Constrained circuit optimization via library table genetic algorithms

Genetic Algorithms (GAs) are presented as a robust method of obtaining optimal or near-optimal solutions to circuit optimization problems. Circuits which must contain devices from a constrained “parts library” are shown to be particularly well-suited for optimization by genetic algorithms. As a practical example of the optimization method, a genetic algorithm implementation was used to optimize a Gilbert Cell mixer with respect to several competing metrics. The simulated power consumption, mixer gain, and IP3 of the mixer were used to construct a cost function. This cost function measure was minimized by the GA, producing several alternative Gilbert Cell mixers as outputs. The solution set was constrained to contain devices chosen from a library of previously characterized MOSFETs