Memory Defect Classification via Block Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect classification methods for memory devices, such as design-based binning (DBB), are ineffective for memory devices with repeating structures like DRAM and Flash memory due to the inability to differentiate defects based on pattern background, leading to limited use in these applications.
Innovation Solution
A computer-implemented method that determines the positions of defects within specific blocks of a memory device area and classifies them based on these positions, using inspection data to differentiate defects within different types of blocks, thereby overcoming the limitations of design-based binning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If design-based binning (DBB) is used to classify defects, then defects can be grouped by pattern background, but defects in memory devices with repeating structures cannot be differentiated because they all have the same pattern background
Solution Approach 1:
The patent segments the defect classification approach by dividing the memory device area into different block types (memory cell blocks, sense amplifier blocks, wordline driver blocks, etc.). Instead of treating all defects uniformly based on pattern background, the system segments defects by their location within specific block types and applies different classification strategies to each segment, enabling effective differentiation even in repeating structures
Solution Approach 2:
The patent transitions from a two-dimensional pattern-based classification (DBB) to a three-dimensional classification by adding the block type location dimension. Defects are now classified not only by their pattern characteristics but also by their position within different functional blocks (memory cell, sense amplifier, wordline driver, etc.), creating a more nuanced classification that works for memory devices with repeating structures
2Quantity of substance
If design rules continue to shrink, then memory device density increases, but design background provides little or no differentiation for defects
Solution Approach 1:
As memory device density increases and design rules shrink, the patent applies segmentation by dividing the high-density structure into distinct functional block types. Each block type (memory cell block, sense amplifier block, etc.) is treated as a separate segment with its own classification criteria, allowing defect differentiation even when overall pattern backgrounds become similar due to scaling
Solution Approach 2:
The patent applies local quality by recognizing that different regions (block types) within the memory device have different functional characteristics and defect patterns. Instead of applying a uniform classification approach across the entire device, the system tailors the classification criteria to each local region's specific function and structure, maintaining defect differentiation capability despite overall density increases
Data Source
AI summary
Computer-implemented methods, computer-readable media, and systems for classifying defects detected in a memory device area on a wafer are provided.