Method, device and system for monitoring status of optical module and optical module

Abstract

The application discloses a state monitoring method, device and system of an optical module and the optical module. The method comprises the following steps: acquiring a current signal of at least one heterogeneous node inside the optical module, and extracting a current characteristic value from the current signal; performing temperature compensation on the current characteristic value to obtain a temperature normalized characteristic value; acquiring a normalized characteristic value baseline, which is generated based on historical data of the temperature normalized characteristic value; comparing the normalized characteristic value baseline of the last period with the temperature normalized characteristic value of the current period, and updating the normalized characteristic value baseline and / or generating a fault alarm signal based on the comparison result. The application extracts the current characteristic value based on the current signal of the heterogeneous node inside the optical module, performs temperature compensation on the current characteristic value, and iteratively updates the dynamic baseline based on the statistical characteristics, thereby converting the fixed absolute threshold into the dynamic baseline drifting with the working condition, so that the dynamic aging process of the optical module can be adapted, and the false positive rate and the false negative rate can be significantly reduced.

Description

Technical Field

[0001] This application belongs to the field of optical module monitoring technology, specifically relating to a method, device, system, and optical module for monitoring the status of an optical module. Background Technology

[0002] As data centers evolve towards 800G and 1.6T speeds, the integration and power density of optical modules are increasing dramatically. The health of the power supply network and optoelectronic conversion link of key internal components (such as DSPs, lasers, modulators, etc.) directly affects system reliability.

[0003] Existing optical module status monitoring methods collect the current parameters of the optical module and make fault judgments based on fixed thresholds, which cannot adapt to the dynamic aging process of the optical module, resulting in a high false alarm rate. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, system and optical module for monitoring the status of optical modules, so as to solve the technical problem that the existing optical module status monitoring methods are based on fixed thresholds for fault determination, which cannot adapt to the dynamic aging process of optical modules and result in a high false alarm rate.

[0005] To achieve the above objectives, the first aspect of this application provides a method for monitoring the state of an optical module, comprising:

[0006] Acquire the current signal of at least one heterogeneous node inside the optical module, and extract the current feature value from the current signal;

[0007] Temperature compensation is applied to the current characteristic value to obtain a temperature-normalized characteristic value;

[0008] Obtain a normalized eigenvalue baseline, which is generated based on historical data of the temperature normalized eigenvalues;

[0009] The normalized characteristic value baseline of the previous cycle is compared with the temperature normalized characteristic value of the current cycle, and the normalized characteristic value baseline is updated and / or a fault alarm signal is generated based on the comparison result.

[0010] In one or more embodiments, the step of temperature compensation for the current characteristic value includes:

[0011] The sampled temperature and the current characteristic value are input into a pre-stored compensation function to calculate the equivalent value of the current characteristic value at a standard temperature, thereby obtaining the temperature normalized characteristic value. The compensation function is used to describe the mapping relationship between the current characteristic value and the temperature.

[0012] In one or more embodiments, the step of comparing the normalized feature value baseline of the previous period with the temperature normalized feature value of the current period, and updating the normalized feature value baseline and / or generating a fault alarm signal based on the comparison result includes:

[0013] Based on the baseline of the normalized eigenvalue of the previous period, determine the steady-state interval; determine whether the normalized temperature eigenvalue of the current period falls within the steady-state interval;

[0014] If so, update the normalized feature value baseline based on the temperature normalized feature value of the current period.

[0015] In one or more embodiments, the step of determining the steady-state interval based on the normalized eigenvalue baseline includes:

[0016] Calculate the statistical mean and standard deviation of the historical data of the temperature normalized eigenvalues ​​corresponding to the baseline of the normalized eigenvalues;

[0017] The steady-state interval is generated based on the statistical mean and the standard deviation.

[0018] In one or more embodiments, if the temperature normalized characteristic value of the current period does not fall within the steady-state range, the method further includes:

[0019] Mark the current period as an abnormal period;

[0020] Determine whether the number of consecutive occurrences of the abnormal cycle reaches the first cycle threshold;

[0021] If so, generate an abnormal current alarm signal.

[0022] In one or more embodiments, if the normalized feature value baseline is updated, the state monitoring method further includes:

[0023] Specific feature values ​​are extracted from the current signal;

[0024] Obtain a baseline for a specific feature value, which is generated based on historical data of the specific feature value;

[0025] The baseline of the specific feature value in the previous period is compared with the specific feature value in the current period, and the baseline of the specific feature value is updated and / or a fault alarm signal is generated based on the comparison result.

[0026] In one or more embodiments, the heterogeneous node includes a laser node and a modulator node;

[0027] The status monitoring method includes:

[0028] Acquire the first current sampling signal of the clock-synchronized laser node and the second current sampling signal of the modulator node;

[0029] The first current feature value and the second current feature value are extracted from the first current sampling signal and the second current sampling signal, respectively.

[0030] Temperature compensation is applied to the first current characteristic value and the second current characteristic value respectively to obtain the first normalized current characteristic value and the second normalized current characteristic value;

[0031] Obtain a first normalized eigenvalue baseline and a second normalized eigenvalue baseline, wherein the first normalized eigenvalue baseline and the second normalized eigenvalue baseline are generated based on historical data of the first normalized current eigenvalue and the second normalized current eigenvalue, respectively.

[0032] Compare the first normalized characteristic value baseline of the previous cycle with the first normalized current characteristic value of the current cycle, and update the first normalized characteristic value baseline and / or generate a fault alarm signal based on the comparison result.

[0033] Compare the baseline of the second normalized characteristic value of the previous cycle with the second normalized current characteristic value of the current cycle, and update the baseline of the second normalized characteristic value and / or generate a fault alarm signal based on the comparison result.

[0034] If both the first normalized eigenvalue baseline and the second normalized eigenvalue baseline are updated, then the peak time delay of the first current sampling signal and the second current sampling signal is calculated as the specific eigenvalue.

[0035] In one or more embodiments, the step of calculating the peak delay of the first current sampling signal and the second current sampling signal includes:

[0036] The first current sampling signal and the second current sampling signal are respectively subjected to mean subtraction to obtain a first zero-mean sequence and a second zero-mean sequence;

[0037] Calculate the cross-correlation function between the first zero-mean sequence and the second zero-mean sequence, and determine the peak delay based on the peak index and sampling period of the cross-correlation function.

[0038] In one or more embodiments, the heterogeneous node includes a core power rail node of a digital signal processor;

[0039] The status monitoring method includes:

[0040] Acquire the third current sampling signal of the core power rail node of the digital signal processor;

[0041] The third current feature value is extracted from the third current sampling signal;

[0042] Temperature compensation is applied to the third current characteristic value to obtain the third normalized current characteristic value;

[0043] Obtain the third normalized eigenvalue baseline, which is generated based on historical data of the third normalized current eigenvalue;

[0044] The baseline of the third normalized characteristic value of the previous cycle is compared with the third normalized current characteristic value of the current cycle, and the baseline of the third normalized characteristic value is updated and / or a fault alarm signal is generated based on the comparison result.

[0045] If the baseline of the third normalized feature value is updated, then a fast Fourier transform is performed on the third current sampling signal to extract the current ripple energy integral in the preset high-frequency range as the specific feature value.

[0046] In one or more embodiments, the step of comparing the specific feature value baseline of the previous period with the specific feature value of the current period, and updating the specific feature value baseline and / or generating a fault alarm signal based on the comparison result includes:

[0047] Based on the baseline of the specific characteristic value of the previous period, the tolerance range is determined;

[0048] Determine whether the specific characteristic value of the current period falls within the tolerance range;

[0049] If so, update the baseline of the specific feature value based on the specific feature value of the current period.

[0050] In one or more embodiments, the specific feature value is extracted from a first current sampling signal of a clock-synchronized laser node and a second current sampling signal of a modulator node;

[0051] If the specific characteristic value of the current period does not fall within the tolerance range, the following is also included:

[0052] Mark the current period as an abnormal period;

[0053] Determine whether the number of consecutive occurrences of the abnormal cycle reaches the second cycle threshold;

[0054] If so, generate an alarm signal for optical path abnormality.

