Temperature-Assisted NAND Flash Management
By using temperature sensors and regression models in NAND flash memory devices, the threshold voltage is optimized in real time, and the delay and bit error rate problems when reading temperature are solved, improving the reliability and life of the device.
Patent Information
- Application Number
- CN202011260215.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2020-09-21
- Filing Date
- 2020-11-12
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2040-11-12
AI Technical Summary
Existing NAND flash memory devices have delay and bandwidth limitations when reading temperatures, resulting in increased bit error rate and shortened device life, especially when temperature changes, making it difficult to optimize threshold voltages in real time, resulting in long and potentially Error convergence.
By introducing temperature sensors into NAND flash memory devices, a multivariate temperature regression model and voltage regression model are established based on the read temperature data, real-time on-chip temperature is predicted and an estimated threshold voltage is generated, and a local scan is performed to obtain the minimum read bit error code The final reference voltage of the rate, avoiding lengthy voltage sweep iterations and preventing erroneous convergence.
It realizes real-time monitoring and optimization of NAND device temperature without sacrificing the data bus bandwidth, reducing the delay and bit error rate of read operations, and improving the reliability and life of the device.
Smart Images

Figure CN112365908B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the management and implementation of NAND flash memory devices, and more particularly to the management of temperature-assisted NAND flash memory devices. Background Art
[0002] The development of modern computing systems has been driven to some extent by the emergence of solid-state drives (SSDs), which offer higher speed and latency performance than traditional hard disks. Unlike hard disks that rely on magnetism to store data, solid-state drives use NAND flash memory devices to achieve data storage. NAND flash memory devices are a series of integrated circuits manufactured through advanced process and packaging technologies that enable multiple layers of storage cells to be stacked vertically into smaller area particles and packages, thereby achieving high storage capacity.
[0003] With the breakthrough of new NAND flash memory process technology, such as multi-level cell (MLC) and triple-level cell (TLC), the storage capacity of SSD has been improved. However, the new technology breakthrough also brings new challenges, namely increased bit error rate and shortened device life due to the reduction of threshold voltage interval. The charge level representing each storage state is usually in a Gaussian statistical distribution with a mean and variance, where the boundary between adjacent distributions represents the optimal level of the read reference voltage that can achieve the minimum read bit error rate.
[0004] The lifetime and reliability of NAND devices also show a great dependence on temperature. On the one hand, retention errors increase with increasing temperature. On the other hand, threshold voltage increases with higher programming temperatures, which helps reduce bit error rates and improve reliability.
[0005] Most existing NAND flash devices include a temperature sensor. The sensor output can be read out by the memory controller with a latency of about a few hundred microseconds. This feature was originally designed to prevent overheating, and it is not very efficient to perform the readout of the NAND temperature sensor by the controller because the use of the data bus will be limited during the readout, resulting in longer delays and reduced bandwidth for regular programming and read operations.
[0006] If the NAND temperature is not read accurately and in a timely manner, when the controller is aware of the potential threshold voltage shift due to temperature changes, the SSD controller usually adopts a brute force read retry method, that is, incrementally scanning the reference voltage until the optimal voltage level with the lowest bit error rate is found. This method is simple, but usually time-consuming, because the voltage scan may usually require 5 to 20 read retry iterations. In addition, when the threshold voltage changes more significantly, the incremental voltage scan may point in the wrong direction and fail to converge to a solution.
[0007] In modern SSDs, there is a need in the art for a solution to set and optimize NAND reference voltage with real-time temperature. Summary of the invention
[0008] The present disclosure provides a temperature-assisted hardware technique to actively monitor and model the temperature of a NAND device without sacrificing NAND data bus bandwidth. In an exemplary embodiment, a method is provided, which may include: obtaining a readout from a temperature sensor in a non-volatile memory system, generating a predicted real-time on-chip temperature for a non-volatile memory device in the memory system based on the readout, generating an estimated threshold voltage for reading data stored in the non-volatile memory device based on the predicted real-time on-chip temperature, and performing a local scan of a reference voltage using the estimated threshold voltage as a starting point to obtain a final read reference voltage with a minimum read bit error rate.
