Three algorithms to exploit the multi-user MIMO channel are examined and their system capacity evaluated in the uplink of a small urban cell scenario using measured channel data. An iterative scheme to determine transmit weights that achieve the system-wide Nash equilibrium (NE) is compared to schemes achieving block-diagonalization (BD) of the overall system channel matrix and successive diagonalization (SD) among users. The BD scheme is shown to noticeably outperform the NE at high SNR - by as much as 9.7 bps/Hz (44%) at 20 dB SNR, whilst the SD yields as little as 33% of the capacity of BD. SD is, however, completely non-iterative which could be important in rapidly changing environments. The NE is found to sometimes yield a higher system capacity with more users; the setup examined here shows a rise of about 2 bps/Hz in the system capacity when going from 3 users to 4. We also find that NE converges quickly, usually in fewer than 6 iterations, and reliably, more than 99% of the time, but BD can have serious slow-convergence problems.
Bibliographical noteAdditional information: With accompanying conference presentation
Sponsorship: The presenters gratefully acknowledge the financial support of EPSRC and Philips
Contributor (Other): IEE Communication Networks and Services Professional Network