Optimisation of Electrical Network Responses

Abstract

The so—called insertion loss method of electric al filter design, developed during the years 1939–50 almost independently in the USA (Darlington) and in Germany (Cauer, Piloty), has come into vogue since it was possible to use comp uters to carry out the very tedious computations required. The method is integral, in the sense that it obtains values of components of the whole network by a sequence of well—defined processes, and it is practical since it starts with a set of operational specifications, such as insertion loss in given frequency bands which can be realised with a high degree of precision. However, certain limitations of this procedure soon became evi dent. The method is limited to reactive networks operating bet ween resistive loads, whereas practical components (e.g. inductances) are associated with stray effects, such as dissipation, with serious consequences in selective filters. Also, the met hod consists of two distinct steps (approximation and realisation) and a choice in the first step necessarily curtails the scope of success in the second. The use of optimisation algor ithms seems to be a good strategy to accomplish design indepen dently of limitations of the network theory.