A precision current-limiting control method and system based on two-state mutual authentication

Abstract

The present application relates to the technical field of current automatic regulation, and particularly relates to a precision current limiting control method and system based on double-state mutual verification. The method comprises the following steps: obtaining initial detection data, wherein the initial detection data comprises current detection voltage data at both ends of a current detection resistor, drain-source voltage data and gate voltage data of a power switch tube; performing synchronous correlation according to the initial detection data to obtain current limiting detection data; performing time domain transition deconstruction on the current limiting detection data to obtain transient characteristic data; performing multi-characteristic out-of-limit judgment and edge trend inference according to the transient characteristic data to obtain first overcurrent state data and second overcurrent state data; performing double-state mutual verification fusion according to the first overcurrent state data and the second overcurrent state data to obtain overcurrent state data; performing stage regulation and control according to the overcurrent state data to obtain gate control data; and performing feedback correction on the gate control data to obtain current limiting control data. The present application improves current limiting precision and protection reliability.

Description

Technical Field

[0001] This invention relates to the field of automatic current regulation technology, and in particular to a precision current limiting control method and system based on dual-state mutual verification. Background Technology

[0002] In power supply systems for servers, communication equipment, hot-swappable boards, and precision electronic devices, load startup, operational state switching, partial faults, or momentary short circuits can all cause a rapid increase in load current. If overcurrent cannot be limited in time, the power switching transistor, power module, and downstream loads can easily be subjected to significant current surges and thermal stress. Therefore, existing power supply systems typically place a current-sensing resistor and a power switching transistor between the power input and the load. The load current is determined by detecting the voltage across the current-sensing resistor, and the gate voltage of the power switching transistor is adjusted when the load current exceeds a set value to achieve current limiting or shutdown protection.

[0003] However, during rapid changes in load current, the parasitic inductance of the current sensing resistor, the trace resistance of the printed circuit board, the gate charge of the power switch, and the sampling delay of the detection channel all affect the overcurrent detection and gate turn-off process. Under sudden short-circuit conditions, the drain-source voltage rises rapidly before the current sensing voltage stabilizes and reaches the current limiting threshold, while the gate voltage of the power switch remains at a high level, resulting in a relatively delayed overcurrent control action. Simultaneously, switching noise, parasitic oscillations, and transient voltage glitches can also cause the current sensing signal to exceed the limit for short periods, leading to false triggering of control methods that rely solely on a single current sensing voltage threshold. Traditional current limiting circuits suffer from transient action delays, detection accuracy affected by parasitic parameters, and a difficulty in balancing current limiting speed and accuracy. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a precision current limiting control method and system based on dual-state mutual verification, thereby resolving at least one of the aforementioned technical problems.

[0005] This application provides a precise current limiting control method based on dual-state mutual verification, including the following steps: Acquire initial detection data, which includes current sensing voltage data across the current sensing resistor, drain-source voltage data of the power switch, and gate voltage data. Synchronous correlation is performed based on the initial detection data to obtain the rate limiting detection data; Temporal transition deconstruction is performed on the current limiting detection data to obtain transient feature data; Based on transient feature data, multi-signal limit judgment and edge situation inference are performed to obtain first overcurrent state data and second overcurrent state data; based on the first overcurrent state data and second overcurrent state data, dual-state mutual verification fusion is performed to obtain overcurrent state data; Gate control data is obtained by performing phased adjustments based on overcurrent state data. Feedback correction is performed on the gate control data to obtain current limiting control data.

[0006] This invention synchronously correlates the current detection voltage, the drain-source voltage of the power switch, and the gate voltage, enabling the representation of the linkage between load current changes, power switch voltage changes, and gate drive actions under a unified time reference. By extracting the transition amplitude, slope, duration, and cross-circuit time difference of each detection signal through time-domain transition deconstruction, the interference of baseline drift and transient glitches on the overcurrent identification results can be effectively reduced. Two types of overcurrent states are formed using multi-signal overcurrent judgment and edge condition inference, respectively. Cross-verification of gradual overcurrent, sudden short circuit, and isolated interference is performed through dual-state mutual verification fusion, thereby reducing the risk of false triggering and missed detection caused by single threshold judgment. Staged gate control is implemented based on the fused overcurrent state, and feedback correction is performed based on the actual gate action and current limiting deviation. This dynamically corrects the gate pull-down strength, control timing, and holding state, enabling the power switch to achieve a more matched turn-off process under different overcurrent conditions. This helps suppress current spikes, shorten the current limiting action time, and improve the stability, accuracy, and circuit protection reliability of current limiting control.

[0007] Optionally, the temporal transition deconstruction includes: Multi-channel timing alignment is performed based on current limiting detection data to obtain transition window data; baseline drift stripping is performed on the transition window data to obtain purified transition data; edge slope analysis is performed on the purified transition data to obtain edge evolution data; cross-channel time difference mapping is performed based on the edge evolution data to obtain transition correlation data; and pulse continuity extraction is performed based on the transition correlation data to obtain transient feature data.

[0008] In this invention, by performing multi-channel timing alignment on the current limiting detection data, the current sensing voltage, drain-source voltage, and gate voltage can be uniformly mapped to the same sampling time axis, avoiding deviations in transient judgment caused by sampling delays from different detection channels. Through baseline drift stripping, interference from temperature drift, device bias, and long-term slowly varying components on the transition signal can be eliminated, retaining the effective change components directly related to the overcurrent event. The direction and rate of change of each signal are obtained using edge slope analysis, and the sequential correlation between current limit exceedance, drain-source voltage surge, and gate voltage drop is established through cross-channel time difference mapping. Furthermore, the duration of abnormal changes is determined by pulse duration extraction. The resulting transient characteristic data can simultaneously characterize the transition amplitude, rate of change, channel time difference, and duration, contributing to improved real-time performance and anti-interference capability of current limiting control.

[0009] Optionally, the multi-feature limit-crossing judgment includes: Feature domain segmentation is performed on transient feature data to obtain over-limit feature data; threshold domain mapping is performed on over-limit feature data to obtain threshold domain associated data; amplitude-time joint judgment is performed on threshold domain associated data to obtain over-limit situation data; multi-signal mutual verification is performed on over-limit situation data to obtain over-limit judgment data; state merging is performed on over-limit judgment data to obtain first overcurrent state data.

[0010] In this invention, by performing feature domain segmentation on transient feature data, features of different dimensions such as transition amplitude, change slope, duration, and cross-circuit time difference can be respectively classified into corresponding judgment domains, avoiding judgment distortion caused by direct mixing of multiple types of features. Through threshold mapping, a clear correspondence is established between various transient features and their corresponding normal intervals, critical intervals, and over-limit intervals. Amplitude-time joint judgment examines both the over-limit amplitude and duration, which can eliminate transient glitches with small amplitudes or excessively short durations, while retaining continuous anomalies with actual overcurrent significance. Multi-signature cross-verification utilizes the coordinated change relationship of current detection voltage, drain-source voltage, and gate voltage to cross-verify the judgment results of a single channel, reducing misjudgments caused by sensor noise, device temperature drift, and local parasitic parameters. Finally, through state merging, the scattered over-limit judgment results are unified into first overcurrent state data with overcurrent type, severity, and duration, thereby improving the accuracy, stability, and engineering applicability of progressive overcurrent and transient short-circuit identification.

[0011] Optionally, the feature domains include: Channel splitting is performed based on transient feature data to obtain path feature data; the path feature data is categorized by dimension to obtain domain candidate data; benchmark normalization is performed based on the domain candidate data to obtain standard domain data; threshold division is performed based on the standard domain data to obtain limit-crossing feature data.

[0012] This invention utilizes channel splitting to extract features such as amplitude changes, slope changes, and durations corresponding to current sensing voltage, drain-source voltage, and gate voltage, avoiding overlap between different detection channels. Subsequently, features are categorized according to physical dimensions such as amplitude, rate of change, duration, and cross-channel time difference, placing similar features into a unified analysis domain. Benchmark normalization weakens the scale effect caused by differences in channel range, sampling gain, and device bias, improving the comparability of multi-dimensional features. Threshold division maps standardized features to normal, critical, and over-limit regions, forming over-limit feature data with clear channel origin, feature category, and over-limit degree, which helps improve the accuracy and stability of overcurrent identification.

