Abstract
The primary objective of this dissertation is to develop new performance function modeling strategies to be used for statistical design of integrated circuits. The major bottle-neck during statistical circuit design is caused by the prohibitive cost of circuit simulation for statistical analysis. The computational burden is reduced by using suitable approximation models of the performance functions of interest. The major emphasis is placed on Design for Quality (DFQ) which aims to minimize the variance of the performances and fulfill all the specifications. This requires more accurate approximation models than would be the case for yield maximization alone. The dissertation is divided into two major sections. The first section deals with the development of the proposed modeling strategies. The major contribution of this dissertation is the development of highly accurate approximation methodologies, which are tailored to individual applications. Three general classes of modeling strategies are developed. The first class is a black-box modeling strategy obtained by combining an interpolation scheme and a self-organizing regression scheme. The second class of methods is applied to improve the accuracy of physically based analytical models (normally preferred by designers and often available, but are usually too simple for statistical analysis) by combining them with a black-box model. The black-box part of the new model is used to provide error correction to the physically based analytical models. The final class of models can statistically characterize large CMOS combinational digital circuits, which, until now, has been beyond the state of the art. The effects of geometric and technological process parameters are included in the solution. The methodology combines analytical/black-box/numerical methods to achieve a high level of statistical accuracy, with speeds of 5 to 6 orders of magnitude faster than circuit simulation (SPICE). For each class, a set of representative examples is used to demonstrate the superiority of the proposed strategy...
Aftab, Syed Amir (1993). Performance function modeling for statistical design of integrated circuits. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -1523640.