Hi. I notice the wavefrom is not continous around the instant when strobeperiod is changed, e.g., a negative drop happens. Then the wavefrom slowly settles back.
I can ensure no circuit action during this instant.
Such slow settlement cause some issue but apparently it doesn't come from circuit itself.
Is there any way to avoid such issue? Or is there any specific explaination?
One simple case is strobeperiod change will not create such a jump on ideal waveform, e.g. the one from ideal vdc.
My spectre version is 18.104.22.1684 isr16.
How are you changing strobeperiod? Can you share the input.scs if this is a simple voltage source case?
This is a complex system simulation with analog (transistor) + digital (verilogA). In fact I change strobeperiod by dynamic parameter.
Another interesting fact: the issue is with SpectreX + MX, but the problem disappear with
A. SpectreX + AX
B. APS + Moderate
However, SpectreX + AX is still too slow dependent on our budget. Is there some any way to improve SpectreX + MX to avoid such a issue?
As far as I know, strobeperiod isn't supported as a dynamic parameter, so I don't know how you're changing it. I tried it, and it had no effect. There's even an enhancement request asking for it to be supported which has not been closed.
Perhaps you're using skipstart/strobestart to turn on strobing at a particular time (and maybe skipstop/strobestop to turn it off)? If so, are you also using dynamic parameters to change something else? It's quite possible that dynamic parameters can cause a discontinuity in the output (for example, changing temperature).
I think I need more detail on your analysis setup.
To answer your questions
1. yes, I use skipstart and skipstop to define a segment window to adjust the strobeperiod
2. I always use paramter set which is defined in a .scs file which is loaded in the corner setup, then inform paramset = xxx in the tran sim option/Misc page.
3. In fact, it works ok so far. But I find the discontinuity dependent on the (err)preset, whatever in SpectreX or APS.
The point is if I use finer preset, the sim time is too long, but looser preset creates the discontinuity issue.
It's absolutely possible that changing the errpreset/steppreset could result in a discontinuity (although I'd expect it to be small in most cases) because you've just altered the convergence criteria and so the tolerated error will be more or less than it was previously, the integration method could change, and so on. Of course, it makes no sense to be taking measurements across the region where you're changing accuracy controls (in fact changing anything in dynamic parameters, because that too could cause a discontinuity).