[0013] Optionally, the amplitude-time joint judgment includes: Amplitude deviation quantization is performed on threshold correlation data to obtain amplitude over-limit data; continuous interval tracking is performed on amplitude over-limit data to obtain over-limit continuous data; amplitude-time pairing is performed on over-limit continuous data to obtain amplitude-time coupled data; spurs are removed from amplitude-time coupled data to obtain over-limit situation data.

[0014] This invention utilizes amplitude deviation quantization on threshold-related data to accurately characterize the extent to which each transient feature exceeds its corresponding threshold boundary, providing a comparable quantitative basis for the over-limit intensity of different detection channels. Continuous interval tracking of amplitude over-limit data allows for the determination of the start and end times of abnormal signals and their duration of continuous operation, avoiding overcurrent judgments based solely on a single sampling point. Amplitude-time pairing binds the over-limit amplitude and duration to the same transient event, creating a joint representation of overcurrent intensity and abnormal duration. Furthermore, glitch removal eliminates short-term isolated pulses caused by switching noise, parasitic inductance, and sampling jitter, retaining the continuous and potentially harmful over-limit process. This results in over-limit situational data reflecting the over-limit level, duration, and evolution trend, thereby improving the anti-interference capability, judgment stability, and reliability of subsequent current limiting control in overcurrent identification.

[0015] Optionally, edge situation inference includes: The transient characteristic data is segmented at the edge to obtain edge sequence data; the edge sequence data is oriented by slope to obtain edge direction data; cross-path time difference tracking is performed based on the edge direction data to obtain edge time difference data; the situation is recursively extrapolated based on the edge time difference data to obtain edge situation data; and the operating conditions are classified based on the edge situation data to obtain the second overcurrent state data.

[0016] This invention utilizes edge segmentation to analyze transient characteristic data, generating continuous sequences of current detection voltage rise edges, drain-source voltage rise edges, and gate voltage fall edges, thus preventing the overlap of local abrupt changes in different channels. Slope orientation clarifies the direction, rate, and steepness of each edge change, enabling the identification of the correspondence between load current growth, power switch voltage rise, and gate turn-off action. Cross-channel time difference tracking determines the trigger sequence and conduction hysteresis of each channel's edges, and situational recursion continuously judges the development trend of abnormal edges, avoiding isolated judgments based on a single moment. Classifying operating conditions based on edge conditions effectively distinguishes between normal fluctuations, gradual overcurrent, sudden short circuits, and gate turn-off hysteresis, yielding second overcurrent state data with time evolution characteristics, thereby improving the real-time performance, anti-glitch capability, and accuracy of operating condition determination in overcurrent identification.

[0017] Optionally, dual-state mutual verification fusion includes: The first and second overcurrent state data are time-stamped to obtain dual-state corresponding data; the dual-state corresponding data is verified to obtain dual-state verification data; the dual-state verification data is used to trace the cause of divergence to obtain state divergence data; the state divergence data is used to determine the protection priority of the state divergence data to obtain protection priority data; and the protection priority data is merged into dual states to obtain overcurrent state data.

[0018] In this invention, by time-stamping the first and second overcurrent state data, the amplitude-duration characteristic results obtained based on multi-signal overcurrent judgment can be associated with the time-series-trend characteristic results obtained based on edge situation inference to the same abnormal event, avoiding erroneous matching between different sampling periods or different fault processes. Furthermore, state verification is used to cross-validate the consistency, level differences, and duration of the two types of judgment results. When the two types of results are inconsistent, divergence tracing can be used to identify sources of divergence such as transient glitches, gradual overcurrent, sudden short circuits, or insufficient turn-off by combining overcurrent amplitude, duration, edge sequence, and gate action hysteresis. Control priority is determined according to the protection priority principle, and unified overcurrent state data is formed through dual-state merging. This reduces missed detections and false triggers caused by single-path judgment, improving the reliability, robustness, and targeted nature of overcurrent identification under complex transient conditions, as well as the subsequent gate control.

[0019] Optionally, phased regulation includes: The gate stage data is obtained by performing stage mapping based on the overcurrent state data; the gate stage data is then pulled down and strengthened to obtain the gate drive data; the gate drive data is then timed to obtain the gate timing data; and the gate timing data is then integrated with the instructions to obtain the gate control data.

[0020] This invention maps overcurrent state data into stages, corresponding different operating conditions such as normal fluctuations, gradual overcurrent, sudden short circuits, and insufficient turn-off to the normal gate drive, pre-discharge, rapid pull-down, and turn-off holding stages, respectively. This avoids over-control or lag caused by using a single turn-off strategy. By strengthening the pull-down, the gate discharge intensity is configured according to the overcurrent level and trend, maintaining stable current limiting under gradual overcurrent conditions and rapidly releasing gate charge under sudden short circuit conditions. Subsequently, the start time, duration, and switching sequence of each control stage are determined through timing arrangement, and the pull-down intensity and control timing are integrated into instructions to form gate control data that can be directly applied to the gate drive branch. This can shorten the turn-off time of the power switch, suppress short-circuit current spikes, reduce the risk of gate voltage oscillation, repeated conduction, and excessive turn-off, and improve the dynamic matching capability, execution stability, and power device protection reliability of the current limiting control.

[0021] Optionally, feedback correction includes: Drive data is sampled back from the gate control data to obtain gate action data; deviation is calculated from the gate action data to obtain current limiting deviation data; hold data is determined from the current limiting deviation data to obtain feedback adjustment data; and instruction correction is applied to the feedback adjustment data to obtain current limiting control data.

[0022] This invention utilizes drive sampling of gate control data to acquire real-time information on the actual gate voltage changes, action times, and holding states of the power switch during pre-discharge, rapid pull-down, and turn-off holding phases, thereby determining whether control commands are accurately executed. Through deviation back-calculation, the actual current-limiting current, gate action delay, and current fallback rate are compared with the target control value to generate current-limiting deviation data reflecting insufficient current limiting, hysteresis, or excessive turn-off. Based on the current-limiting deviation, a holding determination is made to determine whether the current gate pull-down state should be maintained, strengthened, weakened, or exited, preventing the power switch from prematurely resuming conduction before the abnormality is resolved. Through command correction, the gate pull-down strength, duration, and stage switching times are dynamically adjusted, making subsequent control commands more closely match actual load changes. This helps suppress current overshoot and gate oscillation, shortens current-limiting adjustment time, and improves the stability, accuracy, and adaptability of closed-loop current-limiting control.

[0023] Optionally, this application also provides a precision current limiting control system based on dual-state mutual verification, used to execute the precision current limiting control method based on dual-state mutual verification as described above. The precision current limiting control system based on dual-state mutual verification includes: The detection and acquisition module is used to acquire initial detection data, which includes the current sensing voltage data across the current sensing resistor, the drain-source voltage data of the power switch, and the gate voltage data. The synchronization and association module is used to perform synchronization and association based on the initial detection data to obtain the rate limiting detection data. The temporal transition deconstruction module is used to perform temporal transition deconstruction on the current limiting detection data to obtain transient feature data. The dual-state judgment module is used to perform multi-signal limit judgment and edge situation inference based on transient feature data to obtain first overcurrent state data and second overcurrent state data; and to perform dual-state mutual verification fusion based on the first overcurrent state data and the second overcurrent state data to obtain overcurrent state data. The stage control module is used to perform stage control based on the overcurrent state data to obtain gate control data; The feedback correction module is used to perform feedback correction on the gate control data to obtain current limiting control data.

