Fast search agorithm for tolerance design

Abstract

A new fast search technique based on search-space smoothing is proposed for the tolerance design of electronic circuits. By smoothing and enlarging the sampling region of Monte Carlo analysis so that focus is placed on the global features of the acceptance region, the authors looked at the solution space from a larger perspective and reduced the number of local optimal points to be investigated. After the target area is identified the algorithm zooms in to focus on the detailed features of the target area. The yields on spaces with different degree of smoothing are computed and finally the most probable global solution is obtained. The algorithm is simple and efficient. The proposed method is applied to different electronic circuits. Results show that the computational efficiency and the resultant yield achieved are significantly better than the popular centres-of-gravity method. © IEE, 1998.