Temperature compensation-based MOS threshold voltage on-line monitoring and calibration system

Abstract

The embodiment of the application provides a MOS tube threshold voltage online monitoring and calibration system based on temperature compensation, which is applied to the field of power electronics technology, adopts a distributed temperature-sensitive optical sensing array to obtain temperature representation signals of multiple space points of a chip group substrate; a two-dimensional or three-dimensional temperature field distribution is reconstructed through a temperature field reconstruction and hot spot analysis module to identify a local hot spot area; a partition compensation module generates an independent enhanced compensation signal for a specific chip in the hot spot area, and the compensation amount is greater than a reference compensation amount based on an average temperature. Further, the system inversely calculates the actual threshold voltage through the collection of drain current, and forms a closed-loop adaptive correction by using a PID controller. The application realizes a leap from point temperature measurement to field perception, and significantly improves the reliability and consistency of the power module under a non-uniform thermal field through partition independent enhanced compensation and closed-loop correction.

Description

Technical Field

[0001] This application relates to the field of power electronics technology, and in particular to a temperature-compensated online monitoring and calibration system for MOSFET threshold voltage. Background Technology

[0002] Power MOSFETs, as core components of power electronic converters, have a negative temperature coefficient for their threshold voltage, meaning it decreases as the junction temperature rises. This characteristic leads to increased switching losses and changes in on-resistance at high temperatures, potentially causing thermal runaway in severe cases. Traditional threshold voltage compensation methods typically employ single-point temperature monitoring based on the overall average temperature of the chip or the case temperature, applying a uniform compensation amount to the entire power module. However, this method ignores the non-uniform temperature field within the power chipset caused by chip layout, heat dissipation conditions, and uneven current distribution, making it impossible to provide targeted compensation for locally high-temperature areas.

[0003] In practical high-power applications, local hot spots with temperatures far exceeding the average value often form on the substrate of power MOS chipsets. The chip threshold voltage drift is more significant in the hot spot region. If only global average temperature compensation is used, these chips will be in an undercompensated state for a long time, and their electrical stress level will be much higher than that of other areas, becoming a bottleneck in system reliability and accelerating performance degradation and failure.

[0004] Existing temperature monitoring methods mostly employ point sensors such as thermistors or thermocouples. However, due to limitations in wiring complexity and electromagnetic interference, it is difficult to deploy high-density sensor networks within chipsets. The lack of precise sensing capabilities for the spatial distribution of the temperature field makes the identification and differential compensation of hotspot areas a technical challenge, thus restricting the long-term stable operation of power modules under complex operating conditions. Summary of the Invention

[0005] The embodiments of this application provide an online monitoring and calibration system for the threshold voltage of a MOSFET based on temperature compensation, which addresses the threshold voltage mismatch problem caused by uneven temperature distribution. To achieve the above objective, this application adopts the following technical solution: A temperature-compensated online monitoring and calibration system for the threshold voltage of a MOSFET, the system comprising: A distributed temperature-sensitive optical sensing array includes multiple optical sensing units disposed at different positions on the power MOS chipset substrate, used to acquire temperature characterization signals of multiple spatial points within the chipset region; The temperature field reconstruction and hot spot analysis module is used to reconstruct the two-dimensional or three-dimensional temperature field distribution of the power MOS chipset substrate based on the temperature characterization signals of each spatial point, and identify local hot spot regions with temperatures higher than the average value. The partition compensation module is used to generate and apply independent enhancement compensation signals for specific power MOS chips corresponding to the identified local hot spots; wherein the compensation amount for the local hot spots is greater than the reference compensation amount calculated based on the average temperature of the chipset.

[0006] In some possible implementations, the distributed temperature-sensitive optical sensing array is a fiber Bragg grating sensing array, specifically used for: Each optical sensing unit senses temperature changes at its location, generates a corresponding wavelength shift, and produces a reflection spectrum carrying temperature information. The reflection spectrum is demodulated to determine the wavelength shift of each optical sensing unit. According to the preset wavelength-temperature mapping relationship, the wavelength drift is converted into a temperature characterization signal.

[0007] In some possible implementations, the distributed temperature-sensitive optical sensing array is specifically used for: Each of the optical sensing units is fixed on the back side of the power MOS chip substrate in a grid-like arrangement, and the center-to-center distance between adjacent optical sensing units is less than or equal to the preset thermal diffusion feature length, forming a sensing grid covering the chip area. All the optical sensing units collect temperature data at their respective locations to obtain the raw temperature value of each optical sensing unit; Based on the original temperature values, a discrete temperature characterization signal set covering the entire chipset region is generated.

[0008] In some possible implementations, the temperature field reconstruction and hotspot analysis module includes a data preprocessing unit, specifically used for: Receive the discrete temperature characterization signal set; The received signal set is processed by outlier removal and interpolation to generate a clean discrete temperature dataset; The Kriging interpolation algorithm is used to spatially interpolate the clean discrete temperature dataset to generate continuous two-dimensional or three-dimensional temperature field distribution data.

[0009] In some possible implementations, the temperature field reconstruction and hotspot analysis module is specifically used for: Receive temperature field distribution data; The average temperature of the entire temperature field in the temperature field distribution data is calculated as the reference temperature; Calculate the temperature difference between each spatial location and the reference temperature point by point, and mark all grid points whose difference exceeds the preset hotspot threshold as candidate hotspots; Connectivity analysis is performed on the marked candidate hotspots, and connected regions with an area greater than the preset minimum hotspot area are determined as local hotspot regions; Hotspot information is generated based on the local hotspot area, and the hotspot information includes the location coordinates and temperature distribution of the area.

[0010] In some possible implementations, the partition compensation module includes a hotspot mapping unit, specifically used for: Receive hotspot information and obtain preset power MOS chip layout coordinates, the layout coordinates including the position range of each power MOS chip on the substrate; Spatially match the location coordinates of the hotspot areas in the hotspot information with the layout coordinates of the power MOS chips to determine the power MOS chips that fall within the range of each hotspot area and generate a list of hotspot chips. The average temperature of the chipset is calculated based on the temperature distribution in the hotspot information, and the global baseline compensation is calculated based on the average temperature of the chipset. Output the list of hotspot chips and the global baseline compensation amount.

