Defect Binning and Noise Filtering for Semiconductor Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor inspection processes face challenges in efficiently distinguishing between real defects and noise or nuisance events, requiring time-consuming manual setup of data acquisition and processing parameters, and often result in suboptimal defect review processes due to the complexity of adjusting parameters for various defect types.
Innovation Solution
A computer-implemented method that detects noise in inspection data by using detection parameters known to identify noise and nuisance events, eliminates these events, and bins defects based on characteristics and spatial signatures, allowing for optimized data acquisition and processing parameters to improve throughput and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual processes are used to set up inspection processes with data acquisition and data processing parameters, then the inspection process can be customized for different defect types, but the setup time becomes relatively long
Solution Approach 1:
The system automatically determines optimal data acquisition and data processing parameters by analyzing defect characteristics and inspection requirements, eliminating the need for manual parameter setup. The inspection system self-configures by selecting from predefined parameter sets based on the specific defect type and inspection goals, thereby reducing setup time while maintaining customization capability.
2Reliability
If all possible combinations of data acquisition and data processing parameters are tested to find the best inspection process, then the optimal inspection process can be identified, but the time and computational resources required become excessive
Solution Approach 1:
The system pre-establishes multiple sets of data acquisition and data processing parameters corresponding to different defect types and inspection scenarios. Before actual inspection, the system selects the appropriate pre-configured parameter set based on defect characteristics, avoiding the need to test all possible parameter combinations during the inspection process itself.
Solution Approach 2:
The system maintains a library of predefined parameter sets with different data acquisition and processing configurations. Based on the identified defect type and inspection requirements, the system dynamically selects and applies the most suitable parameter set from the library, achieving optimal inspection performance without exhaustive parameter testing.
3Measurement precision
If noise and nuisance events are reduced by using optimized parameters and filtering techniques, then the quality of inspection results improves, but the process becomes more complex and time-consuming
Solution Approach 1:
The system pre-configures optimized parameter sets that are specifically designed to minimize noise and nuisance events for different defect types. By selecting from these pre-optimized parameters rather than performing real-time optimization, the system achieves high inspection quality without the complexity of dynamic parameter tuning and extensive filtering operations.
Data Source
AI summary
Computer-implemented methods for performing one or more defect-related functions are provided. One method for identifying noise in inspection data includes identifying events detected in a number of sets of inspection data that is less than a predetermined number as noise. One method for binning defects includes binning the defects into groups based on defect characteristics and the sets of the inspection data in which the defects were detected. One method for selecting defects for defect analysis includes binning defects into group(s) based on proximity of the defects to each other and spatial signatures formed by the group(s). A different method for selecting defects for defect analysis includes selecting defects having the greatest diversity of defect characteristic(s) for defect analysis. One method includes classifying defects on a specimen using inspection data generated for the specimen combined with defect review data generated for the specimen.