Virtual Test Prediction for Memory Die Manufacturing Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Testing of memory devices during manufacturing is expensive, time-consuming, and reduces yield, limiting the amount of data available for quality control and process adjustments.
Innovation Solution
Employing machine learning to interpolate virtual test results for memory devices based on correlations with other test sets, allowing for more extensive data collection and efficient process adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive testing is performed on memory devices during manufacturing, then quality control data availability improves, but manufacturing cost and time increase
Solution Approach 1:
The patent creates virtual copies of test results through machine learning models that simulate the outcomes of extensive physical testing. Instead of performing actual tests on all devices, the system generates virtual test data that mirrors real testing outcomes, providing comprehensive quality control information without the time and resource costs of actual extensive testing.
Solution Approach 2:
The patent replaces the mechanical testing system with an information-based machine learning system. Physical testing operations are substituted by computational models that predict test outcomes based on process parameters and limited actual test data, eliminating the need for time-consuming physical testing while maintaining quality control effectiveness.
2Measurement precision
If more test samples are prepared for quality control, then measurement accuracy improves, but yield decreases due to sample preparation requirements
Solution Approach 1:
The system generates virtual test samples through machine learning models that replicate the characteristics and outcomes of physical test samples. This allows comprehensive quality control analysis on all production devices without requiring actual sample preparation, maintaining measurement accuracy while preserving manufacturing yield.
Solution Approach 2:
The patent performs preliminary virtual testing through machine learning predictions before physical sample preparation is needed. By obtaining quality control insights through virtual testing first, the system minimizes the need for extensive physical sample preparation, thereby maintaining higher manufacturing yield while preserving measurement accuracy.
3Reliability
If traditional testing methods are used for quality control, then defect detection reliability improves, but manufacturing cost increases
Solution Approach 1:
The patent replaces expensive physical testing operations with computationally-efficient machine learning models. The system uses virtual metrology and predictive analytics to detect defects with reliability comparable to traditional methods, but at significantly lower cost by eliminating the need for expensive test equipment, sample preparation facilities, and manual testing operations.
Solution Approach 2:
The system creates virtual test results that replicate the defect detection capabilities of traditional testing methods. By using machine learning models trained on actual test data, the system generates virtual copies of test outcomes that maintain defect detection reliability while eliminating the high costs associated with performing actual extensive physical testing on all devices.
Data Source
AI summary
To provide more test data during the manufacture of non-volatile memories and other integrated circuits, machine learning is used to generate virtual test values. Virtual test results are interpolated for one set of tests for devices on which the test is not performed based on correlations with other sets of tests. In one example, machine learning determines a correlation study between bad block values determined at die sort and photo-limited yield (PLY) values determined inline during processing. The correlation can be applied to interpolate virtual inline PLY data for all of the memory dies, allowing for more rapid feedback on the processing parameters for manufacturing the memory dies and making the manufacturing process more efficient and accurate. In another set of embodiments, the machine learning is used to extrapolate limited metrology (e.g., critical dimension) test data to all of the memory die through interpolated virtual metrology data values.