MIMO Receiver Based On Stochastic Computing

In order to adapt to the increasing number of wireless terminals and the increasing communication capacity,the development of Multiple Input Multiple Output(MIMO)communication system is imperative,and the design of receiver is very important.In this paper,MIMO receiver is discussed,and a MIMO receiver combined with probabilistic calculation is proposed to improve system performance or reduce hardware complexity or power consumption.Stochastic computing can replace arithmetic unit with logic unit to reduce the cost.This paper mainly studies the following three parts: channel estimation,channel equalization and soft demodulation.The randomness of wireless communication channel is higher,and it is easy to be affected by transmission path,resulting in path loss and multipath effect.The Least Squares Method(LS)and the Minimum Mean Square Error(MMSE)algorithms commonly used in traditional channel estimation are simple in structure and easy to implement.But for MIMO system,which contains complex decomposition of matrix and matrix inversion,this part of the hardware overhead is very large.In this paper,using pulse array structure to calculate the special matrix multiplication,and unit upper triangular matrix LDLT decomposition inversion process.The pulsating array has the advantages of simple wiring and parallel data processing,thus reducing the cost of resources.In order to combat fading and solve Inter Symbol Interference(ISI),channel equalization is also a very important part in receiver design.Methods commonly used include Zero Forcing(ZF)and MMSE forcing,etc.These two methods not only have low precision but also high hardware complexity.Precision in this paper,based on the gradual optimization method the stochastic computing of MIMO equalization,mainly including Adaptive Scaling(Adaptive Scaling Algorithm,ASA)and opposing variable method(Antithetical Variables,AV)these two methods,reduce the logic gate overhead,shortening the critical path,This improves hardware efficiency.In addition,in order to improve channel reliability,it is usually necessary to add encoding and decoding modules,which are usually cascaded after demodulation.For the decoder,soft information is needed for decoding,that is,the Log Likelihood Ratio(LLR)of each bit of information.Generally,LLR is calculated by Maximum a Posteriori Probability(MAP).Because it contains too many logarithms and exponents,it is simplified by using Max-Log-Map.But contains Euclidean distance calculation and minimum search cause huge amount of calculation,especially for high-order modulation,computational complexity is exponential rise.Aiming at this problem,this paper based on the separation of I-Q soft demodulation,converts the calculation of LLR nonlinear function calculation.Combined with Stochastic computing,this paper proposes a multi-thermal code encoding method to realize soft demodulation with low power consumption by reducing bit turnover rate and sharing coding module.