[0011] In some possible implementations, the partition compensation module includes multiple independent compensation signal generation units, specifically used for: Each compensation signal generation unit corresponds to a power MOS chip; The compensation signal generation unit receives a list of hotspot chips and a global reference compensation amount; If the corresponding chip is in the hot spot chip list, then the additional compensation coefficient is calculated based on the difference between the actual temperature of the hot spot where the chip is located and the reference temperature. The enhanced compensation signal is obtained by multiplying the additional compensation coefficient by the global reference compensation amount.

[0012] In some possible implementations, the compensation signal generation unit is specifically used for: Receive the real-time temperature value of the power MOS chip obtained by the temperature field reconstruction and hotspot analysis module based on the temperature field distribution at the current moment through interpolation; Calculate the temperature difference between the real-time temperature value and the average temperature of the chipset; According to a preset compensation function, the temperature difference is linearly converted into an additional compensation voltage value; Obtain the global reference compensation amount, and extract the reference compensation voltage value from the global reference compensation amount; The additional compensation voltage value is added to the reference compensation voltage value to generate an enhanced compensation voltage signal.

[0013] In some possible implementations, the partition compensation module further includes a compensation effect monitoring unit: Collect the drain current of each power MOS chip after applying the enhancement compensation signal to obtain the drain current monitoring value; Obtain the preset rated threshold voltage; The compensation deviation is calculated based on the drain current monitoring value and the rated threshold voltage; Correction coefficients are generated based on the compensation deviation.

[0014] In some possible implementations, the compensation effect monitoring unit is specifically used for: The collected drain current value is used to calculate the current actual threshold voltage according to the preset conversion relationship; The actual threshold voltage is compared with the preset rated threshold voltage to obtain the threshold voltage offset; The threshold voltage offset is input into the PID controller, and the PID controller outputs an adjustment amount as a correction coefficient. The correction coefficient is fed back to the corresponding compensation signal generation unit, which is used to multiply the additional compensation voltage of the temperature difference conversion when calculating the enhanced compensation signal.

[0015] As can be seen from the above technical solution, this application has the following beneficial effects: 1. This invention employs a gridded fiber Bragg grating sensing array. The sensor spacing is optimized based on the length of the thermal diffusion characteristics, ensuring that the spatial sampling rate is higher than the highest spatial frequency of the temperature field. Combined with the Kriging interpolation algorithm, discrete temperature data is spatially reconstructed, enabling the distortion-free reconstruction of high-resolution two-dimensional or three-dimensional continuous temperature field distribution, accurately identifying the location, boundaries, and temperature gradients of hotspot areas. This technological leap from point temperature measurement to field sensing gives the system, for the first time, the quantitative analysis capability of where and how much heat is present, providing precise spatial basis for subsequent zonal compensation.

[0016] 2. This invention spatially matches identified hotspot areas with specific chip layout coordinates using a hotspot mapping unit. The system generates an enhanced compensation signal proportional to the real-time temperature difference for chips within the hotspot area. The compensation amount is significantly greater than the baseline compensation amount based on average temperature, achieving differentiated and precise force application. The system collects drain current in real time and calculates the actual threshold voltage, inputting the deviation into a PID controller to generate a correction coefficient, dynamically adjusting the compensation amount to form a closed-loop negative feedback. This architecture provides hotspot chips with additional compensation matching their temperature exceedance, effectively suppressing excessive threshold voltage drift and balancing the electrical stress distribution across the entire chipset. The closed-loop adaptive mechanism automatically offsets device aging, individual differences, and model errors, ensuring that compensation accuracy remains online throughout the system's entire lifecycle. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 A flowchart of distributed temperature-sensitive optical sensing and temperature data acquisition provided in the embodiments of this application; Figure 2 A flowchart illustrating temperature field reconstruction and hotspot region identification provided in this application embodiment; Figure 3 This is a flowchart of the partition compensation signal generation process provided in an embodiment of this application; Figure 4 This is a flowchart of the closed-loop feedback compensation effect monitoring and correction provided in the embodiments of this application. Detailed Implementation

[0019] The terms "first," "second," and "third," etc., used in this application specification, claims, and drawings are for distinguishing different objects, not for specifying a particular order.

[0020] In the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0021] Research has revealed that existing technologies rely solely on single-point or casing temperature monitoring, which fails to detect localized hotspots within the chipset caused by uneven layout and heat dissipation, resulting in blind spots in temperature sensing. Furthermore, the adopted global unified compensation strategy leaves the chips in hotspot areas in a state of undercompensation for extended periods, leading to excessive threshold voltage drift and reliability issues.

[0022] To solve the above problems, such as Figure 1 - Figure 4 As shown, this application provides a temperature-compensated online monitoring and calibration system for the threshold voltage of a MOSFET: Example 1: System overall architecture and application scenarios.

[0023] This embodiment describes an online monitoring and calibration system for the threshold voltage of MOSFETs in the power module of an electric vehicle's main drive inverter. The inverter module includes a chipset composed of multiple parallel-connected power MOSFET chips for driving a high-power motor.

[0024] (a) System construction and connection relationship.

[0025] This system mainly consists of three core components: a distributed temperature-sensitive optical sensing array, a temperature field reconstruction and hotspot analysis module, and a partition compensation module.

[0026] 1. Deployment of distributed temperature-sensitive optical sensor arrays: In this embodiment, we selected a fiber Bragg grating (FBG) as the optical sensing unit. On the back side of the power MOS chipset substrate, an optical fiber comprising multiple FBGs is tightly bonded together in a serpentine or mesh-like path using a highly thermally conductive insulating adhesive. Each FBG corresponds to a specific spatial location, and the center-to-center spacing is precisely calculated and set to be slightly smaller than the thermal diffusion characteristic length of the power module under thermal equilibrium conditions. For example, if thermal simulation analysis shows that the thermal diffusion characteristic length of the module under typical operating conditions is approximately several millimeters, then the center-to-center spacing of adjacent FBGs is designed to be slightly smaller than this value. In this way, the entire sensing array forms a high-resolution discrete temperature sensing network covering the entire projected area of ​​the chipset on the back side of the substrate. Each FBG unit independently senses the temperature change of its micro-region.

