An Instrumental-Variable QR Decomposition Vector-Fitting Method for Modeling Multiport Networks Characterized by Noisy Frequency Data

This letter presents an efficient method to macromodel large multiport systems characterized by noisy frequency domain data. The proposed method is based on the vector fitting algorithm and uses an instrumental variable approach and QR decomposition to formulate the least squares equations. The instrumental variable method minimizes the biasing effect of the least squares solution caused by the noise of the data samples while QR decomposition decouples the least squares equations of multiport systems described by common set of poles. It is illustrated, that the proposed approach can increase the accuracy of the pole-residue estimates with less iteration when compared to the traditional QR decomposition vector fitting method.