Wafer Array Defect Detection via Robust Image Averaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection methods for wafers face challenges with high wafer-to-wafer noise and die-to-die noise, which decrease the sensitivity of defect detection, especially in array regions where repeater defects are difficult to identify due to noise interference.
Innovation Solution
A computer-implemented method that generates test images and reference images from frame images scanned by an inspection system, using robust averaging and median calculations to enhance the signal-to-noise ratio, allowing for cell-to-cell comparison to detect defects with higher sensitivity, including die repeaters, without requiring separate scans for test and reference images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If standard die image comparison is used for defect detection, then the inspection process is simple, but wafer-to-wafer noise and die-to-die noise interfere with defect detection and decrease sensitivity
Solution Approach 1:
The patent segments the wafer inspection process into multiple individual frame images that are scanned sequentially across the wafer surface. Each frame captures a specific region, and these segments are then processed independently through robust averaging and median calculations before being combined for final defect detection. This segmentation allows noise suppression while maintaining detection sensitivity.
Solution Approach 2:
The patent performs preliminary noise suppression actions by generating robust averaged images and median images from multiple individual frame images before the actual defect detection comparison. This preliminary processing removes wafer-to-wafer noise and die-to-die noise in advance, so that when test die images are compared against reference images, the sensitivity is not degraded by noise interference.
2Measurement precision
If multiple scans are performed to reduce noise, then defect detection sensitivity improves, but inspection time increases and throughput decreases
Solution Approach 1:
The patent merges multiple individual frame images into composite robust averaged images and median images through computational processing. By combining information from multiple frames captured during a single wafer scan, the system achieves noise reduction and enhanced sensitivity without requiring multiple separate physical scans of the wafer, thus maintaining high throughput.
Solution Approach 2:
The patent creates virtual copies of the wafer surface through multiple individual frame images captured during scanning. These frame copies are then processed through robust averaging and median calculations to generate reference images that represent the wafer surface without noise. This copying approach allows noise suppression computation without requiring multiple physical scans, preserving inspection speed.
3Ease of manufacture
If standard die image from another wafer is used as reference, then inspection can proceed, but wafer-to-wafer noise is high and interferes with defect detection
Solution Approach 1:
The patent enables the inspection system to generate its own reference images from the wafer being inspected, rather than relying on external standard die images from other wafers. By processing multiple individual frame images from the current wafer through robust averaging and median calculations, the system creates reference images that are self-consistent with the actual wafer surface characteristics, eliminating wafer-to-wafer noise interference.
4Ease of manufacture
If design data is used to generate standard die image, then reference image is available, but it does not adequately represent actual noise sources on the wafer
Solution Approach 1:
The patent creates reference images by copying and processing actual physical data from multiple individual frame images captured from the wafer surface, rather than generating images from theoretical design data. This copying of real wafer surface information ensures that the reference images accurately represent actual noise sources and surface characteristics present on the specific wafer being inspected, improving reliability.
Data Source
AI summary
Systems and methods for detecting defects on a wafer are provided. One method includes generating test image(s) for at least a portion of an array region in die(s) on a wafer from frame image(s) generated by scanning the wafer with an inspection system. The method also includes generating a reference image for cell(s) in the array region from frame images generated by the scanning of the wafer. In addition, the method includes determining difference image(s) for at least one cell in the at least the portion of the array region in the die(s) by subtracting the reference image from portion(s) of the test image(s) corresponding to the at least one cell. The method further includes detecting defects on the wafer in the at least one cell based on the difference image(s).