[0027] 2. Information flow and module connection: The distributed temperature-sensitive optical sensor array is connected to a fiber optic signal demodulator via optical fiber. The demodulator converts the acquired optical signals into electrical signals and transmits them to the temperature field reconstruction and hotspot analysis module via a high-speed data bus (such as a PXIe bus). This module runs in the main control chip of the inverter controller (such as a high-performance FPGA or DSP). The temperature field distribution data and hotspot area information calculated by the analysis module are sent to the zone compensation module in real time. The zone compensation module is also integrated into the main control chip. Based on the received instructions, it dynamically adjusts the drive circuit parameters connected to the gates of each power MOSFET, generating an independent, precisely calibrated gate drive voltage for each MOSFET, i.e., an enhancement compensation signal.

[0028] (ii) Logical derivation of the system's working principle.

[0029] The core challenge of existing technologies lies in the uneven temperature distribution of power MOSFET chipsets, while the threshold voltage drift is temperature-sensitive. If global compensation based on average temperature is used, chips in hotspot areas will fail due to insufficient compensation, becoming a bottleneck in the system. This invention aims to accurately identify and address this bottleneck.

[0030] This invention transforms the previously invisible and abstract problem of temperature distribution into a visible and concrete problem of temperature data at discrete spatial points through a distributed temperature-sensitive optical sensor array. This is analogous to deploying multiple sentries—optical sensing units—on a battlefield, each capable of accurately reporting the enemy situation at its location, i.e., the temperature.

[0031] Next, the temperature field reconstruction and hotspot analysis module is responsible for integrating these discrete sentinel reports into a complete battlefield situation map (two-dimensional or three-dimensional temperature field distribution). Using this map, the system can intelligently identify the areas of most intense enemy fire, i.e., local hotspots. This resolves the question of where the problem lies.

[0032] Finally, the role of the zone compensation module is to execute precision strikes. Instead of uniformly allocating ammunition (baseline compensation) to the entire army (the entire chipset), it targets besieged hotspots (local hotspots), mobilizing superior forces for saturation reinforcement (enhancing compensation signals). This directly solves the core challenge of how to precisely address problems.

[0033] As can be seen, this invention upgrades the temperature compensation strategy from a rough, egalitarian approach to a refined, precise one by using a closed-loop logic of perception reconstruction and identification of zonal compensation. This fundamentally eliminates the reliability shortcomings caused by uneven temperature distribution and significantly improves the overall lifespan and performance of the power module.

[0034] By constructing a high-resolution temperature field distribution, local hotspot areas were identified, providing a precise spatial basis for subsequent differentiated compensation and avoiding the undercompensation or overcompensation problems of traditional single-point or average temperature compensation. Enhanced compensation was applied to the chips in the most vulnerable hotspot areas, effectively suppressing excessive drift of their threshold voltages and balancing the operating state of the entire chipset. This significantly slowed down the degradation process of system performance and greatly improved the long-term reliability of the power module under complex operating conditions. The temperature field distribution reconstructed in real time by the system can not only be used for threshold compensation but also serve as feedback for thermal design, guiding active thermal management strategies such as adjusting switching frequency and reducing load to achieve electro-thermal synergistic optimization.

[0035] Example 2: Specific implementation of a distributed temperature-sensitive optical sensor array.

[0036] A fiber Bragg grating (FBG) is a periodic refractive index modulation structure that uses ultraviolet light to write into the core of an optical fiber. It can be viewed as a filter that reflects only specific wavelengths of light, known as Bragg wavelengths. When a broadband beam of light is incident on an FBG, wavelengths that satisfy the Bragg condition are reflected back, while the remaining wavelengths are transmitted.

[0037] Wavelength drift: The Bragg wavelength of FBG is extremely sensitive to ambient temperature and strain. When the temperature rises, the fiber material undergoes thermal expansion, and its refractive index also changes, causing the reflected wavelength to shift towards longer wavelengths (redshift); when the temperature decreases, it shifts towards shorter wavelengths (blueshift). The change in wavelength and the change in temperature exhibit a good linear relationship within a certain range.

[0038] Demodulation: The process of converting the wavelength change signal reflected back by the FBG, which carries temperature information, into an electrical signal, such as voltage, current, or digital quantity, that can be recognized and processed by electronic systems.

[0039] In this embodiment, the distributed temperature-sensitive optical sensing array acquires the temperature characterization signal through the following steps: 1. Sensing temperature changes and wavelength shifts: When the power MOS chipset operates, the substrate temperature rises. The temperature of a micro-region containing a FBG cell located directly on the back of the substrate increases. This temperature rise causes thermal expansion (increased grating period) and thermo-optical effects (increased effective refractive index) in the fiber material of that FBG. According to the Bragg wavelength formula... The increase in period and refractive index together cause a shift in the Bragg wavelength of the FBG. The system can monitor the shift in the center wavelength of the reflection spectrum of each FBG in real time using a broadband light source and a spectrometer.

[0040] 2. Demodulation processing of the reflectance spectrum: The fiber optic signal demodulator internally includes a photodetector, a signal amplifier, and an analog-to-digital converter. It converts the received weak reflected light signals from different fiber optic frames (FBGs) into electrical signals, amplifies them, and filters them. Then, using a peak detection algorithm or a centroid algorithm, it accurately determines the peak wavelength of each reflected spectrum, thereby obtaining the precise wavelength shift of each optical sensing unit. .

[0041] 3. Wavelength-temperature mapping conversion: During system initialization or calibration, the wavelength temperature coefficient of each FBG (or the same batch of FBGs) has been experimentally calibrated. This coefficient represents the wavelength shift caused by a unit change in temperature. The system internally uses a wavelength-temperature mapping lookup table or a linear function. By substituting the demodulated wavelength shift into the mapping function, the temperature change at that spatial point can be calculated. Combined with the initial reference temperature Then you can get the current temperature value. The calculated temperature value is the temperature representation signal for that spatial point.

[0042] In the specific scenario described in this invention, the inverter's operating environment is subject to extremely strong electromagnetic interference. Traditional electrical sensors (such as thermistors and thermocouples) are easily affected by this interference, resulting in signal distortion. The FBG sensor, being inherently insulated and unaffected by electromagnetic interference, ensures the accuracy and stability of temperature measurement.