[0009] In another exemplary embodiment, a nonvolatile storage system is provided. The nonvolatile storage system may include a nonvolatile storage device, a temperature sensor, and a processor. The processor may be configured to obtain a readout from the temperature sensor, generate a predicted real-time on-chip temperature for the nonvolatile storage device based on the readout, generate an estimated threshold voltage for reading data stored in the nonvolatile storage device based on the predicted real-time on-chip temperature, and perform a local scan of a reference voltage using the estimated threshold voltage as a starting point to obtain a final read reference voltage with a minimum read bit error rate.
[0010] In yet another exemplary embodiment, a non-transitory machine-readable medium is provided. The non-volatile machine-readable medium may have computer instructions that, when executed by a hardware processor, may cause the hardware processor to perform the steps of obtaining a reading from a temperature sensor in a non-volatile memory system, generating a predicted real-time on-chip temperature for a non-volatile memory device in the memory system based on the reading, generating an estimated threshold voltage for reading data stored in the non-volatile memory device based on the predicted real-time on-chip temperature, and performing a local scan of a reference voltage using the estimated threshold voltage as a starting point to obtain a final read reference voltage with a minimum read bit error rate. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] Figure 1A A nonvolatile storage system according to an embodiment of the present disclosure is schematically shown.
[0012] Figure 1B A nonvolatile storage system according to another embodiment of the present disclosure is schematically shown.
[0013] Figure 1C A nonvolatile storage system according to yet another embodiment of the present disclosure is schematically illustrated.
[0014] Figure 2 is a flow chart of a process for temperature-assisted non-volatile storage device management according to one embodiment of the present disclosure. DETAILED DESCRIPTION
[0015] Now, specific embodiments according to the present application will be described in detail with reference to the accompanying drawings. For consistency, the same elements in various figures are represented by the same reference numerals.
[0016] The present disclosure provides systems and methods for temperature-assisted hardware technology to actively monitor and model the temperature of non-volatile storage devices without sacrificing non-volatile storage data bus bandwidth. As used herein, a non-volatile memory device can be a computer storage device that can retain stored information after power failure and can retrieve the stored information after power is restored (turned off and reopened). Non-volatile storage devices may include NAND flash memory, NOR flash memory, magnetoresistive random access memory (MRAM), resistive random access memory (RRAM), phase change random access memory (PCRAM), Nano-RAM, etc. NAND flash memory can be used as an example to demonstrate the electronic mirroring technology implemented by the controller. However, other types of non-volatile storage devices can be used to implement the technology according to various embodiments of the present disclosure.
[0017] FIG1 schematically illustrates a nonvolatile storage system 100A according to one embodiment of the present disclosure. The nonvolatile storage system 100A may include a nonvolatile storage controller 102 and a nonvolatile storage device 104. When the nonvolatile storage system 100A is coupled to a host, it may provide the host with data storage and / or access to stored data. The one nonvolatile storage device 104 may be a nonvolatile memory (NVM), such as a NAND device. It should be noted that the nonvolatile storage system 100A may include multiple nonvolatile storage devices, and the nonvolatile storage device 104 may be illustrated as a representative of multiple nonvolatile storage devices.
[0018] The non-volatile storage controller 102 may include a processor 106, a non-volatile computer readable storage medium 108, and a temperature sensor 110A. The processor 106 may be a computer processor, such as but not limited to a microprocessor or a microcontroller. The temperature sensor 110A may provide a readout for temperature. The reference threshold voltage for acquiring data stored on the non-volatile storage device 104 may be affected by the real-time on-chip temperature. The on-chip temperature may refer to the temperature on the non-volatile storage device 104.
