A novel digital predistorter technique using an adaptive neuro-fuzzy inference system

This letter presents a novel digital predistorter technique using an adaptive neuro-fuzzy inference system (ANFIS). The proposed approach employs real-time input and output signals of a nonlinear power amplifier as inputs to the ANFIS, so as to approximate the inverse functions of the power amplifier. The antecedent and consequent parameters of the FIS constructed by the ANFIS are tuned using backpropagation and least squares algorithms. Simulation shows that this novel technique has improved the linearity of a WCDMA signal by a further 4 dBc compared to a conventional look-up table (secant) approach. Moreover, this proposed technique is capable of adapting to instantaneous variation in the power amplifier response through time, which is a topic often omitted by researchers in this area.