Experimental Validation of Interference Alignment Techniques using a Multiuser MIMO Testbed ´ O. Gonz´ alez, D. Ram´ ırez and I. Santamaria Dept. of Communications Engineering University of Cantabria 39005 Santander, Spain Email: {oscargf, ramirezgd, nacho}@gtas.dicom.unican.es J. A. Garc´ ıa-Naya and L. Castedo Dept. of Electronics and Systems University of A Coru˜ na 15071 A Coru˜ na, Spain Email: {jagarcia, luis}@udc.es Abstract—Hardware platforms and testbeds are an essential tool to evaluate, in realistic scenarios, the performance of wireless communications systems. In this work we present a multiuser Multiple-Input Multiple-Output (MIMO) testbed made up of 6 nodes, each one with 4 antennas, which allows us to evaluate Interference Alignment (IA) techniques in indoor scenarios. We specifically study the performance of IA for the 3-user interfer- ence channel in the 5 GHz band. Our analysis identifies the main practical issues that potentially degrade the IA performance such as channel estimation errors or collinearity between the desired signal and interference subspaces. Index Terms—MIMO testbed, interference channel, interfer- ence alignment. I. I NTRODUCTION Interference alignment (IA) has recently emerged as an attractive transmission technique for the K-user interference channel [1]. Signals transmitted by all users are designed in such a way that the interfering signals at each receiver fall into a reduced-dimensional subspace. The receivers can then apply an interference-suppression filter to project the desired signal onto the interference-free subspace and, hence, the number of interference-free signalling dimensions of the network is substantially increased. To better understand the impact of IA techniques on practi- cal wireless networks, it is important to evaluate their perfor- mance in real-world scenarios rather than on simplistic channel models often used in simulation-based approaches (e.g. spa- tially uncorrelated channels, perfect synchronization among users, ...). However, experiments for the K-user Multiple- Input Multiple-Output (MIMO) interference channel require an extremely complex set-up made up of 2K MIMO terminals (K transmitters and K receivers). Complexity is considerably larger than in point-to-point MIMO links or in other multiuser MIMO scenarios such as the broadcast multiple-access chan- nels. For instance, the simplest 3-user interference network requires six MIMO nodes with at least two antennas per node. The cost associated to the required hardware set-up explains why the experimental evaluation of IA techniques reported in the literature is scarce, so far. To the best of our knowl- edge, the first experimental work on IA was presented in [2] where a technique that combines Interference Alignment and Cancellation (IAC) was implemented in a testbed made up of 20 GNU Radio Universal Software Radio Peripheral (USRP) nodes with two antennas each. Several practical issues have been evaluated in this work, such as the impact of different modulations, frequency/time synchronization aspects, or the idea of applying the alignment at the sample level (i.e., before timing and frequency offset correction). However, this technique requires an additional wired Ethernet connection to transfer already decoded packets between access points in order that some streams be cancelled. Another experimental study on IA has been recently pre- sented in [3]. Specifically, this work focuses on MIMO Or- thogonal Frequency Division Multiplexing (MIMO-OFDM) 3- user interference channels measured in indoor and outdoor scenarios. The set-up comprises five National Instruments [4] PXI-1045 chassis connected to 3 PCs. Using the measured channels, IA techniques were evaluated in an off-line fashion. This work validates the feasibility of IA techniques and also evaluates the performance degradation under spatially correlated channels. However, notice that in [3] no aligned streams are actually transmitted over the wireless channel and, thus, many practical issues such as synchronization or hardware impairments are not taken into account. In this paper we will present new IA experiments performed in the 5 GHz band. Specifically, we have focused on indoor 3-user interference wireless channels where each user sends one stream and is equipped with two antennas at both sides of the link. For this scenario, which is typically denoted as (2 × 2, 1) 3 , closed-form IA solutions exist and can be obtained solving an eigenvalue problem. The conducted experiments use specifically designed frames and comprise two steps: i) a training step, where all pairwise interference channels are estimated and the corresponding IA precoders are obtained; and ii) a payload transmission step, where all users transmit concurrently using their IA precoders as well as a sequential Time Division Multiple Access (TDMA) mode that is used as a benchmark for comparison purposes. This paper is organized as follows. Section II briefly presents the signal model and the IA technique, including the IA solution for the (2 × 2, 1) 3 scenario. Section III describes the most important practical issues that can affect the IA real-world performance. The testbed hardware used for the experiments is described in Section IV, while the measure- ment set-up and methodology is described in Section V. The