The Effect of Computation and Feedback Delay on the Capacity of Multiuser MIMO Systems in a Small Outdoor Cell

This paper examines how the capacity of three multiuser multiple-input multiple-output (MIMO) algorithms is affected by a delay in computing and feeding back the transmit (and receive) weighting matrices used by the algorithms. Iterative schemes determining transmit weights that achieve the system-wide Nash equilibrium (NE) or block-diagonalization (BD) of the overall system channel matrix are compared as is a scheme applying successive diagonalization (SD) among users. The NE is found to be far the most robust to out-of-date channel information, whilst SD is found to suffer badly at even the shortest delay considered here (about 0.9 ms); BD's losses vary over a wide range depending on the delay. For the two iterative algorithms we also examine the tradeoff between better convergence and less delay. In both cases, the loss in capacity for weaker convergence is very small, and so it is desirable to minimize the delay and thus the capacity loss caused by it