The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Monte Carlo with Temperature using Spectre
https://designers-guide.org/forum/YaBB.pl?num=1138081612

Message started by albchen on Jan 23^rd, 2006, 9:56pm

Title: Monte Carlo with Temperature using Spectre
Post by albchen on Jan 23^rd, 2006, 9:56pm

Hello everyone! Two questions:

1) I am using the *options* statement to control temperature with a parameter. I have the parameter specified in my statistics block that I use in my Monte Carlo. Example:

parameters d_temp=50
setup options temp= d_temp
statistics {
[tab]process {
[tab][tab]vary d_temp dist=unif N=100 percent=yes
[tab]}
}

mc1 montecarlo variations=process {
[tab]tran1 tran start=0 stop=10n
}

However, it seems that the Monte Carlo simulation doesn't do what I expected. In the params file, it has a title line saying "Temperature = 50C". I also tried the following:

setup options temp= 50
statistics {
[tab]process {
[tab][tab]vary temp dist=unif N=100 percent=yes
[tab]}
}

This doesn't work either. Anyone know how I can do a Monte Carlo simulation with temperature as one of the parameters?

2) How do I view the transient files that is produced from the child analysis? The transient files ARE being produced, but no wave viewer has worked. I tried Cscope and the waveform viewer from Cadence. What gives? Is there something I need to set?

Please help and I appreciate your time. Thanks!

Title: Re: Monte Carlo with Temperature using Spectre
Post by achim.graupner on Jan 25^th, 2006, 6:13am

Hi albchen,

why do you want the temperatur to be a random parameter?
I expect you want to do a simulation including both a process variation as well variation in operation conditions. Doing so by including temperature (and maybe supply volteage etc) as random variable will not yield the result you expect. You have to check whether you circuit provides sufficient yield under *all* operating conditions.
Take the example: you circuit provides 100% parametric yield at 0C and 50% parametric yield at 100C. Your simulation with temperature equalliy random distributed between 0C and 100C will estimate a yield of about 75%. But actual you yield is just 50% as your circuit *has* to work at 100C.
So operating conditions (as temperature) have to be included in MC simulations as sweep-parameter.

Have you got the problem or shall I explain the topic more thoroughly?

Regads,
achim

Title: Re: Monte Carlo with Temperature using Spectre
Post by albchen on Jan 25^th, 2006, 2:05pm

Sorry if there was some confusion. This is *NOT* a yield analysis. For that, I completely understand that you need to test the corner cases.

The reason I am using a MonteCarlo is for statistical sampling for the behavior of a circuit, like the gain of an op-amp, without having to resort to a full blown sweep on all PVT parameters. This is intended to be a quick snapshot of the behavior.

So, is it possible? And how can I view my transient files?

Thanks.

Title: Re: Monte Carlo with Temperature using Spectre
Post by Geoffrey_Coram on Feb 10^th, 2006, 6:15am

It still seems you'd be better off distributing your temperatures uniformly over the interval (eg every 10 degrees).
It would be very odd if your circuit didn't work at every multiple of 10 degrees but did work on the 1's and 9's.
If you did statistical, you might get "unlucky" and sample 10 points very close to 27C.