Comparison of MOSFET mismatch models with random physical and random model variables

A comparison between two established MOSFET mismatch modeling approaches, one based upon the use of uncorrelated random physical variables and one based upon the use of correlated random model variables where the correlation is ignored, is presented. An analytical formulation is introduced that enables the designer to easily predict the relative errors in matching performance that will occur if the more popular random model parameters are used instead of the uncorrelated physical parameters. It is shown that neglecting the correlation inherent in the random model variables will result in under-estimate of the area required for meeting a fixed yield target but in the example process considered in this paper, the errors introduced by using the correlated model parameters are small.