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

Abstract

The application provides a power management chip multi-mode modulation switching threshold setting method and system. Wherein, the method obtains the time-domain thermal map sequence of the chip surface area when the load mutates, generates a thermal diffusion matrix based on the temperature change rate, and combines the infrared sensor array to reconstruct the three-dimensional thermal distribution, extracts the migration path of the temperature extreme point as the hot spot trajectory, and marks the turning point timestamp. At the same time, the impact component sequence synchronized with the timestamp is separated from the output ripple voltage, the singular value decomposition is performed on the thermal diffusion matrix to obtain the matrix characteristic quantity, which is fused with the hot spot trajectory path tensor, and the amplitude-frequency characteristics of the impact component are associated to construct a switching cost function. The local minimum point of the function is continuously updated within the chip running period, and the switching threshold boundary between the BUCK and BOOST modes is dynamically adjusted, and the stability and efficiency of the power management chip under load mutation are improved.

Description

Technical Field

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

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

[0003] The current mainstream solution employs a threshold dynamic adjustment technique based on ripple voltage feedback. This technique monitors changes in the amplitude of the output ripple voltage and dynamically corrects the switching threshold using a preset load-ripple mapping table. This solution introduces a digital PID controller to perform frequency domain decomposition of the ripple characteristics. When a high-frequency impulse component is detected, a threshold shift is triggered, thereby shortening the mode switching delay to the microsecond level.

[0004] This solution relies solely on electrical signal feedback and fails to incorporate the internal thermal diffusion characteristics of the chip. First, the local temperature rise caused by hot spot migration during sudden load changes will alter the conduction characteristics of the power device, and pure electrical signal control cannot predict the risk of thermal runaway. Second, ripple frequency domain analysis is insufficient in its ability to resolve multi-physics coupling effects (such as thermoelectric coupling phase shift), which can easily lead to erroneous switching under high-temperature conditions. Summary of the Invention

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

[0006] In a first aspect, this application provides a method for setting a multi-mode modulation switching threshold for a power management chip, including:

[0007] Obtain the time-domain thermal map sequence of the surface region of the power management chip under sudden load changes;

[0008] A thermal diffusion matrix is ​​generated based on the temperature change rate of the time-domain heat map sequence. At the same time, the three-dimensional thermal distribution inside the chip is reconstructed by 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 the hot spot trajectory. The timestamp of the turning point in the hot spot trajectory is marked.

[0009] Separate the impulse component sequence synchronized with the timestamp of the inflection point from the output ripple voltage of the load abrupt change;

[0010] Singular value decomposition is performed on the heat diffusion matrix to obtain matrix features. The matrix features are then fused with the path of the hotspot trajectory using tensors, and the amplitude-frequency features of the impact component sequence are associated to construct a switching cost function.

[0011] The local minimum points of the switching cost function are continuously updated during the chip's operating cycle, and the switching threshold boundary between BUCK mode and BOOST mode is dynamically corrected so that the switching operation occurs in the stable range of the hot spot trajectory.

[0012] Optionally, the local minimum points of the switching cost function are continuously updated during the chip's operating cycle to dynamically correct the switching threshold boundary between BUCK mode and BOOST mode, ensuring that the switching operation occurs within the stable range of the hotspot trajectory, including:

[0013] During the chip's operating cycle, hotspot migration path segments are continuously captured, and the total distance change of adjacent hotspot positions within the migration path segments is calculated. The total distance change is input into the switching cost function to output the cost value. The cost value of multiple cycles is continuously recorded, and cycles with cost values ​​lower than both the previous and next cycle values ​​are selected to extract the maximum offset range of coordinates in each direction during the period with the lowest cost value.

[0014] The boundary of the rectangular stable region is determined according to the maximum offset range. When the output current continues to rise, if the predicted hot spot position exceeds the right boundary of the rectangular region, the current threshold trigger point for BUCK to BOOST is lowered. When the output current continues to fall, if the predicted hot spot position exceeds the lower boundary of the rectangular region, the current threshold trigger point for BOOST to BUCK is raised. Finally, the mode switching operation is performed at the adjusted threshold point.

[0015] Optionally, singular value decomposition is performed on the heat diffusion matrix to obtain matrix features. These features are then tensor-fused with the path of the hotspot trajectory and associated with the amplitude-frequency features of the impact component sequence to construct a switching cost function, including:

[0016] Perform matrix decomposition 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. The fused data tensor is constructed by concatenating the key features as additional dimensions with the coordinate sequence of the hotspot migration path.

[0017] The amplitude and frequency characteristics of the impact component sequence are analyzed to obtain the amplitude-frequency characteristics. The amplitude characteristics are determined by calculating the peak voltage value of each impact component, and the frequency characteristics are determined by calculating the dominant oscillation frequency of each impact component.

[0018] Based on the fused data tensor and the amplitude-frequency features, a cost function is defined to evaluate the cost of power management chip mode switching. The cost function is constructed by linearly combining the norm of the fused data tensor with the amplitude-frequency features, and the weight coefficients are preset to fixed values.

[0019] Optionally, a heat diffusion matrix is ​​generated based on the temperature variability of the time-domain heat map sequence. Simultaneously, the three-dimensional heat distribution inside the chip is reconstructed through spatial coordinate mapping of the infrared sensor array. The migration paths of temperature extreme points in the three-dimensional heat distribution are extracted as hotspot trajectories, and the timestamps of turning points in the hotspot trajectories are marked, including:

[0020] From the time-domain heat map sequence, the temperature change rate sequence at each location point is calculated, wherein the temperature change rate sequence is obtained by dividing the temperature value difference between adjacent time points by the time interval;

[0021] Based on the temperature change rate sequence, a diffusion relationship matrix is ​​constructed, where the rows and columns of the diffusion relationship matrix correspond to the positions on the chip surface, and the matrix element values ​​represent the heat transfer intensity from one position to another. This intensity is determined by the ratio of the temperature change rate difference between the positions to the spatial distance.

[0022] By utilizing the spatial coordinate information of the temperature sensing array and using spatial mapping technology to take the surface temperature value as the boundary condition, combined with the thermal conductivity characteristics of the chip material, the three-dimensional thermal conductivity equation is solved to infer the temperature values ​​of each internal layer, generating the three-dimensional temperature distribution at each time point, so as to reconstruct the three-dimensional temperature distribution sequence inside the power management chip.

[0023] In the three-dimensional temperature distribution sequence, the highest temperature location point at each time point is identified, wherein the highest temperature location point is determined by comparing the temperature values ​​of all internal location points, and the changes of the highest temperature location point over time are connected to form a hotspot migration path, wherein the path is represented by a sequence of coordinates of the highest temperature location at consecutive time points;

[0024] In the hotspot migration path, turning points of directional change are detected. 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.

