Orthogonal polynomial based Hammerstein behavioral model for power amplifiers with strong memory effects

This paper investigates the robustness of Hammerstein based behavioral models for highly nonlinear power amplifiers exhibiting strong memory effects. Augmented Hammerstein models have been proposed to replace the linear function in conventional Hammerstein model by a nonlinear function typically implemented as a memory polynomial. This work illustrates the lack of robustness of such memory polynomial based models and proposes a new type of Hammerstein models based on orthogonal memory polynomials. Experimental validation carried on a high efficiency Doherty amplifier driven by multi-carrier WCDMA signals illustrate that the proposed models maintain the prediction accuracy while considerably improving the identification robustness quantified in terms of the condition number of the Vandermonde matrix.