Mutual coupling among receive antennas can profoundly reduce capacity in multiple-input, multiple-output (MIMO) systems. Prior work has shown that these losses can often be significantly reduced by using sophisticated antenna matching at the receiver. However, previous studies have focused exclusively on frequency non-selective matching, and rely on assumptions that are valid only for small bandwidths. In this paper, we consider the capacity of broadband MIMO-OFDM systems with coupled receive antennas. We present upper and lower bounds on the best capacity that can be achieved by physically realizable, frequency-selective antenna matching. The upper and lower bounds coincide for most antenna configurations considered in the examples, and thus determine the performance of capacity-optimal matching. These results suggest that frequency-selective matching can significantly increase capacity in the presence of strong coupling. Moreover, these gains can be achieved by a simple class of matching networks that are easily realized in hardware.

Mutual coupling among receive antennas can profoundly reduce the ergodic capacity of MIMO systems. Prior work has shown that these losses can often be mitigated by using multi-port antenna matching at the receiver; however, previous studies have focused exclusively on frequency-nonselective matching and rely on assumptions which are valid only for small bandwidths. In this paper, we revisit the ergodic capacity of a MIMO system with coupled receive antennas for a non-negligible bandwidth. In particular, we present upper and lower bounds on the best possible MIMO capacity that can be achieved by physically-realizable, frequency-selective (FS) multi-port and single-port matching networks. These results suggest that FS matching can significantly increase capacity in the presence of strong coupling, for both single-port and multi-port scenarios. Moreover, we show that these gains can be achieved for a 2 × 2 MIMO system by a simple class of matching networks that are easily realized in hardware.