[0025] Optionally, separating the impulse component sequence synchronized with the timestamp of the inflection point from the output ripple voltage of the load abrupt change includes:

[0026] During a load surge event, the output voltage signal of the power management chip is monitored, and the ripple component in the output voltage signal is extracted, wherein the ripple component is obtained by removing the DC component;

[0027] Based on the timestamp of the inflection point, a synchronized time window is located in the ripple component, wherein the time window is centered on each inflection point timestamp and the width is preset to a fixed value. Within the time window, the 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 inflection point timestamp.

[0028] Optionally, the time-domain thermal map sequence of the surface region of the power management chip during a sudden load change is obtained, including:

[0029] During the operation of the power management chip, a load surge event is applied. During the occurrence of the load surge event, a temperature distribution image of the chip surface is captured at a preset time interval using a temperature sensing array. Each sensing unit of the temperature sensing array corresponds to a specific area on the chip surface, and the capture process is performed at a fixed frame rate.

[0030] Each of the temperature distribution images is converted into a two-dimensional temperature value matrix, where each matrix element represents the temperature value detected by the corresponding sensing unit.

[0031] 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 given time point.

[0032] Optionally, utilizing the spatial coordinate information of the temperature sensing array, the surface temperature value is used as a boundary condition through spatial mapping technology. Combined with the thermal conductivity characteristics of the chip material, a three-dimensional thermal conductivity equation is solved to infer the temperature values ​​of each internal layer, generating a three-dimensional temperature distribution at each time point to reconstruct the three-dimensional temperature distribution sequence inside the power management chip, including:

[0033] The position of the surface detection point is determined according to the spatial coordinates of the temperature sensing array. The internal virtual layers are divided at equal intervals along the chip thickness direction, and grid nodes aligned with the surface detection points are generated in each layer. The thermal conductivity coefficient of the chip packaging material is pre-stored, and the thermal resistance value between adjacent grid nodes is set.

[0034] The temperature value of the surface detection point is assigned to the outermost mesh node, a thermal balance equation is established for each internal mesh node, and calculations are performed based on the temperature value and thermal resistance value.

[0035] The operation is performed on each mesh node in order from the surface to the inside of the chip, taking the temperature of the solved nodes in the upper layer as a known quantity, and solving the thermal balance equation of the current layer node until the innermost layer node is solved.

[0036] The temperature values ​​of all grid nodes at each sampling time are obtained to form a three-dimensional temperature distribution, and the three-dimensional temperature distributions at all sampling times are stored continuously in chronological order to form a complete sequence.

[0037] Secondly, this application provides a multi-mode modulation switching threshold setting system for a power management chip, including:

[0038] The acquisition module acquires the time-domain thermal map sequence of the surface region of the power management chip during a sudden load change.

[0039] The marking module generates a heat diffusion matrix based on the temperature change rate of the time-domain heat map sequence. At the same time, it reconstructs the three-dimensional heat 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 heat distribution as hot spot trajectories, and marks the timestamps of the turning points in the hot spot trajectories.

[0040] The separation module separates the impulse component sequence synchronized with the timestamp of the turning point from the output ripple voltage of the load abrupt change;

[0041] The construction module performs singular value decomposition on the heat diffusion matrix to obtain matrix feature quantities, performs tensor fusion of the matrix feature quantities 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.

[0042] The correction module continuously updates the local minimum points of the switching cost function during the chip's operating cycle, dynamically correcting the switching threshold boundary between BUCK mode and BOOST mode, so that the switching operation occurs in the stable range of the hot spot trajectory.

[0043] Thirdly, embodiments of this application provide a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are to be invoked and executed by the processing component to implement a power management chip multi-mode modulation switching threshold setting method as described in the first aspect above.

[0044] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when executed by a computer, implements a power management chip multi-mode modulation switching threshold setting method as described in the first aspect.

[0045] In this embodiment, a time-domain thermal map sequence of the surface region of a power management chip during a load change is obtained. A thermal diffusion matrix is ​​generated based on the temperature change rate of the time-domain thermal map sequence. Simultaneously, the three-dimensional thermal distribution inside the chip is reconstructed through spatial coordinate mapping of an infrared sensor array, and the migration path of the temperature extreme points in the three-dimensional thermal distribution is extracted as a hot spot trajectory. 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 change. Singular value decomposition is performed on the thermal diffusion matrix to obtain matrix features. The matrix features are fused with the path of the hot spot trajectory using tensors, and the amplitude-frequency features of the impact component sequence are associated to construct a switching cost function. The local minimum points of the switching cost function are continuously updated during the chip's operating cycle, and the switching threshold boundary between BUCK mode and BOOST mode is dynamically corrected so that the switching operation occurs in the stable range of the hot spot trajectory.

[0046] This application has the following beneficial effects:

[0047] 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. Quantifying the internal thermal conduction characteristics of the chip allows for precise location of temperature extremes, providing spatial dimensionality information for hotspot migration analysis. Key nodes at risk of thermal runaway are identified, providing a time synchronization benchmark for subsequent electro-thermal synergistic analysis. Transient electrical signal features caused by load mutations are extracted, establishing a correlation with thermal behavior. Dimensionality reduction is used to extract key thermal features, which are then combined with hotspot paths to construct a multi-physics coupled representation, enhancing the robustness of mode switching decisions. Integrated thermo-electric features enable adaptive regulation, ensuring mode switching occurs within the optimal range of thermal / electrical stability, thus improving system reliability.

[0048] Furthermore, this application achieves efficient coupling of thermal and electrical features through matrix decomposition and tensor fusion, and combines amplitude-frequency characteristics to quantify the switching cost, so that the threshold optimization process considers both the spatiotemporal characteristics of thermal diffusion and the transient effects of voltage surges, thereby achieving more accurate mode switching decisions under complex operating conditions.

[0049] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1A flowchart of a method for setting a multi-mode modulation switching threshold for a power management chip provided in this application is shown;

[0052] Figure 2 This paper shows a schematic diagram of the structure of a power management chip multi-mode modulation switching threshold setting system provided in this application;

[0053] Figure 3 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0054] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0055] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0056] Current load surge management schemes for high-power-density power chips primarily rely on ripple voltage feedback mechanisms. While these mechanisms can achieve microsecond-level dynamic threshold adjustment, their purely electrical signal control mode has fundamental limitations: Firstly, the nonlinear coupling between the chip's internal thermal diffusion characteristics and electrical parameters is ignored, making it impossible to predict the risk of thermal runaway caused by hotspot migration under high-temperature conditions. Secondly, ripple frequency domain analysis lacks sufficient analytical capability for thermoelectric coupling phase shifts, resulting in a high false switching rate when the junction temperature exceeds 125°C. These deficiencies stem from the decoupling of the synergistic effects of thermal and electrical multi-physics fields in existing technologies, leading to a lack of global awareness of the chip's actual operating state in mode switching decisions, severely restricting system reliability under high-load scenarios.

