Wafer Spatial Failure Detection Using Transform and Cluster Analysis
Solution Overview
Problem
Current methods for detecting spatial failures in integrated circuit wafers are inefficient, requiring significant resources and time while increasing the risk of missing defects, which can lead to unusable ICs.
Innovation Solution
A method and system utilizing transforms and cluster signature analysis to quickly and accurately detect spatial failures by analyzing only the coordinates of defects, optimizing memory usage and processing speed without increasing error rates, and varying resolution to optimize detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional methods are used to detect spatial failures in wafers, then detection coverage is maintained, but processing time increases and memory requirements increase
Solution Approach 1:
The patent segments the failure detection process into two distinct stages: a first stage that performs comprehensive spatial failure detection using coordinate analysis, and a second stage that performs additional detection on remaining failures. This segmentation allows the system to quickly identify and remove obvious failure patterns in the first stage, reducing the workload for subsequent analysis while maintaining comprehensive detection coverage.
Solution Approach 2:
The patent extracts only the essential coordinate information of spatial failures for analysis, rather than processing complete failure data sets. By taking out only the necessary coordinate parameters and using them to generate histograms and perform pattern matching, the system significantly reduces memory requirements and processing time while maintaining detection accuracy.
2Measurement precision
If comprehensive failure analysis is performed on all spatial failures, then detection accuracy is maintained, but memory requirements increase
Solution Approach 1:
The patent extracts only the essential coordinate information of spatial failures for analysis, rather than processing complete failure data sets. By taking out only the necessary coordinate parameters and using them to generate histograms and perform pattern matching, the system significantly reduces memory requirements while maintaining detection accuracy.
Solution Approach 2:
Instead of analyzing all spatial failures comprehensively from the start, the patent inverts the approach by first analyzing failures in a simplified manner to identify and remove obvious patterns, then performing more detailed analysis only on the remaining failures. This inversion reduces the quantity of data requiring comprehensive analysis while maintaining overall detection precision.
3Measurement precision
If high resolution analysis is used to optimize detection, then detection accuracy improves, but processing time increases
Solution Approach 1:
The patent implements a dynamic two-stage detection process where the analysis resolution and depth are adjusted based on the detection stage and the characteristics of remaining failures. The first stage uses coordinate-based histogram analysis for quick pattern identification, while the second stage applies more detailed analysis only where necessary, optimizing the balance between detection precision and processing speed.
Solution Approach 2:
The patent applies partial analysis actions in the first stage by focusing on identifying and removing common failure patterns through histogram analysis, rather than performing exhaustive analysis on all failures. This partial action approach maintains high detection accuracy for common patterns while significantly reducing processing time, with more detailed analysis applied only to remaining cases in the second stage.
Data Source
AI summary
Detecting spatial failures in a wafer can be performed quickly and accurately by using transformations and cluster signature analysis. For this technique, a system can receive failure coordinates for the wafer, wherein each failure coordinate indicates a spatial failure. A failure array can be generated based on a resolution of those failure coordinates. A first set of objects, e.g. linear and arc objects, can be detected in the failure array using transformations. A second set of objects, e.g. blob and scratch objects, can be detected in the failure array using cluster signature analysis. This technique advantageously reduces the amount of memory required to perform the analysis and increases the processing speed without increasing the error rate of detection.