The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> SpectreRF's Flexible Balanced can't converge
https://designers-guide.org/forum/YaBB.pl?num=1162921214

Message started by pancho_hideboo on Nov 7^th, 2006, 9:40am

Title: SpectreRF's Flexible Balanced can't converge
Post by pancho_hideboo on Nov 7^th, 2006, 9:40am

Hi.

I run a simulation of log envelope detector circuit(RSSI) where only one tone(40MHz) is involved using SpectreRF's Flexible Balanced Analysis.
Already it took three days, but simulation doesn't finish yet.

On the other hand ADS's HB Ana takes only two hours to finish this simulation.
Of course, results of ADS are very reasonable.

Conditions of simulations are almost same for both SpectreRF and ADS.
For example, Order=8 and Oversamples=1.

How to run SpectreRF's Flexible Balanced Analysis effectively ?

I use following Specte.
@(#)$CDS: spectre version 6.1.0 08/15/2006 15:48 (usimlx111) $

It resulted in fail when I invoked PSS Shooting Newton Ana for this circuit simulation.

Title: Re: SpectreRF's Flexible Balanced can't converge
Post by pancho_hideboo on Nov 23^rd, 2006, 2:46am

Hi.

I've evaluated Flexible Balance Analysis in SpectreRF
since "MMSIM 6.0 USR1" and "MMSIM 6.0 USR2".
These were very inferior than HB-Ana of ADS
and helpful only for very small circuits.
Of course FB-QPSS was far helpful than shooting-QPSS.
So I concluded that FB-Analysis of SpectreRF is very inferior
than HB-Ana of ADS, so we've not adopted it until now.

I've expected remarkable improvement in MMSIM6.1.
But my conclusion was never changed.

:'(

Title: Re: SpectreRF's Flexible Balanced can't converge
Post by Andrew Beckett on Jan 15^th, 2007, 2:26am

This is contrary to pretty much every example I've seen, so I suspect something is either wrong with your setup, or you've been extraordinarily unlucky and hit a bug,

Please contact Cadence customer support with your testcase, and then we can investigate it in more detail. It certainly does not match up with what is usually seen - SpectreRF's HB can simulate very large circuits in my experience.

Best Regards,

Andrew.

Title: Re: SpectreRF's Flexible Balanced can't converge
Post by pancho_hideboo on Jan 18^th, 2007, 7:55am

Cadence have continued to claim that
"Shooting Newton of SpectreRF can simulate very large and high nonlinear circuits than HB analysis".
But ............

My log envelope detector circuit(RSSI) is composed of limiter, rectifier, bias etc.
Circuit size is fairly large. So it takes much time even for DC analysis.
Relative new ADS HB can succeed for such large and high nonlinear circuits.
but I think older ADS(e.g. ADS2004) couldn't succeed.

Of course, Shooting of SpectreRF resulted in fail after long long time
and with extraordinary computer memory consumption.

Other than this, there are many circuits where Flexible Balance Analysis of SpectreRF
can't converge or succeed while Agilent ADS can give reasonable results.

For Simple low IF receiver with one RF input and one direct LO input,
FB-Analysis could give reasonable results quickly.

But with frequency divider before mixer local input, FB-Analysis couldn't give
reasonable results, here iteration of FB reached to upper limits.
So I set "maxperiods" parameter which is available also for FB-Analysis.
Of course, I set "freqdivide" parameter and tried to increase "oversample" parameter.
But FB couldn't converge even though it reached to upper iteration limits.