Accurate compact model extraction for on-chip coplanar waveguides

Employing a novel ABCD matrix partitioning technique, the proposed method extracts an accurate frequency-dependent distributed circuit model for the broadband characteristics of coplanar waveguide transmission lines using either measured or simulated scattering parameters. A neural network with a training algorithm is then applied to predict CPW performance for a range of geometry sizes. The circuit model validated using EM (MOMENTUM) and circuit (HSPICE) simulations demonstrates good accuracy, high computational efficiency, and fast model development.