System, method and computer program product for automatic extraction of dynamic parameters of power semiconductor devices

Abstract

The application provides a kind of power semiconductor device dynamic parameter automatic extraction system, method and computer program product, it is related to power semiconductor device test technical field, including obtaining the test waveform data of power semiconductor device in pulse test process;The feature time point corresponding to the switching process of power semiconductor device is identified from test waveform data;Based on each feature time point, the turn-on, turn-off and reverse recovery dynamic parameters of power semiconductor device are calculated, which are output as the dynamic parameter automatic extraction result of power semiconductor device.The beneficial effect is that through the whole-process automation, standardized data processing and analysis method, various errors caused by manual operation are effectively avoided, the accuracy and reliability of dynamic parameter extraction result are greatly improved, the actual dynamic switching characteristics of power semiconductor device can be truly and accurately reflected, and accurate and standardized data basis is provided for device selection, drive design, loss analysis and reliability evaluation.

Description

Technical Field

[0001] This invention relates to the field of power semiconductor device testing technology, and in particular to an automatic extraction system, method and computer program product for dynamic parameters of power semiconductor devices. Background Technology

[0002] Power semiconductor devices are core components of modern power electronic systems, widely used in many key areas such as new energy power generation, electric vehicles, industrial control, and variable frequency home appliances. The dynamic switching characteristics of these devices directly determine the operating efficiency, stability, and lifespan of power electronic equipment. Pulse testing is the mainstream testing method in the industry for evaluating the turn-on, turn-off, and reverse recovery dynamic characteristics of power semiconductor devices. By collecting test waveform data such as voltage and current during the device testing process, multiple core dynamic parameters can be extracted. These parameters are important bases for device selection, drive circuit design, loss analysis, and reliability assessment, and have irreplaceable guiding value for the research and development, testing, and batch quality control of power semiconductor devices.

[0003] Currently, the extraction of dynamic parameters from pulse test waveforms of power semiconductor devices in the industry generally relies on manual analysis. Engineers must manually screen test waveforms, locate various characteristic time points during the device's switching process, and then manually calculate various dynamic parameters. This processing mode is highly dependent on manual operation, resulting in time-consuming analysis of individual waveforms and extremely low overall efficiency. It is unsuitable for R&D and production scenarios involving batch testing and large-scale data analysis. Furthermore, subjective differences in judgment standards and operating habits among different operators can easily lead to deviations in characteristic point location and inconsistent parameter calculation results. This results in poor reproducibility and comparability of test data, significantly reducing the accuracy and standardization of dynamic parameter testing for power semiconductor devices.

[0004] Meanwhile, in actual pulse testing, the output waveform of a device often exhibits non-ideal characteristics due to multiple factors, including device parasitic parameters, drive circuit characteristics, test environment interference, and the device's own switching characteristics. This can lead to complex situations such as waveform oscillation, indistinct characteristic inflection points, and weak peak signals. Traditional manual identification methods struggle to accurately and uniformly identify switching characteristic timing points under complex waveforms, easily resulting in missed or misjudged feature points. This further leads to significant calculation errors in subsequent dynamic parameters such as turn-on, turn-off, and reverse recovery, failing to accurately reflect the true dynamic performance of power semiconductor devices.

[0005] Furthermore, existing technologies lack an integrated automated processing system. Waveform data acquisition, feature recognition, and parameter calculation are fragmented, requiring manual intervention throughout the process. This makes it impossible to form a fully automated processing chain from test waveform data acquisition to automatic extraction and output of dynamic parameters. Manual intervention is costly and prone to errors, making it difficult to meet the current needs of large-scale, high-precision, and standardized dynamic parameter testing and analysis for power semiconductor devices.

[0006] In summary, there is an urgent need for an automatic dynamic parameter extraction system for power semiconductor devices that can automatically acquire pulse test waveform data, accurately identify device switching characteristic time points, and automatically complete dynamic parameter calculation and extraction. This system would solve the technical problems of low efficiency, poor consistency, low accuracy of complex waveform recognition, and large parameter extraction errors in existing technologies, and achieve efficient, accurate, and standardized automatic extraction of dynamic parameters for power semiconductor devices. Summary of the Invention

[0007] To address the problems existing in the prior art, the present invention provides an automatic dynamic parameter extraction system for power semiconductor devices, comprising: The waveform acquisition module is used to acquire test waveform data of power semiconductor devices during pulse testing. A feature recognition module, connected to the waveform acquisition module, is used to identify the feature time points corresponding to the switching process of the power semiconductor device from the test waveform data; The parameter calculation module, connected to the feature recognition module, is used to calculate the turn-on, turn-off, and reverse recovery dynamic parameters of the power semiconductor device based on each of the feature time points, and output the results as the automatic extraction results of the dynamic parameters of the power semiconductor device.

[0008] Preferably, the waveform acquisition module includes: The acquisition unit is used to acquire raw waveform data from heterogeneous data sources of the power semiconductor device during the pulse test process, and parse the raw waveform data into timing data in a unified format; A mapping unit, connected to the acquisition unit, is used to map the physical quantities corresponding to different data columns in each of the time series data to the relevant physical measurement channels to obtain the test waveform data that clearly distinguishes the types of physical quantities. The physical measurement channel includes at least one type of physical quantity selected from gate voltage, collector-emitter voltage, collector current, diode voltage, and diode current.

