The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Montecarlo Simulations for On-chip Process variations for 90nm Technology
https://designers-guide.org/forum/YaBB.pl?num=1240680685

Message started by cmos.analogvala on Apr 25^th, 2009, 10:31am

Title: Montecarlo Simulations for On-chip Process variations for 90nm Technology
Post by cmos.analogvala on Apr 25^th, 2009, 10:31am

I want to analyse a circuit for on-chip process variations. Which all parameters of BSIM4 model file I should vary in Montecarlo simulations ? I think I should vary these five parameters: vth0, dl, dw, lpe0, K3.
Am I correct ?

For 90nm Technology what should be approximate sigma of pdfs of these parameters for on-chip variations ?

Plz Reply
CA

Title: Re: Montecarlo Simulations for On-chip Process variations for 90nm Technology
Post by Geoffrey_Coram on Apr 28^th, 2009, 5:22am

I would expect the sigma is proprietary information of your foundry -- how mature is their process? Also note that the foundry may want to tell you the sigma is much wider than it is in reality, so that you can't claim a wafer is out of spec and they should pay.

Title: Re: Montecarlo Simulations for On-chip Process variations for 90nm Technology
Post by cmos.analogvala on May 11^th, 2009, 6:59pm

Thanks a lot for the guidance. I could get model file for MC from foundry.

-CA

Title: Re: Montecarlo Simulations for On-chip Process variations for 90nm Technology
Post by cmos.analogvala on May 11^th, 2009, 7:10pm

A small query along the same line. Now that I have model files for MC, I want to incorporate correlation between neighboring transistors for Mismatch analysis. I could see an option to specify CC and transistor pairs in ADE/ Statistical Analysis. The question still remains what CC value should I give to the transistors located within the area of 60 lambda X 50 Lambda ? I understand this is also a proprietary information. However I just need a rough estimate.

My chip has two modules which are far apart and the modules are very small. I need to consider the fact that all Module A transistors are identical. All module B transistors are identical. However, transistors of module A and module B are not identical.
Is CC=0.9 an for transistors in single module an educated approximation ?

Title: Re: Montecarlo Simulations for On-chip Process variations for 90nm Technology
Post by mowiehowie on May 14^th, 2009, 1:45am

Hi, I got almost the same doubt.

The only document I got gives me a rule of thumb saying that the accuracy for a MC simulation is Δμ/μ [%] = 100*3.92*(σ/μ)(1/sqrt(n)).

Where n is the number of runs, Δμ/μ is the accuracy, and σ/μ is the deviation/mean ratio.

Okay, I can set the number of runs on the simulator but how do I set the sigma/mean ratio??? :-?

Thanks a lot.