Memory Metric Filtering for Noise-Resistant Read-Write Decisions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Memory management processes in non-volatile memory devices are susceptible to noise and metric variation, leading to inaccurate decision-making and potential errors in read-write operations due to metric-to-metric variation.
Innovation Solution
Employing a low pass filter to mitigate metric variation in memory management processes, reducing noise and improving the accuracy of read-write reliability assessments by using filtered metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If memory management processes use raw metrics to evaluate read-write reliability, then the process can operate with simpler architecture, but the decisions become inaccurate due to noise and metric variation
Solution Approach 1:
A low pass filter is introduced as an intermediary component between the metric collection system and the memory management decisions. The filter processes raw metrics (cell count values) to produce smoothed, noise-reduced metrics that accurately reflect true memory reliability characteristics. This mediator eliminates the direct connection between noisy raw data and decision-making, resolving the contradiction by adding a processing layer that improves accuracy without fundamentally redesigning the entire system.
Solution Approach 2:
The system changes the parameter characteristics of the metrics through filtering operations. By applying the low pass filter, the metric parameters are transformed from raw, noisy values to smoothed, reliable values. The filter modifies the temporal and statistical properties of the metrics, converting them from unstable measurements to stable indicators that accurately represent memory reliability for decision-making.
2Reliability
If the system collects and processes multiple metrics to reduce noise, then read-write reliability assessment improves, but the time required for evaluation increases
Solution Approach 1:
The low pass filter operates continuously on incoming metric data, maintaining a running average that smoothly updates as new measurements arrive. This continuous processing eliminates the need for batch evaluations or repeated measurements, allowing the system to achieve reliable metrics in real-time. The filter's continuous action ensures that reliability assessments are always based on the most current, noise-reduced data without requiring additional time for separate validation steps.
3Measurement precision
If no filtering is applied to metrics, then the system architecture remains simple and fast, but metric-to-metric variation and noise lead to incorrect memory management decisions
Solution Approach 1:
The low pass filter serves as a simple intermediary that sits between metric collection and decision-making logic. It processes each metric value through a straightforward filtering algorithm, producing cleaned metrics that eliminate noise and variation. This mediator approach improves metric accuracy without requiring complex redesign of the entire memory management architecture, resolving the contradiction by adding minimal complexity only where needed.
Data Source
AI summary
In some implementations, a controller of a memory device may obtain a first metric associated with a memory of the memory device using a first memory read configuration. The controller may apply a function to the first metric to obtain a second memory read configuration. The controller may obtain a second metric associated with the memory using the second memory read configuration. The controller may filter the first metric and the second metric to obtain a first filtered metric and a second filtered metric. The controller may provide the first filtered metric and the second filtered metric to a memory management process executing on the controller. The controller may perform an action based on an output of the memory management process, wherein the output is based on the first filtered metric and the second filtered metric.