A gate loop inductance detection method for failure prediction

By performing high-bandwidth acquisition and high-sampling-rate processing on the voltage signal of the gate drive node, the slope dispersion and overshoot/undershoot events of the voltage waveform are calculated, accurately distinguishing the causes of gate circuit faults in the switching power supply. This solves the problem of misjudgment in the prior art and realizes reliable fault prediction and diagnosis.

CN121805811BActive Publication Date: 2026-05-08BEIJING YUEXIN TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202610275438.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-03-09
Publication Date
2026-05-08
Estimated Expiration
2046-03-09

AI Technical Summary

Technical Problem

Existing technologies cannot accurately distinguish the cause of the sluggish gate drive waveform in switching power supplies. This may be due to the parasitic inductance effect of the gate circuit or the insufficient load-carrying capacity of the driver integrated circuit, which leads to the extended dwell time of the power switch in the linear region.

Method used

By acquiring the voltage signal of the gate drive node, converting it into a voltage value and plotting the voltage waveform, calculating the average slope and dispersion of the rising and falling edges, and combining overshoot and undershoot events, the cause of the fault is determined. Non-intrusive data acquisition and processing are performed using a high-bandwidth voltage probe and a high sampling rate analog-to-digital converter.

Benefits of technology

It enables accurate differentiation between two types of faults: parasitic inductance effect in the gate circuit and insufficient load capacity of the driver integrated circuit, avoiding misjudgment, providing reliable early fault prediction, and improving system stability and safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121805811B_ABST
    Figure CN121805811B_ABST
Patent Text Reader

Abstract

The application relates to the technical field of fault prediction, and specifically discloses a gate loop inductance detection method for fault prediction, which comprises the following steps: collecting voltage signals of a gate drive node in continuous multiple switching cycles, converting the voltage signals into voltage values, and drawing a voltage waveform according to the voltage values; counting the number of times that a preset stable high level is greater than in the rising edge Si, and counting the number of times CS in all switching cycles; counting the number of times that a preset stable low level is less than in the falling edge Xi, and counting the number of times CX in all switching cycles; calculating the rising time and the falling time based on the voltage waveform, and averaging the rising slope and the falling slope; calculating the rising index and the falling index; and judging the fault cause based on the rising index, the falling index, the number of times CS and the number of times CX. The application realizes early fault identification and root cause differentiation by counting the gate waveform features of multiple switching cycles, and improves the diagnostic accuracy and the system reliability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of fault prediction technology, and more specifically to a gate circuit inductance detection method for fault prediction. Background Technology

[0002] A gate circuit is a dedicated circuit that controls the operating state of the gate in a transistor or vacuum tube, thereby regulating the current in the main circuit.

[0003] In fault analysis of switching power supplies, a delayed rising or falling edge of the gate drive waveform is a typical phenomenon, and the conventional diagnostic path usually points to the parasitic inductance effect of the gate circuit. However, in practice, there is a more hidden failure mechanism: the actual load-carrying capacity of the driver integrated circuit is lower than its nominal parameters.

[0004] The essence of the problem lies in the fact that the observed "slow charging of the gate capacitor" is a comprehensive result. The peak current value specified in the driver chip datasheet is usually based on ideal conditions, and its actual output capability under complex operating conditions will be reduced. When the driver chip cannot output sufficient current, the charging and discharging process of the gate capacitor will be delayed, directly leading to an increase in the dwell time of the power switch in the linear region.

[0005] Therefore, accurately distinguishing between these two causes has become an urgent problem to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a gate circuit inductance detection method for fault prediction, thereby solving the above-mentioned technical problems.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A method for detecting gate circuit inductance for fault prediction includes the following steps:

[0009] The voltage signal of the gate drive node is acquired in multiple consecutive switching cycles, the voltage signal is converted into a voltage value, and the voltage waveform is plotted based on the voltage value.

