ABSTRACT
A good statistical method can give a speed-up against running all combinations of parameters. Monte-Carlo is so such a technique and practically the most general one. This chapter focuses on MC estimation methods for single real values, like the partial yield or the mean or standard deviation of a certain performance. One important measure for robustness is the production yield, but also others can be of designer’s and quality engineer’s interest. The designer’s real interest is usually in the performance variations, and in between both there is a long often highly nonlinear circuit simulation. MC analysis requires margins due to confidence interval widths. Systematical errors often cannot be reduced so easily by just increasing the MC count.
