A method and system for rapid detection of defects in field effect transistors

Abstract

This invention discloses a rapid defect detection method and system for field-effect transistors (FETs), relating to the field of defect detection technology. The rapid defect detection system for FETs includes: a modal confirmation module, an online detection module, a perturbation detection module, a coupling prediction module, a defect correlation module, and a defect result module. This invention constructs a multimodal correlated defect indicator based on the correlation between the prediction consistency residual and first-type and second-type feature parameters, achieving a fusion expression of electrical behavior, spectral characteristics, and thermal response consistency information, thereby improving the stability and separability of defect diagnosis.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology, and in particular to a rapid defect detection method and system for field-effect transistors. Background Technology

[0002] Field-effect transistors (FETs), as key switching devices in power electronic devices, are widely used in new energy vehicle drives, industrial frequency converters, power conversion, and communication power supplies. Their performance and reliability directly affect system efficiency and safety. To ensure device quality and operational reliability, engineering typically requires defect detection and condition assessment of FETs during production testing or equipment operation and maintenance phases to identify potential issues such as gate dielectric damage, abnormal conduction channels, parasitic parameter drift, and degradation of the package thermal path.

[0003] Existing MOSFET defect detection methods mostly rely on static electrical parameter testing or single-mode dynamic testing. For example, they screen for defects using parameters such as threshold voltage, leakage current, and on-resistance, or determine defects based on a single characteristic among switching waveforms, noise spectra, and thermal transient responses. However, these methods typically suffer from problems such as independent detection modes and a lack of cross-verification of results. On the one hand, early or latent defects are not readily apparent in static parameters, making single-mode testing prone to missed detections. On the other hand, waveform disturbances, noise interference, and thermal environment changes under different operating conditions and loads introduce uncertainties, making single-mode indicators prone to false alarms. Furthermore, if multi-mode testing is simply performed in parallel, it is often difficult to establish consistent constraints between electrical characteristics and thermal responses, making it difficult to ensure both detection speed and reliability. Summary of the Invention

[0004] The core technical problem solved by this invention is how to achieve rapid detection of defects in field-effect transistors (FETs) without significantly increasing testing complexity and stress, and adapting to different operating conditions. Furthermore, by establishing a predictive consistency verification mechanism between electrical equivalent parameters and thermal response, information from different detection modes is correlated and fused to improve the detection capability of early-stage latent defects and reduce the risk of false alarms and missed detections. To achieve the above objectives, the technical solution adopted by this invention is as follows: A rapid defect detection method for field-effect transistors, comprising: Obtain the operating status information of the field-effect transistor under test, including the drive signal type, switching frequency, bus voltage, and load status; The execution mode of the detection mode is determined and the detection is performed based on the working status information. The detection modes include online detection mode, idle perturbation detection mode, and thermal transient detection mode. In the online detection mode, the electrical response signal of the field-effect transistor is synchronously sampled based on the driving signal of the field-effect transistor, the first type of feature parameters are extracted, and the first electrical equivalent parameter set is estimated from the first type of feature parameters; In the idle perturbation detection mode, a small excitation signal is applied to the field-effect transistor and the electrical response signal is acquired. The second type of characteristic parameters are extracted by spectrum analysis, and the second electrical equivalent parameter set is estimated from the second type of characteristic parameters. The electro-thermal coupling prediction model is updated based on the first and second sets of electrical equivalent parameters. The electro-thermal coupling prediction model is used to generate predictive features of temperature-related response signals under thermal transient detection mode. In the thermal transient detection mode, a preset power pulse is applied to the field-effect transistor and the temperature-related response signal is collected to extract the third type of feature parameters. The prediction consistency residual between the third type of feature parameters and the predicted features is calculated, and a multimodal associated defect indicator is constructed based on the correlation between the prediction consistency residual and the first type of feature parameters and the second type of feature parameters. Input the multimodal associated defect indicator into the defect discrimination model, and output the defect detection results of the field-effect transistor.

[0005] As a preferred embodiment of the present invention, the determination of the detection mode includes: judging the current operating state of the field-effect transistor under test based on the operating state information, and determining at least one detection mode among the online detection mode, the idle perturbation detection mode, and the thermal transient detection mode according to the operating state; executing the online detection mode when the drive signal type is periodic switching drive and the switching frequency is within the preset operating frequency range; executing the idle perturbation detection mode when the load state meets the preset idle condition or the bus voltage is within the preset stable range; and executing the thermal transient detection mode when thermal characteristic detection or consistency verification of the field-effect transistor is required.

[0006] As a preferred embodiment of the present invention, the acquisition of the first electrical equivalent parameter set includes: in the online detection mode, synchronously sampling the electrical response signal of the field-effect transistor under test with the driving signal as the synchronization reference, and extracting a first type of feature parameters from the electrical response signal. The first type of feature parameters includes time feature parameters, voltage change feature parameters, or current change feature parameters related to the switching process of the field-effect transistor; establishing a correspondence between the switching behavior of the field-effect transistor and the electrical equivalent parameters based on the first type of feature parameters, and calculating the equivalent parameters characterizing the electrical characteristics of the field-effect transistor based on the correspondence, and combining the equivalent parameters to form the first electrical equivalent parameter set.

[0007] As a preferred embodiment of the present invention, the acquisition of the second electrical equivalent parameter set includes: applying a small-amplitude excitation signal to the field-effect transistor under test in the idle perturbation detection mode, acquiring the electrical response signal of the field-effect transistor, performing spectral analysis on the electrical response signal, and extracting a second type of characteristic parameters, including frequency characteristic parameters or spectral amplitude characteristic parameters reflecting the spectral distribution characteristics of the electrical response signal; establishing a correspondence between spectral characteristics and electrical equivalent parameters based on the second type of characteristic parameters, calculating equivalent parameters characterizing the electrical characteristics of the field-effect transistor based on the correspondence, and combining the equivalent parameters to form the second electrical equivalent parameter set.

