Memory Read Voltage Calibration for Cross-Temperature Drift
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Memory cells in memory systems experience instability due to cross-temperature conditions, leading to inaccurate read voltages and increased refresh frequencies, resulting in read errors and memory access delays, particularly in systems with large temperature variations.
Innovation Solution
A read voltage calibration process that generates offsets based on both slow charge loss and cross-temperature conditions, using linear regression models, AI/ML models, or look-up tables, to improve accuracy and reduce refresh frequency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional read voltage calibration is performed without considering cross-temperature conditions, then the calibration process is simple and fast, but the read voltage accuracy deteriorates under temperature variations
Solution Approach 1:
The patent pre-calibrates read voltage offsets at multiple temperature points (e.g., -40°C, 25°C, 85°C) during manufacturing or initialization, storing these calibration data in lookup tables. During operation, the system retrieves the appropriate calibration data based on current temperature, eliminating the need for complex real-time calculations and achieving high accuracy without increasing operational complexity
Solution Approach 2:
The patent introduces temperature as an additional calibration parameter, creating a multi-dimensional calibration model that accounts for both voltage and temperature variations. This allows the system to adapt read voltage offsets dynamically based on temperature conditions, significantly improving measurement precision under varying thermal environments
2Reliability
If refresh operations are performed frequently to maintain data integrity under cross-temperature conditions, then data reliability is improved, but system performance and productivity deteriorate
Solution Approach 1:
The patent implements a feedback mechanism where the calibrated read voltage offsets are applied during read operations to compensate for temperature-induced variations. This feedback loop ensures that data is read with correct voltage levels even under cross-temperature conditions, maintaining data integrity without requiring excessive refresh operations
Solution Approach 2:
By pre-calculating and storing temperature-compensated read voltage offsets, the system prepares compensation data in advance. This preliminary action allows the system to handle temperature variations efficiently during operation, reducing the need for frequent refreshes and maintaining high productivity while ensuring data reliability
3Measurement precision
If read voltage calibration accounts for both slow charge loss and cross-temperature conditions, then calibration accuracy is improved, but the complexity of determining offsets increases
Solution Approach 1:
The patent segments the calibration process into distinct components: one component handles slow charge loss (time-dependent voltage drift) and another component handles cross-temperature conditions. Each component is calibrated and stored separately, allowing the system to combine these effects additively. This segmentation simplifies the overall complexity by breaking down the complex calibration problem into manageable, independent parts that can be handled separately
Data Source
AI summary
Methods, systems, and devices for cross temperature read voltage calibration for a memory system are described. A memory system may generate read voltage offsets to apply to memory cells during a refresh or calibration operation based on a combination of both a read voltage shift associated with charge loss and cross-temperature conditions of the memory cells. The memory system may program memory cells at a first time and determine a first shift in read voltage for a first logic state as well as a first change in temperature between the first time and second time at which the memory cells are read out. The final read trims to apply during refresh may be calculated by inputting the read voltage shift and the change in temperature to a machine learning (ML) model, a combination of look-up tables (LUTs), or both, each of which may be trained to output read trim values.