[0009] Preferably, the system further includes a drift calibration module, which is connected to the waveform acquisition module and the feature recognition module respectively. The module is used to identify the drift reference interval in the test waveform data acquired by the waveform acquisition module that represents the power semiconductor device in a steady-state operating state. Then, the drift amount is calculated based on the drift reference interval, and the test waveform data is calibrated based on the drift amount to obtain calibrated waveform data, which is then sent to the feature recognition module for feature time point identification.

[0010] Preferably, the feature recognition module includes: The storage unit is used to store the pre-established correspondence between the physical measurement channels and the feature points to be extracted, as well as the identification strategy associated with each of the feature points to be extracted; An identification unit, connected to the storage unit, is used to determine the feature points to be extracted from the test waveform data according to the correspondence, and then match the corresponding identification strategy according to the feature points to be extracted, and then use the identification strategy to identify the feature time points corresponding to the switching process of the power semiconductor device.

[0011] Preferably, the parameter calculation module includes: The extraction unit is used to construct an analysis interval using the feature time points as anchor points; The analysis unit, connected to the extraction unit, is used to perform data analysis within the analysis interval to obtain the turn-on, turn-off, and reverse recovery dynamic parameters of the power semiconductor device.

[0012] Preferably, the data analysis includes at least one of integration, extremum search, difference operation, and sliding window scanning.

[0013] Preferably, it further includes a visualization module, which is connected to the feature recognition module and the parameter calculation module respectively. The visualization module includes: A range adaptation unit is used to match and display the range according to the maximum measured value corresponding to the test waveform data; The display unit, connected to the range adaptation unit, is used to synchronously display test waveform data of different physical quantities on the same image using a multi-Y-axis method based on the display range, and to automatically mark the characteristic time points and the on-off and reverse recovery dynamic parameters on the test waveform data.

[0014] Preferably, it also includes a graphical interaction module for users to perform initial configuration before dynamic parameters are automatically extracted.

[0015] This invention also provides an automatic method for extracting dynamic parameters of power semiconductor devices, using the aforementioned automatic extraction system for dynamic parameters of power semiconductor devices, the method comprising: Step S1: Obtain test waveform data of the power semiconductor device during the pulse test process; Step S2: Identify the characteristic time points corresponding to the switching process of the power semiconductor device from the test waveform data; Step S3: Calculate the turn-on, turn-off, and reverse recovery dynamic parameters of the power semiconductor device based on each of the characteristic time points, and output them as the automatic extraction results of the dynamic parameters of the power semiconductor device.

[0016] The present invention also provides a computer program product, including computer program instructions, which, when executed by a processor, implement the above-described method.

[0017] The above technical solution has the following advantages or beneficial effects: 1) By automatically acquiring test waveform data during the pulse test of power semiconductor devices, the traditional manual import, screening, and sorting of waveform data operation mode is eliminated. No manual data preprocessing is required, which greatly reduces the cost of manual operation and the probability of human error. At the same time, the standardized and stable automatic acquisition of test waveform data is realized, which can provide complete and effective raw data support for subsequent feature recognition and parameter calculation, ensuring the continuity and stability of the overall parameter extraction work. 2) Based on test waveform data, the system automatically identifies the characteristic time points corresponding to the device's turn-on, turn-off, and reverse recovery processes, replacing the traditional identification method of manual visual judgment and manual cursor positioning. This unifies the judgment criteria for characteristic point identification and completely avoids problems such as missed, misjudged, and mispositioned characteristic points caused by differences in subjective judgment among different operators and deviations in manual operation. It significantly improves the accuracy of switch characteristic time point identification and ensures the consistency and reproducibility of test data from multiple batches and scenarios. 3) Based on accurately identified feature time points, the system automatically completes the calculation and output of various core dynamic parameters for device turn-on, turn-off, and reverse recovery. The entire process requires no manual calculation, verification, or data processing, which completely solves the defects of low efficiency and easy calculation errors in traditional manual calculation. It greatly improves the overall efficiency of dynamic parameter extraction and can be adapted to batch testing and analysis scenarios in the R&D and production process of power semiconductor devices.

[0018] 4) Through fully automated and standardized data processing and analysis, various errors caused by manual operation are effectively avoided, and the accuracy and reliability of dynamic parameter extraction results are greatly improved. It can truly and accurately reflect the actual dynamic switching characteristics of power semiconductor devices, and provide accurate and standardized data basis for device selection, drive design, loss analysis and reliability assessment. Attached Figure Description

[0019] Figure 1A schematic diagram of the structure of an automatic dynamic parameter extraction system for power semiconductor devices is shown in a preferred embodiment of the present invention. Figure 2 In a preferred embodiment of the present invention, a schematic diagram is shown in which test waveform data of different physical quantities are simultaneously displayed on the same image using a multi-Y-axis method; Figure 3 This is a flowchart illustrating a method for automatically extracting dynamic parameters of a power semiconductor device, as described in a preferred embodiment of the present invention. Detailed Implementation

[0020] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment; other embodiments that conform to the spirit of the present invention may also fall within the scope of the present invention.