[0055] In one or more embodiments, the specific feature value is extracted from the third current sampling signal of the core power rail node of the digital signal processor;

[0056] If the specific characteristic value of the current period does not fall within the tolerance range, the following is also included:

[0057] Obtain the specific feature value for M consecutive periods, and determine whether all of them are greater than or equal to the statistical mean of the baseline of the specific feature value;

[0058] If the specific characteristic value is greater than or equal to the statistical mean for M consecutive periods, determine whether the specific characteristic value is monotonically increasing within the M consecutive periods;

[0059] If the specific characteristic value increases monotonically over M consecutive periods, obtain the sampled voltage of the core power rail node and determine whether it falls within the rated voltage range.

[0060] If the sampled voltage of the core power rail node falls within the rated voltage range, a capacitor aging alarm signal is generated.

[0061] In one or more embodiments, generating the optical path anomaly alarm signal further includes:

[0062] Determine whether the capacitor aging alarm signal is enabled;

[0063] If so, generate a power failure warning signal.

[0064] In one or more embodiments, the step of updating the normalized eigenvalue baseline includes:

[0065] Using a preset learning rate, the normalized temperature feature value of the current period is weighted and merged with the baseline of the normalized feature value of the previous period to obtain the updated baseline of the normalized feature value. The preset learning rate is used to describe the weight of the normalized temperature feature value of the current period.

[0066] In one or more embodiments, the baseline of the normalized eigenvalues ​​of the previous period is generated based on several historical period data of the temperature normalized eigenvalues ​​stored within a sliding window, and the step of updating the baseline of the normalized eigenvalues ​​includes:

[0067] The normalized temperature feature value of the current period is stored in the sliding window, and the earliest historical period data in the sliding window is removed to obtain the updated sliding window.

[0068] Based on the updated sliding window, an updated normalized feature value baseline is generated.

[0069] To achieve the above objectives, a second aspect of this application provides a state monitoring device for an optical module, comprising:

[0070] The feature extraction module is used to acquire the current signal of at least one heterogeneous node inside the optical module, and extract the current feature value of the current cycle from the current signal.

[0071] The temperature compensation module is used to perform temperature compensation on the current characteristic value of the current cycle to obtain the temperature normalized characteristic value of the current cycle.

[0072] The baseline acquisition module is used to acquire the normalized feature value baseline of the previous period, wherein the normalized feature value baseline is generated based on the historical data of the temperature normalized feature value.

[0073] The monitoring module is updated to compare the normalized feature value baseline of the previous cycle with the temperature normalized feature value of the current cycle, and updates the normalized feature value baseline and / or generates a fault alarm signal based on the comparison result.

[0074] To achieve the above objectives, a third aspect of this application provides a status monitoring system for an optical module, comprising:

[0075] The sensor is configured to sample the current signal of at least one heterogeneous node within the optical module;

[0076] The controller is connected to the sensor signal;

[0077] The controller includes at least one processor and a memory, the memory storing instructions. When the instructions are executed by the at least one processor, the at least one processor performs the optical module status monitoring method as described in any of the above embodiments.

[0078] In one or more embodiments, the sensor is configured to synchronously sample the current signals of the laser node and the modulator node based on a uniform sampling clock; and / or,

[0079] The sensor is configured to sample the current signal from the core power rail node of the digital signal processor; and / or,

[0080] The sensor is configured to synchronously sample the current signals of the laser node, modulator node, and core power rail node of the digital signal processor based on a unified sampling clock.

[0081] To achieve the above objectives, a fourth aspect of this application provides an optical module, including a state monitoring system for the optical module as described in any of the above embodiments, and / or, using a state monitoring method for the optical module as described in any of the above embodiments for state monitoring.

[0082] The advantages of this application, which differ from existing technologies, are:

[0083] This application decouples temperature from current characteristic values ​​by applying temperature compensation, normalizing the current characteristic values ​​under different temperature conditions to their equivalent values ​​at standard temperature, thus eliminating temperature interference with the characteristic values. Simultaneously, a dynamic baseline based on statistical characteristics is constructed using historical data of the temperature-normalized characteristic values, and this dynamic baseline is iteratively updated. The fixed absolute threshold is transformed into a dynamic envelope that drifts with operating conditions, allowing the fault judgment threshold to adaptively follow the slow aging drift of the device and changes in ambient temperature. By statistically filtering the characteristic values ​​based on the steady-state interval of the statistical mean and standard deviation, transient disturbances are excluded from baseline updates, preventing them from contaminating the baseline. This effectively overcomes the problems of false alarms due to excessively narrow tolerance windows and missed alarms due to excessively wide windows during dynamic aging of static thresholds, significantly reducing both the false alarm and missed alarm rates.

[0084] This application uses the timing and phase relationship between the laser drive current and the modulator operating current as a characteristic of optical path health. Since the timing and phase relationship is determined by the RC constant of the circuit and does not depend on any absolute current threshold, it fundamentally avoids the interference of temperature on the current amplitude. Furthermore, since the laser and modulator are located in the same temperature range, the delay effect of temperature changes on both is common-mode. By calculating the peak delay, the common-mode temperature drift can be offset without complex temperature compensation. The peak delay is extremely sensitive to early failure events such as decreased optical path coupling efficiency and increased drive output impedance, which have not yet caused significant changes in current amplitude. Therefore, it can achieve pre-fault prediction rather than post-fault alarm. Combined with the period threshold criterion of multiple consecutive cycles exceeding the tolerance, occasional disturbances can also be eliminated, ensuring the robustness and proactive nature of the monitoring.

[0085] This application monitors the current signal of the DSP core power rail, performs a Fast Fourier Transform on it, and extracts the current ripple energy integral in a preset high-frequency range in the frequency domain. This decouples the high-frequency ripple energy, which reflects power integrity degradation, from the total current. Since the closed-loop compensation of the voltage regulation module only acts on the voltage and is only effective within its loop bandwidth, the DC voltage remains stable during the fault latency period and appears pseudo-normal when the equivalent series resistance of the decoupling capacitor increases or the impedance of the power distribution network deteriorates. However, the high-frequency current ripple energy changes significantly before the voltage drop, which can overcome the failure window period of hundreds of hours in traditional voltage monitoring and achieve early warning of circuit faults. By using a combination criterion that the high-frequency ripple energy integral increases monotonically for several consecutive cycles and the sampled voltage still falls within the tolerance range, it can uniquely point to the increase of the equivalent series resistance of the decoupling capacitor or the deterioration of the impedance of the power distribution network, eliminating interference such as instantaneous load fluctuations and abnormal DC power supply. Furthermore, by using asymmetric sampling of the core power rail and the slow I / O rail, the computing power allocation is optimized, reducing the amount of data and computing power overhead while ensuring diagnostic capabilities.

[0086] This application uses a fusion decision mechanism to jointly diagnose optical path and circuit characteristics, which can distinguish different failure mechanisms such as optical path misalignment, capacitor aging and power supply failure precursors, and achieve accurate fault warning. Attached Figure Description

[0087] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0088] Figure 1 This is a flowchart illustrating one embodiment of the optical module status monitoring method of this application;

[0089] Figure 2 yes Figure 1 A flowchart illustrating one embodiment corresponding to S400;

[0090] Figure 3 yes Figure 2 A flowchart of one embodiment corresponding to S403b;

[0091] Figure 4 yes Figure 1 A flowchart of one embodiment corresponding to S500;

[0092] Figure 5 yes Figure 1 A flowchart illustrating another implementation of S500;

[0093] Figure 6 yes Figure 1 A flowchart illustrating one embodiment of S700;

[0094] Figure 7 yes Figure 6 A flowchart of one embodiment corresponding to S703b.

[0095] Figure 8 yes Figure 6 A flowchart of another embodiment corresponding to S703b.

[0096] Figure 9 This is a schematic diagram of one embodiment of the optical module status monitoring device of this application. Detailed Implementation

[0097] To enable those skilled in the art to better understand the technical solutions in this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this disclosure.

