Correlation-Based Data Filtering for Semiconductor Yield Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor manufacturing, the relevance of inspection and measurement data for evaluating process quality is often reduced due to irrelevant contributions, leading to discarded implicit information and reduced production yield, as current methods fail to efficiently utilize process data effectively.
Innovation Solution
A technique that filters measurement data using predefined criteria to reduce non-relevant 'noise' and merge it with significant 'reference' data, enabling enhanced evaluation and yield loss estimation by determining the correlation between filtered and reference data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If measurement data from multiple manufacturing stages are collected and evaluated, then process quality monitoring capability is improved, but data relevance and signal-to-noise ratio deteriorate due to irrelevant contributions from non-critical processes
Solution Approach 1:
The patent segments the manufacturing process into discrete stages and divides measurement data into groups corresponding to different manufacturing stages. By evaluating correlation between stages, the system identifies which segmented data groups contribute meaningfully to final product quality, filtering out irrelevant segments while maintaining comprehensive monitoring capability.
Solution Approach 2:
The patent extracts and removes irrelevant measurement data from the evaluation process by comparing correlation between different manufacturing stages. Data from stages showing low correlation with final product characteristics are extracted and excluded, thereby eliminating noise while preserving the signal from critical processes.
2Reliability
If comprehensive measurement data from all manufacturing stages are used, then monitoring coverage is improved, but control efficiency deteriorates due to inability to distinguish relevant from irrelevant data
Solution Approach 1:
The patent implements a feedback mechanism where measurement data from later manufacturing stages is used to evaluate and adjust the relevance of data from earlier stages. This feedback loop enables the system to dynamically identify which processes have actual impact on final product quality, improving control efficiency by focusing resources on critical control points while maintaining comprehensive monitoring coverage.
Data Source
AI summary
By performing a contingency-based correlation test of measurement data, such as defect data, with respect to electrical test data after progressively filtering the measurement data, an enhanced analysis of process flow characteristics may be accomplished. Consequently, an efficient yield loss estimation may be performed.


