Passive suppression method of relative carrier envelope phase noise in dual optical comb spectral system

Abstract

This invention relates to the field of dual-comb spectroscopy technology, specifically disclosing a passive suppression method for relative carrier envelope phase noise in a dual-comb spectroscopy system. The method includes the following steps: constructing a hardware optical path; generating a dual-comb signal through a shared pump unit and a differential dispersion compensation unit; and performing physical-level phase matching using a passive optical delay matching unit; inputting the dual-comb signal into a field-programmable gate array (FPGA) core of a custom system-on-a-chip (SoC) to perform synchronous acquisition and passive phase noise suppression processing of the dual-comb signal, obtaining the suppressed signal. In actual industrial measurement processes, this invention first achieves basic phase constraint of the dual-comb signal at the physical level through shared pumping and optical delay matching, then directly performs real-time passive noise suppression within the SoC using an FPGA core, and seamlessly transfers the process to a neural network processor core for multispectral interference removal.

Description

Technical Field

[0001] This invention relates to the field of dual-comb spectroscopy technology, specifically to a passive suppression method for relative carrier envelope phase noise in a dual-comb spectroscopy system. Background Technology

[0002] Dual-comb spectroscopy systems, with their extremely high spectral resolution, ultra-wide measurement bandwidth, and extremely short acquisition time, have demonstrated enormous application potential in the field of online monitoring and precision measurement of industrial gases. Compared to traditional single-wavelength technologies such as tunable semiconductor laser absorption spectroscopy (TDLAS), dual-comb spectroscopy systems can cover a wide infrared absorption band within an extremely short measurement cycle, thereby achieving high-precision simultaneous detection of multiple target gases (such as two or more gas components in mixed industrial waste gas), avoiding the hardware complexity of building multiple monitoring systems.

[0003] However, in practical industrial gas measurement scenarios, such as monitoring chemical plant emission pipelines and online analysis of gas concentrations in high-dust workshops, dual-comb spectral systems face severe challenges in terms of measurement accuracy and stability. Among these challenges, suppressing relative carrier envelope phase noise is crucial for ensuring coherence and measurement accuracy. Existing passive suppression methods typically rely on building complex independent optical phase-locked loop devices or using external host computer software for offline data post-processing to remove phase noise and spectral interference.

[0004] Existing technologies face a conflict between high real-time compensation requirements and complex dynamic environmental interference in practical industrial applications. Specifically:

[0005] First, simple optical-physical matching cannot offset the random phase abrupt changes introduced by the dynamic environment of the industrial site (such as vibration of heavy machinery and temperature changes) in real time, resulting in broadening of the beat spectrum linewidth and reducing the extraction accuracy of gas absorption features.

[0006] Secondly, actual gas measurement sites are often accompanied by transient physical obstruction from large dust particles and broadband cross-absorption from unknown background mixed gases. Pure software algorithms relying on external computers not only suffer from severe data transmission and processing delays, making it impossible to achieve high-frequency closed-loop control and real-time removal of interference signals, but also pure mathematical filtering algorithms often fail to effectively distinguish between non-selective physical energy decay across the entire spectrum and the absorption of real chemical substances, easily leading to the misjudgment and rejection of effective spectral features.

[0007] Finally, the existing system architecture lacks a fault diagnosis and online calibration mechanism at the system level, which means that once the dual-comb gas measurement system leaves the temperature-controlled and vibration-proof laboratory environment, it cannot achieve long-term reliable operation with high precision and high stability in complex industrial sites. Summary of the Invention

[0008] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a passive suppression method for relative carrier envelope phase noise in a dual-comb spectral system, so as to achieve real-time high-precision measurement and closed-loop stable operation of the dual-comb system throughout its entire lifecycle in complex scenarios.

[0009] To achieve the above objectives, a first aspect of the present invention proposes a passive suppression method for relative carrier envelope phase noise in a dual-comb spectral system, comprising the following steps:

[0010] A hardware optical path is constructed, and a dual optical comb signal is generated by sharing a pump unit and a differential dispersion compensation unit. A passive optical delay matching unit is used to perform phase matching at the physical level.

