Solid state disk junction temperature closed loop control method and solid state disk

Abstract

The present application belongs to the technical field of computer storage device thermal management, and discloses a solid state disk junction temperature closed loop control method and a solid state disk; current load characteristic data of the solid state disk is acquired, read-write command queue depth, LDPC error correction trigger type and NAND target channel identification are identified, instantaneous power consumption of a master control core logic module and a NAND storage array is calculated based on a preset heat source power consumption weight matrix, each module power consumption is exponentially weighted integrated in a lead time dimension, and a transient prediction temperature rise increment is obtained. The increment is spatiotemporally aligned and superimposed with basic measurement temperature collected by a digital temperature sensor, and a target prediction junction temperature is obtained. A dynamic junction temperature safety threshold based on a steady-state basic temperature and intervention frequency statistics is adopted, when the prediction junction temperature exceeds the threshold, a hierarchical micro-process parameter intervention instruction is generated according to the deviation degree. The present application realizes advanced prediction and accurate control of the solid state disk junction temperature, and effectively improves system reliability and environmental adaptability.

Description

Technical Field

[0001] This invention relates to the field of thermal management technology for computer storage devices, and more specifically, to a closed-loop control method for junction temperature of a solid-state drive (SSD) and a SSD. Background Technology

[0002] Existing solid-state drive (SSD) temperature control technologies face challenges related to sensor placement and thermal transient latency. Specifically, digital temperature sensors inside SSDs are typically integrated into the edge region of the main control chip or the substrate surface. Due to limitations in chip design and manufacturing processes, sensors often cannot be directly placed in the core heat-generating areas, such as the computing units of LDPC codecs, word line drive circuits of NAND memory arrays, or high-frequency switching transistors of charge pumps. This spatial misalignment results in sensors monitoring "second-hand" or "third-hand" thermal information, rather than the direct temperature state of the heat source. In real-world scenarios such as high-concurrency access in data centers, burst writes to mobile devices, or continuous loads in industrial applications, when the system experiences transient high-load shocks, the temperature of the core heat-generating modules can rise rapidly within milliseconds to seconds. The heat conduction from the heat source to the remote temperature sensor requires passing through multiple layers of silicon-based materials, encapsulation materials, and heat dissipation paths, and this physical conduction process inevitably involves a thermal time constant delay. Therefore, when the sensor detects that the temperature has reached the preset threshold and triggers the temperature control strategy, the actual junction temperature inside the chip has often far exceeded the safety boundary, which may cause soft errors in the logic circuit, data flipping in the storage cell, or even irreversible thermal damage, thus limiting the reliability and performance of solid-state drives under extreme conditions.

[0003] In view of this, the present invention proposes a closed-loop control method for junction temperature of solid-state drives and a solid-state drive to solve the above problems. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a closed-loop control method for junction temperature of a solid-state drive, comprising: The current input and output load characteristic data of the solid-state drive is obtained. Heat source feature mapping and advance integration are performed on the input and output load characteristic data to obtain the transient predicted temperature rise increment of the heat-generating area inside the main control chip and the NAND array at the current moment. The base measurement temperature collected by the digital temperature sensor is obtained, and the transient predicted temperature rise increment is spatiotemporally aligned and superimposed with the base measurement temperature to obtain the target predicted junction temperature at the current moment. If the target predicted junction temperature exceeds the preset dynamic junction temperature safety threshold, then based on the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold, a micro-process parameter intervention instruction for the heat source is generated. The micro-process parameter intervention instruction is used to limit the iteration limit of the LDPC codec and the NAND channel concurrency in the main control chip. After executing the micro-process parameter intervention command, the actual temperature rise slope of the digital temperature sensor is continuously extracted. If the actual temperature rise slope exceeds the preset expected temperature drop slope, a global performance degradation command is generated based on the difference between the actual temperature rise slope and the expected temperature drop slope. The micro-process parameter intervention instructions and global performance degradation instructions are combined and output as a closed-loop control strategy for solid-state drive junction temperature.

[0005] A solid-state drive (SSD) is used to implement the above-mentioned SSD junction temperature closed-loop control method. The SSD includes a main controller chip, a NAND flash memory array connected to the main controller chip, and a digital temperature sensor connected to the main controller chip. A digital temperature sensor is used to collect basic measured temperatures and send them to the main control chip; The main control chip is used to acquire the current input and output load characteristic data of the solid-state drive, perform heat source feature mapping and advance integration on the input and output load characteristic data, and obtain the transient predicted temperature rise increment of the heat-generating area inside the main control chip and the NAND flash memory array at the current moment. The main control chip is also used to acquire the basic measured temperature, and to spatiotemporally align and superimpose the transient predicted temperature rise increment with the basic measured temperature to obtain the target predicted junction temperature at the current moment. The main control chip is also used to generate micro-process parameter intervention instructions for the heat source based on the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold if the target predicted junction temperature exceeds the preset dynamic junction temperature safety threshold. It also limits the iteration limit of the LDPC codec inside the main control chip and the NAND channel concurrency of the NAND flash array according to the micro-process parameter intervention instructions. The main control chip is also used to continuously extract the actual temperature rise slope of the digital temperature sensor after executing the micro-process parameter intervention command. If the actual temperature rise slope exceeds the preset expected temperature drop slope, a global performance degradation command is generated based on the difference between the actual temperature rise slope and the expected temperature drop slope. The main control chip is also used to combine micro-process parameter intervention instructions and global performance degradation instructions to output a closed-loop control strategy for solid-state drive junction temperature.

