Carrier Frequency Offset Estimation Algorithms And Their Performances Analysis In OFDM Systems

In recent years, great interest was focused on OFDM just because its inherent technical advantages in combating multipath fading and its high resource utilization. OFDM also has some disadvantages, for example, it is more sensitive to the carrier frequency offset than single-carrier systems, and very small carrier frequency offset may induce large demodulation errors in the receiver. The main task of this thesis is to study the accurate and high performance algorithms for estimating the carrier frequency offset in OFDM systems. Based on the understanding of some classical carrier frequency offset estimation algorithm in the world, some new training symbol aided carrier frequency offset estimation algorithms are proposed in this thesis, and their performances are demonstrated and evaluated in depth. The main contribution of this thesis includes: First, the effect of frame synchronization errors and carrier frequency offset on the demodulation of OFDM system receivers is analyzed, and a brief equation about the OFDM system demodulation performance as a function of frame synchronization errors and carrier frequency offset is derived. Second, several carrier frequency offset estimation algorithms, which based on the training sequence that is composed of more than two identical small blocks, are proposed. Thses algorithms include: 1) a high performance carrier frequency offset acquisition and tracking scheme that based on a fixed-length training sequence. That training sequence is composed of more than two identical small training symbols. The proposed scheme can shorten each training symbol in a training sequence and select an appropriate estimator simultaneously, as can lead to further reduction of estimation error and increase of acquisition range even with the total training sequence energy being fixed. 2) based on the understanding of BLUE (Best Linear Unbiased Estimator) algorithm proposed by Michele Morelli, two improved algorithms are proposed, i.e., a Weighting and Averaging algorithm and a ML (Maximum Likelihood)...