Predistortion of nonlinear amplifiers using neural networks

High data rate digital radio systems mandate the use of spectrally efficient linear modulation techniques to accommodate multiple users. Unfortunately, the fluctuating envelopes of such systems combined with the nonlinear nature of the commonly used high power RF amplifiers gives rise to spectral spreading, adjacent channel interference, I-Q crosstalk, and degraded bit error rates. A possible solution to these problems is the linearization of the transmitter system by predistorting the baseband digital signal to compensate for amplifier nonlinearities. In this work, a neural networks algorithm is used to implement a predistortion technique which takes advantage of realistic assumptions about the amplifier response. By exploiting a model of RF amplifiers based on AM-AM and AM-PM distortion functions, a neural network adaptive predistortion technique is developed which yields fewer parameters, reduced computational coats, and improved performance over other types of predistortion.