[0098] The health status of the power supply network and photoelectric conversion link of key components inside optical modules such as digital signal processors, lasers, and modulators directly affects the reliability of the system. Conventional monitoring schemes rely on fixed thresholds to determine the faults of optical modules. For example, traditional schemes rely on the "absolute value" of laser bias current to predict lifespan.

[0099] However, temperature and the dynamic aging process of the optical module alter the operating parameters of key components within the module, causing the same health state to exhibit significantly different amplitudes at different temperatures and aging programs. For instance, temperature and aging degree drastically change the laser's threshold current and slope efficiency. Some monitoring schemes incorporate temperature compensation, but these are all based on static lookup tables and cannot adapt to the nonlinear drift of devices during dynamic aging, leading to numerous false alarms. Furthermore, in the early stages of actual failure, threshold mismatch results in missed alarms, affecting monitoring accuracy.

[0100] To address the aforementioned issues, the applicant has developed a novel condition monitoring method for optical modules. This method decouples temperature from parameters by performing temperature compensation on characteristic values ​​and constructs a dynamic baseline based on statistical features. Fault determination is then performed based on the iteratively updated dynamic baseline, thereby effectively avoiding the impact of temperature and nonlinear drift during dynamic aging on condition monitoring and significantly improving the condition monitoring accuracy of optical modules.

[0101] For ease of understanding, the key components in the optical module field covered in this application will be explained in a unified manner: an optical module mainly includes a laser, a modulator, and a digital signal processor (DSP). The laser generates and provides a stable, continuous optical carrier to the modulator; the modulator, as the core electro-optic modulation device, is used to load high-speed electrical signals onto the optical carrier, completing the signal conversion in the electro-optic domain; the DSP is the core of the entire digital signal processing chain, used on the transmitting side for precoding, signal equalization, digital predistortion, and waveform shaping of the data to be transmitted, outputting a high-speed electrical signal that matches the modulator's driving characteristics; on the receiving side, it performs dispersion compensation, channel equalization, clock recovery, and data demodulation on the electrical signal after photoelectric conversion, achieving reliable transmission and restoration of high-speed optical signals.

[0102] The status monitoring method of this application is described in detail below. Please refer to [link / reference]. Figure 1 , Figure 1This is a flowchart illustrating one implementation of the optical module status monitoring method of this application.

[0103] like Figure 1 As shown, the method includes:

[0104] S100: Obtain the current signal of at least one heterogeneous node inside the optical module, and extract the current feature value from the current signal.

[0105] Traditional monitoring solutions only collect the total input current or casing temperature of the optical module and attempt to diagnose the problem through a correlation model between the total current and temperature.

[0106] Since the total input current of an optical module is the sum of the currents of all internal components, in traditional solutions, when both a short circuit in the DSP core and laser aging manifest as an increase in total current, it is impossible to distinguish the specific failed device. Only a general current anomaly alarm can be given, that is, it can only provide an alarm after a fault occurs, but cannot predict a fault before it occurs. Maintenance personnel still need to shut down the system to check one by one, resulting in poor monitoring performance.

[0107] Unlike traditional solutions, this application diagnoses based on the current signals of heterogeneous nodes within the optical module to achieve fault prediction for specific nodes. These heterogeneous nodes can be any node within the optical module, such as laser nodes, modulator nodes, digital signal processor nodes, transimpedance amplifier nodes, microcontroller nodes, beam splitter nodes, and so on.

[0108] In one implementation, the system can acquire the current signal of a specific single heterogeneous node and extract current feature values, which can be characteristic parameters such as the amplitude, frequency, and mean of the current signal; preferably, the current feature value can be the mean of the current signal, such as the mean of the current signal of a laser node.

[0109] In other implementations, the system can also simultaneously monitor the current signals of multiple heterogeneous nodes and extract current feature values ​​by comparing the current signals of multiple heterogeneous nodes.

[0110] S200. Temperature compensation is applied to the current characteristic value to obtain the temperature normalized characteristic value.

[0111] After extracting the current characteristic value in step S100, this embodiment further decouples the current characteristic value by temperature to avoid the influence of temperature on the current characteristic value.

[0112] Specifically, in one implementation, the sampled temperature and current characteristic values ​​can be input into a pre-stored compensation function to calculate the equivalent value of the current characteristic value at the standard temperature, thereby obtaining the temperature normalized characteristic value.

[0113] The compensation function is used to describe the mapping relationship between the current characteristic value and the temperature. In one implementation, the compensation function can be obtained through factory cyclic calibration fitting.

[0114] In one implementation, the standard temperature may be specifically 25°C; in other implementations, it may be adjusted based on actual operating conditions.

[0115] The sampling temperature can be obtained by the temperature sensor built into the optical module and is used to represent the node temperature when the above current signal is sampled.

[0116] It should be noted that the current characteristic value in this embodiment can be as described above. Figure 2 The first current characteristic value shown can also be the one described above. Figure 3 In a preferred embodiment, the system can simultaneously perform temperature compensation on the first and second current characteristic values ​​to obtain the corresponding temperature-normalized characteristic values.

[0117] S300: Obtain the normalized eigenvalue baseline, which is generated based on historical data of temperature normalized eigenvalues.

[0118] In this embodiment, a normalized feature value baseline is maintained, which is generated from historical data based on temperature normalized feature values. This normalized feature value baseline is used to provide an adaptive reference benchmark for the temperature normalized feature value comparison in the current period.

[0119] Unlike traditional solutions that use statically fixed tolerance thresholds, the normalized eigenvalue baseline in this implementation can be dynamically updated based on the temperature normalized eigenvalue for each cycle. The historical data for the temperature normalized eigenvalue can be steady-state data continuously accumulated during normal module operation. By establishing the normalized eigenvalue baseline on historical statistical features, the baseline can be updated synchronously with the slow aging drift of the device, thereby avoiding mismatch of the static threshold during dynamic aging.

[0120] In one implementation, the baseline of the normalized eigenvalues ​​can be fitted based on all historical data of the temperature normalized eigenvalues, with higher weights assigned to temperature normalized eigenvalues ​​with more recent periods.

[0121] In another implementation, the normalized eigenvalue baseline can also be dynamically generated based on the statistical characteristics of historical data within a sliding window. The system can maintain a sliding window historical dataset, storing temperature normalized eigenvalues ​​for multiple historical periods. This sliding window can be updated based on the most recent steady-state data, and the normalized eigenvalue baseline can be updated accordingly.

[0122] S400: Compare the normalized characteristic value baseline of the previous cycle with the temperature normalized characteristic value of the current cycle, and update the normalized characteristic value baseline and / or generate a fault alarm signal based on the comparison result.

[0123] In this embodiment, the normalized temperature characteristic value of the current cycle is compared with the baseline of the normalized characteristic value of the previous cycle. Based on the comparison result, it is determined whether to update the baseline of the normalized characteristic value and whether to generate a fault alarm signal.

[0124] Specifically, please refer to Figure 2 , Figure 2 yes Figure 1 A flowchart of one embodiment corresponding to S400.

[0125] like Figure 2 As shown, methods for comparing the normalized eigenvalue baseline and the temperature normalized eigenvalue may include:

[0126] S401. Determine the steady-state interval based on the normalized eigenvalue baseline of the previous period.

[0127] The steady-state range represents the normal fluctuation range of the temperature normalized characteristic value under healthy operating conditions.

[0128] In one implementation, the statistical mean and standard deviation of historical data of temperature normalized eigenvalues ​​corresponding to the baseline of normalized eigenvalues ​​can be calculated, and a steady-state interval can be generated based on the statistical mean and standard deviation.

[0129] For example, the steady-state interval can be represented as ,in For statistical mean, Standard deviation The statistical coefficients are preset; in one implementation, k can be 3, that is, 3 is used. The criterion generates a steady-state interval.

[0130] S402. Determine whether the normalized temperature characteristic value of the current cycle falls within the steady-state range.

