Models for a New Profit-Based Methodology for Statistical Design of Integrated Circuits

Abstract

A new methodology for the optimization of integrated circuit (IC) parameters which takes into consideration realistic statistical fluctuations occurring in their fabrication, is proposed. The methodology is based on a problem formulation which consists of maximizing a criterion of goodness never before used in statistical circuit design. The criterion is one that recognizes the ultimate economic goal of engineering design--profit expected from the production and sale of the product being designed. In particular, this criterion is motivated in the general context of engineering design and of statistical circuit optimization. The form of profit most appropriate to statistical circuit optimization is specified. The design parameters of the methodology are identified, and the dependence of profit on them modeled.