[0024] The purpose of this invention is to synchronously correlate the current sensing voltage across the current sensing resistor, the drain-source voltage of the power switch, and the gate voltage, so that changes in load current, changes in the voltage across the power switch, and the gate drive action are unified under the same time reference, reducing state mismatch caused by sampling lag between different detection channels. By extracting the transition amplitude, edge slope, duration, and cross-channel time difference of each signal through time-domain transition deconstruction, the interference of baseline drift, parasitic inductance glitches, and sampling jitter on overcurrent identification is reduced. On the one hand, multi-signal over-limit judgment is used to identify the abnormal intensity from the over-limit amplitude and duration; on the other hand, edge situation inference is used to determine the abnormal evolution path from the sequential relationship of changes in current sensing voltage, drain-source voltage, and gate voltage. Then, through dual-state mutual verification fusion, consistency verification and discrepancy resolution of the two types of results are performed, which can more accurately distinguish between normal fluctuations, gradual overcurrent, sudden short circuits, and turn-off hysteresis. The gate pull-down strength, action sequence, and turn-off holding stage are configured according to the fused overcurrent state. Feedback correction is performed based on the actual gate action, current drop-off degree, and current limiting deviation, so that the control command can be dynamically modified according to the actual load state. This balances the rapid turn-off under short-circuit conditions and the smooth current limiting under gradual overcurrent conditions, reduces the risk of current spikes, false triggering, and repeated conduction, and improves the response speed, control accuracy, and power device protection reliability of precision current limiting control. Attached Figure Description

[0025] Other features, objects, and advantages of this application will become more apparent from the following detailed description of the non-limiting embodiments, taken with reference to the accompanying drawings: Figure 1 A flowchart illustrating the steps of a precision current limiting control method based on dual-state mutual verification according to an embodiment is shown. Figure 2 A flowchart illustrating the steps of a temporal transition deconstruction method according to an embodiment is shown; Figure 3 A flowchart illustrating the steps of a multi-feature limit-crossing detection method according to an embodiment is shown. Figure 4 A flowchart illustrating the steps of a stage control method according to one embodiment is shown. Figure 5 A flowchart illustrating the steps of a feedback correction method according to one embodiment is shown. Detailed Implementation

[0026] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0027] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. Functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0028] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0029] The precision current limiting control method is applied to a hot-swappable power supply board with a rated input voltage of 48V, using a current sensing resistor. 10mΩ, target current limiting value The current sensing amplitude threshold is set to 5.0A, the sampling period Δt is set to 0.2μs, and the current sensing amplitude threshold is set to 0.2μs. Use 50mV, current detection slope threshold Take 0.02V / μs as the drain-source voltage slope threshold. The effective duration threshold is set at 2.0 V / μs. Set the threshold for cross-channel edge correlation to 1.0 μs. Take 2.0 μs.

[0030] During testing, the electronic load was switched from a stable 0.3A load state to a short-circuit state. The system synchronously acquired the current sensing voltage, drain-source voltage, and gate-source voltage, and extracted the current sensing voltage transition voltage change within the 17th transition window. Peak slope of the flow detection channel Flow detection channel pulse duration Peak slope of drain-source channel Because the current sensing voltage continuously exceeded the limit and the drain-source voltage rapidly increased, the multi-signal over-limit judgment system classified the event as a sudden short circuit.

[0031] The rise time of the drain-source voltage is 0.4 μs earlier than the rise time of the current sensing voltage, that is: The gate-source voltage falling edge appears 0.8 μs after the drain-source voltage rising edge, that is: The three edges form a complete edge chain of "sudden rise in drain-source voltage - current detection voltage exceeding limit - gate-source voltage drop," and the edge condition inference also classifies the event as a sudden short circuit. Dual-state mutual verification fusion confirms that the two states are consistent, sets the event protection priority to the highest level, and directly enters the fast pull-down phase.

[0032] The system controls the active pull-down transistor to turn on, and the initial fast pull-down duration is... Set the turn-off hold time to 3.0 μs. The time was 5.0 μs. After the initial control, the actual current peak was detected within the feedback window. It is 5.72A, and the gate operation delay is... The current limiting deviation is 1.4 μs. The current draw was 0.72A, which is greater than the allowable current limiting deviation threshold of 0.5A. Therefore, the fast pull-down duration was corrected to 4.0μs. In the subsequent same short-circuit test, the actual peak current decreased to 5.31A, and the gate action delay was shortened to 0.9μs. The load current at the end of the feedback window decreased to 4.2A, the gate-source voltage decreased to 0.72V, and the drain-source voltage fluctuation within 5 consecutive sampling points was 0.11V, which met the overcurrent release condition, and the system exited active pull-down.

[0033] Under the same hardware conditions, compared with a control method using only 50mV single threshold detection and fixed pull-down duration, the average peak current of 20 repeated short-circuit tests was 7.88A, and the average gate action delay was 3.06μs. After adopting the method described in this embodiment, the average peak current was 5.34A, and the average gate action delay was 0.94μs, representing a reduction of approximately 32.2% in peak current and a reduction of approximately 69.3% in gate action delay. The current overshoot ratio relative to the target current limit of 5.0A decreased from 57.6% to 6.8%. Furthermore, in 200 tests with interference pulses lasting less than 1.0μs, the control method generated 23 false triggers, while this embodiment generated only 1 false trigger, reducing the number of false triggers by approximately 95.7%. This achieves rapid suppression of sudden short circuits and improves the anti-interference capability and control stability of precision current limiting control.

[0034] Please see Figures 1 to 5 This application provides a precise current limiting control method based on dual-state mutual verification, including the following steps: 01. Acquire initial detection data, which includes the current sensing voltage data across the current sensing resistor, the drain-source voltage data of the power switch, and the gate voltage data. In one embodiment, in the current sensing resistor The first differential sampling channel is set at both ends of the power switch transistor. A second differential sampling channel is set between the drain and source of the power switch. A third differential sampling channel is set between the gate and source to acquire the current sensing voltage, drain-source voltage, and gate voltage, respectively. The three sampling channels are triggered by the same sampling clock for analog-to-digital conversion, continuously acquiring data at a sampling period Δt, and recording the same sampling sequence number k and timestamp for the three sampling results within the same trigger period. The first differential sampling channel is connected to current sensing resistors using a Kelvin connection. The two main terminals of the current sensing resistor are amplified by a low-offset differential amplifier circuit to amplify the small voltage difference across the current sensing resistor to the analog-to-digital converter input range, which is then used to output the current sensing voltage. The second differential sampling channel is connected to the power switching transistor. The drain and source terminals are connected, and the drain-source voltage is range-shifted through a high input impedance differential attenuation circuit for output drain-source voltage. The third differential sampling channel is connected to the power switching transistors respectively. The gate and source are connected, and the gate-source voltage is acquired through a high input impedance buffer circuit for output gate voltage. Each of the three channels has an independent sample-and-hold unit, triggered simultaneously by the same clock signal, to avoid time differences between channels caused by analog switch polling sampling. The sampling period Δt should be less than one-fifth of the shortest transient change time to be identified, expressed as: ,in, This refers to the shortest characteristic duration during the current sensing voltage rise, drain-source voltage surge, or gate voltage drop process. Considering the microsecond-level transient process of a precision current limiting circuit, in one embodiment, the sampling period can be set to Δt = 0.2, corresponding to a sampling frequency per channel of: This value allows for obtaining 5 sampling points in a 1μs transient process and 10 sampling points in a 2μs transient process, which meets the calculation requirements for edge slope, cross-path time difference, and pulse duration. In actual implementation, Δt is set to 0.2μs.

[0035] After the device is powered on, and with the power switch in a preset stable state, zero-point calibration is performed on the three sampling channels to obtain the zero-point code value. , and The actual voltage of each channel at the k-th sampling time is converted according to the following formula: ,in, ; This represents the analog-to-digital conversion code value of the channel corresponding to the k-th sampling time. This is the zero-point code value for the corresponding channel; The reference voltage for analog-to-digital conversion; n is the number of bits in the analog-to-digital conversion; This is the front-end amplification gain for the corresponding sampling channel; , and These are the current sensing voltage, drain-source voltage, and gate voltage, respectively.

[0036] During the acquisition process, the validity of data from each channel is checked. If the sampling code value reaches the full range of the analog-to-digital conversion, data is continuously missing in two adjacent sampling periods, or any channel under the same sampling number fails to complete acquisition, the corresponding sampling group is marked as invalid and re-acquired; if all three data streams are within their corresponding ranges, , , Sampling sequence number k and timestamp It is packaged into a set of initial detection data.