[0008] As a preferred embodiment of the present invention, the electro-thermal coupling prediction model includes an electrical behavior prediction sub-model and a thermal response prediction sub-model. The electrical behavior prediction sub-model characterizes the electrical behavior of the field-effect transistor under test based on a first set of electrical equivalent parameters and a second set of electrical equivalent parameters, and calculates the power change characteristics of the field-effect transistor under a preset power pulse based on the electrical behavior characteristics. The thermal response prediction sub-model predicts the temperature-related response characteristics of the field-effect transistor based on the power change characteristics, and forms an electro-thermal coupling prediction model through the coupling relationship between the electrical behavior prediction sub-model and the thermal response prediction sub-model.

[0009] As a preferred embodiment of the present invention, the generation of the prediction features includes: inputting a first set of electrical equivalent parameters and a second set of electrical equivalent parameters into an electro-thermal coupling prediction model; calculating the power change process of the field-effect transistor under test under a preset power pulse through an electrical behavior prediction sub-model; and using the power change process as the input of a thermal response prediction sub-model; calculating the temperature change process of the field-effect transistor under test in a thermal transient detection mode through the thermal response prediction sub-model; and extracting prediction features from the temperature change process to characterize the temperature change characteristics.

[0010] As a preferred technical solution of the present invention, the calculation of the prediction consistency residual includes: acquiring the temperature-related response signal under the thermal transient detection mode, and extracting the third type of feature parameters based on the temperature-related response signal; matching the third type of feature parameters with the predicted features, and calculating the difference value between the third type of feature parameters and the predicted features, using the difference value as the prediction consistency residual to characterize the degree of consistency between the actual thermal response of the field-effect transistor under test and the prediction result of the electro-thermal coupling prediction model.

[0011] As a preferred technical solution of the present invention, the construction of the multimodal associated defect indicator includes: jointly processing the prediction consistency residual, the first type of feature parameter and the second type of feature parameter to obtain the association parameter reflecting the consistency characteristics of the electrical behavior and thermal response of the field-effect transistor under test; quantifying the association relationship between the prediction consistency residual and the first type of feature parameter and the second type of feature parameter according to the association parameter, and generating a multimodal associated defect indicator for characterizing the defect state of the field-effect transistor.

[0012] As a preferred technical solution of the present invention, the defect discrimination model includes an analysis module for defect discrimination based on multimodal associated defect indicators; the analysis module compares the multimodal associated defect indicators with preset defect discrimination rules according to the numerical characteristics of the multimodal associated defect indicators, and determines the defect detection result of the field-effect transistor under test according to the comparison result, including the defect type and defect level.

[0013] A rapid defect detection system for field-effect transistors, comprising: Modal verification module: acquires the operating status information of the field-effect transistor under test, determines the execution method of the detection mode based on the operating status information, and performs the detection; Online detection module: Under the online detection mode, the electrical response signal of the field-effect transistor is synchronously sampled based on the driving signal of the field-effect transistor, the first type of feature parameters are extracted, and the first electrical equivalent parameter set is estimated from the first type of feature parameters; Perturbation detection module: In idle perturbation detection mode, a small excitation signal is applied to the field-effect transistor and the electrical response signal is collected. The second type of feature parameters are extracted through spectrum analysis, and the second electrical equivalent parameter set is estimated from the second type of feature parameters. Coupled prediction module: Updates the electro-thermal coupling prediction model based on the first and second electrical equivalent parameter sets, and generates prediction features for temperature-related response signals under thermal transient detection mode through the electro-thermal coupling prediction model; Correlation Defect Module: Under thermal transient detection mode, a preset power pulse is applied to the field-effect transistor and temperature-related response signals are collected to extract the third type of feature parameters, calculate the prediction consistency residual, and construct a multimodal correlation defect indicator; Defect Result Module: Input the multimodal associated defect indicator into the defect discrimination model and output the defect detection results of the field-effect transistor.

[0014] The present invention has the following advantages: This invention acquires the operating status information of the field-effect transistor under test and determines the execution mode of the detection mode accordingly, so as to realize the on-demand selection and execution of the online detection mode, the idle perturbation detection mode and the thermal transient detection mode, and adapt to different operating conditions.

[0015] This invention achieves rapid characterization of the electrical states related to the switching behavior of field-effect transistors by synchronously sampling the electrical response signal and extracting the first type of feature parameters under the online detection mode, thereby estimating the first set of electrical equivalent parameters and reducing the dependence on single static electrical parameter testing. By applying a micro-amplitude excitation signal and performing spectral analysis on the electrical response signal under the idle perturbation detection mode to extract the second type of feature parameters, thereby estimating the second set of electrical equivalent parameters, this invention provides supplementary characterization of the spectrum-related electrical states and improves the sensitivity to early or latent defects.

[0016] This invention achieves a predictable mapping of electrical information to thermal response by updating the electro-thermal coupling prediction model based on a first set of electrical equivalent parameters and a second set of electrical equivalent parameters and generating predictive features of temperature-related response signals, thus forming a consistent constraint basis between different modes.

[0017] By applying a preset power pulse and extracting the third type of feature parameters under the thermal transient detection mode, and combining the prediction features to calculate the prediction consistency residual, the consistency between the actual thermal response and the predicted thermal response can be quantitatively verified, thereby reducing false alarms and missed detections caused by relying solely on a single modal feature.

[0018] This invention constructs a multimodal associated defect indicator based on the correlation between the prediction consistency residual and the first and second type of feature parameters, thereby achieving a fusion expression of electrical behavior, spectral characteristics and thermal response consistency information, thus improving the stability and separability of defect diagnosis. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only schematic diagrams of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort. Figure 1 This is a schematic diagram of a rapid defect detection system for field-effect transistors used in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] Example 1: A rapid defect detection method for field-effect transistors, comprising the following steps: Step S1: Obtain the operating status information of the field-effect transistor under test, including the drive signal type, switching frequency, bus voltage, and load status; determine the execution mode of the detection mode based on the operating status information and perform the detection, wherein the detection mode includes the online detection mode, the idle perturbation detection mode, and the thermal transient detection mode; The operating status information is used to characterize the operating conditions of the MOSFET under test during actual operation and is the basic data set used to determine the execution conditions of the detection mode. The operating status information is obtained by collecting and analyzing the operating signals of the power conversion circuit where the MOSFET is located, specifically including operating data such as drive signal type, switching frequency, bus voltage, and load status.

