A predistortion technique for linear networks

This letter presents an efficient algorithm for the predistortion of networks with arbitrary topology. The method is based on a modification of the Levenberg-Marquardt algorithm using small-change transmission sensitivities. A weighted least square error function is formulated and used subsequently in an iterative optimization scheme. The algorithm determines the errors of the transfer function coefficients due to parasitics of the physical network very accurately. Root-finder methods are used to obtain frequency and Q-deviations of poles and zeros. The predistortion of the network is carried out by perturbing the nominal transfer function poles and zeros correspondingly.