The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> QPSS QPAC Simulation Failure due to insufficient memory
https://designers-guide.org/forum/YaBB.pl?num=1268204443

Message started by aaron_do on Mar 9^th, 2010, 11:00pm

Title: QPSS QPAC Simulation Failure due to insufficient memory
Post by aaron_do on Mar 9^th, 2010, 11:00pm

Hi all,

I'm trying to simulate the IIP3 of a down-conversion system, but the simulation always fails due to insufficient memory. In the QPSS options, I enabled a swapfile (there's around 13 GB free space in that drive), and I enabled compression, and I only saved the selected nets. Is there anything else I can do to get the simulation to finish. Note that my run directory has about 13 GB of free space, and the computer's memory is 3 GB. The output log file always says the following:

Mem Used: 3056.9 MB (Memory Exhausted)
Warning from spectre during IC analysis, during periodic steady state analysis, during Quasi-Periodic Steady State Analysis `sweepqpss-002_qpss', during Sweep analysis `sweepqpss'.
WARNING (SPECTRE-16001): Sweep iteration for `prf' = -60 terminated prematurely because of the following error(s):
Fatal error found by spectre during IC analysis, during periodic steady state analysis, during Quasi-Periodic Steady State Analysis `sweepqpss-002_qpss', during Sweep analysis `sweepqpss'.
FATAL (CMI-2002): Insufficient memory available, To reduce memory for rf analyses, please refer "spectre -h rfmemory".

I can't seem to find this spectre -h rfmemory either...

thanks,
Aaron

Title: Re: QPSS QPAC Simulation Failure due to insufficient memory
Post by pancho_hideboo on Mar 10^th, 2010, 2:10am

aaron_do wrote on Mar 9^th, 2010, 11:00pm:

I'm trying to simulate the IIP3 of a down-conversion system

Which engine do you use for QPSS, Shooting-Newton or HB ?

Show me "Circuit inventory" informations in log file.

Title: Re: QPSS QPAC Simulation Failure due to insufficient memory
Post by aaron_do on Mar 10^th, 2010, 7:37pm

Hi pancho_hideboo,

I am using the default method which seems to be the Shooting-Newton method.

Here is the circuit inventory. It is a post-layout simulation, so its pretty large. Now that I look at it i'm not sure why there are so many inductors. I only did a C+CC extraction.

Circuit inventory:
nodes 4253
balun 1
bsim3v3 8352
bsource_1 14
capacitor 18667
diode 182
diode_va 32
inductor 359
mutual_inductor 6
port 4
quantity 9
res_va 256
resistor 2203
vsource 9

thanks,
Aaron

Title: Re: QPSS QPAC Simulation Failure due to insufficient memory
Post by pancho_hideboo on Mar 11^th, 2010, 12:52am

aaron_do wrote on Mar 10^th, 2010, 7:37pm:

Here is the circuit inventory.
It is a post-layout simulation, so its pretty large.
Now that I look at it i'm not sure why there are so many inductors.
I only did a C+CC extraction.

Not so large.

I'm using HB2SS(=2 large tones with one small signal HB Analysis) of Agilent ADSsim and GoldenGate for IP3 evaluation of more large circuits.

aaron_do wrote on Mar 10^th, 2010, 7:37pm:

I am using the default method which seems to be the Shooting-Newton method.

Use HB not Shooting-Newton.

Title: Re: QPSS QPAC Simulation Failure due to insufficient memory
Post by aaron_do on Mar 11^th, 2010, 4:51am

Quote:

Use HB not Shooting-Newton.

thanks. I'll give it a try

Aaron

Title: Re: QPSS QPAC Simulation Failure due to insufficient memory
Post by Andrew Beckett on Mar 11^th, 2010, 8:46am

The other common mistake people make is related to specification of the tones in QPSS. This works by using the "Frequency Name" on the sources to identify the frequencies of the tones. What you should do is specify any sources with common frequencies (maybe they have different phases) as having the same frequency name. This means that it will not treat them as independent frequencies in QPSS, which will make a big difference to the speed and simulation time (there's absolutely no benefit in treating them separately).

Perhaps you could post the analysis statements from your input.scs too. That could pinpoint any mistakes which would lead to excessive memory consumption.

You could also switch to HB though, as Panchoo Hideboo suggests. In fact if you use the newer "hb" analysis (the algorithm is the same, but the interface to it is a little different) you can directly enter the frequencies of the tones on the GUI, rather than them being looked up from the sources.

Regards,

Andrew.