Limited Feedback Transceiver Design for Downlink MIMO OFDM Cellular Networks

Feedback in wireless communications is tied to a long-standing and successful history, facilitating robust and spectrally efficient transmission over the uncertain wireless medium. Since the application of multiple antennas at both ends of the communication link, enabling multiple-input multiple-output (MIMO) transmission, the importance of feedback information to achieve the highest performance is even more pronounced. Especially when multiple antennas are employed by the transmitter to handle the interference between multiple users, channel state information (CSI) is a fundamental prerequisite. The corresponding multi-user MIMO, interference alignment and coordination techniques are considered as a central part of future cellular networks to cope with the growing inter-cell-interference, caused by the unavoidable densification of base stations to support the exponentially increasing demand on network capacities. However, this vision can only be implemented with efficient feedback algorithms that provide accurate CSI at the transmitter without overloading the uplink channel. In this dissertation, channel state information feedback algorithms for the downlink of MIMO OFDM cellular networks are proposed and evaluated. After developing the basic mathematical description of the data transmission to a user, the first part of the thesis is concerned with single-user MIMO (SU-MIMO) transmission, i.e., spatial multiplexing of multiple data streams to a single user. SU-MIMO is already incorporated and standardized in current cellular technology, such as 3GPP Long Term Evolution (LTE). The SU-MIMO feedback selection algorithms developed in this thesis are hence designed to optimize the achievable throughput of the transmission while keeping standard-compliance and the corresponding architectural constraints of the physical layer in mind. The performance of the proposed methods is evaluated with extensive link level simulations and by comparison to theoretical bounds on the achievable throughput. In the second part of this dissertation, multi-user MIMO (MU-MIMO) is considered. Here only rudimentary specifications are provided in standards, giving a much larger design freedom. The focus is put on block-diagonalization based MU-MIMO. CSI feedback algorithms are proposed that are based on memoryless and predictive quantization on the Grassmann manifold. With predictive quantization, the temporal correlation of the wireless channel can be exploited to enable high fidelity quantization with a reasonable feedback overhead. A subspace quantization based antenna combiner is proposed for the case that the receivers are equipped with excess antennas. With this combiner a significant improvement in the CSI feedback accuracy can be achieved, reducing the residual multi-user interference. The performance of this combiner is analytically analyzed and contrasted to a strategy that maximizes the channel gain of a single user without considering interference. Finally, the performance of SU- and MU-MIMO is investigated by means of simulations, revealing significant improvements with MU-MIMO provided sufficiently accurate CSI is available at the transmitter. Also, the area spectral efficiency of networking architectures using remote radio units and small cells is compared to the classical macro cellular network, demonstrating valuable capacity gains with remote radio units, which enable joint transmission over the distributed antenna system.