[0022] In a complete testing example, the field-effect transistor (FET) under test operates in a switching power conversion circuit. Its gate is controlled by a periodic drive signal, and the level change of the drive signal controls the FET to switch between the on and off states. By sampling this drive signal, the waveform shape of the drive signal is obtained, and the type of drive signal is determined accordingly. The drive signal type is used to characterize the operating mode of the drive signal, such as a periodic switching drive signal.

[0023] While determining the type of drive signal, the periodic characteristics of the drive signal are analyzed, and the switching frequency is calculated from the time interval between adjacent drive pulses. The switching frequency is used to characterize the number of switching actions completed by the field-effect transistor per unit time, and is an important operating parameter reflecting the dynamic operating state of the device.

[0024] Bus voltage is used to characterize the voltage level on the input side of the power conversion circuit, and is obtained by real-time sampling of the voltage signal of the power bus. The bus voltage data reflects the energy supply status of the power circuit, and its stability directly affects the operating conditions of the MOSFET during the detection process.

[0025] Load status is used to characterize the load operation at the output of the power circuit, and is obtained by detecting changes in the load current. Load status is used to determine whether the power circuit is currently in an active output state or in an idle operating state.

[0026] The determination of the detection mode includes: judging the current operating state of the field-effect transistor under test based on the operating state information, and determining at least one of the following detection modes according to the operating state: online detection mode, idle perturbation detection mode, and thermal transient detection mode; when the drive signal type is periodic switching drive and the switching frequency is within the preset operating frequency range, the online detection mode is executed; when the load state meets the preset idle condition or the bus voltage is within the preset stable range, the idle perturbation detection mode is executed; when it is necessary to perform thermal characteristic detection or consistency verification on the field-effect transistor, the thermal transient detection mode is executed.

[0027] The detection mode is used to represent the detection method employed for defect detection of MOSFETs under different operating conditions. The detection mode is selected based on the current operating state of the device to obtain representative electrical or thermal characteristic data without affecting normal operation.

[0028] The aforementioned online detection mode refers to a detection method performed when the field-effect transistor (FET) is in normal switching operation. In this mode, the electrical response changes generated by the FET during actual switching are used as the source of detection information. Synchronous sampling is used to acquire the electrical behavior characteristics of the device under real operating conditions, thereby enabling the detection of changes in the device's electrical performance.

[0029] The idle perturbation detection mode refers to obtaining detection information reflecting the electrical structural characteristics of a field-effect transistor (FET) by applying a small excitation signal to the FET when the power circuit load is in an idle or stable state, thereby causing a measurable change in the device's electrical response. This detection mode obtains the frequency domain response characteristics of the device without changing the normal operating state of the circuit by introducing an external excitation signal with a small amplitude.

[0030] The thermal transient detection mode refers to the controlled temperature change process of a field-effect transistor (FET) by applying a preset power excitation, and obtaining the thermal response characteristics of the device by detecting the temperature-related response signal. The thermal transient detection mode is used to reflect changes in the internal structure and heat conduction path of the device, thereby providing thermal characteristic data for defect detection.

[0031] In a complete testing example, the MOSFET under test is switched on and off in a power conversion circuit with a periodic drive signal. When the drive signal is detected to be a periodic switch drive and the switching frequency is within the preset operating frequency range, the testing process first enters the online testing mode and then obtains the electrical response data of the device during the actual switching process through subsequent steps.

[0032] When the load current is at a low level and the preset idle condition is met, or when the bus voltage is detected to be in a stable range, the detection process enters the idle perturbation detection mode. In this mode, the spectral response characteristics of the device are obtained by applying a small excitation signal.

[0033] When it is necessary to detect the thermal characteristics of a device or to verify the consistency of electrical behavior prediction results, the detection process enters the thermal transient detection mode. A preset power pulse is applied to generate a temperature change process, and the thermal response characteristics of the device are obtained through the temperature-related response signal.

[0034] Step S2: Under the online detection mode, the electrical response signal of the field-effect transistor is synchronously sampled based on the driving signal of the field-effect transistor, the first type of feature parameters are extracted, and the first electrical equivalent parameter set is estimated from the first type of feature parameters; The electrical response signal represents the voltage and current changes generated by the field-effect transistor (FET) during its turn-on and turn-off processes. It originates from real-time sampling data of the FET's relevant electrical nodes, specifically including voltage and current response signals related to the device's switching process. The electrical response signal reflects the dynamic electrical behavior of the device under actual operating conditions and serves as the raw data for extracting the first type of characteristic parameters.

[0035] Synchronous sampling refers to the process of time-aligning sampling of the electrical response signal using the drive signal as a time reference. The drive signal characterizes the gate control signal of the field-effect transistor, and its level change corresponds to the switching action of the device from the off state to the on state or from the on state to the off state. By using the edge change of the drive signal as a trigger reference to synchronously sample the electrical response signal, it is ensured that response data with a consistent time reference is obtained in different switching cycles, thereby making the subsequently extracted feature parameters comparable.

[0036] In a complete testing example, the field-effect transistor (FET) under test (DET) switches on and off under the control of a periodic drive signal. The rising edge of the drive signal corresponds to the FET entering the conduction process, and the falling edge corresponds to the FET entering the turn-off process. Using the edge changes of the drive signal as a synchronization reference, the voltage and current response signals of the FET are synchronously sampled to obtain a sequence of electrical response signals over multiple switching cycles. By aligning the response signals over multiple switching cycles, a stable switching response dataset is formed for subsequent feature extraction.

