Abstract
High data rate communications are limited not only by noise but also by inter-symbol interference (ISI) due to the memory of the dispersive wireless communication channel. Except the conventional channel equalization techniques are used to suppress the ISI caused by channel, an multi-carrier approach, e.g., orthogonal frequency division multiplexing (OFDM), towards transmitting data over a multipath channel also allow us to design a system supporting high data rate. Combining OFDM transmissions with code division multiple access (CDMA) exploit the wideband channel’s inherent frequency diversity by spreading each symbol across multiple sub-carriers. The combination has two major advantages. One is its own capability to lower the symbol rate in each subcarrier enough to have a quasi-synchronous signal reception in uplink. The other is that it can effectively combine the energy of the received signal scattered in the frequency domain. That is, it is possible to prevent the obliteration of certain sub-carriers by deep frequency domain fades. This is achieved by spreading each sub-carrier’s signal with the aid of a spreading code and thereby increasing the achievable error-resilience. In this thesis, we analyze the bit-error-rate (BER) performance of the optimum multiuser detection (MUD) with channel mismatch in multi-carrier code-division-multiple-access (MC-CDMA) systems. However, it is NP-hard to implement an optimum MUD algorithm. To justify the BER performance and to make the optimum MUD feasible, based on Pearl’s belief propagation (BP) scheme, we put together a low-complexity iterative MUD algorithm for MC-CDMA systems. On the other hand, systems that employ multiple antennas in both the transmitter and the receiver of a wireless system have been shown to promise extraordinary spectral efficiency. One way to realize the enormous throughput is to exploit the spatial multiplexing (SM) gain by transmitting several data streams across the wireless MIMO channel simultaneously. Unfortunately, an SM system is sensitive to the rank of its MIMO channel matrix. To prevent ill-conditioned MIMO channel matrix from affecting the system data throughput, e.g., in the downlink scenarios where the number of the transmit antennas is larger than the number of the receive antennas and the corresponding MIMO channel matrix has a non-empty null space, a transmit spatial pre-filtering scheme should be designed to feed the simultaneously-transmitted data streams into the signal space of the MIMO channel matrix, instead of wasting the transmit power in the null space of the MIMO channel matrix. However, the efficiency of the spatial pre-filtering scheme highly depends on the availability of the MIMO channel state information (CSI), which can be estimated at the receiver. Therefore, feedback of sufficiently reliable CSI from the receiver to the transmitter is crucial, especially in the downlink scenarios. Hence, in this thesis, we also focus on the development of efficient coding schemes for channel feedback in downlink scenario, which expects a higher data throughput and is considered the bottleneck in a MIMO system. The thesis contains Four results. First, we analyze the bit-error-rate (BER) performance of the optimum multiuser detection (MUD) with channel mismatch in MC-CDMA systems. To justify the BER performance and to make the optimum MUD feasible, based on Pearl’s belief propagation (BP) scheme, we put together a low-complexity iterative MUD algorithm for MC-CDMA systems. Furthermore, channel mismatch is introduced into the BP-based MUD algorithm to make the scenario general. With channel mismatch, the analytical results of the BP-based MUD algorithm conform perfectly to, and the simulation results of the BP-based MUD algorithm conform very closely to the BER performance of the optimum MUD derived using the replica method, which is a non-trivial extension of the existing replica approach mentioned above. Without channel mismatch, the problem becomes a special case of our contribution. Second, Raleigh and Cioffi proposed a singular-value-decomposition based space-time architecture for multiple-input-multiple-output (MIMO) wireless systems using antenna arrays, discrete matrix multi-tone (DMMT) coding scheme, which claims to achieve near-optimum performance in both signal diversity and channel capacity. However, the DMMT coding scheme suffers high computational complexity in non-stationary channel environments, where channel information update is frequently needed at both the transmitter and the receiver. In addition, the transmitter may need to rely on a wideband feedback channel to obtain the entire set of vector channel information, which is obviously impractical. By exploring the MIMO channel structures, we develop an adaptive version of the DMMT coding scheme for a high-capacity MIMO system with time-varying frequency-selective channels. In the proposed coding scheme, a lowv complexity Jacobi-SVD is utilized to iteratively optimize the transmit signaling by tracking merely the dominant fading paths in the MIMO wireless channel, while only very little feedback information is required. An analytic capacity lower bound considering channel-tracking errors is derived for systems employing the proposed coding scheme. Simulation results reconfirm that the proposed coding scheme works efficiently in indoor wireless applications. Third, for MIMO-OFDM wireless systems, gain in channel throughput educed through sufficient feedback of the CSI is significant, particularly when the number of transmit antennas is larger than the number of receive antennas. In this part, we demonstrate that, in such scenarios, 1) the CSI of each OFDM sub-carrier can be parameterized into a short bit stream by a proposed low-complexity QR decomposition on the corresponding MIMO channel matrix, 2) the overall CSI can be reliably represented by a proposed parameter interpolation on the above bit streams of only a fraction of sub-carriers, and 3) a MIMO-OFDM system with a low-rate CSI feedback parameterized above can provide a channel throughput comparable to the channel capacity. Finally, we propose a novel scheduling mechanism to enhance the throughput in a multiuser MIMO system. As it is known, based on the information theory, that a wireless system with antenna array at both sides of a communication link is able to achieve excellent spectral efficiency. For multiuser multiple-input-multiple-output(MU-MIMO) services, orthogonal multiple accesses, e.g., frequency division multiple access (FDMA) and time division multiple access (TDMA) are popular options to avoid multiuser interference. However, for the FDMA (or TDMA) system, the spatial resources of a frequency band (or time slot) are consumed by a single user. The spatial utility of such a system is very low if it happens to have a common clustered channel structure. A high sum-rate is achievable in an MU-MIMO system where a common frequency or time resource is shared by multiple users if transmitters assume perfect knowledge of the corresponding channels. However, in order to reach this capacity, existing coding schemes suffer not only from high computational complexity but also from the need for excess channel state information (CSI) feedback. It is shown in the previous work that the ergodic capacity of a system employing the simple multiuser angle-frequency coding scheme is close to that of a system employing dirty paper coding but significantly better than that of an orthogonal multiplexing system. In this part of thesis, we focus on the scheduling mechanism for angle-frequency subchannels. With the channel identification, we propose two efficient scheduling strategies to make MU transmit simultaneously without serious packet collisions. With the proposed approaches, the packets transmitted to different subscriber units can be scheduled efficiently at the access point to increase the channel utilization and decrease the average packet delay.