The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Maximum time limit in WaveScan and AWD
https://designers-guide.org/forum/YaBB.pl?num=1238006736

Message started by R. Zakai on Mar 25^th, 2009, 11:45am

Title: Maximum time limit in WaveScan and AWD
Post by R. Zakai on Mar 25^th, 2009, 11:45am

It seems like the waveform viewers, AWD and WaveScan both, in the ADE environment have some sort of maximum time limit. They do not seem to plot results beyond ~1.8seconds even if the simulation was successful in going beyond that and saved the results. Does anyone know how to fix this?

Title: Re: Maximum time limit in WaveScan and AWD
Post by Andrew Beckett on Mar 28^th, 2009, 8:31am

I'm guessing that this might be due to the combination of tool versions you're using. There's no maximum time in the waveform tool, but there are some limits to file size.

In earlier spectre versions (prior to MMSIM61) transient simulation data was broken up into 2Gbyte chunks, and then this could be read in IC5141.

In newer (MMSIM61 and later) spectre supports large results files - more than 2G. These can be read directly in IC61, but not in IC5141.

If using IC5141 with MMSIM61 or later, you need to set some environment variables to tell spectre (or Ultrasim) to use the older "chunk" mode:

For Spectre writing PSF data:

setenv PSF_WRITE_CHUNK_MODE_ON true

For UltraSim writing PSF data:

setenv PSF_LARGE_FILE_ON false

But I am guessing to some extent, because I don't know what tool versions you're using, or which output format (SST2 or PSF), for example.

Regards,

Andrew.

Title: Re: Maximum time limit in WaveScan and AWD
Post by rf-design on Mar 29^th, 2009, 6:07am

The times stamps come in as 64bit-IEEE-754 which has 52-bit mantissa. I assume that this is a programmers decision if for instance 1fs=1e-15sec is used as minimum time step.

I remember to have seen in PSF data that the minimum time step is also stored. But I am not sure. Then the normalization is a wise decision from the programmer.

log(1.8/1e-15)/log(2)=50.6769, very near the 52 mantissa

Try to change either a default number for the waveform tool or detect if some minimum time step is transfered from simulator to waveform.

Title: Re: Maximum time limit in WaveScan and AWD
Post by Andrew Beckett on Mar 30^th, 2009, 8:52am

Actually with PSF, the data is floating point (double precision by default), so you won't have this problem - all that happens is that potentially you lose time resolution on very long simulations with very short timesteps - but you'll hit the same problem in the simulator itself, in fact - so it's quite unusual.

I suspect you're talking about SST2 format - in which case I think (I may be wrong) the time is actually integer, and based around a minimum timestep. There have been some issues with that, but I think these have been resolved. Still this is again guesswork, as I don't know what version or what format the original poster is using.

Regards,

Andrew.

Title: Re: Maximum time limit in WaveScan and AWD
Post by R. Zakai on Apr 1^st, 2009, 10:22am

Thanks Andrew and rf_design for your comments. I am using ICFB 5.10.41 environment and spectre version is MMSIM62. I am 100% sure this is not a file size issue. tran.tran.trn file in the psf directory is only 4k bytes. I have a very simple test case. Take a vpulse, set it up for 2.5kHz operation with 1V peak. Do a long transient analysis, say 5 seconds and see the voltage output of the source either in AWD or WaveScan. The output in my case only goes out to 1.8447s.

Rehan

Title: Re: Maximum time limit in WaveScan and AWD
Post by Andrew Beckett on Apr 4^th, 2009, 2:27pm

I don't have the opportunity to test your example right now, but as a workaround, try doing this before starting ADE:

Code:

envSetVal("spectre.envOpts" "simOutputFormat" 'string "psfbin")

in the CIW, and then start ADE and see if the problem is resolved. I think this might be a problem of finite precision of SST2 (that's what you were using, given the filename you mentioned) - but I'd like to check to be sure.

Can you let me know the subversion of IC5141 you're using (Help->About... in the CIW will tell you this)