Improvement of vector fitting by using a new method for selection of starting poles

摘要

Vector fitting (VF) is known as the most popular method for frequency-dependent modelling. Using VF, frequency-domain data of a network can be converted into an equivalent circuit model. Some initial poles need to be selected before VF can be started. The starting poles affect the accuracy and convergence speed of the vector fitting method. In this paper, a new procedure is proposed to select starting poles. The proposed procedure selects starting poles by partitioning the frequency response, and then ranks the starting poles so that the most dominant poles can be used when low order approximation is desired. Case study results show that the proposed procedure improves the accuracy and the speed of VF significantly.