[0043] FBGs offer extremely high wavelength demodulation accuracy, achieving temperature resolution better than 0.1 degrees Celsius. Furthermore, by cascading multiple FBGs with different center wavelengths onto a single optical fiber, distributed measurements at dozens of points can be easily achieved, greatly simplifying wiring and improving system integration. FBG sensors are passive, corrosion-resistant, and have a long lifespan, making them suitable for long-term stable operation in the high-temperature, high-stress packaging environments of power modules.

[0044] Example 3: Sensor array layout and temperature data acquisition.

[0045] 1. Grid layout design: To achieve high-fidelity temperature field reconstruction, sensor layout is crucial. In this embodiment, the design principles are as follows: Layout method: An orthogonal grid layout is adopted, in which the optical fibers are arranged on the back side of the substrate in a U-shape, so that each FBG unit forms a regular M-row and column matrix on the plane.

[0046] Spacing determination: The characteristic length of heat diffusion refers to the spatial distribution of temperature propagating outward from a significant heat source point during steady-state heat conduction, with its amplitude attenuating to approximately 37% (i.e., The propagation distance at maximum continuous operating power is determined by finite element thermal simulation analysis for the power module in this embodiment. The thermal diffusion characteristic length at maximum continuous operating power is approximately several millimeters. To ensure the capture of details in hotspot areas, the center-to-center spacing between adjacent FBG units is set slightly less than this length, thereby ensuring that the spatial sampling rate of the sensor network is higher than the highest spatial frequency of the temperature field, thus avoiding spatial aliasing.

[0047] Fixing method: The optical fiber is precisely placed in a ceramic substrate with microgrooves, and then encapsulated and fixed to the back of the power MOS chipset substrate using a dispensing process with highly thermally conductive and insulating epoxy resin. This method ensures good thermal conductivity while providing mechanical protection for the fragile optical fiber.

[0048] 2. Generation of discrete temperature characterization signal sets: Data Acquisition: During system operation, the demodulator scans all FBG units sequentially at a sufficiently high frequency, such as tens to hundreds of times per second. Within each scan cycle, the system acquires multiple independent data points corresponding to different spatial coordinates. The original temperature value .

[0049] Dataset Generation: These raw temperature values, carrying their location information, collectively constitute a discrete temperature characterization signal set covering the entire chipset area. This dataset can be represented in matrix form, with each element representing the temperature value of the corresponding grid point.

[0050] The sensor spacing is determined based on the thermal diffusion characteristic length, following a spatial domain generalization of the Nyquist-Shannon sampling theorem. This ensures that the sampling frequency (spatial sampling rate) of the sensor network is higher than the spatial frequency of the temperature field, thus enabling distortion-free reconstruction of the original temperature field and avoiding the omission of hotspot regions with scales comparable to or larger than the thermal diffusion length. Each temperature value is bound to a fixed physical coordinate, allowing subsequent hotspot analysis to not only determine the extent of heat but also pinpoint its exact location, providing a precise spatial basis for zonal compensation.

[0051] Example 4: Temperature field reconstruction and data preprocessing.

[0052] Outlier removal: Due to sensor malfunctions, signal interference, or demodulation errors, the collected data may include "outliers" that clearly do not conform to physical laws. For example, the temperature at a certain point may change abruptly or deviate significantly from that of neighboring points. Removing these outliers is the first step in ensuring data quality.

[0053] Kriging interpolation is a spatial interpolation method based on geostatistics. It considers not only the distance between the interpolation point and known sampling points, but also the spatial autocorrelation (variance) among the known sampling points. Kriging interpolation can provide the best linear unbiased estimate for each interpolation point, and simultaneously provide the variance of the estimate, i.e., the estimation accuracy. It is particularly suitable for physical fields with continuity and a certain spatial structure, such as the reconstruction of temperature fields.

[0054] The variogram describes the relationship between the variance of attribute values ​​between two points in space and the distance between those points. Commonly used theoretical models include the spherical model and the exponential model. By fitting the variogram to sampled data, the spatial correlation of the temperature field can be quantified.

[0055] 1. Data cleaning: The data preprocessing unit first receives a set of discrete temperature characterization signals from the sensor array.

[0056] Outlier removal: The system sets a reasonableness threshold. For example, if the temperature change rate of a single point exceeds a certain number of degrees per second, or if the temperature difference between a point and the average temperature of its surrounding points exceeds a certain range, then that point is determined to be an outlier. Data from points determined to be outliers is marked as invalid.

[0057] Interpolation Processing: Data points marked as invalid cannot be simply ignored, otherwise it will lead to gaps in subsequent interpolation. The system uses simple linear interpolation or nearest neighbor interpolation for initial filling. For example, the average temperature of the valid neighbors around the point is taken as the replacement value for that point. After cleaning and interpolation, a clean, gap-free discrete temperature dataset is finally generated.

[0058] 2. Kriging space interpolation: Input: The cleaned discrete temperature dataset and its corresponding spatial coordinates.

[0059] Variation function modeling: The algorithm first analyzes the relationship between the semivariogram and distance for all known point pairs, and fits a theoretical variation function curve (such as a spherical model). The expression for the spherical model is: Wherein is the nugget constant (representing micro-variation or measurement error). This represents the sill value (representing total variation). The range is denoted by _(maximum distance)_. These parameters are obtained through fitting, and the curve describes the spatial correlation of the temperature field: points that are closer together have stronger temperature correlations and smaller semivariances; points that are farther apart have weaker correlations and larger semivariances, until a stable ground state value is reached.

[0060] Mesh generation: The system divides the entire chipset area into a high-resolution virtual mesh, for example, with a mesh spacing much smaller than the sensor spacing. The temperature of each point on this virtual mesh (called the interpolation point) is unknown and needs to be estimated.

[0061] Point-by-point Kriging estimation: For each interpolation point on the virtual grid, the algorithm searches for known sampling points within a certain range around it, such as the nearest few points. Then, based on the previously established variogram model, it solves a system of Kriging equations to calculate the optimal weighting coefficients for these known points. These coefficients satisfy the conditions of unbiasedness and minimization of estimation variance. Finally, the temperature estimate of the interpolation point is the weighted sum of the temperatures of these known points.

