Fault Detection and Diagnosis for Mixed-Signal Circuits Using Wavelet Based Transient Current Analysis

Purdue University, West Lafayette IN

Investigators

Abstract

As intrinsic leakage in transistor increases with technology scaling the effectiveness of quiescent current (IDDQ) testing reduces significantly. Transient current (IDD) based testing has been often cited and investigated as an alternative and/or supplement to IDDQ testing in Digital CMOS circuits. While the potential of IDD testing for fault detection has been established for digital and analog circuits, there is no known efficient method for fault diagnosis using IDD analysis. We propose a novel integrated method for fault detection and diagnosis in digital CMOS circuits using IDD waveform analysis using wavelet transform. We use wavelet transform to decompose the IDD waveform in both time and frequency domain. The time-frequency resolution of the IDD signal helps us detect as well as localize faults. Initial experimental results on an 8-bit ALU show that wavelet based IDD analysis has the potential to efficiently detect and localize faults considering practical issues like effect of process variation and measurement noise. Transient current (IDD) analysis can also be effective for defect oriented testing of analog circuits. We observe that wavelet transform renders an efficient way for analyzing IDD for fault detection in analog circuits. The property of wavelet for resolving events in both time and frequency domain simultaneously and the property of better sub-banding than Fourier analysis, makes it a effective tool for IDD analysis. Moreover wavelet transform can be easily adapted to current waveforms from different circuits. We have observed that for equivalent number of spectral components, sensitivity of wavelet based fault detection in analog circuits is much higher than fourier or time-domain analysis for both catastrophic and parametric faults. Initial experimental results on a benchmark circuit show that wavelet based method is on average 25 times more sensitive than DFT for parametric faults and can be considered as a promising alternative for analog fault detection amidst measurement hardware noise and process variation. In the proposed research we will develop an integrated fault detection and diagnosis methodology using wavelet based transient current analysis for mixed-signal circuits. We will present the results of our research in domestic and international conferences. International conferences include (International Conference on CAD and CG in Macau, December 2003, and International Conference on CAD and CG 2005; IEEE Design and Test in Europe).