An iterative pruning of 2-D digital predistortion model based on normalized polynomial terms

Abstract

This paper proposes a new method to prune the kernels of 2-D digital preditortion (2-D-DPD) model. The conventional 2-D-DPD model is rewritten as a new form with each of the polynomial terms normalized for further pruning. Then an iteration based characterization process is taken to prune the terms of the model. Two experiments on different PAs were taken to verify the method. Both of them prove the proposed method works effectively to prune the terms of 2-D-DPD model without losing modeling precision. The numbers of coefficients in the model were reduced from 24 to 9 and from 30 to 13 while only about 1dB adjacent channel power ratio (ACPR) and 1dB normalized mean square error (NMSE) were sacrificed.