Distributed Massive MIMO with Low-Resolution ADCs: Enhancing Efficiency through Deep Learning

Abstract

Abstract Distributed massive MIMO, including a central unit (CU) and a large number of spatially distributed antennas, provides more uniform quality of service (QoS) than co-located massive MIMO systems. One of the components used in distributed massive MIMO is the analog-to-digital converters (ADC). However, high-resolution ADCs consume a considerable power. Having a simple structure and a very low power consumption, the low-resolution ADCs, can be used to decrease both the complexity and power consumption. However, using such ADCs, introduces non-linear distortions in the received signals, thus, complicating channel estimation and data detection at the receiver. In this study, a distributed massive MIMO case with one-bit radio-over-fiber fronthaul has been studied where model-driven deep learning structures are used to compensate for the non-linear distortion caused by the low-resolution ADCs used in the communication system. The aim is to improve both channel estimation and data detection in the uplink phase.