[0131] If the normalized temperature eigenvalue of the current period falls within the steady-state range, then:

[0132] S403a: Update the normalized eigenvalue baseline based on the temperature normalized eigenvalue of the current cycle.

[0133] Specifically, in one implementation, when the normalized feature value baseline is obtained by fitting all historical data of the temperature normalized feature value, the temperature normalized feature value of the current period is weighted and merged with the normalized feature value baseline of the previous period using a preset learning rate to obtain an updated normalized feature value baseline. The preset learning rate is used to describe the weight of the temperature normalized feature value of the current period in the update.

[0134] For example, the mean and standard deviation of the updated normalized eigenvalue baseline can be expressed as follows:

[0135] ;

[0136] ;

[0137] In the formula, This is the mean of the baseline normalized eigenvalues ​​from the previous period. This is the normalized characteristic value of the temperature for the current period. The standard deviation of the baseline of the normalized eigenvalues ​​of the previous period. The preset learning rate.

[0138] Among them, the preset learning rate The value can be 0.05~0.3, preferably. It can be 0.15. This value has been verified by a large number of cyclic aging experiments. It can follow the slow drift in the room temperature range with a tracking error of ±5%, while effectively suppressing microsecond-level transient interference.

[0139] It should be noted that, The value of needs to strike a balance between noise reduction and tracking. Values ​​that are too small (e.g.) When the baseline is <0.05, the baseline update is too slow and has extremely strong smoothness. The current value only has a minor impact on the baseline. The baseline is too loose and cannot effectively track the normal physical drift of the device caused by changes in ambient temperature, such as the threshold current of the laser increasing with rising temperature. This will cause the residual D(k) to remain positive for a long time, eventually triggering a large number of false alarms. Values ​​that are too large (e.g.) When the baseline is greater than 0.3, it becomes overly sensitive and incorrectly learns current spikes caused by recent sudden transient loads, such as DSP processing sudden data packets, as new normal states. This raises the baseline, causing the residual D(k) to become smaller when a real fault occurs, resulting in severe missed detections and loss of predictive function. Here, residual D(k) represents the deviation between the current eigenvalue and the mean of the normalized eigenvalue baseline.

[0140] In another embodiment, when the normalized eigenvalue baseline is dynamically generated based on the statistical characteristics of historical data within the sliding window, the method for updating the normalized eigenvalue baseline may also include: storing the temperature normalized eigenvalue of the current period into the sliding window, and removing the earliest historical period data within the sliding window to obtain the updated sliding window; and generating the updated normalized eigenvalue baseline based on the updated sliding window.

[0141] This scheme updates the normalized eigenvalue baseline by updating a sliding window of fixed length, allowing the normalized eigenvalue baseline to focus on data from the next few cycles, thereby further ensuring that the normalized eigenvalue baseline matches the dynamic aging degree of the device.

[0142] Based on the above-mentioned dynamic update method of normalized eigenvalue baseline, the fault judgment threshold can adaptively follow the slow aging drift of the device and the change of ambient temperature, and exclude transient disturbances from the baseline update to avoid contaminating the baseline. This effectively overcomes the problem of false alarms caused by the narrow tolerance window and the missed alarms caused by the wide tolerance window during the dynamic aging process of the static threshold, and significantly reduces the false alarm rate and the missed alarm rate.

[0143] Furthermore, if the normalized temperature eigenvalue of the current period does not fall within the steady-state range, then:

[0144] S403b, Enter the exception judgment logic.

[0145] Specifically, when the temperature normalized characteristic value does not fall within the steady-state range, it is regarded as a sudden disturbance or a precursor to a fault. It is rejected for polluting the baseline of the normalized characteristic value and enters the abnormal counting logic to avoid false alarms caused by transient surge current.

[0146] Specifically, please refer to Figure 3 , Figure 3 yes Figure 2 A flowchart illustrating one embodiment corresponding to S403b is shown below. Figure 3 As shown, the exception detection logic can be specifically as follows:

[0147] S4031. Mark the current cycle as an abnormal cycle.

[0148] S4032. Determine whether the number of consecutive occurrences of the abnormal cycle has reached the threshold of the first cycle.

[0149] If the number of consecutive occurrences of the abnormal cycle reaches the threshold of the first cycle, then:

[0150] S4033, Generates an abnormal current alarm signal.

[0151] In this embodiment, when the normalized temperature characteristic value of the current cycle does not fall into the steady-state range, it is marked as an abnormal cycle. When the number of consecutive occurrences of the abnormal cycle reaches the first cycle threshold, a current abnormality alarm is triggered, thereby eliminating instantaneous out-of-range caused by occasional disturbances and avoiding single-point false alarms.

[0152] The anomaly determination of the temperature normalized eigenvalue for each cycle is based on the baseline of the normalized eigenvalue of the previous cycle.

[0153] Based on the state monitoring methods of the above embodiments, by performing temperature decoupling on the current characteristic value of the current signal and comparing it with the normalized characteristic value baseline of the previous cycle, it can be determined whether the current characteristic value is steady-state data. When the data is steady-state, the baseline is updated using the current characteristic value, and when there are multiple consecutive cycles of non-steady-state data, a current abnormality alarm signal is sent.

[0154] By applying temperature compensation to the current characteristic values, temperature is decoupled from the characteristic values, normalizing the current characteristic values ​​under different temperature conditions to the equivalent values ​​at the standard temperature, thus eliminating the interference of temperature on the characteristic values. Simultaneously, a dynamic baseline based on statistical characteristics is constructed based on historical data of the temperature-normalized characteristic values, and the dynamic baseline is iteratively updated. The fixed absolute threshold is transformed into a dynamic envelope that drifts with operating conditions, enabling the fault judgment threshold to adaptively follow the slow aging drift of the device and changes in ambient temperature. Statistical filtering of the characteristic values ​​based on the steady-state interval of the statistical mean and standard deviation excludes transient disturbances from baseline updates, preventing them from contaminating the baseline. This effectively overcomes the problems of false alarms caused by excessively narrow tolerance windows and missed alarms caused by excessively wide windows during dynamic aging of static thresholds, significantly reducing both false alarm and missed alarm rates.

[0155] In one implementation, when the current characteristic value of the current cycle is determined to be steady-state data, further fault diagnosis can be performed.

[0156] Specifically, such as Figure 1 As shown, the method also includes:

[0157] S500, if the normalized eigenvalue baseline is updated, specific eigenvalues ​​are extracted from the current signal.

[0158] This embodiment further extracts specific feature values ​​from the steady-state current signal to perform fault monitoring of each component.

[0159] Specifically, in one implementation, please refer to Figure 4 , Figure 4 yes Figure 1 A flowchart of one embodiment corresponding to S500.

[0160] like Figure 4As shown, heterogeneous nodes can include laser nodes and modulator nodes, and methods for extracting specific feature values ​​can include:

[0161] S501a: Acquire the first current sampling signal of the clock-synchronized laser node and the second current sampling signal of the modulator node.

[0162] S502a. Calculate the peak time delay of the first current sampling signal and the second current sampling signal to obtain specific characteristic values.

[0163] Specifically, in the steps S100~S400 above, the first current sampling signal of the clock-synchronized laser node and the second current sampling signal of the modulator node are obtained.

[0164] The first current feature value and the second current feature value are extracted from the first current sampling signal and the second current sampling signal, respectively.

[0165] Temperature compensation is applied to the first current characteristic value and the second current characteristic value respectively to obtain the first normalized current characteristic value and the second normalized current characteristic value.

[0166] Obtain the first normalized eigenvalue baseline and the second normalized eigenvalue baseline. The first normalized eigenvalue baseline and the second normalized eigenvalue baseline are generated based on historical data of the first normalized current eigenvalue and the second normalized current eigenvalue, respectively.

[0167] Compare the first normalized characteristic value baseline of the previous cycle with the first normalized current characteristic value of the current cycle, and update the first normalized characteristic value baseline and / or generate a fault alarm signal based on the comparison result;

[0168] Compare the baseline of the second normalized eigenvalue of the previous cycle with the second normalized current eigenvalue of the current cycle, and update the baseline of the second normalized eigenvalue based on the comparison result and / or generate a fault alarm signal.

