Yield optimization using k-means clustering algorithm to reduce Monte Carlo simulations

Abstract

This paper presents an efficient yield optimization approach using k-means clustering algorithm to reduce Monte Carlo (MC) simulations. This approach uses a commercial electrical simulator and PDK models for evaluation purposes. The method was integrated in an analog IC design flow that includes the AIDA-C circuit sizing optimization tool. The proposed yield estimation technique reduces the number of required MC simulations during the optimization process. The simulated solutions are the most likely to populate the Pareto optimal front and result from a selection process based on a modified k-means algorithm. The proposed approach leads 75% reduction in the total number of the MC simulations for the presented case study.