[0006] This invention discloses a closed-loop junction temperature control method for solid-state drives (SSDs) and the technical effects and advantages of SSDs: This invention eliminates the sensing blind spot caused by heat conduction delay, enabling it to detect temperature rise trends and intervene before a thermal hazard actually develops, thus avoiding the risk of potential damage to critical internal circuits under transient high temperatures. This invention ensures thermal safety boundaries while maintaining high-performance output, effectively resolving the contradiction between performance and reliability in traditional solutions. Furthermore, this invention establishes an adaptive adjustment mechanism based on actual operating conditions, automatically optimizing the control strategy according to dynamic changes in the environment, load characteristics, and heat dissipation conditions, avoiding the insufficient adaptability of fixed-parameter control schemes in complex application scenarios. By constructing a refined hierarchical intervention system and a closed-loop feedback verification mechanism, this invention achieves targeted adjustment of different heat sources and real-time verification of control effects, ensuring that every temperature control intervention achieves the expected results and avoiding the risk of thermal runaway due to control measure failure. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of a solid-state drive junction temperature closed-loop control method according to the present invention; Figure 2 The logic diagram for determining the dynamic junction temperature safety threshold of this invention; Figure 3 This is a logic diagram for generating microscopic process parameter intervention instructions according to the present invention; Figure 4 This is a schematic diagram of a solid-state drive according to the present invention. Detailed Implementation

[0008] 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.

[0009] This application provides a closed-loop control method for junction temperature of a solid-state drive (SSD) and an SSD itself. Please refer to [link / reference]. Figure 1 In this embodiment of the invention, a closed-loop control method for solid-state drive junction temperature includes the following steps: Step 1: Obtain the current input / output load characteristic data of the solid-state drive, perform heat source feature mapping and advance integration on the input / output load characteristic data, and obtain the transient predicted temperature rise increment of the heat-generating area inside the main control chip and the NAND array at the current moment. Step 2: Obtain the base measurement temperature collected by the digital temperature sensor, and then perform spatiotemporal alignment and superposition of the transient predicted temperature rise increment with the base measurement temperature to obtain the target predicted junction temperature at the current moment. Step 3: If the target predicted junction temperature exceeds the preset dynamic junction temperature safety threshold, then based on the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold, a micro-process parameter intervention instruction for the heat source is generated. The micro-process parameter intervention instruction is used to limit the iteration limit of the LDPC codec and the NAND channel concurrency in the main control chip. Step 4: After executing the micro-process parameter intervention command, continuously extract the actual temperature rise slope of the digital temperature sensor. If the actual temperature rise slope exceeds the preset expected temperature drop slope, then generate a global performance degradation command based on the difference between the actual temperature rise slope and the expected temperature drop slope. Step 5: Combine the micro-process parameter intervention instructions and the global performance degradation instructions to output a closed-loop control strategy for the junction temperature of the solid-state drive.

[0010] In this embodiment of the invention, the present invention acquires solid-state drive (SSD) input / output load characteristic data, performs heat source feature mapping and advance integral prediction of transient temperature rise increment, combines the basic temperature measurement from a digital temperature sensor for spatiotemporal alignment and superposition to obtain the target predicted junction temperature, generates graded micro-process parameter intervention instructions based on the deviation between the predicted junction temperature and the dynamic safety threshold, and generates global performance degradation instructions by monitoring the actual temperature rise slope and combining them to output a closed-loop control strategy. This effectively overcomes the technical problems of existing SSD temperature control schemes, such as lagging temperature monitoring, inability to predict transient temperature rise leading to untimely thermal protection response, lack of dynamic adaptability of static temperature thresholds, and overly coarse temperature control strategies that cannot finely adjust process parameters. Thus, it achieves the technical effects of advance prediction and precise control of SSD junction temperature, balancing performance and heat dissipation requirements through graded intervention strategies, and adaptive optimization of the temperature control strategy based on a closed-loop feedback mechanism, ultimately ensuring the temperature safety and performance stability of the SSD under high load conditions.

[0011] In a preferred embodiment of the present invention, the above-mentioned process of performing heat source feature mapping and advance integration on the input and output load characteristic data to obtain the transient predicted temperature rise increment of the internal heat-generating area of ​​the main control chip and the NAND array at the current moment includes: Step 2.1 identifies the read / write command queue depth, LDPC erasure trigger type, and NAND target channel identifier in the input / output load characteristic data. Specifically, during SSD operation, the controller chip monitors the input / output command stream received from the host in real time and extracts key load characteristic parameters from the command queue. Read / write command queue depth monitoring includes the number of currently pending read and write commands. Queue depth directly reflects the computational load intensity of the controller chip; for example, a queue depth of 32 commands indicates a high load, while a queue depth of less than 8 commands indicates a light load. LDPC erasure trigger type identification includes different types such as soft-decision error correction, hard-decision error correction, and retry error correction. Different error correction types have significantly different computational complexity and heat generation; soft-decision error correction typically consumes 3-5 times more power than hard-decision error correction. NAND target channel identifier extraction includes the specific NAND channel number currently accessed and the number of concurrently accessed channels. Multi-channel concurrent access significantly increases the power density of the NAND array. By identifying these load characteristic parameters in real time, accurate input data is provided for subsequent heat source power consumption calculations.

