Oscillator phase noise verification accounting for process variations

In this paper, an optimization based methodology is developed to investigate the effect of process variations on oscillator phase noise in transistor level. This approach formulates the worst case oscillator phase noise in the presence of process variations as a bound constrained nonlinear programming (NLP) problem and solves it with a Multi-start Global Optimization (MGO) algorithm. Rigorous comparisons between our method, Monte Carlo (MC), Quasi Monte Carlo (QMC) and Simulated Annealing (SA) methods show that, for the same accuracy, our method gains up to 1724×, 575× and 34× speedup over MC, QMC and SA methods respectively.