Accelerating DC Circuit Simulation with Reinforcement Learning
TimeWednesday, July 13th2pm - 2:30pm PDT
Location3004, Level 3
Event Type
Research Manuscript
Analog Design, Simulation, Verification and Test
DescriptionPost-layout transistor level circuit simulation is the most time-consuming task in the back-end design. Nowadays, fail or significantly slow to find the DC solution is one of the greatest challenges in SPICE-like simulators, especially under the advanced process. Pseudo-transient analysis has a strong ability to eliminate discontinuous brings better convergence, however, it suffers from low simulation efficiency. In this paper, we propose a fast and robust DC analysis method utilizing cutedge machine learning techniques to provide adaptive pseudo-transient analysis for various kinds of circuits depends on circuit properties and behavior.