Group-Based Data Training for Nonvolatile Memory Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-speed and high-capacity storage devices with multiple nonvolatile memory devices face challenges in reducing data training time, which is essential for speedy access and reliability, especially during power-up sequences like booting.
Innovation Solution
A storage device and method that categorize nonvolatile memory devices into groups, perform data training on a representative device within each group, and apply the training results to other devices in the group, thereby reducing the overall data training time by skipping unnecessary training for non-representative devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data training is performed on all nonvolatile memory devices individually, then data reliability is ensured, but data training time increases significantly
Solution Approach 1:
The patent divides nonvolatile memory devices into multiple groups based on their physical characteristics and connection paths. Data training is performed on representative devices from each group, and the training results are applied to all devices within that group. This segmentation approach maintains data reliability while significantly reducing the overall training time by avoiding redundant training of identical devices.
Solution Approach 2:
The patent creates a representative nonvolatile memory device for each group that copies the essential characteristics of all devices in that group. Data training performed on the representative device effectively trains all other devices in the same group, as they share identical or similar electrical and physical properties. This copying principle allows one training operation to serve multiple devices simultaneously.
2Quantity of substance
If the number of nonvolatile memory devices is increased for high capacity, then storage capacity improves, but open timing increases and access speed decreases
Solution Approach 1:
The patent segments the large number of nonvolatile memory devices into multiple groups based on their connection characteristics and physical properties. By performing data training on representative devices from each group rather than individually training all devices, the system maintains high storage capacity while reducing the open timing required during power-up, thereby improving access speed.
3Loss of time
If data training is performed on buffer chips instead of nonvolatile memory devices, then data training time is reduced, but training accuracy may be compromised
Solution Approach 1:
The patent introduces buffer chips as intermediary components between the storage controller and nonvolatile memory devices. Data training is performed on the buffer chips, which act as mediators that can represent the electrical characteristics of multiple memory devices. The training results obtained from buffer chips are then applied to the actual nonvolatile memory devices, achieving both time reduction and maintained accuracy.
Data Source
AI summary
A storage device includes a plurality of nonvolatile memory devices each exchanging data by using a data strobe signal and a data signal, and a storage controller categorizing the plurality of nonvolatile memory devices into a plurality of groups and performing training in units of the plurality of groups. The storage controller performs data training on a first nonvolatile memory device selected in a first group of the plurality of groups and sets a delay of a data signal of a second nonvolatile memory device included in the first group by using a result value of the data training for the first nonvolatile memory device.