NAND flash life dynamic prediction method for extreme temperature variation environment
By generating the center and half-width of the read reference voltage loop, and combining the temperature-dependent drift coefficient and the Arrhenius model, the bias problem of NAND flash memory lifetime prediction under extreme temperature change environment is solved, and the lifetime is accurately quantified and separated, improving the stability of the evaluation and the accuracy of the prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN LARIX TECH CO LTD
- Filing Date
- 2026-03-05
- Publication Date
- 2026-06-09
AI Technical Summary
Existing technologies struggle to accurately separate the effects of reversible temperature hysteresis and irreversible aging of NAND flash memory under complex temperature variations, leading to reduced read efficiency and inaccurate lifetime predictions.
By generating the center and half-width of the bidirectional scanning reference voltage loop, and combining the temperature-dependent drift coefficient and the Arrhenius model, the irreversible net displacement and reversible proportional weights are calculated, and the reversible hysteresis effect is removed, thus achieving accurate quantification and separation of lifetime.
It enables accurate dynamic prediction of NAND flash memory lifespan under extreme temperature change environments, improves the stability and specificity of lifespan assessment, ensures that the prediction results are not sensitive to temperature change processes, and supports reliable applications in complex temperature change scenarios.
Smart Images
Figure CN122177191A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of NAND flash memory technology, and more specifically to a method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments. Background Technology
[0002] NAND flash memory has become a core medium for massive data storage due to its high storage density and low power consumption. In practical applications, NAND flash memory often faces extremely complex temperature change scenarios. For example, industrial-grade storage devices often experience a cold write-to-hot read process where data is written at low temperatures and then rapidly heated to high temperatures for reading. In automotive storage systems, as the vehicle operates, it experiences dynamic and significantly different temperature change processes, such as high-temperature writing in the engine compartment and low-temperature reading in the passenger compartment, or writing in extremely cold environments and high-temperature reading under direct sunlight. These temperature change processes can temporarily affect the prediction of NAND flash memory lifespan. However, the reduction in NAND flash memory lifespan caused by these temperature change processes is usually reversible. However, the threshold voltage distribution of NAND flash memory is highly sensitive to temperature and temperature change history. Existing technologies cannot accurately separate the impact of reversible cross-temperature hysteresis from the impact of actual irreversible long-term aging on lifespan in complex temperature change scenarios, which leads to deviations in read reference voltage adjustment and lifespan assessment.
[0003] The threshold voltage of NAND flash memory cells can change due to irreversible aging drift caused by long-term programming / erasing cycles and data retention, as well as reversible offset caused by different temperature change paths and rates during writing and reading at the same read temperature. In existing technologies, read reference voltage adjustment mostly relies on static calibration at a single temperature or only considers dynamic compensation at the current temperature, ignoring two effects: when NAND flash memory undergoes temperature changes, the threshold voltage of the memory cells will experience reversible offset due to transient non-equilibrium of charge migration. If the read reference voltage based on the current temperature static model or the initial factory calibration is still used, it will lead to a large number of bit errors, forcing the error correction code (ECC) to consume more resources. This not only reduces read efficiency, but also accelerates the misjudgment of irreversible aging due to error accumulation, ultimately affecting the prediction of the actual lifespan of NAND flash memory. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention proposes a dynamic prediction method for NAND flash memory lifespan in extreme temperature environments. This method solves the problem that when NAND flash memory undergoes temperature changes, the threshold voltage of the storage cells experiences a reversible shift due to transient imbalances in charge migration, which affects the prediction of the actual lifespan of the NAND flash memory.
[0005] To achieve the above objectives, the present invention provides the following technical solution: The scanning range of the reading reference voltage is set according to the initial reading reference voltage. The first reading reference voltage is generated based on the scanning direction from the lower limit to the upper limit of the scanning range. The second reading reference voltage is generated based on the scanning direction from the upper limit to the lower limit of the scanning range. The first reading reference voltage and the second reading reference voltage are the reading reference voltages that consume the least amount of error correction code under different scanning directions. The lap center is calculated based on the sum of the first and second reference voltages, and the lap center represents the irreversible threshold voltage drift. The lap half-width is calculated based on the difference between the first and second reference voltages, and the lap half-width represents the reversible threshold voltage drift. The difference between the loop center and the initial reading reference voltage is marked as the observed displacement value. The observed displacement value is the total threshold voltage offset. The isothermal drift baseline generated by combining the reading temperature with the data age is obtained. The isothermal drift baseline is the ideal threshold voltage offset. The difference between the observed displacement value and the isothermal drift baseline is marked as the baseline residual. The irreversible net displacement is calculated based on the loop half-width and the baseline residual. The irreversible net displacement is the irreversible aging threshold voltage net offset.
