This paper proposes an architecture for using multiple-input multiple-output techniques for a multimode fiber broadcast channel, allowing simultaneous transmission of unique streams to different users on the same fiber while using square-law detection. The resulting system throughput scales nearly linearly with the number of transmitters and receivers. The paper also proposes a training scheme appropriate for use with square-law detection.

Multiple-input multiple-output (MIMO) communication systems can theoretically use array superdirectivity to optimally exploit the propagation channel. This superdirectivity leads to very large capacity bounds, particularly when the antennas are closely spaced. However, because superdirective behavior is difficult to achieve in practice, new capacity bounds have been formulated for the case where the level of array superdirectivity is constrained. Existing capacity analyses limit the superdirectivity in a suboptimal way. In this paper, the impact of superdirectivity in the capacity solution is limited by introducing finite ohmic loss in the transmit and receive antenna elements. Computational results reveal that even a small amount of ohmic loss significantly decreases the achievable system capacity