[0021] In a preferred embodiment of the present invention, based on the above-mentioned problems existing in the prior art, an automatic dynamic parameter extraction system for power semiconductor devices is provided. It is applicable to single-pulse, double-pulse and multi-pulse test scenarios. It can perform fully automated analysis of the turn-on, turn-off and reverse recovery dynamic characteristics of various power semiconductor devices such as IGBT, SiCMOSFET, and diodes. It effectively solves the technical defects of traditional manual analysis, such as low efficiency, inconsistent standards, poor identification accuracy and weak data reproducibility. It realizes integrated fully automated processing of test waveform data from acquisition, calibration, feature recognition, parameter calculation to visualization display and result output.

[0022] like Figure 1 As shown, the automatic extraction system for dynamic parameters of power semiconductor devices of the present invention includes a waveform acquisition module 1, a feature recognition module 2, and a parameter calculation module 3. It can also be equipped with a drift calibration module 4, a visualization module 5, and a graphical interaction module 6. The modules work in a modular and collaborative manner, without the need for full manual intervention. Only basic initialization configuration needs to be completed before testing, and the accurate extraction of dynamic parameters of power semiconductor devices can be completed automatically.

[0023] Among them, waveform acquisition module 1 is the data input terminal of this system, used to acquire test waveform data of power semiconductor devices during pulse testing.

[0024] In a preferred embodiment of the present invention, the waveform acquisition module 1 includes: The acquisition unit 11 is used to acquire raw waveform data from heterogeneous data sources of power semiconductor devices during pulse testing, and to parse the raw waveform data into timing data in a unified format. The mapping unit 12 is connected to the acquisition unit 11 and is used to map the physical quantities corresponding to different data columns in each time series data to the relevant physical measurement channels to obtain test waveform data that clearly distinguishes the types of physical quantities. The physical measurement channels include at least one type of physical quantity among gate voltage (Vge), collector-emitter voltage (Vce), collector current (Ic), diode voltage (Vf), and diode current (If).

[0025] Specifically, this invention is compatible with various data sources, including data files exported from external testing devices such as oscilloscopes and data acquisition cards, and real-time data streams acquired through communication interfaces such as GPIB, USB, LAN, and LXI. It is also compatible with local historical data, network-stored data, cloud-stored synchronized data, and simulation data output from simulation software such as SPICE, Saber, and PLECS, completely solving the problems of poor compatibility and highly heterogeneous data formats in traditional analysis tools and equipment. The acquisition unit 11 has built-in multiple parsing engines, compatible with mainstream data formats such as CSV, TXT, XLS, XLSX, and BIN, unifying and standardizing data structures, time formats, and encoding methods to ensure the standardization of data processing from the source.

[0026] Furthermore, considering the differences in naming conventions for waveform data columns from different data sources, it may not be possible to directly determine the waveform type of each channel based solely on the original data column names. Therefore, after standardizing the format, channel mapping is also included to clearly distinguish the physical quantity types of each physical measurement channel for subsequent data retrieval. Preferably, channel mapping can be implemented through three methods: manual configuration via the user interface, automatic matching of preset templates, and intelligent recognition of data column keywords. This adapts to different channel configurations and usage scenarios of different test equipment models, significantly improving the system's adaptability and flexibility.

[0027] The preset template can be configured by default with the first channel as gate voltage (Vge), the second channel as collector-emitter voltage (Vce), the third channel as diode voltage (Vf), the fourth channel as collector current (Ic), the fifth channel as diode current (If), and the sixth channel as gate voltage of the transistor, so that users can fix the above channel order during testing.

[0028] In actual testing, current probes and high-voltage differential probes are susceptible to zero-point drift due to prolonged testing, changes in ambient temperature, and electromagnetic interference. Without calibration, this can lead to systematic deviations in voltage and current measurement data, ultimately resulting in distorted dynamic parameter calculations. Therefore, this invention also includes a drift calibration module 4, connected to the waveform acquisition module 1 and the feature recognition module 2. This module identifies the drift reference interval in the test waveform data acquired by the waveform acquisition module, representing the steady-state operation of the power semiconductor device. The drift amount is then calculated based on this reference interval, and the test waveform data is calibrated based on the drift amount. The calibrated waveform data is then sent to the feature recognition module for feature time point identification.

[0029] Specifically, the aforementioned drift reference range is preferably a stable range without switching transients or waveform oscillations. For different physical measurement channels, the selection logic for the drift reference range is adapted differently as follows: 1) For the collector current Ic measurement channel, the steady-state interval from one-quarter pulse width after the first turn-off time to one-quarter pulse width before the second turn-on time is selected as the drift reference interval to avoid the influence of switching transients.

