A study of the evolutionary approach to network synthesis using coefficient matching.

The standard synthesis techniques are limited in that they cannot deal effectively with either parasitic elements or constraints and in that the range of networks they can adequately synthesise is limited. The computer makes it practical to use methods of directed trial and error which do not have these limitations, such as network evolution. Network evolution is a process by which changes occur in both the network topology and in the values of the network elements in such a way as to drive an objective function (some measure of the error between current and required response) to ever lower values and ultimately solution. In this case the error arises from the matching of the current set of coefficients of the network polynomials with their respective required values. This comparison produces a set of nonlinear equations which on solution give a suitable network topology and element values. These non-linear equations require optimisation techniques for their solution. It is shown that network evolution by coefficient matching is feasible in processes which primarily work either by network growth or by network reduction. The process of network growth works by taking a primitive starting network having the correct network polynomial structure and eliminating and growing elements at the appropriate state of development until a satisfactory solution is obtained. The method of analysis used, in addition to being both accurate and rapid, also gives the sensitivity of the coefficients with respect to virtual zero-valued elements. Use of this information enables a suitable choice of type, place in network and value of element to grow. The network reduction process takes initially a network which produces the required network polynomials, but with redundant common factors, and pares away the excess elements by making them open or short circuit, simultaneously removing excess common factors, until a suitable network is obtained. Suggestions are made on ways of improving the evolutionary process and increasing its scope.