Wafer Image Generation Using Die Tensor from CAD Wireframe
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wafer inspection methods are inefficient and prone to false defect detection due to reliance on comparing gray scale values of neighboring dies, which can be defective, and are time-consuming, especially when process variations and noisy optical conditions are involved.
Innovation Solution
Generating a die tensor from a CAD file by segmenting and transforming a wireframe image using a segmentation engine and image transformation engine, and combining transformed entities into a multi-dimensional image with predefined slices for accurate defect detection, incorporating metrology, design specifications, and optical information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If gray scale values of neighboring dies are compared for defect detection, then defect detection can be performed, but false defect detection occurs when reference values are defective and propagation of false detection happens
Solution Approach 1:
The patent generates a virtual reference image from CAD data before actual wafer inspection. This virtual reference image is created by simulating the manufacturing process and optical effects, providing a defect-free reference that eliminates the propagation of false detections from defective physical reference dies.
Solution Approach 2:
The patent creates a virtual copy of the wafer structure from CAD data rather than using physical reference dies. This digital twin approach allows comparison without the limitations of physical references, eliminating false defects while maintaining detection accuracy.
2Reliability
If gray scale values of all dies are compared with neighboring dies for inspection, then defect detection coverage is achieved, but the inspection process becomes time-consuming
Solution Approach 1:
The virtual reference image is generated in advance from CAD data, containing all necessary reference information. During inspection, only comparison operations are needed without performing complex transformations or simulations during the actual measurement, significantly reducing inspection time while maintaining full coverage.
Solution Approach 2:
The patent divides the wafer inspection into multiple channels (amplitude, phase, gradient) and processes them separately. This segmentation allows parallel processing of different defect detection aspects, improving overall inspection speed while maintaining comprehensive defect coverage.
3Measurement precision
If iterative transform is performed based on optical information for generating reference image, then accurate defect detection is achieved, but computational complexity and processing time increase
Solution Approach 1:
The virtual reference image incorporating optical effects is generated once from CAD data before production. This pre-computed reference captures all optical transformations, allowing rapid comparison during inspection without repeating complex iterative transforms, reducing generation time while maintaining accuracy.
Data Source
AI summary
A method is disclosed of generating a die tensor of a wafer from a Computer-Aided Design (CAD) file. According to the method, a segmentation engine segments a wireframe image obtained from the CAD file into a plurality of entities. An image transformation engine performs a transform on each of the plurality of entities based on at least one of the wireframe image, metrology, a design specification, process information, and optical information. The transform is performed iteratively based on the optical information. A stitch engine generates a die tensor, having a predefined number of slices, by combining each of the transformed plurality of entities.