Magnetic Disk Read Channel Likelihood Correction for Lower Sector Errors

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Magnetic disk devices face challenges in improving sector error rates due to limitations in error correction processing, particularly when using Soft Output Viterbi Algorithm (SOVA) and Low Density Parity Check (LDPC) methods, which struggle with accurately correcting signals and maintaining reliability.

Innovation Solution

The magnetic disk device employs machine learning to convert uncorrected likelihood values into correction likelihood values, integrating these with basic likelihood values for improved error correction processing, specifically using a neural network to adjust likelihood ratios and enhance LDPC processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If machine learning processing is added to correct likelihood values, then sector error rate is improved, but device complexity increases

Engineering Contradiction:

Improvesector error rateVSAvoidprocessing structure

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent combines machine learning processing with existing SOVA and LDPC error correction processing into a unified error correction system. The machine learning unit receives likelihood values from SOVA and outputs corrected likelihood values that are then used by LDPC, merging multiple processing stages into an integrated solution that improves sector error rate while managing complexity through systematic integration.

Inventive Principle:

Principle #5Merging (Combining)

2Measurement precision

If machine learning is used to correct likelihood values, then error correction accuracy is improved, but processing time increases

Engineering Contradiction:

Improveerror correction accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The machine learning unit performs preliminary correction on likelihood values before they are used in the main LDPC error correction process. By pre-processing the likelihood values to improve their quality, the system enables more accurate and potentially faster convergence in the subsequent LDPC decoding stage, balancing accuracy improvement with processing time management.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system uses feedback from the machine learning processing to continuously improve likelihood value correction. The machine learning unit learns from the relationship between input likelihood values and correct bit sequences, and this learned information is fed back into the correction process to enhance accuracy while optimizing processing efficiency through adaptive learning.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11545183B2Magnetic disk device and read processing method

Publication Date: 2023.01.03 KK TOSHIBA

AI summary

According to one embodiment, a magnetic disk device includes a disk, a head that writes data to the disk and reads data from the disk, and a controller that corrects a first signal into a first likelihood value by machine learning based on a correct learning signal set with a likelihood other than 1 and an incorrect learning signal set with a likelihood other than 0 and executes error correction processing based on a second likelihood value according to the first signal and the first likelihood value.