Application of Genetic Algorithms to Nonlinear Microwave Circuit Analysis Methods

Abstract

A genetic algorithm (GA) has been used to demonstrate a new hybrid method for nonlinear microwave circuit analysis: frequency-domain sample balance (FDSB). As in harmonic balance, FDSB updates voltages in the frequency domain, and analyses the linear and nonlinear subcircuits in the frequency-and time-domains respectively. However, the error currents are evaluated in the time domain. This is made possible by the use of a genetic algorithm. It means that unlike conventional harmonic balance, no Fourier transforms are required. This makes the simulation of circuits with a large number of incommensurate multitone inputs feasible. Moreover, the method dispenses with the need for both a Jacobian matrix (as used in most harmonic balance simulators) and a good initial guess. Results are presented for both single-tone and multitone input signals. In all cases, the FDSB algorithm demonstrates robust convergence properties, regardless of starting conditions.