Highly non‐linear and wide‐band mmWave active array OTA linearisation using neural network

Abstract

This paper proposes a neural network (NN)-based over-the-air (OTA) linearisation technique for a highly non-linear and wide-band mmWave active phased array (APA) transmitter and compares it with the conventional memory polynomial model (MPM)based technique. The proposed NN effectively learns the distinctive non-linear distortions, which may not easily fit to existing MPM solutions, and can, therefore, successfully cope with the challenges introduced by the high non-linearity and wide bandwidth. The proposed technique has been evaluated using a state-of-the-art 4 � 4 APA operating in highly non-linear regions at 28 GHz with a 100-MHz-wide 3GPP base-station signal as input. Experimental results show the pre-distortion signal generated by the NN exhibits the peak-to-average power ratio (PAPR) much lower than the one generated by MPM and consequently superior linearisation performance in terms of adjacent channel leakage ratio (ACLR) and error vector magnitude (EVM) for high non-linearity cases. Using the proposed NN-based linearisation technique, an improvement of 5-dB ACLR and 7% points in EVM are achieved, which demonstrates the promising potential of this technique for emerging broadband communication systems such as 5G/6G and low Earth orbit (LEO) satellite networks. K E Y W O R D S 5G mobile communication, active antenna arrays, learning (artificial intelligence), linearisation techniques, millimetre wave amplifiers, millimetre wave antenna arrays, satellite communication 1 | INTRODUCTION Recent wireless communication systems operating at mmWave are using active phased array (APA) transmitters together with multi-input multi-output (MIMO) systems to improve the system capacity and data rates of the wireless networks. The traditional linearisation techniques like digital pre-distortion (DPD) that are mainly based on the memory polynomial model (MPM) are facing new challenges [1] such as the following.� Highly integrated front-ends and a large number of PA's do not allow placement of feedback circuits for each branch so a single-input single-output (SISO) over-the-air (OTA) model is needed.� mmWave frequencies and wide-band linearisation must be handled. � The increasing desire for high power efficiency requires linearisation of power-efficient but highly non-linear APAs.Several solutions using a SISO model and modified DPD algorithms for linearisation have been proposed to combat the above challenges [2][3][4][5][6]. A SISO modelling where the entire transmitter is considered as a two-port system has been presented in [7-10] using an observation receiver in far-field together with using an MPM-based DPD technique for linearisation of the antenna array in the presence of crosstalk. In [11], the potential mismatches between PAs have been This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Bandwidth-Scalable Digital Predistortion of Active Phased Array Using Transfer Learning Neural Network

“…The block diagram of the OTA measurements setup using a compact antenna test range (CATR) is shown in Fig. 3 [20] and the actual laboratory setup is in Fig. 4 The output signal from the pre-amplifier is highly linear and the signal power is sufficient to drive the 4×4 APA, Amotech AAiPK428GC-A0404 [28], close to its saturated region.…”

A Comparison of Surrogate Behavioral Models for Power Amplifier Linearization under High Sparse Data

“…Despite existing works, the need for systems with accurate data analysis, ability to perform automated modeling/linearization processes, and big data management is still present, and in this regard, machine learning (ML) represents an efficient proposal. Artificial neural networks have been used in over-the-air (OTA) linearization for a highly nonlinear and wide-band mmWave active phased array (APA) transmitter, and results indicate that they are capable of learning distinctive nonlinear distortions, which may not easily fit to existing polynomial-based solutions [ 8 ].…”