Model Parameter Identification with SPICE OPUS: a Comparison of Direct Search and Elitistic Genetic Algorithm

SPICE is a widely used tool for simulation of analog circuits. Accurate models of devices are crucial for obtaining realistic results from the simulation. Many devices must still be modelled with subcircuit models in order to achieve sufficient accuracy. Parameter values for subcircuit elements are determined by optimisation. Use of SPICE OPUS optimisation capabilities for model parameter identification is presented. The performance of the simple genetic algorithm when used for model parameter identification is evaluated. The effect of elitism on convergence and results is examined for 9 cases of parameter identification and 2 benchmark optimisation cases. The performance of the simple genetic algorithm is compared to the performance of the simplex algorithm. Finally conclusions and guidelines for future work are presented.