Memory Error Correction With Adaptive LLR Table Shifting

Solution Overview

Problem

Existing memory systems face challenges in enhancing error correction capabilities, particularly in scenarios where threshold voltage distributions of memory cells overlap, leading to unsuccessful hard-decision decoding and increased error rates.

Innovation Solution

The memory system incorporates a semiconductor memory device and a controller that performs error correction using both hard bit and soft bit data. It generates a dynamic LLR estimation table based on corrected data and adjusts this table using a shift value determined by the voltage difference between the read voltage and the tracking voltage, thereby correcting the table to improve decoding accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If hard-decision decoding is used for error correction, then the decoding process is simple and fast, but error correction fails when threshold voltage distributions overlap

Engineering Contradiction:

Improvedecoding speedVSAvoiderror correction capability

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent transitions from static hard-decision decoding to dynamic soft-decision decoding. The system dynamically adjusts LLR values based on voltage differences between read voltage and tracking voltage, and further dynamically updates the LLR table using corrected data from successful decoding operations. This dynamic adaptation enables the system to maintain high reliability across varying threshold voltage conditions while preserving acceptable decoding speed through iterative refinement.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the parameter representation from binary hard decisions to continuous LLR (log-likelihood ratio) values. By calculating voltage differences and using them to adjust LLR values, the system captures nuanced information about threshold voltage distribution overlaps. The LLR table is further refined by incorporating parameters from successfully corrected data, creating a adaptive parameter set that improves error correction capability without sacrificing excessive speed.

Inventive Principle:

Principle #35Parameter changes

2Reliability

If soft-decision decoding is used to improve error correction capability, then decoding accuracy improves, but the complexity of the decoding process increases

Engineering Contradiction:

Improveerror correction capabilityVSAvoiddecoding process complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent performs preliminary action by pre-calculating and storing an LLR table that maps voltage differences to LLR values. This pre-computed table is generated based on expected threshold voltage distributions and is stored for rapid lookup during decoding operations. By performing this complex calculation in advance rather than in real-time, the system reduces the computational burden during actual decoding while maintaining soft-decision accuracy.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback by using data from successful soft-decision decoding operations to update and refine the LLR table. When decoding succeeds, the corrected data is fed back into the system to adjust the LLR table entries, making them more accurate for future operations. This feedback mechanism gradually reduces the complexity gap between hard and soft decision decoding by making the soft decision process more predictable and efficient over time.

Inventive Principle:

Principle #23Feedback

3Productivity

If the LLR table is statically generated, then the decoding process is efficient, but it cannot adapt to changes in threshold voltage distributions

Engineering Contradiction:

Improvedecoding efficiencyVSAvoidadaptability to voltage variations

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent implements feedback by using data from successful soft-decision decoding operations to update and refine the LLR table. When decoding succeeds, the corrected data is fed back into the system to adjust the LLR table entries, making them more accurate for future operations. This feedback mechanism gradually reduces the complexity gap between hard and soft decision decoding by making the soft decision process more predictable and efficient over time.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent transitions from static hard-decision decoding to dynamic soft-decision decoding. The system dynamically adjusts LLR values based on voltage differences between read voltage and tracking voltage, and further dynamically updates the LLR table using corrected data from successful decoding operations. This dynamic adaptation enables the system to maintain high reliability across varying threshold voltage conditions while preserving acceptable decoding speed through iterative refinement.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS20250157564A1Memory system and method

Publication Date: 2025.05.15 KIOXIA CORP

AI summary

A memory system includes a semiconductor memory device including a plurality of memory cells each configured to store data in a non-volatile manner according to a threshold voltage thereof and connected to a word line, and a controller configured to perform an error correction based on hard bit data and soft bit data read from the plurality of memory cells, generate a first table based on corrected data, determine a voltage difference between a first voltage and a second voltage, the first voltage being a voltage applied to the word line when the data being corrected is read, and correct the first table based on the voltage difference.