中文摘要：

为了提高模拟集成电路参数型故障的测试精度,以小波滤波器组实现的子带滤波为理论工具,研究了基于多故障特征提取的模拟集成电路测试方法。研究和对比了奥克塔夫（Octave）和金字塔两种小波分解结构下对Haar和Daubenchies两种小波在时域和频域中的故障分辨特性和诊断能力。对国际标准电路的实验表明：采用Octave结构时,在时域可完成故障检测,以相干函数在频域区分故障困难;而采用金字塔型结构时,在时域和频域皆易完成故障诊断;采用Daubenchies小波的效果优于Haar小波。

英文摘要：

Aiming at improving test accuracy of parametric fault in analog VLSI circuits,by virtue of the theoretical tool,sub-band filtering implemented by wavelet filter banks,an approach on testing analog integrated circuits was proposed based on extraction of multiple fault signature. Different identification performance and diagnosis capacities in both time domain and frequency domain were investigated and compared with respect to two wavelet decomposition structures （Octave and Pyramid） and two types of wavelets（ Haar and Daubenchies）. Experiments to International Benchmark circuit show that fault detection can be realized in time domain and it is difficult to identify fault by coherence function in frequency domain when Octave decomposition is applied. However,fault diagnosis is easy to be finished both in time domain and frequency domain when Pyramid decomposition is practiced. The fact that the effect of Daubenchies wavelet excels that of Haar is proved also.