Analog circuit optimizer based on computational intelligence techniques

Abstract

The computational intelligence techniques for analog circuit optimizer are presented in this paper. This technique uses a diffusion genetic algorithm (DGA) to identify multiple "good" solutions from a multiobjective fitness landscape which are tuned using a local hill-climbing algorithm. The DGA together with fast and accurate circuit performance estimator (CPE) based on neuro-computing technology is used to provide a nature niching mechanism that has considerable computational advantages and generate as many "good" design solutions as possible. The local hill-climbing algorithm restricts the search in the basin of attraction of a design solution, thus tries to tune the design up-to the sub-optimum by using SPICE to validated the performance parameters of synthesized circuits.