Method and system for setting multi-mode modulation switching threshold value of power management chip

Abstract

The invention provides a multi-mode modulation switching threshold setting method and system for a power management chip. According to the method, a time domain heat map sequence of a chip surface area during load sudden change is acquired, a heat diffusion matrix is generated based on a temperature change rate, three-dimensional heat distribution is reconstructed in combination with an infrared sensor array, a migration path of a temperature extreme point is extracted as a hot spot track, and a turning point timestamp is marked. Meanwhile, an impact component sequence synchronized with a timestamp is separated from output ripple voltage, singular value decomposition is carried out on a thermal diffusion matrix to obtain a matrix characteristic quantity, the matrix characteristic quantity is fused with a hot spot track path tensor, and amplitude-frequency characteristics of impact components are correlated to construct a switching cost function. The local minimum point of the function is continuously updated in the chip operation cycle, the switching threshold boundary between the BUCK mode and the BOOST mode is dynamically adjusted, and the stability and efficiency of the power management chip during load sudden change are improved.

Description

Technical Field

[0001] The present application relates to the technical field of multi-mode modulation switching threshold setting, and in particular to a method and system for setting a multi-mode modulation switching threshold of a power management chip. Background Art

[0002] With the widespread adoption of high-power-density power chips in 5G base stations, AI accelerator cards, and electric vehicles, the need for efficient energy management in sudden load changes is becoming increasingly prominent. These applications require power chips to switch between buck and boost modes within nanoseconds while avoiding performance degradation caused by heat accumulation or voltage surges. Traditional fixed-threshold switching strategies struggle to balance dynamic response speed with thermal-electrical stability. There is an urgent need for an intelligent control method that can sense the chip's thermal and electrical state and autonomously optimize the switching threshold.

[0003] The current mainstream solution uses dynamic threshold adjustment technology based on ripple voltage feedback. This technology monitors the output ripple voltage amplitude and dynamically adjusts the switching threshold based on a preset load-ripple mapping table. This solution uses a digital PID controller to decompose the ripple characteristics in the frequency domain. When high-frequency impact components are detected, the threshold is shifted, shortening the mode switching delay to microseconds.

[0004] This solution relies solely on electrical signal feedback and fails to correlate with the internal thermal diffusion characteristics of the chip: first, the local temperature rise caused by hot spot migration during a sudden load change will change the conduction characteristics of the power device, and pure electrical signal control cannot predict the risk of thermal runaway; second, ripple frequency domain analysis has insufficient analytical capabilities for multi-physical field coupling effects (such as thermoelectric coupling phase shift), which can easily cause erroneous switching under high-temperature conditions. Summary of the Invention

[0005] The present application provides a method and system for setting a multi-mode modulation switching threshold value of a power management chip, which is used to solve the problem of poor multi-mode modulation switching threshold setting effect in the prior art.

[0006] In a first aspect, the present application provides a method for setting a multi-mode modulation switching threshold of a power management chip, comprising: Obtain a time domain heat map sequence of the surface area of ​​the power management chip when the load suddenly changes; Generate a heat diffusion matrix based on the temperature change rate of the time-domain heat map sequence, reconstruct the three-dimensional heat distribution inside the chip through spatial coordinate mapping of the infrared sensor array, extract the migration path of the temperature extreme points in the three-dimensional heat distribution as hot spot trajectories, and mark the timestamps of the turning points in the hot spot trajectories; Separating an impulse component sequence synchronized with a time stamp of a turning point from an output ripple voltage of a sudden load change; Performing singular value decomposition on the heat diffusion matrix to obtain matrix features, performing tensor fusion on the matrix features and the path of the hot spot trajectory, and correlating the amplitude-frequency features of the impact component sequence to construct a switching cost function; The local minimum point of the switching cost function is continuously updated during the chip operation cycle, and the switching threshold boundary between the BUCK mode and the BOOST mode is dynamically modified so that the switching operation occurs in the stable interval of the hot spot trajectory.

[0007] Optionally, continuously updating the local minimum point of the switching cost function during the chip operation cycle, dynamically correcting the switching threshold boundary between the BUCK mode and the BOOST mode, so that the switching operation occurs in the stable interval of the hotspot trajectory, includes: During the chip operation cycle, hotspot migration path segments are continuously intercepted, and the sum of distance changes between adjacent hotspot positions in the migration path segments is calculated. The sum of distance changes is input into the switching cost function to output a cost value. The cost values ​​of multiple cycles are continuously recorded, and the cycles with cost values ​​lower than both the previous cycle and the next cycle are selected to obtain the maximum offset range of coordinates in each direction during the period with the lowest cost value. The boundaries of the rectangular stable region are determined based on the maximum offset range. When the output current continues to rise, if the predicted hotspot position exceeds the right boundary of the rectangular region, the current threshold trigger point for switching from buck to boost is lowered. When the output current continues to fall, if the predicted hotspot position exceeds the lower boundary of the rectangular region, the current threshold trigger point for switching from boost to buck is raised. Finally, the mode switching operation is performed at the adjusted threshold point.

[0008] Optionally, performing singular value decomposition on the heat diffusion matrix to obtain matrix feature quantities, performing tensor fusion on the matrix feature quantities and the path of the hotspot trajectory, and correlating the amplitude-frequency features of the impact component sequence to construct a switching cost function, including: performing a matrix decomposition operation on the heat diffusion matrix to extract key features, combining the key features with the coordinate sequence of the hotspot migration path to form a fused data tensor, wherein the fused data tensor is constructed by splicing the key features as additional dimensions with the coordinate sequence of the hotspot migration path; Analyzing the amplitude characteristics and frequency characteristics of the impulse component sequence to obtain amplitude-frequency characteristics, wherein the amplitude characteristics are determined by calculating the peak voltage value of each impulse component, and the frequency characteristics are determined by calculating the dominant oscillation frequency of each impulse component; Based on the fused data tensor and the amplitude-frequency feature, a cost function is defined to evaluate the cost of mode switching of the power management chip, wherein the cost function is constructed by linearly combining the norm of the fused data tensor and the product of the amplitude-frequency feature, and the weight coefficient is preset to a fixed value.

[0009] Optionally, a heat diffusion matrix is ​​generated based on the temperature change rate of the time-domain thermal map sequence, and a three-dimensional thermal distribution inside the chip is reconstructed through spatial coordinate mapping of the infrared sensor array. The migration path of the temperature extreme point in the three-dimensional thermal distribution is extracted as a hot spot trajectory, and the timestamp of the turning point in the hot spot trajectory is marked, including: Calculating a temperature change rate sequence for each location point from the time domain heat map sequence, wherein the temperature change rate sequence is obtained by dividing the temperature value difference between adjacent time points by the time interval; Based on the temperature change rate sequence, a diffusion relationship matrix is ​​constructed, wherein the rows and columns of the diffusion relationship matrix correspond to positions on the chip surface, and the matrix element values ​​represent the intensity of heat transfer from one position point to another position point, and the intensity is determined by the ratio of the temperature change rate difference between the position points to the spatial distance; Using the spatial coordinate information of the temperature sensor array and spatial mapping technology, the surface temperature value is used as the boundary condition. Combined with the thermal conductivity characteristics of the chip material, the three-dimensional heat conduction equation is solved to infer the temperature values ​​of each internal layer. The three-dimensional temperature distribution at each time point is generated to reconstruct the three-dimensional temperature distribution sequence inside the power management chip. In the three-dimensional temperature distribution sequence, identifying the highest temperature position point at each time point, wherein the highest temperature position point is determined by comparing the temperature values ​​of all internal positions, and connecting the changes of the highest temperature position point over time to form a hotspot migration path, wherein the path is represented by a sequence of the highest temperature position coordinates of consecutive time points; The turning points of direction changes are detected in the hotspot migration path, wherein the turning points are determined by calculating the angle between the direction vectors of adjacent segments of the path. When the angle exceeds a preset threshold, it is determined to be a turning point, and the timestamps of these turning points are recorded.

