Monte Carlo in HSPICE

[nirankul2003]

When performing Monte Carlo using HSPICE, does the order in which transistor devices are listed in the netlist, matter?
I am working with certain 90nm design kit where changing the ordering of transistor instances is showing different results. But if that's true,
is it a fault/feature of HSPICE or design kit??

Thanks
-Niranjan

 

[keith1200rs]

The Monte Carlo process isn't truly random. It uses a pseudo random number generator based on a seed value. If you don't change the seed you will get identical results from two groups of Monte Carlo runs. I have never experimented with netlist order but it seems reasonable that changing that will change the results. The important thing with any Monte Carlo runs is to ensure there enough runs for the results to be statistically significant, then such effects should not be noticeable.

Keith

 

[saurabhr8here]

Keith is right. Re-run Monte Carlo simulations with different netlist order, with a large number of iterations/samples and check mean/sigma. You should get similar values.