On the selection of solutions in multiobjective analog circuit design

Abstract

In this work we are sizing the transistors of an CMOS Miller amplifier. More precisely, we are addressing the bi-objective problem of maximizing the DC gain and minimizing the current supply of Miller’s amplifier. As we consider a biobjective problem we get a set of solutions, instead of a single one. Next, we study three schemes to select one or several of those solutions, based on the ones with greater tolerances to variations in the circuit elements, or with lowest sensitivities, or with lowest statistical variations using Monte Carlo method. Also, it is possible to include this tolerance analysis in the optimization loop by adding an objective to the given optimization problem.