[0011] The dual optical comb signal is input into the field-programmable gate array core of the custom system-on-a-chip to perform synchronous acquisition and passive phase noise suppression processing of the dual optical comb signal, and obtain the suppressed signal.

[0012] The suppressed signal is input into the neural network processor core of the system-on-a-chip, and the suppressed signal is processed by executing a multispectral fusion interference elimination algorithm to generate spectral measurement results;

[0013] The system-on-a-chip's security island core is used for full-link fault diagnosis, and a closed-loop self-learning mechanism is used to optimize optical path parameters. The system-on-a-chip's metrology traceability module is used to compare the spectral measurement results with the built-in reference spectrum to generate calibration coefficients to complete dynamic calibration and form a closed-loop suppression of relative carrier envelope phase noise.

[0014] To achieve the above objectives, a second aspect of the present invention proposes a passive suppression system for relative carrier envelope phase noise in a dual-comb spectral system. The system includes: a hardware optical path module and a custom system-on-a-chip, wherein:

[0015] The hardware optical path module is used to generate dual optical comb signals through a shared pump unit and a differential dispersion compensation unit, and to perform physical-level phase matching using a passive optical delay matching unit.

[0016] The customized system-on-a-chip (SoC) is communicatively connected to the hardware optical path module. The customized SoC includes at least a field-programmable gate array (FPGA) core, a neural network processor core, a security island core, and a metrology traceability module; wherein:

[0017] The field-programmable gate array core is used to receive the dual optical comb signal and perform synchronous acquisition and passive phase noise suppression processing of the dual optical comb signal to obtain the suppressed signal;

[0018] The neural network processor core is used to receive the suppressed signal, process the suppressed signal by executing a multispectral fusion interference removal algorithm, and generate spectral measurement results.

[0019] The security island core is used for full-link fault diagnosis and to optimize optical path parameters by linking a closed-loop self-learning mechanism.

[0020] The metrological traceability module is used to compare the spectral measurement results with the built-in reference spectrum, generate calibration coefficients to complete dynamic calibration, and form a closed-loop suppression of relative carrier envelope phase noise.

[0021] To achieve the above objectives, a third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the passive suppression method for relative carrier envelope phase noise in the dual-comb spectral system described above.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] This invention deeply integrates hardware optical path passive matching with chip-level heterogeneous computing. In the actual industrial field measurement process, the basic phase constraint of the dual optical comb signal is first completed at the physical level through shared pump and optical delay matching. Then, the field programmable gate array core is used directly inside the system chip to perform real-time passive noise suppression, and the neural network processor core is seamlessly handed over to perform multispectral interference removal. This effectively eliminates the dependence of traditional solutions on external large computing devices and eliminates data transmission delay.

[0024] Meanwhile, in conjunction with the full-link fault diagnosis and dynamic calibration comparison of the metrological traceability module of the safety island core, the system can achieve adaptive parameter optimization and complete online benchmark calibration when facing harsh and ever-changing physical disturbances in industrial sites. This effectively solves the problem of spectral distortion caused by dynamic environmental interference and ensures the real-time high-precision measurement and closed-loop stable operation of the dual optical comb system in complex scenarios throughout its entire life cycle. Attached Figure Description

[0025] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0026] Figure 1 This is a flowchart illustrating the passive suppression method for relative carrier envelope phase noise in the dual-comb spectral system provided by the present invention.

[0027] Figure 2This is a comparison of the power spectral density before and after phase noise suppression of the dual-comb beat frequency signal in the passive suppression method of relative carrier envelope phase noise in the dual-comb spectral system provided by the present invention.

[0028] Figure 3 This is a feature tensor heatmap before and after multispectral fusion interference removal in the passive suppression method of relative carrier envelope phase noise in the dual-comb spectral system provided by the present invention.

[0029] Figure 4 This is a graph showing the global energy time-varying gradient response under transient physical disturbances in the passive suppression method of relative carrier envelope phase noise in the dual-comb spectral system provided by this invention.