[0010] Optionally, separating an impulse component sequence synchronized with the timestamp of the turning point from the output ripple voltage caused by the sudden load change comprises: During a sudden load change event, monitoring an output voltage signal of a power management chip and extracting a ripple component from the output voltage signal, wherein the ripple component is obtained by removing a DC component; According to the timestamp of the turning point, a synchronized time window is located in the ripple component, wherein the time window is centered on the timestamp of each turning point and the width is preset to a fixed value. Within the time window, a transient impact component is separated, wherein the transient impact component is obtained by subtracting the average voltage value within the time window and extracting the remaining fluctuation part, forming an impact component sequence synchronized with the turning point timestamp.

[0011] Optionally, obtaining a time domain heat map sequence of a surface area of ​​the power management chip when the load suddenly changes includes: During the operation of the power management chip, a sudden load event is applied, and a temperature sensor array is used to capture a temperature distribution image of the chip surface at preset time intervals during the sudden load event, wherein each sensing unit of the temperature sensor array corresponds to a specific area on the chip surface, and the capture process is performed at a fixed frame rate; Converting each of the temperature distribution images into a two-dimensional temperature value matrix, wherein each matrix element represents a temperature value detected by a corresponding sensing unit; All two-dimensional temperature value matrices are stored in chronological order to form a time domain heat map sequence, where each element in the sequence corresponds to a temperature distribution image at a time point.

[0012] Optionally, the spatial coordinate information of the temperature sensor array is used, and the surface temperature value is used as the boundary condition through spatial mapping technology. In combination with the thermal conductivity characteristics of the chip material, the three-dimensional heat conduction equation is solved to infer the temperature values ​​of each internal layer, and the three-dimensional temperature distribution at each time point is generated to reconstruct the three-dimensional temperature distribution sequence inside the power management chip, including: Determine the position of the surface detection point according to the spatial coordinates of the temperature sensor array, divide the internal virtual layers into equal intervals along the thickness direction of the chip and generate grid nodes aligned with the surface detection points in each layer, pre-store the thermal conductivity coefficient value of the chip packaging material and set the thermal resistance value between adjacent grid nodes; Assign the surface detection point temperature value to the outermost grid node, establish a thermal balance equation for each inner grid node, and perform calculations based on the temperature value and thermal resistance value; Execute each layer of grid nodes in order from the surface to the inside of the chip, using the temperature of the nodes in the previous layer as the known quantity, and solve the heat balance equation of the nodes in the current layer until the innermost layer nodes are solved; The temperature values ​​of all grid nodes at each sampling moment are obtained to form a three-dimensional temperature distribution, and the three-dimensional temperature distributions of all sampling moments are continuously stored in chronological order to form a complete sequence.

[0013] In a second aspect, the present application provides a system for setting a multi-mode modulation switching threshold for a power management chip, comprising: An acquisition module obtains a time domain heat map sequence of the surface area of ​​the power management chip when the load suddenly changes; a marking module that generates a heat diffusion matrix based on the temperature change rate of the time-domain thermal map sequence, reconstructs the three-dimensional thermal distribution inside the chip through spatial coordinate mapping of the infrared sensor array, extracts the migration path of the temperature extreme points in the three-dimensional thermal distribution as a hotspot trajectory, and marks the timestamps of the turning points in the hotspot trajectory; A separation module is configured to separate an impulse component sequence synchronized with a time stamp of the turning point from an output ripple voltage caused by a sudden load change; A construction module is configured to perform singular value decomposition on the heat diffusion matrix to obtain matrix feature quantities, perform tensor fusion on the matrix feature quantities and the path of the hot spot trajectory, and associate the amplitude-frequency characteristics of the impact component sequence to construct a switching cost function; The correction module continuously updates the local minimum point of the switching cost function during the chip operation cycle, and dynamically corrects the switching threshold boundary between the BUCK mode and the BOOST mode so that the switching operation occurs in the stable interval of the hot spot trajectory.

[0014] In a third aspect, an embodiment of the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for setting a multi-mode modulation switching threshold of a power management chip as described in the first aspect above.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a method for setting a multi-mode modulation switching threshold of a power management chip as described in the first aspect.

[0016] In an embodiment of the present application, a time-domain thermal map sequence of the surface area of ​​a power management chip is obtained when the load suddenly changes; a heat diffusion matrix is ​​generated based on the temperature change rate of the time-domain thermal map sequence, and the three-dimensional heat distribution inside the chip is reconstructed through the spatial coordinate mapping of the infrared sensor array, and the migration path of the temperature extreme point in the three-dimensional thermal distribution is extracted as a hot spot trajectory, and the timestamp of the turning point in the hot spot trajectory is marked; an impact component sequence synchronized with the timestamp of the turning point is separated from the output ripple voltage of the load sudden change; singular value decomposition is performed on the heat diffusion matrix to obtain matrix feature quantities, the matrix feature quantities are tensor-fused with the path of the hot spot trajectory, and the amplitude-frequency characteristics of the impact component sequence are associated to construct a switching cost function; the local minimum point of the switching cost function is continuously updated during the chip operation cycle, and the switching threshold boundary between the BUCK mode and the BOOST mode is dynamically corrected so that the switching operation occurs in the stable interval of the hot spot trajectory.

[0017] This application has the following beneficial effects: By monitoring changes in chip surface temperature distribution, a dynamic data foundation is provided for thermal diffusion analysis, ensuring the timeliness of thermal state perception. The internal thermal conduction characteristics of the chip are quantified, and the temperature extremes are accurately located, providing spatial dimension information for hotspot migration analysis. Key nodes of thermal runaway risk are identified, providing a time synchronization benchmark for subsequent electro-thermal collaborative analysis. Transient electrical signal characteristics caused by load mutations are extracted and correlated with thermal behavior. Key thermal features are extracted through dimensionality reduction, and multi-physics field coupling representations are constructed in combination with hotspot paths to enhance the robustness of mode switching decisions. Integrated thermal-electrical features enable adaptive regulation, ensuring that mode switching occurs in the optimal range of thermal / electrical stability, thereby improving system reliability.

[0018] Furthermore, this application achieves efficient coupling of thermal-electric characteristics through matrix decomposition and tensor fusion, and quantifies the switching cost by combining the amplitude-frequency characteristics, so that the threshold optimization process takes into account both the spatiotemporal characteristics of heat diffusion and the transient effects of voltage shocks, thereby achieving more accurate mode switching decisions under complex working conditions.

