Machine Learning in Nanometer AMS Design-for-Reliability : (Invited Paper)

Abstract

With continued scaling, the susceptibility of nanometer CMOS to reliability issues has increased significantly in analog/mixed-signal (AMS) circuits. The industrial large-scale AMS circuits bring grand challenges in the efficiency of reliability design and verification. Machine learning (ML) provides one promising direction to achieve significant speedup in design closure. In this paper, we introduce typical reliability issues and review some excellent arts in applying ML approaches to AMS circuits reliability verification and design-for-reliability (DFR). We also discuss some open challenges in the industry and provide potential ML-based solutions. We hope this paper can promote the development of AMS circuits DFR.