无线通信恶劣的信道环境下，均衡能够大大改善接收性能。数字通信中的频域均衡，是在离散频域完成的均衡技术；比传统的时域均衡技术具有更固定的设计和更优异的性能，因而得到了广泛的应用。本文通过分析频域均衡以及非线性均衡的技术，并通过对均衡中噪声传递的研究，做出系统性能的评价与预测，为通信系统的设计提供有力支持。首先，本文分析了线性均衡技术。文章建立了涵盖CP-OFDM、PRP-OFDM以及单载波系统的广义OFDM模型，研究了频域ZF以及LMMSE均衡下的噪声传递，从互信息角度上比较了OFDM与单载波系统在两种均衡器下的性能，证明了ZF均衡下因为均衡DFT与OFDM调制DFT长度不一致时引起的互信息损失，和单载波LMMSE均衡相对于ZF均衡在互信息意义上的优势。其次，本文分析了单载波判决反馈频域均衡技术。文章改进了IBDFE算法中的相关系数估计方法，并基于IBDFE提出了IMF-DFE 的结构。利用IMF-DFE中的噪声传递分析，提出了噪声传递图。噪声传递图能够直观化单载波判决反馈均衡迭代过程，给出的收敛结果能够预测IMF-DFE 的最终性能。最后，本文分析了动态信道下OFDM判决反馈频域均衡技术。文章在动态信道变化线性化的假设下，提出了双PN填充的OFDM系统在动态信道下的循环卷积重构算法，并基于此提出了低复杂度频域均衡算法ZF-FB以及迭代判决反馈频域均衡算法MMSE-DFE。通过对MMSE-DFE噪声传递的分析，跟踪迭代过程中OFDM子载波上的噪声功率，预测迭代算法收敛性能。本文侧重于理论分析，辅以仿真证明理论分析中模型的合理性。文中所研究的算法和技术可以应用到实际的通信、广播系统中，基于这些算法的噪声传递分析可以评估和预测算法性能，指导系统设计。
Equalization is essential to wireless communications to improve the performance under severely frequency-selective channels. The frequency-domain equalization, which operates in the discrete frequency domain, has now been widely adopted in the single-carrier modulation (SCM). Compared to the traditional time-domain techniques, it has a more fixed design with a better output. In this dissertation, we are going to look into the frequency-domain equalizing technologies and evaluate the performance for SC and OFDM via the underlying noise transfer analysis. We believe that the conclusions in the paper would be a powerful tool in assisting the design of a wireless communication system.First, we will look into the linear technologies. By establishing a generalized OFDM model that incorporates CP-OFDM, PRP-OFDM, and UW-SC, we will investigate the noise transfer under the ZF and the LMMSE equalizer. We will compare the mutual information between OFDM-vs-SC and ZF-vs-LMMSE, respectively. The degradation from OFDM to SC in the ZF equalization will be shown to be generated from the mismatch between the lengths of the equalizer's DFT and the modulator's DFT. For SC, the mutual information of the LMMSE equalizer is proved to be greater than the ZF equalizer.Next, we will look into the non-linear methods, i.e., decision feedback equalization (DFE). We will improve the algorithm to estimate the cross-correlation coefficients in IBDFE. Further we will propose the IMF-DFE algorithm, the noise transfer of which will be analyzed. An error transfer chart will be introduced to trace the iteration process and predict the convergence behavior. The simulation will show that the error transfer chart is a powerful tool to visualize the single-carrier frequency-domain DFE and evaluate the performance of the system.Finally, we will look into the DFE of OFDM under time-varying channels. Based on an assumption of the linearized variation of a channel in an OFDM symbol, we will propose a circular convolution reconstruction algorithm for the dual-PN padding OFDM, followed by a low-complexity ICI mitigation algorithm, also known as ZF-FB. We will propose the MMSE-DFE algorithm as an iterative non-linear method to mitigate the ICI. The underlying noise transfer will also be formulated to trace the noise power in the subcarriers and show the convergence behaviour of the iterations.The formulations in the dissertation are theoretical, and we will use simulations to show the validity of the models adopted in the theories. The algorithms and technologies in the paper can be applied in the practical implementations of communication and broadcasting. The underlying noise transfer analysis can be used to evaluate the algorithm and provide a guideline in the system design.
