3D Memory LLR Estimation Using Read-Write Order Effects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing non-volatile memory technologies, such as flash memory, face challenges in improving error correction performance and reducing bit error rates and iteration numbers during LDPC soft decoding due to limitations in estimating LLR values.
Innovation Solution
A read-write method and apparatus for LEPS soft decoding estimation in three-dimensional memory, where the reference memory cell is determined based on the writing and reading orders to account for the influence of adjacent memory cells on the comprehensive distribution state, allowing for improved LLR value determination and soft decoding operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional LEPS estimation method is used for LLR value estimation, then the decoding process can be completed, but the error correction ability is insufficient and the bit error rate is high
Solution Approach 1:
The patent changes the estimation parameters by introducing comprehensive distribution state as an additional parameter beyond the traditional threshold voltage partition. It also introduces storage time as a parameter to track memory cell state changes. These parameter changes enable more accurate LLR estimation by capturing the dynamic behavior of memory cells during storage and reading operations.
Solution Approach 2:
The patent implements feedback mechanisms by using the comprehensive distribution state of reference memory cells to adjust the LLR estimation for current memory cells. The system continuously updates the distribution state information and uses it to refine subsequent estimations, creating a feedback loop that improves estimation accuracy iteratively throughout the decoding process.
2Reliability
If more iterations are performed to improve decoding accuracy, then error correction ability improves, but the number of iterations increases and processing time extends
Solution Approach 1:
The patent performs preliminary actions by pre-calculating and storing the comprehensive distribution states of reference memory cells before the actual decoding process. This preliminary preparation allows the decoder to quickly access accurate reference information during decoding, reducing the need for multiple iterative passes and minimizing processing time while maintaining high decoding accuracy.
3Measurement precision
If reference memory cell influence is considered for comprehensive distribution state, then LLR estimation accuracy improves, but the complexity of determining reference memory cells and processing increases
Solution Approach 1:
The patent segments the memory cell array into distinct groups where certain cells are designated as reference memory cells and others as current memory cells. This segmentation simplifies the processing by clearly defining which cells influence which others, reducing the overall complexity of determining reference relationships while still capturing the necessary inter-cell influences for accurate LLR estimation.
Data Source
AI summary
A read-write method includes: sequentially writing, in a first direction, a code word obtained by information-bit encoding into a target memory cell in each layer of memory cell array in the three-dimensional memory; randomly reading the target memory cell in each layer of memory cell array, or sequentially reading the target memory cell in each layer of memory cell array in a second direction; and determining an LLR value of a current target memory cell according to a storage time corresponding to the current target memory cell when reading, a threshold voltage partition corresponding to the current target memory cell when reading, a comprehensive distribution state corresponding to the current target memory cell when reading, and a pre-established LLR table, so as to perform a soft decoding operation on the code word in the current target memory cell based on the LLR value of the current target memory cell.