On Key Transmitting Technologys In TDD Based Relatively Large Scale Antenna Systems

During the past20years, multiple-input and multiple-output (MIMO) technology has been widely studied and applied in wireless communication systems. Most of the current MIMO systems are small scale antenna systems where the antennas in both of the transmitter and receiver are smaller than8. Theoretically, by increasing the numbers of antennas, a MIMO system obtains more spatial degrees of freedom thereby improving the system data rate, reliability, energy efficiency, robust as well as security. In a word, deploying relative large scale antenna systems to meet the requirement of higher data rate has been regarded as a key feature of the furture broad band wireless cellular systems. Due to the channel reciprocity, time division duplexing (TDD) mode is more suitable than frequency division duplexing (FDD) mode in large scale antenna systems.Taking some practical challenges into consideration, this dissertation focuses on several key technologys at the transmitter side for two kinds of antenna arrays with relatively large scale in the TDD based multiuser scenario. The first kind array is a centralized array using a low cost architecture, in which the number of radio frequency (RF)/Analog Front-Ends in the array is smaller than that of the antennas in the array. The typical system applying this kind of array is mmwave or60GHz wireless system. In the conventional small scale antenna systems where the number of RF/Analog Front-Ends equals that of the antennas, the array signal processing can be entirely implemented in the base-band digital domain. However, when using a less number of RF/Analog Front-Ends, to fully utilize the potential of the array, at least a part of signal processing should be shifted into analog or RF domain. This usually leads to two-stage beamforming or precoding for this kind of array which differs from the conventional one-stage beamforming. In addition, due to the limitation of implementation, analog beamforming is not as flexible as digital beamforming. Hence, conventional digital beamforming can not be directly applied to this low cost large scale array. For this kind of array, we focus on the multiuser scheme for the low cost60GHz wireless system using phased-array as well as multiuser hybrid analog/digital beamforming optimization. The second kind of array is a cooperative virtual array where antennas are geographically distributed. One typical system using this kind of array is the so called network MIMO. In practical TDD based MIMO systems, the antenna mismatch is a widely existing non-ideal factor that destroies channel reciprocity in base band. For this kind of system, the dissertation focuses on studying the impact of antenna mismatch on the multiuser system and antenna calibration method for distributed antennas. The main contributions of this dissertation are concluded as follows:Firstly, for low cost60GHz wireless system using phased-array, this work proposes a low complexity multiuser scheme. The proposed scheme includes three stages, which are scheduling stage, beamforming training stage and digital multiuser interference mitigation stage. The solved key problems in the proposed scheme are the user pairing and determination of the analog and digital beamforming weights. The proposed scheme can take advantage of the single user beamforming training procedure for60GHz wireless system to pair users that are not strongly interfering to each other. The analog beamforming is used to only enhance the link quality between the corresponding RF/Analog Front-End and the user. The residual multiuser interference is mitigated in digital domain on the effective channels. The low complexity of the proposed scheme is reflected in that it requires neither to estimate the large scale MIMO channels between the access point (AP) and users nor to obtain the geographic information of the users. Under the IEEE802.11ad channel model, we evaluate the proposed scheme in line-of-sight (LoS) and non-line-of-sight (NLoS) scenario with different extents of Doppler shift. Simulation results show that the proposed scheduling method can avoid strong multiuser interference, and if the Doppler shift is very small, evident gain can also be observed in low SINR region after digital multiuser interference mitigation. However, the high frequency of60GHz makes the performance of digital multiuser interference mitigation much sensitive to the outdated channel information.Secondly, for general multiuser hybrid analog/digital beamforming, this work further investigates two optimization problems. One is to minimize the average transmit power on each subcarrier subject to the SINR constraints on each subcarrier. The other one is to maximize the minimal weighted SINR subject to the average transmit power per subcarrier constraint. The work in this part of the dissertation can be regarded as an extension of the conventional digital beamforming optimization to the hybrid beamforming and also an extension of the single user hybrid beamforming to multiuser hybrid beamforming. We analyze the feasibility and the properties on the optimal point for the two optimization problems, respectively. In addition, suboptimal solutions are proposed for solving the considered problems due to that the problems are non-convex. Based on the simulation results, the following conclusions are obtained. When channels on different subcarriers are of high correlation, the performance of multiuser hybrid analog/digital beamforming can approach to that of pure digital multiuser beamforming using the same number of antennas. On the contrary, when channels on different subcarriers are of low correlation, the performance degration under multiuser hybrid analog/digital beamforming can be significant compared to that of pure digital multiuser beamforming due to the use of frequency-flat analog beamforming weights. The results apply not only to mmwave wireless systems but also to the systems using low cost architecture arrays operating at low frequency. Under the60GHz line of sight (LoS) channel model, it is also observed that the direction of the optimized analog beamforming per RF/Analog Front-End is related to the spatial separation of the users. When the two users are of large spatial separation, pointing the analog beamformings of different RF/Analog Front-Ends to different users can achieve lower transmit power or higher max-min signal-to-interference-plus-noise-ratio (SINR). On the contrary, when the two users are of small spatial separation, it is benefitial to make the analog beamforming of each RF/Analog Front-End point to all the users while leaving the multiuser interference mitigation for digital beamforming.Thirdly, for the antenna mismatch in TDD based network MIMO system, this work presents some in-depth analyses on the impact of antenna mismatch and propose schemes for calibrating two distributed antennas and arbitrary number of antennas. Existing literatures have pointed that antenna mismatch can cause multiuser interference in MIMO system when regarding the transpose of the uplink base band channel matrix as the downlink base band channel matrix. However, little is known on the statistical property of the caused multiuser interference and the perturbation of multiuser linear precodings. This work analyzes the statistical relationship between the multiuser interference caused by antenna amplitude mismatch and phase mismatch at the base station (BS) side. We also study the perturbation of the precoding vectors caused by antenna mismatch at the user side. These analyses can not only provide in-depth impact of antenna mismatch but also contribute to the design of antenna calibration methods. Antenna calibration is an approach to compensate the antenna mismatch. The array self calibration usually applied in centralized antenna array is costly for calibrating distributd antennas due mainly to its requirement of hardware support. By comparison, pure software antenna calibration which is usually based on air-interface is a better choice for calibrating distributd antennas. We propose an improved calibration method for calibrating two distributed BS antennas which can obtain less calibration error. Besides, we propose an efficient calibration coefficient computation scheme for calibrating arbitrary number of distributed BS antennas. The proposed computation scheme combines temporary calibration coefficients obtained from different references to compute the final calibration coefficient for each antenna. Two combination methods are further proposed. Simulation results validate the analysis and compare the performance of antenna calibration under different parameters.