[0062] Output: Iterates through all virtual grid points, calculates the temperature estimate for each point, and finally generates a high-resolution, continuous two-dimensional or three-dimensional temperature field distribution data consisting of a large number of data points. This data field can be visualized as a temperature cloud map, clearly showing the temperature distribution details on the chipset surface.

[0063] Outlier removal and interpolation processes filter out measurement noise and occasional errors, ensuring the accuracy of subsequent analysis and avoiding the pitfalls of adding and removing irrelevant data. The Kriging interpolation algorithm fully utilizes the spatial correlation of the temperature field, enabling the reconstruction of a continuous and smooth temperature field distribution far exceeding the physical resolution of the sensors using limited discrete sensor information. This is crucial for accurately identifying hotspot boundaries and calculating hotspot areas. Compared to traditional linear interpolation, Kriging interpolation has a clearer physical meaning and produces more accurate and reliable results.

[0064] Example 5: Hotspot area identification algorithm.

[0065] 1. Calculate the reference temperature: The module receives high-resolution continuous temperature field distribution data generated by the previous stage. First, it performs an arithmetic mean on the temperature values ​​of all grid points within the entire temperature field region to obtain a total average temperature, which is used as the chipset average temperature. .

[0066] 2. Mark candidate hotspots: Next, the system iterates through every grid point in the temperature field. Its temperature value The system calculates the difference between the temperature at that point and the average temperature of the chipset. .

[0067] The system has a preset hotspot threshold. This is a temperature difference threshold, representing the degree to which a hot spot deviates significantly from the average temperature, as considered in engineering practice. If... If a grid point is marked as a candidate hotspot, then all the marked candidate hotspots logically constitute a binary image (1 represents a candidate hotspot, and 0 represents a non-hotspot).

[0068] 3. Connected component analysis: Relying solely on a single-point temperature threshold may result in isolated, small-area noise points. A true local hotspot region should be a physically sized region composed of a large number of consecutive, adjacent candidate hotspots.

[0069] Therefore, the system performs connected component analysis on the identified candidate hotspots. This is an image processing algorithm used to identify and label adjacent (four-neighbor or eight-neighbor) pixel clusters in a binary image. The algorithm scans all candidate hotspots and divides them into different groups according to spatial connectivity; each group is an independent connected component.

[0070] 4. Hotspot identification and information generation: The system also presets a minimum hotspot area. It is used to filter out "false hotspots" that are too small and may be caused by noise or localized minor disturbances.

[0071] For each connected component identified through connected component analysis, the system calculates the number of grid points it contains and converts it into actual physical area based on the grid spacing. Only those connected components are ultimately identified as true local hotspots.

[0072] Finally, the system generates hotspot information for each identified local hotspot area, including: Location coordinates: The sequence of geometric center coordinates and boundary coordinates of the hotspot region.

[0073] Temperature distribution: the highest temperature, lowest temperature and average temperature in this area.

[0074] Area Identifier: Assign a unique ID to each hotspot.

[0075] By combining threshold segmentation and connected component analysis, not only were areas with excessive temperatures identified, but their location, size, and temperature distribution characteristics were also precisely quantified. This provides accurate target guidance for subsequent zonal compensation, and by setting a physical constraint of minimum hotspot area, false triggering caused by individual sensor noise or local measurement errors is effectively avoided.

[0076] Example 6: Hotspot mapping in the partition compensation module.

[0077] 1. Data Reception and Preparation: The hotspot mapping unit receives the latest hotspot information from the temperature field reconstruction and hotspot analysis module. Simultaneously, it reads the preset power MOS chip layout coordinates from the system's configuration memory. These layout coordinates are precise engineering drawing data that defines the specific location range of each power MOS chip on the substrate, for example, by sequential numbering, typically represented by a rectangle or a more complex polygon.

[0078] 2. Spatial matching and list generation: This is the core function of the hotspot mapping unit. It performs spatial geometric calculations between the hotspot region (a polygonal region) and the location region of each chip.

[0079] Matching Algorithm: The system employs an algorithm that matches polygon intersections or points within polygons. For each hotspot region, the algorithm checks the layout areas of all chips one by one. If a chip's location area intersects with a hotspot region, or if the chip's geometric center is located inside a hotspot region, then the chip is determined to fall within that hotspot region.

[0080] Generating a list of hotspot chips: After matching each hotspot one by one, the system finally generates a list of hotspot chips. This list records the chip numbers of all chips currently within the hotspot area, as well as their respective hotspot IDs.

[0081] 3. Calculate the baseline compensation amount: While performing spatial matching, the unit also performs preliminary calculations of global parameters. Based on the distribution data of the entire temperature field, it calculates the average temperature of the chipset. (This average temperature may be the same as the previously calculated reference temperature, or it may be a recalculated weighted average temperature that takes into account the chip's current-carrying capacity.) Then, this average temperature is input into a preset reference compensation function. This function is typically a linear or nonlinear formula determined by the device's physical characteristics; it describes the amount of additional global reference compensation that needs to be applied to the gate at the average temperature to maintain rated operation. (Usually a voltage value). For example, the reference compensation amount can be expressed as a voltage value proportional to the change in average temperature relative to a reference temperature: , where is the global temperature compensation coefficient.

[0082] 4. Information Output: The hotspot mapping unit outputs the calculated list of hotspot chips and the global baseline compensation amount as two key data points to multiple independent compensation signal generation units in the subsequent stage.

[0083] This approach successfully maps abstract temperature hotspots to specific, addressable physical chips, enabling the identification of which devices require special attention based on their heat source. It provides precise targets for zonal compensation, separating the calculation of the baseline compensation from the identification and mapping of hotspot regions, resulting in a clear system structure and well-defined logic. The baseline compensation is a global, fundamental adjustment, while subsequent enhancement compensation is a localized correction based on this. The two operate independently, facilitating implementation and debugging.

[0084] Example 7: Generation of independent compensation signals.

