Analogue placement by formulation of macro-components and genetic partitioning

Abstract

A new analogue component placement approach by formulating macro-components and genetic partitioning, utilizing genetic algorithms, is presented. The macro-components are formulated based on knowledge and experience. The slave-components are placed with the exhaustive approach. The genetic partitioning, a central part of the entire placement process, is used for block-level placement. Two circuits are examined for this purpose. The results of the first circuit are optimal in 85% of the cases when compared with those obtained using the exhaustive search approach. Genetic partitioning is competitive with simulated annealing in performance, and faster than simulated annealing in speed, as verified by the second circuit. The conventional genetic algorithms are modified such that the size (number of components) of a solution is a function of partitioning level, and the population is a function of the size.