A new method for solving broadband matching problems

A new method for computing the component values of LC ladder equalizers for broadband matching purposes is presented. The method is inspired by the theory of recursive identification and results in a very compact algorithm. As in the usual real frequency technique, the load and generator impedances have to be known only at a small number of frequencies. Starting from an assumed topology for the equalizer with random values for the components, the equalizer components are iteratively built with a recursive identification algorithm, in order to satisfy asymptotically a least mean squares criterion. In contrast to the usual deterministic least mean squares algorithms used in existing equalization techniques a stochastic algorithm has been developed. This stochastic treatment leads to very simple equalization techniques. On the basis of the solution found with the stochastic recursive identification algorithm a random search of a limited region of the component hyperspace is initiated in order to refine the solution. The paper includes many examples in which the above method has been successfully applied. As a special feature, the method has been applied to equalizers made up of series and parallel resonant circuits.