Statistical methods for the determination of process corners

Abstract

Presents a statistical method to determine the variation of the production process of MOS transistors by finding the wafers that have parameter values on the boundary of the distribution. For the selection of the wafers a location depth method is used. Since it would be too time-consuming to determine the SPICE parameters for all the wafers and compute the boundary wafers in the SPICE domain, we use a different approach. We compute the boundary wafers based on production control parameters and then we transform the production control parameter values to SPICE parameter values. With the SPICE parameter values obtained in this way the circuit simulation is performed and since we use the data of the boundary wafers, we cover the variation of the production process within a certain time period. The applied scheme proves to describe the performance variation of analog/mixed-signal designs very accurately with a small number of simulations. For validation purposes circuit simulations and measurements of benchmark circuits are compared. The statistical methods can easily be integrated into a mixed-signal design environment.