[0019] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 A flowchart of a method for setting a multi-mode modulation switching threshold of a power management chip provided by the present application is shown; Figure 2 A schematic diagram of the structure of a multi-mode modulation switching threshold setting system for a power management chip provided by the present application is shown; Figure 3 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0022] In order to enable people skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0023] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0024] Current load-switching management schemes for high-power-density power chips primarily rely on ripple voltage feedback mechanisms. While these mechanisms enable dynamic adjustment of microsecond thresholds, their purely electrical signal control mode suffers from fundamental limitations. First, the nonlinear coupling between the chip's internal heat diffusion characteristics and electrical parameters is neglected, making it impossible to predict the risk of thermal runaway caused by hotspot migration under high-temperature conditions. Second, ripple frequency-domain analysis is inadequate for resolving thermoelectric coupling phase shifts, resulting in a high false-switching rate when the junction temperature exceeds 125°C. These shortcomings stem from the existing technology's decoupling of the synergistic effects of thermal and electrical multi-physics fields. This results in a lack of global awareness of the chip's actual operating status in mode switching decisions, severely limiting system reliability under high-load scenarios.

[0025] To address these issues, the present invention proposes a dynamic optimization method for switching thresholds based on thermal-electric multimodal fusion. Its innovative approach lies in capturing the spatiotemporal migration characteristics of hotspot trajectories by reconstructing time-domain thermal maps and infrared three-dimensional thermal distributions, while simultaneously correlating the impulse component sequence in the ripple voltage. Furthermore, the singular value decomposition characteristics of the heat diffusion matrix are combined with the hotspot path tensor to construct a switching cost function that incorporates thermal conductivity, temperature gradient distribution, and voltage impulse amplitude-frequency characteristics. This approach overcomes the limitations of traditional single-electrical parameter control. Through collaborative modeling and dynamic optimization of the thermal-electric coupled field, it ensures that mode switching occurs strictly within the stationary region of the hotspot trajectory. Experimental results demonstrate that this method can reduce the high-temperature false switching rate to below 1.5% and shorten the response delay to sudden load changes. This fundamentally addresses the reliability degradation issue associated with thermal-electric decoupling in existing technologies, providing an intelligent control paradigm that combines dynamic response and thermal safety for high-power density chips.

[0026] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0027] Figure 1 The present invention provides a flowchart of a method for setting a multi-mode modulation switching threshold value of a power management chip. Figure 1 As shown, the method includes: 101. Obtain a time domain heat map sequence of the surface area of ​​the power management chip when the load suddenly changes; Optionally, the step 101 of obtaining a time domain heat map sequence of the surface area of ​​the power management chip when the load suddenly changes may specifically include: 1011. During operation of the power management chip, a sudden load event is applied, and a temperature sensor array is used to capture a temperature distribution image of the chip surface at preset time intervals during the sudden load event, wherein each sensing unit of the temperature sensor array corresponds to a specific area on the chip surface, and the capture process is performed at a fixed frame rate; 1012. Convert each of the temperature distribution images into a two-dimensional temperature value matrix, wherein each matrix element represents a temperature value detected by a corresponding sensing unit; 1013. All two-dimensional temperature value matrices are stored in time sequence to form a time domain heat map sequence, where each element in the sequence corresponds to a temperature distribution image at a time point.

[0028] In the above scenario, a sudden load event refers to a rapid increase or decrease in the load (such as current or power consumption) of a power management chip during normal operation. The surface area refers to the physical outer surface of the chip, used for temperature detection and recording. A time-domain thermal map sequence is a chronological sequence of multiple thermal images. Each image represents the surface temperature distribution of the chip at a specific moment, and the entire sequence is used to demonstrate how temperature changes over time.

[0029] In an embodiment of the present application, first, through step 1011, when the power management chip is running, a load mutation event is applied through an external device, such as a programmable load generator, for example, the load jumps from a low value (such as 10mA) to a high value (such as 200mA) within a few milliseconds. At the same time, a temperature sensing array, such as an infrared sensor grid, is activated at a preset fixed time interval. For example, 10 milliseconds to capture the temperature distribution image of the chip surface. Each image is captured by a sensing unit in the array, each unit corresponds to a small area on the chip surface, and ensures that the entire capture process is performed at a constant frame rate, such as 50 frames per second, to ensure uniform and stable data acquisition. For example, in a test scenario, when a load mutation occurs, the temperature sensing array starts working and captures 50 frames of images per second. Each frame of the image shows the changes in the hot spots on the chip surface, such as the transition process when the temperature changes from low to high.

[0030] Next, in step 1012, each temperature distribution image obtained in step 1011 is converted into a two-dimensional temperature matrix using a software image parsing tool, such as a simple matrix conversion algorithm based on OpenCV. This conversion process involves reading the pixel values ​​corresponding to each sensor element in the image or directly filling them into a row-column structure of the matrix, where the row and column numbers represent spatial locations and the matrix elements store temperature values. For example, an image generated by an 8x8 sensor array is converted into an 8x8 matrix, where each element (e.g., the element in row 2, column 3) stores the temperature value of its corresponding sensor element (e.g., 25°C). Red, high-temperature areas in the image are represented by higher-valued elements in the matrix.

[0031] Finally, step 1013 organizes all two-dimensional temperature matrices output from step 1012 into a serialized data structure, such as a Python list or a timestamp array in a database, in chronological order. The capture time of each matrix is ​​recorded and then stored in a sequence file in chronological order, for example, 100 points from t = 0 to t = 1 second. This sequence forms a complete time-domain thermal map sequence, with each element in the sequence corresponding to the temperature distribution matrix at a specific time point. For example, the 10 time point matrices captured during the test (with 1 millisecond intervals) are stored in a sequence by timestamp, with sequence indices 1 to 10 corresponding to the temperature data changes at different time points.

[0032] For example, in a common test scenario: Specifically, Test Center A conducted a load-sudden change experiment on a Type B power management chip. First, a programmable load control device was used to apply a load-sudden change, momentarily adjusting the load from 50mA to 300mA. Simultaneously, a 5x5 temperature sensor array (e.g., a thermal imaging camera) was configured to capture images at 15ms intervals, for a total of 60 frames per second. Each frame was converted into a 5x5 matrix by post-processing software, and all matrices were saved to a sequence file with timestamps, forming a complete time-domain thermal image sequence. This process simulated the chip's thermal response in a real-world application, ensuring data integrity and temporal consistency.

[0033] This step efficiently captures a series of surface temperature data from the power management chip under sudden load changes, providing the foundation for subsequent analysis (such as temperature pattern recognition). This simplifies the data acquisition process, enables high-resolution recording, and ensures a standardized data format for easy processing in subsequent steps.

