The past decade has seen a significant amount of research in the physical layer of multiple-input-multiple-output (MIMO) systems and rapid adoption of MIMO technologies in wireless systems and standards. MIMO systems are significantly superior to its single antenna counterparts, both in the channel capacity and in the link reliability. The difficulty in achieving the gain brought by the increase in spatial dimension in MIMO systems, however, lies in the implementation complexity....[ Read more ]

The past decade has seen a significant amount of research in the physical layer of multiple-input-multiple-output (MIMO) systems and rapid adoption of MIMO technologies in wireless systems and standards. MIMO systems are significantly superior to its single antenna counterparts, both in the channel capacity and in the link reliability. The difficulty in achieving the gain brought by the increase in spatial dimension in MIMO systems, however, lies in the implementation complexity.

One obvious problem is the extra hardware cost. Deploying multiple antennas at both the transmitter and receiver requires multiple sets of expensive radio-frequency (RF) chains including mixers, analog-to-digital converters and amplifiers which dominate the hardware cost. Antenna selection, which uses fewer RF chains than the number of antennas, emerges as an effective solution to realize most of the advantages of MIMO systems with lower cost, lower power and lower hardware complexity. In the first part of this thesis, we study a space-time coded system with joint transmit and receive antenna selection based on outdated channel state information at the transmitter (CSIT). In most of the existing research work, either the receiver or the transmitter uses the estimated CSIT to perform both transmit and receive antenna selections. However, without the knowledge of when the next packet will be transmitted, the derived transmit antenna selection may no longer be optimal at the time of the next transmission, since the channel may have changed. Our work takes into consideration this delay effect and provides a selection algorithm based on both the estimated CSIT and its statistical information. The solution can be adapted to cover the complete spectrum of CSIT quality, ranging from statistical CSIT only to perfect CSIT.

Signaling in a higher spatial dimension also increases the signal processing complexity, in particular, the likelihood ratio estimation algorithm which grows exponentially in the number of transmit antennas. In a MIMO system with channel coding, accurate soft bit likelihood ratio estimates (BLRE) are essential for good performance. Reduced complexity maximum likelihood detection (MLD) and soft BLRE are thus of great interest. State-of-the-art reduced complexity MLD includes sphere decoding and list decoding. We propose a two-stage approach where the first stage pre-processing depends only on the channel information while the second stage MLD and BLRE depends on each received symbol. After the per-channel pre-processing is completed, the search space of the per-received-symbol second stage processing is significantly reduced. Moreover, both stages can take advantage of the lattice structure of the MQAM structure to have further complexity reduction. This approach is simulated with an IEEE 802.11n MIMO frequency selective fading channel and its performance is very close to the optimal scheme.