[0085] The partitioned compensation module integrates a number of compensation signal generation units equal to the number of power MOS chips. Each unit is specifically responsible for driving and compensating a particular power MOS chip; for example, unit U1 is responsible for chip M1, unit U2 is responsible for chip M2, and so on. This one-to-one hardware architecture is the foundation for achieving truly independent partitioned compensation.

[0086] Taking the compensation signal generation unit responsible for a certain chip as an example, its workflow is as follows: 1. Receiving instructions and data: This unit receives the hotspot chip list and global baseline compensation value from the hotspot mapping unit in real time. .

[0087] 2. Determine chip status: The unit first checks the list of hot chips to determine if the chip it is responsible for is on the list.

[0088] Case A (Chip is in the hotspot): If the chip is in the list, the unit enters the enhancement compensation mode.

[0089] Case B (Chip not in hotspot): If the chip is not in the list, the unit enters the normal compensation mode, which directly uses the received global reference compensation amount as the final output signal without amplifying it.

[0090] 3. Calculate the additional compensation coefficient (enhanced compensation mode): Since the chip is located within a hotspot, the unit needs to calculate an additional compensation coefficient for it. The magnitude of this coefficient directly reflects the degree to which the hotspot temperature exceeds the standard.

[0091] This unit obtains the actual average temperature of the hotspot where the chip is located and the average temperature of the chipset from the hotspot information. Calculate the temperature difference between the two. .

[0092] The system has a pre-defined compensation and enhancement function, for example, the simplest linear relationship: The enhancement factor is a pre-calibrated parameter that represents the incremental compensation coefficient required per unit temperature difference. The enhancement factor can be determined through experimental calibration or simulation optimization, with the goal of fully compensating for the threshold voltage drift of the chip within the hot spot.

[0093] The calculated additional compensation coefficient means that for chips within the hotspot, the compensation level needs to be proportionally higher than the global benchmark.

[0094] 4. Generate enhanced compensation signal: Finally, the unit multiplies the calculated additional compensation coefficient by the global reference compensation amount to obtain the final enhanced compensation signal. This signal is sent to the driver stage connected to the chip, thus providing the chip with additional compensation.

[0095] In this scenario, this one-to-one compensation signal generation method based on coefficient amplification achieves true partition independence: independent hardware units ensure that the compensation of each chip is parallel and does not interfere with each other. Chips within the hotspot receive enhanced compensation, while other chips maintain regular compensation, achieving spatial differentiation of compensation intensity. The additional compensation coefficient is proportional to the temperature difference, allowing the compensation amount to be continuously and smoothly adjusted according to the dynamic changes in the hotspot temperature, rather than simple on / off compensation, avoiding system shocks caused by compensation jumps. The calculation method of multiplying the global reference compensation amount by a coefficient has clear logic and low computational load, making it very suitable for implementation in digital logic circuits or simple analog circuits, and has strong engineering practicality.

[0096] Example 8: Calculation of enhanced compensation signal based on real-time temperature value.

[0097] Instead of simply using the average temperature of the hot spot area, a more accurate real-time temperature value is interpolated for the chip.

[0098] 1. Obtain the chip's real-time temperature value: The compensation signal generation unit no longer passively receives hotspot list information, but actively queries the temperature field reconstruction and hotspot analysis module to obtain the chip's real-time temperature value. .

[0099] The temperature field reconstruction and hotspot analysis module possesses a high-resolution temperature field distribution across the entire substrate. Based on the chip's precise physical coordinates, such as its geometric center coordinates, the module utilizes a pre-established Kriging interpolation model or directly queries the current temperature field value at that location to quickly calculate the temperature at the chip's location. This real-time temperature value is a better representation of the chip's junction temperature than the average temperature of the hotspot area.

[0100] 2. Calculate the temperature difference: This unit also obtains the current chipset average temperature from the hotspot mapping unit. Then, the temperature difference between the chip's real-time temperature and its average temperature is calculated. .

[0101] 3. Linear conversion to additional compensation voltage: The system has a pre-defined compensation function, which in this embodiment is a linear function and is referred to as the compensation slope. The compensation slope represents the additional gate compensation voltage required per unit temperature difference. It can be calculated from the temperature coefficient of the device's threshold voltage or obtained through experimental calibration.

[0102] Therefore, the formula for calculating the additional compensation voltage value is: .

[0103] 4. Generate enhanced compensation voltage signal: This unit obtains the global reference compensation amount from the hotspot mapping unit and extracts the reference compensation voltage value from it. .

[0104] Finally, the additional compensation voltage value is algebraically added to the reference compensation voltage value to generate the final enhanced compensation voltage signal. This is a more direct superposition of physical quantities than multiplication, with a clear physical meaning. The magnitude of the compensation value directly reflects the degree of deviation between the chip's real-time temperature and the average level.

[0105] Achieving precise point-to-point compensation: By performing real-time temperature interpolation at each chip location, the calculation basis for the compensation signal is reduced from the regional average value to the point value, achieving true point-to-point precise calibration. This avoids inconsistencies in compensation caused by slight temperature differences between chips at different locations within the same hotspot area.

[0106] The physical meaning is clearer: the method of directly superimposing the compensation voltage based on the temperature difference is highly consistent with the physical drift mechanism of the threshold voltage, that is, the temperature change directly causes the voltage drift. The compensation logic is more intuitive and easier for engineers to understand and debug.

[0107] The compensation curve is flexibly configurable: the compensation slope, as a configurable parameter, brings great flexibility to the system. This slope can be dynamically adjusted according to different MOSFET models, different aging levels, and even different application conditions to achieve the optimal compensation effect.

[0108] Example 9: Monitoring the compensation effect of closed-loop feedback.

[0109] Drain current The current flowing from the drain to the source of a MOSFET when it is turned on. It is one of the most direct physical quantities reflecting the operating state of a MOSFET.

[0110] Rated threshold voltage The minimum gate-source voltage required for a MOSFET to turn on under standard test conditions (e.g., a temperature of 25 degrees Celsius). This value is usually given in the device datasheet and serves as a benchmark for measuring threshold voltage drift.

