基于神经网络的宽带功放动态非线性行为建模

文章主要讨论了如何利用神经网络对宽带功放进行动态非线性行为建模的问题。首先简述了功放的动态非线性特性及行为建模的方法。然后回顾了基于实数时延前馈神经网络、径向基函数神经网络等浅层神经网络构建的功放动态非线性行为模型。在此基础上,针对5G/6G宽带功放具有更强的记忆效应的问题,重点分析了如何使用长短期记忆(LSTM)神经网络对功放的动态非线性进行精确的行为建模。最后展望了构建具有普适性的功放非线性行为模型将是5G/6G通信时代功放非线性建模的一个重要发展方向。     

一种新的用于有记忆功放的查找表预失真器

 针对有记忆非线性功放提出了一种新的基于查找表的记忆型预失真器.该记忆型预失真器采用无记忆预失真器的级联扩展,其横向滤波器结构相比记忆多项式方法降低了复杂度,却能够得到与其相近的线性化效果.并且针对查找表固有的量化误差大和自适应收敛速度慢的缺点,分别采用线性内插和加权窗迭代更新的方法加以改进.在非线性功放的记忆多项式模型下,通过多载波WCDMA宽带信号和OFDM宽带信号验证了基于查找表的记忆型预失真器的良好线性化效果和改进方法的有效性.     

基于广义LSTM神经网络的宽带射频功放数字预失真线性化

在无线通信系统中,射频功放的非线性是信号失真与频谱增生的主要原因,尤其是对于采用64QAM、256QAM 等高峰均功率比的复杂调制系统,对射频功放线性度的要求越来越高;然而宽带射频功放中存在的强记忆效应严重地降低了基于传统非线性模型的数字预失真器的线性化性能。文章提出广义长短期记忆(LSTM)神经网络模型,通过输入的时序特性,从时间轴上进行模型迭代,利用LSTM模型独特的长短时序结构以更好地表征宽带射频功放的记忆效应,同时引入时间超前项以构建广义的LSTM模型,进一步增强其动态非线性建模能力。在不同超参数下的建模结果表明,该模型的归一化均方误差(NMSE)指标可达-42.2895 dB。最后,使用20 MHz 带宽的4 载波WCDMA信号,对中心频率1900 MHz 的50 W Doherty 功放进行预失真线性化实验验证。实验结果证实了基于广义LSTM神经网络模型的数字预失真器可以使互调分量降低达23.27 dB,大大优于记忆多项式等传统非线性模型的非线性校正性能。     

改进型并行双非线性模型的ROF 系统非线性建模

包含光载无线通信(ROF)链路和射频功放的ROF 前传系统存在严重的静态非线性和记忆效应,提出了改进型并行双非线性广义记忆效应模型(APTGMEM),以准确建模ROF 前传系统的强静态非线性和记忆效应。所提出的APTGMEM 包含无记忆和广义记忆效应两个独立部分,其中无记忆部分只考虑高阶静态非线性,广义记忆效应部分只考虑低阶次超前广义非线性。文中采用带宽为20 MHz 的4G LTE(Long Term Evolution 长期演进)信号为测试信号,光模块和峰值输出功率为51dBm 的Doherty 功放被用于模型有效性验证。测试结果表明,APTGMEM 相比广义记忆多项式模型(GMPM),模型系数减少40%,而建模精度相似,同时比并行双非线性两厢模型(PDNTBM)建模精度提高了3. 6 dB,其在建模精度、系数数量和训练复杂度上取得了良好折中。     

适用于宽带通信系统的 DHW功放模型

为有效模拟宽带通信系统中功放的非线性行为,研究了一种双路 Hammerstein-Wiener（DHW）功放模型。分别施加相同的激励信号于该模型的Hammerstein分支和Wiener分支,并将两分支输出信号累加作为最终的响应。利用NXP功放管进行Doherty电路设计并加入多载波WCDMA信号驱动,从ADS中导出强非线性输入输出数据进行模型实验测试。结果显示阶数相同条件下,与 Hammerstein和Wiener模型相比,此双路 Hammerstein －Wiener功放模型能够获得更好的RMSE 。另外,该功放模型可在减少整个记忆多项式参数数量30％～40％的前提下,获得与记忆多项式模型近似相同的性能,对功放系统行为模型的研究具有重要的参考价值。     