[0010] Obtain the rising edge Si and falling edge Xi in the voltage waveform of switching cycle i, count the number of times the rising edge Si is greater than the preset stable high level, and count the number CS in the entire switching cycle; count the number of times the falling edge Xi is less than the preset stable low level, and count the number CX in the entire switching cycle.

[0011] Calculate the rise time and fall time of switching period i based on the voltage waveform of switching period i, and calculate the average rise slope and average fall slope of switching period i.

[0012] The average rising slope and average falling slope of all switching cycles are statistically distributed to obtain the dispersion index of the average rising slope and average falling slope, which are denoted as rising index and falling index, respectively.

[0013] The cause of the fault is determined based on the rising index, falling index, number of times CS and number of times CX. The causes of the fault include the parasitic inductance effect of the gate circuit and the insufficient load capacity of the driver integrated circuit.

[0014] As a further aspect of the present invention: obtaining the voltage value includes:

[0015] The voltage signal of the gate drive node in multiple consecutive switching cycles is acquired by a high-bandwidth voltage probe, which is connected to the gate drive node in a non-invasive manner to capture the analog voltage signal.

[0016] The analog voltage signal is converted into a voltage value by an analog-to-digital converter. The sampling rate of the analog-to-digital converter is set to be a multiple of the switching frequency to ensure that the details of the voltage signal change are captured. The conversion process includes sampling the analog voltage signal at equal intervals. Each sampling point generates a voltage value and attaches a corresponding timestamp, which is based on the system clock synchronization.

[0017] As a further aspect of the present invention: plotting the voltage waveform includes:

[0018] Using timestamps as the horizontal axis and voltage values ​​as the vertical axis, data points are generated in the coordinate system. Adjacent data points are connected by a smooth curve to form a continuous curve, thus generating a voltage waveform. The drawing of the voltage waveform also involves smoothing the voltage values ​​to reduce the impact of noise. The smoothing process uses a moving average method, and the window size of the moving average is adjusted according to the sampling rate.

[0019] As a further aspect of the present invention: the calculation of the rising and falling indicators includes:

[0020] Calculate the initial parameters for all average rising and falling slopes, including variance, standard deviation, and skewness.

[0021] The ascent index is calculated based on the initial parameters of the superior and inferior solution distance method and the average ascent slope.

[0022] The descent index is calculated based on the initial parameters of the superior-inferior solution distance method and the average descent slope.

[0023] As a further aspect of the present invention: before determining the cause of the fault, the method further includes:

[0024] If the rising indicator and / or falling indicator exceed the preset reference range, or the number of times CS and / or the number of times CX are greater than the preset number of times threshold, then a fault is determined to exist;

[0025] If no fault is found, the cause of the fault will not be determined.

[0026] The reference ranges for the upward and downward indicators are different.

[0027] As a further aspect of the present invention: the process of determining the cause of the fault includes:

[0028] When the rising indicator is greater than its preset threshold and the falling indicator is greater than its preset threshold:

[0029] If CS≠0 and / or CX≠0, the fault is determined to be caused by the parasitic inductance effect of the gate circuit; if CS=0 and CX=0, the fault is determined to be caused by insufficient load capacity of the driver integrated circuit.

[0030] When the rising index is greater than its preset threshold and the falling index is less than or equal to its preset threshold, or when the rising index is less than or equal to its preset threshold and the falling index is greater than its preset threshold, the fault is determined to be due to insufficient load capacity of the driver integrated circuit.

[0031] The beneficial effects of this invention compared to the prior art are as follows:

[0032] This invention achieves accurate differentiation between two types of faults: parasitic inductance in the gate circuit and insufficient load-carrying capacity of the driver integrated circuit, by statistically analyzing and extracting transient features from the gate drive waveforms of multiple consecutive switching cycles. By calculating the average slope of the rise and fall processes in each switching cycle and evaluating its dispersion, the stability characteristics of the waveform change rate can be revealed. Combined with the detection of overshoot and undershoot events, inductive oscillation behavior can be identified. This invention uses both "slope dispersion" and "oscillation presence" as dual judgment criteria, enabling the identification of different physical causes even when waveform sluggishness phenomena are similar, thereby effectively avoiding misjudgment.

