基于X86的LTE基带软处理研究

发布时间：2018-08-14 15:59

【摘要】：经过二十多年的发展,移动通信技术取得了巨大的进步,从早期的GSM系统到现在的LTE系统,但是也造成了多种移动通信系统共存的局面。如何能低成本地实现多种网络制式共存,成为了业界研究的课题。随着功耗的降低,加上高性能和编程灵活性的优点,利用基于X86架构的通用处理器实现通信系统吸引越来越多的研究人员。本文即是探究利用基于X86架构的通用处理器实现LTE系统的可能性。本文的主要内容包括以下三个部分:(1)首先简要介绍了LTE系统上行信道处理过程,并基于通用处理器利用C语言进行了实现。经过上千次测试,计算出各个处理模块的的平均处理时间,确定了两个最耗时的模块:传输预编码模块和Turbo译码模块。为了缩短程序的运行时间,研究了基于X86架构的程序优化方法:X86的SSE并行计算指令集优化和查表法优化。(2)传输预编码的数学原理即是离散傅里叶变换(DFT)。由于直接DFT算法需要大量的乘加运算,因此结合LTE系统传输预编码模块的特点,设计了基数为2,3,4和5的混合基快速傅里叶变换(FFT)算法。以长度为300的DFT为例,介绍了混合基算法的实现流程,并详细介绍了利用SSE指令实现长度为3的DFT计算,最后对SSE指令实现的混合基FFT算法进行性能验证。(3)研究了Turbo译码的MAP系列算法的推导过程,进而研究了基于MAXLOG-MAP算法的并行译码算法。利用Matlab对基于MAX-LOG-MAP算法的Turbo并行译码算法进行了仿真,确定了迭代次数和量化方案。介绍了turbo并行译码算法实现框架,并对分支度量、前向度量、后向度量和对数似然比的并行实现中SSE指令的应用进行了具体分析。最后对turbo并行译码算法的性能进行了验证。
[Abstract]:After more than 20 years of development, mobile communication technology has made great progress, from the early GSM system to the present LTE system, but also caused the coexistence of a variety of mobile communication systems. How to realize the coexistence of multiple network standards at low cost has become a research topic in the industry. With the decrease of power consumption and the advantages of high performance and programming flexibility, it is attracting more and more researchers to implement communication systems using universal processors based on X86 architecture. This article explores the possibility of using a general purpose processor based on X 86 architecture to implement a LTE system. The main contents of this paper include the following three parts: (1) the uplink channel processing process of LTE system is introduced briefly and implemented by C language based on general purpose processor. After thousands of tests, the average processing time of each processing module is calculated, and the two most time-consuming modules are determined: the transmission precoding module and the Turbo decoding module. In order to shorten the running time of the program, the optimization methods of program optimization based on X86 architecture, SSE parallel computing instruction set optimization and look-up table method, are studied. (2) the mathematical principle of transmission precoding is discrete Fourier transform (DFT). Because direct DFT algorithm needs a lot of multiplication and addition operation, a hybrid base fast Fourier transform (FFT) algorithm with cardinality of 2 ~ 3 ~ 4 and 5 is designed according to the characteristics of transmission precoding module in LTE system. Taking DFT with length 300 as an example, the realization flow of hybrid basis algorithm is introduced, and the realization of DFT calculation with length 3 by using SSE instruction is introduced in detail. Finally, the hybrid base FFT algorithm implemented by SSE instructions is verified. (3) the derivation process of MAP series algorithm for Turbo decoding is studied, and then the parallel decoding algorithm based on MAXLOG-MAP algorithm is studied. The Turbo parallel decoding algorithm based on MAX-LOG-MAP algorithm is simulated by Matlab, and the iteration times and quantization scheme are determined. This paper introduces the implementation framework of turbo parallel decoding algorithm, and analyzes the application of SSE instruction in the parallel implementation of branch metric, forward metric, backward metric and logarithmic likelihood ratio. Finally, the performance of turbo parallel decoding algorithm is verified.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN929.5

【相似文献】