Variance Weighted Vector Fitting for Noisy Frequency Responses

This letter presents a modification of the Vector Fitting algorithm to estimate rational models of frequency responses affected by noise, e.g. in the case of measured data. It is based on the use of a least-squares weighting function that contains information about the variance of the data samples used in the macromodeling process. The modification is straightforward and does not increase the computational cost of Vector Fitting, but it significantly improves its convergence and accuracy properties in the presence of noise. The residuals test is used to assess the quality of the estimated model. Numerical results validate the proposed modification.