Non-volatile Storage Word Line Defect Recovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Semiconductor memory devices, such as EEPROM and flash memory, can suffer from manufacturing defects that are not detected during the testing phase but cause errors after use, particularly due to defective word lines, which can lead to data corruption and storage errors.
Innovation Solution
A method is introduced to determine whether a word line in a non-volatile storage system is broken by performing both page-by-page and sequential read processes and comparing the results, allowing for data recovery from memory cells connected to a broken word line using a sequential read process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manufacturing testing is performed on memory devices, then manufacturing defects can be detected, but defects in word lines that do not cause immediate failure may still escape detection and cause errors after use
Solution Approach 1:
The patent performs preliminary diagnostic actions by executing read operations on memory blocks during manufacturing or initialization to identify word lines that may have latent defects. This preliminary detection allows the system to flag potentially problematic word lines before they cause actual data errors during normal operation.
Solution Approach 2:
The system implements feedback mechanisms where read results from memory blocks are analyzed to detect patterns indicating word line defects. When defects are detected, the system provides feedback to adjust read voltages or mark affected blocks as requiring special handling, thereby improving future read accuracy and data integrity.
2Productivity
If standard read processes are used, then data can be read quickly, but data from memory cells connected to defective word lines may be corrupted
Solution Approach 1:
The patent segments the memory array into multiple blocks and further divides each block into word line groups. This segmentation allows the system to apply different read strategies to different segments - using standard fast reads for healthy blocks and enhanced error-correcting reads for blocks containing defective word lines, thereby maintaining overall productivity while ensuring data accuracy.
Solution Approach 2:
The system dynamically changes read parameters such as voltage levels and timing based on the detected health status of word lines. For memory blocks identified as containing defective word lines, the system adjusts read voltages and applies additional verification steps, balancing the need for speed with the requirement for accurate data retrieval from problematic areas.
3Reliability
If multiple read processes are performed to detect defective word lines, then data integrity can be maintained, but reading time increases
Solution Approach 1:
The patent performs preliminary identification of defective word lines during manufacturing or initial system operation, creating a map of problematic areas. This preliminary action allows subsequent read operations to use optimized strategies - avoiding time-consuming multiple reads for healthy blocks while applying enhanced verification only where necessary, thus minimizing overall time loss while maintaining data integrity.
Solution Approach 2:
The system creates copies or replicas of data from memory blocks containing defective word lines and performs verification on these copies. This allows the original read operation to complete quickly while backup verification processes run in parallel or asynchronously, ensuring data integrity without significantly impacting the primary read time.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Data, normally read using a page-by page read process, can be recovered from memory cells connected to a broken word line by performing a sequential read process. To determine whether a word line is broken, both a page-by page read process and a sequential read process are performed. The results of both read processes are compared. If the number of mismatches between the two read processes is greater than a threshold, it is concluded that there is a broken word line.