Adaptive Signature Detection for Integrated Circuit Defect Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing inspection methods for integrated circuits, particularly those using a floating reference die, can overlook defects when both the test and reference dies have defects in the same location, as these are misinterpreted as no defects due to zero values in the difference set.
Innovation Solution
A processor-based method that creates an image of the integrated circuit, groups pixels into bins based on common characteristics, calculates mean values, compares these to a threshold, flags bins with significant variations, and performs signature detection on flagged bins without comparing to another image, effectively identifying defects such as under-polish, pattern shifts, and process-induced variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a floating reference die is used in inspection, then inspection coverage is improved, but defects in the same location on both test and reference dies are misinterpreted as no defects
Solution Approach 1:
The patent segments the inspection process into multiple independent stages: image acquisition, feature extraction, histogram analysis, and defect candidate identification. Each stage operates independently on the test die image without requiring reference die comparison, eliminating the false negative problem while maintaining comprehensive defect detection capability
Solution Approach 2:
The patent introduces histograms as an intermediary representation between the raw image data and defect detection. By converting pixel data into histograms that capture statistical characteristics, the system can identify defects without direct comparison to reference dies, resolving the contradiction between inspection coverage and detection precision
2Ease of operation
If image comparison method is used, then defect detection is simplified, but defects on adjacent or near-neighbor dice are overlooked
Solution Approach 1:
The patent extracts defect detection capability from the reference die comparison process. By using histograms derived solely from the test die image pixels, the system can identify defects independently without needing reference dies, thereby eliminating the oversight of defects on adjacent or near-neighbor dice while maintaining operational simplicity
Solution Approach 2:
The system performs self-service defect detection by analyzing the test die image independently through histogram generation and threshold comparison. This self-contained approach eliminates the need for reference dies and external comparison, allowing the system to detect defects on any die without relying on neighboring dice images
Data Source
AI summary
A processor-based method for detecting defects in an integrated circuit, by creating an image of at least a portion of the integrated circuit with a sensor, grouping pixels of the image into bins based at least in part on a common characteristic of the pixels that are grouped within a given bin, creating a histogram of the pixels in each of the bins, calculating a mean value of the histogram for each of the bins, comparing the mean value for each of the bins to a threshold value, flagging as defect candidates those bins where the mean value of the bin varies from the threshold value by more than a predetermined amount, and performing signature detection on the bins that are flagged as defect candidates, where the image of the integrated circuit is not directly compared to any other image of an integrated circuit.