[0006] Furthermore, the write time when data is written to the target storage page in the NAND flash memory and the read time when data is read are obtained. The ambient temperature at the read time is marked as the read temperature. The optimal reference voltage for distinguishing data states is obtained and marked as the initial read reference voltage. The duration obtained by subtracting the write time from the read time is marked as the data age.
[0007] Furthermore, a scan interval covering the initial read reference voltage is set. Starting from the lower limit of the scan interval which is less than the initial read reference voltage, n read reference voltages are generated by increasing the voltage step value by a fixed step value in the direction which is greater than the initial read reference voltage. The value of the last read reference voltage among the n read reference voltages is equal to the upper limit of the scan interval. The resource consumption of the error correction codes of n read reference voltages in the target memory page is detected, and the read reference voltage corresponding to the smallest resource consumption of the error correction code is marked as the first read reference voltage by comparison. The lower limit of the scan interval is equal to the lowest possible threshold voltage of all storage cells in the current target storage page, and the upper limit of the scan interval is equal to the highest possible threshold voltage of all storage cells in the current target storage page.
[0008] Furthermore, the second read reference voltage is generated in the same way as the first read reference voltage, but the scanning direction is as follows: starting from the upper limit of the scanning interval that is greater than the initial read reference voltage, n read reference voltages are generated successively in the direction that is less than the initial read reference voltage with a fixed voltage step value, and the value of the last read reference voltage among the n read reference voltages is equal to the lower limit of the scanning interval.
[0009] Furthermore, the sum of the first and second reference voltages, divided by 2, is marked as the lap center. The value obtained by dividing the difference between the first and second reference voltages by 2 is marked as the loop half-width.
[0010] Furthermore, the threshold voltage of the NAND flash memory cell is obtained as a sequence of changes with storage time under different constant temperatures. The threshold voltage in the sequence is linearly fitted with the logarithm of the corresponding storage time. The slope obtained by the linear fitting is marked as the temperature-dependent drift coefficient, which is the rate of change of the threshold voltage with storage time under different read temperatures. The temperature-dependent drift coefficient of the reading time temperature is obtained, the age of the baseline data is obtained through experiments, and the value obtained by multiplying the temperature-dependent drift coefficient of the reading time temperature by the natural logarithm of the ratio of the data age to the baseline data age is marked as the isothermal drift baseline.
[0011] Furthermore, the invertible proportional weights are calculated based on the loop half-width and baseline residuals, as follows: The value obtained by dividing the absolute value of the loop half-width by the sum of the absolute value of the loop half-width and the absolute value of the baseline residual is labeled as the reversible proportional weight, which is the proportion of the reversible offset in the total non-ideal offset. The value obtained by multiplying the baseline residual by (1 minus the reversible proportional weight) is marked as the irreversible net displacement; The isothermal age correction value is calculated based on the irreversible net displacement. The isothermal age correction value is equivalent to the time when the irreversible net displacement occurs, which is the additional aging time experienced on the basis of the current data age.
[0012] Furthermore, isothermal sensitivity is defined as the temperature-dependent drift coefficient, which is the rate at which the threshold voltage changes with the natural logarithm of the data storage time at a specific temperature. Obtain the isothermal sensitivity of the reading temperature, and mark the value obtained by multiplying the ratio of irreversible net displacement to the isothermal sensitivity of the reading temperature by the data age as the isothermal age correction value. The equivalent age increment at reference temperature is calculated based on the Arrhenius model and isothermal age correction value. The equivalent age increment at reference temperature is the equivalent age increment of the read event at the reference temperature.
