Comparison of contemporary parameterized model order reduction techniques

Abstract

Interconnect effects are one of the dominant factors affecting performance and design of high-speed networks. These effects if not predicted at early stages of the design cycle could lead to costly design iterations. Evaluation of the effects of interconnects requires solving large sets of equations, the direct solution of which is computationally expensive. The need to improve this computational cost brought about the development of model order reduction (MOR) techniques. However, MOR techniques do not consider process and parameter variations that cause inevitable changes in performance at high frequencies. To deal with such performance-accuracy issues, parameterized model order reduction (PMOR) techniques were introduced. In this paper, a comprehensive comparative analysis is presented for contemporary PMOR techniques against a set of established criteria.