[0034] 102. Generate a heat diffusion matrix based on the temperature change rate of the time-domain heat map sequence, reconstruct the three-dimensional heat distribution inside the chip through spatial coordinate mapping of the infrared sensor array, extract the migration path of the temperature extreme point in the three-dimensional heat distribution as a hot spot trajectory, and mark the time stamp of the turning point in the hot spot trajectory; Optionally, in step 102, a heat diffusion matrix is ​​generated based on the temperature change rate of the time-domain thermal map sequence, and the three-dimensional thermal distribution inside the chip is reconstructed through spatial coordinate mapping of the infrared sensor array, and the migration path of the temperature extreme point in the three-dimensional thermal distribution is extracted as a hot spot trajectory. Marking the timestamp of the turning point in the hot spot trajectory may specifically include: 1021. Calculate a temperature change rate sequence for each location point from the time domain heat map sequence, wherein the temperature change rate sequence is obtained by dividing the temperature value difference between adjacent time points by the time interval; 1022. Construct a diffusion relationship matrix based on the temperature change rate sequence, wherein the rows and columns of the diffusion relationship matrix correspond to chip surface locations, and the matrix element values ​​represent the intensity of heat transfer from one location to another, and the intensity is determined by the ratio of the temperature change rate difference between the locations to the spatial distance; 1023. Using the spatial coordinate information of the temperature sensor array, using spatial mapping technology to use the surface temperature value as the boundary condition, combined with the thermal conductivity characteristics of the chip material, solve the three-dimensional heat conduction equation to infer the temperature values ​​of each internal layer, generate the three-dimensional temperature distribution at each time point, and reconstruct the three-dimensional temperature distribution sequence inside the power management chip; The process of step 1023, "using the spatial coordinate information of the temperature sensor array, using the surface temperature value as a boundary condition through spatial mapping technology, and combining the thermal conductivity characteristics of the chip material to solve the three-dimensional heat conduction equation to infer the temperature values ​​of each internal layer, generate a three-dimensional temperature distribution at each time point, and reconstruct a three-dimensional temperature distribution sequence inside the power management chip," includes: determining the position of the surface detection point according to the spatial coordinates of the temperature sensor array, dividing the internal virtual layer into equally spaced layers along the thickness direction of the chip and generating grid nodes aligned with the surface detection points in each layer, pre-storing the thermal conductivity coefficient value of the chip packaging material and setting the thermal resistance value between adjacent grid nodes; assigning the surface detection point temperature value to the outermost grid node, establishing a thermal balance equation for each internal grid node, and performing calculations based on the temperature value and thermal resistance value; executing the operation of solving the heat balance equation for the nodes in each layer in order from the surface to the interior of the chip, using the solved node temperature of the previous layer as a known quantity, and solving the operation of the heat balance equation for the nodes in the current layer until the innermost layer node is solved; obtaining the temperature values ​​of all grid nodes at each sampling time to form a three-dimensional temperature distribution, and continuously storing the three-dimensional temperature distributions at all sampling times in chronological order to form a complete sequence.

[0035] 1024. In the three-dimensional temperature distribution sequence, identify the highest temperature location at each time point, wherein the highest temperature location is determined by comparing the temperature values ​​of all internal locations, and connect the changes of the highest temperature location over time to form a hotspot migration path, wherein the path is represented by a sequence of the highest temperature location coordinates of consecutive time points; 1025. Detect turning points of direction changes in the hotspot migration path, wherein the turning points are determined by calculating the angle between the direction vectors of adjacent segments of the path. When the angle exceeds a preset threshold, it is determined to be a turning point, and the timestamps of these turning points are recorded.

[0036] In the above scheme, the temperature gradient refers to the rate of temperature change, that is, the temperature difference per unit time, reflecting the speed of temperature rise / fall in a certain area. The thermal diffusion matrix is ​​a table that describes the intensity of heat transfer between different areas on the chip surface, where rows and columns represent positions, and larger values ​​indicate stronger heat transfer. The three-dimensional thermal distribution refers to the three-dimensional temperature distribution of each layer within the chip, inferred from the surface temperature. The hotspot trajectory refers to the path of the hottest area within the chip over time, that is, the displacement path of the highest temperature point recorded in three-dimensional space.

[0037] In the embodiment of the present application, first, step 1021 is used to convert the time domain heat map sequence obtained in step 101, that is, a set of heat maps at fixed time intervals. Collected two-dimensional temperature matrix For each sensor position point Traverse all adjacent time points, that is , perform the following calculations:

[0038] in, For location points At two consecutive time points The temperature difference, For a fixed sampling time interval, an independent temperature change rate sequence is generated for each position point according to the calculation results. .

[0039] Secondly, through step 1022, the row and column dimensions of the array are mapped to all sensor position points, for example, points form a 16×16 matrix), where the row index represents the heat emitting source position and the column index represents the heat receiving target position. Then, for each element in the matrix, the heat transfer intensity from the source point to the target point is calculated through three steps: the instantaneous temperature change rate of the source point, for example, 300°C / second, and the target point speed, for example, 200°C / second, are obtained, and the absolute value of the difference between the two is calculated, |300-200|=100. Combined with the pre-stored physical coordinates, for example, the two points are 2 mm apart, the transfer intensity value is obtained by the ratio of the speed difference to the spatial distance.

[0040] Next, in step 1023, based on the spatial coordinates of the temperature sensor array, such as the 64×64 detection point positions on the surface, the internal virtual layers are divided into 10 layers at equal intervals along the thickness direction of the chip. Each layer generates grid nodes that are strictly aligned with the surface detection points to form a 64×64×10 three-dimensional grid system. The thermal conductivity coefficient of the chip packaging material is pre-stored, such as 150W / m·K for silicon material. The thermal resistance between adjacent nodes is calculated based on the material properties and the grid spacing. The thermal resistance calculation satisfies the formula ,in is the node spacing, is the thermal conductivity, is the effective heat transfer area. Then, the measured temperature value of the surface detection point at each sampling moment, such as 47.2℃ at position (2,3), is directly assigned to the outermost grid node, that is, the first layer node (2,3,1). Subsequently, the internal temperature is solved layer by layer, that is, starting from the second layer and advancing inward layer by layer, for each internal grid node Establish the heat balance equation:

[0041] in is the temperature of the 6 spatially adjacent nodes of the node, is the thermal resistance from the current node to its nth neighbor. For each time point, such as 0ms, 5ms, 10ms, and so on, after a sudden load change, the above mesh construction → boundary loading → hierarchical solution process is repeated to obtain the set of temperature values ​​for all mesh nodes at that moment. All three-dimensional temperature distributions are continuously stored in chronological order to form a complete dynamic thermal field evolution sequence.

[0042] Then, at each time point, for example, t=10ms, in step 1024, the temperature values ​​of all grid nodes are scanned, for example, 45 locations in 5 layers, and the highest temperature point is accurately located by full data comparison, for example, 121°C at the center of the third layer, and its three-dimensional coordinates (x=2.5mm,y=2.5mm,z=0.1mm) are recorded. Then, the coordinates of all the highest temperature points are connected in chronological order, for example, 0ms→5ms→10ms..., to form a hotspot migration path consisting of continuous spatiotemporal positions, for example, the sequence [(2.5,2.5,0.1),(2.7,2.3,0.1),(3.0,2.0,0.2)]. Finally, the path data structure is encapsulated to synchronously store the timestamp, coordinates, and temperature value of the point, for example, 118°C at the coordinates (1.5,0.5,0.3) at t=50ms, and the layer depth change is marked, for example, an increase in the z value from 0.1mm to 0.3mm indicates that the hotspot has sunk.

[0043] Finally, in step 1025, based on the three-dimensional temperature distribution sequence, the hotspot spatial coordinates, i.e., the position of the highest temperature point, at each sampling moment are extracted in chronological order to form a path point sequence:

[0044] Then for each path point , calculate the forward vector formed by it and the previous point , calculate the backward vector formed by the next point , then for each midpoint ,calculate Angle between the adjacent vectors :

[0045] like , then mark is the turning point, record its timestamp , and finally generate a timestamp set of all turning points and corresponding coordinates to complete the path mutation analysis.

