3D Mounting Offset Detection via Two-Field Image Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In three-dimensional mounting of semiconductor elements, thermocompression can result in mounting offset, making it difficult to ensure the quality of lamination, especially when inclination occurs, and existing inspection methods are either destructive or limited in scope, preventing thorough evaluation of defects.
Innovation Solution
A three-dimensional mounting method and device using a two-field image recognition system to align and store positional coordinates of electrodes on both faces of joining materials, allowing for non-destructive evaluation of mounting offset between layers by comparing alignment positions, thereby ensuring accurate lamination and preventing material waste.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Strength
If thermocompression is used for joining upper-layer and lower-layer joining materials, then the joining strength is improved, but mounting offset occurs due to inclination during lamination
Solution Approach 1:
The patent performs preliminary positioning of the upper-layer joining material relative to the lower-layer joining material before applying thermocompression. The recognition means captures images, calculates positional relationships, and adjusts positions to ensure accurate alignment. This preliminary positioning action prevents mounting offset from occurring during the subsequent thermocompression process, thereby maintaining both joining strength and mounting precision.
2Difficulty of detecting and measuring
If X-ray imaging is used to inspect mounting offset, then the detection capability is improved, but the images of various layers overlap making correct evaluation difficult
Solution Approach 1:
The patent segments the inspection process by layer. The recognition means inspects each layer separately by capturing images of individual joining materials and their alignment marks. The positional relationship between layers is calculated by comparing coordinates from separate layer inspections, rather than attempting to inspect all layers simultaneously. This segmentation eliminates image overlap issues while maintaining comprehensive detection capability.
3Difficulty of detecting and measuring
If cross section observation is used for inspection, then the defect detection capability is improved, but the test becomes destructive limiting the number of inspectable products
Solution Approach 1:
The patent replaces the mechanical cross-section observation method with an optical imaging system. The recognition means uses cameras to capture images of alignment marks on the surfaces of joining materials. This non-contact optical measurement system substitutes for the destructive mechanical sectioning process, enabling full inspection of all products without limiting productivity, while still providing sufficient defect detection capability through precise positional measurement.
4Manufacturing precision
If precise positioning is performed for each lamination, then the manufacturing precision is improved, but the time required for recognition and positioning increases
Solution Approach 1:
The patent merges the recognition and positioning operations into a single integrated process. The recognition means captures images that contain both alignment mark information and positioning information simultaneously. The control unit processes these images to perform both recognition and positioning calculations in one operation rather than as separate sequential steps. This merging reduces the total time required while maintaining precise positioning accuracy for each lamination.
Data Source
AI summary
A three-dimensional mounting method for successively laminating N number of upper-layer joining materials includes positioning a first upper-layer joining material relative to a lowermost-layer joining material by recognizing an alignment position of the lowermost-layer joining material and a lower face alignment position of the first upper-layer joining material by a two-field image recognition unit, storing positional coordinates of the alignment position of the lowermost-layer joining material, positioning an (n+1)-th upper-layer joining material relative to an n-th upper-layer joining material by recognizing an upper face alignment position of the n-th upper-layer joining material and a lower face alignment position of the (n+1)-th upper-layer joining material, storing positional coordinates of the upper face alignment position of the n-th upper-layer joining material, recognizing an upper face alignment position of the N-th uppermost-layer joining material, and storing positional coordinates of the upper face alignment position of the N-th uppermost-layer joining material.