NAND Flash Read Threshold Estimation for Lower SSD Read Errors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SSDs face challenges in minimizing bit error rates due to threshold voltage shifts, leading to increased latency and reduced Quality of Service (QoS) during data read operations, particularly in NAND flash memories, as existing Quick Training (QT) algorithms rely on linear fits that do not accurately capture the nonlinear relationship between mock histogram values and read thresholds.
Innovation Solution
Implementing nonlinear transformations of mock histogram features and using a deep-neural network (DNN) or iterative weighted least squares algorithm to estimate optimal read thresholds, combined with feature selection to reduce computational complexity and improve HB decoding success rates, thereby reducing the need for soft decision decoding and minimizing bit error rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If linear fit algorithms are used for threshold estimation, then computational complexity is reduced, but threshold estimation accuracy deteriorates due to inability to capture nonlinear relationships
Solution Approach 1:
The patent transforms the input parameters by applying nonlinear transformations (power transformations with exponent γ, logarithmic transformations) to the mock histogram values before feeding them to the estimation algorithm. This allows the system to capture nonlinear relationships between threshold voltages and histogram features while maintaining computational efficiency through closed-form solutions or pre-computed lookup tables.
Solution Approach 2:
The patent introduces curvature into the estimation model by using power transformations (raising mock histogram values to power γ) and logarithmic transformations. These transformations bend the linear relationship into a curved one, allowing the model to capture the nonlinear behavior of threshold voltage shifts while still using relatively simple computational operations.
2Measurement precision
If more features are used in the transformation, then threshold estimation accuracy improves, but computational complexity increases
Solution Approach 1:
The patent uses a selective feature approach where only the most relevant transformed features are included in the final estimation model. By applying power transformations and logarithmic transformations to specific moments of the histogram (mean, variance, skewness) rather than all possible features, the system achieves good accuracy with reduced computational burden.
Solution Approach 2:
The patent extracts only the essential features from the mock histogram distribution that are most relevant to threshold voltage estimation. By focusing on key statistical moments (mean, variance, skewness) and applying transformations only to these selected features, the system avoids the computational overhead of processing all possible histogram features while maintaining estimation accuracy.
3Speed
If hard decision decoding is used, then processing speed is improved, but bit error rate increases due to threshold voltage shifts
Solution Approach 1:
The patent performs preliminary threshold voltage estimation using the transformed mock histogram features before the actual data reading and decoding process. By pre-calculating accurate threshold values based on the observed histogram distribution, the system enables hard decision decoding to work effectively without suffering from threshold voltage shifts, thus maintaining both speed and reliability.
Solution Approach 2:
The patent uses the mock histogram data obtained during the read process as feedback to dynamically adjust the threshold voltage estimates. This feedback mechanism allows the system to adapt to threshold voltage shifts caused by wear and environmental conditions, maintaining low bit error rates while using fast hard decision decoding.
Data Source
AI summary
A method for reading data from a solid-state drive (SSD) configured to store data in a plurality of memory cells arranged in memory blocks comprising rows, the method performed by a controller in communication with the plurality of memory cells. The method comprises retrieving data from a target row of memory cells of the plurality of memory cells associated with a read request received from a host using initial threshold voltages. The method also includes decoding the data using a hard decision stage. Additionally the method comprises estimating read threshold voltages of the target row of memory cells based on a transformation of a distribution of threshold voltages of cells in a memory block containing the target row when the hard decision decoding stage fails. The method further includes retrieving data from the target row using the estimated read threshold voltages.