Multiantenna Wireless Architectures with Low Precision Converters

Licentiatavhandling, 2022

One of the main key technology enablers of the next generation of wireless communications is massive multiple input multiple output (MIMO), in which the number of antennas at the base station (BS) is scaled up to the order of tens or hundreds. It provides considerable energy and spectral efficiency by spatial multiplexing, which enables serving multiple user equipments (UEs) on the same time and frequency resource. However, the deployment of such large-scale systems could be challenging and this thesis is aimed at studying one of the challenges in the optimal implementation of such systems. More specifically, we consider a fully digital setup, in which each antenna at the BS is connected to a pair of data converters through a radio-frequency (RF) chain, all located at the remote radio head (RRH), and there is a limitation on the capacity of the fronthaul link, which connects the RRH to the baseband unit (BBU), where digital signal processing is performed. The fronthaul capacity limitation calls for a trade-off between some of the design parameters, including the number of antennas, the resolution of data converters and the over-sampling ratio. In this thesis, we study the aforementioned trade-off considering the first two design parameters.
First, we consider a quasi-static scenario, in which the fading coefficients do not change throughout the transmission of a codeword. The channel state information (CSI) is assumed to be unknown at the BS, and it is acquired through pilot transmission. We develop a framework based on the mismatched decoding rule to find lower bounds on the achievable rates. The bi-directional rate at 10% outage probability is selected as the performance metric to determine the recommended architecture in terms of number of antennas and the resolution of data converters.
Second, we adapt our framework to a finite blocklength regime, considering a realistic mm-wave multi-user clustered MIMO channel model and a well suited channel estimation algorithm. We start our derivations by considering random coding union bound with parameter s (RCUs) and apply approximations to derive the corresponding normal approximation and further, an easy to compute outage with correction bound. We illustrate the accuracy of our approximations, and use the outage with correction bound to investigate the optimal architecture in terms of the number of antennas and the resolution of the data converters.
Our result show that at low signal to noise (SNR) regime, we benefit from lowering the resolution of the data converters and increasing the number of antennas, while at high SNR for a practical scenario, the optimal architecture could move to 3 or 4 bits of resolution since we are not in demand of large array gain anymore.