Adaptive Read Reference Voltage for Flash Memory Error Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Non-volatile memory technologies, such as flash memory, face reliability issues due to analog charge storage leading to errors during data read operations, especially in smaller geometries and multiple-bits-per-cell structures, where errors may not be correctable by existing error checking and correction processes.
Innovation Solution
The system adjusts the reference voltage based on error analysis, comparing the number of 0-to-1 and 1-to-0 errors to determine if and how to adjust the reference voltage, thereby attempting to minimize errors in subsequent read operations, particularly useful in non-volatile memory systems that read sequential data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a fixed reference voltage is used for reading data from non-volatile memory, then the reading process is simple and fast, but errors occur due to charge leakage and analog variation over time
Solution Approach 1:
The reference voltage is transformed from a static fixed value to a dynamic adjustable parameter. The system continuously or periodically adjusts the reference voltage based on detected error patterns, allowing the reference level to adapt to charge leakage and analog variations over time, thereby maintaining reading accuracy without requiring a completely complex new architecture
Solution Approach 2:
An error detection and feedback mechanism is implemented where the system monitors reading errors, analyzes error patterns (such as predominant 0-to-1 or 1-to-0 errors), and uses this feedback to adjust the reference voltage accordingly. This closed-loop control enables the system to self-correct and maintain reliability without external intervention
Solution Approach 3:
The reference voltage parameter is made adjustable rather than fixed. By changing the reference voltage level dynamically based on error analysis, the system compensates for analog charge variations and leakage, effectively resolving the reliability issue without fundamentally changing the memory cell structure or reading architecture
2Quantity of substance
If smaller geometries and multiple-bits-per-cell structures are used to increase storage density, then more data can be stored, but tolerance for charge variation decreases leading to more errors
Solution Approach 1:
The reference voltage is dynamically adjusted to compensate for the reduced noise margin inherent in smaller geometries and multiple-bits-per-cell structures. By adapting the reference level based on detected error patterns, the system maintains adequate separation between voltage thresholds even when charge storage capacity is reduced, thereby preserving reliability while enabling higher density
Solution Approach 2:
The reading system becomes adaptive rather than static, continuously adjusting reference voltages to match the actual charge distribution in high-density cells. This dynamic adaptation allows the system to maintain reliable reading despite the tighter voltage margins required for high-capacity storage
3Reliability
If error checking and correction (ECC) is used to detect and correct reading errors, then some errors can be corrected, but uncorrectable errors still occur and ECC adds processing overhead
Solution Approach 1:
The reference voltage is adjusted in advance based on error pattern analysis before errors become uncorrectable. By proactively adapting the reference level to prevent error accumulation, the system reduces the burden on ECC and minimizes uncorrectable errors, allowing ECC to focus on handling only residual random errors rather than systematic biases
Data Source
AI summary
In a non-volatile memory that reads a binary value from a storage cell by comparing the voltage level of a stored charge in that cell against a reference voltage, the accumulated errors in a range of memory locations may be analyzed to determined if there are more errors in one direction than the other (for example, more 0-to-1 errors than 1-to-0 errors). If so, the reference voltage may be adjusted up or down so that subsequent reads from that range may produce approximately the same number of errors in each direction. For multiple-bits-per-cell memories, where there are multiple reference voltages for each cell, each reference voltage may be adjusted separately by keeping track of the errors related to that particular threshold.