[0037] 02. Synchronize and correlate the initial detection data to obtain the rate limiting detection data; In one embodiment, the sampling sequence number k and timestamp are read from the initial detection data. Current detection voltage Drain-source voltage and gate voltage Using the sampling sequence number k as the primary association index, three voltage signals with the same sampling sequence number are grouped into the same detection group. For the same detection group, the maximum difference between the three timestamps is calculated: .

[0038] when At that time, it is determined that the three data streams meet the synchronization conditions, among which, To preset the synchronization tolerance, which is no greater than one sampling period Δt, such as 0.05μs, the median value of the three timestamps is determined as the unified timestamp. And form a detection unit: .

[0039] When duplicate data exist under the same sampling number, retain the data with the closest timestamp. Data; when any data path is missing or At that time, the detection group is marked as an asynchronous group and will not participate in subsequent processing. The valid detection units are then arranged in ascending order according to a unified timestamp to obtain the current limiting detection data. Where N is the number of effective detection units, thus ensuring the time correspondence of the three voltages in the same transient event.

[0040] 03. Perform time-domain transition deconstruction on the current limiting detection data to obtain transient feature data; In one embodiment, the current limiting detection data is read. ,in A common time axis is constructed using a uniform sampling interval Δt. Isolated missing points are interpolated using linear interpolation between adjacent points; if any channel has more than two consecutive missing periods, the corresponding interval is discarded. For each channel... Calculate the adjacent difference when When the k-th sampling point is determined as a candidate transition point, the preceding P point and the following Q point are extracted to form a transition window; candidate points with a time interval less than QΔt are merged into the same window to obtain transition window data. The threshold value for the adjacent sampling voltage transition in the x-th detection channel is, for example... The voltage transition threshold for the current sensing channel, such as 2mV. This is the drain-source channel voltage transition threshold, such as 0.2V. This is the threshold voltage transition value for the gate-source channel, such as 0.3V.

[0041] For each transition window, the mean of P=20 samples before the transition point is used as the channel baseline: , ,in, Let x be the local baseline of the x-th detection channel within the r-th transition window. The cleanup voltage after baseline drift removal. Let the sampling sequence number be the candidate transition point corresponding to the r-th transition window; thus, the purified transition data is obtained. The edge slope is calculated using center difference. The first time the condition is met within the window... The time is recorded as the edge start time. If the subsequent R=3 consecutive sampling points all satisfy The last moment before that is denoted as the edge termination moment. , Let be the edge slope threshold of the x-th detection channel. Detection channel edge slope threshold Take 0.01 V / μs as the drain-source channel edge slope threshold. The threshold slope of the gate-source channel edge is set to 1.0 V / μs. Use 1.5 V / μs. Calculate the transition amplitude based on this. Peak slope And the edge direction; where positive values ​​represent rising edges and negative values ​​represent falling edges. Calculate the time difference between the three edges: , , And calculate the pulse duration of each channel. The corresponding transition windows , Edge direction The transient feature data is obtained by associating and encapsulating the three sets of cross-path time differences.

[0042] 04. Based on transient feature data, perform multi-signal limit judgment and edge situation inference to obtain the first overcurrent state data and the second overcurrent state data; based on the first overcurrent state data and the second overcurrent state data, perform dual-state mutual verification fusion to obtain the overcurrent state data; In one embodiment, each transition window is treated as a transient event. Read the current sensing voltage transition amplitude from the transient characteristic data. Drain-source voltage transition amplitude Peak slope of each channel Pulse duration Edge direction and cross-road time difference , and ,in The rising edge is denoted as The falling edge is denoted as When the system performs multi-feature limit detection, it compares each feature with a preset current detection amplitude threshold. Detection slope threshold Drain-source voltage slope threshold and effective duration threshold Compare, It is 50mV. It is 0.02V / μs. It is 2.0V / μs. It is 1.0 μs. When , When the event is determined to have persistent flow detection limit violation characteristics / persistent limit violation; when both conditions are met... When the event exhibits a rapid rise in drain-source voltage, it is determined to have this characteristic; when only a single path exceeds the limit, it is classified as an isolated limit exceedance; when the limit exceedance duration is less than [a certain value], it is classified as an isolated limit exceedance. When only a single path characteristic exceeds the limit, it is judged as a transient glitch / short-term glitch. When the current sensing voltage continuously exceeds the limit but the drain-source voltage does not rise rapidly, the event is classified as a gradual overcurrent; when the current sensing voltage continuously exceeds the limit and the drain-source voltage rises rapidly, the event is classified as a sudden short circuit; thus, the first overcurrent state data is obtained.

[0043] When the system performs edge situation inference, it establishes a triggering sequence based on the direction and start time of the three edges. When the conditions are met... , , At that time, a complete edge chain is considered to be formed: current detection rise—drain-source rise—gate fall. If , If the current sensing voltage rises before the drain-source voltage, it is classified as a gradual overcurrent; , This indicates that the drain-source voltage rises sharply before the current sensing voltage, which is classified as a sudden short circuit; if an edge appears on a certain path, within a preset correlation time... If no secondary edges of other channels appear within 2.0 μs, it is classified as an isolated interference; the second overcurrent state data is obtained.

[0044] The first and second overcurrent state data are time-stamped using the event start time and transition window number. When the two state results are consistent, the corresponding overcurrent state is directly confirmed; when the two results are inconsistent, protection priority is determined according to the rules of complete cross-circuit edge chains taking precedence over isolated overcurrents, continuous overcurrents taking precedence over short-term spikes, and sudden short circuits taking precedence over gradual overcurrents. Specifically, if the second overcurrent state indicates that the drain-source voltage rises sharply first, and... If the first overcurrent state indicates that the current sensing voltage continuously exceeds the limit, while the second overcurrent state does not form a complete edge chain, then it is determined to be a gradual overcurrent; if both states have a duration of less than If an isolated edge occurs, it is identified as a transient disturbance. The event number, overcurrent type, occurrence time, duration, and protection priority are encapsulated into overcurrent status data.

[0045] 05. Perform phased regulation based on overcurrent status data to obtain gate control data; In one embodiment, the event number r and the overcurrent type are read from the overcurrent status data. The start time of the event Duration and protection priority ,in, This includes transient interference, gradual overcurrent, and sudden short circuit; the overcurrent status data also includes additional states. The additional state This includes situations without additional states or gate turn-off hysteresis. According to... Establish phase mapping rules: when During transient interference, maintain normal gate drive and disable the gate drive acceleration branch conduction; when For gradual overcurrent, a pre-discharge stage and a rapid pull-down stage are configured sequentially; when In the event of a sudden short circuit, the pre-discharge stage is skipped, and the system directly enters the rapid pull-down stage; thus, gate stage data is obtained.

[0046] During the pre-discharge phase, the gate discharge path of the current limiting controller itself releases part of the gate charge; during the rapid pull-down phase, the active pull-down transistor is controlled. Turn on, enabling the power switching transistor The gate through the gate resistor The gate discharge current is expressed as follows: ,in, This is the gate discharge current. This refers to the gate-source voltage of the power switch. Transistor On-state voltage drop, This is the gate discharge resistor. Progressive overcurrent corresponds to a shorter active pull-down duration. Sudden short circuit corresponding to continuous conduction duration ,and The duration of shutdown is: ,in, To preset the minimum holding time, such as 5.0 μs, This represents the duration of the current overcurrent event. The turn-off hold duration is used to maintain the gate drive acceleration branch or turn-off state after the power switch completes a rapid pull-down, preventing the power switch from prematurely resuming conduction before the load current has stabilized.

[0047] According to protection priority Determine the start time and switching sequence of each stage; when multiple overcurrent events overlap, prioritize the fast pull-down instruction corresponding to the event with the higher protection level. Encapsulate the stage identifier, start time, active pull-down enable state, discharge duration, and turn-off hold time into gate control data. ,in, This is the conduction control quantity for the gate drive acceleration branch.

[0048] 06. Feedback correction is performed on the gate control data to obtain current limiting control data.

