Addressing Load Sensitivity of Rational Macromodels

Behavioral models are effective tools used to relieve the computational burden of large-scale system-level simulations. In electrical and electronic applications, Vector Fitting (VF) iteration often represents algorithm choice for generating low order equivalent circuits complex multiport components in a data-driven setting. Although accurate reliable general, macromodels generated via VF inherently represented terms rational approximation one specific input-output transfer function structure under modeling, e.g., its scattering matrix. However, accuracy representation does not necessarily imply good when solving macromodel setting, different termination conditions. fact, sensitivity with respect loading conditions may be large needs addressed controlled. this work, we present modified scheme that overcomes issue, by introducing requirement remains interconnected known class admissible networks. The proposed formulation is based on an augmentation cost minimized at each iteration; further, it require additional expensive data gathering steps compared standard approaches. effectiveness tested over set relevant examples, particular Power Integrity applications.