[0037] The acquisition of the first electrical equivalent parameter set includes: under the operation of online detection mode, synchronously sampling the electrical response signal of the field-effect transistor under test with the driving signal as the synchronization reference, and extracting the first type of feature parameters from the electrical response signal. The first type of feature parameters includes time feature parameters, voltage change feature parameters, or current change feature parameters related to the switching process of the field-effect transistor; establishing the correspondence between the switching behavior of the field-effect transistor and the electrical equivalent parameters based on the first type of feature parameters, and calculating the equivalent parameters characterizing the electrical characteristics of the field-effect transistor based on the correspondence, and combining the equivalent parameters to form the first electrical equivalent parameter set.

[0038] The first type of feature parameters is used to represent the dynamic feature data related to the switching behavior of the field-effect transistor (FET) extracted from the electrical response signal. These first-type feature parameters characterize the voltage, current, and time variations of the FET during turn-on and turn-off processes, and are a key data set reflecting the dynamic electrical characteristics of the device.

[0039] The first type of characteristic parameters comes from the time-domain characteristic analysis of the electrical response signal, and specifically includes the following three types of data: Time characteristic parameters are used to characterize the time-varying properties of a field-effect transistor (FET) during switching operations. These parameters are calculated from the time relationship between the drive signal and the electrical response signal. For example, within a switching cycle, by analyzing the time interval between the edge change of the drive signal and the change in the electrical response signal, characteristic parameters reflecting the duration of the device's turn-on or turn-off process can be obtained.

[0040] Voltage variation characteristic parameters are used to characterize the dynamic characteristics of voltage changes in a field-effect transistor (FET) during switching. These parameters are extracted from the voltage waveform variation patterns in the electrical response signal. By analyzing the changes in the voltage response signal over time, characteristic data reflecting the voltage change behavior during the device's switching process are obtained.

[0041] Current variation characteristic parameters are used to characterize the dynamic characteristics of current changes during the conduction and turn-off processes of a field-effect transistor (FET). These parameters are derived from the change process of the current response signal. By analyzing the current variation over time, characteristic data reflecting the device's conduction process and current build-up process are obtained.

[0042] After obtaining the first type of characteristic parameters, parameter mapping calculations are performed on these parameters based on the correspondence between the switching behavior of the field-effect transistor (FET) and its electrical parameters. The electrical equivalent parameters represent the set of parameters describing the FET's electrical behavior through an equivalent electrical model. These parameters characterize the FET's conduction, switching, and dynamic response characteristics during actual operation. By calculating the correspondence between the first type of characteristic parameters and the electrical equivalent parameters, a set of equivalent parameters characterizing the FET's electrical behavior is obtained. These equivalent parameters are then combined according to a predetermined format to form the first set of electrical equivalent parameters.

[0043] Step S3: In the idle perturbation detection mode, apply a small excitation signal to the field-effect transistor and collect the electrical response signal. Extract the second type of characteristic parameters through spectrum analysis, and estimate the second electrical equivalent parameter set from the second type of characteristic parameters. In the idle perturbation detection mode, the power circuit in which the MOSFET under test is located meets the preset idle condition or the bus voltage is in a stable range. At this time, the device does not undertake the main power switching task, and the circuit is in a stable or low-load operation state. Under this operating condition, by applying a small external excitation signal to the MOSFET, the device produces a measurable change in electrical response, thereby obtaining response information reflecting the electrical structure characteristics of the device.

[0044] The micro-amplitude excitation signal represents a low-amplitude test signal applied to the field-effect transistor (FET) in the idle perturbation detection mode. The amplitude of the micro-amplitude excitation signal is kept relatively low compared to the amplitude of the FET's normal operating drive signal, introducing a small perturbation without altering the device's original operating state, thereby obtaining the device's electrical response characteristics under micro-excitation conditions. The signal form of the micro-amplitude excitation signal is determined by a preset detection strategy and is used to excite the FET to produce a response change suitable for spectral analysis.

[0045] In a complete testing example, when the power circuit is determined to be in an idle operating state, the testing process enters the idle perturbation detection mode. At this time, a small excitation signal is introduced at the control terminal or relevant electrical node of the field-effect transistor (FET), causing the device to produce a small change in electrical response while maintaining its original operating state. This small excitation signal maintains a stable variation over time, thus forming a recognizable frequency component in the electrical response signal. After applying the small excitation signal, the electrical response signal of the FET is acquired. This electrical response signal is the voltage or current change signal generated by the device under perturbation conditions, originating from the electrical response data obtained by sampling the relevant electrical nodes of the FET. The electrical response signal contains dynamic change information of the device under small excitation, providing basic data for subsequent spectrum analysis. Frequency domain analysis is performed on the acquired electrical response signal to obtain its frequency distribution characteristics. The frequency domain analysis process involves performing a spectral transformation on the time series of the electrical response signal to obtain the energy distribution of the signal at different frequency components. By analyzing the variation pattern of the electrical response signal in the frequency domain, characteristic data reflecting the electrical structural characteristics of the device are extracted.

[0046] The acquisition of the second electrical equivalent parameter set includes: applying a small-amplitude excitation signal to the field-effect transistor under test in the idle perturbation detection mode, acquiring the electrical response signal of the field-effect transistor, performing spectral analysis on the electrical response signal, and extracting second-type feature parameters, including frequency feature parameters or spectral amplitude feature parameters that reflect the spectral distribution characteristics of the electrical response signal; establishing a correspondence between spectral features and electrical equivalent parameters based on the second-type feature parameters, calculating equivalent parameters characterizing the electrical characteristics of the field-effect transistor based on the correspondence, and combining the equivalent parameters to form the second electrical equivalent parameter set.

[0047] The second type of characteristic parameters represents the frequency domain characteristic data obtained through spectral analysis of the electrical response signal. This second type of characteristic parameter describes the frequency response characteristics of the field-effect transistor under micro-amplitude excitation conditions and is an important data set reflecting the internal electrical structure of the device.

[0048] The second type of feature parameters mainly includes the following two types of data: Frequency characteristic parameters are used to characterize the distribution characteristics of frequency components in the spectrum of an electrical response signal. These parameters are derived from frequency position data in the spectral analysis results. By identifying the main frequency components appearing in the spectrum, frequency characteristic data reflecting changes in the electrical structure of the device are obtained.