[0049] In one embodiment, gate control data is executed. Then, to control the start time. Starting from the preset feedback window The current sensing voltage is reacquired within 20.0 μs. Drain-source voltage and gate voltage and in accordance with Calculate the actual load current. Extract the peak current within the feedback window. and gate turn-off time ,in, This is the preset turn-off gate voltage threshold for the power switch, such as 1.0V; the gate operation delay is... Based on the target rate limit value (e.g., 5.0A) Calculate the current limiting deviation: .when or When the current pull-down is insufficient, the gate drive acceleration branch is kept on, and the pull-down duration is adjusted to... , For current limiting deviation, To allow for a current limiting deviation threshold, such as 0.5A, The maximum allowable gate operation delay is, for example, 2.0 μs. This refers to the duration of a single pull-down, such as 1.0 μs; when , Furthermore, the drain-source voltage fluctuation within M consecutive sampling points at the end, Not greater than the stability threshold When the overcurrent condition is determined to be resolved, the pull-down will correct the energy to... And end the shutdown and maintain it. This represents the actual load current at the k-th sampling time. To preset the recovery current threshold, such as 4.5A, Set a preset gate-source turn-off voltage threshold for the power switch, such as 1.0V. This is the drain-source voltage stability threshold, such as 0.2V; otherwise, it remains constant. And let The event number r and the actual peak current are... Gate operation delay Current limiting deviation The revised pull-down reduces energy. Drop-down duration Off-hold time The maintained or exited state is encapsulated as current limiting control data.

[0050] Optionally, the temporal transition deconstruction includes: 31. Perform multi-channel timing alignment based on the current limiting detection data to obtain the transition window data; In one embodiment, the current limiting detection data is read. ,in, To standardize timestamps, , and These represent the current sensing voltage, drain-source voltage, and gate voltage at the k-th sampling time, respectively, where N is the number of effective sensing units. A common time axis is constructed using the sampling interval Δt; when any channel is missing only one sampling point, linear interpolation between adjacent points is used to fill the gap; when there are more than two consecutive missing sampling periods, the corresponding time period is discarded. For each channel... ,when When, the k-th sampling point is determined as a candidate transition point, where, This represents the preset transition threshold for the corresponding channel. P sampling points and Q=20 sampling points are extracted centered on the candidate transition point; adjacent candidate transition points with an interval less than QΔt are merged to form the r-th transition window. The three voltage levels, timestamps, and candidate transition positions within each transition window are encapsulated into transition window data.

[0051] 32. Baseline drift stripping is performed on the transition window data to obtain cleaned transition data; In one embodiment, for each transition window in the transition window data The three voltage sequences were read separately. Using the P samples preceding the candidate transition point as a stable reference segment, the local baseline of channel x was calculated: ,in, Let be the candidate transition point number for the r-th transition window. Baseline correction is performed on each sampled value within the window to eliminate zero-point bias, temperature drift, and slowly changing components within the window. The corrected three-channel purification voltage sequence, corresponding timestamp, and window number r are then encapsulated as purification transition data.

[0052] 33. Perform edge slope analysis on the purification transition data to obtain edge evolution data; In one embodiment, based on the purification transition data, the edge slope of each channel at the j-th sampling point is calculated using central difference: The first time the condition is met within the window... The sampling time is determined as the edge start time. ,in, This is the preset slope threshold for the corresponding channel. When the subsequent R consecutive sampling points all satisfy... When the last sampling time before the continuous interval is defined as the edge termination time, Calculate the transition amplitude and peak slope: , .when When, the direction of the edge is denoted as ascending; when When the edge direction is denoted as descent, the slope sequence, edge start and end times, transition amplitude, peak slope, and edge direction are encapsulated as edge evolution data.

[0053] 34. Perform cross-path time difference mapping based on edge evolution data to obtain transition correlation data; In one embodiment, based on edge evolution data, the edge start times of the current detection voltage, drain-source voltage, and gate voltage are associated with the same transition window number r, and calculated respectively: , , A positive time difference indicates that the edge of the preceding channel occurred first, while a negative time difference indicates that the edge of the following channel occurred first. If no valid edge is detected in a channel within its corresponding transition window, the time difference involving that channel is marked as missing and will not be included in subsequent edge priority determination. The three edges and their cross-channel time differences are associated according to the transition window number to obtain transition association data.

[0054] 35. Based on the transition correlation data, pulse-based continuous extraction is performed to obtain transient feature data.

[0055] In one embodiment, the pulse duration of each channel within the r-th transition window is calculated based on the edge start and edge end times of each channel in the transition correlation data. According to the transition window number r, the transition amplitudes corresponding to the current detection voltage, drain-source voltage, and gate voltage are... Peak slope Edge direction, pulse duration and cross-road time difference , and By performing correlation encapsulation, transient feature data is obtained.

[0056] Optionally, the multi-feature limit-crossing judgment includes: 41. Perform feature segmentation based on transient feature data to obtain limit-crossing feature data; In one embodiment, the transient events corresponding to each transition window are used. As a processing unit, the transition amplitude is read from the transient characteristic data. Peak slope Edge direction Duration and cross-road time difference, among which .Will and The peak slopes of each channel are then classified into the slope domain, and the amplitude domain is assigned to the slope domain. To be included in the persistent domain , and The data is incorporated into the time difference domain, and the corresponding event number r, channel identifier, and edge direction are retained to obtain the out-of-limit feature data.

[0057] 42. Perform threshold mapping on the out-of-limit feature data to obtain threshold-related data; In one embodiment, a one-to-one mapping is established between each feature in the out-of-limit feature data and its corresponding threshold. The following settings are configured: , , Continuous signage was installed at each channel. The conditions are Otherwise, take 0. Associate the feature values, corresponding thresholds, and out-of-limit identifiers according to the event number to obtain threshold-related data.

[0058] 43. Perform amplitude-time correlation judgment on threshold-related data to obtain over-limit situation data; In one embodiment, based on threshold correlation data, the out-of-limit amplitude or out-of-limit slope of the same event and the same channel is paired with its duration. The valid out-of-limit identifier for the current detection channel is: The indicator for a rapid rise in drain-source voltage is: When a certain feature reaches a corresponding threshold, but the duration is less than... If no other channels exhibit valid limit exceedances within the same event, the feature is marked as an isolated spike; if the amplitude or slope exceeds the limit and the duration reaches [a certain value], the spike is considered isolated. If an exception occurs, it is retained as a valid exception. The valid limit violation identifier, isolated spurious identifier, limit violation start and end times, and duration are encapsulated into limit violation status data.

[0059] 44. Perform multi-sense cross-verification based on the over-limit situation data to obtain over-limit judgment data; In one embodiment, logical cross-verification is performed on different features based on the out-of-limit situation data. When , When the current sensing voltage continuously exceeds the limit but the drain-source voltage does not rise rapidly, the event is marked as a gradual overcurrent; when , When a short circuit occurs, the event is marked as a sudden short circuit; when only isolated glitches exist, the event is marked as a transient interference. Additionally, if the gate voltage is a falling edge and the following conditions are met: This confirms that the corresponding overcurrent characteristic has triggered a gate turn-off action. The time limit is 2μs; otherwise, it is marked as pending verification. The overcurrent type, valid over-limit channel, gate follower state, and determination time are encapsulated into over-limit judgment data.

[0060] 45. Based on the over-limit judgment data, the state is merged to obtain the first overcurrent state data.

[0061] In one embodiment, the data for exceeding the limit is arranged according to the start time of the event. For states of the same type and adjacent time intervals not exceeding [a certain value], [the data is then processed]. Events are grouped into a single overcurrent event. The start time of the merged event is the earliest time of each event, the end time is the latest time, and the duration is the difference between the two. For events with overlapping time intervals but different state types, the state with the higher protection level is retained in the order of "sudden short circuit takes precedence over gradual overcurrent, and gradual overcurrent takes precedence over transient interference". The event number, state type, occurrence time, duration, effective over-limit channel, and gate follower state are encapsulated into the first overcurrent state data, which is consistent with the input of the aforementioned dual-state mutual verification fusion step.

