Automatic SPICE- Integrated Reinforcement Learning for Decap Optimization for EMI and Power Integrity

Abstract

The automatic SPICE-integrated reinforcement learning (RL) is proposed for decap optimization for radiated electromagnetic interference (EMI) and power integrity. A power distribution network (PDN) structure is modeled in a circuit fashion to be solved in a SPICE solver. For EMI optimization, the branch currents for radiated EMI calculation were obtained from ac simulations. For PI optimization, the voltage fluctuations in time domain were obtained from transient simulations. Finally, it is demonstrated that a consistent RL environment integrated with SPICE solvers can be utilized in the optimization for both radiated EMI and PI.