针对宽带功率放大器设计的人工神经网络算法

doi:10.19665/j.issn1001-2400.2019.06.017

摘要/Abstract

摘要：

为了实现准确快速设计宽带匹配网络的目的, 提出了基于人工神经网络设计宽带高效功率放大器的新方法。通过对匹配网络及晶体管阻抗特性的分析,借助人工神经网络对功率放大器的匹配网络进行建模。结合训练模型与优化方法设计宽带匹配网络初值,在指定的频带内满足晶体管最优阻抗随频率变化的曲线。选择商用的氮化镓高电子迁移率晶体管,分别设计了六阶低通网络为输入和输出匹配网络,实现了一款工作在0.2~1.6 GHz的宽带高效率功率放大器。仿真结果表明,在0.2~1.6 GHz(轴比带比约为156%)的带宽范围内,功率放大器达到64.5%~80.5%的漏极效率,输出功率为40.0~41.6 dBm,频带内增益为11.1~12.6 dB。该方法提升了宽带功率放大器匹配网络的设计速度与准确性。

关键词: 功率放大器, 神经网络, 阻抗匹配, 宽带网络, 漏极效率

Abstract:

The artificial neural network modeling approach to designing matching networks for the broadband high efficiency power amplifiers (PAs) is proposed for the first time. The effects of input and output matching networks on the performance of PAs are analyzed. The neural network model is exploited to represent the relationship between matching networks and the optimal impedances, and the design process is combined with the optimization method. The developed ANN model allows the RF amplifier designers to realize specified matching networks conveniently and effectively. Circuit examples are used to demonstrate our proposed method. Simulation results show that a broadband high-efficiency PA is realized from 0.2 to 1.6 GHz (fractional bandwidth = 156%) with the simulated drain efficiency of 64.5%~80.5% and the output power of 40.4~41.6 dBm (10~14.5W).

Key words: power amplifiers, neural network, impedance matching, broadband networks, drain efficiency

中图分类号:

TN41

张明哲,邬海峰,魏世哲. 针对宽带功率放大器设计的人工神经网络算法[J]. 西安电子科技大学学报, 2019, 46(6): 118-124.

ZHANG Mingzhe,WU Haifeng,WEI Shizhe. Design of a broadband high-efficiency power amplifier using the artificial neural network[J]. Journal of Xidian University, 2019, 46(6): 118-124.

图/表 13

图1

图2

图3

图4

图5

表1

表2

图6

表3

图7

图8

图9

图10

参考文献 18

[1]	MENG X, YU C, LIU Y , et al. Implementation of Flat Gain Broadband Power Amplifier with Impedance Rotation Compensation[J]. IEEE Access, 2019,7(1):13304-13316. doi: 10.1109/Access.6287639
[2]	张瑛, 马凯学, 张翼 , 等. 2.5~14.5GHz分布式功率放大器设计[J] . 西安电子科技大学学报, 2018,45(1):88-92.
	ZHANG Ying, MA Kaixue, ZHANG Yi , et al. Design of a 2.5~14.5GHz Distributed Power Amplifier[J]. Journal of Xidian University, 2018,45(1):88-92.
[3]	LIU C, CHENG Q F . Analysis and Design of High-efficiency Parallel-circuit Class-E/F Power Amplifier[J]. IEEE Transactions on Microwave Theory and Techniques, 2019,67(6):2382-2392. doi: 10.1109/TMTT.22
[4]	YANG M, XIA J, GUO Y , et al. Highly Efficient Broadband Continuous Inverse Class-F Power Amplifier Design Using Modified Elliptic Low-pass Filtering Matching Network[J]. IEEE Transactions on Microwave Theory and Techniques, 2016,64(5):1515-1525. doi: 10.1109/TMTT.2016.2544318
[5]	TUFFY N, GUAN L, ZHU A D , et al. A Simplified Broadband Design Methodology for Linearized High-efficiency Continuous Class-F Power Amplifiers[J]. IEEE Transactions on Microwave Theory and Techniques, 2012,60(6):1952-1963. doi: 10.1109/TMTT.2012.2187534
[6]	MENG F, ZHU X W, XIA J , et al. A New Approach to Design a Broadband Doherty Power Amplifier via Dual-transformation Real Frequency Technique[J]. IEEE Access, 2018,6(9):48588-48599. doi: 10.1109/ACCESS.2018.2868238
[7]	王乐, 赵志鹏, 李亚楠 , 等. 多阶阻抗匹配与宽频带天线的设计分析[J]. 西安电子科技大学学报, 2018,45(5):75-80.
	WANG Le, ZHAO Zhipeng, LI Yanan , et al. Design and Analysis of Multi-tuned Impedance Matching and Broadband Antennas[J]. Journal of Xidian University, 2018,45(5):75-80.
[8]	ZHUANG Y, FEI Z X, CHEN A Q , et al. Design of Multioctave High-efficiency Power Amplifiers Using Stochastic Reduced Order Models[J]. IEEE Transactions on Microwave Theory and Techniques, 2018,66(2):1015-1023. doi: 10.1109/TMTT.2017.2750164
[9]	YARMAN B S, CARLIN H J . Simplified “Real Frequency” Technique Applied toBroad-band Multistage Microwave Amplifiers[J]. IEEE Transactions on Microwave Theory and Techniques, 1982,30(12):2216-2222. doi: 10.1109/TMTT.1982.1131411
[10]	CHEN P, MERRICK B M, BRAZIL T J . Bayesian Optimization for Broadband High-efficiency Power Amplifier Designs[J]. IEEE Transactions on Microwave Theory and Techniques, 2015,63(12):4263-4272. doi: 10.1109/TMTT.2015.2495360
[11]	ZHANG C, JIN J, NA W C , et al. Multivalued Neural Network Inverse Modeling and Applications to Microwave Filters[J]. IEEE Transactions on Microwave Theory and Techniques, 2018,66(8):3781-3797. doi: 10.1109/TMTT.2018.2841889
[12]	ZHANG Q J, GUPTA K C, DEVABHAKTUNI V K . Artificial Neural Networks for RF and Microwave Design-from Theory to Practice[J]. IEEE Transactions on Microwave Theory and Techniques, 2003,51(4):1339-1350. doi: 10.1109/TMTT.2003.809179
[13]	FENG F , GONGAL-REDDY V M R, ZHANG C, et al. Parametric Modeling of Microwave Components Using Adjoint Neural Networks and Pole-residue Transfer Functions with EM Sensitivity Analysis[J]. IEEE Transactions on Microwave Theory and Techniques, 2017,65(6):1955-1975. doi: 10.1109/TMTT.2017.2650904
[14]	ZHU L, ZHANG Q, LIU K , et al. A Novel Dynamic Neuro-space Mapping Approach for Nonlinear Microwave Device Modeling[J]. IEEE Microwave and Wireless Components Letters, 2016,26(2):131-133. doi: 10.1109/LMWC.2016.2516761
[15]	陈昀, 蔡晓东, 梁晓曦 , 等. 权重量化的深度神经网络模型压缩算法[J]. 西安电子科技大学学报, 2019,46(2):132-138.
	CHEN Yun, CAI Xiaodong, LIANG Xiaoxi , et al. Compression Algorithm for Weights Quantized Deep Neural Network Models[J]. Journal of Xidian University, 2019,46(2):132-138.
[16]	MENG X, YU C, LIU Y , et al. Design Approach for Implementation of Class-J Broadband Power Amplifiers Using Synthesized Band-pass and Low-pass Matching Topology[J]. IEEE Transactions on Microwave Theory and Techniques, 2017,65(12):4984-4996.
[17]	JIA P, YOU F, HE S , et al. A 0.25-1.25-GHz High-efficiency Power Amplifier with Computer-aided Design Based on Optimized Impedance Solution Continuum[J]. IEEE Microwave and Wireless Components Letters, 2018,28(5):443-445. doi: 10.1109/LMWC.2018.2814483
[18]	VAI M, PRASAD S . Automatic Impedance Matching with a Neural Network[J]. IEEE Microwave and Guided Wave Letters, 1993,3(10):353-354. doi: 10.1109/75.242258

变量	训练数据			测试数据
变量	最小值	最大值	数目/个	最小值	最大值	数目/个
L₁/nH	0.2	6	10 000	0.5	5.5	1 000
L₂/nH	0.2	6	10 000	0.5	5.5	1 000
L₃/nH	0.2	6	10 000	0.5	4.5	1 000
C₁/pF	0.2	5	10 000	0.5	4.5	1 000
C₂/pF	0.2	5	10 000	0.5	4.5	1 000
C₃/pF	0.2	5	10 000	0.5	4.5	1 000

输入阻抗(Z_in)	神经元数/个	训练误差/%	测试误差/%
Z_in@0.3GHz	14	0.03	0.04
Z_in@0.5GHz	14	0.13	0.13
Z_in@0.7GHz	20	0.17	0.16
Z_in@1.0GHz	26	0.36	0.35
Z_in@1.2GHz	32	0.51	0.51

序号	神经网络初值						反向优化后结果
序号	L₁/nH	C₁/pF	L₂/nH	C₂/pF	L₃/nH	C₃/pF	L₁/nH	C₁/pF	L₂/nH	C₂/pF	L₃/nH	C₃/pF
1	5.79	0.67	2.13	1.55	1.94	2.79	4.81	0.86	1.05	1.13	1.21	1.32
2	0.53	3.66	4.60	2.15	4.91	2.23	5.03	0.85	1.05	1.94	1.98	0.85
3	0.96	1.68	2.53	1.29	3.19	3.62	4.25	0.85	1.38	0.89	1.56	1.69
4	2.70	3.65	3.94	1.10	3.68	1.67	5.42	3.52	3.77	1.06	3.82	0.99
5	4.90	0.89	0.95	2.02	3.02	1.19	4.72	0.89	1.05	1.94	3.16	1.06