[0169] If both the first normalized eigenvalue baseline and the second normalized eigenvalue baseline are updated, then specific eigenvalues ​​can be further extracted from the first current sampling signal and the second current sampling signal.

[0170] The first current sampling signal and the second current sampling signal are guaranteed to be clock synchronized, and a specific characteristic value is obtained by calculating the peak delay of the two sampling signals.

[0171] Traditional methods determine optical link health by monitoring the amplitudes of the laser drive current and modulator current. However, this approach has fundamental flaws: current amplitude is coupled with multiple factors including temperature, bias point, and aging, requiring complex nonlinear temperature compensation, which always leaves residual errors; the sensitivity of the amplitude method is extremely low in the early stages of failure, only triggering an alarm when the failure has progressed to the point of causing a significant current change, making it a reactive alarm; when the optical path is slightly misaligned, the driver chip automatically increases the current to maintain optical power, potentially causing the current amplitude to rise instead of fall, completely masking the fault; the amplitude method can only identify excessively high current amplitudes, failing to accurately distinguish between specific faults such as laser aging, optical path contamination, or driver chip failure. In summary, the traditional amplitude method cannot accurately monitor the operating status of the optical path.

[0172] Therefore, this application abandons the traditional approach and uses the timing and phase relationship between the laser drive current and the modulator current as a specific characteristic value to determine the optical path operating state. Specifically, under normal operating conditions, changes in the laser drive current are coupled to the modulator current through the RC path of the driver chip output stage and the modulator PN junction, and there is a fixed delay between the two determined by the circuit RC constant. When optical path abnormalities occur, such as decreased optical coupling efficiency or increased drive output impedance, the RC constant changes, leading to this fixed delay. The offset occurs, and this delayed offset is extremely sensitive to early failure events that have not yet caused a significant change in current amplitude. Therefore, it can achieve pre-fault prediction, and the monitoring effect is significantly better than the traditional amplitude method.

[0173] In particular, since the laser and modulator are located in the same temperature region, the delay effect of temperature change on both is common-mode. By obtaining the peak delay of the current signals of both, the common-mode temperature drift can be offset without complex compensation.

[0174] Specifically, methods for calculating peak latency may include:

[0175] The first current sampling signal and the second current sampling signal are respectively subjected to mean subtraction to obtain the first zero-mean sequence and the second zero-mean sequence;

[0176] Calculate the cross-correlation function of the first zero-mean sequence and the second zero-mean sequence, and determine the peak delay based on the peak index and sampling period of the cross-correlation function.

[0177] This implementation first eliminates the bias of the DC component on the cross-correlation function by means subtraction, and then determines the peak time delay by calculating the cross-correlation function.

[0178] The cross-correlation function can be specifically expressed as: Rxy(τ) = In the formula, This represents the time-domain signal of the laser drive current. The time-domain signal representing the modulator's operating current. For time variables, For time delay variables, representing Compared to Time offset.

[0179] Based on this cross-correlation function, the cross-correlation values ​​of the first current sampling signal and the second current sampling signal when they are offset by different sampling periods can be calculated, and the number of offset sampling periods corresponding to the maximum cross-correlation value can be determined, thus obtaining the peak index. By calculating the product of the peak index and the sampling period, the peak delay can be obtained.

[0180] By monitoring the offset of the peak delay relative to the baseline, the health of the optical path can be monitored, and an early warning can be issued when an optical path abnormal event occurs (i.e. when the offset of the peak delay relative to the baseline exceeds the threshold), ensuring the robustness and proactiveness of the monitoring.

[0181] In another implementation, the system can also monitor the core power rail nodes of the digital signal processor to achieve circuit health monitoring.

[0182] Traditional monitoring solutions for DSP power rails often focus on voltage dips or ripple monitoring. However, in real-world scenarios, when decoupling capacitors age or PCB via impedance increases, the voltage remains stable during the fault latency period due to the closed-loop compensation characteristics of the voltage regulator module (VRM). Relying solely on voltage monitoring results in a failure window of hundreds of hours, making it impossible to provide early warning before a fault occurs.

[0183] Therefore, this implementation method bypasses traditional voltage-based methods and achieves early warning of circuit faults by monitoring the current signal of the DSP's core power rail node. Since the compensation effect of the VRM only applies to voltage and is effective only within its loop bandwidth, it cannot eliminate high-frequency ripple components in the power rail current. Physically, an increase in the equivalent series resistance (ESR) of the decoupling capacitor raises the equivalent impedance of the power rail at high frequencies, causing switching actions to generate larger high-frequency ripples in the current. Therefore, early warning of circuit faults can be achieved by monitoring the current signal of the DSP's core power rail.

[0184] The system can simultaneously monitor the current signals of multiple different core power rails, such as 1.2V and 0.8V, or it can monitor the current signal of a single core power rail, both of which can achieve the effect of this implementation.

[0185] In particular, in one implementation, the system can also synchronously monitor the slow I / O rails of the DSP. The core power rail can be configured with a sampling rate of ≥10MHz, and the slow I / O rail can be configured with a sampling rate of ≤1MHz to achieve asymmetric sampling. This optimizes the allocation of computing power, allowing limited edge computing power to be concentrated on the most critical fault characteristics, and significantly reduces the amount of data and computing power without sacrificing diagnostic capabilities.

[0186] Specifically, please refer to Figure 5 , Figure 5 yes Figure 1 A flowchart illustrating another implementation of S500.

[0187] like Figure 5 As shown, heterogeneous nodes may include the core power rail nodes of a digital signal processor, and methods for extracting specific feature values ​​may include:

[0188] S501b: Acquire the third current sampling signal of the core power rail node of the digital signal processor.

[0189] S502b: Perform a fast Fourier transform on the third current sampling signal, extract the current ripple energy integral in the preset high-frequency range, and obtain specific characteristic values.

[0190] Specifically, in steps S100 to S400 above, the third current sampling signal of the core power rail node of the digital signal processor is obtained; the third current feature value is extracted from the third current sampling signal; the third current feature value is temperature compensated to obtain the third normalized current feature value; the baseline of the third normalized feature value is obtained, which is generated based on the historical data of the third normalized current feature value; the baseline of the third normalized feature value of the previous cycle is compared with the third normalized current feature value of the current cycle, and the baseline of the third normalized feature value is updated based on the comparison result and / or a fault alarm signal is generated.

[0191] If the baseline of the third normalized eigenvalue is updated, then specific eigenvalues ​​can be further extracted from the third current sampling signal.

[0192] The third current sampling signal of the core power rail node includes a DC component (corresponding to the average power consumption of the DSP), switching noise (generated by the switching of the internal circuitry of the DSP), and ripple (from the switching frequency of the DC-DC converter and its harmonics). Since it is difficult to intuitively distinguish between normal switching noise and abnormal ESR ripple rise in the time-domain voltage / current waveform, the method of this embodiment performs a fast Fourier transform on the core rail current to convert it into frequency domain features, and extracts the current ripple energy integral in a preset high-frequency range as a specific feature value to determine abnormal circuit events.

[0193] In one implementation, the preset high-frequency range can be specifically 1MHz–10MHz; in other implementations, it can be adjusted based on actual needs.

[0194] In the above Figure 4 and Figure 5 In the embodiments shown, specific feature values ​​are extracted based on the current signals of the laser and modulator, respectively, and specific feature values ​​are extracted based on the core power rail current signal of the DSP. In a preferred embodiment, the system can simultaneously sample the current signals of the core power rails of the laser, modulator, and DSP, and extract two specific feature values ​​respectively, so as to synchronously realize the status monitoring of the optical path and circuit of the optical module. Preferably, the core power rails of the laser, modulator, and DSP can share a unified clock to ensure that the current signals of different nodes are strictly aligned in the time domain.