[0012] Step 2.2: Based on the preset heat source power consumption weight matrix, calculate the current instantaneous power consumption of the main control core logic module and the NAND storage array. Specifically, this includes: pre-calibrating the heat source power consumption weight matrix based on the hardware architecture characteristics of the solid-state drive and a large amount of test data. This matrix describes the contribution weight of different load characteristic parameters to the power consumption of each heat-generating module. The power consumption weight matrix of the main control core logic module includes power allocation coefficients for sub-modules such as the read / write processor, LDPC codec, and interface controller. For example, the weight coefficient of LDPC soft-decision error correction for the codec power consumption is 0.85, and the base power consumption weight of the read / write processor at high queue depth is 0.6. The power consumption weight matrix of the NAND storage array includes power allocation for different channels and different operation types. For example, the base power consumption weight for a single-channel write operation is 0.4, and the power consumption superposition coefficient for multi-channel concurrency is 1.2. During runtime, the identified load characteristic parameters are multiplied with the corresponding power consumption weight matrix to obtain the instantaneous power consumption value of the main control core logic module and the NAND storage array under the current load state, providing basic power consumption data for heat source temperature rise prediction.

[0013] Step 2.3: Using the pre-calibrated thermal conduction delay time constants from each module to the digital temperature sensor, perform an exponentially weighted integral of the current instantaneous power consumption of each module in the lead time dimension. Convert the integral result into the transient predicted temperature rise increment of the heat-generating area inside the main control chip and the NAND array at the current moment. Specifically, this includes calibrating the thermal conduction delay time constants from each heat-generating module to the digital temperature sensor through thermal simulation analysis and actual testing. The thermal conduction time constant from the heat-generating area inside the main control chip to the temperature sensor is typically 2-5 seconds, and the thermal conduction time constant from the NAND array to the temperature sensor is typically 3-8 seconds. The difference in time constants is mainly determined by the length of the thermal conduction path and the heat dissipation structure. Based on the thermal conduction delay time constants, perform an exponentially weighted integral of the current instantaneous power consumption of each module in the lead time dimension. The integral formula is: Temperature rise increment (t) = ∫ Power consumption (τ) × exp(-(t-τ) / Time constant) dτ. This formula considers the leading effect of power consumption changes on the temperature at future moments. The integral calculation covers historical power consumption data from the past heat conduction cycle, and uses exponentially decaying weights to reflect the influence of time distance on temperature rise. Finally, the integral result is converted into temperature units through a heat-to-electric conversion model, outputting the transient predicted temperature rise increment of the heat-generating area inside the main control chip and the NAND array at the current moment. This increment reflects the prediction of future temperature changes based on the current and historical load conditions.

[0014] In this embodiment of the invention, by employing the technical means of identifying the read / write command queue depth, LDPC erasure trigger type, and NAND target channel identifier in the input / output load characteristic data, calculating the current instantaneous power consumption of the main control core logic module and the NAND storage array based on a preset heat source power consumption weight matrix, and using a pre-calibrated heat conduction delay time constant to perform an advanced exponential weighted integral on the power consumption of each module and convert it into a transient predicted temperature rise increment, the invention overcomes the technical problems of existing solid-state drive temperature monitoring systems that rely solely on real-time temperature measurement, cannot predict the advanced impact of load changes on temperature leading to lag in temperature control response, and lack accurate heat source power consumption modeling and heat conduction delay compensation mechanisms leading to insufficient temperature prediction accuracy. This achieves the technical effect of advanced temperature rise prediction based on load characteristics, accurate quantification of the impact of different loads on each heat-generating module through a power consumption weight matrix, and improved prediction timeliness through heat conduction delay compensation, providing accurate and forward-looking temperature information for solid-state drive junction temperature control.

[0015] Please see Figure 2 The logic diagram for determining the dynamic junction temperature safety threshold is shown below. In a preferred embodiment of the present invention, the dynamic junction temperature safety threshold is determined in the following manner: Step 3.1: Obtain the steady-state baseline temperature of the digital temperature sensor in the idle state. Specifically, after the SSD system initializes or has been idle for a long time, when the input / output load characteristic data shows that the read / write command queue depth is zero, the LDPC erasure trigger count is lower than a preset minimum threshold, and the NAND channel access frequency is at its lowest level, the SSD is determined to have entered the idle state. In the idle state, the digital temperature sensor continuously collects temperature data with a sampling period of 100 milliseconds and a continuous monitoring time of no less than 30 seconds to ensure temperature stability. The collected temperature sequence is processed using a moving average algorithm. When the temperature change of 10 consecutive sampling points is less than 0.5 degrees Celsius, the temperature is determined to have reached a steady state. The arithmetic mean of the temperature measurements during the steady-state period is taken as the steady-state baseline temperature. This temperature reflects the environmental heat dissipation equilibrium temperature of the SSD under no-load conditions, providing a benchmark reference for subsequent dynamic threshold calculations.

