Skip to Main Content
Multi-input multi-output (MIMO) channels have been shown in the literature to present a significant capacity increase over single-input single-output ones in some situations. To achieve this theoretical capacity, the constituent parallel subchannels arising from the MIMO channel have to be properly used. Many practical schemes are being currently developed to achieve this goal. We first show that, from an information-theoretic point of view, beamforming becomes asymptotically optimal as the spatial correlation of the channel fading increases. In light of this result, wideband beamvectors are jointly derived for both transmission and reception. We allow a controlled partial response and design zero-forcing and minimum mean-squared error transmit-receive filters. Conceptually, the beamforming scheme is shown to decompose into two stages: the first one corresponds to a spatial flattening of the MIMO channel, i.e., choosing the subchannel with the highest gain at each frequency; the second stage depends on the particular design criterion and performs a power distribution at the transmitter and defines the equalizer at the receiver. These methods are further extended to the general case of multiple beamforming, i.e., when more than one subchannel are used. An exact and practical implementation of a modified "waterfilling" solution required for the filter design is proposed. All derived methods are assessed and compared in terms of capacity and bit-error rate.