Monte Carlo simulation for W/L+variaition oercent

[an_82]

I want to do Monte Carlo simulation with Gaussian distribution for W/L (width/length) in HSPICE in 0.35um CMOS technology.

How much variation percent should I consider ? and why?

 

[csolis_GT]

You mainly need to consider two types of variations: vth and K (transconductance parameter) variations. Since your technology node is relatively large, you can assume that vth variations will be dominant in most situations. You can use the approximation that σV_TH = 0.1*tox/√(W*L). This discussion is on Razavi's book on section 13.2. If you want more precise precise simulations refer to your process documentation and see if they model mismatches, they will usually tell you standard deviations parameters and formula to model both types of variations.