[0016] Step 3.2 involves querying a pre-defined temperature-threshold offset mapping table based on the steady-state base temperature to obtain the initial junction temperature safety threshold. Specifically, this includes pre-establishing a temperature-threshold offset mapping table based on the SSD's thermal design power (TDP), chip junction temperature specifications, and heat dissipation capabilities under different ambient temperatures. This mapping table takes the steady-state base temperature as input and outputs the corresponding threshold offset. For example, when the steady-state base temperature is 25 degrees Celsius, the threshold offset is 45 degrees Celsius; when the steady-state base temperature is 35 degrees Celsius, the threshold offset is 40 degrees Celsius, reflecting the design principle that the safety margin decreases as the ambient temperature increases. The mapping table is established based on the highest operating temperature specifications of the controller chip and NAND flash memory chips. Typically, the highest junction temperature of the controller chip is 85-95 degrees Celsius, and the highest operating temperature of the NAND flash memory is 70-85 degrees Celsius. A safety margin of 10-15 degrees Celsius is set based on these specifications. During runtime, the acquired steady-state base temperature is used as the query key value. The corresponding threshold offset is found in the temperature-threshold offset mapping table. The steady-state base temperature and the threshold offset are added together to obtain the initial junction temperature safety threshold, which serves as the initial judgment benchmark for temperature control.

[0017] Step 3.3: Count the frequency of micro-process parameter intervention commands triggered within a preset sliding time window. If a certain number of pre-defined frequencies exceed the preset frequency upper limit, the initial junction temperature safety threshold is adjusted downwards by a preset safety step size based on the extent to which the intervention frequency exceeds the preset frequency upper limit. The adjusted initial junction temperature safety threshold is then used as the dynamic junction temperature safety threshold. Specifically, this includes setting the preset sliding time window to 10 minutes and counting the cumulative number of micro-process parameter intervention commands triggered within this time window in real time. The preset frequency upper limit is determined based on the normal operating mode of the solid-state drive. For example, setting it to 5 times means that if more than 5 intervention commands are triggered within 10 minutes, the current heat dissipation conditions are considered to have deteriorated or the load is too heavy. When the counted intervention frequency exceeds the preset frequency upper limit, the extent to which the intervention frequency exceeds the preset frequency upper limit is calculated. For example, if there are 8 actual interventions, the excess is 3 times. The threshold correction amount is determined based on the excess amount: Correction amount = Excess amount × Preset safety step size, where the preset safety step size is set to 2 degrees Celsius, meaning that the safety threshold is adjusted downwards by 2 degrees Celsius for each intervention frequency exceeding the limit. The corrected junction temperature safety threshold is obtained by subtracting the correction amount from the initial junction temperature safety threshold. For example, if the initial threshold is 70 degrees Celsius and the correction amount is 6 degrees Celsius, then the corrected dynamic junction temperature safety threshold is 64 degrees Celsius. The correction mechanism reflects an adaptive adjustment strategy based on the system's operating state. When temperature control intervention is frequently triggered, the safety threshold is proactively lowered to improve the sensitivity of heat dissipation protection.

[0018] In this embodiment of the invention, by employing the technical means of acquiring the steady-state base temperature of a digital temperature sensor in an idle state, querying a preset temperature-threshold offset mapping table based on the steady-state base temperature to obtain the initial junction temperature safety threshold, and statistically analyzing the intervention frequency within a sliding time window and adjusting the initial threshold downwards according to the exceedance situation to obtain a dynamic junction temperature safety threshold, this invention overcomes the technical problems of existing solid-state drive temperature control systems that use fixed safety thresholds, cannot adaptively adjust according to changes in ambient temperature and system operating status leading to overly conservative or aggressive temperature control strategies, and lack a threshold optimization mechanism based on historical intervention frequencies resulting in poor temperature control performance. This achieves the technical effect of benchmark adaptation of the safety threshold based on ambient temperature, dynamic optimization adjustment of the threshold through intervention frequency statistics, effectively balancing system performance and thermal safety protection requirements, and providing a scientific and reasonable dynamic judgment standard for solid-state drive junction temperature control.

[0019] Please see Figure 3 This is a logic diagram for generating micro-process parameter intervention instructions. In a preferred embodiment of the present invention, the above-mentioned generation of micro-process parameter intervention instructions for the heat source based on the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold includes: Step 4.1: If the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold is within the first deviation range, a first-level intervention instruction is generated. This first-level intervention instruction reduces the drive current of the charge pump inside the main controller to a preset safety lower limit and inserts microsecond-level idle intervals between NAND write operations. Specifically, it includes: calculating the temperature deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold; when the deviation is within the first deviation range (1-5 degrees Celsius), it is determined to be a slight overheating state, and the first-level intervention instruction generation process is initiated. The first-level intervention instruction finely adjusts the charge pump circuit inside the main controller chip. The charge pump is a key circuit for NAND flash memory programming and erasing operations, and its drive current directly affects power consumption and heat generation. By reducing the charge pump's drive current from the normal operating current to the preset safety lower limit through firmware instructions, for example, from 100mA to 80mA, a 20% reduction in current can achieve a power consumption reduction of approximately 35%. Meanwhile, microsecond-level idle intervals are inserted into the timing control of NAND write operations. Specifically, this is achieved by delaying the execution of the next operation by 10-50 microseconds after each write operation cycle is completed. These idle intervals allow the NAND array to dissipate heat, effectively reducing heat accumulation caused by continuous operations. The execution of the first-level intervention command has a minimal impact on the user's perceived performance, typically with a performance degradation of less than 5%.