基于改进型径向基函数神经网络的功放线性化

提出了一种基于改进型径向基函数神经网络(MRBFNN)的数字预失真线性化模型,用于更为精确地矫正宽带射频功率放大器的动态非线性。该神经网络模型的输入层使用传统的延时抽头以补偿功放的线性记忆效应,同时对每个抽头进行级数展开用于补偿功放的非线性记忆效应,从而更好地抑制功放的动态非线性失真。文中使用WCDMA 三载波信号对一个460MHz 的Doherty 功率放大器进行数字预失真线性化实验。实验结果表明,与传统数字预失真线性化模型相比,基于改进型径向基神经网络的数字预失真线性化模型能更好地抑制宽带功放动态非线性引起的带外频谱再生,其三阶互调(IMD3)失真最多可以抑制23dB,大大提高了功放的线性度,验证了所提出的数字预失真线性化模型的有效性。     

基于时间功率混合分段模型的磁共振功放非线性建模

为了能够对磁共振功放的强分叉非线性进行精确建模,提出一种时间功率混合分段模型。时间功率混合分段模型是针对磁共振功放的激励为辛格脉冲信号时的精确非线性模型。所提出的模型基于辛格脉冲信号的特点进行时域和幅度域分段,从而能够对磁共振功放的分叉非线性进行精确建模。采用辛格脉冲信号作为测试信号,测试信号的占空比为5%,设计输出峰值为18 kW 的1.5 T 磁共振功放(工作频率63.89 MHz)作为测试功放进行模型的有效性验证。实测磁共振功放和模型的AM/AM 和AM/PM 曲线的比较结果表明,时间功率混合分段模型能够对磁共振功放的分叉非线性进行精确建模。同时,所提出模型的归一化均方误差(NMSE)比记忆多项式、广义记忆多项式模型和时分多项式模型分别提高10.7 dB, 10.3 dB 和14.7 dB。     

宽带卫星功率放大器失真补偿技术研究进展及应用

针对宽带卫星通信系统中高功率放大器的记忆非线性失真问题,对不同补偿技术进行 了综述,重点研究了预失真的学习结构、数据预失真、信号预失真、时频域预失真和自适应 均衡等补偿技术并对它们进行了相应的优缺点比较,最后提出了宽带卫星功放实际应用环境 下的4种补偿方案,为我国宽带卫星功放失真补偿研究提供理论参考。     

一种基于模糊辨识的OFDM系统预失真器设计

将有记忆的高功率放大器(HPA)描述成一个线性动态系统级联一个无记忆非线性系统的Wiener系统模型,由Sugeno模糊模型自适应跟踪HPA的无记忆非线性曲线,通过离线过程设计OFDM系统数字预失真器。为了消除HPA记忆效应,首先在频域得到信号的均衡补偿,然后通过傅里叶逆变换后得到时域记忆性补偿。该预失真器具有收敛速度快、系统稳定、补偿效果好的特点。     

改进的动态记忆多项式功放模型及预失真应用

针对无线通信系统中射频功放的强非线性、强记忆效应特征,提出一种改进的多支路组合模型——改进的动态记忆多项式(G-DMP)模型。该模型以动态记忆多项式模型为基础,引入记忆时刻输入信号包络对当前输入信号的影响项和当前输入信号包络对记忆时刻输入信号的影响项,更灵活地对功放强记忆效应建模。同时采用改进的RLS_LMS联合算法对G-DMP模型进行自适应预失真系统仿真。结果表明:G-DMP模型比DMP模型精度优于0.5dB时,系数减少14.29%,邻信道功率比(ACPR)改善了4dB,由此可知,G-DMP模型能获得更好的建模精度和线性化效果。     

A vector intermodulation analyzer applied to behavioral modeling of nonlinear amplifiers with memory

A large signal vector intermodulation network analyzer with a dynamic range of 90 dB and phase resolution of better than 2/spl deg/ is reported. The analyzer is used in conjunction with a multislice behavioral model to characterize memory effects in three different RF power amplifiers: an MOSFET instrumentation amplifier, a multistage GaAs/silicon-based broadband microwave integrated-circuit amplifier, and an SiGe HBT monolithic-microwave integrated-circuit amplifier. The multislice behavioral model architecture builds on conventional single-tone AM-AM and AM-PM modeling extended to capture long-term memory effects that are characterized by asymmetric intermodulation distortion (IMD). Phase asymmetries of upper and lower IMD are captured. A systematic procedure for extracting the model is presented.     

