Deep Learning Enhanced Time-Step Control in Pseudo Transient Analysis for Efficient Nonlinear DC Simulation

With the development of VLSI, circuit simulation plays an important role in circuit design, and DC analysis, as the basis of circuit behavior analysis, is the foundation for nonlinear electronic circuit simulation. Pseudo transient analysis(PTA) methods have gained great success among various continuation algorithms. However, PTA tends to be computationally intensive without proper time-step control method. In this paper, we harness the latest advancing in deep learning to resolve this issue. Particularly, a coarse and fine grained hybrid sampling strategy is used to search the optimal time- step, which resolves the problem that the optimal time-step is no precise definition in PTA theory. Afterwards, the long short-term memory(LSTM) algorithm can be utilized to learn an optimal time-step control method through feature selection and two-stage data preprocessing strategy accelerating DC analysis ultimately. Experimental results demonstrate a significant speedup of up to 41.98X.