Identification Methods with Different Digital Predistortion Models for Power Amplifiers with Strong Nonlinearity and Memory Effects

摘要

Different methods have been used to identify a digital predistortion (DPD) model coefficients in power amplifier (PA) linearization, especially direct and indirect learning architecture (DLA/ILA), and iterative learning control (ILC) technique, etc. A comparison is made in this paper under the scenario of a PA with strong nonlinearity. Analysis of DLA and ILC shows their similarity. However DLA works only with linear-in-parameter models, excluding the new emerging neural network models which has shown a good linearization accuracy with ILC according to experimental results.