Neural Networks and Volterra series for modeling new wireless communication devices

In this work, a TDNN based behavioral model is proposed for accurately reproducing the nonlinear and dynamic behavior of a wireless communications device. Moreover, an efficient procedure to extract a Volterra model from the parameters of the behavioral model is explained, thus providing a simple way to construct very compact and accurate Volterra models, which provide open information about device performance, and their implementation in RF CAD circuit simulators is generally less time-consuming. Two tests try to demonstrate the validity of the proposed approach, both for one-tone and two-tones test characterization in RF of PAs showing strong nonlinearities and memory effects.