[0020] Step 4.2: If the deviation is in the second deviation range, which is greater than the first deviation range, a second-level intervention instruction is generated based on the first-level intervention instruction. The second-level intervention instruction limits the maximum number of iterations for the current LDPC soft-decision decoding to a preset lower limit and activates the NAND multi-channel alternating sleep mechanism. Specifically, when the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold is within the second deviation range (5-10 degrees Celsius), it is determined to be a moderate over-temperature state. In addition to continuing to execute the first-level intervention instruction, an additional second-level intervention instruction is generated to strengthen temperature control. The second-level intervention instruction first restricts the LDPC soft-decision decoding algorithm. LDPC decoding is one of the most power-consuming computational modules of the main control chip. The number of iterations for soft-decision decoding is usually set to 8-16 times. By limiting the maximum number of iterations to a preset lower limit, such as 4 times, the computational complexity of decoding and the corresponding power consumption and heat generation can be significantly reduced. Although reducing the number of iterations may slightly decrease the error correction capability, data reliability can still be guaranteed in most scenarios. The second-level intervention command also activates the NAND multi-channel alternating sleep mechanism. This mechanism changes the original concurrent access of multiple NAND channels to an alternating access mode. For example, it changes the concurrent access of 8 channels to the alternating access of 4 channels. During the operation of each channel group, other channel groups are in a low-power sleep state. By using time-division multiplexing, the number of channels working at the same time is reduced, thereby reducing the instantaneous power density and heat concentration of the NAND array.

[0021] In this embodiment of the invention, because the overheating level is determined by the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold, a first-level intervention instruction is generated in the first deviation interval to reduce the charge pump drive current and insert the NAND write operation idle interval, and a second-level intervention instruction is generated in the second deviation interval to limit the number of LDPC soft decision decoding iterations and start NAND multi-channel alternating sleep. Therefore, this invention overcomes the technical problems of existing solid-state drive temperature control systems, which use simple frequency reduction or performance limitation, lack fine adjustment for specific heat sources, resulting in coarse temperature control effects, and lack of graded strategies for temperature control measures, making it impossible to flexibly adjust according to the overheating level. Thus, it achieves graded and precise temperature control based on the degree of temperature deviation, achieves efficient cooling by adjusting the micro-process parameters of key heat sources such as charge pumps, LDPC decoding, and NAND channels, and minimizes the impact on user performance experience, providing a fine thermal management strategy for solid-state drives.

[0022] In a preferred embodiment of the present invention, after executing the micro-process parameter intervention command, the actual temperature rise slope of the digital temperature sensor is continuously extracted. If the actual temperature rise slope exceeds the preset expected temperature drop slope, a global performance degradation command is generated based on the difference between the actual temperature rise slope and the expected temperature drop slope, including: Step 5.1: Within the preset feedback observation window after executing the micro-process parameter intervention command, the temperature sequence from the digital temperature sensor is acquired at a microsecond-level cycle. Specifically, this includes: immediately initiating the temperature feedback monitoring process after the micro-process parameter intervention command is executed, setting the preset feedback observation window to 30 seconds. This window length is determined based on the thermal conductivity characteristics of the solid-state drive, sufficient to observe the temperature response effect of the intervention measures. Within the observation window, high-frequency sampling is performed on the digital temperature sensor at a microsecond-level cycle, with a sampling period set to 50 milliseconds to ensure accurate capture of subtle temperature changes. A high-precision digital temperature sensor with a temperature resolution of 0.1 degrees Celsius is used for temperature acquisition, and the sampled data is filtered by hardware to eliminate high-frequency noise interference. The acquired temperature data is stored chronologically to form a temperature sequence, with each data point containing a timestamp and the corresponding temperature value, providing an accurate data foundation for subsequent slope calculations. The temperature sequence storage employs a circular buffer mechanism to ensure the continuity of data acquisition and the effective utilization of system resources.

[0023] Step 5.2 involves calculating the actual temperature rise slope based on temperature sequence fitting, and simultaneously calculating the expected temperature drop slope based on the anticipated power consumption reduction after intervention. Specifically, this includes: linearly fitting the collected temperature sequence using the least squares method to calculate the actual temperature rise slope. The fitting algorithm considers the weight of all sampling points within the time window, and obtains the slope of the temperature-time relationship through linear regression analysis. A positive slope indicates a temperature rise, and a negative slope indicates a temperature fall; the absolute value of the slope reflects the rate of temperature change. To improve fitting accuracy, a sliding window approach is used for piecewise fitting, with each sliding window covering 10 seconds of data. A stable actual temperature rise slope is obtained by averaging the slopes of multiple windows. Simultaneously, the expected power consumption reduction is calculated based on the executed micro-process parameter intervention commands. For example, the first-level intervention command is expected to reduce power consumption by 15%, and the second-level intervention command is an additional 25% reduction on top of the first level. The expected temperature drop rate, i.e., the expected temperature drop slope, is calculated based on the power consumption reduction and the heat conduction model. This slope reflects the cooling effect that the intervention measures should achieve under ideal heat dissipation conditions.