[0013] Furthermore, by setting a reference temperature, the activation energy factor of the NAND flash memory is obtained; The formula for calculating the Arrhenius acceleration factor is derived based on the Arrhenius model, and the specific formula is as follows: in, Temperature during reading The Arrhenius acceleration factor under the following conditions As the activation energy factor, Boltzmann's constant, Let exp(·) be the reference temperature, exp(·) be the exponential function, and Arrhenius acceleration factor be the acceleration factor. Temperature during reading The aging rate relative to the reference temperature The acceleration factor of the aging rate; Temperature during reading Acceleration factor The value obtained by multiplying by the isothermal age correction value is marked as the reference temperature equivalent age increment.
[0014] Furthermore, the N reference temperature equivalent age increments generated by the N read events that read data from the target storage page are accumulated in chronological order, and the accumulated value is used as the cumulative lifetime ledger of the target storage page; The longest time that data can be retained in NAND flash memory after N P / E cycles at a reference temperature is obtained, and this time is marked as the ideal lifetime time. The ratio of the cumulative lifespan to the ideal lifespan is marked as the lifespan consumption coefficient, with a value range of [0,1]. The difference between the ideal lifetime time and the cumulative lifetime ledger is marked as the remaining lifetime margin, which reflects the remaining reliable retention time of data in the target storage page.
[0015] Compared with existing technologies, it has the following advantages: This proposed method for dynamic lifetime prediction of NAND flash memory under extreme temperature variations achieves precise quantification and separation of reversible hysteresis and irreversible aging along different paths at the same temperature under extreme temperature variations through end-to-end collaboration. First, by calculating the center and half-width of the read reference loop, the optimal read reference voltage under the two temperature variation paths obtained from bidirectional scanning is abstracted as the loop center and loop half-width. This innovatively transforms cross-temperature path dependence into quantifiable parameters, providing a direct basis for subsequent separation of irreversible aging and reversible hysteresis. This solves the problem that existing technologies cannot distinguish between the two effects, upgrading lifetime assessment from fuzzy calculation of mixed effects to precise quantification of individual effects. Second, when constructing a pathless isothermal baseline, relying on the temperature-related threshold drift rate coefficient and the isothermal early-stage retained baseline, combined with the baseline residual and reversible proportional weight, the interference of reversible cross-temperature hysteresis on the threshold displacement is precisely removed, obtaining only the net displacement reflecting irreversible aging. This process completely eliminates the influence of transient temperature variations on the input of lifetime calculation, ensuring that subsequent lifetime prediction focuses on long-term irreversible aging. By understanding the aging nature of NAND flash memory, the stability and specificity of lifetime assessment are improved. Utilizing the Arrhenius acceleration factor, irreversible net displacement at different read temperatures is uniformly converted to the equivalent time increment at the reference temperature, achieving a unified temporal dimension for aging effects across temperature scenarios. This provides an accumulative basis for lifetime accumulation under multiple temperature variations, solving the problem of inconsistent calculation of aging effects at different temperatures in existing technologies. Finally, based on the accumulated equivalent time increment at the reference temperature and the nominal guaranteed retention time, a dynamic lifetime consumption coefficient and remaining margin are output. Furthermore, because reversible hysteresis is removed in the early stages, the prediction results are insensitive to errors in transient scenarios such as significant temperature changes, truly achieving accurate and dynamic lifetime prediction under extreme temperature variations. This provides technical support for the reliable application of NAND flash memory in complex temperature variation scenarios such as industrial and automotive applications. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Please see Figure 1 This application provides a method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments; The method specifically includes the following steps: Step 1: Obtain the page identifier used to uniquely identify the target storage page in the NAND flash memory; obtain the time when data is written to the target storage page and mark this time as the write time; obtain the time when data is read from the target storage page and mark this time as the read time; obtain the ambient temperature corresponding to the read time and mark this temperature as the read temperature; obtain the optimal reference voltage determined by the NAND flash memory at the factory to distinguish the data state of each storage cell in the initial state and mark this voltage as the initial read reference voltage. The duration obtained by subtracting the write time from the read time is marked as the data age. Specifically, the unit of data age is time, such as seconds or minutes, depending on the actual scenario. By quantifying the time span from writing to reading, the data age can introduce the time dimension into subsequent analysis, providing time dimension support for subsequent research on early retention and self-recovery.