[0030] 2) For the diode current If measurement channel, the steady-state interval from half the pulse width after the second turn-on time to one-quarter of the pulse width before the second turn-off time is selected as the drift reference interval to avoid the influence of waveform oscillation; 3) For voltage measurement channels such as Vce and Vf, the steady-state interval from one-quarter pulse width after the first turn-on time to one-quarter pulse width before the first turn-off time is selected as the drift reference interval.

[0031] After determining the drift reference interval, the arithmetic mean, median, mode, and other statistical characteristics of all sampling points within the drift reference interval are selected, with the arithmetic mean being preferred as the final drift amount. Subsequently, a global drift subtraction operation is performed on the entire waveform time series data, and data calibration is completed using the formula x[i] = x[i] - d (d is the drift amount, i is the sampling point index). High-precision, baseline-offset-free calibrated waveform data is output and sent to feature recognition module 2.

[0032] More preferably, each physical measurement channel is equipped with an independent on / off calibration function, which users can configure according to the probe accuracy and test accuracy requirements. For example, calibration can be turned off for high-quality probes and turned on for aging probes or high-precision tests, balancing test flexibility and accuracy.

[0033] In this embodiment, the feature recognition module 2 is connected to the waveform acquisition module 1 and is used to identify the feature time points corresponding to the switching process of the power semiconductor device from the calibrated test waveform data.

[0034] In a preferred embodiment of the present invention, the feature recognition module 2 includes: Storage unit 21 is used to store the pre-established correspondence between physical measurement channels and feature points to be extracted, as well as the identification strategy associated with each feature point to be extracted; The identification unit 22 is connected to the storage unit 21 and is used to determine the feature points to be extracted from the test waveform data according to the correspondence. Then, it matches the corresponding identification strategy according to the feature points to be extracted and then uses the identification strategy to identify the feature time points corresponding to the switching process of the power semiconductor device.

[0035] Specifically, in this embodiment, the feature points to be extracted include, but are not limited to, the core feature points of the device switch such as k1, g1, k2, g2, Vcp, Icp, Ic, Irrm, and current tail position. The above correspondence includes: the Vge waveform corresponds to the extraction of gate switch feature points k1 (start point of the first rising edge), g1 (end point of the first falling edge), k2 (start point of the second rising edge), and g2 (end point of the second falling edge); the Ic waveform corresponds to the extraction of Icp, steady-state Ic, and current tail position; the Vce waveform corresponds to the extraction of Vcp turn-off peak feature point; and the If waveform corresponds to the extraction of Irrm reverse recovery peak feature point.

[0036] Furthermore, the recognition unit 22 can automatically match the feature points to be extracted according to the physical channel corresponding to the waveform, and then adaptively call a multi-level, multi-method composite recognition strategy to accurately locate the feature time point. Specifically, this invention integrates three core recognition methods: threshold comparison method, coarse-fine positioning strategy, and slope detection method, and adapts them to different feature points to solve the problem that traditional single recognition methods are easily affected by noise and waveform distortion, leading to misjudgment and missed judgment of feature points.

[0037] The threshold comparison method is mainly used for identifying gate switch feature points, determining the start and end times of the switch based on preset proportional thresholds. The system dynamically calculates thresholds based on user-input gate positive and negative voltage parameters. The turn-on threshold k1 is selected as 0.5-0.9 times the gate negative voltage value (preferably 2 / 3 times), and the turn-off threshold g1 is selected as 0.5-0.9 times the gate positive voltage value (preferably 4 / 5 times). If the gate negative voltage is zero, it automatically switches between 1V and 2V fixed thresholds to adapt to different test conditions. By comparing waveform sampling points with preset thresholds, gate switch feature points such as k1, g1, k2, and g2 are accurately located.

[0038] The coarse-fine positioning strategy is mainly used to avoid waveform oscillations and noise interference, and to adapt to the identification of complex feature points such as current tailing. Taking the identification of the second turn-on time as an example, coarse positioning is first completed 2μs after the first turn-off time to avoid interference from gate negative voltage oscillation after turn-off. Then, starting from the coarse positioning point, a fine search is conducted on the sampling point that first exceeds the turn-on threshold to accurately determine the second turn-on time, which greatly improves the identification stability under complex waveforms.

[0039] The slope detection method calculates the first derivative of the waveform and identifies the zero-crossing point of the derivative to accurately locate the inflection point and peak point of the waveform. It is mainly suitable for peak-type feature point recognition such as Icp, Irrm, and Vcp, and effectively solves the problem of feature point recognition deviation in non-ideal waveforms, weak peaks, and high-frequency oscillation scenarios.

[0040] The entire recognition process is fully automated, with unified recognition standards, completely eliminating subjective human judgment bias and ensuring the consistency and reproducibility of feature point localization across multiple batches of tests.

[0041] In this embodiment, the parameter calculation module 3 is connected to the feature recognition module 2 and is used to calculate the dynamic parameters of the power semiconductor device's turn-on, turn-off, and reverse recovery based on each feature time point, and output the results of the automatic extraction of dynamic parameters of the power semiconductor device.

