Research And Implementation Of Channel Coding Technology Based On General Purpose Processor

With the establishment of the 5th generation mobile communication(5G)standard,wireless communication has gradually entered the 5G era.Compared with the previous four generations of mobile communication system,5G introduces more new technologies,including low-density parity-check codes(LDPC),Polar codes,massive multiple-input multiple output(MIMO)and other technologies.Although Turbo codes were not adopted by 5G,its excellent performance is beyond doubt.At present,most codecs are realized based on FPGA and other hardware technologies.These kinds of implementation have shortcomings such as long development cycle,inflexibility,and poor scalability.With the development of multi-core parallel technology,such as multi-threading technology and single-instruction multiple-data(SIMD),it is feasible to implement channel codecs based on general-purpose processors(GPP).In this thesis,we focus on the design and implementation of codecs based on GPP,and test the implemented software codecs on the MIMO verification prototyping platform,after the indepth study of channel coding technology.Firstly,a GPP-based MIMO verification prototyping system is introduced,which is composed of software-defined radio(SDR)nodes and multi-core GPPs.The software process of this system is to divide one data frame into many subframes and process in parallel based on multi-threading technology.Compared with the conventional prototyping system architectures,our system architecture adopts an FPGA coprocessor and a high-speed Ethernet interface to enhance the capability of real-time signal processing and data transfer between SDRs and GPPs.Secondly,we study the implementation of software turbo decoder based on GPPs.The encoder of Turbo codes is built by two identical recursive systematic convolutional(RSC)codes with parallel concatenation,and the decoder is composed of two maximum a posteriori probability(MAP)component decoders.In this thesis,a parallel decoding scheme based on SIMD instructions set is proposed by analyzing the parallel butterfly structure between Turbo codes state transitions.The test results show that the throughput of the software decoder can achieve 55 Mbps,while maintaining a good bit error rate(BER)performance.Nextly,we focus on the implementation of the LDPC code in 5G new radio(NR),which is quasi-cyclic LDPC(QC-LDPC)code with cyclic structure.Both encoder and decoder are implemented on GPPs with SIMD instructions set,and over 100 Mbps throughput is achieved on our MIMO prototyping testbed.Some scenario tests are performed and the results prove the feasibility of the proposed implementation.Time cost analysis indicates the practical use and real-time processing ability of the implemented QC-LDPC encoder and decoder.Finally,we address the SIMD-based encoder and decoder of Polar codes adopted by 5G NR.Polar codes are based on a phenomenon called channel polarization,which transforms the physical channel into virtual outer channels with a multiple recursive concatenation of a short kernel code.The common decoding algorithm of Polar codes is successive cancellation(SC)algorithm.In this thesis,we realize the Polar encoder with Intel math kernel library(MKL),whose throughput can achieve up to 133 Mbps.The Polar decoder implemented based on SIMD can achieve a throughput up to 274 Mbps with a good BER performance.