[0057] To address the aforementioned issues, this invention proposes a dynamic optimization method for switching thresholds based on thermal-electric multimodal fusion. Its innovation lies in: capturing the spatiotemporal migration characteristics of hotspot trajectories through temporal thermal map sequences and infrared 3D thermal distribution reconstruction, and simultaneously correlating the impulse component sequence in the ripple voltage; then, utilizing the singular value decomposition characteristics of the thermal diffusion matrix and the fusion of the hotspot path tensor, constructing a switching cost function encompassing thermal conduction characteristics, temperature gradient distribution, and voltage impulse amplitude-frequency characteristics. This scheme overcomes the limitations of traditional single-parameter control, ensuring that mode switching strictly occurs within the stable range of the hotspot trajectory through collaborative modeling and dynamic optimization of the thermal-electric coupling field. Experiments show that this method can reduce the high-temperature false switching rate to below 1.5%, while shortening the response delay to load abrupt changes, fundamentally solving the reliability degradation problem caused by thermal-electric decoupling in existing technologies, and providing a smart control paradigm for high-power-density chips that combines dynamic response and thermal safety.

[0058] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0059] Figure 1 This application provides a flowchart of a method for setting a multi-mode modulation switching threshold for a power management chip, as illustrated in the embodiments of this application. Figure 1 As shown, the method includes:

[0060] 101. Obtain the time-domain thermal map sequence of the surface region of the power management chip during a sudden load change;

[0061] Optionally, obtaining the time-domain thermal map sequence of the surface region of the power management chip under sudden load changes in step 101 may specifically include:

[0062] 1011. During the operation of the power management chip, a load change event is applied, and a temperature distribution image of the chip surface is captured at a preset time interval during the occurrence of the load change event using a temperature sensing array, wherein each sensing unit of the temperature sensing array corresponds to a specific area of ​​the chip surface, and the capture process is performed at a fixed frame rate.

[0063] 1012. Convert each of the temperature distribution images into a two-dimensional temperature value matrix, where each matrix element represents the temperature value detected by the corresponding sensing unit;

[0064] 1013. Store all two-dimensional temperature value matrices 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 given time point.

[0065] In the above scheme, load mutation refers to a rapid change in the load (such as current or power consumption) of the power management chip during normal operation, characterized by a sudden increase or decrease. The surface region refers to the physical outer surface of the chip's top, used for temperature detection and recording. A time-domain thermal image sequence is a sequence of multiple thermal images arranged chronologically, where each image represents the surface temperature distribution of the chip at a specific moment, and the sequence as a whole is used to demonstrate the temperature change process over time.

[0066] In this embodiment, firstly, during the operation of the power management chip in step 1011, a load surge event is applied via an external device, such as a programmable load generator. For example, the load jumps from a low value (e.g., 10mA) to a high value (e.g., 200mA) within milliseconds. Simultaneously, a temperature sensing array, such as an infrared sensor grid, is activated to capture temperature distribution images of the chip surface at preset fixed time intervals, such as 10 milliseconds. Each image is captured by a sensing unit in the array, with each unit corresponding to a small area of ​​the chip surface. The entire capture process is performed at a constant frame rate, such as 50 frames per second, ensuring uniform and stable data acquisition. For example, in a test scenario, when a load surge occurs, the temperature sensing array starts working, capturing 50 frames per second. Each frame shows the hotspot changes on the chip surface, such as the transition process as the temperature increases.

[0067] Secondly, in step 1012, for each temperature distribution image obtained in step 1011, a software image analysis tool, such as a simple matrix transformation algorithm based on OpenCV, is used to convert it into a two-dimensional temperature value matrix. The conversion process involves reading the pixel value or direct sensor reading corresponding to each sensing unit in the image and filling it into the row and column structure of the matrix. The row and column numbers of the matrix 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 the 2nd row and 3rd column) stores the temperature value (e.g., 25°C) of its corresponding sensing unit. High-temperature red areas in the image are represented by higher-value elements in the matrix.

[0068] Finally, in step 1013, all the two-dimensional temperature value matrices output in step 1012 are organized into a serialized data structure in chronological order, such as a Python list or a timestamp array in a database. The capture time point of each matrix is ​​recorded, and then these 100 points are stored in a sequence file in chronological order, for example, from t=0 to t=1 seconds. This sequence forms a complete time-domain heatmap sequence, where each element corresponds to a temperature distribution matrix at a given time point. For example, the 10 time-point matrices captured in the test (intervals of 1 milliseconds) are stored in a sequence with timestamps, and sequence indices 1 to 10 correspond to the temperature data changes at different time points.

[0069] For example, in a general test scenario:

[0070] Specifically, Test Center A conducted a load surge experiment on a Type B power management chip. First, a programmable load control device was used to apply a load surge event, instantly adjusting the load from 50mA to 300mA. Simultaneously, a 5x5 temperature sensor array (such as a thermal imaging camera) was configured to capture images at 15-millisecond intervals, acquiring 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 map sequence. This process simulated the chip's thermal response in real-world applications, ensuring data integrity and time consistency.

[0071] This step efficiently acquires the surface temperature change data sequence of the power management chip under sudden load changes, providing a basic input for subsequent analysis (such as temperature pattern recognition). Benefits include simplified data acquisition, high temporal resolution recording, and standardized data format for easier processing in subsequent steps.

[0072] 102. A heat diffusion matrix is ​​generated based on the temperature change rate of the time-domain heat map sequence. At the same time, the three-dimensional heat distribution inside the chip is reconstructed by 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 the hot spot trajectory, and the timestamp of the turning point in the hot spot trajectory is marked.

[0073] Optionally, in step 102, a heat diffusion matrix is ​​generated based on the temperature variation rate of the time-domain heat map sequence. Simultaneously, the three-dimensional heat distribution inside the chip is reconstructed through spatial coordinate mapping of the infrared sensor array, and the migration paths of temperature extreme points in the three-dimensional heat distribution are extracted as hotspot trajectories. The timestamps marking the turning points in the hotspot trajectories may specifically include:

[0074] 1021. From the time-domain heat map sequence, calculate the temperature change rate sequence for each location point, wherein the temperature change rate sequence is obtained by dividing the temperature value difference between adjacent time points by the time interval;

[0075] 1022. 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 the positions on the chip surface, and the matrix element values ​​represent the heat transfer intensity from one position point to another, which is determined by the ratio of the temperature change rate difference between the positions to the spatial distance.