[0049] Spectral amplitude characteristic parameters are used to characterize the amplitude distribution of an electrical response signal across different frequency components. These parameters are extracted from amplitude information obtained from spectral analysis results. By analyzing the amplitude values ​​corresponding to different frequency components, characteristic data reflecting changes in the frequency response intensity of the device are obtained.

[0050] After obtaining the second type of characteristic parameters, parameter mapping calculations are performed on these parameters based on the correspondence between spectral characteristics and electrical equivalent parameters. The correspondence between spectral characteristics and electrical equivalent parameters represents the relationship between frequency domain response characteristics and the device's equivalent electrical behavior. By analyzing the spectral distribution characteristics, equivalent electrical parameters reflecting the device's electrical structure state are obtained. A set of equivalent parameters characterizing the electrical properties of the field-effect transistor is calculated using the aforementioned correspondence, and these equivalent parameters are combined in a predetermined manner to form the second set of electrical equivalent parameters.

[0051] Step S4: Update the electro-thermal coupling prediction model based on the first and second electrical equivalent parameter sets, and generate prediction features for the temperature-related response signal under the thermal transient detection mode through the electro-thermal coupling prediction model; The first set of electrical equivalent parameters characterizes the dynamic electrical behavior of the field-effect transistor (FET) under actual switching conditions, while the second set characterizes its frequency-domain electrical structure under perturbation conditions. These two sets of electrical equivalent parameters reflect the device's electrical state information from different perspectives. By jointly inputting these two sets of parameters into an electro-thermal coupling prediction model, a prediction mechanism is constructed to describe the relationship between the device's electrical behavior and thermal response.

[0052] The electro-thermal coupling prediction model is a computational model used to describe the correlation between changes in the electrical behavior of a field-effect transistor (FET) and its temperature response. This model is used to infer the temperature change process under controlled power excitation based on the device's electrical behavior characteristics, thus generating a prediction of the temperature response characteristics before thermal transient detection is performed. The electro-thermal coupling prediction model consists of an electrical behavior prediction sub-model and a thermal response prediction sub-model, and a complete electro-thermal coupling relationship is formed through the data association between the two. Here, "coupling" refers to the influence of power changes generated by the device's electrical behavior on temperature changes, and the establishment of a transfer relationship between electrical behavior and thermal response through this power-temperature relationship.

[0053] The electro-thermal coupling prediction model includes an electrical behavior prediction sub-model and a thermal response prediction sub-model. The electrical behavior prediction sub-model characterizes the electrical behavior of the field-effect transistor under test based on a first set of electrical equivalent parameters and a second set of electrical equivalent parameters, and calculates the power change characteristics of the field-effect transistor under a preset power pulse based on the electrical behavior characteristics. The thermal response prediction sub-model predicts the temperature-related response characteristics of the field-effect transistor based on the power change characteristics, and forms the electro-thermal coupling prediction model through the coupling relationship between the electrical behavior prediction sub-model and the thermal response prediction sub-model.

[0054] The electrical behavior prediction sub-model describes the power change behavior of the field-effect transistor (FET) under specific excitation conditions based on its electrical equivalent parameters. The sub-model takes two sets of electrical equivalent parameters as input: a first set and a second set. Joint analysis of these two sets of parameters forms a comprehensive characterization of the device's electrical behavior. The first set of electrical equivalent parameters is derived from the switching behavior analysis under the online detection mode, primarily reflecting the dynamic on / off characteristics of the device in actual switching operation. The second set of electrical equivalent parameters is derived from the frequency domain response analysis under the idle perturbation detection mode, primarily reflecting the electrical structural characteristics of the device under small-signal perturbation conditions. Combining these two sets of parameters provides a unified description of the overall electrical behavior of the device.

[0055] In a complete detection example, the first and second sets of electrical equivalent parameters are used as input data to predict the electrical behavior of the field-effect transistor (FET) under the action of a preset power pulse. The preset power pulse represents the power excitation signal applied to the FET in the subsequent thermal transient detection mode, and its amplitude and duration are set before the detection begins.

[0056] The electrical behavior prediction sub-model calculates and analyzes the electrical equivalent parameters to obtain the power change process of the field-effect transistor (FET) under a preset power pulse. This power change process represents the time-varying power generated by the FET during the power pulse. This data reflects the power change trend of the device under controlled excitation conditions and serves as crucial intermediate data connecting electrical behavior and thermal response. After obtaining the power change process, it is used as input data for the thermal response prediction sub-model. The thermal response prediction sub-model predicts the temperature change behavior of the FET based on the device's power change. It converts the power change process into a temperature change process by describing the internal heat conduction path and heat diffusion process of the device.

[0057] In this detection example, the thermal response prediction sub-model calculates the temperature change process of the field-effect transistor (FET) under thermal transient excitation conditions based on the input power change process. The temperature change process represents the device's response curve as temperature changes over time under power excitation. This process reflects the combined effects of the device's internal heat conduction path and thermal capacity characteristics, serving as a crucial data source for extracting predicted temperature features. After obtaining the temperature change process, feature extraction is performed to acquire predictive features describing the temperature change behavior.

[0058] The generation of the predicted features includes: inputting the first set of electrical equivalent parameters and the second set of electrical equivalent parameters into the electro-thermal coupling prediction model; calculating the power change process of the field-effect transistor under test under the action of a preset power pulse through the electrical behavior prediction sub-model; and using the power change process as the input of the thermal response prediction sub-model; calculating the temperature change process of the field-effect transistor under test under the thermal transient detection mode by the thermal response prediction sub-model; and extracting predicted features from the temperature change process to characterize the temperature change characteristics.