功放数字预失真技术现状及未来发展

功放是无线通信发射机的重要组成部分,随着无线通信的不断发展,对宽带线性功放的要求也越来越高。然而,功放的非线性特性和记忆效应会导致信号严重失真。为了减小带内失真和邻道干扰,数字预失真成为功放线性化技术中最有发展潜力的技术。介绍了功放模型,数字预失真的学习结构、数字预失真器参数的辨识算法等研究现状,探讨了数字预失真的发展趋势,为今后数字预失真技术的研究提供参考。     

Wide-band dynamic modeling of power amplifiers using radial-basis function neural networks

A radial-basis function neural network (RBFNN) has been used for modeling the dynamic nonlinear behavior of an RF power amplifier for third generation. In the model, the signal's envelope is used. The model requires less training than a model using IQ data. Sampled input and output signals were used for identification and validation. Noise-like signals with bandwidths of 4 and 20 MHz were used. The RBFNN is compared to a parallel Hammerstein (PH) model. The two model types have similar performance when no memory is used. For the 4-MHz signal, the RBFNN has better in-band performance, whereas the PH is better out-of-band, when memory is used. For the 20-MHz signal, the models have similar performance in- and out-of-band. Used as a digital-predistortion algorithm, the best RBFNN with memory suppressed the lower (upper) adjacent channel power 7 dB (4 dB) compared to a memoryless nonlinear predistorter and 11 dB (13 dB) compared to the case of no predistortion for the same output power for a 4-MHz-wide signal.     

A Gaussian Approximation of High-Order Distortion Spectrum in Broadband Amplifiers

A novel analytical Gaussian approximation is developed for the evaluation of the distortion spectrum introduced by a nonlinear amplifier. This method allows to consider high-order distortion contributions when the device is driven by a broadband signal with Gaussian amplitude distribution. The results are applied to a ninth-order power series model based on well known single-tone and two-tone analysis parameters. The cascade of two or more amplifiers is investigated as well providing a complete set of tools for the early system specification of broadband transceivers. Simulations shows excellent agreement even for high input power levels, when former third-order and fifth-order approximations fail to yield accurate predictions.      

A data predistortion technique with memory for QAM radio systems

The authors present an efficient data predistortion technique with memory for compensation of high-power amplifier (HPA) nonlinearities in digital microwave radio systems employing quadrature amplitude modulation (QAM) signal formats. A practical implementation method is described which trades off performance against complexity and which makes it possible to implement this kind of predistorter in 256-QAM, and higher-level QAM systems. Using the 16-, 64-, and 256-QAM signal constellations, it is shown that the proposed technique achieves a considerably higher performance than that of conventional memoryless data predistortion of the predistortion technique with memory based on finite-order inverses of nonlinear systems. Specifically, numerical results show that the proposed technique achieves a gain that is in excess of 2 dB over conventional memoryless data predistortion     

Optimization of Third-Order Intermodulation Product and Output Power from an X-Band MESFET Amplifier Using Volterra Series Analysis

A third-order analysis for accurately predicting large-signal power and intermodulation distortion performance for GaAs MESFET amplifiers is presented. The analysis is carried out for both single- and two-tone input signals using the Volterra series representation and is based only on small-signal measurements. Simple expressions for the nonlinear power gain frequency response, the output power, the gain compression factor, and the third-order intermodulation (IM/sub 3/) power are presented. The major sources of gain compression and intermodulation distortion are identified. Based on the developed nonlinear analysis in conjunction with the device nonlinear model, a systematic procedure for designing a MESFET amplifier under large-signal conditions for optimum output power and IM/sub 3/ performance is proposed. The method utilizes out of band computed matching compensation through a nonlinear model of the amplifier. The accuracy of the device large-signal and IM/sub 3/ distortion characterization and the practicability of the proposed method are illustrated through comparison between measured and predicted results.