Statistical circuit simulation with measurement-based active device models: implications for process control and IC manufacturability

This paper presents a new approach to statistical active circuit design which unifies device parametric-based process control and non-parametric circuit simulation. Predictions of circuit sensitivity to process variation and yield-loss of circuits fabricated in two different GaAs IC processes are described. The simulations make use of measurement-based active device models which are not formulated in terms of conventional parametric statistical variables. The technique is implemented in commercially available simulation software (HP MDS).