Offline Method to Determine Optimal Complexity in Predistortion of RF Power Amplifiers

This paper presents an effective approach to determine the optimal DPD model complexity to an RF power amplifier. Thanks to the iterative learning control technique, a set of different predistorter models can be evaluated offline to determine the trade-off between model error and complexity. This allows to maximize the ACPR correction with lowest DPD complexity model. The technique is demonstrated on an 80W broadband integrated Doherty MMIC in C-Band. A Vector-switched GMP is used as predistorter. A signal with 500 MHz instantaneous bandwidth is used. ACLR after DPD reach −47.2/−48.5 dBc ACPR with saving 49% of FLOPs compare to the most complex model. This FLOPs reduction significantly decreases the DSP resources while linearity is still above 3GPP requirement with margin. This optimal model has been found in an effective way thanks to the offline method introduced in this work.