Are you the 1%?
Figure 2. Root Cause Deconvolution determines the root cause distribution and devices most likely to fail for each root cause. Click on image to enlarge.
Layout-aware diagnosis is performed on a set of die that failed manufacturing test (1). Each diagnosis result contains a set of root causes that are potential explanations for the failure. If we sum all the root causes and count the number of die whose failure could be explained by each root cause, we get a diagram that includes all the real root causes as well as the noise (2). RCD then eliminates this noise and identifies the underlying root cause distribution (3). From this distribution, the user can focus on the most significant root cause, on a root cause that hasn't been seen before, or for other unexpected reasons. Along with the root cause distribution, RCD assigns a probability for root cause per diagnosis suspect (4). This means that the user can easily identify the die that has the highest probability of representing a particular root cause, and use that as way to select die for FA (failure analysis). When comparing the RCD results to the original diagnosis report for one failing die, we see that RCD has eliminated several of the original root causes, thus effectively improved the resolution for that individual result (5). In this particular example, the original report contained seven possible root causes for one failing die, while RCD limited this to one single result. The layout snapshots show the defect bounding boxes before and after RCD (6).
The value of RCD is obvious when analyzing a single collection of fail data, such as a single wafer or a single lot. This technology has also proven useful for longer term yield monitoring. RCD defect distributions for multiple lots and even multiple devices can be compared to identify defect trends and variation. A recent paper published by GLOBALFOUNDRIES stated that "To best leverage RCD, the analysis population needs to be carefully prepared. By accumulating the RCD results across time and designs, helpful yield analysis can be carried out with minimal effort."
In conclusion, diagnosis-driven yield analysis with RCD is a quick and cost effective way to determine the underlying root causes represented in a population of failing devices from test data alone. This technology can, for instance, ferret out the final 1% yield loss in mature technologies, thus providing significant value to the yield and failure analysis process at fabless semiconductor companies because the test data is readily available.
1. W. Yang, C. Hao, Diagnosis-Driven Yield Analysis Improves Mature Yield, Chip Design Magazine, Fall 2011.
2. B. Benware, et.al., Determining a Failure Root Cause Distribution From a Population of Layout-Aware Scan Diagnosis Results, IEEE D&T of Computers, Volume 29, Issue 1.
3. Y. Pan, et.al., Leveraging Root Cause Deconvolution Analysis for Logic Yield Ramping, International Symposium for Test and Failure Analysis 2013.
No Article Found