Instrumental Variable Vector Fitting Method for Rational Approximation of Noisy Data

This work presents a novel vector fitting (VF)-instrumental variable (IV) method to approximate systems characterized by noisy tabulated frequency responses. The proposed method uses a two-stage least-squares (2SLS) formulation and a Gaussian weighted Savitzky–Golay (SG) filter to formulate instruments. This enables the proposed method to be less sensitive to initial poles, to improve conditioning, to converge in fewer iterations, and to be efficient for multiport problems.