Given a logic circuit that fails a test, diagnosis is the process of narrowing down the possible locations of the defects. Diagnosis to locate defects in VLSI circuits has become very important during the yield ramp up process, especially for 90 nm and below technologies where physical failure analysis machines become less successful due to reduced defect visibility by the smaller feature size and larger leakage currents. Successful defect isolation relies heavily on the guidance from fault diagnosis and will depend even more for the future technologies.
To assist a designer or a failure analysis engineer, the diagnosis tool tries to identify the possible locations of the failure effectively and quickly. While many defects reside in the logic part of a chip, defects in scan chains have become more and more common recently as typically 30%-50% logic gates impact the operation of scan chains in a scan design. Logic diagnosis and scan chain diagnosis are the two main fields of diagnosis research. The quality of diagnosis directly impacts the time-to-market and the total product cost. Volume diagnosis with statistical learning is important to discover systematic defects. An accurate diagnosis tool is required to diagnose large numbers of failing devices to aid statistical yield learning. In this work, we propose techniques to improve diagnosis accuracy and resolution, techniques to improve run-time performance.
We consider the problem of determining the location of defects in scan chains and logic. We investigate a method to improve the diagnosability of production compressed test patterns for multiple scan chain failures. Then a method to generate special diagnostic patterns for scan chain failures was proposed. The method tries to generate a complete test pattern set to pinpoint the exact faulty scan cell when flush tests tell which scan chain is faulty.
Next we studied the problem of diagnosis of multiple faults in the logic of circuits. First we propose a method to diagnose multiple practical physical defects using simple logic fault models. At last we propose a method based on fault-tuple equivalence trees to further improve diagnosis quality.
| Identifer | oai:union.ndltd.org:uiowa.edu/oai:ir.uiowa.edu:etd-2079 |
| Date | 01 December 2010 |
| Creators | Tang, Xun |
| Contributors | Reddy, Sudhakar M. |
| Publisher | University of Iowa |
| Source Sets | University of Iowa |
| Language | English |
| Detected Language | English |
| Type | dissertation |
| Format | application/pdf |
| Source | Theses and Dissertations |
| Rights | Copyright 2010 Xun Tang |
Page generated in 0.0013 seconds