Loewner Matrix interpolation for noisy S-parameter data

Abstract

Loewner Matrix (LM) interpolation technique was proposed as an efficient macromodeling approach compared to state of the art technologies. However, the method becomes inaccurate in presence of noise as it interpolates noise itself. In this paper, we propose a LM interpolation technique suitable for extracting an accurate and passive macromodel from noisy S-parameter data. An order searching algorithm to find the most accurate model maintaining stability is proposed first. Then we propose a least-square approximation based correction on the macromodel. Finally, the passivity of the model is ensured by using a Hamiltonian Matrix Pencil perturbation scheme. The advantages of the proposed approach is illustrated using one full-wave example.