Predicting system performance variances using enhanced propagation of variance

In the computer aided design (CAD) area, design for quality (DFQ) is becoming more and more important in the industry. In designing integrated circuits (ICs), the major objective of DFQ is to design ICs characterized by small performance variability, high reliability and high manufacturing yield. A method of predicting a system performance variance called enhanced propagation of variance (EPOV) is proposed. This method combines the linear and quadratic POV with the potential of dealing with higher orders. The EPOV method adjusts the usage of different models depending upon the statistics computed. Using a specific IC circuit as a system example, significant reduction of error and more accuracy of variance estimation are reported.