Photomask Inspection Using Codebook Pattern Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional photomask and wafer inspection techniques face challenges such as high computational requirements, precision alignment issues, and the need for multiple dice on a mask, which complicates defect detection and increases hardware complexity and cost, especially for large-die ICs like advanced microprocessors and field programmable gate arrays.
Innovation Solution
A method and system for inspecting photomasks using a codebook to compare data representative of images of circuit patterns, employing lossy compression algorithms and a knowledge base to identify potentially defective patterns without requiring a reference image, allowing for self-inspection and reducing computational overhead by eliminating typical patterns from further analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional D:DB inspection is used to compare captured optical images with database images, then defect detection capability is improved, but computational requirements and processing time increase tremendously
Solution Approach 1:
The patent extracts and removes typical, normal circuit patterns from the inspection process by comparing captured images against a database of known good patterns. Only atypical patterns that differ from the database are flagged for further analysis, eliminating the need to process entire high-resolution database images for every inspection.
Solution Approach 2:
The inspection process is segmented into two stages: first, a rapid comparison against a compressed codebook identifies potential defects; second, only those specific regions flagged as atypical undergo detailed analysis. This segmentation divides the computational workload into coarse filtering and fine inspection phases.
2Measurement precision
If high-precision alignment between mask position and database is required for D:DB inspection, then measurement accuracy is improved, but system complexity and hardware requirements increase
Solution Approach 1:
The patent creates a compressed codebook representation of the database patterns that captures essential features without requiring exact pixel-level correspondence. This codebook serves as a simplified copy that can be rapidly compared against captured images without demanding high-precision alignment hardware.
3Ease of manufacture
If D:D inspection is used to compare two optical images from different dice, then implementation ease is improved, but the ability to detect mask defects is reduced
Solution Approach 1:
The patent performs preliminary action by pre-processing and compressing database patterns into a codebook before inspection. This pre-computed reference enables rapid comparison that is more reliable than D:D methods, as it compares against a comprehensive database of known good patterns rather than relying on the assumption that two dice will be identical.
Data Source
AI summary
There are many inventions described and illustrated herein. In one aspect, the present invention is directed to a technique of, and system for inspecting integrated circuits, including, for example, patterns projected, provided or formed on a wafer using photomasks, or patterns on the photomask itself. The inspection system and technique of this aspect includes first identifying, determining and/or detecting areas and/or patterns that are potentially defective by removing, filtering and/or eliminating from a set of potential defects any and/or all typical, regular or normal patterns. The identification, determination and/or detection of potential defects may be performed relatively quickly by a rapidly executing algorithm. In this way, a first or “coarse” analysis is performed rapidly and some, many, all or substantially all of the regular, normal or typical patterns are eliminated from further analysis. Thereafter, a second more detailed analysis is performed. This second analysis focuses on the set of potential defects that were identified, determined and/or detected during the first analysis of the photomask or wafer (i.e., the “coarse” analysis). The second analysis may be considerably a more detailed or a “fine” analysis relative to the first or “coarse” analysis. Indeed, in one embodiment, the second analysis may implement a more computational intensive process, without sacrificing throughput, since only a small portion of the photomask or wafer is inspected in the second analysis. In this way, the detailed analysis of the defect candidates may identify (i) all or substantially all of the actual defects and/or (ii) only the actual defects from the potential defects identified during the first analysis.