模拟VLSI电路故障诊断的子带特征提取方法

为了降低模拟电路参数型故障的测试难度,提出了一种基于奥克塔夫(Octave)-Haar小波结构的模拟VLSI电路故障诊断方法。将测试响应经小波滤波器组完成子带滤波,随后对各子带滤波序列计算故障子序列与正常子序列的互相关系数,对每一故障,可确定出互相关系数最小的子带,并将此数值作为该故障的特征,对应子带的正常响应序列的自相关系数作为无故障特征,用故障特征与正常特征的对比可诊断故障。对国际标准电路的实验表明,该方法对参数型故障的诊断已具有高分辨率。

Subband Signature Extraction for Fault Diagnosis of Analog VLSI Circuits

An approach for diagnosing analog VLSI circuits based on Octave-Haar wavelet was presented in order to lower the difficulty level of testing analog parametric fault.Sub-band filtering of testing response signal was accomplished by wavelet filter banks.The cross-correlation coefficient(CCC) was computed to sequence acquired from sub-band filtering.Consequently,for each fault,subband with lowest CCC value was determined and this value could be used as faulty signature for this fault,and auto-correlation coefficient of normal response in corresponding sub-band was referred to as fault-free signature.Hence,fault diagnosis can be finished by comparing the faulty signature with its fault-free counterpart.Experiments of international Benchmark circuit showed that methodology had high precision for diagnosing parametric fault.