摘要: |
针对当前数字预失真系统的软核处理能力较差、速度较慢等问题,提出了一种基于Zynq和AD9361联合预失真的架构。该架构具有高集成度、高计算效率的特点。功率放大器的数字模型采用记忆多项式模型,并采用新提出的基于查找表和卷积模块的硬件实现方法,更易于工程实现。自适应算法采用归一化最小均方误差算法,结合正交频分复用技术调制信号,分析、仿真和实验证明,在记忆深度为Q、非线性阶数为K、运算单元延迟分别为P与M时,新提出的预失真器结构的硬件乘法器消耗相较于目前常用的多项式与查找表结构分别减少了Q个和2Q+3个,同时其运算周期相较于目前常用的记忆多项式与查找表实现方式分别降低了(3K-1)P与M个。通过频谱仪观察预失真后的信号,其邻信道功率比改善了10 dBc左右。 |
关键词: 数字预失真系统 记忆多项式 归一化最小均方 功率放大器 |
DOI: |
|
基金项目:重庆市技术创新与应用发展项目(cstc2019jscx-msxmX0079) |
|
A new architecture of digital pre-distortion system and its implementation |
ZHANG Hongsheng,YI Shenghong,MENG Jin,WEI Zhongyang,MA Xiaodong |
(a.School of Optoelectronics Engineering;b.Key Laboratory of Microelectronics Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China) |
Abstract: |
A joint pre-distortion architecture based on Zynq and AD9361 is proposed to address the problems of poor soft core processing capability and slow speed of current digital pre-distortion systems,which is featured by high integration and high computational efficiency.The digital model of the power amplifier adopts memory polynomial model and the newly proposed hardware implementation method of lookup table and convolution,which is more optimal for engineering implementation.The system chooses normalized least mean square(NLMS) algorithm as the adaptive algorithm.Combining with the orthogonal frequency division multiplexing(OFDM) signal,simulations,analysis and experiments prove that with the memory depth of Q,nonlinear order of K,and operation cell delay of P and M,respectively,the hardware multiplier consumption of the newly proposed predistorter structure is reduced by Q and 2Q+3 compared with the currently used polynomial and lookup table structures,respectively,while its operation period is reduced by (3K-1)P and M compared with the currently used memory polynomial and lookup table implementations,respectively.The pre-distortion signal is observed by a spectrometer,and the adjacent channel power ratio is improved by about 10 dBc. |
Key words: digital pre-distortion system memory polynomial normalized least mean square power amplifier |