Robust Parametric Macromodeling Using Multivariate Orthonormal Vector Fitting

Abstract

A robust multivariate extension of the orthonormal vector fitting technique is introduced for rational parametric macromodeling of highly dynamic responses in the frequency domain. The technique is applicable to data that is sparse or dense, deterministic or a bit noisy, and grid-based or scattered in the design space. For a specified geometrical parameter combination, a SPICE equivalent model can be calculated.