Self-organizing multivariate constrained meta-modeling technique for passive microwave and RF components

A self-organizing algorithm is developed for multivariate constrained modeling of general passive components. The algorithm builds compact, analytical circuit models and represents the scattering parameters of the passive components as a function of its geometrical parameters and as a function of the frequency. Multiple constraints, or relationships between the geometrical parameters, may exist. The model generation algorithm combines iterative sampling and modeling techniques. It groups a number of full-wave electromagnetic simulations in one multivariate analytic model. The modeling accuracy level is user-defined. The analytical circuit models can easily be implemented and used in commercial circuit simulators. The model extraction provides EM-accuracy and generality at traditional circuit simulation speed.