[0059] The predicted features represent the temperature response characteristics calculated by the electro-thermal coupling prediction model. These features are derived from feature extraction results of the temperature change process and characterize the temperature change behavior of the device under preset power pulse conditions. In the complete detection example described above, a set of predicted features reflecting the device's temperature change characteristics is obtained through the collaborative calculation of the electrical behavior prediction sub-model and the thermal response prediction sub-model. These predicted features serve as the thermal response prediction result of the device under ideal electrical behavior conditions, providing a reference benchmark for consistent comparison of actual temperature response data in subsequent steps.

[0060] Step S5: Under the thermal transient detection mode, a preset power pulse is applied to the field-effect transistor and the temperature-related response signal is collected to extract the third type of feature parameters. The prediction consistency residual between the third type of feature parameters and the predicted features is calculated, and a multimodal associated defect indicator is constructed based on the correlation between the prediction consistency residual and the first type of feature parameters and the second type of feature parameters. To verify the consistency between the predicted feature and the actual thermal response of the device, this step enters the thermal transient detection mode. The actual temperature response of the device is acquired by applying controlled power excitation, and the third type of feature parameters for consistency analysis are extracted accordingly. The thermal transient detection mode represents the detection method of inducing a measurable temperature change in the field-effect transistor (FET) through controlled power excitation. In this detection mode, a preset power pulse is applied, causing the FET to dissipate power of a certain magnitude within a short period, thereby inducing a temperature change inside the device. This temperature change reflects the internal structural state and heat conduction path characteristics of the device, and is an important basis for judging the device's health status.

[0061] The preset power pulse represents the controlled power excitation signal applied to the field-effect transistor in the thermal transient detection mode. The preset power pulse is set before the detection begins, and its excitation mode remains consistent throughout the detection process to ensure uniform excitation conditions across different detection cycles. The preset power pulse is used to excite the device to generate a temperature change under controllable conditions, thereby acquiring response data for thermal characteristic analysis.

[0062] In a complete testing instance, when it is determined that thermal characteristic testing or conformance verification needs to be performed, the testing process enters the thermal transient testing mode. At this time, a preset power pulse is applied to the field-effect transistor (FET), causing a power dissipation change in the device within a short period, resulting in a temperature change over time. During the power pulse, the temperature-dependent response signal of the FET is acquired. The temperature-dependent response signal represents the response data reflecting the device's temperature change state, and it originates from the detection results of the device's temperature change behavior. The temperature-dependent response signal changes over time to form a temperature response sequence, which reflects the actual temperature change process of the FET under power excitation. By analyzing the temperature-dependent response signal, the temperature change data of the device during the thermal transient process is obtained, and characteristic parameters characterizing the temperature change behavior are extracted based on this.

[0063] The calculation of the prediction consistency residual includes: acquiring the temperature-related response signal under the thermal transient detection mode, and extracting the third type of feature parameters based on the temperature-related response signal; matching the third type of feature parameters with the predicted features, and calculating the difference value between the third type of feature parameters and the predicted features, using the difference value as the prediction consistency residual to characterize the degree of consistency between the actual thermal response of the field-effect transistor under test and the prediction result of the electro-thermal coupling prediction model.

[0064] The third type of feature parameters represents the temperature change feature data extracted from the temperature-related response signal. These parameters originate from the feature extraction results of the temperature change process and characterize the actual thermal response behavior of the field-effect transistor under power excitation conditions. In the complete detection example described above, a set of feature data reflecting temperature change behavior is obtained by analyzing the temperature response sequence, thus forming the third type of feature parameter set. After obtaining the third type of feature parameters, they are matched with the predicted features obtained in S4, and the difference between the two is calculated to obtain the prediction consistency residual. The prediction consistency residual represents the degree of deviation between the actual thermal response of the device and the prediction result of the electro-thermal coupling prediction model.

[0065] When the device is in normal condition, the actual temperature change characteristics and the predicted characteristics maintain a high degree of consistency, and the prediction consistency residual remains within a small range. When defects occur in the internal structure of the device or its performance degrades, the actual temperature change behavior of the device will deviate from the model prediction results, thus causing a significant change in the prediction consistency residual. After obtaining the prediction consistency residual, it is jointly processed with the first type of feature parameters obtained in S2 and the second type of feature parameters obtained in S3 to construct defect evaluation parameters that comprehensively reflect the device state.

[0066] The construction of the multimodal correlated defect indicator includes: jointly processing the prediction consistency residual, the first type of feature parameters, and the second type of feature parameters to obtain correlation parameters reflecting the consistency characteristics of the electrical behavior and thermal response of the field-effect transistor under test; quantifying the correlation relationship between the prediction consistency residual and the first and second type of feature parameters based on the correlation parameters, and generating a multimodal correlated defect indicator to characterize the defect state of the field-effect transistor. The multimodal correlated defect indicator represents the defect evaluation data obtained by fusing multi-source detection data. By fusing electrical feature data and thermal response consistency data, the multimodal correlated defect indicator provides a unified characterization of the comprehensive deviation between changes in the device's electrical behavior and changes in its thermal response.

[0067] In this embodiment, "multimodal" refers to data sources from different detection modalities, including first-type feature parameters obtained under the online detection modality, second-type feature parameters obtained under the idle perturbation detection modality, and prediction consistency residuals obtained under the thermal transient detection modality. The correlation parameter represents intermediate evaluation data formed after jointly processing data from multiple modalities. The correlation parameter, by applying a unified scale to data from different modalities, makes feature data from different sources comparable and provides a comprehensive representation of multi-source feature data based on a unified scale.

[0068] After standardizing the prediction consistency residuals, first-type feature parameters, and second-type feature parameters, corresponding weights are assigned based on the sensitivity of each type of feature data to device defect changes. Weighted fusion of the normalized feature data forms a correlation parameter describing the overall degree of change in multi-source features. After obtaining the correlation parameter, a unified evaluation process is performed to convert the multi-source feature change information into a single evaluation quantity, resulting in a multimodal correlated defect indicator used to characterize the degree of device state change. This multimodal correlated defect indicator comprehensively reflects the overall deviation between changes in the electrical behavior, frequency response, and thermal response consistency of the field-effect transistor. In the complete detection example described above, when the device is in a normal state, the multimodal correlated defect indicator remains within a stable range; when defects or performance degradation occur within the device, the multimodal correlated defect indicator will show significant changes.

