为了满足下一代无线移动宽带业务需求，3GPP（3rd Generation Partnership Project）启动了LTE（Long Term Evolution）研究项目。作为LTE物理层关键技术之一的OFDM（Orthogonal Frequency Division Multiplexing）是一种多载波调制技术，因此它对时间和频率偏移非常敏感，获得更加精确的时频同步对OFDM系统来说是非常重要的。 本文研究了TDD-LTE系统下行同步的流程及相应的同步算法。LTE下行同步过程分同步捕获过程和同步跟踪过程。在同步的捕获阶段，终端利用主同步信号PSS（Primary Synchronization Signal）获得定时位置、频偏估计和扇区ID（Identifier）。常用的定时算法包括了相关算法、互相关算法和自互相关结合的算法；粗频偏估计研究了三种算法。帧同步和小区ID的获得是利用辅同步信号SSS（Secondary Synchronization Signal），SSS的检测算法中包括了相干检测算法和非相干检测算法。在同步跟踪阶段，终端可以利用循环前缀（CP）或是导频（RS）获得时间频率的调整量，相应的估计算法包括了时域频偏估计算法和频域频偏估计算法。 本文搭建了TDD-LTE链路仿真平台并对下行同步过程中的各种算法进行了仿真。仿真结果表明：在同步捕获阶段，PSS粗定时的三种同步算法中的互相关算法的性能最好，PSS粗频偏估计算法1和算法2都受定时的影响性能有所下降，粗频偏估计算法3，在存在较大频偏的情况下，可以提高性能，但此算法的缺点就是只能做整数倍频偏估计。辅同步信号检测算法中，通过仿真可以发现在小频偏的情况下，相干检测比非相干检测性能更好。在大频偏的情况下，随着信噪比的增加，相干检测算法的性能比不上非相干检测算法中的部分相干算法，差分相干检测的算法的性能比不上部分相干检测算法。在同步的跟踪阶段，先对频域频偏估计算法进行了仿真，通过仿真对比了几种导频符号相关组合下频偏估计性能的高低。最后针对时域频偏估计算法和频域频偏算法做了仿真，并对比了加入平滑算法前后各频偏算法的性能，通过仿真可以发现无论平滑前还是平滑后，时域（TTI）算法优于时域（symbol）算法，频域算法都优于时域算法。

3GPP initiated research projects of Long Term Evolution(LTE) for satisfied the demands of next-generation wireless mobile broadband services.OFDM is a technology of multi-carrier modulation as one of the key technologies of the LTE physical layer. It is very important for OFDM system to obtain more accurate time and frequency synchronization, because OFDM is very sensitive to time and frequency offset. Downlink synchronization procedure and the corresponding synchronization algorithms in TDD-LTE are researched in this paper. LTE downlink synchronization procedure includes synchronous capture procedure and synchronous tracking procedure. In the synchronous capture stage, terminal obtain the regular position, the frequency offset estimation and sector ID by using primary synchronization signal (PSS). Commonly used timing algorithms includes correlation algorithm, cross-correlation algorithm and algorithm integrated autocorrelation and cross-correlation algorithm. Coarse frequency offset estimation algorithm includes three algorithms. Frame synchronization and cell ID are obtained by using secondary synchronous signals SSS, SSS detection algorithm includes coherent detection algorithm and non-coherent detection algorithm. In the synchronous tracking stage, terminal obtain the amount of time and frequency adjustment by using Cyclic prefix (CP) or pilot (RS) , the related estimation algorithms include frequency offset estimation algorithms of time domain and frequency domain. In this paper, TDD-LTE link simulation platform is constructed and a variety of algorithms in downlink synchronization process are simulated. The result of simulation shows that in the synchronous capture stage, cross-correlation algorithm achieves the best performance in three kinds of synchronization algorithms of PSS coarse timing, PSS coarse frequency offset estimation algorithm 1 and algorithm 2 are drop in performance affected by timing, coarse frequency offset estimation algorithm 3 can improve the performance in case of larger frequency offset, but the disadvantage of this algorithm is that it can only complete integer times frequency offset estimation. In the secondary synchronization signal detection algorithm, we can find that the performance of coherent detection is better than the non-coherent detection by simulation in case of small frequency offset. In the case of large frequency offset, with the increase of SNR, the performance of coherent detection algorithm is worse than partial coherence algorithm of non-coherent detection algorithms, the performance of differential coherent detection algorithm is worse than partial coherence detection algorithm. In the synchronous tracking stage, firstly, frequency offset estimation algorithm of frequency domain is simulated, and the performances of frequency offset estimation algorithms under several kinds of pilot symbols combination are compared by simulation. Finally, frequency offset estimation algorithms of time domain and frequency domain are simulated, and the performances of frequency offset algorithm before and after joining the smoothing algorithm are compared. According to the simulation, we can find that , whether smoothed or not, time domain（TTI）algorithm is better than time （symbol）domain algorithm, and frequency domain algorithm is better than time domain algorithm.