[0030] Figure 5 This is a comparison of the effective spectral absorbance recovery before and after the global compensation multiplier action in the passive suppression method of relative carrier envelope phase noise in the dual-comb spectral system provided by this invention.

[0031] Figure 6 This is a graph showing the evolution of driving current and core junction temperature during the switching process of the main and backup pump modules in the passive suppression method of relative carrier envelope phase noise in the dual optical comb spectral system provided by this invention.

[0032] Figure 7 This is a graph showing the offsetting and cancellation curves of the standby pump cross-threshold transient chirp phase jump and the feedforward correction in the passive suppression method of relative carrier envelope phase noise in the dual-comb spectral system provided by the present invention.

[0033] Figure 8 This is a schematic diagram illustrating the implementation of the passive suppression system for relative carrier envelope phase noise in the dual optical comb spectral system provided by the present invention.

[0034] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0035] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0036] The passive suppression method, system, and electronic device for relative carrier envelope phase noise in a dual-comb spectral system according to embodiments of the present invention are described below with reference to the accompanying drawings.

[0037] Example 1:

[0038] Based on practical industrial precision spectral measurement scenarios, this embodiment provides a passive suppression method for relative carrier envelope phase noise in a dual-comb spectral system. This method combines miniaturized hardware optical path design with a customized system-on-a-chip (SoC) underlying heterogeneous computing architecture, forming a complete closed loop from physical signal acquisition to low-level hardware acceleration, high-level intelligent algorithm processing, and finally to system safety management and legal metrological traceability. Specifically, the method in this embodiment includes the following steps:

[0039] Step 1: Construct the hardware optical path to complete the basic phase matching and signal generation at the physical level.

[0040] Specifically, step one includes constructing a hardware optical path, generating a dual-comb signal through a shared pump unit and a differential dispersion compensation unit, and performing physical-level phase matching using a passive optical delay matching unit. Dual-comb spectral systems have extremely high requirements for the coherence of the light source. In traditional schemes, the non-common-mode noise introduced by independent pump sources is the main cause of carrier envelope phase drift. Therefore, this scheme uses a shared pump and passive matching architecture as the signal source at the physical layer.

[0041] For example, the common pump unit in the hardware optical path uses a single-frequency semiconductor laser. Single-frequency semiconductor lasers are characterized by narrow linewidth and compact structure. In actual operation, even slight fluctuations in ambient temperature or ripple in the driving power supply can cause the laser output frequency to drift. Therefore, the pump power of the single-frequency semiconductor laser is adjusted in steps via a digital-to-analog converter (DAC), and power monitoring data is acquired through the DAC. Due to the complex industrial environment, the DAC receives digital control commands from subsequent system-on-a-chip (SoC) chips and converts them into high-precision analog drive currents to adjust the laser's lasing state with extremely small current steps. Simultaneously, the DAC acquires the photodiode current corresponding to the laser's backscattered light in real time, forming a preliminary closed loop in the physical layer to ensure that the pump source output power operates within the set rated parameter range.

[0042] The differential dispersion compensation unit in the hardware optical path employs a combination of dispersion-compensating fiber and a chirped fiber grating. In a dual-comb system, because the physical paths of the measurement and reference combs in the fiber are difficult to make perfectly symmetrical, minute differences in length or refractive index can cause different degrees of dispersion broadening in the two pulses. Using dispersion-compensating fiber alone is insufficient to accurately cancel higher-order dispersion, while the chirped fiber grating can provide a customized dispersion slope by designing a specific grating period chirp rate. This combined scheme controls the dispersion difference between the two combs within a preset dispersion threshold. This preset dispersion threshold is set according to the tolerance for pulse width in the specific application, ensuring that severe frequency domain aliasing or time domain broadening does not occur before the beat frequency interference of the dual-comb pulses.

