Complex valued neural network modelling of the balun of a biconical antenna

A methodology is presented aimed at extracting the parameters of a lumped equivalent circuit model of a network from measurements of its complex network (scattering) parameters. The core of the extraction procedure is a modified complex valued neural network that can directly accept the complex network parameters as inputs. The methodology is here applied to the derivation of the circuit model of the balun of a biconical antenna. The complete process of parameter extraction is described paying particular attention to the description of the physical model of the network, the measurement setup and the operation and training of the neural network.