Signal processing and bounds for fully digital mmWave architectures with low-resolution converters
Low-resolution analog to digital converters (ADCs) and digital to analog converters (DACs) are the key to power-efficient fully digital massive multiple input multiple output transceivers operating at large bandwidths. The use of low-resolution converters, however, results in severe distortion in the signal model which requires significantly different analysis and signal processing techniques compared to high-resolution systems. In this dissertation, we analyze the performance and design algorithms for sensing and wireless communication systems equipped with low-resolution converters. In the first half of this dissertation, we focus on the sensing application where we consider a fully digital architecture with 1-bit ADCs on each radio-frequency chain. We characterize the effect of the 1-bit ADCs on the radar parameter estimation by the Cramér-Rao bound and show that at low per-antenna signal-to-noise ratios the 1-bit converters result in a loss of 2 dB compared to a system with ideal ∞-resolution ADCs. We then design a low-complexity analog preprocessing unit, realizable through low-cost low-resolution phase-shifters, that reduces the performance gap of the 1-bit system from the ∞-resolution system to 1.16 dB. Our numerical results demonstrate the potential of the proposed architecture to meet the requirements of high-resolution sensing under the low-resolution hardware constraints. In the second half of this dissertation, we focus on the communication application and consider the multi-user (MU) downlink (DL) beamforming (BF) problem under constant envelope quantizer (CEQ) constraints at the base station. We provide a linear precoding based solution to the MU-DL-BF problem by extending the well known uplink-downlink duality principle for ∞-resolution systems to systems constrained by CEQs for both flat and frequency selective channels. Our results show that the proposed formulation significantly outperforms state of the art linear precoding strategies in terms of the ergodic sum rate and ergodic minimum rate. The proposed solution further reduces the performance gap of linear precoding strategies from non-linear precoding methods and actually outperforms them in terms of the coded bit error rate over a wide range of system parameters.