[0111] Constant current method: A commonly used method for measuring the threshold voltage of a MOSFET. Its basic principle is: under the condition that the drain-source voltage of the MOSFET is fixed, the gate-source voltage is adjusted... This allows the drain current to be equal to a preset small current value, such as 250 microamps. The gate-source voltage measured at this point is an approximation of the threshold voltage. For greater precision, multiple sets of [voltage parameters] can be measured. The data, through linear fitting of the relationship between the two, the intersection of the extension line and the axis is the threshold voltage.

[0112] To achieve higher precision closed-loop control, this embodiment introduces a compensation effect monitoring unit.

[0113] 1. Collect drain current monitoring values: High-precision current sensors, such as Hall effect current sensors or precision sampling resistors, are integrated into the source or drain circuit of each power MOS chip. The compensation effect monitoring unit acquires the drain current of the chip after the application of the enhanced compensation signal at an extremely high sampling rate, for example, millions of times per second, to obtain the current monitoring value. .

[0114] 2. Calculate the compensation deviation: The monitoring unit knows the current gate voltage value applied to the chip. (Including baseline compensation and additional compensation). To evaluate the compensation effect, it is necessary to deduce the chip's current actual threshold voltage.

[0115] The constant current method is implemented by briefly adjusting the gate voltage (during a period of time within the switching cycle that does not affect the main function) while simultaneously monitoring changes in the drain current. When the drain current is exactly equal to a preset small current value, the gate voltage at this moment is recorded, which is the actual threshold voltage at the current junction temperature. Alternatively, the formula for the saturation current of a MOSFET can be used. ,in This is the gain factor of the device. By measuring two sets... Data can be used to calculate and .because Due to temperature variations, pre-calibrated [equipment / systems] are often used in practice. Value or online estimation.

[0116] Obtain the rated threshold voltage: Read the chip's voltage at the reference temperature from memory. Rated threshold voltage at (e.g., 25 degrees Celsius) Meanwhile, based on the current average temperature of the chipset... Using a temperature model of the threshold voltage, the desired ideal threshold voltage at the current average temperature is calculated as the target value. .

[0117] Calculate the deviation: Compare the actual threshold voltage with the expected target value to obtain the threshold voltage offset. This refers to the compensation deviation. If the compensation is completely accurate, this deviation should be close to zero.

[0118] 3. Generate correction coefficients: Based on the calculated compensation deviation, the monitoring unit generates a correction coefficient to adjust the compensation strategy. A classic and effective implementation is: The threshold voltage offset is used as the error signal. The input is fed into a digital PID controller. The PID controller then adjusts the output based on the proportional gain. ,integral ,differential The three stages process the error signal and output an adjustment amount. This adjustment can be converted into a correction factor, for example... .

[0119] (III) Feedback application of correction coefficients The generated correction coefficients are fed back to the corresponding compensation signal generation unit in real time. Taking the calculation method of Example 8 as an example, the original enhanced compensation voltage is... After the correction is introduced, the compensation voltage at the next moment becomes... If the PID controller detects that the actual threshold voltage is still too low (indicating insufficient compensation), its output correction coefficient will be greater than 1, thereby amplifying the additional compensation voltage value calculated based on the temperature difference until the compensation deviation disappears, achieving zero steady-state error compensation.

[0120] In this scenario, this closed-loop monitoring and feedback mechanism enables adaptive calibration: the system no longer relies solely on a preset temperature and voltage model, but instead forms a complete closed-loop control system by directly monitoring the key output parameters (drain current) of the MOSFET. This allows the system to automatically adapt to device aging, individual differences, and model errors, always maintaining optimal compensation. As the power module's usage time increases, the characteristics of the MOSFET will change. The introduction of the compensation effect monitoring unit gives the compensation strategy self-correction capabilities, effectively offsetting the effects of long-term aging and ensuring the stability and reliability of the system throughout its entire lifecycle. The PID controller is one of the most mature and widely used algorithms in industrial control. Its parameters have clear physical meanings (proportional eliminates current error, integral eliminates cumulative error, and derivative predicts error trends), enabling it to quickly, smoothly, and accurately converge the threshold voltage deviation to zero, exhibiting good dynamic response characteristics.

[0121] Example 10: Precise correction based on PID controller.

[0122] The PID controller integrated within the compensation effect monitoring unit has the following input, output, and feedback connections: 1. Input (controlled variable and setpoint): Controlled quantity ( The actual threshold voltage of the chip is calculated in real time using the constant current method or other methods. .

[0123] Setting value ( This refers to the expected threshold voltage that the chip should achieve under current operating conditions. Since the threshold voltage varies with temperature, the set value should not be a fixed rated value at 25°C, but rather an ideal value that takes into account the current temperature. Typically, the threshold voltage exhibits a good linear relationship with temperature. , where is the reference temperature The rated threshold voltage at (e.g., 25°C) is available from the device datasheet. This is the temperature coefficient of the threshold voltage (usually a negative value, in millivolts per degree Celsius, and can also be obtained from datasheets or experimental calibration). Therefore, at the current chipset average temperature, the expected target threshold voltage is [value missing]. .

[0124] Error value : This error directly reflects the degree of deviation between the current compensated actual threshold voltage and the ideal target.

[0125] 2. PID controller calculation: PID controller based on error Perform the following calculations: in: It is the adjustment value output by the PID controller.

[0126] It is a proportionality coefficient that determines the strength of the response to the current error.

[0127] It is the integral coefficient, used to eliminate the steady-state error of the system. As long as the error exists, the integral term will continue to increase until the error is zero.

[0128] These are differential coefficients, which make predictive adjustments based on the trend of error changes, playing a damping role and preventing system overshoot.

[0129] The tuning of these three coefficients can be achieved through trial and error based on engineering experience or... Wait until the classic method is completed.

[0130] 3. Output (Correction Factor): Adjustment value of PID controller output It is converted into a correction factor. This factor can be defined as follows: When the error When positive (actual threshold voltage is lower than target value), For positive, Greater than This enhances compensation; conversely, it also enhances it.

[0131] 4. Feedback to the compensation signal generation unit: The correction coefficient is sent in real time to the corresponding compensation signal generation unit of the chip. When calculating the additional compensation voltage at the next moment, the unit multiplies it by this coefficient, as described in Example 9, thereby dynamically adjusting the compensation intensity to form a complete closed-loop negative feedback control loop.

