Detection Algorithm Research And Hardware Implementation Based On Asynchronous SCMA Uplink Systems

In the 5G era,with the continuous increase of connected devices,how to improve the access capability of the system has become a hot topic.As a member of non-orthogonal multiple access technologies,SCMA is an excellent candidate for the future generations of wireless communication systems.In recent years,more and more scholars have studied SCMA systems,including the design of SCMA codebooks,SCMA detection algorithms,and how to combine SCMA and OFDM.They usually assume that the signals of each user are ideally synchronized.In an actual SCMA uplink system,each user may be located in a different geographic location,and the distance between the user and the base station may change with time.In this situation,a common solution is to use the auxiliary signal to estimate the time delay parameters of each user at the receiving end,then transmit the parameters back to the sending end,and perform compensation at the sending end.However,due to the influence of system synchronization accuracy,this method is still difficult to ensure that all user signals arrive at the receiving end at the same time.Once the arrival times of the user signals are not perfectly synchronized,then we can call the system a non-ideally synchronized system.After verification in this paper,under the condition of non-ideal synchronization,the performance curve of the most commonly used Message Passing Algorithm(MPA)detection scheme cannot converge.Therefore,how to effectively complete multi-user signal detection in the SCMA system with non-ideal synchronization conditions and ensure the stability of the actual SCMA uplink communication system is a problem worth studying.Based on the actual SCMA uplink system model,this paper deduces the expression of the sampling signal,and analyzes the reasons for non-ideal synchronous multi-user access interference.In order to solve this problem,we start with the traditional MPA detection method and analyze the principle,performance and complexity of MPA,MAXLOG-MPA and Stochastic-MPA algorithms respectively.Then,on the basis of the traditional MPA detection method,a solution is given.The main innovations of this thesis include:(1)For the non-ideal synchronous SCMA uplink system,a multi-sampling point joint detection method(Sampling Selection-MPA,SS-MPA)is proposed,the algorithm adds a signal pre-processing step,obtains the best sampling point according to the different time delays of each user,expands the function nodes,and changes the update rules,so as to support the iterative update of multi-sampling points.The performance of the algorithm is achieve a huge improvement.It is consistent with the performance of the MPA algorithm under ideal synchronization conditions.It can also converge quickly under non-ideal synchronization conditions.Compared with MPA under ideal synchronization conditions,it only degrades 0.5 dB under AWGN conditions.(2)In order to reduce the complexity of the SS-MPA,this paper transforms it into the logarithmic domain and the probability domain respectively,and obtains the SS-MAXLOGMPA and the SS-Stochastic-MPA.Compared with the SS-MPA,the two obtained 40% and90% complexity gains at the cost of performance loss of 1.5 dB and 0.5 dB,respectively.(3)In order to further enhance the performance of the SS-MPA algorithm and solve the problem of its slow convergence speed under harsh channel conditions,this paper combines SS-MPA with parallel interference cancellation and proposes the SS-MPA-PIC algorithm.Under non-ideal synchronization conditions,the BER of SS-MPA-PIC is only degraded by 0.1 dB compared with ideal synchronization conditions.(4)Considering the practicability,we chose the SS-Stochasitc-MPA with the lowest complexity for hardware implementation and described the low-complexity design method of the core module in detail.The normalized hardware resource overhead of this scheme on the FPGA is only 20% of that of the traditional MAXLOG-MPA method,and the performance is only degraded by 0.3 dB compared to the floating-point performance.