[0033] The detection process of this invention requires no external excitation or complex modeling; it relies solely on the acquisition and processing of actual operating waveforms to achieve online, non-invasive diagnostics. This method can identify potential inductive interference or driving capability degradation trends at an early stage, providing a reliable basis for the health assessment and predictive maintenance of power devices, and improving the stability and safety of the system. Attached Figure Description

[0034] The invention will now be further described with reference to the accompanying drawings.

[0035] Figure 1 This is a schematic flowchart of a gate circuit inductance detection method for fault prediction according to the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Please see Figure 1 As shown, the present invention is a gate circuit inductance detection method for fault prediction, comprising the following steps:

[0038] The voltage signal of the gate drive node is acquired in multiple consecutive switching cycles, the voltage signal is converted into a voltage value, and the voltage waveform is plotted based on the voltage value.

[0039] In a preferred embodiment of the present invention, obtaining the voltage value includes:

[0040] The voltage signal of the gate drive node in multiple consecutive switching cycles is acquired by a high-bandwidth voltage probe, which is connected to the gate drive node in a non-invasive manner to capture the analog voltage signal.

[0041] The analog voltage signal is converted into a voltage value by an analog-to-digital converter. The sampling rate of the analog-to-digital converter is set to be a multiple of the switching frequency to ensure that the details of the voltage signal change are captured. The conversion process includes sampling the analog voltage signal at equal intervals. Each sampling point generates a voltage value and attaches a corresponding timestamp, which is based on the system clock synchronization.

[0042] It should be noted that the non-invasive acquisition of the voltage signal of the gate drive node using a high-bandwidth voltage probe, followed by conversion into discrete voltage values ​​via a high-sampling-rate analog-to-digital converter, is designed to fully preserve information about the high-speed transient changes during gate driving. Since the gate signal completes charging and discharging on a nanosecond timescale, any measurement method with insufficient bandwidth or a low sampling rate will result in the loss of waveform details, smoothing out minute oscillations, overshoots, and fluctuations that reflect the fault mechanism, making them unrecognizable. A high-bandwidth probe ensures that no additional parasitic effects are introduced during signal acquisition, thus accurately reflecting the inherent dynamic characteristics of the gate circuit. The high-sampling-rate analog-to-digital conversion guarantees the continuity and accuracy of the voltage waveform in the time domain, providing a reliable data foundation for subsequent calculations of rising and falling edge slopes and dispersion statistics. This design essentially protects the integrity of waveform information. By improving time resolution, the transient characteristics of the fault can be accurately quantified, ensuring that subsequent judgments are based on real, resolvable physical signals, thus making the fault prediction results reliable and discriminative.

[0043] In a preferred embodiment of this invention, plotting the voltage waveform includes:

[0044] Using timestamps as the horizontal axis and voltage values ​​as the vertical axis, data points are generated in the coordinate system. Adjacent data points are connected by a smooth curve to form a continuous curve, thus generating a voltage waveform. The drawing of the voltage waveform also involves smoothing the voltage values ​​to reduce the impact of noise. The smoothing process uses a moving average method, and the window size of the moving average is adjusted according to the sampling rate.

[0045] It should be noted that the smoothing process is a method of continuously correcting the sampled discrete voltage values ​​during the voltage waveform plotting stage. This is achieved by using a moving average to weight and average adjacent data points over the time series. This can include: before generating the voltage waveform, forming a window by combining the voltage value of each sample point with the voltage values ​​of several sample points before and after it according to the sampling interval. All voltage values ​​within the window are averaged, and the resulting average replaces the original voltage value at that position. As the time axis progresses, this window continuously slides across the entire data series, thus achieving smoothing correction of the entire voltage signal. Since the window size is related to the sampling rate, when the sampling rate is high, the time interval covered by the window is correspondingly smaller to avoid over-smoothing the signal and losing transient details; when the sampling rate is low, the window range is relatively larger to obtain sufficient smoothing effect between sparser data points.

