LTE系统MIMO球形译码算法的低复杂度实现

发布时间：2018-10-26 19:34

【摘要】：未来无线数据业务推动着无线通信系统的发展。MIMO(Multiple-input multiple-output)技术满足了口益增长的无线数据业务的需求,是未来无线通信系统中的关键技术,尤其是MIMO空分复用系统,因为能够在不增加带宽的情况下成倍的提高系统容量,所以能够实现较高的频谱利用率。然而,因为接收机需要把空分复用的数据流分开,所以高的频谱利用率也给接收机带来了高的检测复杂度。这就要求MIMO检测必须在保证性能的前提下,具有尽可能低的复杂度。本文对常用的MIMO空分复用检测算法进行了分析和比较,决定以深度优先的球形译码算法为切入点,重点研究LTE(Long Term Evolution)系统软输出球形译码算法的低复杂度实现。本文在详细分析了软输出球形译码算法后,对涉及的一些基本概念,包括部分欧式距离、SE(Schnorr-Euchner)枚举、球限、有效遍历集合、单树搜索、对数似然比、排序的QR分解进行分析和讨论。在此基础上,本文着重提出提高系统性能的方法：对于球形译码吞吐率不固定的问题,采用了限制最大搜索点数的方法,同时提出了“基于统计”的方法设置最大搜索值；对于最大搜索点数可能导致的BER(Bit Error Rate)性能降低的问题,提出了“基于权重”、“可调权重”、优化的“LTE块处理”等一系列方法。对于4x464-QAM系统,和没有设置权重的情况相比,三种方法的性能分别提高2db@1e-3、2.2db@1e-3、2.5db@1e-3。结果证明提出的方法对于BER性能有明显的提高作用。本文提出了基于单树搜索的软输出球形译码架构,同时在该架构的基础上进行了定点分析。本文对架构涉及的时序、搜索复杂度、功耗三方面的问题提出了低复杂度的实现方法：对于搜索复杂度问题,本文提出了"RIP-SE(Reduce Invalid Point-SE)"的枚举方法和"LSR(Layer based sphere restriction)”的球限设置方法,从而有效减少了“无效点”,在性能降低很少的情况下吞吐率提高了2倍；对于时序问题,本文对限制最大搜索点数情况下的定点进行优化,同时提出了球限并行计算方法,最终面积降低了50%;对于功耗问题,本文采用了“格雷码编码”、“流水线结构”、“门控时钟”的低功耗设计方法,功耗减少了65%。结果证明以上的方法对于低复杂度的实现有很大的帮助。
[Abstract]:The future wireless data service promotes the development of wireless communication system. MIMO (Multiple-input multiple-output) technology meets the increasing demand of wireless data service. It is the key technology of future wireless communication system, especially the MIMO space division multiplexing system. Because it can increase the system capacity exponentially without increasing the bandwidth, it can achieve high spectral efficiency. However, because the receiver needs to separate the spatial division multiplexed data stream, the high spectral efficiency also brings high detection complexity to the receiver. This requires that MIMO detection must have as low a complexity as possible on the premise of guaranteed performance. Based on the analysis and comparison of the common MIMO space-division multiplexing (SDDM) detection algorithms, it is decided to focus on the low-complexity implementation of the soft-output spherical decoding algorithm in LTE (Long Term Evolution) systems by using the depth-first spherical decoding algorithm as the starting point. After detailed analysis of the soft-output spherical decoding algorithm, some basic concepts, including partial Euclidean distance (, SE (Schnorr-Euchner) enumeration, sphere limit, effective traversal set, single tree search, logarithmic likelihood ratio, are discussed in this paper. The sorted QR decomposition is analyzed and discussed. On this basis, this paper puts forward the methods to improve the system performance: for the problem of the indefinite throughput of spherical decoding, the method of limiting the maximum number of search points is adopted, and the method based on statistics is put forward to set the maximum search value; A series of methods, such as "based on weight", "adjustable weight" and "LTE block processing", are proposed to solve the problem of BER (Bit Error Rate) performance degradation caused by the maximum number of search points. In the case of 4x464-QAM systems, the performance of the three methods is improved by 2DB @ 1e-32.2db1e-3e 2.5db @ 1e-3compared to those without weights. The results show that the proposed method can obviously improve the performance of BER. In this paper, a single tree search based soft-output spherical decoding architecture is proposed, and the fixed-point analysis is carried out on the basis of this architecture. In this paper, a low complexity implementation method is proposed for the time sequence, search complexity and power consumption of the architecture. In this paper, the enumeration method of "RIP-SE (Reduce Invalid Point-SE" and the method of setting the ball limit of "LSR (Layer based sphere restriction)" are proposed, which can effectively reduce the "invalid point", and increase the throughput by 2 times when the performance is very low. For the time series problem, this paper optimizes the fixed points under the condition of limiting the maximum number of search points, and puts forward the parallel calculation method of sphere limit, which reduces the final area by 50%. In this paper, the low power design method of "Graycode coding", "pipelined structure" and "gated clock" is adopted, and the power consumption is reduced by 65%. The results show that the above method is of great help to the implementation of low complexity.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN929.5

【相似文献】