[0019] In some embodiments, a thermal model may be built based on the readings from the temperature sensor 110A to predict the real-time on-die temperature. That is, the readings from the temperature sensor 110A may be used to obtain the predicted real-time on-die temperature of the one non-volatile storage device 104 .
[0020] In one embodiment, a multivariate temperature regression model may be developed to model the on-chip temperature of a non-volatile memory device as a function of the temperature with corrections (e.g., T read-out + correction). For example, the readout of the temperature sensor 110A at the latest timestamp can be used as the initial temperature value T to be corrected. read-out . The temperature correction can depend on a variety of factors, including the readout of the temperature sensor, the physical distance from the temperature sensor 110A to the non-volatile storage device 104, the thermal properties of the thermally conductive materials in the system, the frequency of access to the non-volatile storage device 104, and the intensity of read / write access (the number of accesses in a fixed time interval). These factors may be various variables in the regression model. The respective coefficients of each variable and the mathematical relationship of each variable (e.g., linear, polynomial, logarithmic, etc.) can be obtained by accurately fitting the simulation results to the measured data (e.g., least squares, least absolute deviation regression, etc.). For example, in a controlled setting (e.g., a laboratory or experiment), a temperature sensor can be used to monitor the actual real-time on-chip temperature, and the regression model can be fit using the actual real-time on-chip temperature.
[0021] In some embodiments, the temperature regression model may be stored as computer instructions in the non-transitory computer-readable storage medium 108 and may be executed by the processor 106 during operation of the system 100A to obtain a predicted real-time on-chip temperature of the non-volatile storage device 104 based on a readout from the temperature sensor 110A.
[0022] Using the real-time predicted temperature of the non-volatile memory device 104, an estimated threshold voltage can be obtained. In some embodiments, a multivariate voltage regression model can be used to obtain the estimated threshold voltage. The estimated threshold voltage can be modeled as an initial reference voltage V0 with a voltage correction ΔV. The initial reference voltage V0 can be the threshold voltage immediately after the cell in the non-volatile memory device 104 is programmed. During programming, a higher programming temperature may increase the median of each charge state, and therefore also increase the reference voltage that may have the minimum read bit error rate. In one embodiment, the initial reference voltage can be obtained based on the on-chip temperature immediately after the cell is programmed and the number of program / erase (P / E) cycles of the cell. For example, an empirical model can be established to estimate the initial reference voltage based on the on-chip temperature immediately after the cell is programmed and the number of P / E cycles of the cell. The voltage correction may depend on a variety of factors, including the predicted real-time on-chip temperature, self-recovery effects, data retention time, and P / E cycle number. These factors may be various variables in the voltage regression model.
[0023] In one embodiment, the storage controller 102 may save a time stamp each time a storage block in the non-volatile storage device 104 is programmed. Thus, the dwell time may be calculated as the time period between two consecutive programming accesses. After a predetermined number of P / E cycles, an average dwell time may be obtained over multiple P / E cycles. Self-recovery influencing factors may be represented by the average dwell time in a multivariate regression model. Typically, the average dwell time obtained from a newer P / E cycle may have a more effective self-recovery effect than an earlier P / E cycle. For example, the average dwell time used for the multivariate regression model may be obtained in the last 20 to 50 P / E cycles or the last 10% of P / E cycles (e.g., from 20 to up to hundreds of P / E cycles). For each storage block of the non-volatile storage device 104, the storage controller 102 may keep a record of the P / E cycle count. For data retention time, a timestamp of programming the storage block of the non-volatile storage device 104 may be used. The data retention time may be obtained as the time difference between a read access and the last programming access. The reference voltage may also increase with increasing dwell time due to self-recovery effects. Longer dwell times help extend the life of NAND devices. Generally, the longer the dwell time, the slower the retention loss and the slower the change in reference voltage.
