Memory Error Correction Using Syndrome Computation for LDPC Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current error correction methods in memory subsystems, particularly those using low-density parity check (LDPC) codes and staircase codes, are vulnerable to high reliability errors (HREs) and suffer from performance degradation due to the need to retrieve additional data for decoding, which is impractical for moderate to high error rates.
Innovation Solution
The implementation of syndrome-based decoders in memory devices that transmit encoded data along with syndromes to an error corrector, allowing for robust error correction without the need for additional data transfer, thereby improving performance and reducing data transmission while maintaining high correction capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If syndrome-based decoders are implemented to improve error correction robustness, then error correction capability is improved, but device complexity increases
Solution Approach 1:
The error correction system is segmented into distinct functional components: syndrome computation unit that calculates syndromes from received data, syndrome storage unit that stores computed syndromes, and syndrome-based decoder that uses stored syndromes for error correction. This segmentation allows each component to be optimized independently while working together to achieve robust error correction.
Solution Approach 2:
Syndromes are computed and stored in advance before actual error correction is needed. The syndrome computation unit calculates syndromes from received encoded data and stores them in the syndrome storage unit, so that when error correction is required, the decoder can immediately access pre-computed syndromes without waiting for additional data retrieval or computation.
2Reliability
If additional data is retrieved for decoding to maintain error correction capability, then error correction robustness is improved, but data transmission volume increases
Solution Approach 1:
The essential error correction information (syndromes) is extracted from the encoded data and transmitted separately. Instead of retrieving additional encoded data blocks for decoding, the system extracts and transmits only the syndrome values, which contain all necessary information for error correction. This extraction principle significantly reduces the amount of data that must be transmitted and stored.
Solution Approach 2:
The system changes the parameter representation of error correction information from transmitting full encoded data blocks to transmitting compressed syndrome values. Syndromes are mathematical parameters that encapsulate the essential error-detection and correction information in a compact form, reducing transmission volume while maintaining correction capability.
3Reliability
If traditional LDPC or staircase codes are used to maintain error correction capability, then correction robustness is improved, but performance degrades due to additional data retrieval requirements
Solution Approach 1:
Syndromes serve as an intermediary between received encoded data and the error correction process. Instead of directly processing large volumes of encoded data or retrieving additional data blocks, the syndrome computation unit transforms the received data into compact syndrome parameters that mediate the error correction process, improving decoding performance by reducing computational complexity and data access requirements.
Data Source
AI summary
Methods, systems, and apparatus for error correction with syndrome computation in a memory device are described. A first syndrome for first encoded data is generated in a memory device. The first syndrome and the first encoded data are transmitted to a controller that is coupled with the memory device. A second syndrome for first and second encoded data is generated. The first encoded data and the second encoded data are interrelated according to an error correction code. The second syndrome is transmitted to the controller without the second encoded data and the controller is to decode the first encoded data based on at least one of the first syndrome, the second syndrome, or a combination thereof.