Application of statistical experimental design to the development of a one micron CMOS process

The concepts of statistical experimental design have been applied to the development of a 1- mu m CMOS process for the manufacturing of integrated circuits with optimized performance, while maintaining sufficient margin for both manufacturability and reliability as design goals. The application of these concepts, along with the use of an expert system (RS/Discover) as the basis for selection, design, and analysis of each experiment, is discussed. Examples are presented of the characterization of short-channel MOSFETs developed using the RS/Discover system as the basis for experimental design. Examples of parametric targeting, process optimization, and sensitivity analysis are given.<
>