Yield Model Extraction of Attribute Fail Rates in Semiconductor Manufacturing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor manufacturing, identifying attribute fail rates in complex systems is challenging due to the convoluted nature of processes and designs, which hinders yield optimization and profitability.
Innovation Solution
A computer-implemented method using yield modeling and statistical analysis to extract attribute fail rates by generating a yield model from test results, applying a Poisson yield model, and performing statistical analysis to determine failure rates of specific attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If semiconductor manufacturing processes and designs become more complex to improve functionality, then device capabilities are enhanced, but identifying and determining the cause of failures becomes more difficult
Solution Approach 1:
The patent segments the complex manufacturing process into distinct process steps and attributes, creating a structured breakdown of the manufacturing flow. This segmentation allows the system to isolate and analyze specific process steps individually, making failure identification manageable despite overall system complexity.
Solution Approach 2:
The patent introduces yield modeling and statistical analysis as intermediary tools between the complex manufacturing process and failure analysis. These intermediaries process the convoluted process data, transforming it into actionable insights about failure causes without requiring direct analysis of the entire complex system.
2Reliability
If traditional failure analysis techniques are used in complex systems, then some failure information can be obtained, but the convoluted nature of processes and designs hinders effective yield optimization
Solution Approach 1:
The patent changes the analytical parameters by applying statistical methods and yield modeling to process data. Instead of using traditional qualitative failure analysis, the system transforms process data into quantitative yield metrics and failure rates, enabling effective yield optimization despite process complexity.
Solution Approach 2:
The patent replaces traditional mechanical failure analysis techniques with computational and statistical methods. By substituting physical inspection and analysis with data-driven modeling and statistical analysis, the system overcomes the limitations of traditional techniques in handling convoluted processes.
3Loss of information
If detailed process data is collected for comprehensive analysis, then more information is available, but the convoluted nature of the data makes it difficult to extract meaningful failure rates
Solution Approach 1:
The patent introduces yield modeling as an intermediary layer between raw process data and failure rate extraction. This intermediary process filters, structures, and analyzes the convoluted data, transforming it into precise failure rate measurements without losing critical information.
Solution Approach 2:
The patent implements a feedback mechanism where statistical analysis results are used to refine the yield model and improve subsequent failure rate extractions. This iterative feedback process continuously enhances measurement precision by learning from accumulated data and analysis results.
Data Source
AI summary
A method, system or computer usable program product for extracting attribute fail rates for manufactured devices including testing manufactured devices having a set of attributes to provide a set of test results stored in memory; generating a yield model of the manufactured devices parsed by the set of attributes; populating the yield model based on the set of test results; and utilizing a processor to perform statistical analysis of the populated yield model to extract fail rates of the selected subset of attributes.