[0024] In some embodiments, as the data retention time increases, the estimated threshold voltage V0 may be lower than the initial reference voltage. That is, the adjustment ΔV may be a negative value, and the absolute value of the adjustment is larger when the data retention time is longer. In addition, in one embodiment, a higher predicted real-time on-chip temperature may cause the reference voltage to decrease more quickly, and the voltage regression model generates a lower estimated threshold voltage. However, in some embodiments, an increase in the dwell time may cause the reference voltage to increase, and the voltage regression model generates a higher estimated threshold voltage. This may be due to self-recovery effects. In addition, a higher programming temperature will increase the median value of each charge state, and therefore may also increase the optimal reference voltage and cause the voltage regression model to generate a higher estimated threshold voltage. These effects of various factors on the estimated threshold voltage may be examples, and the regression fitting process may generate corresponding coefficients for these effects to be reflected in the generated estimated threshold voltage.
[0025] In some embodiments, the voltage regression model may also be stored as computer instructions in the non-transitory computer-readable storage medium 108 and may be executed by the processor 106 during operation of the system 100A to obtain an estimated reference voltage for the non-volatile storage device 104 .
[0026] Using the estimated threshold voltage, the processor 106 can be configured to perform a local scan process to find the final reference voltage with the minimum read bit error. In some embodiments, the processor 106 can use the estimated reference voltage generated by the regression model as the starting point for performing a local scan of the reference voltage. In one embodiment, the processor 106 can be configured to adjust the reference voltage according to the estimated reference voltage. When performing a local scan, the processor 106 can be configured to control the storage controller 102 to adjust the reference voltage for reading without blindly performing an incremental scan of the reference voltage. For example, the processor 106 can be configured to control the storage controller 102 to send a command to the non-volatile storage device 104 to instruct the non-volatile storage device 104 to read from the non-volatile storage device 104 using the adjusted reference voltage. The controller can use the reference voltage adjusted by each voltage step to perform a read or read retry operation. Based on the data received from the non-volatile storage device 104, a bit error rate or error count can be calculated for each read, and the processor 102 can determine the final read reference voltage with the minimum read error. The final read reference voltage may be the new best reference voltage because it may have the lowest bit error rate. Embodiments using this method not only eliminate lengthy voltage scan iterations in conventional methods, but also prevent the controller 102 from falling into erroneous results due to significant threshold voltage variations.
[0027] In some embodiments, the local scan process may also be stored as computer instructions in the non-transitory computer-readable storage medium 108 and may be executed by the processor 106 to perform the local scan during operation of the system 100A. The processor 106 may control the controller 102 to send a command to the non-volatile storage device 104 using the locally adjusted reference voltage to read data from the non-volatile storage device 104.
[0028] Figure 1B FIG. 1 schematically shows a nonvolatile storage system 100B according to another embodiment of the present disclosure. Figure 1B In the illustrated embodiment, the system 100B may include a temperature sensor 110B on a printed circuit board (PCB) that carries the storage controller 102 and the non-volatile storage device 104. Except that the location of the temperature sensor 110B is different from that of the temperature sensor 110A, the other components of the system 100B may be the same as their respective counterparts of the system 100A. Similar to the temperature sensor 110A, the readout from the temperature sensor 110B may be used to obtain a predicted real-time on-chip temperature of the non-volatile storage device 104. It should be noted that the temperature regression model developed using the temperature sensor 110B may contain the same set of variables and may be constructed in the same manner as using the readout from the temperature sensor 110A. However, some coefficients of certain variables (e.g., factors) may be different. The voltage regression model and the local scanning process may be the same as those of the system 100A.
[0029] Figure 1C FIG. 1 schematically shows a nonvolatile storage system 100C according to another embodiment of the present disclosure. Figure 1C In the illustrated embodiment, the system 100C may include a temperature sensor 110C located on the non-volatile storage device 104. Except that the location of the temperature sensor 110C is different from that of the temperature sensor 110A, the other components of the system 100C may be the same as their respective counterparts of the system 100A. Similar to the temperature sensor 110A, the reading from the temperature sensor 110C may be used to obtain a predicted real-time on-chip temperature of the non-volatile storage device 104.
