James Grimbleby

Automatic Circuit Synthesis

Genetic algorithms (GAs) are a closely-related group of optimisation methods which attempt to exploit the remarkable success of natural selection in adapting organisms to suit their environment. A population is maintained, each individual of which has a unique genotype and fitness. All the information (the genotype) needed to describe an individual is coded in a “chromosome”, and the fitness is a measure of its success in meeting some objectives. GAs aim to improve the general fitness of the population by selective breeding.

GAs can be used to synthesize circuits to meet required specifications. It is only necessary to be able to analyse a circuit and compare its specification with the target specification in order to apply GAs. No expert knowledge of circuit synthesis need be written into the program.

Initial work concentrated on passive circuits: Galesia95

This has now been extended to active circuits: Galesia97

Circuit design involves the selection of both suitable circuit topology and component values. The key to efficient synthesis is to use a hybrid-GA approach: GAs are used to generate new circuit topologies and numerical optimisation is used to select suitable component values.

In some cases circuits have been generated that have been more efficient than those developed by formal design methods. For example, the elliptic filter is often thought to give the most economical filter design for sharp cut-off filters but hybrid-GA synthesis may produce a better design: CEC99 and IEE00

Lately a new circuit representation method has been devised which can be used with Genetic Programming (GP) to create new circuit configurations. Building block 4-terminal circuits, each containing a single passive component, are assembled into a candidate circuit using a set of unary and binary operations. This work has not yet been published but is available as a preprint.

Automatic Circuit Synthesis Demonstration Program


Page last updated July 24, 2012
email: j.b.grimblebyXremX@rdg.ac.uk (remove XremX)