An Inter-Modulation Oriented Learning Based Digital Pre-Distortion Technique via Joint Intermediate and Radio Frequency Optimization

Abstract

Pre-distortion is a key technique to compensate for the nonlinear distortions caused by the transmitter in wireless communication systems. Generally, pre-distortion can be classified into digital pre-distortion (DPD) and analog pre-distortion (APD), which focus on optimizing and assessing the nonlinearity in their own areas. In this paper, we propose a new DPD approach to optimize the performance metric of the analog RF-domain (i.e., inter-modulation distortion (IMD) or adjacent channel power ratio (ACPR)) and that of the digital IF-domain (i.e., mean square error (MSE)) simultaneously. To make the joint design feasible, we derive a new hybrid performance metric, where the analog preferred metric is defined in the form of digital signals to bridge the gap between digital and analog signal processing. On top of that, an effective DPD scheme is developed based on a new dual time-delayed neural network (TDNN) learning architecture. The coefficients of the TDNN for power amplifier (PA) modeling can be trained offline with a PA dataset, while those of the TDNN for pre-distortion are obtained adaptively by optimizing the proposed joint design metric. Experimental results show that the proposed scheme is able to significantly improve the IMD/ACPR performance without compromising the MSE, compared to conventional DPD schemes.