摘要
提出了运用神经网络对模拟IC进行芯片合格分类和故障检测的方法。通过BP 型神经网络,运用误差反向传播算法,对CMOS 运算放大器输入脉冲测试信号,以正常和故障芯片供电电流的时域响应和频域响应作为样本反复训练网络。检测IC 故障实验和仿真结果都表明:BP 型神经网络可以用来有效、方便地测试模拟IC。
A new approach for detecting defects and classification in analog integrated circuits using the feed-forward neural network with the resilient error back-propagation method is presented. The experiment and simulation were performed for a CMOS operational amplifier ICs by adding input impulse test signal. Networks were trained with supply current responses of good and faulty ICs in time domain and frequency domain. Experiments and simulation results show that the BP Neural Network is a very efficient and versatile approach for testing and detection of analog circuits.
出处
《半导体技术》
CAS
CSCD
北大核心
2005年第3期41-44,40,共5页
Semiconductor Technology
关键词
模拟IC
人工神经网络
芯片检测
运算放大器
analogue IC
artificial neural networks
detection
operational amplifier
