PSS and transient in Spectre
Post by cheap_salary on Apr 28^th, 2016, 5:00am

megaknaller wrote on Apr 26^th, 2016, 7:41am:

I am designing a mm-wave Colpitts VCO
.................................................................
I use harmonic balance for performing my PSS simulations with an harmonic number of 10.

Oscillation frequency of your VCO is very high.
So it need very long time to reach steady state.

Can you understand algorithm of HB Analysis ?
Result of HB-PSS is proper in your case.

Use Shooting-PSS instead of HB-PSS and then confirm whether it can converge within reasonable iterations.

Title: Re: Discrepancy between oscillation frequency with PSS and transient in Spectre
Post by megaknaller on May 5^th, 2016, 7:12am

I've run a transient simulation to check for stabilization, the oscillator amplitude stabilizes aroudn 3 ns, therefore I set a tstab in PSS of 5 ns, but I've already tried with values as high as 50 ns. I've found this issue not only in Colpitts but also Clapp oscillators, which I typically work with. This might indicate for me it is not an architecture related issue.

The transient saved by PSS agrees sometimes quite well with my transient waveform, but the PSS time wave does not agree with both of them at the end. I understand more or less how the HB works, but this does not explain why I get the discrepancy at the end.

Any feedback is really appreciated.

Thank you! :)

Title: Re: Discrepancy between oscillation frequency with PSS and transient in Spectre
Post by cheap_salary on May 5^th, 2016, 9:30am

megaknaller wrote on May 5^th, 2016, 7:12am:

I understand more or less how the HB works,
but this does not explain why I get the discrepancy at the end.

Show me netlist or schematic.
Does your oscillator include distributed-lines or s-parameter box ?

Have you confirmed Shooting-PSS ?