Implementation Of 5G-NR Detection Technology Based On Vector Processor

The development of the fifth-generation mobile communication technology is becoming more and more mature,and it has begun to be commercialized at this stage.Massive multiple-input multiple-output technology has become one of the key technologies of 5G mobile communication system due to its good performance in improving the channel capacity,transmission rate and anti-interference ability of communication systems.However,the application of large-scale antenna technology makes a large number of user signals superimpose each other during the transmission process,which greatly increases the signal processing difficulty of the uplink receiver.How to quickly and correctly detect all user signals has become a research hot spot in the field of mobile communications at home and abroad.In order to meet the requirements of 5G systems for ultra-low latency and high reliability,this dissertation conducts in-depth research on the signal detection technology of uplink receivers to find signal detection algorithms suitable for 5G systems.This dissertation first introduces the basic principles of signal detection technology,mainly studies linear detection algorithms such as matched filter detection,zero-forcing detection and minimum mean square error detection,as well as nonlinear detection algorithms,such as maximum likelihood detection,serial interference cancellation detection,and interference cancellation based on QR decomposition.Through theoretical analysis and algorithm simulation,the algorithm complexity analysis and detection performance comparison of the above MIMO detection algorithms are carried out.The analysis results show that,compared with the nonlinear detection algorithm,the linear detection algorithm has the advantages of low implementation complexity and low detection delay,and can also obtain good detection performance in massive MIMO systems,so it is more suitable for 5G systems.For the high-order matrix inversion problem caused by large-scale antennas,this dissertation introduces the exact inversion method based on matrix inversion lemma and Cholesky decomposition,and the approximate solution method based on iterative method.It is verified that the minimum mean square error detection algorithm based on the iterative method can obtain the detection performance approaching the detection algorithm based on the exact inversion with a small number of iterations under the condition that the scale ratio of the transceiver antenna is large.This dissertation uses the CEVA-XC4500 platform as the development platform to complete the realization of the signal detection technology based on the vector processor.Firstly,the hardware resources and overall architecture of the platform are introduced,and then two parallel processing schemes of the minimum mean square error detection algorithm based on vector processor are designed according to the operation mechanism and execution characteristics of the vector computing unit.In order to reduce the detection delay of the signal detection module,the calculation process of the algorithm is adjusted and split,the algorithm data processing sequence is optimized,and the correlation between the data is reduced,so as to make full use of the platform hardware resources and speed up the overall operation speed of the algorithm.Finally,the implementation scheme is simulated and verified.The results show that the MMSE detection parallel implementation scheme proposed in this dissertation is correctly designed and has good detection performance.The parallel processing of data and instructions greatly improves the data processing capability of the communication system and effectively reduces the detection delay.