[0019] Step 2: When reading data from the target storage page, set the scan interval of the read reference voltage. The scan interval must cover the initial read reference voltage. Starting from the lower limit of the scan interval which is less than the initial read reference voltage, n read reference voltages are generated by increasing the voltage step value by a fixed step value in the direction which is greater than the initial read reference voltage. The value of the last read reference voltage among the n read reference voltages is equal to the upper limit of the scan interval. The resource consumption of the error correction codes of n read reference voltages in the target memory page is detected, and the read reference voltage corresponding to the smallest resource consumption of the error correction code is marked as the first read reference voltage by comparison. Specifically, NAND flash memory represents data states through the threshold voltage distribution of storage cells. When the threshold voltage of a storage cell is higher than the read reference voltage, it is determined to be in one state; when it is lower than the read reference voltage, it is determined to be in another state. In extreme temperature environments, the threshold voltage will change dynamically, so a new read reference voltage must be generated to match the current threshold distribution. The lowest and highest possible threshold voltages are determined based on the NAND flash memory's read reference voltage design range, thus determining the scan range of the read reference voltage. The lower limit of the scan range is equal to the lowest possible threshold voltage of all storage cells in the current target storage page, and the upper limit of the scan range is equal to the highest possible threshold voltage of all storage cells in the current target storage page, ensuring coverage of all potential optimal read reference voltage boundaries. In this example, considering the voltage accuracy requirements and scan efficiency requirements of NAND flash memory, a fixed voltage step value of 5mV can be set. It should be noted that the smaller the fixed voltage step value... Higher scanning accuracy results in longer read times per operation, requiring specific settings based on actual conditions. Starting from the lower limit of the scanning interval as the first read reference voltage, a fixed voltage step value is incremented each time to generate a new read reference voltage. Each time a read reference voltage is generated, a read operation is performed on the target memory page, and bit errors in the data read under that read reference voltage are detected using error correction codes. The last read reference voltage of the scanning process is the upper limit of the scanning interval. After scanning is complete, the read reference voltage corresponding to the smallest bit error detected by the error correction code is used as the first read reference voltage. The first read reference voltage is the read reference voltage that minimizes the computational resources consumed by the error correction code. The number of bit errors in the data read from the target memory page is minimized under the first read reference voltage. The fewer the errors, the lower the resource overhead of the error correction code. Therefore, the scanning process of determining the first read reference voltage is the process of finding the read reference voltage that minimizes the number of read errors. In addition, starting from the upper limit of the scan interval in the scan interval that is greater than the initial read reference voltage, the read reference voltage is adjusted step by step in the direction that is less than the initial read reference voltage with a fixed voltage step value. After each adjustment, the consumption of error correction code of the target memory page is detected. By comparison, the corresponding read reference voltage with the least consumption of error correction code is marked as the second read reference voltage. Specifically, the process of marking the second reference voltage is the same as that of marking the first reference voltage, except that the scanning direction of the scanning interval is different. The first reference voltage is the optimal reference voltage under the temperature-varying path from low temperature to high temperature, which provides the first dimension of observation for subsequent quantification of cross-temperature path dependence. The second reference voltage is the optimal reference voltage under the temperature-varying path from high temperature to low temperature. The second reference voltage and the first reference voltage form a bidirectional path observation pair, which provides the second dimension of observation for quantification of cross-temperature asymmetry. The sum of the first and second reference voltages divided by 2 is marked as the lap center. Specifically, the lap center represents the irreversible threshold voltage drift, that is, the shift of the threshold voltage distribution center related to long-term irreversible aging such as data retention and wear. The value obtained by dividing the difference between the first and second reading reference voltages by 2 is marked as the loop half-width. Specifically, the loop half-width represents the reversible threshold voltage drift caused by a reversible cross-temperature path, such as when a cold write is performed to a hot read temperature change path. Specifically, the first read reference voltage (the optimal read reference voltage for scanning from low to high) and the second read reference voltage (the optimal read reference voltage for scanning from high to low) correspond to the optimal read boundaries under two temperature-changing paths: cold write to hot read and hot write to cold read. Due to the asymmetry between cross-temperature write and read and the effect of different paths at the same temperature, the two will not coincide. Their difference includes both reversible temperature-changing path hysteresis (such as transient offset) and irreversible long-term aging (such as permanent threshold drift). Based on this, the calculated hysteresis center is the arithmetic mean of the two optimal read references. The hysteresis center physically represents the result after removing the directional dependence of reversible hysteresis. The core threshold position corresponding to irreversible long-term aging can eliminate the back-and-forth shift of reversible effects and focus on the core drift of irreversible aging. The calculated loop half-width is half the difference between the two optimal read reference voltages. The loop half-width can physically quantify the hysteresis of the reversible temperature change path and intuitively reflect the degree of reversible deviation caused by different historical paths at the same temperature. The calculation method of loop center and loop half-width is based on the measured law of cross-temperature asymmetry and the unidirectional cumulative characteristics of irreversible aging. It provides an innovative observation method for the subsequent NAND flash memory lifetime accounting to separate the reversible part and calculate irreversible aging, and realizes the accurate distinction between the two effects.