[0062] Optionally, the feature domains include: Channel splitting is performed based on transient characteristic data to obtain branch characteristic data; In one embodiment, the transient event corresponding to the r-th transition window is used. As a processing unit, it reads the transition amplitudes of the three signals from the transient characteristic data. Peak slope Edge direction Pulse duration and cross-road time difference, among which, Construct them separately according to the signal source: , , Save the above three sets of data according to the event number r to obtain the branch characteristic data.

[0063] The path feature data is dimensionally categorized to obtain domain candidate data; In one embodiment, dimensional classification is performed based on the physical meaning of the branch feature data. Substitute into the amplitude feature domain, and Incorporate into the edge feature domain, Submitting to the persistent feature domain, the three sets of cross-road time differences are subsumed into the time difference feature domain and expressed as follows: , , , During the classification process, event numbers and channel identifiers are retained to obtain candidate data for different domains, thus avoiding the mixing of the same-dimensional features from different events or channels.

[0064] Based on the candidate data of each domain, a baseline normalization is performed to obtain the standard domain data; In one embodiment, dimensionless processing is performed based on existing judgment criteria corresponding to each feature. The current sensing voltage amplitude, current sensing channel slope, and drain-source voltage slope are normalized according to the following formula: , , The transition amplitudes of the drain-source voltage and gate voltage are normalized using the transition thresholds of their respective channels. The gate edge slope, duration, and cross-path time difference are expressed as follows: , , Edge direction The original values ​​are preserved without numerical normalization. The original feature values, normalized feature values, event numbers, and channel identifiers are encapsulated together to obtain standard domain data.

[0065] Threshold regions are divided based on standard domain data to obtain over-limit feature data.

[0066] In one embodiment, threshold boundaries for each dimension are established based on a normalized reference value 1. For the amplitude and slope domains, [0,1) is set as the reference domain, and... Set as a candidate bounded domain; for a persistent domain, set [0,1) as a short-time domain, and set as... Set as a persistent domain; for time difference domains, set [-1,1] as the associated time difference domain and the rest as the non-associated time difference domain. Edge direction follows... and Separately define the ascending and descending directions. Encapsulate the standard domain data for each event along with the aforementioned threshold boundaries as follows: ,in, Let be the set of threshold regions for each dimension corresponding to the r-th event, from which the over-limit feature data is obtained.

[0067] Optionally, the amplitude-time joint judgment includes: Amplitude deviation quantization is performed based on threshold correlation data to obtain amplitude exceeding the limit data; In one embodiment, the transient event corresponding to the r-th transition window is used. The processing unit reads the current sensing voltage transition amplitude from the threshold correlation data. Peak slope of the flow detection channel Drain-source voltage peak slope And their corresponding thresholds. Calculate them separately: , , ,in, , and These represent the over-threshold deviations of the current sensing amplitude, current sensing slope, and drain-source voltage slope, respectively; a deviation of 0 indicates that the corresponding feature has not exceeded the limit. The event number, feature type, original feature value, corresponding threshold, and deviation are encapsulated into amplitude over-limit data.

[0068] By continuously tracking the amplitude exceeding the limit data, the continuous exceeding limit data can be obtained. In one embodiment, the corresponding transition window is returned based on the amplitude exceeding the limit data. The limit-crossing state is tracked point by point in the purification transition data and edge slope sequence. The current detection amplitude limit-crossing condition is: The over-limit conditions for current detection slope and drain-source voltage slope are as follows: , The moment when the corresponding condition is first met is recorded as the start time of the over-limit. The last moment before the end of the continuous over-limit is recorded as the over-limit termination time. ,in, Interruptions should not exceed one sampling period. Adjacent out-of-bounds intervals are merged into a single interval, and the following calculations are performed: The start and end times of the out-of-limit event, the duration, and the corresponding deviation are encapsulated as out-of-limit continuous data.

[0069] Amplitude-time pairing is performed on the out-of-limit sustained data to obtain amplitude-time coupled data; In one embodiment, the deviation amount is paired one-to-one with the corresponding over-limit duration interval according to the event number r and the feature type q to form an amplitude-time unit: When the same feature has multiple discontinuous over-limit intervals within a transition window, amplitude-time cells are generated separately, and the durations of different intervals are not accumulated; when the over-limit intervals of current detection amplitude, current detection slope, or drain-source voltage slope overlap in time, or the difference between the corresponding start times is not greater than At that time, the same linkage identifier is set for the relevant amplitude-time units to obtain amplitude-time coupling data.

[0070] By removing glitch data from the amplitude-time coupling data, the out-of-limit situation data is obtained.

[0071] In one embodiment, the validity of each amplitude-time unit is determined based on amplitude-time coupling data. When When, the corresponding amplitude-time unit is retained as a continuous over-limit unit; when And in If there are no associated out-of-limit units in other channels, it is identified as an isolated spurious ... However, when it forms a linkage identifier with a continuous over-limit unit in another channel, it is retained as an associated transient unit. The effective feature type, deviation amount, over-limit start and end time, duration, linkage channel, and glitch removal results are encapsulated according to the event number to obtain the over-limit situation data.

[0072] Optionally, edge situation inference includes: 46. ​​Perform edge-sequence segmentation on transient feature data to obtain edge sequence data; In one embodiment, the transient event corresponding to the r-th transition window is used. For the processing unit, the edge start and end times, transition amplitude, peak slope, and duration corresponding to the current sensing voltage, drain-source voltage, and gate voltage are extracted from the transient feature data to construct the edge unit: .according to The edge cells of the same channel are sorted in ascending order. When the time interval between adjacent edge cells is no greater than [a certain value], the sorting is performed accordingly. When the jump directions are the same, they are programmed into the same edge trajectory; when the time interval is greater than If the jump direction changes, the current trajectory ends and a new trajectory is established. The three edge trajectories, event numbers, and time sequence are associated to obtain edge sequence data.

[0073] 47. Perform slope orientation on the edge sequence data to obtain edge direction data; In one embodiment, the edge direction is determined based on the transition amplitude of each edge cell in the edge sequence data: ,in, Indicates the rising edge. This indicates the falling edge. The rising edge of the current sensing voltage characterizes an increase in load current, the rising edge of the drain-source voltage characterizes an increase in the voltage across the power switch, and the falling edge of the gate voltage characterizes the gate turn-off action. If adjacent edges within the same trace have different directions, the trace is re-splitted based on the moment of direction change. The edge direction, peak slope, edge start and end times, and trace number are encapsulated as edge direction data.

[0074] 48. Perform cross-path time difference tracking based on edge direction data to obtain edge time difference data; In one embodiment, based on edge direction data, using the same event number r as an index, cross-channel pairing is performed for the current detection voltage rising edge, drain-source voltage rising edge, and gate voltage falling edge, and the following calculations are performed: , , When the absolute value of any time difference is not greater than When the absolute value is greater than 1, the corresponding edge is identified as the associated edge; when the absolute value is greater than 1, the corresponding edge is identified as the associated edge. When a valid edge is not found in the corresponding channel, the corresponding time difference is marked as missing. The time difference, association status, and edge direction of each cross-channel are encapsulated according to the event number to obtain the edge time difference data.

[0075] 49. Based on the edge time difference data, the situation is recursively extrapolated to obtain the edge situation data; In one embodiment, an event situation state is established based on edge time difference data. When the following conditions are met: , , and , At that time, This is recorded as a proactive approach to flow detection; when ,and At that time, This is denoted as the drain-source first-emergence condition; when the current sensing voltage or drain-source voltage shows a rising edge, but... If no gate voltage falling edge occurs, This is denoted as a shutdown hysteresis state; when there is only a single channel edge and its duration is less than At that time, This is denoted as an isolated pulse state. For states of the same type and adjacent event intervals not exceeding [a certain value], [the following applies]. For continuous events, the original state is maintained and its duration is extended to the time when the current event ends; when the state type changes, the original situation is terminated and a new situation is established, thereby obtaining edge situation data.

[0076] 4a. Classify the operating conditions based on the edge situation data to obtain the second overcurrent state data.