[0046] Taking advantage of the physical continuity between adjacent data points in time, and assuming that the actual changes in the signal will not change abruptly in a very short period of time, the influence of random noise can be offset by averaging between nearby samples. However, gate drive signals contain high-frequency noise and parasitic oscillations. Directly analyzing the slope or edge features without smoothing can easily lead to misjudgments due to local interference. Moving averages, while maintaining the overall signal shape and edge transition characteristics, reduce the transient bias introduced by noise, making the subsequently extracted rise time, fall time, and average slope more reflective of the true electrical dynamic behavior.

[0047] Obtain the rising edge Si and falling edge Xi in the voltage waveform of switching cycle i, count the number of times the rising edge Si is greater than the preset stable high level, and count the number CS in the entire switching cycle; count the number of times the falling edge Xi is less than the preset stable low level, and count the number CX in the entire switching cycle.

[0048] It is worth noting that the number CS is the sum of the number of times the rising edge Si is greater than the preset stable high level in all switching cycles; the number CX is the sum of the number of times the falling edge Xi is less than the preset stable low level in all switching cycles.

[0049] Calculate the rise time and fall time of switching period i based on the voltage waveform of switching period i, and calculate the average rise slope and average fall slope of switching period i.

[0050] It should be noted that the rise time is calculated based on the transition of the gate voltage waveform from a stable low level to a stable high level within one switching cycle. The time interval is determined by identifying the start and end points of the rising edge in the waveform. The start point corresponds to the position where the voltage begins to leave the stable low level, and the end point corresponds to the position where the voltage reaches a stable high level and tends to stabilize. The time difference between these two points is the rise time. The fall time is calculated in the same way, except that it corresponds to the transition of the waveform from a stable high level to a stable low level. The time difference is obtained by determining the time positions when the voltage begins to fall and reaches a stable low level. The average rise slope is the result of integrating or weighting the ratio of the voltage change at the rising edge to the rise time over the time series, reflecting the overall rate of change of the voltage during the rise phase. The average fall slope is calculated in the same way, characterizing the rate of change of the fall process by the average ratio of the voltage change at the falling edge to the fall time.

[0051] The average rising slope and average falling slope of all switching cycles are statistically distributed to obtain the dispersion index of the average rising slope and average falling slope, which are denoted as rising index and falling index, respectively.

[0052] In another preferred embodiment of the present invention, the calculation of the rising index and the falling index includes:

[0053] Calculate the initial parameters for all average rising and falling slopes, including variance, standard deviation, and skewness.

[0054] The ascent index is calculated based on the initial parameters of the superior and inferior solution distance method and the average ascent slope.

[0055] The descent index is calculated based on the initial parameters of the superior-inferior solution distance method and the average descent slope.

[0056] The cause of the fault is determined based on the rising index, falling index, number of times CS and number of times CX. The causes of the fault include the parasitic inductance effect of the gate circuit and the insufficient load capacity of the driver integrated circuit.

[0057] In another preferred embodiment of the present invention, the method further includes, before determining the cause of the fault:

[0058] If the rising indicator and / or falling indicator exceed the preset reference range, or the number of times CS and / or the number of times CX are greater than the preset number of times threshold, then a fault is determined to exist;

[0059] If no fault is found, the cause of the fault will not be determined.

[0060] The reference ranges for the upward and downward indicators are different.

