A novel broadband macro-modeling using piecewise vector fitting

For broadband frequency domain responses of highly dynamical systems, because vector fitting always has low accuracy especially at resonant frequency response points, a new piecewise vector fitting algorithm is proposed. The complete broadband frequency range is divided into a number of sections. Vector fitting is then applied to each section to identify the poles and residues. Through frequency partitioning, the numerical conditioning of the pole and residue identification equations is improved. Then in each section a sub-macromodel is obtained. Finally, by adding up the sub-macromodels in each section, a novel broadband macro-model with high accuracy is obtained for the whole broadband frequency domain. At the same time, to reduce redundant poles, in each section the number of poles is controlled by a predefined fitting error. An illustrative example involving a low-pass filter is presented for validation of the proposed method.