[0042] In a preferred embodiment of the present invention, the parameter calculation module 3 includes: Extraction unit 31 is used to construct analysis intervals using feature time points as anchor points; The analysis unit 32 is connected to the extraction unit 31 and is used to perform data analysis within the analysis interval to obtain the turn-on, turn-off and reverse recovery dynamic parameters of the power semiconductor device.

[0043] The extraction unit 31 uses each feature time point as the core anchor point and combines the fixed offset method and the dynamic threshold method to automatically construct a dedicated analysis interval, avoiding interference from switching noise, waveform distortion, and steady-state intervals, thus ensuring the accuracy of the analysis interval. Specifically, the fixed offset method uses a preset time offset to avoid oscillation interference before and after the switching transient and delineates a stable analysis interval, while the dynamic threshold method adaptively finds the threshold inflection point based on the waveform morphology, adapting to non-ideal waveforms with different degrees of distortion.

[0044] Within the defined precise analysis interval, analysis unit 32 automatically completes the calculation of dynamic parameters for all product categories through four core data analysis algorithms: integration (e.g., Eon / Erec, calculating ∫(Ic ×Vce) dt), extreme value search (e.g., Icp / Irrm, searching for the array max), difference operation (e.g., di / dt, calculating ΔI / Δt), and sliding window scanning (e.g., maximum di / dt, calculating the maximum value by sliding within the interval with a 1ns step and a 10ns window). The calculation logic of each parameter closely matches the actual switching physical characteristics of the device, specifically including the following three categories: 1. Activation parameters The activation parameters are calculated based on the activation waveforms of the second and subsequent pulses, avoiding the problems of unstable waveforms and inability to establish a steady state during the first pulse. Specifically, these parameters include: 1) Turn-on peak current Icp: Defined as the maximum value of the collector current during the turn-on process, which usually occurs during the reverse recovery phase. When the reverse recovery peak is too small, the calculation range can be adjusted by selecting on the main interface to avoid misjudgment.

[0045] Specifically, the extraction unit 31 locks the first steady-state interval after Ic rises from 10% steady-state current to the peak value. The analysis unit 32 traverses the sampled values ​​point by point within this interval and searches for the maximum current value as Icp, effectively filtering out false peak values ​​caused by early oscillation spikes.

[0046] 2) Turn-on delay time tdon: Defined as the time difference between the moment the gate voltage rises to a preset percentage (e.g., 10% of the positive gate voltage Vge_a) and the moment the collector current rises to a first current threshold (e.g., 10% Ica, where Ica is the turn-off current of the previous pulse, which the program will automatically read). It accurately characterizes the device's turn-on response delay. The calculation formula is as follows: tdon = t(Ic > 10% Ic_a) - t(Vge > 90% Vge_a) 3) Current rise time tr: Defined as the time difference between the collector current rising from the first current threshold (10% Ica) to the second current threshold (90% Ica), to characterize the device's turn-on current rise rate characteristics. The calculation formula is as follows:

[0047] 4) Turn-on loss Eon: Defined as the integral of the product of collector-emitter voltage and collector current over time during the turn-on process. The integration interval typically extends from the start of the collector current increase (e.g., current rises to 10% Ica) to the end of the turn-on transient (e.g., voltage drops to 2% Vce, where Vce is the voltage value input to the interface). The integral formula is Eon = ∫(Vce(t) × Ic(t))dt. By continuously integrating the transient voltage and current product, the dynamic loss of the device during the entire turn-on process can be accurately calculated, eliminating the interference of steady-state conduction loss.

[0048] Alternatively, the cumulative method can be used for calculation, and the formula is as follows: k1 is 10% Ica, k2 is 2% Vce, and Δt is the sampling interval time, which the program will automatically obtain from the waveform.

[0049] 5) Average rate of change of turn-on current di / dt(on): Defined as the average rate of change of collector current during the turn-on process. Starting from turn-on time k1, find the first point Ip_1 where the current is greater than 50% of Ica. Then, starting from that point, find the first point Ip_2 where the current is greater than 50% of (Icp-Ica) + Ica. The time from Ip_1 to Ip_2 is t. The calculation formula is:

[0050] 6) Maximum rate of change of on-current di / dt(on)max: The maximum rate of change can be calculated using the sliding window method. The sliding window method uses a step size of 1ns and a window width of 10ns. It calculates the average rate of change within the specified interval Ip1 to Ip2, and takes the maximum value as the final result.

[0051] 7) Average turn-on voltage change rate du / dt(on): Defined as the average rate of change of collector-emitter voltage during the turn-on process. Starting from the turn-on time k1, the first point less than 90% Vce is taken, ending at the first point less than 10% Vce. (Vce is the voltage value input from the interface). The voltage difference in this segment is divided by the time difference as the result.

[0052] 8) Maximum turn-on voltage change rate du / dt(on)max: Defined as the rate of change of collector-emitter voltage during turn-on. The sliding window method uses a step size of 1ns and a window width of 10ns to calculate the average rate of change within the specified interval from 90%Vce to 10%Vce, and takes the maximum value as the final result.