[0030] It should be noted that although the temperature sensor 100C may be located on the non-volatile storage device 104, the system 100C does not need to poll the temperature sensor 100C in real time to obtain the real-time on-chip temperature. Because the use of the data bus may be limited, conventional programming and reading operations may be given higher priority than obtaining a readout from the temperature sensor 110C. Therefore, as with the readout of the temperature sensor 110A in the system 100A, the previous readout obtained at the last timestamp may be used to predict the real-time on-chip temperature. The temperature regression model and factors (e.g., variables in the model) may be constructed in the same manner as using the readout from the temperature sensor 110A, but some coefficients for certain variables (e.g., factors) may be different. The voltage regression model and local scanning process may be the same as that of the system 100A.
[0031] It should be noted that some embodiments of the non-volatile storage system may have more than one temperature sensor. For example, an embodiment of the non-volatile storage system may include any two or all of the temperature sensors 110A, 110B, and 110C. In some embodiments where there are multiple temperature sensors in the system, the readout of each individual temperature sensor may be used as a factor in a temperature regression model to predict the real-time on-chip temperature of the NAND flash memory device.
[0032] In an embodiment of a non-volatile storage system 100C where the non-volatile storage device 104 includes a temperature sensor 110C, the temperature sensor 110C may be periodically polled to proactively adjust the predicted real-time on-chip temperature. This may help improve the accuracy of the predicted temperature while minimizing the impact on the storage device data bus and system throughput.
[0033] It should be noted that in various embodiments, the non-volatile storage system (e.g., 100A, 100B, or 100C) may include multiple non-volatile storage devices 104, and the predicted real-time on-chip temperature may be obtained individually for any one of the non-volatile storage devices 104 to read the data stored in each non-volatile storage device.
[0034] In at least one embodiment, some or all of the processor 106 , the storage medium 108 , and the temperature sensor 110A (if any) may be integrated into one integrated circuit (IC) chip.
[0035] Figure 22 is a flow chart of a process 200 for obtaining a final reference voltage in a non-volatile storage device according to one embodiment of the present disclosure. In block 202, a readout may be obtained from a temperature sensor in a non-volatile storage system. For example, a readout from a temperature sensor 110A, 110B, or 110C may be obtained from a non-volatile storage system 100A, 100B, or 100C. In block 204, a predicted real-time on-chip temperature may be generated for a non-volatile storage device in the storage system based on the readout. In various embodiments, a temperature regression model may be used to determine a temperature correction for the readout taking into account a plurality of factors. In block 206, an estimated threshold voltage for reading data stored in a non-volatile storage device may be generated. In various embodiments, an estimated threshold voltage may be generated based on a predicted real-time on-chip temperature. For example, a voltage regression model may be used to generate an estimated threshold voltage based on a predicted real-time on-chip temperature and taking into account a plurality of factors.
[0036] In block 208, a local scan of the reference voltage may be performed using the estimated threshold voltage as a starting point to obtain a final read reference voltage with a minimum read bit error rate. In one embodiment, the local scan of the reference voltage may be started by decreasing from the estimated threshold voltage.
[0037] Process 200 may be implemented using software (e.g., executable by a computer processor (CPU, GPU, or both)), hardware (e.g., a field programmable gate array (FPGA) or an application specific IC (ASIC)), firmware, or any suitable combination of the three. In one embodiment, for example, process 200 may be implemented in a hardware circuit (e.g., an FPGA or an ASIC). In another embodiment, process 200 may be implemented in firmware and stored in a read-only memory (e.g., memory 108), and executed by processor 106 executing the firmware. In addition, process 200 may be programmed in computer processor executable instructions, stored in a non-transitory machine-readable medium (e.g., a hard drive, a CD, a DVD, etc.), and executed by a computer processor (e.g., a microprocessor or processor) executing the executable instructions.
