Model Parameter Identification with SPICE OPUS a Comparison of Direct Search and Elitistic.pdfVIP

Model Parameter Identification with SPICE OPUS: a Comparison of Direct Search and Elitistic Genetic Algorithm Arpad B ŰRMEN * Tadej TUMA * Abstract - SPICE is a widely used tool for This paper deals with the simple genetic algorithm simulation of analog circuits. Accurate models of (SGA) [3] and the effect of elitism on convergence of the devices are crucial for obtaining realistic results from model parameter identification. Genetic algorithms have the simulation. Many devices must still be modelled successfully been applied to a wide variety of with subcircuit models in order to achieve sufficient optimisation and identification problems. Their main accuracy. Parameter values for subcircuit elements advantage is the fact that they avoid getting caught in are determined by optimisation. Use of SPICE OPUS local minima. The main handicap of genetic algorithms optimisation capabilities for model parameter is their slow convergence. One way to improve this is to identification is presented. The performance of the add elitism to the genetic algorithm. simple genetic algorithm when used for model The SGA is implemented as one of the available parameter identification is evaluated. The effect of optimisation methods in SPICE OPUS and some elitism on convergence and results is examined for 9 encouraging results on benchmark circuit optimisations cases of parameter identification and 2 benchmark we