Designer’s Guide Community

Evgenii

Community Member

Offline

Posts: 34

Different step size for different analisys time in ADE spectre transient
Dec 03^rd, 2010, 5:22am

Hi Folks!

How to set different step size in spectre. Say, i need simulate freq tuning of quartz model with precision time step (maxstep 1ps for measurements) after his settling time (10ms). Settling time may be computed with too big or variable step size...
How to realise is? It is possible?

Thanks!

Back to top

IP Logged

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Re: Different step size for different analisys time in ADE spectre transient
Reply #1 - Dec 5^th, 2010, 6:20am

I don't think this is feasible.

Run spectre -help tran, I find the following.
Quote:

Dynamic Parameters during Transient Analysis
The parameters defined in the `Dynamic parameters' section allow users to change temperature,design
parameters or some option parameters (reltol, residualtol, vabstol, iabstol, isnoisy) during transient
simulation.

Obviously, maxstep is not a parameter that can be changed during a tran analysis in spectre simulator until now. The spectre sub-version is 10.1.0.213.isr1.

Back to top

IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2386
Silicon Valley

Re: Different step size for different analisys time in ADE spectre transient
Reply #2 - Dec 5^th, 2010, 11:17pm

You can easily use verilog-a's boundstep to do this, but you probably shouldn't. Manually controlling the time-steps is one of the most inefficient ways of controlling accuracy. If you don't think you answer is accurate enough, have you tried just tightening reltol? Or perhaps you can set errpreset to conservative. Those are generally the best first steps.

-Ken

Back to top

IP Logged

Evgenii Community Member Offline Posts: 34	Re: Different step size for different analisys time in ADE spectre transient Reply #3 - Dec 6^th, 2010, 2:19am Thanks!:)
Back to top	IP Logged

Evgenii Community Member Offline Posts: 34	Re: Different step size for different analisys time in ADE spectre transient Reply #4 - Dec 6^th, 2010, 9:19am Unfortunaly I don't �find section `Dynamic parameters' �in spectre ver. 5.10.41.020608 i tried next: p paramset { �time reltol �0 �1e-3 �5m 1e-6 } error.... 'time' is built-in parameter....
Back to top	« Last Edit: Dec 7^th, 2010, 2:53am by Evgenii » IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: Different step size for different analisys time in ADE spectre transient
Reply #5 - Jan 2^nd, 2011, 2:28am

Why are you using such an old version of spectre? spectre in the IC5141 has not been updated for some considerable time. At the end of 2004 (soon after IC5141 came out), spectre was split off into a separate stream, MMSIM. Since then there have been MMSIM60, MMSIM61, MMSIM62, MMSIM70, MMSIM71, MMSIM72 and MMSIM101. That's 7 major releases, so you are using a really old version of spectre, lacking in a lot of improvements and enhancements. There's no reason why you can't use a recent MMSIM version with IC5141 - that's really the recommendation.

That said, I don't think dynamic parameters is the solution in this case, as Ken said - controlling the accuracy of the simulator would be a better option. I just wanted to ensure you're actually taking advantage of 6 more years of simulator evolution and improvement!

Regards,

Andrew.

Back to top

IP Logged