Name: Rana A. Abdelaal
Date: January 30, 2017
Time: 1:00 PM
Location: Engineering Hall 3106
Committee: Professor Ahmed Eltawil
Multi Input Multi Output (MIMO) technology has seen prolific use to achieve higher data rates and an improved communication experience for cellular systems. However, one of the challenging problems in MIMO systems is interference. Interference limits the system performance in terms of rate and reliability. In this thesis, we analyze methods that provide high performance over interference-limited wireless networks such as Long Term Evolution (LTE) and WiFi. In this thesis, we tackle different sources of interference. One of the interference sources is the neighboring interference, we propose methods that include an optimized solution that models the interference as correlated noise, and uses its statistical information to jointly optimize the base station precoding and user receiver design of LTE systems. We study the benefits of exploiting interference in terms of both probability of error and signal-to-noise ratio (SNR). In addition, we compare the proposed method with the conventional beamforming and maximum ratio combining (MRC). One of the key challenges to enable high data rates in the downlink of LTE is the precoding and receiver design. We focus primarily on the UE and the base station (BS) processing, particularly on estimating and using the interference resulting from neighboring stations. In this thesis, we propose a receiver design that performs well in the presence of interference. Furthermore, we present a precoding scheme that the BS can use to maximize the signal-to-interference plus noise-ratio (SINR). An interference free scenario is used as a benchmark to evaluate the proposed system performance. In this thesis, we optimize the performance of LTE by tackling practical considerations that affects the system performance. We present a suboptimal practical way of estimating the interference and utilizing this information on the processing techniques used at both the UE and the eNodeB sides. We focus on managing both MU-MIMO interference and other cell interference. The proposed study improves system performance even under non-perfect channel knowledge, enabling the throughput gains promised by MU-MIMO.
Other types of interference exist in In-band full-duplex (IBFD) communication systems. IBFD is very promising in enhancing wireless LANs, where full-duplex access points (APs) can support simultaneous uplink (UL) and downlink (DL) flows over the same frequency channel. One of the key challenges limiting IBFD benefits is interference. In this thesis, we also propose a scheduling technique to manage interference in wireless LANs with full-duplex capability. We focus primarily on scheduling UL and DL stations (STAs) that can be efficiently served simultaneously.
It is very important to apply system knowledge to reduce power and/or improve performance. Thus, we also aim at exploring energy and power management techniques for practical wireless communication systems. An important topic for practical communication systems is handling the interference due to the power amplifier nonlinearities. Managing this type of interference is of very high importance, especially in Orthogonal Frequency-Division Multiple Access (OFDMA) based communication systems. Although, OFDMA is the modulation of choice due to its robustness to time-dispersive radio channels, low-complexity receivers, and simple combining of
signals from multiple transmitters in broadcast networks. The transmitter design for OFDMA is more costly, as the Peak-to-Average Power Ratio (PAPR) of an OFDMA signal is relatively high, resulting in the need for highly linear RF power amplifiers (PA). This problem becomes more compounded when a large number of PAs is required, as in Massive MIMO for example. In this thesis, we discuss the impact of PAs on cellular systems. We show the constraints that PAs introduce, and we take these constraints into consideration while searching for the optimum set of transmitter and receiver filters. Moreover, we highlight how Massive MIMO cellular networks can relax PAs constraints resulting in low cost PAs, while maintaining high performance. The performance is evaluated by showing the probability of error curves and signal-to-noise-ratio curves for different transmit powers and different number of transmit antennas. Another promising topic for efficient practical communication systems is Associative processors (APs). AP is a good candidate for in-memory computation, however it has been deemed too costly and energy hungry in the past. The advent of ultra-dense resistive memories is changing this paradigm, allowing for efficient in-memory associative processors. However, with high levels of integration, issues related to power density become the bottleneck. In this thesis, we show the potential use of the approximate computing in wireless communication systems, specifically, we present Fast Fourier Transform (FFT) implemented by associative processor. Results confirm that approximate computing for in-memory associative processors is a viable approach to reduce power consumption while maintaining good performance. A promising approach to save energy is through reducing the bit width, however reducing the bit width introduces errors that may affect the performance. In this thesis, our goal is to adjust the bit width based on the channel SNR, aiming at achieving good performance at reduced energy consumption. The mathematical approach that analytically describes the system performance under the reduced bit width noise is presented. Based on this model, an adaptive bit width adjustment algorithm is presented that utilizes the received SNR estimates to find the optimal bit width that achieves performance goals at reduced energy consumption. Simulation results show that the proposed algorithms can achieve up to 45% energy savings as compared to wireless communication systems with conventional FFT.