An Inmproved Digital Predistortion for Model Simplification of Power Amplifiers

Abstract

The paper introduces a new segmented dual-feedback digital predistortion (DPD) construction for the front-end of a broadband power amplifier (PA). Different from the traditional indirect learning architecture of DPD that directly estimates the post-inverse model of the PA, the proposed architecture firstly simplifies DPD model through the optimal basis set selection module based on compressed sensing (CS). Then the model coefficients are estimated through the behavior model identification algorithm and replicated to the DPD. It can reduce model parameters while reducing the computational complexity and improving the stability of the system to achieve the effect of simplifying the model of the PA. Finally, the effect of improving the linearization of the PA is achieved. To achieve our proposed structure, we use a 20 MHz long-term evolution (LTE) signal to drive a 35 dBm Class-F power amplifier. The experimental results show that the adjacent channel power ratio (ACPR) and the normalized mean squared error (NMSE) is significantly improved compared with no DPD.