Non-Volatile Memory Die Defect Detection via Adaptive Read Current Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to reliably detect defective non-volatile memory cells during manufacturing, as the range of read current values for good cells varies and the use of a single predetermined threshold can lead to inaccurate identification of defective dies.
Innovation Solution
A method involving two read operations to determine outlier cells by setting a target current or voltage threshold relative to the lowest or highest current/voltage of a die's memory cells, with a predetermined margin, to accurately identify defective dies based on the distribution of read currents or threshold voltages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single predetermined read current threshold is used for all dies, then the testing process is simple and fast, but the accuracy of identifying defective dies deteriorates due to variations in read current distributions across different dies
Solution Approach 1:
The patent applies dynamics by making the read current threshold adaptive rather than fixed. The threshold is dynamically adjusted based on the actual read current distribution characteristics of each die being tested, allowing the testing method to adapt to variations between different dies while maintaining high testing speed
Solution Approach 2:
The patent changes the parameter of the threshold from a fixed predetermined value to a variable value that depends on the statistical characteristics (mean and standard deviation) of the read current distribution. This parameter change enables accurate defective die identification across different dies with varying read current distributions
2Reliability
If the predetermined read current threshold is set high to reduce false negatives, then defective dies are more likely to be detected, but false positives increase causing good dies to be erroneously marked as defective
Solution Approach 1:
The patent uses feedback by incorporating the statistical characteristics (mean and standard deviation) of the read current distribution into the threshold determination. The threshold is set based on feedback from the actual measurement data, specifically using a formula that accounts for both the mean and standard deviation to achieve an optimal balance between detecting defective dies and avoiding false positives
Solution Approach 2:
The patent performs preliminary statistical analysis (calculating mean and standard deviation) on the read current distribution before setting the threshold. This preliminary action allows the threshold to be optimally positioned to detect defective dies while minimizing false positives, based on the actual characteristics of the being-tested die
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A method of testing non-volatile memory cells formed on a die includes erasing the memory cells and performing a first read operation to determine a lowest read current RC1 for the memory cells and a first number N1 of the memory cells having the lowest read current RC1. A second read operation is performed to determine a second number N2 of the memory cells having a read current not exceeding a target read current RC2. The target read current RC2 is equal to the lowest read current RC1 plus a predetermined current value. The die is determined to be acceptable if the second number N2 is determined to exceed the first number N1 plus a predetermined number. The die is determined to be defective if the second number N2 is determined not to exceed the first number N1 plus the predetermined number.