Design And Implementation Of LDPC Decoding Algorithm For 5G System

With the increasing demand for high-speed services in the fifth generation of mobile communication systems(5G),new channel coding schemes are expected to provide significant performance gain for 5G key performance parameters.The 3rd Generation Partnership Project(3GPP)needs to evaluate coding gain,rate-compatible,hardware implementation,latency and throughput in the discussions of channel coding techniques of 5G New Radio(NR).At the RAN1#87 conference,the Low Density Parity Check(LDPC)codes with various advantages become the channel coding scheme of the data channel to replace Turbo codes for the enhanced Mobile Broadband scenario,and then,the encoding and decoding algorithms and decoder design of LDPC codes have become the focus,they have wide application values and significant economic benefits.This thesis summarizes the development of LDPC codes and analyzes the advantages and characteristics of LDPC codes in the evaluation of channel coding schemes in 5G technical specifications.The definition and coding method of LDPC codes are introduced based on the check matrix and graph model of LDPC codes.The parameters affecting the error correction performance are analyzed,and the structural characteristics of the block LDPC codes benifiting hardware implementation are summarized.This thesis studies the special structure of the check matrix for 5G NR LDPC codes and discusses the generation method.In addition,the advantages of flexible scalability and rate compatibility of 5G LDPC codes are analyzed.This thesis focus on the Belief Propagation(BP)algorithm of LDPC,and discusses the iterative decoding processes of BP algorithm with different measures based on the Tanner graph.In order to reduce complexity,the evolutions of approximation algorithms such as Min-Sum algorithm and Normalized Min-Sum algorithm are discussed,and then the low complexity Density Evolution single minimum Min-sum(DE-smMS)algorithm was proposed.The proposed algorithm combines the density evolution theory with the weighted average scheme to correct the normalization factor in the iterative process and then caculates a single minimum value to reduce the complexity.The simulation results show that the proposed algorithm has a performance gain of about 0.2dB compared with the existing single minimum algorithm,especially,when the code rate is low,the proposed algorithm can achieve the same performance as the NMS algorithm.The proposed algorithm achieves a better trade-off between complexity and decoding performance.Based on the block structure characteristics of 5G NR LDPC codes,the corresponding Layered scheduling scheme is proposed.The design and implementation of the transmitter and receiver of the physical downlink shared channel were completed based on TMS320C6678 DSP platform,and the design was verified to be consistent with the technical specification of 5G NR on signal&spectrum analyzer of Rohde&Schwarz.The implementation of LDPC decoder is completed,and the implementation architecture of a parallel decoder is proposed for the proposed DE-smMS algorithm.The 5G Physical Downlink Shared Channel is implemented on the DSP to analyze the error correction performance and the latency of the decoder.The parallelism of the decoder is implemented on the C6678 with the OpenMP framework based on multi-core shared memory.The results show that the correction performance of the proposed algorithm approximates the traditional NMS algorithm and the latency is reduced by about 12%.