Automated Optical Inspection for Unit-Specific Semiconductor Patterning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automated optical inspection (AOI) methods are inadequate for semiconductor packages with unit-specific patterning, as they rely on average or golden images that cannot account for unique, customized patterns, leading to incomplete or ineffective defect detection.
Innovation Solution
A method for AOI of unique semiconductor packages involves creating unique reference standards for each unit-specific pattern, using pre-processing techniques such as converting images to binary formats, extracting geometry, and comparing these standards to acquired images to detect defects, allowing for efficient inspection of dynamic and customized patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional AOI uses average or golden images for inspection, then the inspection process is simple and fast, but it cannot effectively detect defects in packages with unit-specific patterning
Solution Approach 1:
The patent segments the inspection process into two distinct phases: (1) a learning phase where the system acquires images of known good packages with unit-specific patterning to build a reference model, and (2) an inspection phase where the learned model is applied to detect defects. This segmentation allows the system to handle customized patterns effectively while maintaining operational efficiency.
Solution Approach 2:
The patent performs preliminary action by pre-acquiring and processing images of known good packages during a learning phase before actual production inspection begins. This preliminary step creates a customized reference model that accounts for unit-specific patterning variations, enabling accurate defect detection without requiring complex real-time adaptations during high-speed inspection.
2Adaptability or versatility
If conventional AOI uses a single golden image for all packages, then the manufacturing process is simple, but it cannot account for unique customized patterns on each die
Solution Approach 1:
The system performs preliminary image acquisition and processing of known good packages during a learning phase before production inspection. This pre-processing creates customized reference models that capture unit-specific patterning characteristics, allowing the inspection phase to proceed efficiently without real-time customization delays.
Solution Approach 2:
The patent creates reference copies of known good packages with unit-specific patterning during the learning phase. These reference copies serve as customized templates for comparison during inspection, allowing the system to adapt to unique patterns on each die type while maintaining fast inspection speeds through template matching rather than real-time analysis.
3Measurement precision
If AOI creates a unique reference standard for each unit-specific pattern, then defect detection becomes accurate, but the processing complexity increases
Solution Approach 1:
The patent segments the reference standard creation process into automated learning phase processing, where the system acquires and processes images of known good packages to generate customized reference models. This segmentation handles the computational complexity during an initial setup phase rather than during high-speed production inspection, maintaining both precision and efficiency.
Solution Approach 2:
The system performs self-service by automatically acquiring, processing, and generating customized reference models during the learning phase without requiring manual intervention. The automated processing handles image acquisition, alignment, and model generation, reducing the burden of complex processing while ensuring high measurement precision through consistent algorithmic operations.
Data Source
AI summary
A method of automated optical inspection (AOI) for a plurality of unique semiconductor packages can comprise providing a plurality of semiconductor die formed as a reconstituted wafer. A plurality of unit specific patterns can be formed by forming a unit specific pattern over each of the plurality of semiconductor die, wherein each of the unit specific patterns is customized to fit its respective semiconductor die. A plurality of images can be acquired by acquiring an image for each of the plurality of unit specific patterns. A plurality of unique reference standards can be created by creating a unique reference standard for each of the plurality of unit specific patterns. Defects can be detected in the plurality of unit specific patterns by comparing one of the plurality of unique reference standards to a corresponding one of the plurality of images for each of the plurality of unit specific patterns.