[0053] 2. Shutdown parameters The turn-off parameters are calculated based on the turn-off waveforms of each pulse, with a focus on adaptive optimization to address the non-ideal characteristics of device turn-off current tailing and voltage spike oscillations. Specifically, this includes: 1) Turn-off peak voltage Vcp: defined as the maximum value of the collector-emitter voltage during the turn-off process.

[0054] The preferred approach is to use adaptive interval extreme value search to lock the complete range from Vce rising from 10% bus voltage to the overshoot peak and falling back to the steady state. The maximum voltage value is extracted by traversing the sampling points, which effectively identifies the turn-off peak voltage and reflects the device's turn-off overvoltage characteristics.

[0055] 2) Turn-off delay time tdoff: Defined as the time difference between the moment when the collector current drops to a specified threshold (e.g., 90% of Ica) from point Ica and the moment when the gate voltage drops to a specified threshold (e.g., 90% of the positive gate voltage Vgea), precisely matching the device's turn-off trigger delay characteristics. Calculation formula:

[0056] 3) Fall time tf: Defined as the time difference between the moment when the collector current drops to a specified threshold (e.g., 90% Ica) from point Ica backwards, and the moment when the gate voltage drops to a specified threshold (e.g., 10% Ica). It characterizes the device's turn-off current turn-off rate. Calculation formula:

[0057] 4) Turn-off loss Eoff: Defined as the integral of the product of collector-emitter voltage and collector current over time during the turn-off process. The integration interval typically extends from the start of the collector-emitter voltage rise (e.g., voltage rises to 10% Vce) to the end of the turn-off transient (e.g., collector current decays to 2% Ica). The transient loss is calculated by time-domain integration of Eoff = ∫(Vce(t) × Ic(t))dt, accurately calculating the dynamic loss throughout the entire turn-off process.

[0058] The optimal method can also be calculated using the summation method. The calculation formula is as follows: Δt is the sampling interval time, which the program will automatically obtain from the waveform.

[0059] 5) Average turn-off current change rate di / dt(off): Defined as the rate of change of collector current during turn-off. The calculation threshold is 90%Ica point - 10%Ica point of the turn-off waveform. Considering the possible tailing phenomenon of the turn-off current, the current change rate is also calculated with a threshold of 90%Ica point - 50%Ica point. The calculation method is the same as the turn-on voltage change rate.

[0060] 6) Maximum rate of change of turn-off current di / dt(off)max: The calculation method is based on the maximum rate of change of turn-on current, with the window moving between 90% Ica and 10% Ica of the turn-off waveform.

[0061] 7) Turn-off voltage change rate du / dt(off): Defined as the rate of change of collector-emitter voltage during turn-off. The calculation method is the same as the average turn-on change rate. The calculation threshold is 90%Vce point - 10%Vce point of the turn-off waveform.

[0062] 8) Turn-off voltage change rate du / dt(off) max: Defined as the rate of change of collector-emitter voltage during turn-off. The calculation method is the same as the maximum turn-on change rate. The calculation threshold is 90%Vce point - 10%Vce point of the turn-off waveform.

[0063] To address the common current tailing problem in turn-off waveforms, the system can automatically identify waveform inflection points and adaptively switch analysis intervals: the system automatically identifies whether there is tailing distortion in the Ic falling waveform. If there is no tailing, the system calculates di / dt(off) using the 90%Ic to 10%Ic interval; if a tailing feature of slow current decay is detected, the system automatically switches to the 90%Ic to 50%Ic interval to calculate the rate of change in the effective falling interval, completely avoiding parameter distortion and rate underestimation problems caused by the smooth tailing segment.

[0064] 3. Reverse recovery parameters When the device under test includes a diode, the reverse recovery parameters are calculated based on the diode If and Vf waveforms to adapt to the physical process of diode reverse recovery during the turn-on phase of the power device.

[0065] The system first precisely locates the moment when the If current crosses zero as the starting anchor point for reverse recovery. From this moment, it extracts the entire reverse recovery process interval until the reverse current decays to 2% of the Irrm amplitude as the termination boundary.

[0066] The reverse recovery parameters specifically include: 1) Reverse recovery peak current Irrm: Defined as the maximum absolute value of the diode current during reverse recovery. In the If waveform, find the maximum value of the reverse current after the forward current crosses zero.

[0067] 2) Reverse recovery voltage Vrp: Defined as the maximum value of the diode voltage during the reverse recovery process. Find the maximum voltage value during the reverse recovery period in the Vf waveform.

[0068] 3) Reverse recovery time trr: Defined as the time difference between the moment the diode current crosses zero and the moment it decays to a specified current threshold (e.g., 2% Irrm). Calculation formula:

[0069] 4) Reverse recovery charge Qr: Defined as the integral of the diode current with respect to time during the reverse recovery process. The integration interval extends from when the diode current crosses zero to when it decays to near zero (e.g., 2% Irrm). Calculation formula: .

[0070] 5) Reverse recovery loss Erec: Defined as the integral of the product of diode voltage and current over time during the reverse recovery process. The integration interval is the same as Qr. It can also be calculated using the accumulation method, with the following formula: .

