Abstract :
Analogue circuit design optimisation has for a long time been treated classically through a number of conventional optimisation techniques such as gradient methods or Hill-Climbing techniques. The component values produced by these conventional methods are typically assumed to be ideal with unrestricted values in a circuit configuration that is pre-defined. Therefore it would be very desirable to perform the search for an optimum design in a solution space that more closely reflects partial applications. Here, components may be restricted to have values that are selected from a menu of pre-defined preferred values, and have associated imperfections such as individual parasitics. It is further desirable that the circuit structure not be pre-defined since in general it would not be known to be optimal, but be included with the optimisation process. In this form, the optimisation problem is very complex with many local minima. These difficulties have suggested the use of evolutionary optimisation methods. The aim of this paper is to review progress to date and to describe some work in hand, on the evolutionary design of analogue electronic circuits. The use of genetic algorithms is more widely reported and dealt within this paper
