Is there any method to change simulator options as dynamic parameters?
For example I want to start simulation with errpreset="conservative" and after some simulation time change it to "liberal".
Another question how can I set simulation time to be stopped after some event?
For example if I run parametric or corner analysis for POR until POR event happened the simulation running very fast because internal oscillator is stopped. But after POR event the oscillator is ON and simulation runs much slower. I need to get only POR delay time, but it's change very wide for different corners, so I have to set the maximum simulation time and wait very long time to finish my corner analysis.
I tried the way with paramset file like Andrew recommended /forums/p/10265/414454.aspx#414454
with time and reltol change only but got error message:
ERROR: (SFE-1703) : "/user/fomin/pset1" 1. Wrong number of nodes. Port instance has 2 or 3 terminals.
ERROR: (SFE-874) : "/user/fomin/pset1" 2. Unexpected end of line. Expected equals sign, numeric value or string value.
In reply to fomin:
Because your filename does not have the ".scs" suffix, it's assuming it is in SPICE rather than spectre syntax. So either change the filename to be called "pset1.scs" or add a line:
before the paramset definition
In reply to Andrew Beckett:
Thank you Andrew!
Now it works.
But I changed 2 parameter "reltol" and "strobeperiod". But spectre changes only "reltol". For "strobeperiod" it gives warning that
"... The parameter 'strobeperiod' defined in 'pset1' paramset is not found inthe netlist and will be ignored."
In the netlist I see that "reltol" is declared as simulator option but 'strobeperiod' is transient analysis parameter.
If there any other way to set different 'strobeperiod' for different time slot? I noticed that for larger 'strobeperiod' simulation is much faster and CPU usage is higher.
Now I have to use combination of 'skipstart/stop/count' and 'strobeperiod' setting to speed up my simulation.
I thought I used combination of 'skipstart/stop/count' and 'strobeperiod' setting to speed up my simulation. And as I remember it worked, but now spectre responded with error message that SkipCount and StrobePeriod cannot be specified simultaneously. Actually I didn't use them simulataneously. I set 'strobedelay' after 'skipstop'
I tried another way.
In paramset I set reltol=0.01, vabstol=1e-5, iabstol=1e-11 for time from 0 to 300mS,
then they2 should be changed to 0.0001, 1e-7 and 1e-12 from 300mS.
First it inform me about option setting to reltol=0.01, vabstol=1e-5 and iabstol=1e-11 and than
"FATAL (CMI-2002). Insufficient memory available. To reduce memory for rf analyses, please refer "aps -h rfmemory""
When I changed these options to default 0.001, 1e-6 and 1e-12 simulation start running without any error message.
I am confused. I thought just opposite: more tough options need more memory
If I remember rightly, reltol is handled specially by the dynamic parameter stuff - it knows about a few global options which can be changed this way. I'm not sure you can change any old parameter throughout the simulation. You might be able to do it by making the strobeperiod set to "mystrobeperiod" and then varying that, but I think that may not work too - give it a try. Otherwise you'll probably need to contact customer support (support.cadence.com) so it can be investigated properly and an enhancement request filed.
You should be able to use skipstart and strobeperiod (this allows control over when the strobing starts). However, I suspect the other combinations may make spectre think that both strobing and skipping are in operation together. Anyway, if this is not behaving properly, I think customer support is the best bet so we can get it to R&D.