Skip to Main Content
This paper presents an efficient methodology to improve the convergence properties of vector fitting (VF) when the frequency data is contaminated by noise. The proposed algorithm uses an instrumental variable approach, which minimizes the biasing effect of the least squares solution caused by the noise of the data samples. These instruments are generated using the rational approximation of the previous iteration and does not increase the computational complexity of the VF algorithm. Numerical examples are provided to illustrate the validity of the proposed method.