Pseudo-Random Quantization Based Detection in One-Bit Massive MIMO Systems

Abstract

Analog-to-digital converter (ADC) units are one of the most power-hungry devices in the radio-frequency (RF) chains of massive multiple-input multiple-output (MIMO) systems. Therefore, utilizing low-resolution ADCs in uplink massive MIMO systems is a practical solution to decrease power consumption. However, when high modulation orders are employed for high-rate communication, the achievable rate saturates after a finite SNR value due to the stochastic resonance (SR) phenomenon. This thesis proposes a novel pseudo-random quantization (PRQ) scheme that can help compensate for the effects of SR and makes communication with high-order modulation schemes with one-bit quantization possible. The ADC thresholds at the receiver side of uplink one-bit massive MIMO systems are changed to work with the PRQ scheme. We modify linear detectors for one-bit non-zero threshold quantization and propose new detection methods for the frequency-flat and frequency-selective fading scenarios. For flat fading, we offer a two-stage detector that works with PRQ. The first stage is an iterative method called Boxed Newton Detector (BND) that utilizes Newton's method to maximize the log-likelihood. The second stage, Nearest Codeword Detector (NCD), exploits the first stage to create a small set of most likely candidates based on sign constraints to increase performance. For frequency-selective fading, we design a new frequency-domain equalization (FDE) scheme, called the projected quasi-Newton detector (PQND), to optimize the log-likelihood using a quasi-Newton approach that works with PRQ in both orthogonal frequency division multiplexing (OFDM) and single carrier (SC) systems. The proposed methods outperform the existing detectors with comparable complexity.

Date of Award	18 Jan 2023
Original language	English
Awarding Institution	Middle East Technical University (METU)
Supervisor	Ali Özgür Yılmaz (Supervisor 1)