[0069] Step S6: Input the multimodal associated defect indicator into the defect discrimination model and output the defect detection result of the field-effect transistor.

[0070] The defect discrimination model is used to determine the health status of a field-effect transistor (FET) based on multimodal correlated defect indicators. The model performs regularization analysis on the defect indicators, comparing them with pre-established defect discrimination rules to determine whether a defect exists in the device and the severity of that defect.

[0071] The defect discrimination model includes an analysis module for defect discrimination based on multimodal correlated defect indicators. The analysis module compares the multimodal correlated defect indicators with preset defect discrimination rules based on their numerical characteristics, and determines the defect detection result of the field-effect transistor under test (FET) based on the comparison result, including the defect type and defect level. Specifically, the analysis module performs the state discrimination process for the multimodal correlated defect indicators. The analysis module parses the input multimodal correlated defect indicators and classifies the device state according to the preset defect discrimination rules.

[0072] The preset defect discrimination rule is used to represent the state judgment standard established based on the changing patterns of device characteristics. The preset defect discrimination rule is obtained through statistical analysis of a large amount of device testing data and is used to describe the range of variation of multimodal associated defect indicators of the device under normal conditions and different defect states.

[0073] In a complete testing example, multiple tests are performed on a field-effect transistor (FET) in a normal state to obtain the range of multimodal associated defect indicators corresponding to the normal device. A normal state discrimination interval is established based on this range. When the detected multimodal associated defect indicator falls within this interval, the device is determined to be in normal working condition. When the multimodal associated defect indicator exceeds the normal state discrimination interval, it indicates that the device's electrical behavior or thermal response characteristics have changed. At this point, the device state is classified into different defect levels based on the degree of change in the defect indicator. The defect type indicates the category of abnormal characteristic changes in the FET. The defect type is determined by analyzing the changing characteristics of the multimodal associated defect indicator. For example, when the change in the defect indicator is mainly caused by changes in electrical behavior characteristics, it can be identified as an electrical characteristic-related defect; when the change in the defect indicator is mainly caused by changes in thermal response consistency, it can be identified as a thermal characteristic-related defect. The defect level indicates the severity of device performance degradation or structural abnormalities. The defect level is classified based on the degree of deviation of the multimodal associated defect indicator from the normal range. As the deviation of the defect indication increases, the corresponding defect level gradually increases, thus reflecting the process of the device state changing from a slight abnormality to a severe abnormality.

[0074] In the complete testing example described above, after the testing process is completed, the multimodal associated defect indicator is input into the defect discrimination model, and the analysis module performs state discrimination according to preset defect discrimination rules. Through this discrimination process, the defect detection results of the field-effect transistor under test are obtained, including the defect type and the corresponding defect level.

[0075] Example 2: A rapid defect detection system for field-effect transistors (FETs), see [link to example]. Figure 1 As shown, it includes the following modules: Modal verification module: acquires the operating status information of the field-effect transistor under test, determines the execution method of the detection mode based on the operating status information, and performs the detection; Online detection module: Under the online detection mode, the electrical response signal of the field-effect transistor is synchronously sampled based on the driving signal of the field-effect transistor, the first type of feature parameters are extracted, and the first electrical equivalent parameter set is estimated from the first type of feature parameters; Perturbation detection module: In idle perturbation detection mode, a small excitation signal is applied to the field-effect transistor and the electrical response signal is collected. The second type of feature parameters are extracted through spectrum analysis, and the second electrical equivalent parameter set is estimated from the second type of feature parameters. Coupled prediction module: Updates the electro-thermal coupling prediction model based on the first and second electrical equivalent parameter sets, and generates prediction features for temperature-related response signals under thermal transient detection mode through the electro-thermal coupling prediction model; Correlation Defect Module: Under thermal transient detection mode, a preset power pulse is applied to the field-effect transistor and temperature-related response signals are collected to extract the third type of feature parameters, calculate the prediction consistency residual, and construct a multimodal correlation defect indicator; Defect Result Module: Input the multimodal associated defect indicator into the defect discrimination model and output the defect detection results of the field-effect transistor.

[0076] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A rapid defect detection method for field-effect transistors, characterized in that, include: Obtain the operating status information of the field-effect transistor under test, including the drive signal type, switching frequency, bus voltage, and load status; The execution mode of the detection mode is determined and the detection is performed based on the working status information. The detection modes include online detection mode, idle perturbation detection mode, and thermal transient detection mode. In the online detection mode, the electrical response signal of the field-effect transistor is synchronously sampled based on the driving signal of the field-effect transistor, the first type of feature parameters are extracted, and the first electrical equivalent parameter set is estimated from the first type of feature parameters; In the idle perturbation detection mode, a small excitation signal is applied to the field-effect transistor and the electrical response signal is acquired. The second type of characteristic parameters are extracted by spectrum analysis, and the second electrical equivalent parameter set is estimated from the second type of characteristic parameters. The electro-thermal coupling prediction model is updated based on the first and second sets of electrical equivalent parameters. The electro-thermal coupling prediction model is used to generate predictive features of temperature-related response signals under thermal transient detection mode. In the thermal transient detection mode, a preset power pulse is applied to the field-effect transistor and the temperature-related response signal is collected to extract the third type of feature parameters. The prediction consistency residual between the third type of feature parameters and the predicted features is calculated, and a multimodal associated defect indicator is constructed based on the correlation between the prediction consistency residual and the first type of feature parameters and the second type of feature parameters. Input the multimodal associated defect indicator into the defect discrimination model, and output the defect detection results of the field-effect transistor.

