This work considers the computationally challenging problem of optimizing systems of broadband electromagnetic (EM) devices fed by strongly non-linear circuits. Traditionally, fine-tuning this system would require self-consistent fully coupled analysis of the EM-circuit network in the time domain. Optimization is then dependent on the cost of repeatedly evaluating the tightly-integrated broadband device and strongly non-linear connected circuit.A novel method for transient port parameter extraction enables a circumvention of this computational cost. It constructs a reduced-order EM-circuit representation in which the equivalent EM model is circuit-agnostic. This method of port extraction has been demonstrated to produce equivalent results to that of self-consistent analysis. The circuit-agnostic approach is amenable to any optimization framework given the EM system is invariant. An optimization scheme is developed for application to tightly-coupled broadband EM and non-linear circuit systems that is otherwise untenable with a self-consistent coupled method.This work demonstrates the use of a genetic algorithm to optimize such a system. The optimization scheme is applied to both linear and non-linear circuit systems for objectives including reflection coefficient minimization, reduction of distortion in a nonlinear amplifier feeding a broadband antenna, and reverse beam-steering for an array with time-varying incident interference.
Read
