Sensitivity analysis in the optimal sizing of analog circuits by evolutionary algorithms

The sensitivity analysis based on finite differences and Richardson approximation for the optimal sizing of analog circuits by applying the multi-objective evolutionary algorithm NSGA-II, is presented. The proposed sensitivity analysis approximation allows to calculate the partial derivatives of the circuit performance without an explicit equation. In this manner, both the optimization task and the sensitivity analysis are executed by linking HSPICE in order to compute the circuit performances. The case of study is the optimal sizing of the voltage follower, for which it is shown the usefulness of applying the sensitivity analysis approach to identify the solutions that are not reliable, i.e. those degrading the circuit performances.