论文部分内容阅读
低压差线性稳压器(LDO)在电磁干扰影响下会发生不同程度的性能受损,进而影响到整个系统的电磁兼容性能。为解决这一问题,提出了一种基于误差反向传播(BP)神经网络的建模方法,并使用遗传算法优化网络初始权值与阈值矩阵。采用直接功率注入法设计电路板,在100 MHz~1 GHz频率范围、-15~25 d Bm W功率范围内对LDO进行电磁干扰注入实验;采样LDO的输出作为训练数据,对不同结构的BP神经网络预测性能进行对比,选取合适的网络结构,进而构建LDO的电磁干扰损伤模型。从多个角度使用模型预测了电磁干扰对LDO输出数据和传导电磁敏感性的影响,并进行实验验证;最终采用该模型预测了LDO的传导电磁敏感度,并对比分析了模型预测数据和实验测试数据。结果表明,在100 MHz~2 GHz频率范围内,模型仿真输出与LDO测试输出的最大相对误差<8%,模型仿真所得电磁敏感度与实验测试数据的最大相对误差<9%。
Low Dropout Linear Regulator (LDO) under the influence of electromagnetic interference will have varying degrees of performance damage, thus affecting the overall system of electromagnetic compatibility. To solve this problem, a modeling method based on error back propagation (BP) neural network is proposed, and genetic algorithm is used to optimize the network initial weight and threshold matrix. The circuit board was designed by direct power injection method, and the experiment of electromagnetic interference injection was carried out in the power range of 100 MHz ~ 1 GHz and -15 ~ 25 d Bm W. The output of sampled LDO was used as training data to measure the BP neural Network prediction performance comparison, select the appropriate network structure, and then build LDO electromagnetic interference damage model. The effects of electromagnetic interference on the output data and conducted electromagnetic susceptibility of LDO were predicted from several perspectives and verified by experiments. Finally, the model was used to predict the conducted electromagnetic susceptibility of LDO, and the model predictive data and experimental tests data. The results show that the maximum relative error between model simulation output and LDO test output is less than 8% in the frequency range from 100 MHz to 2 GHz, and the maximum relative error between electromagnetic susceptibility and experimental test data is <9%.