[0020] Step 3: Mark the difference between the loop center and the initial reading reference voltage as the observed displacement value. Specifically, the observed displacement value is the total offset of the threshold voltage from the initial reading reference voltage to the current loop center, including the offset caused by irreversible aging and the offset caused by reversible transtemperature hysteresis. By continuously monitoring the change sequence of the threshold voltage of the memory cell with storage time at multiple constant temperatures for NAND flash memory samples of the same type, and based on the fact that the threshold drift in the early retention stage conforms to the logarithmic time dependence law, the threshold voltage at each constant temperature and the corresponding logarithmic storage time curve are linearly fitted, and the slope obtained is the temperature-related drift coefficient at that temperature. Based on this, the temperature-related drift coefficient of the read time temperature is obtained. The initial time at which the threshold drift begins to be significantly observable is obtained through experiments, and this initial time is marked as the baseline data age. For example, if the experiment observes that the threshold drift exhibits a stable logarithmic pattern starting 100 seconds after the data is written, then 100 seconds can be used as the baseline data age. The value obtained by multiplying the temperature-dependent drift coefficient of the reading temperature by the natural logarithm of the ratio of the data age to the baseline data age is marked as the isothermal drift baseline. Specifically, the isothermal drift baseline is the path-independent ideal threshold voltage offset determined by the reading temperature and the data age, providing a reference standard for stripping reversible transtemperature hysteresis. The difference between the observed displacement value and the isothermal drift baseline is marked as the baseline residual. Specifically, the baseline residual can reflect the extent to which the actual threshold voltage shift deviates from the path-independent ideal evolution, including additional deviations from reversible transthermal hysteresis and irreversible aging. By stripping away the portion of the total shift that can be explained by the isothermal ideal evolution, the remaining non-ideal deviations can be focused on, providing an intermediate amount for subsequent separation of reversible hysteresis. The invertible proportional weights are calculated based on the loop half-width and baseline residuals, as follows: The value obtained by dividing the absolute value of the loop half-width by the sum of the absolute values of the loop half-width and the baseline residual is marked as the reversible proportional weight. Specifically, the reversible proportional weight is the proportion of the magnitude of the reversible transtemperature hysteresis, i.e., the absolute value of the loop half-width, in the sum of the magnitude of the reversible hysteresis and the absolute value of the baseline residual. This quantifies the proportion of reversible offset in the total non-ideal offset, provides a weight basis for the subsequent stripping of the reversible part, and realizes a quantitative description of the reversible effect. The irreversible net displacement is calculated by baseline residual × (1 - reversible proportional weight). Specifically, the irreversible net displacement is derived from the baseline residual by removing reversible transtemperature hysteresis and only reflects the net offset of the threshold voltage due to irreversible aging. By accurately separating the contributions of reversible transtemperature hysteresis and irreversible aging to the threshold offset, the threshold displacement related only to long-term irreversible aging is obtained. This provides core data for subsequent steps to convert the irreversible net displacement into time dimension and incorporate it into the NAND flash memory lifetime calculation. Specifically, the observed displacement value includes the ideal offset of irreversible aging retained in the early isothermal period, the additional offset of irreversible aging, and the offset caused by reversible transtemperature hysteresis, and is the total offset composed of these three factors; the baseline residual is the total non-ideal deviation of the observed displacement value after deducting the isothermal drift baseline, i.e., the isothermal ideal irreversible offset, including the additional offset of irreversible aging and the offset caused by reversible transtemperature hysteresis; the reversible proportional weight quantifies the proportion of the reversible transtemperature hysteresis amplitude in the sum of the absolute values of the reversible amplitude and the baseline residual; and the irreversible net displacement is the net offset belonging only to irreversible aging obtained after removing the offset caused by reversible transtemperature hysteresis from the total non-ideal deviation. The irreversible net displacement integrates additional irreversible factors other than the isothermal ideal irreversible offset, eliminates the interference of reversible transtemperature hysteresis, and thus provides a single irreversible threshold voltage offset input for the subsequent conversion of irreversible effects into time dimensions and inclusion in lifetime accounting.