[0077] In one embodiment, operating condition mapping is performed based on edge condition data: current detection-preemptive conditions are classified as gradual overcurrent, drain-source-preemptive conditions are classified as sudden short circuit, turn-off hysteresis conditions are classified as gate turn-off hysteresis, and conditions with durations less than [missing information] are classified as [missing information]. Isolated pulse states that do not form cross-path associations are classified as transient interference; other events that do not meet the overcurrent link conditions are classified as normal fluctuations. For continuous similar operating conditions, the earliest edge start time is taken as the operating condition occurrence time, and the last edge termination time is taken as the operating condition termination time, with the difference between the two determining the operating condition duration. The event number, operating condition type, occurrence time, duration, edge triggering order, and cross-path association status are encapsulated into second overcurrent state data.

[0078] Optionally, dual-state mutual verification fusion includes: 4b. Time-stamp alignment is performed based on the first overcurrent state data and the second overcurrent state data to obtain the corresponding data for the two states; In one embodiment, the state type in the first overcurrent state data is read. When it happens Termination time and duration And read the status type from the second overcurrent status data. When it happens Termination time and duration Calculate the overlap time between the two state intervals: .when ,or When two states are identified as the same anomalous event, if multiple candidate states exist, the one with the longest overlap time is selected first; if there is no overlap, the one with the smallest difference in occurrence time is selected. The corresponding states, time intervals, and original judgment criteria are associated to obtain the corresponding data for the two states.

[0079] 4c. Perform state verification on the corresponding data of the two states to obtain the two-state verification data; In one embodiment, a consistency check is performed on the two types of state results in the bi-state correspondence data. When When the two states are of different types, but the first overcurrent state is a gradual overcurrent or a sudden short circuit, and the second overcurrent state is a gate turn-off hysteresis, they are marked as supplementary states to characterize the simultaneous existence of overcurrent conditions and their turn-off anomalies; other state combinations are marked as divergent states. Set verification flag: The verification identifier, two types of states, and corresponding evidence are encapsulated into bi-state verification data.

[0080] 4d. Based on the dual-state verification data, perform bifurcation tracing to obtain state bifurcation data; In one embodiment, only for The dual-state verification data is used for bifurcation analysis. If the first overcurrent state is a gradual overcurrent and the second overcurrent state is a transient disturbance, then check whether the duration of the current sensing voltage reaches the specified value. And whether there is cross-path association; to achieve However, if a complete edge chain is not formed, the cause of the divergence is determined to be a missing edge chain. If the first overcurrent state is a transient disturbance, and the second overcurrent state is a sudden short circuit, and the drain-source voltage peak slope satisfies... If the first state is determined to be overcurrent, while the second state is... If no falling edge of the gate is detected, the cause of the divergence is determined to be a missing gate action. The divergence type, corresponding characteristic value, and original state result are encapsulated into state divergence data.

[0081] 4e. Perform protection priority determination on the state divergence data to obtain protection priority data; In one embodiment, priority is determined according to deterministic protection rules based on state divergence data. Sudden short circuit corresponds to the highest protection priority P=3; gate turn-off hysteresis accompanied by continuous overcurrent corresponds to P=3; gradual overcurrent corresponds to P=2; transient interference corresponds to P=1; normal fluctuation corresponds to P=0. When two state types have different priorities, the one with the higher priority is retained; when the priorities are the same, priority is reserved for periods up to [duration missing]. Furthermore, it has a cross-path associated state; if neither of these conditions is met, it is determined to be transient interference. The target state, protection priority, and determination criteria are encapsulated into protection priority data.

[0082] 4f. Perform dual-state merging on the protection priority data to obtain the overcurrent status data.

[0083] In one embodiment, the consistency state, supplementary state, and priority determination state in the protection priority data are merged. The occurrence time after merging is taken as: The termination time is taken as: The duration is: For supplementary states, the overcurrent type is retained as the main state, and gate turn-off hysteresis is treated as an additional state; for states identified only by a single path but with a duration reaching [a certain threshold], [the following applies]. Events that form a complete chain across road edges are still retained as valid overcurrent events. The event number r and overcurrent type are specified. When it happens Duration Protection priority The additional state is encapsulated as overcurrent state data.

[0084] Optionally, phased regulation includes: 51. Perform stage mapping based on the overcurrent state data to obtain the gate stage data; In one embodiment, according to the type of overcurrent And establish the gate control stage mapping relationship with the additional states. When In the case of transient interference, the current event is mapped to the normal driving phase, keeping the gate drive acceleration branch off; when For gradual overcurrent, it is sequentially mapped to a pre-discharge stage, a rapid pull-down stage, and a turn-off holding stage; when In the event of a sudden short circuit, the pre-discharge phase is skipped, and the circuit is directly mapped to a fast pull-down phase and a turn-off holding phase. When the additional state is gate turn-off hysteresis, an enhanced pull-down phase is added after the original control phase. The event number r, phase type, phase sequence, and occurrence time are specified. Duration and protection priority Encapsulated as gate stage data.

[0085] 52. Pull down the gate stage data to obtain the gate drive data; In one embodiment, the conduction state and discharge duration of the gate drive acceleration branch are configured based on gate stage data. During the normal drive stage, the following is specified: During the pre-discharge phase, the current limiting controller releases part of the gate charge through its own discharge path; the rapid pull-down and enhanced pull-down phases... Control the active pull-down transistor Turn on. The gate discharge current is: ,in, This refers to the gate-source voltage of the power switch. for The on-state voltage drop. The pre-discharge duration corresponds to the gradual overcurrent. Sudden short circuit corresponds to rapid pull-down duration ,and If gate turn-off hysteresis exists, maintain active pull-down until subsequent feedback correction makes an exit decision. (The stage type, ...) , , or Packaged as gate drive data.

[0086] 53. Perform timing arrangement based on the gate drive data to obtain the gate timing data; In one embodiment, the start time, end time, and switching relationship of each stage are determined based on gate drive data. The pre-discharge stage of the gradual overcurrent event starts from... The start and end times are: The rapid pull-down phase begins immediately after the pre-discharge phase; the rapid pull-down phase for sudden short-circuit events starts directly from... Start-up. The shutdown hold time is determined according to the following formula: When multiple events overlap in time, protection priority is applied. Arranged from high to low priority; after a high-priority event is initiated, the normal drive or pre-discharge instructions of low-priority events are aborted, and the system switches to the corresponding fast pull-down or enhanced pull-down stage. The start time, duration, switching order, and pull-down enable state of each stage are encapsulated as gate timing data.

[0087] 54. Instructions are integrated based on the gate timing data to obtain gate control data.

[0088] In one embodiment, the stage type, start-up time, pull-down enable energy, discharge duration, and turn-off hold time in the gate timing data are integrated according to the event number r as follows: ,in, Indicates selection based on event type. or .when When, output the active pull-down conduction command; when When the time comes, output the drop-down branch cutoff command. Set the corresponding events... Arranged according to the startup time, the gate control data is obtained and used as the direct input for subsequent drive retrieval and feedback correction.

[0089] Optionally, feedback correction includes: 61. Drive and sample back the gate control data to obtain the gate action data; In one embodiment, based on gate control data The control gate drive acceleration branch executes the corresponding pre-discharge, fast pull-down, or turn-off holding actions, based on the event occurrence time. Starting from the preset feedback window The three voltage channels are re-acquired according to the sampling period Δt. The actual load current is calculated based on the current sensing voltage. Extract the peak current within the feedback window: The first time it will be satisfied The sampling time is determined as the gate turn-off time: ,in, This is a preset gate voltage threshold; if the threshold is not reached within the feedback window, then... Calculate the gate operation delay: The original gate control parameters, three-channel retrieval sequence, , and Encapsulated as gate action data.

[0090] 62. Perform deviation back calculation based on gate operation data to obtain current limiting deviation data; In one embodiment, the actual peak current is determined based on gate operation data. With target current limit Compare and calculate the current limiting deviation: Read the load current at the end of the feedback window. and gate voltage ,in, The last sampling number in the feedback window; and the drain-source voltage fluctuation is calculated within a consecutive M=5 sampling points at the end: The event number r, , , , and It is encapsulated as current limiting deviation data to determine whether the current current limiting action is sufficient.