[0076] 1023. Using the spatial coordinate information of the temperature sensing array, the surface temperature value is used as the boundary condition through spatial mapping technology. Combined with the thermal conductivity characteristics of the chip material, the three-dimensional thermal conductivity equation is solved to infer the temperature values ​​of each internal layer and generate the three-dimensional temperature distribution at each time point to reconstruct the three-dimensional temperature distribution sequence inside the power management chip.

[0077] The process of step 1023, "using the spatial coordinate information of the temperature sensing array, using spatial mapping technology to take the surface temperature value as the boundary condition, and combining the thermal conductivity characteristics of the chip material to solve the three-dimensional thermal conductivity equation to infer the temperature values ​​of each internal layer, generating the three-dimensional temperature distribution at each time point, so as to reconstruct the 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 sensing array; dividing the internal virtual layers at equal intervals along the chip thickness direction 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 calculating it according to the temperature value and thermal resistance value; performing the operation on each layer of grid nodes in the order from the surface to the inside of the chip, taking the temperature of the solved node in the upper layer as a known quantity, and solving the thermal balance equation of the current layer node until the innermost 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 of all sampling times in chronological order to form a complete sequence.

[0078] 1024. In the three-dimensional temperature distribution sequence, identify the highest temperature location point at each time point, wherein the highest temperature location point is determined by comparing the temperature values ​​of all internal location points, and connect the changes of the highest temperature location point over time to form a hotspot migration path, wherein the path is represented by a sequence of coordinates of the highest temperature location at consecutive time points;

[0079] 1025. Detect turning points of directional change in the hotspot migration path, wherein the turning point is 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.

[0080] In the above scheme, the temperature change rate refers to the rate of temperature change, i.e., the temperature difference per unit time, reflecting how quickly a certain area heats up / cools down. The heat diffusion matrix is ​​a table describing 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. Three-dimensional thermal distribution refers to the three-dimensional temperature distribution of each layer inside the chip inferred from the surface temperature. Hot spot trajectory refers to the path of the hottest area inside the chip over time, i.e., recording the displacement path of the highest temperature point in three-dimensional space.

[0081] In this embodiment of the application, firstly, step 1021 is used to process the time-domain heatmap sequence obtained in step 101, that is, a set of heatmaps ordered at fixed time intervals. Acquired two-dimensional temperature matrix In the middle, for each sensor location point Iterate through all adjacent time points, i.e. Perform the following calculations:

[0082]

[0083] in, Location point At two consecutive time points temperature difference, A fixed sampling time interval is used. Based on the calculation results, an independent temperature change rate sequence is generated for each location point.

[0084] .

[0085] Secondly, in step 1022, the row and column dimensions of the matrix are mapped to all sensor location points (e.g., 16×16 matrix is ​​formed by 16 points), where the row index represents the heat source location and the column index represents the heat receiving target location. Then, for each element in the matrix, the heat transfer intensity from the source point to the target point is calculated through three steps: obtain the instantaneous temperature change rate of the source point (e.g., 300℃ / second) and the speed of the target point (e.g., 200℃ / second), calculate the absolute value of the difference between the two (|300-200|=100), and combine it with the pre-stored physical coordinates (e.g., the two points are 2 mm apart) to obtain the transfer intensity value by the ratio of the speed difference to the spatial distance.

[0086] Next, based on the spatial coordinates of the temperature sensing array (e.g., the 64×64 detection point positions on the surface) in step 1023, internal virtual layers are divided at equal intervals along the chip thickness direction, for example, 10 layers. Each layer generates grid nodes that are strictly aligned with the surface detection points, forming a 64×64×10 three-dimensional grid system. The thermal conductivity coefficient of the chip packaging material is pre-stored, for example, 150 W / m·K for silicon. The thermal resistance between adjacent nodes is calculated based on the material properties and grid spacing. The thermal resistance calculation satisfies the formula... ,in The node spacing, Thermal conductivity, The effective heat transfer area is determined. Then, the measured temperature value of each surface detection point at each sampling time, for example, 47.2℃ at position (2,3), is directly assigned to the outermost mesh node, i.e., the first layer node (2,3,1). Subsequently, the internal temperature is solved layer by layer, starting from the second layer and progressing inwards, for each internal mesh node. Establish the heat balance equation:

[0087]

[0088] in The temperatures of the six spatially adjacent nodes of this node are... Let be the thermal resistance from the current node to its nth neighbor. For each time point, such as 0ms, 5ms, 10ms after a sudden load change, repeat the above mesh construction → boundary loading → layered solution process to obtain the set of temperature values ​​of all mesh nodes at that moment. Store all three-dimensional temperature distributions continuously in chronological order to form a complete dynamic thermal field evolution sequence.

[0089] Then, in step 1024, at each time point, for example, at t=10ms, the temperature values ​​of all grid nodes are scanned in the three-dimensional temperature distribution. For example, there are 45 locations across 5 layers. The highest temperature point is accurately located through full data comparison, for example, the center point of the 3rd layer at 121℃, and its three-dimensional coordinates (x=2.5mm, y=2.5mm, z=0.1mm) are recorded. Next, the coordinates of all the highest temperature points are connected in time order, for example, 0ms→5ms→10ms… to form a hotspot migration path composed of continuous spatiotemporal locations, 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, and the timestamp, coordinates, and temperature value of the point are stored synchronously. For example, at t=50ms, the coordinates (1.5,0.5,0.3) have a temperature of 118℃, and the layer depth change is marked. For example, the z value increases from 0.1mm to 0.3mm, indicating that the hotspot sinks.

[0090] Finally, based on the three-dimensional temperature distribution sequence in step 1025, the spatial coordinates of the hotspots, i.e. the locations of the highest temperatures, are extracted in chronological order at each sampling moment to form a pathpoint sequence:

[0091]

[0092] Then for each path point Calculate the forward vector formed by the point and the previous point. Calculate the backward vector formed by the next point. Then for each intermediate point ,calculate The angle between the adjacent vectors of the included angle :

[0093]

[0094] like Then mark Record the timestamp as a turning point. Finally, a set of timestamps for all turning points is generated. Based on the corresponding coordinates, complete the path mutation analysis.

[0095] For example, when a power management chip (5mm×5mm) experiences a sudden load change, a 4×4 temperature sensor array is used to capture the surface temperature sequence.

[0096] Specifically, firstly, the temperature change rate at each point was calculated (reaching a maximum of 520°C / s in the central region), and a diffusion matrix was constructed, revealing that the heat transfer from the left region to the center was strongest (matrix element value 450). Secondly, 3D reconstruction revealed that the initial hotspot reached 121°C at a depth of 0.1 mm below the surface, and later decreased to 112°C when it sank to a depth of 0.3 mm. Based on this, the extracted hotspot migration path showed that the highest temperature region shifted from the center to the lower right. When a sudden change in path direction was detected at t=15 ms (the angle between adjacent movement vectors was 110°), this time point was marked as the turning point in heat dissipation. The entire process provided key thermal dynamic characteristics for multi-mode modulation switching (such as avoiding hotspots passing through sensitive areas).

