Optimization of multi-stage amplifiers in deep-submicron CMOS using a distributed/parallel genetic algorithm

Abstract

This paper presents a framework for time-domain optimization of amplifiers employing a parallel genetic algorithm based on a message passing interface. This methodology achieves a considerable reduction in the optimization time (up to 19 times faster than a serial implementation). Increasing the processing capacity allows searching within a larger design space using complex transistors models, yielding more accurate results. The optimization, based on transient simulations, is possible due to the integration of a genetic algorithm optimizer together with the open-source simulator NGSPICE source-code.