摘要
提出了一种基于广义记忆型神经网络（GMNN）的数字预失真器非线性模型,以更好地抑制由于射频功放动态非线性导致的带内失真以及带外频谱扩展等问题。通过引入时间上的超前项,使得功放模型的记忆效应建模能力得以扩展,通过添加高阶非线性级数,使得功放非线性建模精度进一步提高。文中使用带宽为20 MHz的4载波WCDMA信号作为测试信号,对一个中心频率为460 MHz的60W Doherty射频功放进行数字预失真线性化实验。实验结果表明,广义记忆型神经网络数字预失真器的带外抑制可达19 d B,能更有效地抑制射频功放的带外频谱扩展,相比于其他几种预失真器展现出更好的线性化效果,验证了广义记忆型神经网络数字预失真器的有效性。
This paper proposes a digital predistorter nonlinear model based on a generalized memory neural network（ GMNN）,so as to suppress in-band distortion and out-of-band spectrum expansion caused by the dynamic nonlinearity of a radio frequency power amplifier（ RFPA）. By adding some leading terms in time domain,the memory effect modeling capabilities of the PA model is extended. By adding high-order nonlinear series,the accuracy of the power amplifier nonlinear model is further improved. A 4-carrier Wideband Code Division Multiple Access（ WCDMA） signal with 20 MHz bandwidth is applied to a 460 MHz 60 W Doherty RFPA for experimental validation of the digital predistorter. The experimental results illustrate that the out-of-band suppression of the GMNN predistorter can be up to 19 d B. So the predistorter can more effectively suppress the out-of-band spectral spread of an RFPA and has better linearization performance than other digital predistorters.The experimental results verify the effectiveness of the GMNN digital predistorters.
作者
尹思源
刘太君
叶焱
许高明
杨东旭
YIN Si-yuan1, LIU Tai-jun1, YE Yan1, XU Gao-ming1,YANG Dong-xu2(1. Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo 315211, China; 2. Tourism College of Zhejiang, Hangzhou 310013, Chin)
出处
《微波学报》
CSCD
北大核心
2018年第2期47-50,共4页
Journal of Microwaves
基金
国家自然科学基金项目(61571251,61501272)
浙江省公益技术应用研究项目(2015C34004,2016C34003)
关键词
广义记忆型神经网络
射频功放
数字预失真器
非线性模型
generalized memory neural network, radio frequency power amplifier, digital predistorter, nonlinearmodel
