Application of Wiener polynomial decomposition to power amplifier linearization problem

The mathematical model of nonlinear device (ND) plays an essential role in the power amplifier (PA) linearization by means of digital signal processing. In this paper we propose a model derived by the Wiener orthogonalization method. The one important feature of this method is that the resulted output decomposition depends on the statistics of the input signal, and initially it was derived by Wiener in application to Brownian input. Thus, we first establish the derivation of Wiener decomposition for the band-limited Gaussian inputs, and next we generalize the results to some of the non-Gaussian cases. Using the computer simulation, we show that the property of the Wiener decomposition to discriminate uncorrelated components in the output of ND holds also for the considered type of non-Gaussian input. In the experimental part of this work we mainly focus on the memoryless nonlinearity but we also show how the results may be used to construct a reduced-complexity memory nonlinearity model. We compare convergence behavior of the adaptive parameter identification process using the proposed model and the standard power series model. The results show the major benefit in the convergence speed.