[0091] 63. Based on the current limiting deviation data, a maintenance judgment is made to obtain feedback adjustment data; In one embodiment, a graded judgment is made based on the current limiting deviation data. When the following conditions are met: If the actual peak current exceeds the allowable deviation or the gate turn-off action is delayed, the adjustment state is set to enhanced pull-down, and the following settings are configured: , , ,in, To allow for current limiting deviation, To the maximum allowable gate operation delay, This is the extension amount for a single pull-down operation. When both of the following conditions are met: , , When the load current has decreased, the power switch has been turned off, and the drain-source voltage has stabilized, the adjustment state is set to exit pull-down, and the following is set: Stop continuing the shutdown hold. In all other cases, set the adjustment state to hold down, and let... , and keep The adjustment status and correction parameters are encapsulated as feedback adjustment data.

[0092] 64. Correct the feedback adjustment data with instructions to obtain the current limiting control data.

[0093] In one embodiment, the current gate control command is corrected based on feedback adjustment data. When the adjustment state is enhanced pull-down, the corrected command is executed. Extended active pull-down transistor The conduction time; when the adjustment state is to maintain the pull-down, the original pull-down strength is maintained and the off-hold is continued; when the adjustment state is to exit the pull-down, the control... Cut off and release the shutdown hold. Encapsulate the feedback result as: ,in, To enhance, maintain, or exit the dropdown adjustment state, the corresponding events are... The current limiting control data is obtained by arranging the events according to their numbers and times of occurrence. This data is used for the continuous control of the current overcurrent event and serves as the basis for calling the gate control parameters in the next feedback cycle.

[0094] Optionally, this application also provides a precision current limiting control system based on dual-state mutual verification, used to execute the precision current limiting control method based on dual-state mutual verification as described above. The precision current limiting control system based on dual-state mutual verification includes: The detection and acquisition module is used to acquire initial detection data, which includes the current sensing voltage data across the current sensing resistor, the drain-source voltage data of the power switch, and the gate voltage data. The synchronization and association module is used to perform synchronization and association based on the initial detection data to obtain the rate limiting detection data. The temporal transition deconstruction module is used to perform temporal transition deconstruction on the current limiting detection data to obtain transient feature data. The dual-state judgment module is used to perform multi-signal limit judgment and edge situation inference based on transient feature data to obtain first overcurrent state data and second overcurrent state data; and to perform dual-state mutual verification fusion based on the first overcurrent state data and the second overcurrent state data to obtain overcurrent state data. The stage control module is used to perform stage control based on the overcurrent state data to obtain gate control data; The feedback correction module is used to perform feedback correction on the gate control data to obtain current limiting control data.

[0095] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended application documents rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the application documents be incorporated into the invention.

[0096] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A precise current limiting control method based on dual-state mutual verification, characterized in that, Includes the following steps: Acquire initial detection data, which includes current sensing voltage data across the current sensing resistor, drain-source voltage data of the power switch, and gate voltage data. Synchronous correlation is performed based on the initial detection data to obtain the rate limiting detection data; Temporal transition deconstruction is performed on the current limiting detection data to obtain transient feature data; Based on transient feature data, multi-signal limit judgment and edge situation inference are performed to obtain first overcurrent state data and second overcurrent state data; based on the first overcurrent state data and second overcurrent state data, dual-state mutual verification fusion is performed to obtain overcurrent state data; Gate control data is obtained by performing phased adjustments based on overcurrent state data. Feedback correction is performed on the gate control data to obtain current limiting control data.

2. The precision current limiting control method based on dual-state mutual verification according to claim 1, characterized in that, The deconstruction of time-domain transitions includes: Multi-channel timing alignment is performed based on current limiting detection data to obtain transition window data; baseline drift stripping is performed on the transition window data to obtain purified transition data; edge slope analysis is performed on the purified transition data to obtain edge evolution data; cross-channel time difference mapping is performed based on the edge evolution data to obtain transition correlation data; and pulse continuity extraction is performed based on the transition correlation data to obtain transient feature data.

3. The precision current limiting control method based on dual-state mutual verification according to claim 1, characterized in that, Multi-signature exceeding limit judgment includes: Feature domain segmentation is performed on transient feature data to obtain over-limit feature data; threshold domain mapping is performed on over-limit feature data to obtain threshold domain associated data; amplitude-time joint judgment is performed on threshold domain associated data to obtain over-limit situation data; multi-signal mutual verification is performed on over-limit situation data to obtain over-limit judgment data; state merging is performed on over-limit judgment data to obtain first overcurrent state data.

4. The precision current limiting control method based on dual-state mutual verification according to claim 3, characterized in that, Feature domains include: Channel splitting is performed based on transient feature data to obtain path feature data; the path feature data is categorized by dimension to obtain domain candidate data; benchmark normalization is performed based on the domain candidate data to obtain standard domain data; threshold division is performed based on the standard domain data to obtain limit-crossing feature data.

5. The precision current limiting control method based on dual-state mutual verification according to claim 3, characterized in that, The simultaneous judgment includes: Amplitude deviation quantization is performed on threshold correlation data to obtain amplitude over-limit data; continuous interval tracking is performed on amplitude over-limit data to obtain over-limit continuous data; amplitude-time pairing is performed on over-limit continuous data to obtain amplitude-time coupled data; spurs are removed from amplitude-time coupled data to obtain over-limit situation data.

6. The precision current limiting control method based on dual-state mutual verification according to claim 1, characterized in that, Marginal situation inference includes: The transient characteristic data is segmented at the edge to obtain edge sequence data; the edge sequence data is oriented by slope to obtain edge direction data; cross-path time difference tracking is performed based on the edge direction data to obtain edge time difference data; the situation is recursively extrapolated based on the edge time difference data to obtain edge situation data; and the operating conditions are classified based on the edge situation data to obtain the second overcurrent state data.

7. The precision current limiting control method based on dual-state mutual verification according to claim 1, characterized in that, Dual-state mutual verification fusion includes: The first and second overcurrent state data are time-stamped to obtain dual-state corresponding data; the dual-state corresponding data is verified to obtain dual-state verification data; the dual-state verification data is used to trace the cause of divergence to obtain state divergence data; the state divergence data is used to determine the protection priority of the state divergence data to obtain protection priority data; and the protection priority data is merged into dual states to obtain overcurrent state data.

8. The precision current limiting control method based on dual-state mutual verification according to claim 1, characterized in that, Phased regulation includes: The gate stage data is obtained by performing stage mapping based on the overcurrent state data; the gate stage data is then pulled down and strengthened to obtain the gate drive data; the gate drive data is then timed to obtain the gate timing data; and the gate timing data is then integrated with the instructions to obtain the gate control data.

9. The precision current limiting control method based on dual-state mutual verification according to claim 1, characterized in that, Feedback correction includes: Drive data is sampled back from the gate control data to obtain gate action data; deviation is calculated from the gate action data to obtain current limiting deviation data; hold data is determined from the current limiting deviation data to obtain feedback adjustment data; and command correction is applied to the feedback adjustment data to obtain current limiting control data.

10. A precision current limiting control system based on dual-state mutual verification, characterized in that, For executing the precision current limiting control method based on dual-state mutual verification as described in claim 1, the precision current limiting control system based on dual-state mutual verification includes: The detection and acquisition module is used to acquire initial detection data, which includes the current sensing voltage data across the current sensing resistor, the drain-source voltage data of the power switch, and the gate voltage data. The synchronization and association module is used to perform synchronization and association based on the initial detection data to obtain the rate limiting detection data. The temporal transition deconstruction module is used to perform temporal transition deconstruction on the current limiting detection data to obtain transient feature data. The dual-state judgment module is used to perform multi-signal limit judgment and edge situation inference based on transient feature data to obtain first overcurrent state data and second overcurrent state data; and to perform dual-state mutual verification fusion based on the first overcurrent state data and the second overcurrent state data to obtain overcurrent state data. The stage control module is used to perform stage control based on the overcurrent state data to obtain gate control data; The feedback correction module is used to perform feedback correction on the gate control data to obtain current limiting control data.

Images