[0097] This step constructs a thermal diffusion matrix based on the temperature change rate generated from the time-domain thermal map sequence, intuitively presenting the main path of heat transfer; at the same time, through infrared sensor coordinate mapping and three-dimensional thermal conduction calculation, the surface temperature is used as a boundary condition, combined with the material thermal resistance parameters, to reconstruct the three-dimensional temperature distribution inside the chip, effectively breaking through the limitations of surface monitoring; by extracting the migration path of three-dimensional temperature extreme points, a complete hot spot trajectory is formed; finally, based on the detection of abrupt changes in path direction, the turning point of heat dissipation state is accurately identified and marked.

[0098] 103. Separate the impulse component sequence synchronized with the timestamp of the turning point from the output ripple voltage of the load change;

[0099] Optionally, separating the impulse component sequence synchronized with the timestamp of the inflection point from the output ripple voltage of the load abrupt change may specifically include:

[0100] 1031. During a load surge event, monitor the output voltage signal of the power management chip and extract the ripple component from the output voltage signal, wherein the ripple component is obtained by removing the DC component;

[0101] 1032. Based on the timestamp of the inflection point, locate the synchronous time window in the ripple component, wherein the time window is centered on each inflection point timestamp and the width is preset to a fixed value. Within the time window, separate the transient impact component, 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 inflection point timestamp.

[0102] In the above scheme, the output ripple voltage refers to the minute fluctuations in the output voltage of the power management chip, i.e., the high-frequency oscillation component superimposed on the stable voltage due to sudden load changes. The inflection point timestamp refers to the time marker when the movement direction of the highest temperature region inside the power management chip changes significantly during the load change process. The impulse component sequence refers to the voltage change signal segment that is strictly synchronized with the inflection point, reflecting the instantaneous impact of thermal dynamic changes on the circuit output.

[0103] In this embodiment, firstly, during the duration of the load surge event (e.g., 0-50ms), a high-precision voltage sensor is used to acquire the raw output voltage signal of the power management chip. Then, a sliding window mean filtering technique is employed to separate the ripple component: a fixed time window is first set, with a width of 0.5ms corresponding to 500 sampling points. The arithmetic mean of all voltage values ​​within the window is calculated as the DC component for that time period. Then, the raw voltage value of each sampling point within the window is subtracted from this DC component, and the difference is the ripple voltage value at the corresponding moment, retaining only the high-frequency components. Fluctuation component; This process gradually covers the entire period of load change by sliding the window. That is, after the previous window ends, the sampling point is immediately shifted to the right to process the new window. Finally, the continuous ripple waveform with the same length as the original signal is output. For example, if the original voltage sampling value is [3.301V, 3.298V, 3.305V], after calculating the DC component of 3.301V at three adjacent points at t=10ms through the window, the ripple component [0.000V, -0.003V, +0.004V] is obtained. Its magnitude is usually in the millivolt range (±30mV), which accurately reflects the power supply noise caused by load change.

[0104] Finally, after acquiring the ripple voltage signal in step 1032, a precise synchronization analysis is performed based on the inflection point timestamps: First, a time window is located centered on each inflection point, for example, the window width is preset to a fixed value such as ±1ms, which covers... Within the specified range, the window is strictly aligned with the moment of thermal event abrupt change. Transient impact separation is then performed within the window: first, the arithmetic mean of all ripple voltage points within the window is calculated (representing the background baseline for that period); then, the ripple voltage at each sampling point is subtracted from this mean, and the resulting residual signal is the pure transient impact component, retaining only the voltage abrupt change characteristics triggered by the thermal event. Finally, the impact component value corresponding to the center moment of the window is extracted and encapsulated into an impact component sequence in timestamp order, for example, a sequence obtained from three inflection points:

[0105]

[0106]

[0107]

[0108] For example, in a charging startup load surge event, the power chip of a certain mobile device:

[0109] Specifically, a sudden change in the hotspot migration path direction was detected at 15.2ms, i.e., the movement vector of the previous stage. and the next stage The included angle reaches 85°, and the voltage ripple window is simultaneously positioned centered on this timestamp. Extract the original ripple signal Subtract window mean The synchronous impact component of -21.8mV (corresponding to 15.2ms) was separated, forming direct evidence of the correlation between "thermal inflection point → strong negative voltage pulse".

[0110] This step achieves precise spatiotemporal matching between thermal dynamic events (direction change) and circuit transient response (voltage impulse). By filtering out background noise (window mean cancellation) and purifying the signal (impulse component separation), a voltage impulse feature library indexed by thermal inflection points is constructed (e.g., an 85° direction change must be accompanied by a ≥20mV negative pulse), providing high-value input for the thermal-electric coupling analysis model.

[0111] 104. Perform singular value decomposition on the heat diffusion matrix to obtain matrix features, perform tensor fusion of the matrix features with the path of the hot spot trajectory, and associate the amplitude-frequency features of the impact component sequence to construct a switching cost function.

[0112] Optionally, step 104, which involves performing singular value decomposition on the heat diffusion matrix to obtain matrix features, performing tensor fusion of the matrix features with the path of the hotspot trajectory, and associating the amplitude-frequency features of the impact component sequence to construct a switching cost function, may specifically include:

[0113] 1041. Perform matrix decomposition 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 concatenating the key features as additional dimensions with the coordinate sequence of the hotspot migration path.

[0114] 1042. Analyze the amplitude and frequency characteristics of the impact component sequence to obtain the amplitude-frequency characteristics, wherein the amplitude characteristics are determined by calculating the peak voltage value of each impact component, and the frequency characteristics are determined by calculating the dominant oscillation frequency of each impact component.

[0115] 1043. Based on the fused data tensor and the amplitude-frequency feature, a cost function is defined to evaluate the cost of power management chip mode switching. 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 coefficients are preset to fixed values.

[0116] In the above scheme, the matrix characteristic quantity refers to the principal components (such as the first singular value) extracted from the heat diffusion matrix through singular value decomposition (SVD), reflecting the core mode intensity of heat transfer in the chip. The hotspot trajectory path refers to the complete spatiotemporal record of the dynamic migration of the core heat-generating area of ​​the power management chip during load surges. It consists of the locations of the highest internal temperatures captured at continuous time points (e.g., every 5ms), with each location containing three-dimensional data accurate to the millimeter level. The amplitude-frequency characteristic refers to the peak voltage of the impact component and its dominant oscillation frequency, characterizing the transient response of the circuit. The switching cost function is a mathematical expression that quantifies the comprehensive risk of mode switching; a larger value indicates a higher switching cost.

