Abstract
A method using orthogonal array (OA)-based design of experiments is applied to screening and stochastic gradient evaluation for a generalized yield/Taguchi 'loss' function and a general CMOS statistical model. In comparison with the Taguchi approach, the method does not require 'inner' and 'outer' table combinations. The method is shown to be much more efficient than an equivalent Monte Carlo approach on a practical CMOS circuit design example.

This publication has 4 references indexed in Scilit: