Switching power supply IC integrated fault protection method

Abstract

This application relates to the field of power management integrated circuit technology and discloses a method for integrated fault protection of switching power supply ICs. The method includes: in calibration mode, acquiring the switching node waveforms of the switching power supply at multiple load states, extracting a health fingerprint vector containing parameters such as the inductor current rise slope and ringing frequency, and establishing a dynamic mapping table between the load and the fingerprint; in normal operation mode, monitoring the current load state in real time to call the corresponding reference fingerprint vector, extracting the real-time fingerprint vector, and comparing the two to generate a deviation vector; finally, comparing the deviation vector with preset warning and fault thresholds, and executing graded fault protection actions such as non-interruptible warnings or interruptible protection. This invention solves the problem that traditional fixed threshold methods cannot adapt to changes in operating conditions and have a delayed response. By constructing a dynamic health benchmark, it achieves predictive diagnosis of component performance degradation, improving the adaptability and accuracy of protection.

Description

Technical Field

[0001] This invention relates to the field of power management integrated circuit technology, specifically to a fault protection method for switching power supply ICs. Background Technology

[0002] As an indispensable power conversion unit in modern electronic devices, the reliability and safety of switching power supplies are of paramount importance. Therefore, integrating comprehensive fault protection functions into power management integrated circuits is a standard design requirement.

[0003] Currently, the protection technologies widely used in switching power supply ICs are mainly based on comparing the real-time values ​​of key electrical parameters (such as peak inductor current and output voltage) with one or more fixed preset thresholds. When the monitored parameter value exceeds the preset threshold, the IC's internal control logic will trigger corresponding protection actions, such as cycle-by-cycle current limiting, hiccup mode restart, or latch-up shutdown, to prevent permanent damage to the power supply itself or the downstream load.

[0004] However, this fixed threshold-based protection mechanism is essentially a reactive protection, which only responds when the fault has developed to the point where macroscopic parameters such as voltage or current significantly exceed limits. For progressive "soft" faults caused by the gradual degradation of power components (such as power inductors and output capacitors), this mechanism lacks effective early warning capabilities.

[0005] Furthermore, the operating characteristics of switching power supplies, especially the voltage and current waveforms of their switching nodes, change significantly with variations in load conditions. Using fixed protection thresholds cannot adapt to these dynamic changes, often leading to overly sensitive protection actions and false triggering under certain operating conditions, or excessive threshold margins reducing protection sensitivity under other conditions, making it difficult to achieve optimal protection across the entire operating range.

[0006] More importantly, traditional overvoltage or overcurrent protection methods struggle to directly monitor the health status of critical passive components in the power stage. Parameter drift in these components (e.g., decreased inductance due to high temperatures, increased equivalent series resistance in capacitors due to aging) is a common cause of eventual power system failure. These early changes often manifest as subtle alterations in the dynamic waveform characteristics of switching nodes, which traditional protection methods are insensitive to. Therefore, there is an urgent need in the field for a novel protection scheme capable of fault prediction and adapting to different operating conditions. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides an integrated fault protection method for switching power supply ICs. This method solves the problem that existing fault protection methods for switching power supply ICs, which use fixed thresholds, are unable to perform early prediction and adaptive protection against progressive faults such as component performance degradation under different load conditions.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for integrated fault protection of a switching power supply IC, comprising:

[0009] The first aspect of the present invention provides a method for integrated fault protection of a switching power supply IC, which is executed inside the IC.

[0010] The method includes operation within a preset calibration mode. This operation acquires the switching node waveforms of the switching power supply at at least one load state point and extracts one or more health fingerprint vectors from the switching node waveforms. Each health fingerprint vector characterizes the healthy operation characteristics of the switching power supply at the corresponding load state point.

[0011] The method further includes operation in normal operating mode. This operation monitors the current load state of the switching power supply in real time and extracts the real-time fingerprint vector of the switching node. Then, based on the monitored current load state, a reference fingerprint vector for current comparison is determined from one or more pre-stored health fingerprint vectors. Next, the real-time fingerprint vector is compared with the reference fingerprint vector to generate a deviation vector. Finally, based on this deviation vector, a preset fault protection action is executed on the switching power supply.

[0012] As a specific implementation, both the health fingerprint vector and the real-time fingerprint vector can include the following characteristic parameters: inductor current rise slope, switching node ringing frequency, and switching node ringing attenuation ratio.

[0013] Wherein, the inductor current rise slope ( The calculation of ) can follow the following formula:

[0014] ;

[0015] In this formula, Indicates the slope of the inductor current rise; and These are two different moments during the main switch conduction period. and The inductor current value collected below.

[0016] Among them, the ringing attenuation ratio of the switching node ( The calculation of ) can follow the following formula:

[0017] ;

[0018] In this formula, Indicates the ringing attenuation ratio of the switching node; and These are the first and second positive peak voltages that appear consecutively on the waveform at the switching node during the dead time after the main switch is turned off.

[0019] As a preferred implementation, the operation in the preset calibration mode can be further refined as follows: First, the driving switching power supply is stably operated sequentially at multiple different load state points; second, at each load state point, a corresponding health fingerprint vector is extracted; third, each load state point is associated with its corresponding health fingerprint vector to establish and store a dynamic fingerprint mapping table. Accordingly, the step of determining the reference fingerprint vector specifically involves: based on the current load state, querying and retrieving the corresponding health fingerprint vector from the dynamic fingerprint mapping table as the reference fingerprint vector.

[0020] The operation of real-time monitoring of the current load status can be achieved by monitoring the average output current or the PWM duty cycle.

[0021] The step of executing a preset fault protection action based on the deviation vector may include: comparing the deviation vector with a preset warning threshold and a fault threshold; if the deviation vector exceeds the warning threshold but does not exceed the fault threshold, then executing a non-interruptible warning action, such as reporting a status flag bit to the system main controller through a communication interface, or recording it in an internal fault log register; if the deviation vector exceeds the fault threshold, then executing an interruptible protection action.

[0022] The interruption protection action may include performing adaptive intervention, or putting the switching power supply into hiccup mode or latch-off mode. The adaptive intervention may include temporarily lowering the cycle-by-cycle current limit point, limiting the maximum duty cycle, or reducing the switching frequency.

[0023] A second aspect of the present invention provides an integrated fault protection system for a switching power supply IC, comprising:

[0024] The fingerprint extraction module is configured to extract a healthy fingerprint vector from the waveform of the switch node in a preset calibration mode, and to extract the real-time fingerprint vector of the switch node in real time in normal operation mode.

[0025] A load monitoring module is configured to monitor the current load status of the switching power supply in real time during normal operation.

[0026] The storage and query module is configured to store one or more of the health fingerprint vectors and to determine the reference fingerprint vector for the current comparison based on the current load status output by the load monitoring module.

[0027] A deviation analysis module, which is connected to the fingerprint extraction module and the storage and query module, is configured to compare the real-time fingerprint vector with the reference fingerprint vector to generate a deviation vector;

[0028] The protection execution module, which is connected to the deviation analysis module, is configured to perform a preset fault protection action on the switching power supply based on the deviation vector.

[0029] This invention provides a method for integrated fault protection of switching power supply ICs. It has the following beneficial effects:

[0030] 1. This invention extracts the real-time fingerprint vector of the switching node in normal operating mode and compares it with a reference fingerprint vector representing the health status. This allows for the quantification of subtle waveform feature changes caused by component performance degradation, such as decreased inductance and increased capacitor ESR. This state deviation-based analysis method enables the identification of early anomalies in the system's operating state before serious faults such as inductor saturation or component failure occur, thus achieving predictive fault diagnosis.

[0031] 2. This invention solves the technical problem that traditional fixed threshold protection methods cannot adapt to changes in operating conditions by establishing a dynamic fingerprint mapping table covering multiple load status points in calibration mode and dynamically calling the matching health fingerprint vector as a comparison benchmark based on the real-time monitored current load status in normal operating mode. The introduction of this adaptive benchmark ensures that comparisons are always performed under similar operating conditions, effectively distinguishing normal waveform changes from actual fault precursors, avoiding misjudgments caused by load fluctuations or missed judgments due to insufficient sensitivity under specific operating conditions, and improving the accuracy of protection.

[0032] 3. This invention generates a deviation vector containing multi-dimensional information such as inductor current rise slope deviation and ringing frequency deviation, and performs hierarchical threshold judgment based on this vector. This not only determines whether a fault exists but also provides a basis for preliminary fault type localization. Furthermore, by setting two-level thresholds for warning and fault, and correspondingly executing non-interrupted warning actions and interrupted protection actions, the protection response becomes more flexible and refined, achieving an improvement from simple shutdown protection to multi-level health management. Attached Figure Description

[0033] Figure 1 This is a flowchart of the method of the present invention;

[0034] Figure 2 This is a system architecture diagram of the present invention. Detailed Implementation

[0035] 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.

[0036] Example:

[0037] Please see the appendix Figure 1 This invention provides a method for integrated fault protection of a switching power supply IC, comprising the following steps:

[0038] S1: In the preset calibration mode, the switching node waveform of the switching power supply at at least one load state point is acquired, and the health fingerprint vector is extracted from the switching node waveform. The health fingerprint vector characterizes the working characteristics of the switching power supply in a healthy state.

[0039] In this embodiment, under a preset calibration mode, the switching node waveform of the switching power supply at at least one load state point is acquired, and a health fingerprint vector is extracted from the switching node waveform. The health fingerprint vector characterizes the operating characteristics of the switching power supply in a healthy state. The specific implementation method may include the following steps:

[0040] The preset calibration mode is a special operating state triggered by the internal control logic of the switching power supply IC under specific conditions. Preferably, this mode can be activated upon the IC's first power-on, system reset, or upon receiving a specific calibration command from an external main control unit. This mode is designed to establish an accurate and reliable health baseline for subsequent online real-time monitoring.

[0041] Once in calibration mode, to ensure the accuracy and adaptability of subsequent online monitoring, the IC's internal control logic drives the switching power supply system to operate stably at multiple preset, representative load state points in sequence. For example, these load state points may include light load, medium load, and heavy load states, thereby covering the main operating range of the power supply.

[0042] At each stable load point, an integrated waveform analysis engine within the IC will acquire waveforms from the switching nodes of the switching power supply for one or more switching cycles. The switching nodes are chosen as the information acquisition source because the voltage and current waveforms of these nodes contain rich transient characteristic information determined by power stage components such as the main inductor, main switching transistor, and related capacitors, and are extremely sensitive to changes in the health status of these components.

[0043] After acquiring the waveform at the switching node, the waveform analysis engine extracts and calculates a multi-dimensional health fingerprint vector. This vector is a digital "snapshot" representing the healthy operating characteristics of the power supply at that load point.

[0044] Specifically, the first key feature parameter of the health fingerprint vector is the inductor current rise slope. This parameter directly reflects the rate of change of current flowing through the main inductor during the conduction of the main switch. According to the working principle of the switching power supply, this slope is directly related to the input voltage, output voltage, and the inductance value of the main inductor. Therefore, by accurately quantifying this slope, the electrical characteristics of the main inductor can be effectively characterized, providing a crucial basis for subsequent monitoring of potential faults such as inductor saturation. Inductor current rise slope ( The calculation of ) can follow the following formula:

[0045] ;

[0046] in, Indicates the slope of the inductor current rise; and These are two different moments during the main switch conduction period. and The inductor current value is collected by the current sampling circuit inside the IC.

[0047] The second key feature parameter of the health fingerprint vector is the switching node ringing frequency. This parameter is generated during the dead time after the main switch is turned off, when the main inductor and the equivalent capacitance of the switching node (mainly composed of the output capacitance of the main switch) form an LC resonant circuit, producing ringing. The natural frequency of this ringing is mainly determined by the inductance and equivalent capacitance values. Therefore, monitoring changes in this frequency can simultaneously reflect the changing trends of the main inductance and the parasitic parameters of the main switch. Switching node ringing frequency ( The time taken by the fixed number of ringing waveform cycles can be measured within the dead time using a high-speed counter inside the IC.

