Modeling MOSFET process variation using PSP

Abstract

The PSP model has been extensively evaluated for its ability accurately match IV and CV characteristics MOSFETs measured on a single die. This enables accurate prediction of circuit behavior for circuits made of transistors which are very similar to the typical devices which were measured. However during manufacturing variations occur in the process which leads to difference in MOSFET characteristics. We examine various random sources of variation which affect all devices on a chip and attempt to model them by varying the PSP model parameters. We focus on determining a minimum set of parameter distributions for Monte Carlo modeling believing that fixed corner and user defined corners should be derived from accurate Monte Carlo models.