On Time-Frequency Synchronization For Next Generation Wireless Communication Systems

In the last decade, the fast growing demands for high-data-rate and high-quality communi-cations over wireless channels make the design and implementation of the next-generationwireless communication systems more and more challenging. In physical layer, to eliminatethe inter-symbol interference (ISI) caused by the high-rate wireless transmission effectively,and to solve the basic contradiction between the high-rate wireless communication require-ments of large number of subscribes and the scarcity of electromagnetic spectrum, multi-carrier technique, such as orthogonal frequency division multiplexing (OFDM) technique,and multiple-input multiple-output (MIMO) technique, which is able to exploit the potentialspatial resource effectively, attracts much attention and extensive research in both academicand industrial society. They have already been widely recognized as the most promisingcandidates for the next-generation wireless communications systems.Synchronization, which consists of timing synchronization and carrier synchronization,is a key technique of digital communication systems. Various modulation and multiple-access techniques based on OFDM, such as SISO-OFDM, MIMO-OFDM and orthogonalfrequency-division multiple-access (OFDMA), are very sensitive to carrier frequency offset(CFO). On the other hand, cooperative communication system, as a special form of dis-tributed MIMO system, is also inherently time-frequency asynchronous. Consequently, it isvery important to do the research on the time-frequency synchronization for these systems.This dissertation focuses on the synchronization issue for next generation wireless communi-cation systems. Time-frequency synchronization in various modulation and multiple-accesstechniques based on OFDM, such as SISO-OFDM, MIMO-OFDM and OFDMA, and co-operative communication systems are investigated. Several aspects of communications andsignal processing are involved in the dissertation, including parameter estimation, signaldetection and space-time coding design. The main contributions of the dissertation lie in:1. The CFO estimation issue in SISO-OFDM, MIMO-OFDM and OFDMA systems areinvestigated. It has been indicated that the CFO estimation is essentially the frequenciesestimation issue of multiple harmonics with different amplitudes and initial phases in noise.Consequently, the CFO estimation of various OFDM systems can be discussed in an unifiedframework. Time-domain, frequency-domain and spatial-domain resources of these OFDM systems can be fully exploited and a series of CFO estimation approaches are proposed,which provide a new perspective of the CFO estimation issue. Especially:1) Taking full advantage of the frequency-domain resource. The redundancy of CFOinformation in the spectrum structure of OFDM signals has been fully taken advan-tage of, and a novel CFO estimation method based on virtual subcarriers (zero pilots)has been proposed applying to both SISO- and MIMO-OFDM systems. It has beenshown that the CFO information is redundant in the spectrum structure of the receivedOFDM blocks. By allocating virtual subcarriers uniformly and down-sampling the re-ceived blocks, the method remodels each received OFDM block and provides a signalmatrix with lower dimension. The spectrum aliasing induces by the down-samplingdoes not cause the CFO information loss but compresses the dimension of signal spaceand enhances the estimation performance. Moreover, in frequency-selective channels,this spectrum aliasing provides frequency diversity and hence avoids the performancedegradation due to the deep fading (even spectrum-zeros) of the channel. The pro-posed method retains the high-resolution feature and the computational complexity isalso very low.2) Taking full advantage of the time-domain resource. It has been shown in practicalOFDM systems that, sufficiently long cyclic prefix (CP) is always adopted to elimi-nate the inter-block interference (IBI) completely, hence there always exists"clean"CP samples, i.e., CP samples do not corrupted by the previous OFDM block. A noveladaptive joint block timing and CFO estimation algorithm for SISO-OFDM systemshas been proposed. Compared with similar methods exploiting CP, the adaptive ap-proach can utilize clean CP samples to avoid IBI effectively in a way without knowingthe exact length and location of them. Consequently, the algorithm is robust in time-dispersive channels. Furthermore, when the adaptive process converges, we show thealgorithm provides the maximum likelihood (ML) estimation of the CFO.3) Taking full advantage of the spatial-domain resource. The adoption of uniform lineararray (ULA) at the base station of OFDMA systems has been fully taken advantageof, and a novel multiuser CFOs estimation method has been proposed for OFDMA up-link systems. Based on the narrow-band signal assumption, the new approach exploitsthe spatial division capability of ULA to separate signals from different users in thespatial-domain, and hence is able to estimation the CFOs of multiusers independently.Due to the utilization of spatial resource, the novel method has a series of merits: a) Itdoes not rely on any specific carrier assignment scheme (CAS) and applies to systemswith generalized CAS, so it is more ?exible and dynamic channel assignment is avail-able; b) It applies to fully loaded systems, i.e., all subcarriers can be assigned to users,which results in higher bandwidth efficiency; c) In addition, the directions of arrival(DOAs) of all active users are also obtained, which can be used for downlink selective transmission and mobile user orientation in frequency division duplex (FDD) systems.2. The reliable communication issue in cooperative communication systems with both timeand frequency asynchronization is investigated. From the transmitting end to the receivingend, distributed space-time coding schemes and signal detection algorithms are discussedand proposed. Especially:1) The cooperative transmission scheme design for frequency-asynchronous cooperativecommunication systems with multiple CFOs is investigated. The existence of multi-ple CFOs in cooperative systems makes the equivalent channels time-varying, whichwill destroy the elaborately designed structure of the distributed space-time codes, anddegrade the performance of the system. To deal with this, by time-frequency trans-form, a ?at fading channel with CFO in the time domain has been converted to a blocktime-invariant ISI channel in the frequency domain. Based on such channel, two dis-tributed space-frequency codes (SFC), frequency-reversal SFC and frequency-domainlinear convolutive SFC, are proposed to achieve the cooperative spatial diversity. Ithas been shown that even under the frequency-asynchronous case, full cooperative di-versity can be achieved with more computationally efficient linear receivers, such aszero-forcing (ZF) and minimum mean square error (MMSE) receivers, instead of thecomputationally exhausted maximum likelihood (ML) decoding. The design schemeis then generalized to the case with any number of relay nodes. A family of bandwidthefficient SFCs which can achieve full diversity with linear receivers under frequency-asynchronous case is proposed. The new family of codes achieves a better tradeoffamong performance, decoding complexity and spectrum efficiency.2) The reliable communication issue in time-frequency asynchronous cooperative com-munication systems is investigated. To deals with the time-selective and frequency-selective equivalent channel caused by the multiple time and frequency offsets in co-operative communication systems, the design of transmitting and receiving schemesare addressed. At the transmitter end, distributed linear convolutive space-time code isadopted to collect full time-asynchronous cooperative diversity with only linear re-ceivers. Meanwhile at the receiver end, computationally efficient fast equalizationmethods are proposed. The time-selectivity of the equivalent channel is only intro-duced by the multiple CFOs, which can be determined once these CFOs are correctlyestimated. This implies that there exists special relationship between the channel ma-trices in the consecutive two symbol durations. Based on this observation, iterativeminimum mean square error (MMSE) equalizer and MMSE decision feedback equal-izer (MMSE-DFE) are proposed, where the equalizer at current symbol duration canbe designed using the one obtained at the last symbol duration instead of re-designcompletely, which therefore significantly reduces the computational complexity for the equalization.