PASSIOT: A pareto-optimal multi-objective optimization approach for synthesis of analog circuits using Sobol' indices-based engine

Abstract

This paper presents a new approach for multi-objective optimization synthesis of analog circuits based on computing Sobol' indices for vectors of input variable parameters of the circuits, PASSIOT. By adopting Sobol' sensitivity analysis, the tool quantifies the amount of variance that each parameter (or a set of parameters) contributes to the uncertainty of the model output presented in the objective functions of the circuit. Assessing different higher order sensitivities effects of different sets of input parameters on the model output results in quantifying the parameters having the highest impact on the output uncertainty, traverses the output design space efficiently, and converges to higher circuit performance in a reasonable runtime compared to other approaches. PASSIOT adopts corner-driven Pareto-Optimal solution model for efficient multi-objective optimization of analog circuits parameters. PASSIOT is simulated using real objective functions and is proven to be competitive in terms of runtime and solution quality.