Power Scalable Neural Network Model for Wideband Digital Predistortion

In this letter, a power scalable neural network (PSNN) model for wideband digital predistortion (DPD) is proposed. The proposed model extracts the common behavior characteristic of power amplifier (PA) under different input power conditions offline and updates the specific power behavior characteristic online by least-squares (LSs) algorithm. The PSNN model significantly reduces the complexity of online updating DPD coefficients and preserves the excellent performance of the neural network model in the case of wideband DPD. Experimental results show that the proposed PSNN model can reduce the number of update coefficients by 95% while maintaining very good linearization performance.