Channel Parameters Estimation And Peak-to-Average Power Ratio Reduction For OFDM Systems

As a multi-carrier technique with high bandwidth efficiency, low-complexityequalization, capability of combating multi-path, good design flexibility andcompatibility, orthogonal frequency division multiplexing (OFDM) has been widelyused in many wireless wideband transmission realms. More specifically, beingconsidered as the key technique for the fourth generation wireless communications,OFDM has been adopted by several emerging international wireless standards, e.g.,WLAN, WiMAX and LTE-A. However, since OFDM was first proposed, its inherentproblems of high peak-to-average power ratio (PAPR) and susceptibility to frequencyoffset have restricted its application to a large extent in some specific scenarios, such ashigh-mobility and power-limited environments. Aiming at these two major problems, anin-depth investigation of OFDM is implemented in this dissertation from theperspectives of PAPR reduction, time-frequency synchronization and channel estimation,respectively. The primary contributions included in this dissertation are summarizedbelow:1. To alleviate the inherent high-PAPR problem of OFDM systems, twospectral-shaping based PAPR reduction algorithms (spectral-shaping based PAPRreduction without bandwidth expansion and time-domain locating-pulse peakcancellation assisted spectral-shaping based PAPR reduction) are proposed. Thecommon characteristic of these two algorithms is that matrix-based precoding andfrequency-domain spectral shaping are integrated. Since fewer FFT/IFFF operationsand no side information are required, not only the complexity is greatly reduced, butalso the bit-error-rate performance degradation due to an erroneous detection of theside information is avoided. Moreover, since the design of the adoptedspectral-shaping functions is independent of the original OFDM signals, the increasein the complexity due to a search of the optimal function can be avoided. Thedistinction between these two algorithms lies in the extra subcarriers which arereserved for the latter one. By using the time-domain locating-pulse generated fromthe IFFT of the reserved subcarriers to iteratively cancel the peak signals afterspectral shaping, the PAPR can be further reduced. Both theoretic analyses andsimulations verify the validity of the proposed algorithms.2. Attempts to upgrade the OFDM systems with low-accuracy channel parameters estimation in high-mobility scenarios led to a series of novel channel parametersestimation algorithms which are based on cross ambiguity functions (CAF). i) Anovel CAF energy detection based estimation metric is proposed for integerfrequency offset (IFO) estimation. By designing a training sequence with theambiguity function (AF) which can approximate the ideal AF, a robust, full-range,high-accuracy and low-complexity IFO estimation is obtained. ii) A noveltwo-dimensional time-lag/frequency-offset metric function is proposed for IFOestimation in an OFDM system with residue time offset (RTO). By implementing atwo-dimensional search, both IFO and RTO estimation can be obtained. iii) A novelChu-sequence based joint estimation of IFO and channel impulse response (CIR) isproposed. Due to the adoption of Chu-sequence, the degradation of channelestimation due to the high PAPR can be avoided. By utilizing the specialcharacteristics of the CAF Doppler cuts of the Chu-sequence, the zero-th CAFDoppler cut is sufficient to achieve a super-low-complexity and high-accuracy jointestimation of IFO and CIR at the cost of a moderately reduced estimation range.3. Also proposed in this dissertation are a series of novel low-PAPR and high-accuracyCIR estimation approaches based on circularly orthogonal Chu-sequences forMIMO-OFDM systems. i) Through a thorough investigation of the idealperiodic-autocorrelation characteristic of Chu-sequences, a time-domain circularlyorthogonal Chu-sequences based CIR estimation is proposed for frequency selectivefading channels. By decoupling the multiple taps for each branch, a valid CIRestimation can be obtained. ii) An improved CIR estimation which is based onsuccessively repetitive circularly short orthogonal Chu-sequences is proposed. Notonly the same estimation performance can be obtained, but also the computationalcomplexity of the algorithm can be greatly reduced. iii) Further, a novel CIRestimation for sparse multi-path channels is investigated. Specifically, generalizedAkaike’s information criterion is adopted to determine the channel order and thetime delay of each tap. By preserving the effective taps, the effect of the noise on theCIR estimation can be mitigated. As a result, higher-accuracy CIR estimation forsparse multi-path channels is achieved.