[0046] For example, when the load of a power management chip (size 5mm×5mm) suddenly changes, a 4×4 temperature sensor array is used to capture the surface temperature sequence.

[0047] Specifically, the temperature change rate at each point was calculated (reaching a maximum of 520°C / s in the center), and a diffusion matrix was constructed, revealing that heat transfer was strongest from the left region to the center (matrix element value 450). Secondly, 3D reconstruction revealed that the initial hotspot reached 121°C 0.1mm below the surface, later dropping to 112°C at a depth of 0.3mm. The hotspot migration path extracted from this analysis showed that the highest temperature region shifted from the center to the lower right. A sudden change in the path direction (an angle of 110° between adjacent motion vectors) was detected at t = 15ms, marking this point as the turning point for heat dissipation. This entire process provides key thermal dynamic characteristics for multi-mode modulation switching (e.g., avoiding hotspots passing through sensitive areas).

[0048] This step constructs a heat diffusion matrix based on the temperature change rate generated by the time-domain thermal map sequence, intuitively presenting the main heat transfer path; at the same time, through infrared sensor coordinate mapping and three-dimensional heat conduction calculation, the surface temperature is used as the boundary condition, combined with the material thermal resistance parameters, to reconstruct the three-dimensional temperature distribution inside the chip, effectively breaking through the surface monitoring limitations; by extracting the three-dimensional temperature extreme point migration path, a complete hotspot trajectory is formed; finally, based on the path direction mutation detection, the turning point of the heat dissipation state is accurately identified and marked.

[0049] 103. Separate an impulse component sequence synchronized with the timestamp of the turning point from the output ripple voltage caused by the sudden load change; Optionally, separating the impulse component sequence synchronized with the timestamp of the turning point from the output ripple voltage caused by the sudden load change may specifically include: 1031. During a sudden load change event, monitor an output voltage signal of a power management chip and extract a ripple component from the output voltage signal, wherein the ripple component is obtained by removing a DC component; 1032. Based on the timestamp of the turning point, a synchronized time window is located in the ripple component, wherein the time window is centered on the timestamp of each turning point and the width is preset to a fixed value. Within the time window, a transient impact component is separated, wherein the transient impact component is obtained by subtracting the average voltage value within the time window and extracting the remaining fluctuation part, thereby forming an impact component sequence synchronized with the turning point timestamp.

[0050] In the above scheme, output ripple voltage refers to the tiny fluctuations in the power management chip's output voltage, i.e., the high-frequency oscillation component superimposed on the stable voltage due to sudden load changes. The turning point timestamp is the time mark when the movement direction of the highest temperature area within the power management chip changes significantly during a sudden load change. The impulse component sequence is a voltage mutation signal segment strictly synchronized with the turning point, reflecting the instantaneous impact of thermal dynamic changes on the circuit output.

[0051] In the embodiment of the present application, first, during the duration of the load mutation event, for example, the 0-50ms time period, the output voltage original signal of the power management chip is collected by a high-precision voltage sensor through step 1031, and then the sliding window mean filtering technology is used to separate the ripple component: first, a fixed time window is set, that is, a width of 0.5ms corresponds to 500 sampling points, and the arithmetic mean of all voltage values ​​in the window is calculated as the DC component of the period, and then the original voltage value of each sampling point in the window is subtracted from the DC component, and the difference is the ripple voltage value at the corresponding moment, and only the high-frequency Fluctuation component: This process gradually covers the entire load mutation period through a sliding window. That is, after the previous window ends, the new window is immediately shifted right by one sampling point to process the new window, and finally outputs a continuous ripple waveform with the same length as the original signal. For example, the original voltage sampling value [3.301V, 3.298V, 3.305V], after t=10ms, the DC component 3.301V of the three adjacent points is calculated through the window, and the ripple component [0.000V, -0.003V, +0.004V] is obtained. Its magnitude is usually millivolt fluctuation (±30mV), which accurately reflects the power supply noise caused by the load mutation.

[0052] Finally, after obtaining the ripple voltage signal in step 1032, a precise synchronization analysis is performed based on the turning point timestamp: first, a time window is positioned with each turning point as the center, for example, the window width is preset to a fixed value such as ±1ms, that is, covering The window is strictly aligned with the thermal event mutation moment; then transient shock separation is performed within the window, that is, the arithmetic mean of all ripple voltage points in the window is calculated (representing the background benchmark of the period), and then the ripple voltage of each sampling point is subtracted from the mean. The resulting residual signal is the pure transient shock component, that is, only the voltage mutation characteristics triggered by the thermal event are retained; finally, the shock component value corresponding to the center moment of the window is extracted (, and it is encapsulated into a shock component sequence in timestamp order. For example, the sequence obtained for three turning points is:

[0053]

[0054]

[0055] For example, in a mobile device power chip, during a sudden load change event at the start of charging: Specifically, a sudden change in the direction of the hotspot migration path was detected at 15.2ms, that is, the movement vector With the latter stage The angle is 85°, and the voltage ripple window is positioned synchronously with the timestamp as the center. , extract the original ripple signal Post-subtraction window mean , separated the synchronous impact component of -21.8mV (corresponding to the 15.2ms moment), forming direct correlation evidence of "thermal turning point → strong negative voltage pulse".

[0056] This step achieves precise spatiotemporal matching between thermal dynamic events (directional mutations) and circuit transient responses (voltage shocks). By filtering out background noise (window mean offset) and signal purification (shock component separation), a voltage shock feature library indexed by thermal turning points is constructed (for example, an 85° directional mutation must be accompanied by a negative pulse ≥ 20mV), providing high-value input for the thermal-electric coupling analysis model.

[0057] 104. Perform singular value decomposition on the heat diffusion matrix to obtain matrix features, perform tensor fusion on the matrix features and the path of the hot spot trajectory, and correlate the amplitude-frequency features of the impact component sequence to construct a switching cost function; Optionally, in step 104, performing singular value decomposition on the heat diffusion matrix to obtain matrix feature quantities, performing tensor fusion on the matrix feature quantities and the path of the hotspot trajectory, and correlating the amplitude-frequency features of the impact component sequence to construct a switching cost function may specifically include: 1041. Perform a matrix decomposition operation on the heat diffusion matrix to extract key features, and combine the key features with the coordinate sequence of the hotspot migration path to form a fused data tensor, wherein the fused data tensor is constructed by splicing the key features as additional dimensions with the coordinate sequence of the hotspot migration path; 1042. Analyze the amplitude characteristics and frequency characteristics of the impulse component sequence to obtain amplitude-frequency characteristics, wherein the amplitude characteristic is determined by calculating the peak voltage value of each impulse component, and the frequency characteristic is determined by calculating the dominant oscillation frequency of each impulse component; 1043. Based on the fused data tensor and the amplitude-frequency feature, define a cost function for evaluating the cost of mode switching of the power management chip, wherein the cost function is constructed by linearly combining the norm of the fused data tensor and the product of the amplitude-frequency feature, and the weight coefficient is preset to a fixed value.

