Isomorphism elimination to enhanced design centering of analog circuits using GA and the regionalization method

摘要

A two-stage analog circuit design centering technique, using Genetic Algorithms (GA) and the Regionalization method, has previously been described. Circuits are produced that satisfy specified frequency response specifications whose elements are chosen from a set of preferred values incorporating associated parasitic effects. In the first stage, approximate feasible regions are obtained using relaxed specification of the desired response. During the second stage, the search concentrates in these approximate regions. In this paper an enhancement to the technique is introduced, in which the problem of isomorphism among genetic populations of solutions is tackled. This has proven to be a major problem as the solution space becomes more limited in stage two of the search. A simple and efficient technique is used to determine isomorphic solutions, resulting in a faster convergence and producing a broader window into the solution space of the required circuit design.