[0117] In this embodiment of the application, the thermal diffusion matrix is ​​first extracted through singular value decomposition (SVD) in step 1041. That is, the key feature quantities of the row and column corresponding to the chip surface location points, and fused with the hotspot migration path coordinate sequence to construct a tensor: first, matrix factorization is performed. ,in It is a left singular vector matrix. It is a singular value diagonal matrix. Transpose the right singular vector matrix and extract the first singular value. ,Right now The largest diagonal element characterizes the principal intensity of thermal diffusivity and the first left singular vector. ,Right now The first column represents the dominant heat transfer direction, for example, from a 4×4 matrix decomposition. Secondly, the hotspot path coordinate sequence, i.e., containing three-dimensional location... and timestamp Convert to a matrix For example, 3 waypoints:

[0118]

[0119] Finally, a fusion tensor is constructed by splicing and expanding the feature quantities. ,in It is a column vector of all 1s. For vector tiling operations, generate dimensions ,in for Dimension, for example, results in a 9-column tensor:

[0120]

[0121] To achieve cross-dimensional integration of thermodynamic properties and spatiotemporal paths.

[0122] Secondly, the amplitude-frequency characteristics of the impact component synchronized with the thermal inflection point are extracted in step 1042, and the impact component sequence within each time window is analyzed. Calculate the absolute peak voltage ,in For the first The impulse voltage value at each sampling point The number of sampling points in the window is determined, and then the dominant oscillation frequency is calculated using Fast Fourier Transform (FFT) and spectral peak detection.

[0123]

[0124] Take the frequency corresponding to the maximum value of the amplitude spectrum:

[0125] in, Sampling rate, The Nyquist frequency is used as the final output, and the amplitude-frequency eigenvectors at each inflection point are then defined. All points constitute the feature matrix The size is The number of turning points, where To quantify impact intensity, To reveal the oscillation characteristics.

[0126] Finally, the fused data tensor is calculated in step 1043. ,size Includes hotspot path coordinates and heat diffusion characteristics Norm:

[0127]

[0128] in, For tensor elements, The number of path points. The feature dimension is used as the feature dimension, and then the amplitude-frequency feature matrix is ​​calculated. The size is Energy weighted sum:

[0129]

[0130] in, For peak impulse voltage, As the dominant frequency, As a frequency-weighted term, it is ultimately determined by preset weighting coefficients. (heat risk weight) and Construct a linear combination cost function using (electric risk weights):

[0131]

[0132] Output value The cost of switching is directly quantified; the higher the value, the higher the combined thermal and electrical risks caused by the switching.

[0133] For example, when a server power chip experiences a sudden load change, three key points are extracted from the hotspot trajectory path:

[0134] Specifically, at t=10ms, the coordinates are (1.0, 1.0, 0.1) (center region of the surface); at t=30ms, the coordinates are (1.5, 0.8, 0.1) (shifted 1.5mm to the right); and at t=50ms, the coordinates are (1.8, 0.5, 0.3) (sunk to a depth of 0.3mm). Based on this path and the characteristics of the thermal diffusion matrix (singular value σ1=15.3), a 9-dimensional fusion tensor (path coordinates + singular value + singular vector) is constructed, while simultaneously correlating the voltage impact characteristics (peak value) at the three inflection points. ,frequency The final synthesis cost function is:

[0135] This step breaks through the limitations of single-dimensional analysis, realizing three-dimensional collaborative modeling of thermodynamic path (spatial migration), thermal conduction characteristics (matrix features), and electrical response (impact amplitude and frequency). Through tensor fusion and function weighting, heterogeneous data is transformed into a unified thermo-electric comprehensive risk assessment index (such as switching suppression triggered by a cost value > 25), providing a cross-domain joint optimization basis for dynamic modulation decision-making.

[0136] 105. During the chip's operating cycle, continuously update the local minimum points of the switching cost function, dynamically correct the switching threshold boundary between BUCK mode and BOOST mode, and ensure that the switching operation occurs in the stable range of the hot spot trajectory.

[0137] Optionally, step 105, which involves continuously updating the local minimum points of the switching cost function during the chip's operating cycle and dynamically correcting the switching threshold boundary between BUCK mode and BOOST mode to ensure that the switching operation occurs within the stable range of the hotspot trajectory, includes:

[0138] 1051. During the chip's operating cycle, continuously capture hotspot migration path segments and calculate the total distance change of adjacent hotspot positions within the migration path segments. Input the total distance change into the switching cost function to output the cost value. Continuously record the cost value for multiple cycles and filter out cycles where the cost value is lower than both the previous and next cycle values ​​to extract the maximum offset range of coordinates in each direction during the period with the lowest cost value.

[0139] 1052. Determine the boundary of the rectangular stable region according to the maximum offset range. If the predicted hot spot position exceeds the right boundary of the rectangular region when the output current continues to rise, lower the current threshold trigger point for BUCK to BOOST. If the predicted hot spot position exceeds the lower boundary of the rectangular region when the output current continues to fall, raise the current threshold trigger point for BOOST to BUCK. Finally, perform the mode switching operation at the adjusted threshold point.

[0140] In the above scheme, the stable interval refers to the period with the smallest movement distance in the hotspot migration path. 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 for the next cycle estimated based on the current change trend. Threshold correction is the dynamic adjustment of the current value at the BUCK / BOOST switching trigger point.

[0141] In this embodiment of the application, firstly, in step 1051, hotspot migration path segments are continuously captured during chip operation (e.g., each 10ms segment contains 3 path points), and the sum of distance changes between adjacent hotspot positions is calculated for each segment:

[0142]

[0143] in For the first The three-dimensional coordinates of the path points The number of path points within the segment, Input the preset switching cost function in step 1043 to output the cost value, record multiple cycles continuously, and then filter the time period that meets the local minimum condition (the cost value is lower than the previous and subsequent cycle values ​​at the same time): for example, the value of the second cycle is 19.3 < the previous value 21.5 and < the subsequent value 22.1 → meets the condition, and the value of the fourth cycle is 20.0 > the subsequent value 23.6 → does not meet the condition;

[0144] Extract the maximum offset range of coordinates in each direction within the selected optimal time period (e.g., the second cycle):

[0145]

[0146] Finally, in step 1052, the boundary of the rectangular stable region is first defined based on the maximum offset range in step 1051—centered on the current hotspot location. Direction range , Direction range Then, monitor the current change trend and predict the hot spot location, i.e., extrapolate when the current rises. Coordinates, extrapolated during descent Coordinates, implementing key threshold correction rules: when the output current continues to rise, i.e., entering a high load trend and the predicted hotspot X position exceeds the right boundary, for example... When switching from BUCK mode to BOOST mode, the current trigger threshold is proactively lowered, for example, from the original threshold of 1.2A to 1.18A, to avoid thermal risks by switching in advance; when the output current continues to decrease, indicating a low load trend and predicting hot spots... The position exceeds the lower boundary, for example When switching from BOOST mode to BUCK mode, the current trigger threshold is proactively increased, for example, from the original threshold of 0.8A to 0.82A, delaying the switch to ensure thermal stability. Finally, the mode switching operation is performed at the adjusted current threshold, ensuring that the switching time always falls within the stable range with minimal hotspot fluctuations. and .