[0048] To further enhance the dimensionality and accuracy of the diagnosis, the health fingerprint vector may preferably include a third key feature parameter: the switching node ringing attenuation ratio. This parameter quantifies the damping characteristics of the aforementioned LC resonance process, i.e., the rate of attenuation of the ringing waveform amplitude. This attenuation rate is closely related to energy dissipation factors such as the equivalent series resistance in the resonant circuit, and the equivalent series resistance (ESR) of the output capacitor is a crucial component. Therefore, changes in this parameter can effectively indicate the aging state of components such as the output capacitor. In one specific implementation, this parameter can be obtained by calculating the ratio of two consecutive positive peak voltages of the switching node ringing waveform, and its calculation follows the formula:

[0049] ;

[0050] in,

[0051] Indicates the ringing attenuation ratio of the switching node;

[0052] and These are the first and second positive peak voltages that appear consecutively on the switching node waveform, captured by the peak voltage detection circuit inside the IC during the dead time after the main switch is turned off.

[0053] Through the above steps, a corresponding health fingerprint vector is generated for each traversed load state point, containing information in multiple dimensions such as the inductor current rise slope, switching node ringing frequency, and switching node ringing attenuation ratio. After fingerprint extraction for all preset load points is completed, the calibration mode ends. These collected and stored health fingerprint vectors collectively constitute a comprehensive health benchmark database correlated with load states, providing a solid foundation for subsequent high-precision, adaptive fault warning and diagnosis under normal operating conditions.

[0054] S2: In normal operating mode, monitor the current load status of the switching power supply in real time and extract the real-time fingerprint vector of the switching node.

[0055] In this embodiment, under normal operating mode, the current load state of the switching power supply is monitored in real time, and the real-time fingerprint vector of the switching node is extracted in real time. The specific implementation method may include the following steps:

[0056] After the switching power supply IC completes the preset calibration mode and enters the normal operating mode for powering the load, its internal monitoring and analysis system will switch to continuous, periodic online operation. The core task at this stage is to perform two key operations in parallel: first, to identify the current load condition of the system in real time; and second, to capture waveform fingerprints that reflect the current health level of the system in real time under this condition.

[0057] Firstly, the purpose of real-time monitoring of the current load status of the switching power supply is to provide an accurate contextual environment for subsequent fingerprint comparison. Since the various switching waveform parameters of the switching power supply naturally shift with changes in load, comparing them using a single health benchmark without distinguishing load status will not accurately differentiate between normal operating condition changes and abnormal fault precursors.

[0058] To achieve this, a load monitoring module within the switching power supply IC continuously monitors one or more electrical parameters that macroscopically characterize the load size. Preferably, this parameter can be the average duty cycle of the PWM control signal or the average current flowing through the output terminal. These parameters are easily obtained internally by filtering and sampling circuits and are highly correlated with the actual load state. The load monitoring module compares the real-time monitored parameter values ​​with a load state range established in calibration mode to identify the closest current load state point and output a corresponding current load state identifier.

[0059] Meanwhile, the IC's internal waveform analysis engine will extract features from the transient waveforms of the switching nodes in real time using the exact same method as in calibration mode, generating a real-time fingerprint vector. This ensures consistency between real-time data and health baseline data in terms of acquisition and calculation methods, thereby guaranteeing the validity of subsequent comparisons.

[0060] Specifically, the real-time fingerprint vector generation process is as follows: The waveform analysis engine calculates the inductor current rise slope in real time. This is achieved by sampling the inductor current twice at high speed during the main switch conduction time in each or every few switching cycles, and calculating its rate of change. This real-time slope value reflects the instantaneous electrical characteristics of the inductor under the current operating state. This real-time inductor current rise slope ( The calculation of ) can follow the following formula:

[0061] ;

[0062] in,

[0063] This represents the real-time inductor current rise slope.

[0064] and These are two different moments during the main switch conduction period in the current switching cycle. and The inductor current value is collected in real time.

[0065] Next, the waveform analysis engine will calculate the ringing frequency of the switching node in real time. By measuring the period of the ringing waveform using an internal high-speed counter during the main switch off dead time in each or every few switching cycles, the resonant frequency reflecting the current inductance and the equivalent capacitance of the switching node can be obtained.

