NAND Flash Soft Information Generation via Adaptive Bin Splitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the threshold voltage of memory cells in flash memory devices are not robust or efficient, particularly when reference voltages fall out of sequence, leading to corruption of soft information generation and inability to handle read-to-read variations due to noise and measurement inaccuracies.
Innovation Solution
The system applies a reference signal to a memory cell, measures its output, and uses a bin index generation unit to divide the range of threshold voltage values into bins, splitting them based on the output and voltage value, and assigns indices to these bins to improve resolution and handle invalid measurements, ensuring accurate soft information generation even with out-of-order reference voltages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-pass read with monotonically increasing or decreasing reference voltages is used, then soft information can be generated, but the method is corrupted when reference voltages fall out of order
Solution Approach 1:
The patent applies dynamics by making the read operation adaptive rather than static. The system dynamically adjusts the read process based on actual threshold voltage measurements from previous reads. When reference voltages fall out of monotonically increasing or decreasing order, the system can handle these variations by using the measured threshold voltage distribution to correct or reinterpret the soft information, rather than following a rigid predetermined sequence.
Solution Approach 2:
The patent implements feedback by using the measured output from previous reads to inform subsequent read operations. The system measures the actual threshold voltage distribution of memory cells and uses this information to adjust the interpretation of soft information in later reads. This feedback mechanism allows the system to correct for out-of-order reference voltages and read-to-read variations, maintaining reliability while preserving measurement precision.
2Ease of operation
If current soft information generation methods are used, then reading process is simple, but read-to-read variation due to noise and measurement inaccuracies cannot be handled
Solution Approach 1:
The system uses feedback from measured threshold voltage distributions to correct for noise and measurement variations. By comparing actual measurements against expected distributions and adjusting subsequent reads accordingly, the system maintains measurement precision without significantly complicating the read process.
Solution Approach 2:
The patent changes parameters dynamically based on measured conditions. Instead of using fixed reference voltage sequences, the system adjusts the interpretation parameters based on the actual threshold voltage distribution measured in previous reads. This allows the system to maintain simplicity while adapting to variations in measurement conditions.
3Productivity
If bins are not split based on measured output, then processing is faster, but bin resolution is insufficient for accurate soft information
Solution Approach 1:
The patent applies segmentation by dividing the threshold voltage range into bins and further splitting these bins based on measured outputs. This creates a hierarchical structure where coarse bins provide fast initial classification, and finer split bins provide accurate soft information where needed. The segmentation is adaptive, splitting bins only when measurement precision requirements demand it.
Solution Approach 2:
The system performs partial bin splitting only when necessary for accurate soft information generation. Instead of splitting all bins in all cases, the system selectively applies bin splitting based on the measured output and the specific read conditions, maintaining processing speed while achieving sufficient resolution where required.
Data Source
AI summary
Systems and methods are provided to generate soft information related to the threshold voltage of a memory cell. A range of threshold voltages for the memory cell is divided into subregions of threshold voltage values herein referred to as bins. An output of the memory cell in response to an applied reference signal is measured. The applied reference signal includes a voltage value and position information. A single bin is identified based on the position information of the reference signal. The identified bin is split into more than one bin based on the output of the memory cell and the voltage value of the reference signal. The newly split bins and all the other bins that were not split are assigned new bin indices.