Omnidirectional Transmission And Channel Estimation In MIMO Systems

By employing multiple antennas at both the transmitter and receiver, multiple-input multiple-output (MIMO) technology can provide a greater degree of freedom in addition to time and frequency dimensions in wireless channels. Therefore, a significant performance improvement can be obtained in terms of spectral efficiency, power efficiency, and reliability. MIMO has been regarded as one of the key technologies in modern cellular mobile communication systems. In this dissertation, the theories and approaches of omnidirectional transmission and channel estimation in MIMO systems are investigated.Firstly, we investigate omnidirectional transmission in massive MIMO sys-tems. We propose omnidirectional precoding based transmission, where the high-dimensional signal vector transmitted from the base station (BS) is composed by an omnidirectional precoding matrix and a low-dimensional signal vector. The user terminal (UT) just needs to estimate the precoded effective channel. Hence, the downlink pilot overhead can be reduced significantly. For the omnidirectional precoding based transmission, we present three necessary conditions that the pre-coding matrix should satisfy, to meet the requirements for omnidirectional trans-mission for reliable cell-wide coverage, equal-average-power on each antenna to sufficiently utilize all the available power amplifier capacities of BS antennas, and achievable ergodic rate maximization for the independent identically distributed (i.i.d.) channel, respectively. Several examples of the precoding matrix satisfying the three necessary conditions simultaneously are designed by utilizing the Zadoff-Chu (ZC) sequence and its properties. We also analyze the system performance in terms of achievable ergodic rate, outage probability, and peak-to-average pow-er ratio (PAPR) for these designs. It is shown that a dedicated design has the following advantages. It asymptotically maximizes the achievable ergodic rate, maximizes the achievable diversity order, and minimizes the outage probability in the large-scale array regime, not only for the i.i.d. channel, but also for spa-tially correlated channels; It preserves the PAPR of the transmitted signal after precoding.Secondly, we investigate omnidirectional space-time block coding in massive MIMO systems. In order to reduce the burden of the downlink channel estima-tion and achieve partial spatial diversity from BS antennas, we propose channel-independently precoded low-dimensional space-time block code (STBC). The pre-coding matrix and the signal constellation in the low-dimensional STBC are jointly designed to guarantee omnidirectional transmission and equal-power on each an-tenna at any instant time, and at the same time, achieve the full diversity of the low-dimensional STBC. Under this framework, several designs are presented. To provide transmit diversity order 2, a precoded Alamouti code is proposed, which has a fast maximum-likelihood (ML) symbol-wise decoding. To provide transmit diversity order 4, a precoded quasi-orthogonal STBC is proposed, which has a pair-wise ML decoding. Moreover, a precoded no-zero-entry Toeplitz code and a precoded no-zero-entry overlapped Alamouti code are also proposed. These two codes can achieve a higher diversity order with linear receivers.Then, we investigate downlink synchronization in millimetre wave massive MIMO systems. When the number of radio frequency chains are limited at both the BS and UT, the transmission model of synchronization signals is presented. Under this model, we derive the optimal synchronization detector based on gen-eralized likelihood ratio test. Then we present several necessary conditions that the transmit and receive beamforming matrices should satisfy, to achieve omnidi-rectional coverage, minimize the asymptotic false alarm probability, and minimize the asymptotic missed detection probability for both the i.i.d. channel and the single-path channel. Then, by utilizing the ZC sequence and the Golay sequence, we design several examples for the transmit and receive beamforming matrices sat-isfying these necessary conditions simultaneously. The proposed approach shows significant performance gains when compared with the traditional beam scanning method.Next, we investigate channel estimation in MIMO-OFDM systems when con-sidering co-channel interferences. The received pilot signals are utilized to estimate the channel frequency response (CFR) matrix and the interference-plus-noise co-variance (INC) matrix on each subcarrier. The estimation of the CFR matrix is obtained under least square criterion by utilizing the sparsity in time domain. Then we estimate the autocorrelation function of interference-plus-noise in time domain, instead of estimating the ICM in frequency domain directly. The auto-correlation function estimation has two steps. Firstly, we present the relationship between the sample autocorrelation function of the residual and the true auto-correlation function. Based on this, we propose a compensating method. Then, since the compensated sample autocorrelation function of the residual cannot be guaranteed to be an autocorrelation sequence (ACS), we utilize semidefinite pro-gramming (SDP) to find the closest ACS. The SDP is solved in its dual form, which yields a significantly reduced complexity. Finally, the estimated ICM is reutilized to revise the CFR estimation. The estimations of CFR and ICM show excellent interference suppression performance when being applied in an interfer-ence rejection combining receiver.Finally, we investigate semi-blind channel estimation in MIMO systems when considering co-channel interferences. The received pilot and data signals are jointly utilized to estimate the channel vector and the INC matrix. When the length of the data signals is sufficiently large, the sample covariance matrix of the received data signals can be approximated by the sum of the covariance matrix of the channel vector and the INC matrix. By utilizing this property, we propose to estimate the channel vector and the INC matrix jointly under conditional ML criterion, and model it as a non-convex optimization problem. There are two steps to solve this problem. Firstly, the optimal channel vector is derived in a quasi-closed form, where there is only one unknown scalar. Then, this unknown scalar is determined by solving polynomial equations. Although the theoretical derivation of this conditional ML estimator is based on the assumption that the length of the data signals is sufficiently large, simulation results show that as long as the length of the data signals is not too small, this conditional ML estimator shows significant performance gains when compared with the traditional ML estimator.