[0038] In an exemplary embodiment, a method is provided, which may include obtaining a readout from a temperature sensor in a nonvolatile memory system, generating a predicted real-time on-chip temperature for a nonvolatile memory device in the memory system based on the readout, generating an estimated threshold voltage for reading data stored in the nonvolatile memory device based on the predicted real-time on-chip temperature, and performing a local scan of a reference voltage using the estimated threshold voltage as a starting point to obtain a final read reference voltage with a minimum read bit error rate.
[0039] In one embodiment, the temperature sensor may be located on a memory controller of the nonvolatile memory system, a printed circuit board of the nonvolatile memory system, or on the nonvolatile memory device.
[0040] In one embodiment, the predicted real-time on-chip temperature may be generated by a temperature regression model using the readout as input.
[0041] In one embodiment, the temperature regression model can be a multivariate regression model based on multiple factors, which may include the readout, the physical distance between the temperature sensor and the non-volatile storage device, the thermal properties of the thermally conductive material in the non-volatile storage system, and the access frequency and read / write access intensity of the non-volatile storage reserve.
[0042] In one embodiment, the estimated threshold voltage may be generated by a voltage regression model using the predicted real-time on-chip temperature as an input.
[0043] In one embodiment, the voltage regression model may be a multivariate regression model based on multiple factors, and the multiple factors may include the predicted real-time on-chip temperature, average dwell time, data retention time, and P / E cycle number.
[0044] In one embodiment, the estimated threshold voltage generated by the voltage regression model may decrease as the data retention time increases.
[0045] In one embodiment, the plurality of factors may include a programming temperature, and the estimated threshold voltage generated by the voltage regression model may increase as the programming temperature increases.
[0046] In another exemplary embodiment, a nonvolatile storage system is provided. The nonvolatile storage system may include a nonvolatile storage device, a temperature sensor, and a processor. The processor may be configured to obtain a readout from the temperature sensor, generate a predicted real-time on-chip temperature for the nonvolatile storage device based on the readout, generate an estimated threshold voltage for reading data stored in the nonvolatile storage device based on the predicted real-time on-chip temperature, and perform a local scan of a reference voltage using the estimated threshold voltage as a starting point to obtain a final read reference voltage with a minimum read bit error rate.
[0047] In one embodiment, the temperature sensor may be located on a memory controller of the nonvolatile memory system, a printed circuit board of the nonvolatile memory system, or on the nonvolatile memory device.
[0048] In one embodiment, the predicted real-time on-chip temperature may be generated by a temperature regression model using the readout as input, and the temperature regression model may be a multivariate regression model based on multiple factors, the multiple factors may include the readout, the physical distance between the temperature sensor and the non-volatile storage device, the thermal properties of the thermally conductive material in the non-volatile storage system, and the access frequency and read / write access intensity of the non-volatile storage reserve.
[0049] In one embodiment, the estimated threshold voltage may be generated by a voltage regression model using the predicted real-time on-chip temperature as an input, wherein the voltage regression model may be a multivariate regression model based on multiple factors including the predicted real-time on-chip temperature, average dwell time, data retention time, and number of P / E cycles.
[0050] In one embodiment, the estimated threshold voltage generated by the voltage regression model may decrease as the data retention time increases.
[0051] In one embodiment, the plurality of factors may include a programming temperature, and the estimated threshold voltage generated by the voltage regression model may increase as the programming temperature increases.
[0052] In yet another exemplary embodiment, a non-transitory machine-readable medium is provided. The non-volatile machine-readable medium may have computer instructions that, when executed by a hardware processor, may cause the hardware processor to perform the steps of obtaining a reading from a temperature sensor in a non-volatile memory system, generating a predicted real-time on-chip temperature for a non-volatile memory device in the memory system based on the reading, generating an estimated threshold voltage for reading data stored in the non-volatile memory device based on the predicted real-time on-chip temperature, and performing a local scan of a reference voltage using the estimated threshold voltage as a starting point to obtain a final read reference voltage with a minimum read bit error rate.
