Propagation of variance using a new approximation in system design of integrated circuits

Abstract

In this paper we are concerned with Design For Quality (DFQ), which has recently attracted much attention in system design. Our aim is to introduce effective methods for estimating the variability of a system's performance. The linear approximation method is perhaps the most often used method for predicting variance propagation in a system. In this paper, we exploit the Moment Method in statistical analysis to extend the linear approximation to quadratic approximation, which will be referred to as Quadratic Propagation of Variance (QPOV). The derivation of the new method is provided and some results of its application in a real circuit are given. Our examples indicate that the new method QPOV gives much more accurate estimates for performance variability. Also this method can be used in other CAD fields.