Lightweight BES Approximation for Read Threshold Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing read threshold calibration operations in data storage devices, such as SSDs and iNAND storage devices, are time-consuming and expensive, leading to increased latency and decreased quality of service due to shifted read thresholds caused by factors like read disturb effects and physical degradation.
Innovation Solution
A data storage device employs a controller to perform read threshold calibration on less than all pages of a representative wordline, using a machine learning model to correlate read thresholds of one page with another page and account for physical conditions, thereby predicting read thresholds for unsensed pages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If read threshold calibration is performed on all pages of a representative wordline, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent applies partial action by performing BES operations on only a subset of pages (e.g., first, second, and third pages) rather than all pages of the representative wordline. The machine learning model then predicts read thresholds for the remaining fourth page, reducing the number of actual measurements needed while maintaining calibration accuracy through intelligent prediction.
Solution Approach 2:
The patent uses copying by creating a machine learning model that learns the relationship between read thresholds of different pages. Once the model is trained on data from some pages, it can generate predicted read thresholds for other pages, effectively copying the calibration information across pages without performing complete BES operations on every page.
2Productivity
If machine learning model is used to predict read thresholds, then productivity is improved, but device complexity increases
Solution Approach 1:
The patent applies preliminary action by performing offline training of the machine learning model using historical BES data from multiple pages. This training phase is completed beforehand, allowing the model to be stored in memory and used for rapid predictions during actual operation, separating the complex training process from the time-critical calibration process.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary between the physical measurement process and the read threshold determination. The model acts as a mediator that takes BES results from sampled pages and physical condition data as input, and outputs predicted read thresholds for all pages, simplifying the overall calibration workflow.
Data Source
AI summary
A data storage device includes a memory device and a controller coupled to the memory device. When a read threshold calibration operation occurs, less than all of the pages of a representative wordline is sensed, such that read thresholds of less than all of the pages of the representative wordline is obtained. The obtained read thresholds and one or more physical conditions of the representative wordline are provided to a model to obtain the other read thresholds of the remaining pages of the representative wordline that were not sensed. The model correlates read thresholds of one page to another page of the same representative wordline and accounts for the one or more physical conditions of the representative wordline.