In the multi-user MIMO uplink channel, the presence of multiple receive antennas enables the base station to serve a number of users simultaneously, thus increasing overall system throughput. In practice, users have to be scheduled instead of being served all at once. In this work, a novel scheduling metric is proposed based on ergodic sum capacity. A low-SNR Taylor series of the capacity expression is combined with a receiver-correlation channel model to obtain a robust measure of spatial compatibility. Long-term channel knowledge at the base station is sufficient for metric evaluation, which significantly reduces the amount of required channel sounding, compared to previously proposed schemes.

In this paper, the multi-user MIMO downlink channel is considered, where a single transmitter (base station) sends data to several users within the same resource unit. Assuming perfect channel state information (CSI) at the transmitter, different linear precoding schemes are studied, aiming at mean square error minimizing transmission to all users. The exact MMSE solution can be obtained by exploiting uplink/downlink duality. A suboptimal solution based on necessary conditions for MMSE optimality may also be computed directly in the downlink domain. Our results indicate that the performance difference between the direct and duality-based method that is visible in uncoded BER curves vanishes when channel coding is introduced to the system. We also compare the MMSE-optimal approaches to other existing linear precoding schemes (BD, S-MMSE)

In this paper, we consider frequency-selective coherent multiple-input multiple-output (MIMO) multiple-access fading channels. Assuming that each of the users employs orthogonal frequency-division multiplexing (OFDM), we introduce a multiple-access scheme that gradually varies the amount of user collision in signal space by assigning different subsets of the available OFDM tones to different users. The corresponding multiple-access schemes range from frequency-division multiple access (FDMA) (each OFDM tone is assigned to at most one user) to CDMA (each OFDM tone is assigned to all the users). We quantify the effect of signal space collision between the users by computing the ergodic capacity region for the entire family of multiple-access schemes. It is shown that the ergodic capacity region obtained by a fully collision-based scheme (CDMA) is an outer bound to that corresponding to any other multiple-access strategy. In practice, however, minimizing the amount of user collision in frequency is desirable as this minimizes the receiver complexity incurred by having to separate the interfering (colliding) signals. Our analysis shows that the impact of collision on spectral efficiency depends critically on the channel's spatial fading statistics and the number of antennas

This paper gives an overview of some so-called link performance models used in system level simulations to determine the link packet error rate (PER) at reduced complexity. A subset of link performance models is evaluated in terms of PER prediction accuracy focusing on a single receive and transmit antenna OFDM link with different coding options and channel characteristics. The results demonstrate that a mutual-information based metric which accounts for the modulation alphabet is preferable in the considered cases and, furthermore, applicable to the large class of MIMO-OFDM transmission techniques with linear pre- and post-processing.

We consider frequency selective Multiple-Input Multiple-Output (MIMO) multiple access fading channels with perfect channel knowledge in the receiver and no channel state information in the transmitters. Assuming that each of the users employs Orthogonal Frequency Division Multiplexing (OFDM), we introduce a multiple access scheme which allows to gradually vary the amount of user collision in signal space by assigning different subsets of the available OFDM tones to different users. The corresponding multiple access schemes range from FDMA (each OFDM tone is assigned to at most one user) to CDMA (each OFDM tone is assigned to all the users). We quantify the effect of signal space collision between users by computing ergodic capacity regions for joint and single user decoding. In the case of joint decoding, we prove that irrespectively of the spatial receive fading correlation, the ergodic capacity region obtained by a fully collision-based scheme is an outer bound to any other OFDM-based multiple-access strategy, where the users collide only on subsets of the available tones. For single user decoding, we find that the amount of collision maximizing the capacity region depends critically on the number of transmit and receive antennas and the users' receive fading correlation.

For a transmitter that has a perfect knowledge of the MIMO channel, the maximum achievable capacity corresponds to the waterfilling solution. In practice, the available knowledge may only be partial due to the time selectivity of the channel, and the delay or absence of feedback from the receiver. However, exploiting the partial knowledge leads to a significant improvement when compared to the capacity without any channel knowledge. We analyze the MIMO capacity with various types of partial knowledge of the channel under practical frequency flat channel models.