[0024] Step 5.3: Calculate the slope deviation between the actual temperature rise slope and the expected temperature drop slope. If the slope deviation exceeds a preset slope deviation tolerance threshold, the current heat dissipation environment is deemed degraded. Based on the slope deviation, a global performance degradation instruction is generated to forcibly suspend the non-core I / O queue. Specifically, this includes: calculating the slope deviation between the actual temperature rise slope and the expected temperature drop slope through numerical comparison. Deviation = Actual temperature rise slope - Expected temperature drop slope. A positive value indicates that the actual temperature rise exceeds expectations, and a negative value indicates that the actual temperature drop meets or exceeds expectations. The preset slope deviation tolerance threshold is set to 0.2 degrees Celsius / second. When the deviation exceeds this threshold, it indicates that the micro-process parameter intervention measures have failed to achieve the expected cooling effect, and the current heat dissipation environment is deemed degraded, possibly due to factors such as increased ambient temperature, decreased heat sink performance, or internal dust accumulation. Based on the magnitude of the slope deviation, a global performance degradation instruction of corresponding strength is generated. When the deviation is 0.2-0.5 degrees Celsius / second, a mild degradation instruction is generated to suspend non-critical background tasks such as garbage collection and wear leveling. When the deviation exceeds 0.5 degrees Celsius / second, a severe degradation instruction is generated to forcibly suspend non-core I / O queues and retain only necessary read and write operations, thereby achieving rapid cooling by significantly reducing the load.

[0025] In this embodiment of the invention, the technical means of acquiring the temperature sequence of a digital temperature sensor at a microsecond-level cycle within the feedback observation window after the execution of the micro-process parameter intervention command, calculating the actual temperature rise slope based on the temperature sequence fitting, and calculating the expected temperature drop slope based on the expected power consumption reduction after intervention, and determining the heat dissipation environment status by comparing the deviation of the two slopes and generating corresponding global performance degradation commands, overcomes the technical problems of existing solid-state drive temperature control systems, such as lack of intervention effect feedback verification mechanism, inability to determine whether temperature control measures are effective leading to missed opportunities for further cooling, and lack of dynamic performance adjustment strategy based on actual heat dissipation capacity leading to inflexible temperature control response. This achieves the technical effect of verifying the intervention effect based on actual temperature feedback, timely detecting heat dissipation environment degradation through slope deviation analysis and initiating corresponding global performance adjustments, ensuring effective temperature control under various heat dissipation conditions, and providing an adaptive temperature control protection mechanism for solid-state drives.

[0026] In a preferred embodiment of the present invention, after the above-mentioned combination and output of micro-process parameter intervention instructions and global performance degradation instructions into a solid-state drive junction temperature closed-loop control strategy, it further includes: Step 6.1: After triggering the global performance degradation command, continuously monitor the temperature changes of the digital temperature sensor, identify the moment the temperature reaches its peak, and record the measured peak temperature. Specifically, after the global performance degradation command is executed, the load on the solid-state drive decreases significantly, and the system temperature should show a trend of first rising briefly and then continuously decreasing. The brief rise is due to the heat accumulation and release caused by the temperature control response delay. Continuously monitor the output of the digital temperature sensor and use a sliding window peak detection algorithm to identify the temperature peak moment. Specifically, when the temperature values ​​of 5 consecutive sampling points are lower than the current detection point, the current detection point is determined to be the peak point. Set the peak identification time window to within 60 seconds after the execution of the global performance degradation command. This time window covers the complete process from load reduction to temperature stabilization. Record the identified peak moment and the corresponding measured peak temperature value. This temperature value reflects the actual highest temperature that the system can reach under the current heat dissipation conditions and temperature control strategy.

[0027] Step 6.2: Extract the target predicted junction temperature at the moment the global performance degradation instruction is generated, and calculate the prediction residual between the measured peak temperature and the target predicted junction temperature. Specifically, this includes extracting the target predicted junction temperature value corresponding to the moment the global performance degradation instruction is generated from the system log. This value is a temperature prediction calculated based on the load characteristic data, heat source characteristic mapping, and lead integration at that time. The prediction residual is calculated as: Measured Peak Temperature - Target Predicted Junction Temperature. A positive residual indicates that the actual temperature is higher than the predicted temperature, suggesting that the temperature rise prediction model is too optimistic; a negative residual indicates that the actual temperature is lower than the predicted temperature, suggesting that the temperature rise prediction model is too conservative. The magnitude of the prediction residual reflects the accuracy of the temperature prediction model; a smaller residual indicates a more accurate prediction, while a larger residual indicates that the prediction model needs parameter correction.

[0028] Step 6.3: Statistically analyze the prediction residuals across multiple consecutive temperature control events. If the prediction residuals consistently exceed a preset residual threshold, initiate the heat conduction delay time constant correction process. This includes: establishing a statistical mechanism for prediction residuals, recording the prediction residual values ​​for the most recent 10 temperature control events, and monitoring the long-term trend of prediction accuracy using a sliding statistical method. A preset residual threshold of 3 degrees Celsius is set. When the absolute value of the prediction residuals for 5 consecutive temperature control events exceeds this threshold, it is determined that the prediction residuals consistently exceed the threshold, indicating that the current heat conduction delay time constant is no longer suitable for the actual system state and model parameter correction is required. Reasons for consistently exceeding the residual threshold may include: changes in radiator performance, long-term changes in ambient temperature, and changes in heat conduction characteristics due to system aging. Adaptive parameter correction can maintain the long-term accuracy of the temperature prediction model.