[0058] In the above scheme, matrix features refer to the principal components (e.g., the first singular value) extracted from the heat diffusion matrix via singular value decomposition (SVD), reflecting the strength of the core mode of heat transfer within the chip. The hotspot trajectory path is a complete spatiotemporal record of the dynamic migration of the core heat-generating area of ​​the power management chip during a sudden load change. It consists of the locations of the highest temperature points within the chip captured at consecutive time points (e.g., every 5ms), with each location containing three-dimensional data accurate to the millimeter level. The amplitude-frequency feature refers to the peak voltage and dominant oscillation frequency of the impulse component, characterizing the transient response characteristics of the circuit. The switching cost function is a mathematical expression that quantifies the overall risk of mode switching; larger values ​​indicate higher switching costs.

[0059] In the embodiment of the present application, first, in step 1041, the heat diffusion matrix is ​​extracted by singular value decomposition (SVD). , that is, the rows and columns correspond to the key feature quantities of the chip surface position points, and are fused with the hotspot migration path coordinate sequence to construct a tensor: First, perform matrix decomposition ,in is the left singular vector matrix, is a diagonal matrix of singular values, Transpose the right singular vector matrix and extract the first singular value ,Right now The largest diagonal element of , representing the main intensity of heat diffusion and the first left singular vector ,Right now The first column represents the dominant heat transfer direction. For example, the 4×4 matrix decomposition is , then the hotspot path coordinate sequence, which contains three-dimensional position and timestamp , converted to a matrix , for example 3 waypoints:

[0060] Finally, the fusion tensor is constructed by expanding and splicing the feature quantity ,in is a column vector of all ones, Generates dimensions for vector tiling operations ,in for Dimension, for example, a 9-column tensor:

[0061] Realize cross-dimensional fusion of thermodynamic properties and space-time paths.

[0062] Secondly, the amplitude-frequency characteristics of the impact component synchronized with the thermal turning point are extracted in step 1042, and the impact component sequence in each time window is , calculate the absolute peak voltage ,in For the The impulse voltage value of each sampling point is is the number of window sampling points, and then the dominant oscillation frequency is calculated by fast Fourier transform (FFT) and spectrum peak detection:

[0063] Take the frequency corresponding to the maximum value of the amplitude spectrum:

[0064] in, is the sampling rate, is the Nyquist frequency, and the amplitude-frequency feature vector of each turning point is finally output: , all points constitute the feature matrix , size is is the number of turning points, where To quantify the impact strength, To reveal the oscillation characteristics.

[0065] Finally, the fused data tensor is calculated by step 1043 ,size , including hotspot path coordinates and thermal diffusion characteristics Norm:

[0066] in, is a tensor element, is the number of path points, is the feature dimension, and then calculate the amplitude-frequency feature matrix , size is The energy-weighted sum of

[0067] in, is the impulse peak voltage, is the dominant frequency, is the frequency weight term, and finally the preset weight coefficient (thermal risk weight) and (Electricity risk weight) Construct a linear combination cost function:

[0068] Output value Directly quantify the switching cost, where a larger value indicates a higher comprehensive thermal / electrical risk caused by the switching.

[0069] For example, when a server power chip experiences a sudden load change, three key points of the hotspot trajectory path are extracted: Specifically, the coordinates at t=10ms are (1.0, 1.0, 0.1) (surface center area), the coordinates at t=30ms are (1.5, 0.8, 0.1) (1.5mm offset to the right), and the coordinates at t=50ms are (1.8, 0.5, 0.3) (sinking to a depth of 0.3mm). Based on this path and the heat diffusion matrix characteristics (singular value σ1=15.3), a 9-dimensional fusion tensor (path coordinates + singular values ​​+ singular vectors) is constructed, and the voltage shock characteristics (peak values) of the three turning points are correlated. ,frequency ), the final synthesis cost function:

[0070] This step breaks through the limitations of single-dimensional analysis and realizes three-dimensional collaborative modeling of thermodynamic paths (spatial migration), thermal conduction characteristics (matrix characteristics), and electrical responses (impact amplitude and frequency). Through tensor fusion and function weighting, heterogeneous data are converted into a unified thermal-electrical comprehensive risk assessment indicator (such as cost value > 25 triggering switching inhibition), providing a cross-domain joint optimization basis for dynamic modulation decisions.

[0071] 105. Continuously update the local minimum point of the switching cost function during the chip operation cycle, and dynamically modify the switching threshold boundary between the BUCK mode and the BOOST mode, so that the switching operation occurs in the stable interval of the hotspot trajectory.

[0072] Optionally, the step 105 of continuously updating the local minimum point of the switching cost function during the chip operation cycle and dynamically correcting the switching threshold boundary between the BUCK mode and the BOOST mode so that the switching operation occurs in the stable interval of the hotspot trajectory includes: 1051. During the chip operation cycle, continuously intercept hotspot migration path segments, calculate the sum of distance changes between adjacent hotspot positions within the migration path segments, input the sum of distance changes into the switching cost function to output a cost value, continuously record the cost values ​​for multiple cycles, and select cycles where the cost value is lower than both the previous cycle value and the next cycle value, to obtain the maximum offset range of coordinates in each direction during the period with the lowest cost value; 1052. Determine the boundary of the rectangular stable region according to the maximum offset range. When the output current continues to rise, if the predicted hotspot position exceeds the right boundary of the rectangular region, lower the current threshold trigger point for switching from buck to boost. When the output current continues to fall, if the predicted hotspot position exceeds the lower boundary of the rectangular region, raise the current threshold trigger point for switching from boost to buck. Finally, perform the mode switching operation at the adjusted threshold point.

[0073] In the above scheme, the stable interval refers to the period of time in the hotspot migration path with the minimum movement distance. The rectangular stable region refers to the rectangular boundary defined by the maximum coordinate offset range. The predicted hotspot location refers to the hotspot coordinates estimated for the next cycle based on the current change trend. Threshold correction dynamically adjusts the current value at the buck / boost switching trigger point.

[0074] In the embodiment of the present application, first, in step 1051, hotspot migration path segments are continuously intercepted during chip operation (e.g., each 10ms segment contains 3 path points), and the sum of the distance changes between adjacent hotspot positions is calculated for each segment:

[0075] in For the The three-dimensional coordinates of the path points, is the number of path points in the fragment, Input the cost value output by the switching cost function preset in step 1043, record multiple cycles continuously, and then select the time period that meets the local minimum condition (the cost value is lower than both the previous and next cycle values). For example, if the value of the second cycle is 19.3 < the previous value 21.5 and < the next value 22.1, it meets the condition; if the value of the fourth cycle is 20.0 > the next value 23.6, it does not meet the condition. Extract the maximum offset range of coordinates in each direction within the selected optimal period (such as the second cycle):

[0076] Finally, step 1052 first defines the rectangular stable region boundary based on the maximum offset range in step 1051 - set the current hotspot position as the center Direction range , Direction range , then monitor the current change trend and predict the hot spot location, that is, extrapolate when the current rises Coordinates, extrapolated when descending Coordinates, implement the key threshold correction rule: when the output current continues to rise. That is, it enters a high load trend and the predicted hotspot X position exceeds the right boundary, for example When the output current continues to decrease, it enters a low load trend and predicts hot spots. Position exceeds the lower boundary, e.g. When the current trigger threshold for switching from BOOST mode to BUCK mode is raised, for example, the original threshold of 0.8A is raised to 0.82A, and the switching is delayed to ensure thermal stability; finally, the mode switching operation is performed at the adjusted current threshold point, so that the switching time always falls in the stable range with the smallest fluctuation of the hot spot space, for example and .

