Multilevel Parameterized Model Order Reduction for Variability Analysis of Circuits

A multilevel parameterized model order reduction technique is proposed for variability analysis of large circuits. The original large-scale equations are kept into the multilevel scheme, which allows a full error control about the estimation of stochastic moments. A high computational efficiency is also provided by a set of parameterized reduced order models properly combined with the original large-scale model. Pertinent numerical results are presented to validate the proposed technique.