CMOS Analog Circuit Automation Using Multi-Objective LearnableEvolution Model

This paper presents a new simulation-based CMOSanalog circuits design automation method which uses a multiobjectivealgorithm based on Learnable Evolution Model(LEM). Strength Pareto Evolutionary Algorithm (SPEA) is usedfor handling multi-objective property of the problem. LEM as aNon-Darwinian-type evolutionary algorithm includes a machinelearning method like decision trees to discriminate betweenhigh and low fitness areas in the design space. The learningprocess can find the right directions of the evolution and causehigh steps in the individuals’ evolution. So, the learning phaseshortens the evolution process and makes remarkable reductionin the number of individuals’ evaluations. Also, a knowledgebase of designer’s expert about circuit is created and used inthe analog circuit design process. Circuit evaluation is done byHSPICE simulation. Bsim۳v۳ CMOS transistor model is used inthe proposed design method in order to improve designaccuracy. The proposed method has been tested on a CMOSanalog operational amplifier circuit. Obtained results verifyperformance of the proposed method in CMOS analog circuitdesign in comparison with evolutionary strategy algorithm.

کلیدواژه ها:

CMOS analog circuits ، automated design ، learnable evolution model ، multi objective