[0053] In one embodiment, the temperature sensor may be located on a memory controller of the nonvolatile memory system, a printed circuit board of the nonvolatile memory system, or on the nonvolatile memory device.
[0054] In one embodiment, the predicted real-time on-chip temperature may be generated by a temperature regression model using the readout as input, and the temperature regression model may be a multivariate regression model based on multiple factors, the multiple factors may include the readout, the physical distance between the temperature sensor and the non-volatile storage device, the thermal properties of the thermally conductive material in the non-volatile storage system, and the access frequency and read / write access intensity of the non-volatile storage reserve.
[0055] In one embodiment, the estimated threshold voltage may be generated by a voltage regression model using the predicted real-time on-chip temperature as an input, wherein the voltage regression model may be a multivariate regression model based on multiple factors, wherein the multiple factors may include the predicted real-time on-chip temperature, average dwell time, data retention time, and number of P / E cycles.
[0056] In one embodiment, the estimated threshold voltage generated by the voltage regression model may decrease as the data retention time increases.
[0057] In one embodiment, the plurality of factors may include a programming temperature, and the estimated threshold voltage generated by the voltage regression model may increase as the programming temperature increases.
[0058] Any disclosed methods and operations may be implemented as computer-executable instructions (e.g., software code of the operations described herein) stored on one or more computer-readable storage media (e.g., non-transitory computer-readable media, such as one or more optical disk media, volatile storage components (e.g., DRAM or SRAM), or non-volatile storage components (e.g., hard disk drives) and executed on a device controller (e.g., firmware executed by an ASIC). Any computer-executable instructions for implementing the disclosed techniques and any data created and used during implementation of the disclosed embodiments may be stored on one or more computer-readable media (e.g., non-transitory computer-readable media).
[0059] Although various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art. The various aspects and embodiments disclosed herein are for purposes of illustration and are not intended to be limiting, with the true scope and spirit being indicated by the appended claims.
Claims
1. A method comprising: obtaining a readout from a temperature sensor in a nonvolatile storage system; generating a predicted real-time on-chip temperature for a non-volatile storage device in the storage system based on the reading; generating an estimated threshold voltage for reading data stored in the nonvolatile memory device based on the predicted real-time on-chip temperature; and The estimated threshold voltage is used as a starting point to perform a partial scan of the reference voltage to obtain a final read reference voltage with a minimum read bit error rate.
2. The method according to claim 1, characterized in that The temperature sensor is located on a storage controller of the nonvolatile storage system, a printed circuit board of the nonvolatile storage system, or the nonvolatile storage device.
3. The method according to claim 1, characterized in that The predicted real-time on-chip temperature is generated by a temperature regression model using the readout as input.
4. The method according to claim 3, characterized in that The temperature regression model is a multivariate regression model based on multiple factors, including the readout, the physical distance between the temperature sensor and the non-volatile storage device, the thermal properties of the thermally conductive material in the non-volatile storage system, and the access frequency and read / write access intensity of the non-volatile storage device.
5. The method according to claim 1, characterized in that The estimated threshold voltage is generated by a voltage regression model using the predicted real-time on-chip temperature as an input.
6. The method according to claim 5, characterized in that The voltage regression model is a multivariate regression model based on multiple factors, including the predicted real-time on-chip temperature, average dwell time, data retention time, and number of P / E cycles.
7. The method according to claim 6, characterized in that The estimated threshold voltage generated by the voltage regression model decreases as the data retention time increases.
8. The method according to claim 6, characterized in that The plurality of factors include a programming temperature, and the estimated threshold voltage generated by the voltage regression model increases as the programming temperature increases.
9. A non-volatile storage system, characterized in that: include: Non-volatile storage devices; Temperature sensor; and A processor configured to execute the method according to any one of claims 1 to 8.
10. A non-transitory machine-readable medium having computer instructions, characterized in that: When the computer instructions are executed by a hardware processor, the hardware processor is caused to perform the method according to any one of claims 1 to 8.