[0061] It is important to note that in this invention, the presence of a fault is determined by the rise index, fall index, and the number of overshoots and undershoots. This is based on the stability assessment principle of waveform statistical characteristics. Under normal operating conditions, the rise and fall of the gate drive waveform are determined by stable drive capability and loop parameters. The variation patterns between different switching cycles should exhibit high consistency, the voltage change rate should statistically maintain a concentrated distribution, and the waveform should be stable without significant transient deviations. When the system operates stably, these statistical characteristics are reflected in the rise and fall indices being within a preset reference range, indicating that the rise and fall slopes of the waveform fluctuate little between different cycles, and the system's dynamic behavior is controlled and repeatable.

[0062] If a disturbance or parameter anomaly occurs during system operation, the time response of the waveform will lose consistency, manifesting as a diffusion in the statistical distribution of the rising or falling slope, causing the corresponding indicators to exceed the reference range. Simultaneously, if multiple voltage deviations exceeding the steady-state level appear in the waveform, i.e., frequent overshoot or undershoot events, it indicates that the system's transient characteristics have undergone significant abnormalities.

[0063] In another preferred embodiment of the present invention, the process of determining the cause of the fault includes:

[0064] When the rising indicator is greater than its preset threshold and the falling indicator is greater than its preset threshold:

[0065] If CS≠0 and / or CX≠0, the fault is determined to be caused by the parasitic inductance effect of the gate circuit; if CS=0 and CX=0, the fault is determined to be caused by insufficient load capacity of the driver integrated circuit.

[0066] When the rising index is greater than its preset threshold and the falling index is less than or equal to its preset threshold, or when the rising index is less than or equal to its preset threshold and the falling index is greater than its preset threshold, the fault is determined to be due to insufficient load capacity of the driver integrated circuit.

[0067] The overall concept of this invention is to fundamentally change the traditional approach of relying on empirical judgment based on the shape of a single gate drive waveform. Instead, it uses statistical dynamics and transient characteristic analysis as its core to systematically model and extract features from the gate drive signals of multiple consecutive switching cycles, thereby achieving differentiation at the fault mechanism level. First, the voltage signal of the gate drive node is non-invasively acquired using a high-bandwidth probe, and the analog waveform is converted into discrete voltage value data, generating a continuous voltage waveform based on timestamps. Next, the rise time, fall time, and average slope of the rising and falling edges of the waveform in each switching cycle are extracted to quantify the voltage change rate characteristics in each cycle. Subsequently, the average rise slope and average fall slope of all cycles are statistically analyzed to calculate their dispersion index, reflecting the consistency or instability of the waveform's dynamic changes. Simultaneously, transient events exceeding the stable high level or falling below the stable low level in the waveform are detected, and the number of overshoots and undershoots is counted to determine the presence of inductive oscillations.

[0068] The principle for determining the cause of a fault is based on an understanding of the physical mechanisms of two typical fault types: If parasitic inductance exists in the gate circuit, a reverse induced electromotive force will be generated when the drive current changes abruptly, forming LC oscillations. This manifests as significant waveform overshoot, undershoot, and slope fluctuations, i.e., high dispersion and oscillation events. Conversely, when the drive integrated circuit has insufficient load-carrying capacity, its output current is limited, resulting in a slow rise or fall of the waveform, but overall smooth and stable, without oscillations, only potentially increased slope dispersion. Based on this mechanism, the system uses dual-channel judgment logic to comprehensively analyze the collected data: if the slope dispersion is high and overshoot or undershoot is detected, it is determined to be the effect of parasitic inductance in the gate circuit; if the slope dispersion is high but no oscillation event is detected, it is determined to be insufficient load-carrying capacity of the drive circuit; if only a single edge shows abnormal dispersion, it is also attributed to the asymmetric degradation of the drive chip. Through this process, the present invention extracts quantitative features reflecting the root cause of the fault from the appearance of "waveform sluggishness", replaces manual experience judgment with calculable statistical indicators, and constructs a systematic diagnostic method that can automatically identify and distinguish gate faults caused by different physical mechanisms.