[0077] For example, when the fast charging chip is running, the hotspot path fragment is intercepted: Specifically, t=50-60ms three-point sequence , calculate the sum of distances between adjacent points Input cost function (Cycle minimum value), extract coordinate offset range Based on this, a rectangular stable zone (X: [1.49, 1.51] mm, Z: [0.295, 0.305] mm) is defined. When a rising current trend is detected and the next hotspot (X = 1.52 mm) is predicted (beyond the right boundary), the buck-to-boost current threshold is dynamically lowered from 1.5 A to 1.48 A, bringing the hotspot back into the stable zone after switching (measured X = 1.50 mm).

[0078] This step accurately locates the thermodynamic stability window through the dual indicators of moving distance and cost value; actively adjusts the switching threshold based on the predicted position out-of-bounds warning; strictly constrains the mode switching to the extremely small range of hot spot spatial fluctuations, eliminates thermal stress shock during the switching period, and simultaneously improves chip life and output stability.

[0079] The following is a complete example of steps 101 to 105: A fast-charging device uses a PMIC-X power management chip (4mm×4mm×0.6mm). When the user plugs in the charger, the load current suddenly increases from 100mA to 500mA, triggering a multi-mode modulation optimization process. First, when the PMIC-X chip load on the fast-charging device suddenly changes (current 100mA → 500mA), the surface temperature is captured at 50 frames / second using a 4×4 infrared sensor array. The 60-frame thermal image is converted into a 4×4 temperature matrix sequence. For example, the matrix captured at t = 20ms is: And stored in chronological order.

[0080] Secondly, the temperature change rate is calculated based on the thermal map sequence and the thermal diffusion matrix is ​​constructed. , while locating internal hotspots through 6-layer 3D reconstruction, while locating internal hotspots through 6-layer 3D reconstruction:

[0081] The turning point is marked at t=40ms (vector angle 85°>70° threshold).

[0082] Next, the output voltage ripple (±50mV) is synchronously collected, and the impact component is separated within the turning point t=40ms±1ms window: the original ripple [-12mV, -8mV, -28mV, +15mV] is deducted to obtain a peak value of -24.8mV.

[0083] Then, the main feature of heat diffusion (the first singular value σ1 = 18.5) and the hotspot path are combined to construct a 10-dimensional tensor, which is then correlated with the shock amplitude-frequency feature [24.8 mV, 480 kHz] to calculate the switching cost:

[0084] Finally, identify the stable interval (t=80-100ms moving distance , demarcate the stable zone When the current rises to 480mA and the predicted hotspot X=2.53mm exceeds the limit, the BUCK→BOOST threshold is dynamically lowered from 500mA to 480mA to execute the switch, stabilizing the hotspot at 2.50mm (reducing the fluctuation range by 60%) and achieving improved thermal stress avoidance.

[0085] Figure 2 The present invention provides a schematic diagram of a multi-mode modulation switching threshold setting system for a power management chip. Figure 2 As shown, the system includes: An acquisition module 21 acquires a time domain heat map sequence of a surface area of ​​the power management chip when the load suddenly changes; a marking module 22 that generates a heat diffusion matrix based on the temperature change rate of the time-domain thermal map sequence, reconstructs the three-dimensional thermal distribution inside the chip through spatial coordinate mapping of the infrared sensor array, extracts the migration path of the temperature extreme points in the three-dimensional thermal distribution as a hotspot trajectory, and marks the timestamps of the turning points in the hotspot trajectory; A separation module 23 is configured to separate an impulse component sequence synchronized with a timestamp of the turning point from the output ripple voltage caused by a sudden load change; A construction module 24 performs singular value decomposition on the heat diffusion matrix to obtain matrix features, performs tensor fusion on the matrix features and the path of the hot spot trajectory, and associates the amplitude-frequency features of the impact component sequence to construct a switching cost function; The correction module 25 continuously updates the local minimum point of the switching cost function during the chip operation cycle, and dynamically corrects the switching threshold boundary between the BUCK mode and the BOOST mode, so that the switching operation occurs in the stable interval of the hotspot trajectory.

[0086] Figure 2 The power management chip multi-mode modulation switching threshold setting system can be executed Figure 1 The implementation principles and technical effects of the method for setting a multi-mode modulation switching threshold value for a power management chip described in the illustrated embodiment will not be elaborated upon. The specific manner in which each module and unit performs operations in the multi-mode modulation switching threshold setting system for a power management chip in the aforementioned embodiment has been described in detail in the embodiments of the method and will not be elaborated upon here.

[0087] In one possible design, Figure 2 The power management chip multi-mode modulation switching threshold setting system of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0088] The processing component 32 is used for the above Figure 1 The embodiment of the present invention provides a method for setting a multi-mode modulation switching threshold value for a power management chip.

[0089] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0090] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0091] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0092] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0093] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0094] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0095] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is a method for setting a multi-mode modulation switching threshold of a power management chip.

[0096] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0097] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0098] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for setting a multi-mode modulation switching threshold for a power management chip, characterized in that: include: Obtain a time domain heat map sequence of the surface area of ​​the power management chip when the load suddenly changes; Generate a heat diffusion matrix based on the temperature change rate of the time-domain heat map sequence, reconstruct the three-dimensional heat distribution inside the chip through spatial coordinate mapping of the infrared sensor array, extract the migration path of the temperature extreme points in the three-dimensional heat distribution as hot spot trajectories, and mark the timestamps of the turning points in the hot spot trajectories; Separating an impulse component sequence synchronized with a time stamp of a turning point from an output ripple voltage of a sudden load change; Performing singular value decomposition on the heat diffusion matrix to obtain matrix features, performing tensor fusion on the matrix features and the path of the hot spot trajectory, and correlating the amplitude-frequency features of the impact component sequence to construct a switching cost function; The local minimum point of the switching cost function is continuously updated during the chip operation cycle, and the switching threshold boundary between the BUCK mode and the BOOST mode is dynamically modified so that the switching operation occurs in the stable interval of the hot spot trajectory.

2. The method according to claim 1, characterized in that Continuously updating the local minimum point of the switching cost function during the chip operation cycle, dynamically correcting the switching threshold boundary between the BUCK mode and the BOOST mode, so that the switching operation occurs in the stable range of the hotspot trajectory, including: During the chip operation cycle, hotspot migration path segments are continuously intercepted, and the sum of distance changes between adjacent hotspot positions in the migration path segments is calculated. The sum of distance changes is input into the switching cost function to output a cost value. The cost values ​​of multiple cycles are continuously recorded, and the cycles with cost values ​​lower than both the previous cycle and the next cycle are selected to obtain the maximum offset range of coordinates in each direction during the period with the lowest cost value. The boundaries of the rectangular stable region are determined based on the maximum offset range. When the output current continues to rise, if the predicted hotspot position exceeds the right boundary of the rectangular region, the current threshold trigger point for switching from buck to boost is lowered. When the output current continues to fall, if the predicted hotspot position exceeds the lower boundary of the rectangular region, the current threshold trigger point for switching from boost to buck is raised. Finally, the mode switching operation is performed at the adjusted threshold point.