[0021] Step 4: Once the irreversible net displacement is obtained, it needs to be converted into an equivalent time increment at the reference temperature. This ensures that the irreversible aging offsets at different temperatures and for different reading events can be accumulated using a unified time dimension, as detailed below: Isothermal sensitivity is defined based on the temperature-dependent drift coefficient. Isothermal sensitivity = temperature-dependent drift coefficient. The isothermal sensitivity represents the rate of change of the threshold voltage with the natural logarithm of the data age at a specific temperature, reflecting the modulating effect of temperature on the relationship between threshold voltage displacement and logarithmic age. The isothermal sensitivity of the reading temperature is obtained, and the value obtained by multiplying the ratio of irreversible net displacement to the isothermal sensitivity of the reading temperature by the data age is marked as the isothermal age correction value. The isothermal age correction value is the data age correction amount equivalent to the irreversible net displacement at the reading temperature, and the unit is time. Specifically, the ratio of irreversible net displacement to isothermal sensitivity can be viewed as the displacement rate, resulting in a logarithmic time increment with the dimension ln(time), rather than the actual time dimension. To convert the logarithmic time increment to the actual time dimension, it is necessary to scale it proportionally based on the current data age. Since logarithmic time and actual time have an exponential relationship, in incremental analysis, the current actual data age can be approximated by multiplying the logarithmic time increment to obtain the equivalent increment of actual time. The isothermal age correction value obtained by multiplying the ratio of irreversible net displacement to isothermal sensitivity at the reading temperature by the current data age is equivalent to the time for irreversible net displacement to be the additional aging time experienced on the basis of the current data age. For example, when the data age = 100 hours, the isothermal sensitivity = 2mV / ln(hours), and the irreversible net displacement = 1mV, the isothermal age correction value = 50 hours is obtained, which means that at the reading temperature, an irreversible net displacement of 1mV is generated, which is equivalent to an additional aging of 50 hours on the basis of the existing 100-hour data age. The reference temperature is set according to actual needs. The reference temperature is the benchmark temperature selected in the scheme study and is used to unify the life calculation of different temperature change scenarios. The activation energy factor of NAND flash memory was obtained. The activation energy factor is the minimum energy required for charge to cross the energy barrier during the thermal activation aging process. It is used to describe the sensitivity of the aging rate to temperature changes. The larger the activation energy factor, the more significant the effect of temperature on the aging rate. It was obtained through device-level characterization experiments. The formula for calculating the Arrhenius acceleration factor is derived based on the Arrhenius model, and the specific formula is as follows: in, Temperature during reading The acceleration factor below, As the activation energy factor, Boltzmann's constant, The reference temperature is exp(·), which is an exponential function and the acceleration factor is exp(·). Temperature during reading The aging rate relative to the reference temperature The acceleration factor is the factor that increases the aging rate at higher reading temperatures. Temperature during reading Acceleration factor The value obtained after multiplying by the isothermal age correction value is labeled as the reference temperature equivalent age increment. The reference temperature equivalent age increment is the equivalent age increment of this reading event at the reference temperature. For example, when =2.14, isothermal age correction value =50 hours, at this time the reference temperature equivalent age increment =107 hours, which means that the irreversible net displacement generated in 50 hours at the reading temperature would require 107 hours to generate the same irreversible net displacement at the reference temperature with a slower aging rate. Specifically, the Arrhenius model is used to uniformly convert the isothermal age correction values at different reading temperatures to the reference temperature, so that irreversible aging under different temperature change scenarios can be accumulated with the same temperature reference time, providing a unified dimension for the cross-temperature accumulation of the subsequent lifespan ledger.