[0029] Step 6.4: Determine the time constant correction based on the direction and magnitude of the predicted residuals, perform a reverse decay correction on the heat conduction delay time constant, and update the corrected heat conduction delay time constant in the subsequent exponentially weighted integral calculation. Specifically, this includes: calculating the time constant correction based on the statistical characteristics of the predicted residuals, using a reverse decay mechanism. When the predicted residual is positive (actual temperature is higher than predicted), it indicates that the temperature rise rate is faster than expected, requiring a reduction in the heat conduction delay time constant. The correction formula is: New time constant = Original time constant × (1 - Correction coefficient × Residual magnitude / Threshold), where the correction coefficient is set to 0.1. When the predicted residual is negative (actual temperature is lower than predicted), it indicates that the temperature rise rate is slower than expected, requiring an increase in the heat conduction delay time constant. The correction formula is: New time constant = Original time constant × (1 + Correction coefficient × Residual magnitude / Threshold). The corrected heat conduction delay time constant is written into the subsequent exponentially weighted integral calculation module through a parameter update mechanism, ensuring that the temperature prediction model can adaptively track changes in the system's thermal characteristics and continuously maintain prediction accuracy.

[0030] In this embodiment of the invention, by employing a technique of continuously monitoring a digital temperature sensor to identify the measured peak temperature after triggering a global performance degradation instruction, calculating the prediction residual between the measured peak temperature and the target predicted junction temperature when the global performance degradation instruction is generated, statistically analyzing the prediction residuals of multiple consecutive temperature control events, and performing reverse decay correction on the heat conduction delay time constant based on the direction and magnitude of the residuals when the temperature continuously exceeds a threshold, the technical problems of existing solid-state drive temperature prediction models—such as using fixed parameters, being unable to adaptively track changes in system thermal characteristics leading to a decline in long-term prediction accuracy, and lacking a model feedback correction mechanism based on actual temperature control effects leading to the accumulation of prediction deviations—are overcome. This achieves the technical effect of self-learning optimization of the temperature prediction model based on actual temperature control effects, maintaining long-term prediction accuracy through dynamic correction of heat conduction parameters, ensuring the prediction reliability and control effectiveness of the solid-state drive junction temperature closed-loop control system during long-term operation, and providing continuous and accurate prediction support for temperature control strategies.

[0031] The above describes a closed-loop control method for junction temperature of a solid-state drive (SSD) according to an embodiment of this application. The following describes an SSD according to an embodiment of this application. Please refer to [link / reference]. Figure 2 One embodiment of a solid-state drive in this application includes: The main control chip, the NAND flash memory array connected to the main control chip, and the digital temperature sensor connected to the main control chip; A digital temperature sensor is used to collect basic measured temperatures and send them to the main control chip; The main control chip is used to acquire the current input and output load characteristic data of the solid-state drive, perform heat source feature mapping and advance integration on the input and output load characteristic data, and obtain the transient predicted temperature rise increment of the heat-generating area inside the main control chip and the NAND flash memory array at the current moment. The main control chip is also used to acquire the basic measured temperature, and to spatiotemporally align and superimpose the transient predicted temperature rise increment with the basic measured temperature to obtain the target predicted junction temperature at the current moment. The main control chip is also used to generate micro-process parameter intervention instructions for the heat source based on the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold if the target predicted junction temperature exceeds the preset dynamic junction temperature safety threshold. It also limits the iteration limit of the LDPC codec inside the main control chip and the NAND channel concurrency of the NAND flash array according to the micro-process parameter intervention instructions. The main control chip is also used to continuously extract the actual temperature rise slope of the digital temperature sensor after executing the micro-process parameter intervention command. If the actual temperature rise slope exceeds the preset expected temperature drop slope, a global performance degradation command is generated based on the difference between the actual temperature rise slope and the expected temperature drop slope. The main control chip is also used to combine micro-process parameter intervention instructions and global performance degradation instructions to output a closed-loop control strategy for solid-state drive junction temperature.

[0032] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0033] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0034] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A closed-loop control method for junction temperature of a solid-state drive, characterized in that, include: The current input / output load characteristic data of the solid-state drive is obtained, and the heat source characteristic data is mapped and integrated in advance to obtain the transient predicted temperature rise increment of the heat-generating area inside the main control chip and the NAND array at the current moment. The base measurement temperature collected by the digital temperature sensor is obtained, and the transient predicted temperature rise increment is spatiotemporally aligned and superimposed with the base measurement temperature to obtain the target predicted junction temperature at the current moment. If the target predicted junction temperature exceeds the preset dynamic junction temperature safety threshold, then based on the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold, an intervention instruction for the microscopic process parameters of the heat source is generated. After executing the micro-process parameter intervention command, the actual temperature rise slope of the digital temperature sensor is continuously extracted. If the actual temperature rise slope exceeds the preset expected temperature drop slope, a global performance degradation command is generated based on the difference between the actual temperature rise slope and the expected temperature drop slope. The micro-process parameter intervention command and the global performance degradation command are combined and output as a solid-state drive junction temperature closed-loop control strategy.