[0195] In one implementation, the current signals of multiple heterogeneous nodes can be sampled in a clock-synchronized manner by a multi-node synchronous current sensing network. This network consists of multiple high common-mode rejection ratio (CMRR) differential amplifiers, precision shunt resistors, and a unified clock source. By connecting a low-temperature coefficient precision shunt resistor in series at each node, the current is converted into a voltage drop, which is then amplified by a high CMRR differential amplifier. Since the differential amplifier has a strong ability to suppress common-mode noise, it only amplifies the useful differential-mode voltage across the shunt resistor. Therefore, the small current of each node can be converted into a voltage signal with a high signal-to-noise ratio, providing clean raw data for subsequent condition monitoring.

[0196] S600: Obtain a baseline for a specific feature value. The baseline for a specific feature value is generated based on historical data of a specific feature value.

[0197] In this embodiment, a baseline of specific feature values ​​is maintained, which is generated from historical data based on specific feature values. This specific feature value is used to provide an adaptive reference benchmark for the comparison of specific feature values ​​in the current period.

[0198] Similar to the normalized eigenvalue baseline, the specific eigenvalue baseline in this embodiment can be dynamically updated based on the temperature normalized eigenvalue for each cycle. The historical data for the specific eigenvalue is based on steady-state data continuously accumulated by each module during normal operation. By establishing the specific eigenvalue baseline on historical statistical features, the specific eigenvalue baseline can be updated synchronously with the slow aging drift of the device, thereby avoiding mismatch of static thresholds during dynamic aging.

[0199] The method for generating specific eigenvalue baselines can be the same as that for normalized eigenvalue baselines, and will not be elaborated here.

[0200] S700 compares the specific characteristic value baseline of the previous cycle with the specific characteristic value of the current cycle, and updates the specific characteristic value baseline and / or generates a fault alarm signal based on the comparison result.

[0201] In this embodiment, the specific characteristic value of the current period is compared with the baseline of the specific characteristic value of the previous period, and the comparison result determines whether to update the baseline of the specific characteristic value and whether to generate a fault alarm signal.

[0202] Specifically, please refer to Figure 6 , Figure 6 yes Figure 1 A flowchart of one embodiment corresponding to S700.

[0203] like Figure 6 As shown, methods for comparing a specific eigenvalue baseline from the previous period with a specific eigenvalue from the current period may include:

[0204] S701. Determine the tolerance range based on the baseline of specific characteristic values ​​from the previous period.

[0205] S702. Determine whether a specific characteristic value of the current period falls within the tolerance range.

[0206] If so, then:

[0207] S703a: Update the baseline of a specific eigenvalue based on the specific eigenvalue of the current period.

[0208] The method for determining the tolerance interval can be the same as the method for determining the steady-state interval, that is, the tolerance interval is generated based on the statistical mean and standard deviation of a specific characteristic value baseline, which will not be elaborated here.

[0209] If a specific characteristic value in the current period does not fall within the tolerance range, then:

[0210] S703b, Enter the exception detection logic.

[0211] Furthermore, depending on the type of heterogeneous node, the anomaly detection logic is as follows:

[0212] In one implementation, the heterogeneous nodes include laser nodes and modulator nodes, with a specific characteristic value being peak delay. Correspondingly, a baseline for the specific characteristic value is generated based on historical data of the peak delay, and optical path anomalies can be determined based on the specific characteristic value.

[0213] Please see Figure 7 , Figure 7 yes Figure 6 A flowchart of one embodiment corresponding to S703b is shown.

[0214] like Figure 7 As shown, methods for determining optical path anomalies may include:

[0215] S7031a, Mark the current cycle as an abnormal cycle.

[0216] S7032a. Determine whether the number of consecutive occurrences of the abnormal cycle has reached the threshold of the second cycle.

[0217] If the number of consecutive occurrences of an abnormal cycle reaches the cycle threshold, then:

[0218] S7033a, Generates an alarm signal for optical path abnormality.

[0219] In this embodiment, when a specific characteristic value of the current period does not fall into the steady-state range, it is marked as an abnormal period. When the number of consecutive occurrences of the abnormal period reaches the threshold of the second period, an optical path abnormality alarm is triggered, thereby eliminating instantaneous out-of-tolerance caused by occasional disturbances and avoiding single-point false alarms.

[0220] The anomaly determination of a specific feature value in each cycle is based on the baseline of the specific feature value of the previous cycle.

[0221] The optical path monitoring method based on the above optical module uses the peak delay between the laser drive current and the modulator operating current as the optical path health characteristic. Since the peak delay is determined by the RC constant of the circuit and does not depend on any absolute current threshold, it fundamentally avoids the interference of temperature on the current amplitude. Moreover, the laser and modulator are located in the same temperature range, and the effect of temperature change on the delay of both is common-mode. By calculating the peak delay, the common-mode temperature drift can be offset without complex temperature compensation. The peak delay is extremely sensitive to early failure events such as decreased optical path coupling efficiency and increased drive output impedance, which have not yet caused significant changes in current amplitude. Therefore, it can achieve pre-fault prediction rather than post-fault alarm. Combined with the period threshold criterion of multiple consecutive cycles exceeding the tolerance, occasional disturbances can also be eliminated, ensuring the robustness and proactive nature of the monitoring.

[0222] In another embodiment, the heterogeneous node includes the core power rail node of the digital signal processor. The specific characteristic value is the current ripple energy integral in a preset high-frequency range. Correspondingly, the temperature normalized characteristic value is obtained by temperature decoupling the current ripple energy integral. The baseline of the normalized characteristic value is generated based on the historical data of temperature decoupling of the current ripple energy integral. Circuit anomaly can be determined based on the temperature normalized characteristic value.

[0223] Please see Figure 8 , Figure 8 yes Figure 6 A flowchart illustrating another embodiment corresponding to S703b.

[0224] like Figure 8 As shown, methods for determining circuit anomalies may include:

[0225] S7031b: Obtain specific characteristic values ​​for M consecutive periods and determine whether they are all greater than or equal to the statistical mean of the baseline of the specific characteristic values.

[0226] If a specific characteristic value is greater than or equal to the statistical mean for M consecutive periods, then:

[0227] S7032b: Determine whether a specific characteristic value monotonically increases over M consecutive periods.

[0228] If a specific characteristic value monotonically increases over M consecutive periods, then:

[0229] S7033b: Obtain the sampled voltage of the core power rail node and determine whether it falls within the rated voltage range.

[0230] The rated voltage range can be defined based on the DSP's specifications, for example, it can be ±5% of the rated voltage.

[0231] If the sampled voltage of the core power rail node falls within the rated voltage range, then:

[0232] S7034b generates a capacitor aging alarm signal.

[0233] Specifically, when the ESR of the decoupling capacitor increases or the impedance of the power distribution network deteriorates, the high-frequency current ripple energy continues to rise. However, the closed-loop compensation of the VRM keeps the DC voltage within the rated tolerance range. Therefore, the combination of monotonically increasing high-frequency energy and normal voltage can uniquely point to the gradual degradation of the decoupling capacitor or the power distribution network, while excluding interference such as instantaneous load fluctuations or abnormal DC power supply.

[0234] Therefore, in this embodiment, when the high-frequency current ripple energy integral EHF shows a monotonically increasing trend for several consecutive cycles, and EHF significantly exceeds the dynamic tolerance window, while the corresponding DSP power rail DC voltage value is still within the rated voltage range, it is determined that the equivalent series resistance (ESR) of the decoupling capacitor has increased or the impedance of the power distribution network has deteriorated, thereby triggering a capacitor aging alarm.

[0235] Based on this anomaly detection logic, early warning of circuit faults can be achieved, and it is not affected by temperature or dynamic aging of components, thus ensuring the robustness of optical module status monitoring.

