Parametric model order reduction technique for design optimization

Model order reduction has proven to be an effective tool for dealing with the computational complexity that arises during the simulation of large interconnect networks. However, in the case of parametric reduced order models, the effectiveness of traditional reduction methods is dependent on the number of moments and cross moments required to construct the orthonormal basis used in the congruence transformation. This can result in a relatively large reduced system in cases when the number of parameters is large. We propose a new approach for constructing the orthonormal basis that is not directly dependent on the moments. This new technique reduces a circuit with respect to many parameters by using singular value decomposition as a tool to filter out redundant information from the original subspaces. The result is a parametric reduced order model that is smaller, but still conserves the essential behavior of the original circuit as a function of frequency and other circuit parameters.