Modeling of distance-dependent mismatch and across-chip variations in semiconductor devices

Abstract

We present a simple and general method of modeling distance-dependent mismatch and across-chip variations (ACV). We are able to model various shapes of spatial correlation of a process/device/circuit parameter exactly in a compact device model. Examples include Gaussian and exponential types of spatial correlations. There are no grid points, groups, and brackets in our method. The resulting spatial correlation is both translational invariant and continuous. The correlation range of the ACV part of spatial correlation can be much smaller than chip size, about the chip size, or much larger than chip size. The method can model either isotropic or anisotropic spatial correlations. © 2013 IEEE.