Nonlinear microwave circuit analysis using the genetic algorithms toolbox in MATLAB

This paper will show how it is possible to use a relatively inexpensive MATLAB toolbox together with MATLAB to apply a genetic algorithm to a both the sample balance and harmonic balance techniques, and then subsequently solve a variety of nonlinear microwave circuits. The extension of this method to multitone circuit analysis (in which the tones are noncommensurate) will also be highlighted. One of the strengths of genetic algorithms is their relative transparency to the problem being solved. Auxiliary information (such as a Jacobian) is not required and there is great flexibility in the choice of the fitness evaluation function. This allows the development of a frequency-domain sample balance technique, in which the error is formed in the time-domain but the voltages are updated in the frequency-domain. It will be shown that this novel application of genetic algorithms yields a nonlinear microwave circuit solver that is robust, can ultimately handle any number of noncommensurate input tones, and does not assume any a priori knowledge of circuit behaviour in the steady state.