question on montecarlo correlate

Hello *,

I'm currently re-implementing an inhouse tool which allows e.g. to
produce "custom corners" by manually varying process parameters...
The former tool handled only correlations with cc=1, by a hard
coupling of the parameter sliders in the GUI. Hmm, that's not
exactly how MonteCarlo works, is it?

Now I just want to understand, how (Cadence) MonteCarlo exactly handles
parameter correlations and how exactly the process parameters (and an
effective sigma of the current setting) is computed.

Assume we have following statements:

statistics {
process {
vary A dist=gauss std=0.1
vary B dist=gauss std=0.1
}

correlate param=[A B] cc=0.5
}

Without correlations:

A = Amean + gauss(sigma_a,std_a)
B = Bmean + gauss(sigma_b,std_b)

sigma_eff^2 = sigma_a^2 + sigma_b^2

sigma_a, sigma_b given by GUI (or Montecarlo random number... e.g. [-3...+3])

What is the formula with correlations?

Are there good links, books? ...but I'm not a mathematician

Thanks for any help,
Sylvio

 

Hi Sylvio,

look at Sourcelink for solution number 1841461.

Kind Regards,
Marc