Skip to Main Content
This paper presents a new methodology for automated broadband model generation from S-parameter data for interconnects and passive components. The new methodology is based on augmenting an existing equivalent circuit model with a macromodel (black-box) network described by rational functions while simultaneously perturbing the equivalent circuit component values. The macromodel network is determined using standard least-squares or vector-fitting approaches. The perturbation of the equivalent circuit parameter values is achieved during the macromodel generation by means of global optimization based on intelligent search algorithms. The new approach is demonstrated on several two-port test example structures including a broadband probe tip structure and a CMOS spiral inductor.