Manufacturing process variations have become a major issue in today's semiconductor chip design. In order to improve the robustness of nano-scale chip design, it is highly desirable to develop novel simulators to efficiently quantify the uncertainties of integrated circuits (ICs) and microelectromechanical systems (MEMS). Mainstream simulators (such as HSPICE, Cadence Spectre and MEMS+) employ Monte Carlo for statistical analysis, requiring a huge number of repeated simulations. In this talk, I will present some novel stochastic spectral methods for stochastic circuit and MEMS simulation.
I will present two frameworks for fast stochastic IC and MEMS simulation. First, a variant of stochastic collocation method called stochastic testing will be presented for fast stochastic DC, transient and AC analysis of nonlinear integrated circuits. Advanced periodic steady state simulation for efficient uncertainty analysis of analog/RF circuits (such as electronic oscillators) will also be discussed. Then, I will report a hierarchical uncertainty quantification approach that can handle complex systems with several subsystems and high-dimensional random parameters. The simulation results on many IC and MEMS examples have shown that our simulator can achieve 100x to 1000x speedup over state-of-the-art circuit and MEMS simulators.
Thesis Supervisor: Professor Luca Daniel