Analysis Of Signal Detection Algorithm In MIMO System And Optimized Design Of Area

The combination of multiple input multiple output(MIMO)and orthogonal frequency division multiplexing(OFDM)technologies improves the data transmission rate and system capacity.However,the high-performance signal detection algorithm of the MIMO-OFDM system has high complexity.In practical applications,the area-optimized signal detection design has practical engineering value.We mainly model and analyze the signal detection algorithm of the 2×2 MIMO system,supporting 256 Quadrature Amplitude Modulation(QAM)and 80 M bandwidth.Perform fixed-point simulation of minimum mean square error(MMSE),maximum likelihood(ML)and approximate maximum likelihood signal detection algorithms based on QR orthogonal triangle decomposition in different channel environments,and compare packet error rates for different signal-to-noise ratios.We found that the error performance of the approximate ML algorithm based on QR decomposition is close to the ML algorithm.Then,the area-optimized MMSE signal detection pipeline architecture design and sequential circuit design are made.By simplifying the matrix,the resources of the multiplier are reduced,and numbers with a large range of numerical values are represented by mantissas and order codes,which reduces the bit width of the multiplier.Then,design a pipeline architecture for time-multiplexing hardware resources to achieve area-optimized approximate ML signal detection and softly outputs the Bit Log Likelihood Rate(BLLR)of different subcarrier modulation methods.We propose to reduce the area by reducing the number of distance calculation units and the multiplier resources of the distance calculation unit.Considering the error caused by the unstable value of QR decomposition,we propose to use a single register to detect the two signals of each subcarrier to mark whether it is necessary to traverse all the constellation points to improve the performance of approximate ML signal detection.In order to reduce the implementation complexity of square root and reciprocal operation and improve the accuracy,we propose a 12-bit mantissa lookup table implementation with input greater than 1 and less than 2.After that,we conduct VCS simulation and verification on various input conditions of MMSE and approximate ML based on QR decomposition for two signal detection scheme modules.In the end,the data achieved by register transfer level(Register Transfer Level,RTL)is completely the same as the algorithm data realized by C language.Under the22 nm process,the area of MMSE signal detection evaluated by the comprehensive tool Design Compiler is27329.55,and the area of the approximate ML signal detection module based on QR decomposition is 857753.53.The area-optimized MMSE signal detection module requires 46 real multipliers;the area-optimized approximate ML signal detection module requires 524 real multipliers.At 320 MHz clock frequency,the approximate ML signal detection throughput rate is 710.5Mb/s.Finally,Our design meet the requirements.This design provides a feasible solution for area-optimized MMSE signal detection and approximate ML signal detection,which can realize an area-optimized signal detection circuit.It also proposes a signal detection scheme that combines the MMSE algorithm and the approximate ML algorithm.The packet error rate performance is better than the MMSE signal detection,which is inferior to the approximate ML signal detection,and the complexity is tremendously reduced.The multiplier resource of the approximate ML signal detection is reduced to 276;For the 3/4 coding rate,the throughput with a packet error rate of 1% reaches 724.4 Mb/s.