[0071] In summary, the parameter calculation module relies entirely on refined, differentiated, and adaptive proprietary calculation logic. Only basic fixed parameters such as bus voltage, gate positive and negative voltage, and steady-state current need to be configured before testing. After starting the analysis, the entire process of interval delineation, algorithm calculation, and parameter output is completed automatically without manual intervention or experience correction. This completely solves the defects of traditional manual measurement, such as chaotic standards, low accuracy, poor adaptability of trailing waveforms, inaccurate capture of transient extrema, and misalignment of integration intervals. It can perfectly adapt to batch, high-precision, and standardized device dynamic parameter testing scenarios.

[0072] In a preferred embodiment of the present invention, a visualization module 5 is further included, which is connected to the feature recognition module 2 and the parameter calculation module 3 respectively. The visualization module 5 includes: The range adaptation unit 51 is used to match the display range according to the maximum measured value corresponding to the test waveform data; The display unit 52 is connected to the range adapter unit 51. It is used to synchronously display test waveform data of different physical quantities on the same image using a multi-Y-axis method based on the display range, and automatically mark characteristic time points and on / off and reverse recovery dynamic parameters on the test waveform data.

[0073] Specifically, in this embodiment, the range adaptation unit 51 automatically adapts to the optimal display range based on the measured maximum values ​​of the voltage and current of the test waveform and according to the engineering general 1-2-5 base range classification table. This covers multiple current ranges from 12A, 30A, 60A, 120A to 6000A, and multiple voltage ranges from 12V, 30V to 6000V. This avoids waveform display overflow, compression, and distortion problems, ensuring that waveforms of various amplitudes can be displayed completely and clearly. At the same time, it supports users to manually force the specification of the range to adapt to personalized display needs.

[0074] The preferred current range classification is as follows: Ic measured maximum value range adapts to display range 0A < Ic ≤ 5A 12A 5A < Ic ≤ 15A 30A 15A < Ic ≤ 30A 60A 30A < Ic ≤ 60A 120A 60A < Ic ≤ 150A 300A 150A < Ic ≤ 300A 600A 300A < Ic ≤ 600A 1200A 600A < Ic ≤ 900A 2000A 900A < Ic ≤ 1500A 3000A Ic > 1500A 6000A Similarly, the voltage range classification includes the following range values: 12V, 30V, 60V, 120V, 300V, 600V, 1200V, 2000V, 3000V, and 6000V.

[0075] Furthermore, the display unit 52, based on the adapted standard range, adopts, as follows: Figure 2 The multi-Y-axis synchronous display technology shown displays waveforms of various physical quantities such as Vge, Vce, Ic, and If in layers within the same image, distinguishing each curve with differentiated colors to intuitively present the waveform changes throughout the device's switching process. Simultaneously, it automatically completes full-dimensional annotation, including vertical lines marking feature time points, horizontal dashed lines marking thresholds, border annotations for analysis intervals, text annotations for dynamic parameter results, legends, axis labels, and test condition titles, achieving integrated visualization of waveforms, features, intervals, and parameters. The module supports high-definition output in multiple formats such as PNG, SVG, PDF, and EPS, meeting the needs of various scenarios including scientific research archiving, report presentation, and paper publication.

[0076] In a preferred embodiment of the present invention, a graphical interaction module 6 is further included, which allows the user to perform initialization configuration before the dynamic parameters are automatically extracted.

[0077] Specifically, in this embodiment, the graphical interaction module 6 is only used for initial configuration before the automatic extraction of dynamic parameters. It does not interfere with the automatic analysis process of the system throughout the entire process, thus reducing human intervention and enhancing automation performance. This module builds a visual human-computer interaction interface, requiring no programming skills from the user, and significantly reducing the threshold for test analysis operations.

[0078] The user interface employs a dual-page layout, comprising a main configuration page and an advanced configuration page. The main configuration page provides basic configuration functions such as file path, output path, test condition parameters, oscilloscope type, and anomaly mode selection, supporting batch file selection and one-click analysis startup. The advanced configuration page offers personalized configuration functions such as channel mapping, coordinate range customization, drift calibration switch, screenshot options, and report generation options, adapting to different testing equipment, accuracy requirements, and analysis habits. All configurations are only initial setup settings; the system runs automatically after analysis begins, requiring no continuous manual intervention.

[0079] In summary, this invention, through modular architecture design, constructs a fully closed-loop automated processing system encompassing data acquisition, error calibration, intelligent feature recognition, high-precision parameter calculation, visualization, and automatic output. Compared to traditional manual analysis and single-script tools, it represents a significant technological advancement: First, it achieves universal adaptation of heterogeneous data, resolving device format incompatibility issues; second, it eliminates system errors through drift calibration, greatly improving data accuracy; third, it employs a multi-method composite recognition strategy to adapt to various non-ideal, distorted, and noisy waveforms, significantly improving feature recognition stability and accuracy; fourth, it achieves fully automated computation without manual intervention, greatly improving batch testing efficiency; and fifth, it integrates waveform visualization and parameter analysis, solving the problems of fragmented processes and cumbersome operations in traditional methods, providing standardized, high-precision, and reproducible data support for power semiconductor device R&D, selection, performance evaluation, loss design, and reliability analysis.