[0147] For example, during the operation of a fast charging chip, hotspot path segments can be captured:

[0148] Specifically, the three-point sequence at t=50-60ms Calculate the sum of distances between adjacent points. Input cost function (Minimum value of the period), extract the coordinate offset range Based on this, the rectangular stable region is defined as X: [1.49, 1.51] mm and Z: [0.295, 0.305] mm. When an upward trend in current is detected and the next hot spot is predicted to be X=1.52 mm (exceeding the right boundary), the BUCK to BOOST current threshold is dynamically reduced from 1.5A to 1.48A, so that the hot spot after switching falls back into the stable region (actually measured X=1.50 mm).

[0149] This step precisely locates the thermodynamic stability window using both movement distance and cost; proactively adjusts the switching threshold based on the predicted location out-of-bounds warning; and strictly constrains mode switching within the range of minimal hotspot spatial fluctuations, eliminating thermal stress shocks during switching and simultaneously improving chip lifespan and output stability.

[0150] The following is a complete embodiment of steps 101-105:

[0151] A fast charging device uses a PMIC-X power management chip (4mm×4mm×0.6mm in size). 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 load on the PMIC-X chip in 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 frames of thermal images are then converted into a 4×4 temperature matrix sequence. For example, the matrix is ​​captured at t=20ms. And store them in chronological order.

[0152] Secondly, the temperature change rate is calculated based on the heat map sequence, and a heat diffusion matrix is ​​constructed. Simultaneously, internal hotspots are located through 6-layer 3D reconstruction.

[0153]

[0154] And mark the turning point at t=40ms (the threshold of vector angle 85°>70°).

[0155] Next, the output voltage ripple (±50mV) was synchronously acquired, and the impulse component was separated within the window of the inflection point t=40ms±1ms: the original ripple [-12mV, -8mV, -28mV, +15mV] was reduced to a mean value of -24.8mV peak value.

[0156] Then, a 10-dimensional tensor is constructed by fusing the main thermal diffusion feature (first singularity σ1 = 18.5) with the hotspot path, and correlated with the impact amplitude-frequency feature [24.8mV, 480kHz] to calculate the switching cost:

[0157] Finally, identify the stationary interval (distance moved between t=80-100ms). Delineate stable zones When the current rises to 480mA and the predicted hot spot X=2.53mm exceeds the limit, the BUCK→BOOST threshold is dynamically reduced from 500mA to 480mA to perform the switching, so that the hot spot is stabilized at 2.50mm (the fluctuation range is reduced by 60%), thereby improving thermal stress avoidance.

[0158] Figure 2 This application provides a schematic diagram of a power management chip multi-mode modulation switching threshold setting system, as shown in the embodiment of the present application. Figure 2 As shown, the system includes:

[0159] Module 21 acquires the time-domain heat map sequence of the surface region of the power management chip during a sudden load change;

[0160] The marking module 22 generates a heat diffusion matrix based on the temperature change rate of the time-domain heat map sequence. At the same time, it reconstructs the three-dimensional heat distribution inside the chip through spatial coordinate mapping of the infrared sensor array, extracts the migration path of the temperature extreme point in the three-dimensional heat distribution as the hot spot trajectory, and marks the timestamp of the turning point in the hot spot trajectory.

[0161] Separation module 23 separates the impulse component sequence synchronized with the timestamp of the turning point from the output ripple voltage of the load change;

[0162] The construction module 24 performs singular value decomposition on the heat diffusion matrix to obtain matrix feature quantities, performs tensor fusion of the matrix feature quantities 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.

[0163] The correction module 25 continuously updates the local minimum points of the switching cost function during the chip's operating cycle, dynamically corrects the switching threshold boundary between BUCK mode and BOOST mode, and ensures that the switching operation occurs in the stable range of the hot spot trajectory.

[0164] Figure 2 The power management chip multi-mode modulation switching threshold setting system can perform... Figure 1 The implementation principle and technical effects of the power management chip multi-mode modulation switching threshold setting method described in the illustrated embodiment will not be repeated here. The specific operation methods of each module and unit in the power management chip multi-mode modulation switching threshold setting system in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0165] 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;

[0166] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0167] The processing component 32 is used for the above Figure 1 The embodiment describes a method for setting the multi-mode modulation switching threshold for a power management chip.

[0168] 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-described method. Alternatively, the processing component may 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-described method.

[0169] 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 storage 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 storage, flash memory, magnetic disk, or optical disk.

[0170] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0171] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0172] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0173] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0174] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown illustrates a method for setting the multi-mode modulation switching threshold of a power management chip.

[0175] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0176] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0177] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part 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, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for setting a multi-mode modulation switching threshold for a power management chip, characterized in that, include: Obtain the time-domain thermal map sequence of the surface region of the power management chip under sudden load changes; A thermal diffusion matrix is ​​generated based on the temperature change rate of the time-domain heat map sequence. At the same time, the three-dimensional thermal distribution inside the chip is reconstructed by 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 the hot spot trajectory. The timestamp of the turning point in the hot spot trajectory is marked. Separate the impulse component sequence synchronized with the timestamp of the inflection point from the output ripple voltage of the load abrupt change; Singular value decomposition is performed on the heat diffusion matrix to obtain matrix features. The matrix features are then fused with the path of the hotspot trajectory using tensors, and the amplitude-frequency features of the impact component sequence are associated to construct a switching cost function. The local minimum points of the switching cost function are continuously updated during the chip's operating cycle, and the switching threshold boundary between BUCK mode and BOOST mode is dynamically corrected so that the switching operation occurs in the stable range of the hot spot trajectory. Specifically, continuously updating the local minimum points of the switching cost function during the chip's operating cycle, and dynamically correcting the switching threshold boundary between BUCK mode and BOOST mode, ensures that the switching operation occurs within the stable range of the hotspot trajectory, including: During the chip's operating cycle, hotspot migration path segments are continuously captured, and the total distance change of adjacent hotspot positions within the migration path segments is calculated. The total distance change is input into the switching cost function to output the cost value. The cost value of multiple cycles is continuously recorded, and cycles with cost values ​​lower than both the previous and next cycle values ​​are selected to extract the maximum offset range of coordinates in each direction during the period with the lowest cost value. The boundary of the rectangular stable region is determined according to the maximum offset range. When the output current continues to rise, if the predicted hot spot position exceeds the right boundary of the rectangular region, the current threshold trigger point for BUCK to BOOST is lowered. When the output current continues to fall, if the predicted hot spot position exceeds the lower boundary of the rectangular region, the current threshold trigger point for BOOST to BUCK is raised. Finally, the mode switching operation is performed at the adjusted threshold point.

