Vector Fitting-Based Calculation of Frequency-Dependent Network Equivalents by Frequency Partitioning and Model-Order Reduction

This paper presents a methodology to identify a frequency-dependent network equivalent from a highly resonant frequency response and in a broad range of frequencies. Primarily, the identification process consists on dividing the complete frequency range into a number of subranges either uniformly spaced or by containing an equal number of resonance peaks. Vector fitting (VF) is then applied to each subrange, yielding an accurate local approximation. A refinement of the local set of partial fractions is made through model-order reduction to remove spurious (redundant) poles. The final equivalent comprises a reduced number of poles, keeping the computational load lower than when applying VF to the whole range of frequencies while giving a similar rms error. An illustrative example involving a transmission system is presented for validation of the procedures above.