[0080] This invention also provides an automatic method for extracting dynamic parameters of power semiconductor devices, which utilizes the aforementioned automatic extraction system for dynamic parameters of power semiconductor devices, such as... Figure 3 As shown, the method includes: Step S1: Obtain test waveform data of the power semiconductor device during the pulse test process; Step S2: Identify the characteristic time points corresponding to the switching process of the power semiconductor device from the test waveform data; Step S3: Calculate the turn-on, turn-off, and reverse recovery dynamic parameters of the power semiconductor device based on each characteristic time point, and output the results as the automatic extraction results of the dynamic parameters of the power semiconductor device.

[0081] The present invention also provides a computer program product, including computer program instructions, which, when executed by a processor, implement the above-described method.

[0082] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. An automatic dynamic parameter extraction system for power semiconductor devices, characterized in that, include: The waveform acquisition module is used to acquire test waveform data of power semiconductor devices during pulse testing. A feature recognition module, connected to the waveform acquisition module, is used to identify the feature time points corresponding to the switching process of the power semiconductor device from the test waveform data; The parameter calculation module, connected to the feature recognition module, is used to calculate the turn-on, turn-off, and reverse recovery dynamic parameters of the power semiconductor device based on each of the feature time points, and output the results as the automatic extraction results of the dynamic parameters of the power semiconductor device.

2. The automatic extraction system for dynamic parameters of power semiconductor devices according to claim 1, characterized in that, The waveform acquisition module includes: The acquisition unit is used to acquire raw waveform data from heterogeneous data sources of the power semiconductor device during the pulse test process, and parse the raw waveform data into timing data in a unified format; A mapping unit, connected to the acquisition unit, is used to map the physical quantities corresponding to different data columns in each of the time series data to the relevant physical measurement channels to obtain the test waveform data that clearly distinguishes the types of physical quantities. The physical measurement channel includes at least one type of physical quantity selected from gate voltage, collector-emitter voltage, collector current, diode voltage, and diode current.

3. The automatic extraction system for dynamic parameters of power semiconductor devices according to claim 1, characterized in that, It also includes a drift calibration module, which is connected to the waveform acquisition module and the feature recognition module respectively. The module is used to identify the drift reference interval in the test waveform data acquired by the waveform acquisition module that represents the power semiconductor device in a steady-state operating state. Then, the drift amount is calculated based on the drift reference interval, and the test waveform data is calibrated based on the drift amount to obtain calibrated waveform data, which is then sent to the feature recognition module for feature time point identification.

4. The automatic extraction system for dynamic parameters of power semiconductor devices according to claim 2, characterized in that, The feature recognition module includes: The storage unit is used to store the pre-established correspondence between the physical measurement channels and the feature points to be extracted, as well as the identification strategy associated with each of the feature points to be extracted; An identification unit, connected to the storage unit, is used to determine the feature points to be extracted from the test waveform data according to the correspondence, and then match the corresponding identification strategy according to the feature points to be extracted, and then use the identification strategy to identify the feature time points corresponding to the switching process of the power semiconductor device.

5. The automatic extraction system for dynamic parameters of power semiconductor devices according to claim 1, characterized in that, The parameter calculation module includes: The extraction unit is used to construct an analysis interval using the feature time points as anchor points; The analysis unit, connected to the extraction unit, is used to perform data analysis within the analysis interval to obtain the turn-on, turn-off, and reverse recovery dynamic parameters of the power semiconductor device.

6. The automatic dynamic parameter extraction system for power semiconductor devices according to claim 5, characterized in that, The data analysis includes at least one of integration, extremum search, difference operation, and sliding window scanning.

7. The automatic extraction system for dynamic parameters of power semiconductor devices according to claim 2, characterized in that, It also includes a visualization module, which is connected to the feature recognition module and the parameter calculation module respectively. The visualization module includes: A range adaptation unit is used to match and display the range according to the maximum measured value corresponding to the test waveform data; The display unit, connected to the range adaptation unit, is used to synchronously display test waveform data of different physical quantities on the same image using a multi-Y-axis method based on the display range, and to automatically mark the characteristic time points and the on-off and reverse recovery dynamic parameters on the test waveform data.

8. The automatic dynamic parameter extraction system for power semiconductor devices according to claim 1, characterized in that, It also includes a graphical user interface module for users to perform initial configuration before dynamic parameters are automatically extracted.

9. A method for automatically extracting dynamic parameters of power semiconductor devices, characterized in that, The method, applied to the automatic dynamic parameter extraction system for power semiconductor devices as described in any one of claims 1-9, comprises: Step S1: Obtain test waveform data of the power semiconductor device during the pulse test process; Step S2: Identify the characteristic time points corresponding to the switching process of the power semiconductor device from the test waveform data; Step S3: Calculate the turn-on, turn-off, and reverse recovery dynamic parameters of the power semiconductor device based on each of the characteristic time points, and output them as the automatic extraction results of the dynamic parameters of the power semiconductor device.

10. A computer program product comprising computer program instructions, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in claim 9.

Images