[0022] Step 5: Accumulate the N reference temperature equivalent age increments generated by the N read events that read data from the target storage page in chronological order, and use the accumulated value as the cumulative lifetime ledger of the target storage page; The longest time that data can be retained in NAND flash memory after N P / E cycles at a reference temperature is obtained, and this time is marked as the ideal lifetime time. The ideal lifetime time is obtained experimentally or according to design standards. The ratio of the cumulative lifespan to the ideal lifespan is marked as the lifespan consumption coefficient, with a value range of [0,1]. The closer the lifespan consumption coefficient is to 1, the more severe the lifespan consumption. The difference between the ideal lifetime time and the cumulative lifetime ledger is marked as the remaining lifetime margin, which reflects the remaining reliable retention time of data in the target storage page. Specifically, by using the accumulated lifetime ledger as the core and combining it with the inherent lifetime attributes of NAND flash memory, the calculated lifetime consumption coefficient and remaining lifetime margin are dual indicators that predict the lifetime status of NAND flash memory from different dimensions. This achieves the quantification of irreversible aging from a single read event to the dynamic lifetime prediction of the target page. In addition, because the reversible hysteresis is removed in the early stage, the prediction results can accurately reflect the impact of long-term irreversible aging under extreme temperature change environment.
[0023] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments, characterized in that, include: The scan range of the read reference voltage is set according to the initial read reference voltage. The first read reference voltage is generated based on the scan direction from the lower limit to the upper limit of the scan range. The second read reference voltage is generated based on the scan direction from the upper limit to the lower limit of the scan range. The first read reference voltage and the second read reference voltage are the read reference voltages that consume the least error correction code under different scan directions. The read reference voltage is the preset voltage reference when reading data from NAND flash memory. The lap center is calculated based on the sum of the first and second reference voltages. The lap center represents the irreversible threshold voltage drift. The loop half-width is calculated based on the difference between the first and second reading reference voltages. The loop half-width represents the reversible threshold voltage drift. The difference between the loop center and the initial reading reference voltage is marked as the observed displacement value. The observed displacement value is the total threshold voltage offset. The isothermal drift baseline generated by combining the reading temperature with the data age is obtained. The isothermal drift baseline is the ideal threshold voltage offset. The difference between the observed displacement value and the isothermal drift baseline is marked as the baseline residual. The irreversible net displacement is calculated based on the loop half-width and the baseline residual. The irreversible net displacement is the irreversible aging threshold voltage net offset.
2. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 1 further includes: The system obtains the write time when data is written to the target storage page in the NAND flash memory and the read time when data is read. It marks the ambient temperature at the read time as the read temperature, obtains the optimal reference voltage to distinguish the data state, marks the optimal reference voltage as the initial read reference voltage, and marks the duration obtained by subtracting the write time from the read time as the data age.
3. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 1, characterized in that, The methods for generating the first reading reference voltage include: Set a scan interval that covers the initial read reference voltage. Starting from the lower limit of the scan interval that is less than the initial read reference voltage, generate n read reference voltages by increasing the voltage step value by a fixed step value in the direction that is greater than the initial read reference voltage. The value of the last read reference voltage among the n read reference voltages is equal to the upper limit of the scan interval. The resource consumption of the error correction codes of n read reference voltages in the target memory page is detected, and the read reference voltage corresponding to the smallest resource consumption of the error correction code is marked as the first read reference voltage by comparison. The lower limit of the scan interval is equal to the lowest possible threshold voltage of all storage cells in the current target storage page, and the upper limit of the scan interval is equal to the highest possible threshold voltage of all storage cells in the current target storage page.
4. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 3, characterized in that, The second read reference voltage is generated in the same way as the first read reference voltage, but the scanning direction is as follows: starting from the upper limit of the scanning interval that is greater than the initial read reference voltage, n read reference voltages are generated by decreasing the voltage step value by a fixed step value in the direction that is less than the initial read reference voltage. The value of the last read reference voltage among the n read reference voltages is equal to the lower limit of the scanning interval.
5. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 4, characterized in that, The calculation methods for the center line and half-width of the loop line include: The value obtained by dividing the sum of the first and second reference voltages by 2 is marked as the lap center. The value obtained by dividing the difference between the first and second reference voltages by 2 is marked as the loop half-width.
6. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 5, characterized in that, Methods for generating isothermal drift baselines include: Obtain the sequence of threshold voltage of NAND flash memory cell changing with storage time under different constant temperatures. Linearly fit the threshold voltage in the sequence with the logarithm of the corresponding storage time. Mark the slope obtained by linear fitting as the temperature-dependent drift coefficient. The temperature-dependent drift coefficient is the rate of change of threshold voltage with storage time under different read temperatures. The temperature-dependent drift coefficient of the reading time temperature is obtained, the age of the baseline data is obtained through experiments, and the value obtained by multiplying the temperature-dependent drift coefficient of the reading time temperature by the natural logarithm of the ratio of the data age to the baseline data age is marked as the isothermal drift baseline.
7. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 6, characterized in that, Methods for calculating irreversible net displacement include: The invertible proportional weights are calculated based on the loop half-width and baseline residuals, as follows: The value obtained by dividing the absolute value of the loop half-width by the sum of the absolute value of the loop half-width and the absolute value of the baseline residual is labeled as the reversible proportional weight, which is the proportion of the reversible offset in the total non-ideal offset. The value obtained by multiplying the baseline residual by (1 minus the reversible proportional weight) is marked as the irreversible net displacement; The isothermal age correction value is calculated based on the irreversible net displacement. The isothermal age correction value is equivalent to the time when the irreversible net displacement occurs, which is the additional aging time experienced on the basis of the current data age.
8. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 7, characterized in that, The calculation methods for isothermal age correction values include: Define isothermal sensitivity as the temperature-dependent drift coefficient, which is the rate at which the threshold voltage changes with the natural logarithm of the data storage time at a specific temperature. Obtain the isothermal sensitivity of the reading temperature, and mark the value obtained by multiplying the ratio of irreversible net displacement to the isothermal sensitivity of the reading temperature by the data age as the isothermal age correction value. The equivalent age increment at reference temperature is calculated based on the Arrhenius model and isothermal age correction value. The equivalent age increment at reference temperature is the equivalent age increment of the read event at the reference temperature.
9. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 8, characterized in that, The calculation method for the equivalent age increment based on reference temperature includes: Set a reference temperature and obtain the activation energy factor of the NAND flash memory; The formula for calculating the Arrhenius acceleration factor is derived based on the Arrhenius model, and the specific formula is as follows: in, Temperature during reading The Arrhenius acceleration factor under the following conditions As the activation energy factor, Boltzmann's constant, Let exp(·) be the reference temperature, exp(·) be the exponential function, and Arrhenius acceleration factor be the acceleration factor. Temperature during reading The aging rate relative to the reference temperature The acceleration factor of the aging rate; Temperature during reading Acceleration factor The value obtained by multiplying by the isothermal age correction value is marked as the reference temperature equivalent age increment.
10. The method for dynamically predicting the lifetime of NAND flash memory in extreme temperature environments according to claim 9, characterized in that, The N reference temperature equivalent age increments generated by the N read events that read data from the target storage page in chronological order are accumulated, and the accumulated value is used as the cumulative lifetime ledger of the target storage page. The longest time that data can be retained in NAND flash memory after N P / E cycles at a reference temperature is obtained, and this time is marked as the ideal lifetime time. The ratio of the cumulative lifespan to the ideal lifespan is marked as the lifespan consumption coefficient, with a value range of [0,1]. The difference between the ideal lifetime time and the cumulative lifetime ledger is marked as the remaining lifetime margin, which reflects the remaining reliable retention time of data in the target storage page.