Statistical Circuit Design: Characterization and Modeling for Statistical Design

In statistical design work involving discrete passive elements, nominal parameter values and production distributions give an adequate description. The situation is more complicated for active devices in which the equivalent circuit used to describe these devices requires parameters which are interrelated. With integrated circuits, conditions are further compounded by the fact that the parameters of different devices on an IC chip are interdependent. Experimentally, however, it has been found that the integrated circuit case can be conveniently decoupled to a manageable degree of complexity even when temperature effects are considered. In the following sections, the philosophy underlying both our modeling approach and the measurement techniques is outlined. This is followed by a description of the transistor model favored for statistical analysis work, highlighting some popular misconceptions arising from inadequate past measurements. The implications for device modeling in an integrated circuit environment are then considered, followed by a description of the practical methods which have been successfully used to predict the variability in circuit performance arising from variations in the manufacturing process. 1105