Throughput of Realistic Multi-User MIMO-OFDM Systems Bartosz Mielczarek and Witold Krzymie´ n TRLabs / University of Alberta, Edmonton, Alberta, Canada Email: {bmielcza, wak}@trlabs.ca Abstract— In this paper we present an architecture for a real- istic multi-user multi-antenna system operating on a frequency selective fading channel. We argue that it is highly unlikely that the base station will have access to full channel state information and we design our system accordingly. By splitting the signal processing burden between the base station and the individual receivers, we take into consideration possible upgrades of the system and likely heterogeneity of the receivers. The proposed system divides the available transmission band- width into sub-bands and uses layered MIMO approach to transmit information to all users, which in turn employ some form of interference cancellation. Based on simulations we show the influence of different system parameters on upper bounds on the downlink throughput of the system. Index Terms— Multiple-input multiple-output, orthogonal fre- quency division multiplexing, multi-user diversity. I. I NTRODUCTION The development of the modern Internet-based data com- munication systems and ever increasing demand for bandwidth have spurred an unprecedented progress in fixed network capacity, enabling cabled users to access the network at bit rates well in excess of 100 Mbits/s. The development of wireless systems that can provide such high transmission rates has yet to catch up with that progress. Current trends in wireless system design focus on the use of multiple antennas to provide capacity gains on fading channels [1], [2], orthogonal frequency division multiplexing (OFDM) to facilitate the utilization of these capacity gains on frequency- selective channels [3], [4], spread spectrum techniques to enhance robustness to multiuser interference [5] and powerful channel codes (such as turbo codes [6]) to combat both high interference and noise levels. There is a very large body of research on single user multiple-input multiple-output systems (MIMO) in different scenarios and it can be safely stated that such systems are in general well understood. The use of MIMO in multi-user systems has also attracted interest of the research community and resulted in many papers (a very good overview of relevant issues and methods can be found in [7]). There are, however, certain problems that still remain to be solved, especially in the case of broadcast channels (the downlink). The optimal way of operating in such an environment is the so called dirty paper coding [8]. However, due to its complexity and requirement of perfect channel knowledge at the transmitter suboptimal approaches are preferable. In this paper we aim at showing the upper bounds on throughput of realistic multi-user systems operating on fre- quency selective channels. First, we try to predict how the future systems are likely to be implemented. We treat the downlink capacity as most important since, even with the advent of peer-to-peer communication systems, it is likely that in the foreseeable future the wireless communication networks will have to handle higher traffic loads on the downlink. We assume that the base station will not have a full channel state information from all its users. Moreover, we argue that even though the receivers may have multiple antennas, to keep complexity low, the system should treat every receiver as if it had only one antenna. If the receivers have more antennas, they can use them to perform some form of interference cancellation without burdening the base station with the exact parameters of their algorithms (see [9], [10], [11]). Using simulations on a realistic channel model, we determine how different design parameters (number of antennas, number of users, different interference cancellation algorithms) influence the expected throughput of the system. In Section II, we present the detailed assumptions for the proposed system. The operation of the receiver is analyzed in Section III and Section IV discusses the scheduling at the base station that takes into account the specifics of the proposed approach. In Section V, we discuss the simulation results and compare them with theoretical upper bounds on capacity (cooperative MIMO systems). We conclude the paper and discuss directions of future research work in Section VI. II. SYSTEM MODEL A. Main system assumptions The system discussed in this paper consists of one base station with N antennas and K users equipped with n (k) ∈ {1, 2, 4},k =0, 1, ...K - 1, antennas each. The signal from the base station is transmitted using OFDM format and any form of PSK/QAM modulation. The number of sub-carriers N OFDM , the guard interval T g and the OFDM symbol duration T OFDM are adjusted to the known frequency selective channel model with the maximum Doppler shift f D so that no inter-carrier interference (ICI) is present. All users are assumed to receive the same mean power P on each OFDM sub-carrier (the total transmission power is equally divided between each base station antenna and sub-carrier) and the noise variance is assumed to be 1.