2. The method according to claim 1, characterized in that, Singular value decomposition is performed on the heat diffusion matrix to obtain matrix features. These features are then tensor-fused with the paths of the hotspot trajectories and associated with the amplitude-frequency features of the impact component sequence to construct a switching cost function, including: Perform matrix decomposition on the heat diffusion matrix to extract key features, and combine the key features with the coordinate sequence of the hotspot trajectory path to form a fused data tensor. The fused data tensor is constructed by concatenating the key features as additional dimensions with the coordinate sequence of the hotspot trajectory path. The amplitude and frequency characteristics of the impact component sequence are analyzed to obtain the amplitude-frequency characteristics. The amplitude characteristics are determined by calculating the peak voltage value of each impact component, and the frequency characteristics are determined by calculating the dominant oscillation frequency of each impact component. Based on the fused data tensor and the amplitude-frequency features, a cost function is defined to evaluate the cost of power management chip mode switching. The cost function is constructed by linearly combining the norm of the fused data tensor with the amplitude-frequency features, and the weight coefficients are preset to fixed values.

3. The method according to claim 1, characterized in that, A heat diffusion matrix is ​​generated based on the temperature variability of the time-domain heat map sequence. Simultaneously, the three-dimensional heat distribution inside the chip is reconstructed through spatial coordinate mapping of the infrared sensor array. The migration paths of temperature extreme points in the three-dimensional heat distribution are extracted as hotspot trajectories, and the timestamps of turning points in the hotspot trajectories are marked, including: From the time-domain heat map sequence, the temperature change rate sequence at each location point is calculated, 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, where the rows and columns of the diffusion relationship matrix correspond to the positions on the chip surface, and the matrix element values ​​represent the heat transfer intensity from one position to another. This intensity is determined by the ratio of the temperature change rate difference between the positions to the spatial distance. By utilizing the spatial coordinate information of the temperature sensing array and using spatial mapping technology to take the surface temperature value as the boundary condition, combined with the thermal conductivity characteristics of the chip material, the three-dimensional thermal conductivity equation is solved to infer the temperature values ​​of each internal layer, generating the three-dimensional temperature distribution at each time point, so as to reconstruct the three-dimensional temperature distribution sequence inside the power management chip. In the three-dimensional temperature distribution sequence, the highest temperature location point at each time point is identified, wherein the highest temperature location point is determined by comparing the temperature values ​​of all internal location points, and the changes of the highest temperature location point over time are connected to form a hotspot migration path, wherein the path is represented by a sequence of coordinates of the highest temperature location at consecutive time points; In the hotspot migration path, turning points of directional change are detected. 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.

4. The method according to claim 1, characterized in that, Separating the impulse component sequence synchronized with the timestamp of the inflection point from the output ripple voltage of the load abrupt change includes: During a load surge event, the output voltage signal of the power management chip is monitored, and the ripple component in the output voltage signal is extracted, wherein the ripple component is obtained by removing the DC component; Based on the timestamp of the inflection point, a synchronized time window is located in the ripple component, wherein the time window is centered on each inflection point timestamp and the width is preset to a fixed value. Within the time window, the 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 inflection point timestamp.

5. The method according to claim 1, characterized in that, Obtain the time-domain thermal map sequence of the surface region of the power management chip during a sudden load change, including: During the operation of the power management chip, a load surge event is applied. During the occurrence of the load surge event, a temperature distribution image of the chip surface is captured at a preset time interval using a temperature sensing array. Each sensing unit of the temperature sensing array corresponds to a specific area on the chip surface, and the capture process is performed at a fixed frame rate. Each of the temperature distribution images is converted into a two-dimensional temperature value matrix, where each matrix element represents the temperature value detected by the 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 given time point.

6. The method according to claim 3, characterized in that, Utilizing the spatial coordinate information of a temperature sensing array, surface temperature values ​​are used as boundary conditions through spatial mapping technology. Combined with the thermal conductivity characteristics of the chip material, a three-dimensional heat conduction equation is solved to infer the temperature values ​​of each internal layer, generating a three-dimensional temperature distribution at each time point. This reconstructs the three-dimensional temperature distribution sequence inside the power management chip, including: The position of the surface detection point is determined according to the spatial coordinates of the temperature sensing array. The internal virtual layers are divided at equal intervals along the chip thickness direction, and grid nodes aligned with the surface detection points are generated in each layer. The thermal conductivity coefficient of the chip packaging material is pre-stored, and the thermal resistance value between adjacent grid nodes is set. The temperature value of the surface detection point is assigned to the outermost mesh node, a thermal balance equation is established for each internal mesh node, and calculations are performed based on the temperature value and thermal resistance value. The operation is performed on each mesh node in order from the surface to the inside of the chip, taking the temperature of the solved nodes in the upper layer as a known quantity, and solving the thermal balance equation of the current layer node until the innermost layer node is solved. The temperature values ​​of all grid nodes at each sampling time are obtained to form a three-dimensional temperature distribution, and the three-dimensional temperature distributions at all sampling times are stored continuously in chronological order to form a complete sequence.

7. A power management chip multi-mode modulation switching threshold setting system, used to execute the method for setting a power management chip multi-mode modulation switching threshold as described in any one of claims 1 to 6, characterized in that, include: The acquisition module acquires the time-domain thermal map sequence of the surface region of the power management chip during a sudden load change. The marking module generates a heat diffusion matrix based on the temperature change rate of the time-domain heat map sequence. At the same time, it reconstructs the three-dimensional heat 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 heat distribution as hot spot trajectories, and marks the timestamps of the turning points in the hot spot trajectories. The separation module separates the impulse component sequence synchronized with the timestamp of the turning point from the output ripple voltage of the load abrupt change; The construction module performs singular value decomposition on the heat diffusion matrix to obtain matrix feature quantities, performs tensor fusion of the matrix feature quantities 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 continuously updates the local minimum points of the switching cost function during the chip's operating cycle, dynamically correcting the switching threshold boundary between BUCK mode and BOOST mode, so that the switching operation occurs in the stable range of the hot spot trajectory.

8. 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 invoked 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 6.

9. A computer storage medium, characterized in that, The device stores a computer program, which, when executed by a computer, implements the power management chip multi-mode modulation switching threshold setting method as described in any one of claims 1 to 6.

Images