Capacity of wide-band MIMO channels via space-time diversity of scattered fields

In this paper the information capacity of a multiple-input multiple-output (MIMO) Gaussian channel is computed via the theory of non-redundant sampling of scattered fields. Unlike previous works relying on the same approach, it is assumed that communication can occur over a wide frequency band. This leads to a trade-off in the exploitation of diversity, due to the mutual coupling of space and time spectra of the field: maximizing time diversity leads to the impossibility of exploiting space diversity and viceversa. It follows that the ultimate physical capacity limit along both space and time dimensions corresponds either to the maximum achievable information rate of an arbitrarily large MIMO narrowband system, or to the one of a single-input single-output (SISO) ultra-wide band (UWB) system. In contrast, in the practical case of MIMO systems of finite size and finite frequency bands used for communication, space and time diversities can be optimally combined by allocating linearly the signal power in the frequency domain.