[0066] Furthermore, to obtain more comprehensive status information, the waveform analysis engine can preferably calculate the ringing attenuation ratio of the switching node in real time. This is achieved by detecting and calculating the ratio of two consecutive positive peak voltages of the ringing waveform within the same dead time, thus quantifying the damping characteristics of the current resonant circuit. This real-time switching node ringing attenuation ratio (… The calculation of ) can follow the following formula:

[0067] ;

[0068] in,

[0069] This indicates the real-time ringing attenuation ratio of the switching nodes;

[0070] and These are the first and second consecutive positive peak voltages captured in real time on the waveform of the switching node during the dead time of the current switching cycle.

[0071] Through the aforementioned parallel real-time monitoring and extraction operations, at any given moment during normal operation, the switching power supply IC can obtain two key output pieces of information: one is the current load state identifier representing the current operating environment, and the other is the real-time fingerprint vector representing the current physical state of the system. These two pieces of information will serve as inputs to the subsequent deviation analysis and protection decision-making modules, thus constituting the logical premise for the adaptive and predictive fault protection of this invention.

[0072] S3: Based on the current load status, determine the reference fingerprint vector for the current comparison from one or more pre-stored health fingerprint vectors;

[0073] In this embodiment, a reference fingerprint vector for the current comparison is determined from one or more pre-stored health fingerprint vectors based on the current load status. The specific implementation may include the following steps:

[0074] In the normal operating mode of the switching power supply IC, immediately after the load monitoring module determines the current load state and the waveform analysis engine extracts the real-time fingerprint vector, the control logic inside the IC will execute a crucial benchmark invocation step. This step is the core element for realizing the adaptive protection function of this invention; its purpose is to ensure that subsequent deviation analysis is performed on a fair and relevant benchmark, i.e., to achieve comparison under the same operating conditions.

[0075] Specifically, the IC integrates a storage and query module that pre-stores one or more health fingerprint vectors established in calibration mode.

[0076] In a preferred embodiment, these health fingerprint vectors are stored in the form of a dynamic fingerprint mapping table. This mapping table establishes and solidifies a one-to-one correspondence between multiple load state point identifiers and their respective corresponding health fingerprint vectors, which contain characteristic parameters such as inductor current rise slope, switching node ringing frequency, and switching node ringing attenuation ratio.

[0077] When the storage and query module receives the current load status identifier output by the load monitoring module, it will use this identifier as the index or keyword for the query.

[0078] Subsequently, the storage and query module will perform a retrieval and matching operation in the dynamic fingerprint mapping table. When an entry is found that completely matches the input current load status identifier, the module will read and output the complete health fingerprint vector corresponding to that entry.

[0079] The health fingerprint vector retrieved and retrieved is then used as the reference fingerprint vector for this comparison cycle. It represents the idealized waveform characteristics that a healthy power supply system should exhibit under load conditions that are the same as or very similar to the current operating conditions.

[0080] By executing this dynamic invocation process, this invention effectively solves the technical problems faced by traditional fixed threshold protection methods. It acknowledges and adapts to the objective law that the waveform characteristics of switching power supplies will undergo normal, non-fault-related changes with load conditions. By consistently selecting the health fingerprint that best matches the current load as the comparison benchmark, subsequent deviation analysis can more accurately identify the true deviations caused by abnormal factors such as component aging and potential saturation risks, thereby significantly improving the accuracy and reliability of fault warnings.

[0081] In a more refined implementation, if the current load state falls precisely between two calibrated load state points, the storage and query module can also be configured to perform an interpolation calculation. For example, the module can simultaneously retrieve the health fingerprint vectors of two load points adjacent to the current load state, and perform linear or nonlinear interpolation on each component of the two health fingerprint vectors based on the relative position of the current load point between these two calibration points, thereby generating a more accurate, dynamically synthesized reference fingerprint vector.

[0082] Ultimately, whether through direct query matching or interpolation calculation, the reference fingerprint vector determined in this step will be immediately transmitted to the subsequent deviation analysis module for element-by-element quantitative comparison with the real-time fingerprint vector extracted within the same period, providing direct data input for the final fault diagnosis and protection decision.

[0083] S4: Compare the real-time fingerprint vector with the reference fingerprint vector to generate an offset vector;

[0084] In this embodiment, the real-time fingerprint vector is compared with the reference fingerprint vector to generate an offset vector. The specific implementation method may include the following steps:

[0085] After the deviation analysis module of the switching power supply IC receives the reference fingerprint vector representing the current ideal operating state and the real-time fingerprint vector representing the current actual operating state, determined by the previous steps, it will immediately initiate an element-by-element quantization comparison process. This process is designed to transform the difference between the two high-dimensional feature vectors into a structured deviation metric that can be directly used for fault determination.

[0086] Specifically, the deviation analysis module will operate on the two received vectors. The real-time fingerprint vector can be represented as... The reference fingerprint vector can be represented as .

[0087] The core operation of the comparison is to calculate the relative deviation of each feature parameter component in the real-time fingerprint vector relative to the corresponding component in the reference fingerprint vector. Using relative deviation instead of absolute difference has the advantage of normalizing the deviation, allowing the subsequent threshold decision logic to be independent of the absolute value of the feature parameter at different load points. This simplifies threshold setting and enhances the universality of the decision.

[0088] This calculation process will generate a deviation vector with the same dimension as the input vector. Each component of this deviation vector independently characterizes the deviation from a specific physical property. The deviation vector can be represented as... .

[0089] Among them, the first component of the deviation vector is the deviation of the inductor current rising slope ( Its calculation can follow the following formula:

[0090] ;

[0091] In this formula, The calculated deviation value; The slope of the inductor current rise in the real-time fingerprint vector; and This represents the rising slope of the inductor current in the reference fingerprint vector. The magnitude of this component directly reflects the real-time change in the electrical characteristics of the main inductor.

[0092] The second component of the deviation vector is the deviation of the ringing frequency of the switching node ( Its calculation can follow the following formula:

[0093] ;

[0094] In this formula, The calculated deviation value; The ringing frequency of the switch nodes in the real-time fingerprint vector; and This refers to the ringing frequency of the switching node in the reference fingerprint vector. The magnitude of this component can indicate potential changes in parameters such as the main inductance or the parasitic capacitance of the main switch.

[0095] Preferably, the third component of the deviation vector, namely the deviation of the ringing attenuation ratio of the switching node ( Its calculation can follow the following formula:

[0096] ;

[0097] In this formula,

[0098] The calculated deviation value;

[0099] The ringing attenuation ratio of the switch nodes in the real-time fingerprint vector;

[0100] and This is the ringing attenuation ratio of the switch node in the reference fingerprint vector.

[0101] The magnitude of this component can be used to characterize the performance degradation trend of damping-related components such as output capacitor ESR.

[0102] By performing the above calculations, the deviation analysis module converts the original waveform characteristic parameters, which have different physical meanings, into dimensionless relative values ​​that characterize the degree of deviation.

[0103] Ultimately, the complete deviation vector output by this module will be transmitted as a structured diagnostic data packet to the subsequent protection execution module. This vector not only reflects the overall deviation of the system's health status, but the different combinations of its internal components can also provide a basis for distinguishing specific fault types, thus laying the data foundation for implementing refined, tiered protection response strategies.

[0104] S5: Based on the deviation vector, perform preset fault protection actions on the switching power supply.

[0105] In this embodiment, based on the deviation vector, a preset fault protection action is performed on the switching power supply. The specific implementation method may include the following steps:

[0106] After the protection execution module of the switching power supply IC receives the deviation vector generated by the deviation analysis module, which quantifies the deviation of the system state, it will initiate a hierarchical and intelligent decision-making and response process. This process is designed to transform abstract deviation data into precise and appropriate physical intervention or status reporting, thereby achieving a leap from simple fault shutdown to refined health management.

[0107] Specifically, the protection execution module pre-stores at least two sets of multi-dimensional threshold vectors: a warning threshold vector and a fault threshold vector. The dimensions of these two threshold vectors correspond exactly to the deviation vector, and each component within them represents a decision threshold set for the degree of deviation from a specific fingerprint feature. Typically, the component values ​​of the fault threshold vector are greater than the corresponding component values ​​of the warning threshold vector.

[0108] The protection execution module will process the received deviation vector ( The absolute values ​​of each component of the threshold vector are compared in real time with the corresponding components of the two threshold vectors mentioned above. To prevent misjudgment caused by transient noise or disturbance, preferably, the comparison result is confirmed by a digital filter or a fault counter, that is, the subsequent response action is officially triggered only after the abnormal state has lasted for a preset confirmation time.

[0109] Based on the comparison results, the protection execution module will perform the following graded fault protection actions:

[0110] Level 1 Response: Non-disruptive early warning actions

[0111] When the absolute value of any component of the deviation vector exceeds its corresponding warning threshold but does not reach the fault threshold, the system determines that there is a potential, slowly developing performance degradation trend or a "soft" fault. In this case, to avoid affecting the normal operation of the system, the module will execute a non-disruptive warning action.

[0112] Preferably, the early warning action may include: sending one or more status flag bits to the system's main controller or monitoring unit through the communication interface inside the IC to prompt the system level to pay attention or arrange preventive maintenance.

[0113] In another implementation, the warning action may further include writing the current complete deviation vector, the identified out-of-limit parameter type, and timestamp information into a dedicated, non-volatile fault log register within the IC. This provides valuable data records for subsequent offline fault diagnosis and lifecycle analysis.

[0114] Level 2 response: Disruptive protective action

[0115] When the absolute value of any component of the deviation vector instantaneously or within a very short time exceeds its corresponding, more stringent fault threshold, the system determines that a "hard" fault that may damage the hardware, such as inductor saturation, is about to occur or has already occurred. At this point, the module must execute interruptible protection actions to ensure safety.

[0116] To improve system availability and robustness, interruption protection actions can be further subdivided into two sub-phases:

[0117] First, implement adaptive intervention. As the first line of defense, the module can attempt proactive, localized adjustments instead of immediately shutting down the entire system. For example, it can temporarily and slightly lower the cycle-by-cycle current limit to directly suppress abnormal current growth, clamp the maximum duty cycle to limit energy input, or appropriately reduce the switching frequency to increase the magnetic reset time. This aims to proactively intervene in the fault's evolution, attempting to allow the system to recover to a safe operating range without interrupting power supply to the load.

[0118] Secondly, in cases where adaptive intervention is ineffective or the initial deviation value is extremely large, a safety shutdown is executed. If the aforementioned adaptive intervention measures fail to restore the deviation vector to a safe range within a preset time, or if the deviation value indicates an irreversible serious fault, the module will immediately activate traditional safety protection mechanisms. For example, it may put the switching power supply into an automatically recoverable hiccup restart mode, or into a latch-off shutdown mode that requires external intervention to recover, thereby completely cutting off the power transmission path to maximize the protection of the switching power supply itself and the downstream loads.

[0119] Through the aforementioned hierarchical decision-making and multi-strategy response mechanism based on deviation vectors, this invention can not only distinguish the severity of faults, but also take a series of progressive protection actions that match the risk level, from information prompts and proactive intervention to safety shutdown, thereby constructing a more comprehensive, intelligent and efficient integrated fault protection system.

[0120] Please see the appendix Figure 2 A fault protection system integrated into a switching power supply IC, comprising:

[0121] The fingerprint extraction module is used to extract a healthy fingerprint vector from the waveform of the switch node in a preset calibration mode, and to extract the real-time fingerprint vector of the switch node in real time in normal operation mode.

[0122] The load monitoring module is used to monitor the current load status of the switching power supply in real time during normal operation.

[0123] The storage and query module is used to store one or more health fingerprint vectors and can determine the reference fingerprint vector for the current comparison based on the current load status.

[0124] The deviation analysis module is used to compare the real-time fingerprint vector with the reference fingerprint vector to generate a deviation vector;

[0125] The protection execution module is used to perform preset fault protection actions on the switching power supply based on the deviation vector.

[0126] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for integrated fault protection of a switching power supply IC, characterized in that, Includes the following steps: S1: In a preset calibration mode, the switching node waveform of the switching power supply is acquired at at least one load state point, and a health fingerprint vector is extracted from the switching node waveform. The health fingerprint vector characterizes the operating characteristics of the switching power supply in a healthy state. S2: In normal operating mode, monitor the current load status of the switching power supply in real time and extract the real-time fingerprint vector of the switching node in real time. S3: Based on the current load state, determine the reference fingerprint vector for the current comparison from one or more pre-stored health fingerprint vectors; S4: Compare the real-time fingerprint vector with the reference fingerprint vector to generate an offset vector; S5: Based on the deviation vector, perform a preset fault protection action on the switching power supply; The health fingerprint vector and the real-time fingerprint vector both include the following feature parameters: inductor current rise slope, switching node ringing frequency, and switching node ringing attenuation ratio.

2. The method according to claim 1, characterized in that, The method for calculating the rising slope of the inductor current is limited to: ； in, This indicates the slope of the inductor current rise; and These are two different moments during the main switch conduction period. and The inductor current value collected below.

3. The method according to claim 1, characterized in that, The calculation method for the ringing attenuation ratio of the switching node is limited to: ； in, This indicates the ringing attenuation ratio of the switching node; and These are the first and second positive peak voltages that appear consecutively on the waveform of the switching node during the dead time after the main switch is turned off.

4. The method according to claim 1, characterized in that, The step of generating the deviation vector in S4 specifically includes: The relative deviation of each feature parameter in the real-time fingerprint vector from the corresponding feature parameter in the reference fingerprint vector is calculated to generate the deviation vector composed of the relative deviation.

5. The method according to claim 1, characterized in that, The steps in the preset calibration mode further include: The switching power supply is driven to operate stably at multiple different load state points in sequence; At each load state point, a corresponding health fingerprint vector is extracted; Each load state point is associated with its corresponding health fingerprint vector to establish and store a dynamic fingerprint mapping table. Specifically, the step of determining the reference fingerprint vector for the current comparison involves: querying and retrieving the corresponding healthy fingerprint vector from the dynamic fingerprint mapping table based on the current load status, using it as the reference fingerprint vector.

6. The method according to claim 1, characterized in that, The step of real-time monitoring of the current load state of the switching power supply is achieved by monitoring the average output current or PWM duty cycle.

7. The method according to claim 1, characterized in that, The step of performing a preset fault protection action on the switching power supply based on the deviation vector includes: The deviation vector is compared with preset warning thresholds and fault thresholds; If the deviation vector exceeds the warning threshold but does not exceed the fault threshold, a non-disruptive warning action is executed. If the deviation vector exceeds the fault threshold, an interruption protection action is performed.

8. The method according to claim 7, characterized in that, The non-disruptive early warning actions include: reporting the status flag bit to the system main controller through the communication interface, or recording it in the internal fault log register.

9. The method according to claim 7, characterized in that, The interruption protection actions include: performing adaptive intervention, or putting the switching power supply into hiccup mode or latch-off mode; The adaptive intervention includes: temporarily lowering the cycle-by-cycle current limit point, limiting the maximum duty cycle, or reducing the switching frequency.

10. A switching power supply IC integrated fault protection system, comprising a switching power supply IC integrated fault protection method according to any one of claims 1-9, characterized in that, Includes the following steps: The fingerprint extraction module is used to extract a healthy fingerprint vector from the waveform of the switching node in a preset calibration mode, and to extract the real-time fingerprint vector of the switching node in real time in a normal operation mode; both the healthy fingerprint vector and the real-time fingerprint vector contain the following characteristic parameters: inductor current rise slope, switching node ringing frequency, and switching node ringing attenuation ratio. The load monitoring module is used to monitor the current load status of the switching power supply in real time during normal operation. The storage and query module is used to store one or more of the health fingerprint vectors and to determine the reference fingerprint vector for the current comparison based on the current load status. A deviation analysis module is used to compare the real-time fingerprint vector with the reference fingerprint vector to generate a deviation vector; The protection execution module is used to perform preset fault protection actions on the switching power supply based on the deviation vector.

Images