Monte Carlo simulation for integrated circuits

A method for generating multidimensional normal random numbers that is suited to Monte Carlo simulation of integrated circuits is presented. The method makes it possible to generate random numbers with accurate correlations in a short time and using a smaller core size than that required by direct application of Moonan's method. In actual simulation, it is made possible to change mean values, total variance, and correlation of element parameters, as well as structure of these factors, and to easily examine results of making such changes. The main features of a program for this simulation and an example of its application are described. In the method two factors, inter-pellet factors and in-pellet factors, are considered as factors responsible for variability element parameters, but the method may be easily expanded to deal with multifactor structures taking into account interbatch, or interslice factors, etc.