The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> SpectreRF Parametric Sweep convergence problems
https://designers-guide.org/forum/YaBB.pl?num=1073695033

Message started by OregonStateGrad on Jan 9^th, 2004, 4:37pm

Title: SpectreRF Parametric Sweep convergence problems
Post by OregonStateGrad on Jan 9^th, 2004, 4:37pm

When I run a PSS simulation in a parametric sweep, if any of the simulations in the sweep fail to converge, all the following runs fail also. They give the error that the output is too small to reliably detect the period of the oscillator. I can look at the outputs in a transient analysis and see that they both swing about a volt. What gives? ???

BTW, I'm simulating differential LC oscillators with MOS varactors for tuning. This problem only seems to happen when the parameter I'm sweeping changes the operating point of the oscillator significantly. Sweeping the bias current for example changes the output voltage swing which in turn changes the oscillation frequency a little.

The only thing i notice that is wierd is that when the second and subsequent pss analyses run in the sweep, they use the frequency from the previous pss run as the guessed fundamental in the current run. Maybe if one doesnt converge the next one cant find the fundamental to guess or maybe it is also storing initial conditions.....

Has anyone heard of this kind of thing before?

thanks

martin

Title: Re: SpectreRF Parametric Sweep convergence problem
Post by Y Z on Jan 12^th, 2004, 11:40am

Martin,

This was fixed more than half year ago. In that fix, the PSS convergence for oscillators was improved, and the sweep won't fail if one in it does.

The fix was done for version 4.4.6 and all the later release. If you can find the following message after the tstab integration (the first transient integration) "The Estimaged Oscillating Frequency from Tstab Tran is ... (Hz)", then you are using the fixed release. Otherwise, you should update your Spectre.

If you do have those messages and your sweep still fails, please report it to Cadence. We are very interested to fix your problem.

>The only thing i notice that is wierd is that when the
>second and subsequent pss analyses run in the
>sweep, they use the frequency from the previous pss
> run as the guessed fundamental in the current run.

Yes. the previous converged period is used as guess period for the current run. But for each run, Spectre does a tstab integration from 0 to tstab+4*guess_period, and estimates a more accurate period from the tstab integration.

YZ