A statistically optimal macromodeling framework with application in process variation analysis of MEMS devices

Abstract

Macromodels are used extensively in circuit and process analysis for higher computation efficiency, and better insight into system behaviors. A statistically optimal and elegant framework for macro-modeling is proposed in this paper, which can successfully handle the modeling challenges created by the highly customized fabrication/design paradigm of MEMS devices. Without requirements for a priori knowledge and experience of fast emerging and highly diversified MEMS fabrication and design style, the proposed framework can adapt to arbitrary distribution and correlation by optimally scaling the order and dimension of the process variation models for trade-off between accuracy and efficiency. The effectiveness of the proposed framework is demonstrated by process variation modeling and analysis of MEMS devices.