[0236] Based on the circuit monitoring method of the aforementioned optical module, the current signal of the DSP core power rail is monitored, and a fast Fourier transform is performed on it. The current ripple energy integral of the preset high-frequency range is then extracted in the frequency domain, decoupling the high-frequency ripple energy reflecting power integrity degradation from the total current. Since the closed-loop compensation of the voltage regulation module only acts on the voltage and is only effective within its loop bandwidth, when the equivalent series resistance of the decoupling capacitor increases or the impedance of the power distribution network deteriorates, the DC voltage remains stable during the fault latency period and presents a pseudo-normal state. However, the high-frequency current ripple energy will change significantly before the voltage drop. Therefore, it can overcome the failure window period of hundreds of hours that exists in traditional voltage monitoring and achieve early warning of circuit faults.

[0237] By using a combination criterion that the high-frequency ripple energy integral increases monotonically for multiple consecutive cycles while the sampling voltage still falls within the rated voltage range, it can uniquely point to an increase in the equivalent series resistance of the decoupling capacitor or a deterioration in the impedance of the power distribution network, thus eliminating interference such as instantaneous load fluctuations and abnormal DC power supply. Furthermore, by using asymmetric sampling of the core power rail and the slow I / O rail, the computing power allocation is optimized, reducing the amount of data and computing power overhead while ensuring diagnostic capabilities.

[0238] In a preferred embodiment, the steps of optical path anomaly determination and circuit anomaly determination can be performed simultaneously, and the outputs of the two can be merged for determination.

[0239] For example, when based on Figure 7 The steps shown identify optical path anomalies, based on Figure 8 When the circuit is deemed normal, the system can generate an optical path abnormality alarm signal; when both the optical path and the circuit are deemed normal, the system can generate a capacitor aging alarm signal; when both the optical path and the circuit are deemed abnormal at the same time, the system can determine that the current situation is a precursor to a total power supply failure, and the system can then generate a power failure warning signal.

[0240] Through the aforementioned cross-validation fusion decision mechanism, this application can distinguish between different failure mechanisms such as optical path misalignment, capacitor aging, and precursors to total power supply failure, avoiding misjudgments caused by single-dimensional criteria and achieving accurate early warning at the fault mechanism level.

[0241] In the above embodiments, anomaly determination is performed based on the deviation of specific feature values ​​to avoid errors introduced by data processing such as temperature normalization. In other embodiments, temperature normalization can be performed on specific feature values ​​first, and then anomaly determination can be performed on the baseline generated by historical data based on the normalized specific feature values, which can also achieve the effect of this embodiment.

[0242] In particular, for timing features such as peak delay corresponding to laser nodes and modulator nodes that require high waveform phase fidelity, since laser nodes and modulator nodes are in the same temperature range, the specific feature values ​​themselves have already offset the influence of common-mode temperature to a certain extent. Therefore, it is preferable to directly determine anomalies based on specific feature values ​​and specific feature value baselines, avoiding phase distortion and waveform distortion introduced by temperature compensation fitting calculations. This can accurately preserve the timing offset features caused by early failures and significantly improve the sensitivity of pre-detection of optical path faults.

[0243] For the high-frequency ripple energy type numerical characteristics corresponding to the core power rail of a digital signal processor, specific characteristic values ​​can be normalized and compared with the baseline to determine circuit anomalies.

[0244] In the dual-baseline hierarchical architecture of this embodiment, the normalized eigenvalue baseline undertakes the first-level steady-state condition verification function, which is used to eliminate abnormal conditions such as drastic temperature fluctuations and transient load shocks, ensuring that all subsequent baseline iterations and fault determinations are based on valid steady-state data with stable temperature. The specific eigenvalue baseline serves as the second-level judgment benchmark, and only performs its own dynamic updates and anomaly identification under steady-state conditions that have passed the first-level verification.

[0245] This application also provides a status monitoring device for an optical module; please refer to [link / reference]. Figure 9 , Figure 9 This is a schematic diagram of one embodiment of the optical module status monitoring device of this application.

[0246] like Figure 9 As shown, the optical module status monitoring device includes a feature extraction module 21, a temperature compensation module 22, a baseline acquisition module 23, and an update monitoring module 24.

[0247] The feature extraction module 21 is used to acquire the current signal of at least one heterogeneous node inside the optical module, and extract the current feature value of the current period from the current signal.

[0248] The temperature compensation module 22 is used to perform temperature compensation on the current characteristic value of the current cycle to obtain the normalized temperature characteristic value of the current cycle.

[0249] The baseline acquisition module 23 is used to acquire the normalized eigenvalue baseline of the previous period. The normalized eigenvalue baseline is generated based on historical data of temperature normalized eigenvalues.

[0250] The update monitoring module 24 is used to compare the normalized characteristic value baseline of the previous cycle with the temperature normalized characteristic value of the current cycle, and update the normalized characteristic value baseline and / or generate a fault alarm signal based on the comparison result.

[0251] As per the above reference Figures 1 to 8This specification describes a method for monitoring the state of an optical module according to embodiments thereof. The details mentioned in the above description of the method embodiments also apply to the state monitoring device for the optical module according to embodiments thereof. The above-described state monitoring device for the optical module can be implemented in hardware, software, or a combination of hardware and software.

[0252] This application also provides a status monitoring system for an optical module, including sensors and a controller.

[0253] The sensor is configured to sample the current signal of at least one heterogeneous node inside the optical module.

[0254] The controller is connected to the sensor signal and includes at least one processor and a memory. The memory stores instructions, and when the instructions are executed by at least one processor, the at least one processor performs the optical module status monitoring method as described in any of the above embodiments.

[0255] In one implementation, the sensor is configured to synchronously sample the current signals of the laser node and the modulator node based on a unified sampling clock.

[0256] In another implementation, the sensor is configured to sample the current signal of the core power rail node of the digital signal processor.

[0257] In yet another implementation, the sensor is configured to synchronously sample the current signals of the laser node, modulator node, and core power rail node of the digital signal processor based on a unified sampling clock.

[0258] In yet another implementation, the sensor is configured to synchronously sample the current signals of the laser node, modulator node, and core power rail node and slow I / O rail node of the digital signal processor based on a unified sampling clock.

[0259] Specifically, the sensor can be a multi-node synchronous current sensing network, consisting of multiple high common-mode rejection ratio (CMRR) differential amplifiers, precision shunt resistors, and a unified clock source;

[0260] The controller can correspond to an edge computing unit, which can integrate a Cortex-M3 core or an equivalent coprocessor, with a built-in hardware-accelerated FFT unit and cross-correlation operation engine, and communicates with the main DSP through an isolated bus to ensure that the monitoring behavior does not interfere with the integrity of high-speed signals.

[0261] This application also provides an optical module, including a status monitoring system for the optical module as described in any of the above embodiments, and / or, a status monitoring method for the optical module as described in any of the above embodiments.

[0262] It will be apparent to those skilled in the art that this disclosure is not limited to the details of the exemplary embodiments described above, and that this disclosure can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of this disclosure is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this disclosure. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0263] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for monitoring the status of an optical module, characterized in that, include: Acquire the current signal of at least one heterogeneous node inside the optical module, and extract the current feature value from the current signal; Temperature compensation is applied to the current characteristic value to obtain a temperature-normalized characteristic value; Obtain a normalized eigenvalue baseline, which is generated based on historical data of the temperature normalized eigenvalues; The normalized characteristic value baseline of the previous cycle is compared with the temperature normalized characteristic value of the current cycle, and the normalized characteristic value baseline is updated and / or a fault alarm signal is generated based on the comparison result.

2. The condition monitoring method according to claim 1, characterized in that, The step of temperature compensation for the current characteristic value includes: The sampled temperature and the current characteristic value are input into a pre-stored compensation function to calculate the equivalent value of the current characteristic value at a standard temperature, thereby obtaining the temperature normalized characteristic value. The compensation function is used to describe the mapping relationship between the current characteristic value and the temperature.

3. The condition monitoring method according to claim 1, characterized in that, The steps of comparing the normalized feature value baseline of the previous period with the temperature normalized feature value of the current period, and updating the normalized feature value baseline and / or generating a fault alarm signal based on the comparison result include: Based on the normalized eigenvalue baseline of the previous period, the steady-state interval is determined; Determine whether the normalized temperature characteristic value of the current period falls within the steady-state range; If so, update the normalized feature value baseline based on the temperature normalized feature value of the current period.