3. The method according to claim 1, characterized in that Performing singular value decomposition on the heat diffusion matrix to obtain matrix features, performing tensor fusion on the matrix features and the path of the hot spot trajectory, and correlating the amplitude-frequency features of the impact component sequence to construct a switching cost function, including: performing a matrix decomposition operation on the heat diffusion matrix to extract key features, combining the key features with the coordinate sequence of the hotspot migration path to form a fused data tensor, wherein the fused data tensor is constructed by splicing the key features as additional dimensions with the coordinate sequence of the hotspot migration path; Analyzing the amplitude characteristics and frequency characteristics of the impulse component sequence to obtain amplitude-frequency characteristics, wherein the amplitude characteristics are determined by calculating the peak voltage value of each impulse component, and the frequency characteristics are determined by calculating the dominant oscillation frequency of each impulse component; Based on the fused data tensor and the amplitude-frequency feature, a cost function is defined to evaluate the cost of mode switching of the power management chip, wherein the cost function is constructed by linearly combining the norm of the fused data tensor and the product of the amplitude-frequency feature, and the weight coefficient is preset to a fixed value.

4. The method according to claim 1, wherein A heat diffusion matrix is ​​generated based on the temperature change rate of the time-domain heat map sequence, and the three-dimensional heat distribution inside the chip is reconstructed through spatial coordinate mapping of the infrared sensor array. The migration path of the temperature extreme point in the three-dimensional heat distribution is extracted as a hot spot trajectory, and the time stamp of the turning point in the hot spot trajectory is marked, including: Calculating a temperature change rate sequence for each location point from the time domain heat map sequence, wherein the temperature change rate sequence is obtained by dividing the temperature value difference between adjacent time points by the time interval; Based on the temperature change rate sequence, a diffusion relationship matrix is ​​constructed, wherein the rows and columns of the diffusion relationship matrix correspond to positions on the chip surface, and the matrix element values ​​represent the intensity of heat transfer from one position point to another position point, and the intensity is determined by the ratio of the temperature change rate difference between the position points to the spatial distance; Using the spatial coordinate information of the temperature sensor array and spatial mapping technology, the surface temperature value is used as the boundary condition. Combined with the thermal conductivity characteristics of the chip material, the three-dimensional heat conduction equation is solved to infer the temperature values ​​of each internal layer. The three-dimensional temperature distribution at each time point is generated to reconstruct the three-dimensional temperature distribution sequence inside the power management chip. In the three-dimensional temperature distribution sequence, identifying the highest temperature position point at each time point, wherein the highest temperature position point is determined by comparing the temperature values ​​of all internal positions, and connecting the changes of the highest temperature position point over time to form a hotspot migration path, wherein the path is represented by a sequence of the highest temperature position coordinates of consecutive time points; The turning points of direction changes are detected in the hotspot migration path, wherein the turning points are determined by calculating the angle between the direction vectors of adjacent segments of the path. When the angle exceeds a preset threshold, it is determined to be a turning point, and the timestamps of these turning points are recorded.

5. The method according to claim 1, characterized in that Separating an impulse component sequence synchronized with a time stamp of a turning point from an output ripple voltage of a sudden load change, including: During a sudden load change event, monitoring an output voltage signal of a power management chip and extracting a ripple component from the output voltage signal, wherein the ripple component is obtained by removing a DC component; According to the timestamp of the turning point, a synchronized time window is located in the ripple component, wherein the time window is centered on the timestamp of each turning point and the width is preset to a fixed value. Within the time window, a transient impact component is separated, wherein the transient impact component is obtained by subtracting the average voltage value within the time window and extracting the remaining fluctuation part, forming an impact component sequence synchronized with the turning point timestamp.

6. The method according to claim 1, wherein Obtain a time domain heat map sequence of the surface area of ​​the power management chip when the load suddenly changes, including: During the operation of the power management chip, a sudden load event is applied, and a temperature sensor array is used to capture a temperature distribution image of the chip surface at preset time intervals during the sudden load event, wherein each sensing unit of the temperature sensor array corresponds to a specific area on the chip surface, and the capture process is performed at a fixed frame rate; Converting each of the temperature distribution images into a two-dimensional temperature value matrix, wherein each matrix element represents a temperature value detected by a corresponding sensing unit; All two-dimensional temperature value matrices are stored in chronological order to form a time domain heat map sequence, where each element in the sequence corresponds to a temperature distribution image at a time point.

7. The method according to claim 4, characterized in that By using the spatial coordinate information of the temperature sensor array and spatial mapping technology, the surface temperature value is used as the boundary condition. Combined with the thermal conductivity characteristics of the chip material, the three-dimensional heat conduction equation is solved to infer the temperature values ​​of each internal layer. The three-dimensional temperature distribution at each time point is generated to reconstruct the three-dimensional temperature distribution sequence inside the power management chip, including: Determine the position of the surface detection point according to the spatial coordinates of the temperature sensor array, divide the internal virtual layers into equal intervals along the thickness direction of the chip and generate grid nodes aligned with the surface detection points in each layer, pre-store the thermal conductivity coefficient value of the chip packaging material and set the thermal resistance value between adjacent grid nodes; Assign the surface detection point temperature value to the outermost grid node, establish a thermal balance equation for each inner grid node, and perform calculations based on the temperature value and thermal resistance value; Execute each layer of grid nodes in order from the surface to the inside of the chip, using the temperature of the nodes in the previous layer as the known quantity, and solve the heat balance equation of the nodes in the current layer until the innermost layer nodes are solved; The temperature values ​​of all grid nodes at each sampling moment are obtained to form a three-dimensional temperature distribution, and the three-dimensional temperature distributions of all sampling moments are continuously stored in chronological order to form a complete sequence.

8. A multi-mode modulation switching threshold setting system for a power management chip, characterized in that: include: An acquisition module obtains a time domain heat map sequence of the surface area of ​​the power management chip when the load suddenly changes; a marking module that generates a heat diffusion matrix based on the temperature change rate of the time-domain thermal map sequence, reconstructs the three-dimensional thermal distribution inside the chip through spatial coordinate mapping of the infrared sensor array, extracts the migration path of the temperature extreme points in the three-dimensional thermal distribution as a hotspot trajectory, and marks the timestamps of the turning points in the hotspot trajectory; A separation module is configured to separate an impulse component sequence synchronized with a time stamp of the turning point from an output ripple voltage caused by a sudden load change; A construction module is configured to perform singular value decomposition on the heat diffusion matrix to obtain matrix feature quantities, perform tensor fusion on the matrix feature quantities and the path of the hot spot trajectory, and associate the amplitude-frequency characteristics of the impact component sequence to construct a switching cost function; The correction module continuously updates the local minimum point of the switching cost function during the chip operation cycle, and dynamically corrects the switching threshold boundary between the BUCK mode and the BOOST mode so that the switching operation occurs in the stable interval of the hot spot trajectory.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the power management chip multi-mode modulation switching threshold setting method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the method for setting a multi-mode modulation switching threshold of a power management chip according to any one of claims 1 to 7 is implemented.

Images