The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> PSS: insufficient Memory
https://designers-guide.org/forum/YaBB.pl?num=1168863700

Message started by Kudie on Jan 15^th, 2007, 4:21am

Title: PSS: insufficient Memory
Post by Kudie on Jan 15^th, 2007, 4:21am

Hi,

I'm trying to simulate a chopping amplifier with fIn=200Hz and fChop=100k, so I have to include 501 side-bands. When I start a PSS / PNoise with this configuration I get the error message "insufficient memory". When I'm doing the same simulation with fIn=1kHz and 101 side-bands it's working.
Has anybody an idea what I could do to make the simulation running?

Up on now I'm simulating on a Dual Core with 4GB of RAM. There is plenty of swap space available so memory shouldn't be a problem....

Thank you very much
Markus

Title: Re: PSS: insufficient Memory
Post by ACWWong on Jan 15^th, 2007, 4:54am

Spectre (32-bit) has a limit of 4G size (2G on some OS). Whilst running your sim you can "top" to see the spectre executable growing in size until it bombs out at 4G (or 2G) size. So having loads of extra RAM won't help.
Firstly ensure your running the new spectre front-end (MMSIM stream) rather than IC5x, because the new version is much more effiecient in memory handling (it was 10x less memory for me!).
do
which spectre
to see. If you're already on MMSIM then try the 64 bit spectre version on a 64 bit machine...

Tackling from a different view point, try using array of sidebands rather than every sideband upto 501... maybe you can reduce spectre memory size with little impact on simulation accuarcy by only including salient sidebands?

Title: Re: PSS: insufficient Memory
Post by Andrew Beckett on Jan 15^th, 2007, 5:52am

Also, try using the "swapfile" parameter on the pss analysis. This allows the data captured during the shooting interval to be memory mapped to a set of files, which allows more than 4G of memory to be accessed (in effect). There was also a memory problem with higher numbers of harmonics, if my memory is correct - but I recall that being fixed in an MMSIM60 hotfix - so MMSIM61 USR1 should be a good thing to aim for.

If you have enough memory though, you could use the 64 bit version by doing:

Code:

setenv CDS_AUTO_64BIT spectre

Regards,

Andrew.