Statistical methods in semiconductor design and manufacturing
by
Steven G. Duvall
Intel Australia Pty Ltd

The application of statistical methods has been critical to the success of the semiconductor industry in achieving high levels of production of increasingly complex integrated circuit products. Following the initial applications of statistical methods in manufacturing in the 1980’s, they spread rapidly throughout manufacturing engineering groups and are now routinely used to understand and reduce process variation, diagnose manufacturing problems, optimize process settings, etc. Despite these successes in manufacturing, statistical methods have seen little use in other engineering disciplines, notably circuit design. However, the industry’s continued drive to higher levels of integration and performance has increased the sensitivity of circuits to process variations and increased the risk process variations pose to circuit functionality. Although managing this risk has required ongoing reductions in process variation, manufacturing solutions have not been sufficient. Product design groups have been increasingly forced to treat process variation explicitly, driving efforts to develop statistical methods for integrated circuit design. These efforts face significant challenges, among them the complexity of design, the inherent uncertainty in design, and the concurrency of process development and product design. In this talk, we will look at these challenges and some of the statistical methods being developed to address them.

Date received: April 3, 2002