2. The method for rapid defect detection in a field-effect transistor according to claim 1, characterized in that, The determination of the detection mode includes: judging the current operating state of the field-effect transistor under test based on the operating state information, and determining at least one of the following detection modes according to the operating state: online detection mode, idle perturbation detection mode, and thermal transient detection mode; when the drive signal type is periodic switching drive and the switching frequency is within the preset operating frequency range, the online detection mode is executed; when the load state meets the preset idle condition or the bus voltage is within the preset stable range, the idle perturbation detection mode is executed; when it is necessary to perform thermal characteristic detection or consistency verification on the field-effect transistor, the thermal transient detection mode is executed.

3. The method for rapid defect detection in a field-effect transistor according to claim 1, characterized in that, The acquisition of the first electrical equivalent parameter set includes: under the operation of online detection mode, synchronously sampling the electrical response signal of the field-effect transistor under test with the driving signal as the synchronization reference, and extracting the first type of feature parameters from the electrical response signal. The first type of feature parameters includes time feature parameters, voltage change feature parameters, or current change feature parameters related to the switching process of the field-effect transistor; establishing the correspondence between the switching behavior of the field-effect transistor and the electrical equivalent parameters based on the first type of feature parameters, and calculating the equivalent parameters characterizing the electrical characteristics of the field-effect transistor based on the correspondence, and combining the equivalent parameters to form the first electrical equivalent parameter set.

4. The method for rapid defect detection in a field-effect transistor according to claim 1, characterized in that, The acquisition of the second electrical equivalent parameter set includes: applying a small-amplitude excitation signal to the field-effect transistor under test in the idle perturbation detection mode, acquiring the electrical response signal of the field-effect transistor, performing spectral analysis on the electrical response signal, and extracting second-type feature parameters, including frequency feature parameters or spectral amplitude feature parameters that reflect the spectral distribution characteristics of the electrical response signal; establishing a correspondence between spectral features and electrical equivalent parameters based on the second-type feature parameters, calculating equivalent parameters characterizing the electrical characteristics of the field-effect transistor based on the correspondence, and combining the equivalent parameters to form the second electrical equivalent parameter set.

5. The method for rapid defect detection in a field-effect transistor according to claim 1, characterized in that, The electro-thermal coupling prediction model includes an electrical behavior prediction sub-model and a thermal response prediction sub-model. The electrical behavior prediction sub-model characterizes the electrical behavior characteristics of the field-effect transistor under test based on the first electrical equivalent parameter set and the second electrical equivalent parameter set, and calculates the power change characteristics of the field-effect transistor under the action of a preset power pulse based on the electrical behavior characteristics. The thermal response prediction sub-model predicts the temperature-related response characteristics of the field-effect transistor based on the power change characteristics, and forms an electro-thermal coupled prediction model through the coupling relationship between the electrical behavior prediction sub-model and the thermal response prediction sub-model.

6. The method for rapid defect detection in a field-effect transistor according to claim 5, characterized in that, The generation of the predicted features includes: inputting the first set of electrical equivalent parameters and the second set of electrical equivalent parameters into the electro-thermal coupling prediction model; calculating the power change process of the field-effect transistor under test under the action of a preset power pulse through the electrical behavior prediction sub-model; and using the power change process as the input of the thermal response prediction sub-model; calculating the temperature change process of the field-effect transistor under test under the thermal transient detection mode by the thermal response prediction sub-model; and extracting predicted features from the temperature change process to characterize the temperature change characteristics.

7. The method for rapid defect detection in a field-effect transistor according to claim 1, characterized in that, The calculation of the prediction consistency residual includes: acquiring the temperature-related response signal under the thermal transient detection mode, and extracting the third type of feature parameters based on the temperature-related response signal; matching the third type of feature parameters with the predicted features, and calculating the difference value between the third type of feature parameters and the predicted features, using the difference value as the prediction consistency residual to characterize the degree of consistency between the actual thermal response of the field-effect transistor under test and the prediction result of the electro-thermal coupling prediction model.

8. The method for rapid defect detection in a field-effect transistor according to claim 1, characterized in that, The construction of the multimodal associated defect indicator includes: jointly processing the prediction consistency residual, the first type of feature parameter, and the second type of feature parameter to obtain the association parameter reflecting the consistency characteristics of the electrical behavior and thermal response of the field-effect transistor under test; quantifying the association relationship between the prediction consistency residual and the first type of feature parameter and the second type of feature parameter based on the association parameter, and generating a multimodal associated defect indicator for characterizing the defect state of the field-effect transistor.

9. The method for rapid defect detection in a field-effect transistor according to claim 1, characterized in that, The defect discrimination model includes an analysis module for defect discrimination based on multimodal correlated defect indicators. The analysis module compares the multimodal correlated defect indicators with preset defect discrimination rules according to the numerical characteristics of the multimodal correlated defect indicators, and determines the defect detection results of the field-effect transistor under test based on the comparison results, including the defect type and defect level.

10. A rapid defect detection system for field-effect transistors, characterized in that, The system employs a rapid defect detection method for field-effect transistors as described in any one of claims 1 to 9, comprising: Modal verification module: acquires the operating status information of the field-effect transistor under test, determines the execution method of the detection mode based on the operating status information, and performs the detection; Online detection module: Under the online detection mode, the electrical response signal of the field-effect transistor is synchronously sampled based on the driving signal of the field-effect transistor, the first type of feature parameters are extracted, and the first electrical equivalent parameter set is estimated from the first type of feature parameters; Perturbation detection module: In idle perturbation detection mode, a small excitation signal is applied to the field-effect transistor and the electrical response signal is collected. The second type of feature parameters are extracted through spectrum analysis, and the second electrical equivalent parameter set is estimated from the second type of feature parameters. Coupled prediction module: Updates the electro-thermal coupling prediction model based on the first and second electrical equivalent parameter sets, and generates prediction features for temperature-related response signals under thermal transient detection mode through the electro-thermal coupling prediction model; Correlation Defect Module: Under thermal transient detection mode, a preset power pulse is applied to the field-effect transistor and temperature-related response signals are collected to extract the third type of feature parameters, calculate the prediction consistency residual, and construct a multimodal correlation defect indicator; Defect Result Module: Input the multimodal associated defect indicator into the defect discrimination model and output the defect detection results of the field-effect transistor.

Images