[0132] In this scenario, a PID controller-based correction method achieves zero steady-state error and precise compensation: through the action of the integral term, the PID controller eliminates any steady-state error, ensuring that the final actual threshold voltage is precisely controlled at the desired target value, regardless of model deviation or external disturbances. Proper PID parameter tuning allows the system to respond quickly to temperature changes and load surges while avoiding drastic oscillations in the compensation voltage, ensuring the smoothness of the compensation process and the dynamic stability of the system. The PID algorithm has a simple structure, clear parameter physical meanings, and extensive engineering application experience. Even under complex electromagnetic environments and operating conditions, it provides stable and reliable control performance.

[0133] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A temperature-compensated online monitoring and calibration system for the threshold voltage of a MOSFET, characterized in that, The system includes: A distributed temperature-sensitive optical sensing array includes multiple optical sensing units disposed at different positions on the power MOS chipset substrate, used to acquire temperature characterization signals of multiple spatial points within the chipset region; The temperature field reconstruction and hot spot analysis module is used to reconstruct the two-dimensional or three-dimensional temperature field distribution of the power MOS chipset substrate based on the temperature characterization signals of each spatial point, and identify local hot spot regions with temperatures higher than the average value. The partition compensation module is used to generate and apply independent enhancement compensation signals for specific power MOS chips corresponding to the identified local hot spots; wherein the compensation amount for the local hot spots is greater than the reference compensation amount calculated based on the average temperature of the chipset.

2. The system according to claim 1, characterized in that, The distributed temperature-sensitive optical sensing array is a fiber Bragg grating sensing array, specifically used for: Each optical sensing unit senses temperature changes at its location, generates a corresponding wavelength shift, and produces a reflection spectrum carrying temperature information. The reflection spectrum is demodulated to determine the wavelength shift of each optical sensing unit. According to the preset wavelength-temperature mapping relationship, the wavelength drift is converted into a temperature characterization signal.

3. The system according to claim 1, characterized in that, The distributed temperature-sensitive optical sensing array is specifically used for: Each of the optical sensing units is fixed on the back side of the power MOS chip substrate in a grid-like arrangement, and the center-to-center distance between adjacent optical sensing units is less than or equal to the preset thermal diffusion feature length, forming a sensing grid covering the chip area. All the optical sensing units collect temperature data at their respective locations to obtain the raw temperature value of each optical sensing unit; Based on the original temperature values, a discrete temperature characterization signal set covering the entire chipset region is generated.

4. The system according to claim 1, characterized in that, The temperature field reconstruction and hotspot analysis module includes a data preprocessing unit, specifically used for: Receive the discrete temperature characterization signal set; The received signal set is processed by outlier removal and interpolation to generate a clean discrete temperature dataset; The Kriging interpolation algorithm is used to spatially interpolate the clean discrete temperature dataset to generate continuous two-dimensional or three-dimensional temperature field distribution data.

5. The system according to claim 1, characterized in that, The temperature field reconstruction and hotspot analysis module is specifically used for: Receive temperature field distribution data; The average temperature of the entire temperature field in the temperature field distribution data is calculated as the reference temperature; Calculate the temperature difference between each spatial location and the reference temperature point by point, and mark all grid points whose difference exceeds the preset hotspot threshold as candidate hotspots; Connectivity analysis is performed on the marked candidate hotspots, and connected regions with an area greater than the preset minimum hotspot area are determined as local hotspot regions; Hotspot information is generated based on the local hotspot area, and the hotspot information includes the location coordinates and temperature distribution of the area.

6. The system according to claim 1, characterized in that, The partition compensation module includes a hotspot mapping unit, specifically used for: Receive hotspot information and obtain preset power MOS chip layout coordinates, the layout coordinates including the position range of each power MOS chip on the substrate; Spatially match the location coordinates of the hotspot areas in the hotspot information with the layout coordinates of the power MOS chips to determine the power MOS chips that fall within the range of each hotspot area and generate a list of hotspot chips. The average temperature of the chipset is calculated based on the temperature distribution in the hotspot information, and the global baseline compensation is calculated based on the average temperature of the chipset. Output the list of hotspot chips and the global baseline compensation amount.

7. The system according to claim 6, characterized in that, The partition compensation module includes multiple independent compensation signal generation units, specifically used for: Each compensation signal generation unit corresponds to a power MOS chip; The compensation signal generation unit receives a list of hotspot chips and a global reference compensation amount; If the corresponding chip is in the hot spot chip list, then the additional compensation coefficient is calculated based on the difference between the actual temperature of the hot spot where the chip is located and the reference temperature. The enhanced compensation signal is obtained by multiplying the additional compensation coefficient by the global reference compensation amount.

8. The system according to claim 7, characterized in that, The compensation signal generation unit is specifically used for: Receive the real-time temperature value of the power MOS chip obtained by the temperature field reconstruction and hotspot analysis module based on the temperature field distribution at the current moment through interpolation; Calculate the temperature difference between the real-time temperature value and the average temperature of the chipset; According to a preset compensation function, the temperature difference is linearly converted into an additional compensation voltage value; Obtain the global reference compensation amount, and extract the reference compensation voltage value from the global reference compensation amount; The additional compensation voltage value is added to the reference compensation voltage value to generate an enhanced compensation voltage signal.

9. The system according to claim 8, characterized in that, The partition compensation module also includes a compensation effect monitoring unit: Collect the drain current of each power MOS chip after applying the enhancement compensation signal to obtain the drain current monitoring value; Obtain the preset rated threshold voltage; The compensation deviation is calculated based on the drain current monitoring value and the rated threshold voltage; Correction coefficients are generated based on the compensation deviation.

10. The system according to claim 9, characterized in that, The compensation effect monitoring unit is specifically used for: The collected drain current value is used to calculate the current actual threshold voltage according to the preset conversion relationship; The actual threshold voltage is compared with the preset rated threshold voltage to obtain the threshold voltage offset; The threshold voltage offset is input into the PID controller, and the PID controller outputs an adjustment amount as a correction coefficient. The correction coefficient is fed back to the corresponding compensation signal generation unit, which is used to multiply the additional compensation voltage of the temperature difference conversion when calculating the enhanced compensation signal.

Images