【作者】 金龙；

【导师】 阮成礼；

【作者基本信息】 电子科技大学 ， 电磁场与微波技术， 2006， 博士

【摘要】 本文是对W频段接收机前端进行研究。受器件和工艺的影响，该系统采用混合集成的方法研制。接收机前端的任务是将天线送来的92-96GHz高频信号变频到中频8-12GHz，要求系统有较高的灵敏度和频率稳定度。系统主要由高频带通滤波器、本振信号源和亚谐波混频器组成。文中对构成电路的鳍线及其不连续性和构成系统的三个部分分别做了研究。 编制了一套用谱域法和经验公式法分析鳍线的计算软件。提出了采用径向基函数神经网络作为鳍线不连续性的网络模型。通过电磁仿真软件分析鳍线不连续性，获得神经网络需要的样本数据。鳍线不连续性的物理尺寸和频率作为输入变量，不连续性等效电路元件参数作为神经网络的输出变量，经过对神经网络的训练，最后可以由神经网络快速、有效地获得鳍线不同槽宽的跳变、凹槽和凸带的等效电路参数。经过训练后的网络可以模拟鳍线不连续性的网络特性，计算速度比直接用电磁仿真软件分析要快得多，使之能用于含有鳍线不连续性的电路设计和优化。 对于带通滤波器，采用E面滤波器的方式实现，首先用传统的模式匹配法设计了用于本振滤波的U频段滤波器，所设计的滤波器经现有商业软件的仿真，性能都符合要求。提出了可以用于复杂结构的神经网络模型综合方法设计E面滤波器，并提出了用反转神经网络输入输出变量的网络模型以简化滤波器的综合设计，避免了一次求解方程获得物理尺寸的过程，并以该方法设计了用于射频滤波的W频段E面带通滤波器，测试结果跟预期的结果符合的比较好。 本振源是采用高稳定的微波锁相，然后倍频到毫米波的方式实现。微波锁相源采用数字锁相集成电路PE3236实现。压控振荡器输出10.5GHz的微波信号经过HEMT器件倍频器获得所需要的42GHz信号，信号经过毫米波单片功率放大器以获得推动混频器的本振功率。输出采用对极鳍线的波导微带过渡，通过仿真认为影响过渡性能最重要的参数是对极鳍线余弦过渡线的长度。通过测试计算研制的本振源输出功率达到19dBm。 亚谐波混频器采用单脊鳍线的方式实现。先分析了肖特基势垒二极管对的特性，然后通过分别独立出射频网络、本振网络和中频网络的方法设计该混频器。即先用大信号方法设计本振网络，使之匹配混频用的肖特基二极管对，用谐波平更多还原

【Abstract】 The W-band receiver front-end is developed in this dissertation. The system is realized using hybrid integrated circuits because limited by the device and manufacture technique. The mission of the receiver front-end is converting the signal with frequency from 92 to 96 GHz, which received by the antenna to intermediate frequency signal with frequency from 8 to 12 GHz. The receiving sensitivity and frequency stability are important. The system consists of radio frequency band pass filter (BPF), local oscillator (LO) and sub-harmonic mixer. Finlines, finline discontinuities and the three parts as mentioned above are discussed respectively.A software which calculates the characteristic impedance and propagation constant of finline is developed. The radial basis network is used as the finline discontinuities EM neural network models. EM software anlysis is employed to characterize finline discontinuities. EM neural network models are then trained using physical parameters and frequency as inputs and equivalent electric circuit element parameters of finline discontinuities as outputs. Once trained , the EM neural network models can simulate equivalent electric circuit element parameters of finline step, notch and strip very fast and efficiently. These trained models allow for circuit design, simulation, and optimization. Circuit simulated with these models.The BPFs are realized using E-plane filter. The U band filter is designed by mode-matching method, and simulating result given by commercial simulation software achieves the aim. A method for synthesis filter with the EM ANN models, which can be used in complex structure, is given to design E-plane filter. The EM ANN model with exchange the input/output is given to simplify the process of synthesis filter. The method can obtain the actual dimension avoiding solved equations. A E-plane BPF that uses in the RF network is developed by this method, and the measured results coincide with the simulated.The local oscillator source is realized by a method, which a high stability microwave phase locked loop oscillator is achieved, then the microwave signal is converted to millimeter wave by a quad-multiplier. The microwave locked phase更多还原