[SOLVED] Monte Carlo - Cadence Virtuoso IC 6.1.5

[kenambo]

Hi,
What is the need for the "monte carlo seeds" in virtuoso IC 6.1.5?

thanks

 

[erikl]

Even if MC simulation selects "random" start values, by using a seed value there is reproducibility. See this general discussion on MC simulation:
View attachment Cadence_Monte_Carlo_simulation_tutorial.pdf

 

[kenambo]

Hi erikl,
i just go through this document you provided..

so i come to a point.. that seeds are used to reduce the iteration and find the same variation as if we assign more iterations.. Is it right?

And on which basis the seeds have to be chosen?

thanks

 

[erikl]

... seeds are used to reduce the iteration and find the same variation as if we assign more iterations.. Is it right?

No, that's not the point. MC simulation doesn't use true randomness, but so-called pseudo-randomness, which allows for reproducibility, as mentioned above. So if you started an MC simulation with seed=1 and you want to repeat e.g. simulation results from #10 run, you can restart the same MC sims with a seed=10, and its 1^st simulation will reproduce the same result as the previous 10^th simulation, because it will use the same set of "random" variations.

If you think your results' spread isn't wide - or significant - enough (because you didn't use enough iterations), by this same method you can continue and extend the number of iterations without the need of repeating the previous ones, i.e. the previous deviation sets won't be repeated. In the example above you'd then start with seed=11 .

So using a larger seed number doesn't help for better significance - just for reproducibility resp. continuation of a previous simulation.