4. The condition monitoring method according to claim 3, characterized in that, The step of determining the steady-state interval based on the normalized eigenvalue baseline includes: Calculate the statistical mean and standard deviation of the historical data of the temperature normalized eigenvalues ​​corresponding to the baseline of the normalized eigenvalues; The steady-state interval is generated based on the statistical mean and the standard deviation.

5. The condition monitoring method according to claim 3, characterized in that, If the normalized temperature eigenvalue of the current period does not fall within the steady-state range, the following is also included: Mark the current period as an abnormal period; Determine whether the number of consecutive occurrences of the abnormal cycle reaches the first cycle threshold; If so, generate an abnormal current alarm signal.

6. The condition monitoring method according to any one of claims 1 to 3, characterized in that, If the normalized feature value baseline is updated, the state monitoring method further includes: Specific feature values ​​are extracted from the current signal; Obtain a baseline for a specific feature value, which is generated based on historical data of the specific feature value; The baseline of the specific feature value in the previous period is compared with the specific feature value in the current period, and the baseline of the specific feature value is updated and / or a fault alarm signal is generated based on the comparison result.

7. The condition monitoring method according to claim 6, characterized in that, The heterogeneous nodes include laser nodes and modulator nodes; The status monitoring method includes: Acquire the first current sampling signal of the clock-synchronized laser node and the second current sampling signal of the modulator node; The first current feature value and the second current feature value are extracted from the first current sampling signal and the second current sampling signal, respectively. Temperature compensation is applied to the first current characteristic value and the second current characteristic value respectively to obtain the first normalized current characteristic value and the second normalized current characteristic value; Obtain a first normalized eigenvalue baseline and a second normalized eigenvalue baseline, wherein the first normalized eigenvalue baseline and the second normalized eigenvalue baseline are generated based on historical data of the first normalized current eigenvalue and the second normalized current eigenvalue, respectively. Compare the first normalized characteristic value baseline of the previous cycle with the first normalized current characteristic value of the current cycle, and update the first normalized characteristic value baseline and / or generate a fault alarm signal based on the comparison result. Compare the baseline of the second normalized characteristic value of the previous cycle with the second normalized current characteristic value of the current cycle, and update the baseline of the second normalized characteristic value and / or generate a fault alarm signal based on the comparison result. If both the first normalized eigenvalue baseline and the second normalized eigenvalue baseline are updated, then the peak time delay of the first current sampling signal and the second current sampling signal is calculated as the specific eigenvalue.

8. The condition monitoring method according to claim 7, characterized in that, The step of calculating the peak delay of the first current sampling signal and the second current sampling signal includes: The first current sampling signal and the second current sampling signal are respectively subjected to mean subtraction to obtain a first zero-mean sequence and a second zero-mean sequence; Calculate the cross-correlation function between the first zero-mean sequence and the second zero-mean sequence, and determine the peak delay based on the peak index and sampling period of the cross-correlation function.

9. The condition monitoring method according to claim 6, characterized in that, The heterogeneous node includes the core power rail node of the digital signal processor; The status monitoring method includes: Acquire the third current sampling signal of the core power rail node of the digital signal processor; The third current feature value is extracted from the third current sampling signal; Temperature compensation is applied to the third current characteristic value to obtain the third normalized current characteristic value; Obtain the third normalized eigenvalue baseline, which is generated based on historical data of the third normalized current eigenvalue; The baseline of the third normalized characteristic value of the previous cycle is compared with the third normalized current characteristic value of the current cycle, and the baseline of the third normalized characteristic value is updated and / or a fault alarm signal is generated based on the comparison result. If the baseline of the third normalized feature value is updated, then a fast Fourier transform is performed on the third current sampling signal to extract the current ripple energy integral in the preset high-frequency range as the specific feature value.

10. The condition monitoring method according to claim 6, characterized in that, The step of comparing the specific feature value baseline of the previous period with the specific feature value of the current period, and updating the specific feature value baseline and / or generating a fault alarm signal based on the comparison result includes: Based on the baseline of the specific characteristic value of the previous period, the tolerance range is determined; Determine whether the specific characteristic value of the current period falls within the tolerance range; If so, update the baseline of the specific feature value based on the specific feature value of the current period.

11. The condition monitoring method according to claim 10, characterized in that, The specific feature value is extracted from the first current sampling signal of the clock-synchronized laser node and the second current sampling signal of the modulator node; If the specific characteristic value of the current period does not fall within the tolerance range, the following is also included: Mark the current period as an abnormal period; Determine whether the number of consecutive occurrences of the abnormal cycle reaches the second cycle threshold; If so, generate an alarm signal for optical path abnormality.

12. The condition monitoring method according to claim 10, characterized in that, The specific feature value is extracted from the third current sampling signal of the core power rail node of the digital signal processor; If the specific characteristic value of the current period does not fall within the tolerance range, the following is also included: Obtain the specific feature value for M consecutive periods, and determine whether all of them are greater than or equal to the statistical mean of the baseline of the specific feature value; If the specific characteristic value is greater than or equal to the statistical mean for M consecutive periods, determine whether the specific characteristic value is monotonically increasing within the M consecutive periods; If the specific characteristic value increases monotonically over M consecutive periods, obtain the sampled voltage of the core power rail node and determine whether it falls within the rated voltage range. If the sampled voltage of the core power rail node falls within the rated voltage range, a capacitor aging alarm signal is generated.

13. The condition monitoring method according to claim 11, characterized in that, The generation of the optical path anomaly alarm signal also includes: Determine whether the capacitor aging alarm signal is enabled; If so, generate a power failure warning signal.

14. The condition monitoring method according to claim 1, characterized in that, The step of updating the normalized eigenvalue baseline includes: Using a preset learning rate, the normalized temperature feature value of the current period is weighted and merged with the baseline of the normalized feature value of the previous period to obtain the updated baseline of the normalized feature value. The preset learning rate is used to describe the weight of the normalized temperature feature value of the current period; or, The baseline of the normalized eigenvalues ​​for the previous period is generated based on several historical periods of temperature normalized eigenvalues ​​stored within a sliding window. The step of updating the baseline of the normalized eigenvalues ​​includes: The normalized temperature feature value of the current period is stored in the sliding window, and the earliest historical period data in the sliding window is removed to obtain the updated sliding window. Based on the updated sliding window, an updated normalized feature value baseline is generated.

15. A status monitoring device for an optical module, characterized in that, include: The feature extraction module is used to acquire the current signal of at least one heterogeneous node inside the optical module, and extract the current feature value of the current cycle from the current signal. The temperature compensation module is used to perform temperature compensation on the current characteristic value of the current cycle to obtain the temperature normalized characteristic value of the current cycle. The baseline acquisition module is used to acquire the normalized feature value baseline of the previous period, wherein the normalized feature value baseline is generated based on the historical data of the temperature normalized feature value. The monitoring module is updated to compare the normalized feature value baseline of the previous cycle with the temperature normalized feature value of the current cycle, and updates the normalized feature value baseline and / or generates a fault alarm signal based on the comparison result.

16. A status monitoring system for an optical module, characterized in that, include: The sensor is configured to sample the current signal of at least one heterogeneous node within the optical module; The controller is connected to the sensor signal; The controller includes at least one processor and a memory, the memory storing instructions that, when executed by the at least one processor, cause the at least one processor to perform the optical module status monitoring method as described in any one of claims 1 to 14. The sensor is configured to synchronously sample the current signals of the laser node and the modulator node based on a unified sampling clock; and / or, The sensor is configured to sample the current signal from the core power rail node of the digital signal processor; and / or, The sensor is configured to synchronously sample the current signals of the laser node, modulator node, and core power rail node of the digital signal processor based on a unified sampling clock.

17. An optical module, characterized in that, The status monitoring includes the optical module status monitoring system of claim 16 and / or the optical module status monitoring method of any one of claims 1 to 14.

Images