Research And Implementation Of Parallel Timing Synchronization Algorithm Of OFDM

Due to the strong anti-fading capability, the strong anti-inter-symbol interference (ISI) capability, the high spectral efficiency and so on, orthogonal frequency division multiplexing (OFDM) technology are widely used in the field of wireless mobile communication. However, the OFDM technology also has the inherent limitation, which is reflected in that it is very sensitive to the synchronization errors. Without the accurate synchronization algorithm, there can not guarantee to transmit the reliable data in this system. The synchronization technology is directly related to the performance of the communication system, and the OFDM timing synchronization is the premise for all other synchronization. So how to implement the timing synchronization also has become an important issue in this field. And with the workload of storge, transmission, processing of data is getting larger and larger, so how to improve the running speed of the timing synchronization algorithm is becoming one of the hot issues in todays wireless communication system.The rapid development of Graphics Processing Unit (GPU) brings us a new solution to this issue. Because the GPU’s computing power and the bandwidth of memory have reached a hundred times of the contemporaneous Central Processing Unit’s (CPU), or even more. So GPU provides us an efficient platform for the high-performance computing. But because the limitations of the GPU’s hardware structure, programmers are difficult to use the GPU’s resources to implement the general-purpose computing. However, the Compute Unified Device Architecture (CUDA) which is introduced by NVIDIA has changed the status quo. CUDA can use the strong GPU’s computing power for the general purpose computing, and it is based on the C language. So programmers can easily use it to develop programs on GPU.This thesis discusses how to use the CUDA architecture to implement the parallelization of the timing synchronization algorithm, and to improve the running speed of the algorithm. The main work of this paper includes:(1) Realize the serial code of the timing synchronization algorithm based on the cyclic prefix on CPU. (2) Generalize the parallel parts of the serial algorithm and rewrite them, and then run it on GPU to achieve the parallelization of the timing synchronization algorithm. (3) Write the timing synchronization program with Matlab, and verify the correctness of the serial and parallel algorithms with the simulation results, and then make a comparison between the execution time by the serial algorithm and the parallel one. The results show that the time consumed of parallel algorithm based on GPU is much less than the serial one based on CPU. (4) Finally, according to these results, noting bottlenecks to realize the parallel algorithm on GPU, and give some corresponding optimization strategies.