Optimal order estimation for modeling and predistortion of power amplifiers

Power amplifier is an essential component in communication systems. Digital baseband predistortion is a cost effective approach to linearize a power amplifier. Models derived from Volterra series are often used. For a given type of model, two questions have to be solved: estimation of the model coefficients, and determination of the model structure, e.g. orders of non-linearity, memory lengths, etc. There have been many works on the estimation of model coefficients but few works have focused on the determination of the model structure. To choose proper orders of series is not evident in order to keep low complexity with acceptable results. In the article, we propose to use an integer genetic algorithm for the determination of orders of polynomial series for predistortion or power amplifier modeling. The used fitness function aims to achieve a compromise between accuracy and complexity of the model. The method is evaluated on real measured power amplifier. The obtained results show that the proposed integer genetic algorithm is able to determine in a very small number of iterations a good model structure. Its complexity is quite low and it can be applied whether offline or online for adaptive determination of the structure.