Massive MIMO wireless systems under an RF chains constraint

Massive MIMO is an exciting new technology which is revolutionizing the wireless community. Massive MIMO serves multiple users with increased data rates using a massively large number of antenna elements at the base station using simple beamforming techniques. In digital beamforming, each antenna element is connected to a dedicated radio-frequency chain. With so many antenna elements, it is not practical to have a dedicated RF chain for each antenna element due to high fabrication cost and power consumption. We start by considering the pilot contamination problem in which inter-cell interference occurs because the channel estimate of a user sending a particular training sequence in one cell is corrupted by the transmissions of the users using the same training sequence in other cells. To mitigate pilot contamination, we consider an adaptive least-squares algorithm in massive MIMO which employs bi-directional training to optimize precoders and receive filters without doing channel estimation. We characterize the impact of using different types of training sequences on sum-rate performance of a multi-cell system. We demonstrate analytically that usage of random training sequences across cells, offers better performance than usage of identical training sequences for a fixed number of interferers which is critical because it avoids having to synchronize the users across cells. The thesis next considers the challenges arising from the the limited number of RF chains. There are a variety of approaches to tackle the problem of RF chains constraint. The first approach is to study generalized spatial modulation (GSM) in which RF chains are connected to a subset of antennas. We propose novel compressive sensing (CS) aided detection algorithm which offers superior performance to existing algorithms. For imperfect channel state information, we use CS in conjunction with total least-squares to mitigate the effect of contaminated channel estimates. Our CS framework is premised on the assumption of frequency-selective fading for GSM to account for high data-rate applications. We derive the achievable rates for GSM and provide closed-form expressions for upper and lower bounds of the achievable rates to investigate the capacity gains offered by GSM under finite alphabet constraint. The second approach we take is to investigate the antenna selection for massive MIMO whereby the limited number of RF chains are connected to the best subset of antennas. We derive a theorem which provides a concise formula for the limiting normalized channel power gain in the large system dimensions. Finally, the application of massive MIMO in millimeter-wave (mmWave) band under RF chains constraint is analysed. The performance bottleneck in mmWave massive MIMO is beam-alignment due to narrow beams. We propose beam alignment algorithms under RF chains constraint to improve upon existing schemes. We derive theorems which provide closed-form expressions for probability of beam-misalignment of energy and Bayesian detectors for analogue beam-steering. Moreover, we perform asymptotic analysis for Bayesian detector and demonstrate that probability of beam misalignment tends to zero in the limit of infinite numberof BTS antennas under RF chains constraint. The results demonstrate that Bayesian detection offers superior performance over energy detection in terms of mmWave beam-alignment which is imperative because it minimizes beam pointing losses.