2. The method according to claim 1, characterized in that, The step of performing heat source feature mapping and advance integration on the input / output load characteristic data to obtain the transient predicted temperature rise increment of the internal heat-generating area of ​​the main control chip and the NAND array at the current moment includes: Identify the read / write command queue depth, LDPC erasure trigger type, and NAND target channel identifier in the input / output load characteristic data; Based on the preset heat source power consumption weight matrix, calculate the current instantaneous power consumption of the main control core logic module and the NAND storage array; Using the pre-calibrated thermal conduction delay time constant of each module to the digital temperature sensor, the current instantaneous power consumption of each module is subjected to an exponentially weighted integral in the lead time dimension, and the integral result is converted into the transient predicted temperature rise increment of the heat-generating area inside the main control chip and the NAND array at the current moment.

3. The method according to claim 1, characterized in that, The dynamic junction temperature safety threshold is determined in the following way: Obtain the steady-state baseline temperature of the digital temperature sensor in the idle state; Based on the steady-state base temperature, a preset temperature-threshold offset mapping table is queried to obtain the initial junction temperature safety threshold.

4. The method according to claim 3, characterized in that, After obtaining the initial junction temperature safety threshold, the process also includes: The frequency of micro-process parameter intervention commands triggered within a preset sliding time window is counted. If the intervention frequency is greater than a preset frequency upper limit, the initial junction temperature safety threshold is adjusted downward by a preset safety step size based on the extent to which the intervention frequency exceeds the preset frequency upper limit. The adjusted initial junction temperature safety threshold is then used as the dynamic junction temperature safety threshold.

5. The method according to claim 1, characterized in that, The step of generating microscopic process parameter intervention instructions for the heat source based on the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold includes: If the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold is within the first deviation range, a first-level intervention instruction is generated. The first-level intervention instruction is used to reduce the drive current of the main controller's internal charge pump to a preset safety lower limit and insert a microsecond-level idle interval between NAND write operations. If the deviation is in the second deviation range, which is greater than the first deviation range, then a second intervention instruction is generated based on the execution of the first intervention instruction. The second intervention instruction is used to limit the maximum number of iterations of the current LDPC soft decision decoding to a preset lower limit and to start the NAND multi-channel alternating sleep mechanism.

6. The method according to claim 1, characterized in that, After executing the micro-process parameter intervention command, the actual temperature rise slope of the digital temperature sensor is continuously extracted. If the actual temperature rise slope exceeds the preset expected temperature drop slope, the following applies: Within the preset feedback observation window after executing the micro-process parameter intervention command, the temperature sequence of the digital temperature sensor is collected at a microsecond-level cycle. The actual temperature rise slope is calculated based on the temperature sequence fitting, and the expected temperature drop slope is calculated based on the expected power consumption reduction after intervention.

7. The method according to claim 6, characterized in that, The generation of global performance degradation instructions based on the difference between the actual temperature rise slope and the expected temperature drop slope includes: Calculate the slope deviation between the actual temperature rise slope and the expected temperature drop slope. If the slope deviation is greater than a preset slope deviation tolerance threshold, the current heat dissipation environment is determined to be degraded. Based on the slope deviation, a global performance degradation instruction to forcibly suspend the non-core I / O queue is generated.

8. The method according to claim 2, characterized in that, After combining the micro-process parameter intervention command and the global performance degradation command to output a solid-state drive junction temperature closed-loop control strategy, the following is also included: After triggering the global performance degradation instruction, obtain the measured peak temperature when the digital temperature sensor reaches its peak temperature. Calculate the prediction residual between the measured peak temperature and the target predicted junction temperature when generating the global performance degradation instruction.

9. The method according to claim 7, characterized in that, If the predicted residual continues to exceed the preset residual threshold, the heat conduction delay time constant is corrected by reverse decay based on the predicted residual, and the corrected heat conduction delay time constant is updated in the subsequent exponential weighted integral calculation.

10. A solid-state drive (SSD) for implementing the method according to any one of claims 1 to 9, characterized in that, It includes a main control chip, a NAND flash memory array connected to the main control chip, and a digital temperature sensor connected to the main control chip; The digital temperature sensor is used to collect the basic measured temperature and send it to the main control chip; The main control chip is used to acquire the current input and output load characteristic data of the solid-state drive, perform heat source feature mapping and advance integration on the input and output load characteristic data, and obtain the transient predicted temperature rise increment of the internal heat-generating area of ​​the main control chip and the NAND flash memory array at the current moment. The main control chip is also used to acquire the basic measured temperature, and to spatiotemporally align and superimpose the transient predicted temperature rise increment with the basic measured temperature to obtain the target predicted junction temperature at the current moment. The main control chip is also used to generate micro-process parameter intervention instructions for the heat source based on the deviation between the target predicted junction temperature and the dynamic junction temperature safety threshold if the target predicted junction temperature exceeds the preset dynamic junction temperature safety threshold. The main control chip is also used to limit the iteration limit of the LDPC codec inside the main control chip and limit the NAND channel concurrency of the NAND flash array according to the micro-process parameter intervention instructions. The main control chip is also used to continuously extract the actual temperature rise slope of the digital temperature sensor after executing the micro-process parameter intervention command. If the actual temperature rise slope exceeds the preset expected temperature drop slope, a global performance degradation command is generated based on the difference between the actual temperature rise slope and the expected temperature drop slope. The main control chip is also used to combine the micro-process parameter intervention instructions and the global performance degradation instructions to output a solid-state drive junction temperature closed-loop control strategy.

Images