[0069] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A method for detecting gate circuit inductance for fault prediction, characterized in that, Includes the following steps: The voltage signal of the gate drive node is acquired in multiple consecutive switching cycles, the voltage signal is converted into a voltage value, and the voltage waveform is plotted based on the voltage value. Obtain the rising edge Si and falling edge Xi in the voltage waveform of switching cycle i, count the number of times the rising edge Si is greater than the preset stable high level, and count the number CS in the entire switching cycle; count the number of times the falling edge Xi is less than the preset stable low level, and count the number CX in the entire switching cycle. Calculate the rise time and fall time of switching period i based on the voltage waveform of switching period i, and calculate the average rise slope and average fall slope of switching period i. The average rising slope and average falling slope of all switching cycles are statistically distributed to obtain the dispersion index of the average rising slope and average falling slope, which are denoted as rising index and falling index, respectively. The cause of the fault is determined based on the rising index, falling index, number of times CS and number of times CX. The causes of the fault include the parasitic inductance effect of the gate circuit and the insufficient load capacity of the driver integrated circuit. The process of determining the cause of a fault includes: When the rising indicator is greater than its preset threshold and the falling indicator is greater than its preset threshold: If CS≠0 and / or CX≠0, the fault is determined to be caused by the parasitic inductance effect of the gate circuit; if CS=0 and CX=0, the fault is determined to be caused by insufficient load capacity of the driver integrated circuit. When the rising index is greater than its preset threshold and the falling index is less than or equal to its preset threshold, or when the rising index is less than or equal to its preset threshold and the falling index is greater than its preset threshold, the fault is determined to be due to insufficient load capacity of the driver integrated circuit.

2. The gate circuit inductance detection method for fault prediction according to claim 1, characterized in that, Obtaining voltage values ​​includes: The voltage signal of the gate drive node in multiple consecutive switching cycles is acquired by a high-bandwidth voltage probe, which is connected to the gate drive node in a non-invasive manner to capture the analog voltage signal. The analog voltage signal is converted into a voltage value by an analog-to-digital converter. The sampling rate of the analog-to-digital converter is set to be a multiple of the switching frequency to ensure that the details of the voltage signal change are captured. The conversion process includes sampling the analog voltage signal at equal intervals. Each sampling point generates a voltage value and attaches a corresponding timestamp, which is based on the system clock synchronization.

3. The gate circuit inductance detection method for fault prediction according to claim 2, characterized in that, Drawing voltage waveforms includes: Using timestamps as the horizontal axis and voltage values ​​as the vertical axis, data points are generated in the coordinate system. Adjacent data points are connected by a smooth curve to form a continuous curve, thus generating a voltage waveform. The drawing of the voltage waveform also involves smoothing the voltage values ​​to reduce the impact of noise. The smoothing process uses a moving average method, and the window size of the moving average is adjusted according to the sampling rate.

4. The gate circuit inductance detection method for fault prediction according to claim 1, characterized in that, The calculation of rising and falling indicators includes: Calculate the initial parameters for all average rising and falling slopes, including variance, standard deviation, and skewness. The ascent index is calculated based on the initial parameters of the superior and inferior solution distance method and the average ascent slope. The descent index is calculated based on the initial parameters of the superior-inferior solution distance method and the average descent slope.

5. The gate circuit inductance detection method for fault prediction according to claim 1, characterized in that, Before determining the cause of the malfunction, the following should also be included: If the rising indicator and / or falling indicator exceed the preset reference range, or the number of times CS and / or the number of times CX are greater than the preset number of times threshold, then a fault is determined to exist; If no fault is found, the cause of the fault will not be determined. The reference ranges for the upward and downward indicators are different.

Citation Information

Patent Citations

  • RC-IGBT fault detection and diagnosis method

    CN119270020A

  • Method and system for testing direct-current offset compensation circuit of optical transceiving driving chip, and medium

    CN121186576A