Research On Digital Predistortion Technology Based On QRD-RLS

As the key equipment in the wireless communication transmitter,power amplifiers are responsible for amplifying power of small signals and increasing its transmission distance.However,the inherent non-linear characteristics of power amplifiers will cause the spectrum of transmitted signals to broaden and seriously interfere with the normal communication of adjacent channel signals.To solve this problem,this paper proposes a digital predistortion(DPD)technology based on QR decomposition recursive least squares(QRD-RLS)method,which can not only effectively reduce out-of-band spectrum leakage,but also improve the calculation efficiency of predistortion parameters.The main work consists of following three parts:The theory of QRD-RLS algorithm is analyzed.First of all,this article focuses on the mathematical analysis of the least squares algorithm and recursive least squares algorithm in digital predistortion parameter learning.Then a recursive least squares algorithm based on QR decomposition is proposed,and the systolic array learning structure is used for parallel calculation.Finally,the CORDIC calculation unit is used to implement vector rotation,which avoids division and square root operations in Givens rotation.The digital predistortion link based on QRD-RLS is simulated.First,the predistortion parameter learner is designed in detail,which mainly includes the kernel combination module,the systolic array learning module,and the Gaussian elimination module.Then in order to verify the correctness of the predistortion parameters calculated by the QRD-RLS algorithm,an indirect learning structure is used to build a DPD floating-point simulation link.Finally,a DPD fixed-point simulation link is built to evaluate the influence of the fixed-point number on the predistortion performance,and to calculate the error range between the floating-point output and the fixed-point output of each module.The digital predistortion based on QRD-RLS is experimentally verified.First,the top-level structure of each sub-module in the predistortion parameter learner is designed,and the detailed configuration of the internal key intellectual property core is given.Then based on the software radio experimental platform,the digital predistorter and the predistortion parameter learner are implemented by digital logic.Finally,a closed-loop learning method is used to build a QRD-RLS-based digital predistortion system,and the predistortion performance test is performed.When the baseband signal sampling rate is153.6Msps,for a 16 QAM signal with a 20 MHz bandwidth,its out-of-band leakage can be reduced from-38 d B to-49 d B by the proposed method.This paper studies a digital predistortion technology based on QRD-RLS,and verifies the correctness of the proposed algorithm through theoretical analysis,simulation experiments,and physical testing.The research results can adaptively adjust predistortion parameters according to input and output signals of power amplifiers,which is suitable for closed-loop digital predistortion systems with FPGA or ASIC as the computing core.