Neuro-fuzzy approach to adaptive digital predistortion

A neuro-fuzzy approach to adaptive digital predistortion is presented. The hybrid approach is called adaptive neuro-fuzzy inference system and uses real-time modelling of the power amplifier responses. Measurement results show that the linearity of a two-tone signal is improved by 26.3 dB, while the noise floor of a WCDMA signal is improved by at least 12 dB.