NAND Flash Decoding With Deep Learning Interference Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current memory systems face challenges in accurately decoding data due to interference from neighboring memory cells, particularly in NAND-type flash memory, where traditional error correction methods like LDPC decoding may fail, leading to errors in data retrieval.
Innovation Solution
A memory system with a decoder that performs multiple levels of soft decoding, including initial soft decoding, interference correction using a quantizer, and enhanced decoding via a deep learning model to accurately estimate log-likelihood ratio (LLR) values, ensuring reliable data retrieval even when initial methods fail.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional LDPC decoding is used, then the decoding process is simple and fast, but data correction accuracy deteriorates due to interference from neighboring memory cells
Solution Approach 1:
The decoding process is segmented into multiple levels: first soft decoding, then interference correction decoding, and finally deep learning-based decoding. Each level handles specific aspects of the decoding task, with the deep learning model specifically addressing interference from neighboring cells that traditional methods cannot correct.
Solution Approach 2:
A deep learning model is introduced as an intermediary component between the traditional LDPC decoder and the memory system. This model processes LLR values and predicts corrections for bits that failed initial decoding, acting as a bridge that enhances traditional decoding without replacing it entirely.
2Reliability
If multiple levels of soft decoding with deep learning are applied, then data correction accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The system performs deep learning-based decoding only for bits that fail initial soft decoding, rather than applying it to all bits. This partial application of the complex decoding method reduces overall processing time while maintaining high reliability for difficult-to-decode bits.
Solution Approach 2:
The system performs preliminary soft decoding using traditional LDPC methods first, identifying which bits require additional correction. This preliminary action filters out easily decodable bits, allowing the time-intensive deep learning process to focus only on problematic bits.
3Measurement precision
If interference correction is applied to all bits, then measurement precision is improved, but processing efficiency deteriorates
Solution Approach 1:
Interference correction is applied locally only to specific bits that exhibit high probability of error, rather than uniformly to all bits. The system identifies problematic bits through initial decoding metrics and applies the computationally intensive deep learning correction only where needed.
Solution Approach 2:
The system dynamically changes the decoding approach based on bit-specific parameters. Bits with LLR values indicating high uncertainty trigger the deep learning interference correction process, while bits with confident decoding results proceed directly to data output without additional processing.
Data Source
AI summary
Memory systems, controllers, decoders and methods execute decoding with a mufti-level interference correction scheme. A decoder performs first soft decoding to generate log likelihood ratio (LLR) values of a select bit and bits of memory cells neighboring a memory cell of the select bit. A quantizer obtains an estimated LLR value of the select bit based on the LLR values of the select bit and the bits of the memory cells neighboring the memory cell of the select bit, when the first soft decoding fails. The decoder performs second soft decoding using the estimated LLR value when the first soft decoding fails, and performs third soft decoding using information obtained from application of a deep learning model to provide a more accurate estimate of the LLR value of the select bit when the second soft decoding fails.