基于电流的集成电路故障诊断研究

【作者】 张兰；

【导师】 徐红兵；

【作者基本信息】 电子科技大学 ， 模式识别与智能系统， 2004， 硕士

【摘要】 集成电路的故障诊断是集成电路生产中的一个重要问题。传统的故障电压诊断方法对于目前的生成工艺下的故障类型并不能有效的实现故障诊断,且大规模电路中包含冗余电路,很多冗余故障是无法检测的。新型的电流测试故障包括静态电流检测技术(IDDQ)和动态电流检测技术(IDDT)。这些检测技术是从电路电源线中的电流信号中提取电路的故障信息实现故障诊断。电流检测技术是电压检测技术的重要补充。本文对静态电流的故障定位算法以及相应的故障诊断生成算法进行了研究,结合电路的拓扑结构和静态电流信息实现定位,避免了使用故障字典,并且仿真结果说明该算法能对多故障电路实现故障诊断。在相应的诊断生成算法中,根据静态电流定位算法的特点,运用遗传算法使得静态电流响应序列差异的最大化,实现了故障的诊断生成。此外,对动态电流故障诊断算法提出特征空间和特征曲线的概念,以实现对动态电流信号中的故障信息的有效提取。先通过小波变换、主成分分析、归一化处理将电路馈电线中各类故障动态电流包含的电路信息映射到特征空间中形成故障特征曲线,然后再结合DUT的IDDT在特征空间的分布对电路进行诊断。运用SPICE对上述各种算法进行了模拟,并与文献中的算法进行了比较,仿真结果表明了算法的有效性。更多还原

【Abstract】 Fault diagnosis is a critical problem in IC chip production. It is difficult for traditional voltage-diagnosis method to detect all kinds of faults under current IC production art condition, moreover, with the progress of the VLSI, the pin number of a chip is comparatively limited that it is hard to detect the redundant fault in the chip. Recently, novel fault diagnosis technology including IDDQ and IDDT has developed rapidly. These new methods adopt the circuit information contained in the power supply line current to realize the fault diagnosis. And these methods have now been an important supplement to the diagnosis method based on voltage detection. This research is mainly concerned with the fault diagnosis using current signal. In IDDQ research, the circuit structure and the quiescent current were used to analyze and locate the fault in the chip which avoids to compiling the fault dictionary. Following work is concerned with the location algorithm’s diagnosis generation. Genetic algorithm is applied to realize this generation. Moreover, in IDDT diagnosis, the conception of fault feature curve and feature space is proposed to diagnose the fault in the analog and mixed-signal circuit. Fault signature contained in the transient current is projected into fault feature space. Preprocess technique such as wavelet transform, PCA, normalization, are used to plot the fault feature curve. The fault diagnosis is realized through comparing the distribution of the IDDT in the feature space and the fault feature curve. Simulative experiments have realized our algorithms, and we compared the simulative results with the literatures. These experiments have shown the effectiveness of our algorithms.更多还原