Machine Learning Assisted Circuit Sizing Approach for Low-Voltage Analog Circuits with Efficient Variation-Aware Optimization

Abstract

Low-power analog design is a hot topic for various power efficient applications. Sizing low-power analog circuits is not easy because the increasing uncertainties from low-voltage techniques magnify process variation effects on the design yield. Simulation-based approaches are often adopted for analog circuit sizing because of its high accuracy and adaptability in different cases. However, if process variation is also considered, the huge number of simulations becomes almost infeasible for large circuits. Although there are some recent works that adopt machine learning (ML) techniques to speed up the optimization process, the process variation effects are still hard to be considered in those approaches. Using the popular evolutionary algorithm (EA) as an example, this paper proposes an ML-assisted prediction model to speed up the variation-aware circuit sizing technique for low-voltage analog circuits. By predicting the likelihood for a design that has worse performance, the enhanced EA process is able to skip many unnecessary simulations to reduce the convergence time. Moreover, a novel force-directed model is proposed to guide the optimization toward better yield. Based on the performance of prior circuit samples in the EA optimization, the proposed force model is able to predict the likelihood of a design that has better yield without time-consuming Monte Carlo simulations. Compared with prior works, the proposed approach significantly reduces the number of simulations in the yield-aware EA optimization, which helps to generate practical low-voltage designs with high reliability and low cost.