Measurement Based Evaluation of Interference Alignment on the Vienna MIMO Testbed

Most modern wireless multi-user networks suffer from undesired interference that impairs the data transmission over the individual radio links. In order to maximize the data throughput in such systems, several interference mitigation schemes have been investigated recently. Interference alignment stands out as one of the most promising ones, able to attain the maximum data throughput over interference disturbed links in theory, given the right conditions. As the theoretical research progresses, practical implementations have to be considered in order to confirm the theory and discover limitations introduced by hardware. Interference alignment utilizes linear filtering at each transmitter and receiver of the network. The transmit filters thereby partition the signal space at the receiver into two subspaces, a desired signal subspace containing the signal from the desired transmitter and an interference subspace accumulating all the interfering signals. The aligned interference is then forced to zero by the receive filter and only the desired signal is retained. For this to be accomplished, cooperation of all users in the network is required. This work first deals with the theoretical foundations of interference alignment by introducing the relevant system model and discussing feasibility and filter computation. It then advances to the characterization of the Vienna MIMO testbed on which interference alignment was implemented throughout this work. The testbed employs two outdoor transmitters on rooftops, one indoor transmitter and one indoor receiver. The radio channels in the considered setup are quasi-static. Hardware, software and the used signals are described. Performance measures for evaluation are introduced. Finally, measurement results are presented. The feasibility of interference alignment is shown, and the performance measures are evaluated over variable signal to noise ratio and variable signal to interference ratio. The results are discussed, and impairments introduced by hardware are highlighted.