Research And Implementation Of The Precoding Technology Of Downlink LTE-Advanced Systems

The rapid growth of mobile communication requires wider bandwidth and higer spectral efficiency. LTE-Advanced(LTE-A) proposed by 3GPP which is evoled from LTE extends the downlink transmission to 8x8 MIMO and increases channel capacity. However due to the MIMO technology, it’s necessary to do research on the precoding to eliminate the interferences between different data streams. This paper researchs on pre-coding algorithms based on matrix decomposition, focuses on the complexity and achieves high throughput and precision 8x8 precoding.Firstly, the MIMO system model is analysed, the simulation platform and parameters based on PDSCH are briefly illustrated. According to traditional singular value decomposition(SVD), LDLH decomposition, geometric mean decomposition(GMD) and uniform channel decomposition(UCD) are analysed. BER performance and complexity of each algorithm are given at the end.SVD algorithm is the foundation of pre-coding algorithms proposed above. The key to reduce the complexity of pre-coding is to reduce the complexity of SVD. So the complexity of the pre-coding algorithm can be redused by optimizing SVD algorithm. The main three kinds of algorithms discussed are as follows: 1) QR iteration algorithm; 2) Two bilateral Jacobi algorithm; 3) faster convergence superlinear convergence algorithm, they all have defects. Afterwards, low-complexity bilateral Jacobi algorithm is proposed based on Jacobi algorithm and Hertimian matrix. The new algorithm can greatly reduce the complexity and meet different dimensions of the matrix by using the characteristics of Hertimian matrix. Then simulation and theory analysis for the proposed algorithm are given out. The results indicate that this algorithm can achieve the lowest complexity and fastest convergence in the case of same accuracy compared to the other three algorithms.The algorithm proposed in this paper is implemented by FPGA on the Xilinx XC7VX690 T platform, the matrix dimension is 8x8. Setting the threshold in the fixed-point simulation can improve the precision. In the design of hardware circuit, the parallel rotation is improved which can reduce the consumption of resources, relieve the pressure of layout and increase the clock frequency. The maximum frequency of this precoding circuit can reach 200.120 MHz and throughput rate can reach 5.99 M matrices/s, which fulfills the need for dynamic channel pre-coding.