[0043] It is also important to note that the passive optical delay matching unit includes an fiber stretcher and a piezoelectric ceramic driven mirror assembly for compensating for differential optical path delay. When the dual-comb signals interfere in spatial coincidence, extremely high-precision alignment of the two pulses on the time axis is required. The fiber stretcher provides millimeter-scale, wide-range, slow optical path delay adjustment by mechanically stretching the fiber to change its physical length and refractive index. The piezoelectric ceramic driven mirror assembly utilizes the inverse piezoelectric effect of piezoelectric materials to generate rapid nanometer-scale micro-displacements based on the applied voltage. The combination of these two components achieves composite phase compensation with a large dynamic range and high resolution, providing a solid physical signal foundation for subsequent chip-level passive suppression.

[0044] Step 2: Utilize the underlying hardware core of the system-on-a-chip (SoC) for signal acquisition and real-time passive suppression processing.

[0045] Specifically, step two includes inputting the dual optical comb signal generated by the hardware optical path into the field-programmable gate array core of the custom system-on-a-chip, performing synchronous acquisition and passive phase noise suppression processing of the dual optical comb signal, and obtaining the suppressed signal.

[0046] For example, the system-on-a-chip (SoC) architecture features multiple heterogeneous computing cores, where a synchronous data acquisition module and a real-time phase noise compensation module are deployed within the field-programmable gate array (FPGA) core for dual-channel synchronous sampling and timing jitter control. In industrial measurement, the beat frequency signal generated by dual-comb interferometry is typically in the radio frequency band, requiring the acquisition system to have extremely high sampling rates and high synchronization between channels. The synchronous data acquisition module utilizes the global clock tree distribution network within the FPGA core to ensure that the clock skew of the input signals from the primary and backup detectors approaches zero during analog-to-digital converter (ADC) sampling. The real-time phase noise compensation module, based on the principle of digital quadrature demodulation, extracts the carrier envelope phase fluctuations in the beat frequency signal and generates a digital local oscillator signal with opposite phase fluctuation trends through a direct digital frequency synthesizer. This signal is then mixed and subtracted in the digital domain to offset high-frequency random phase jitter in the physical environment in real time, outputting a suppressed signal with high phase stability.

[0047] like Figure 2 This is a comparison of the power spectral density of the dual-comb beat frequency signal before and after phase noise suppression. The horizontal axis represents the frequency shift in Hertz, and the vertical axis represents the power spectral density in dB per Hertz.

[0048] The figure contains two curves displayed on a logarithmic coordinate system, where the blue curve represents the power spectral density before suppression and the red curve represents the power spectral density after suppression.

[0049] As can be seen from the blue curve, when the dual-comb signal is processed by the field-programmable gate array core without being connected to the system-on-a-chip, there is significant phase random jitter in the low-frequency band. In the frequency offset range of 1 Hz to 100 kHz, its power spectral density value is high and shows a steep downward trend, which reflects the serious low-frequency broadening phenomenon introduced by the dynamic environment of the industrial site.

[0050] After the real-time passive suppression processing of the underlying hardware as described in this embodiment of the invention, the red curve exhibits a significant decrease and flattening waveform transformation in the control bandwidth band within 100 kHz. Its power spectral density value drops dramatically from -60 dB / Hz to around -115 dB / Hz at the 10 Hz frequency offset. This waveform transformation objectively confirms that the digital quadrature demodulation and mixing subtraction operations performed by the field-programmable gate array core can effectively offset phase drift in the low-frequency band.

[0051] The comparison of the two curves shows that the linewidth of the beat spectrum corresponding to the suppressed red curve is limited to within 100 Hz, which physically supports the system accuracy performance of a coherent accumulation time of more than one second. This confirms the effectiveness of the hardware optical path passive matching and system-on-chip heterogeneous acceleration processing mechanism of this invention in improving the actual measurement signal-to-noise ratio of the dual-comb spectral system.

[0052] Step 3: Use the neural network processor core to execute the multispectral fusion interference removal algorithm to generate spectral measurement results.

[0053] Specifically, step three involves inputting the suppressed signal into the neural network processor core of the system-on-a-chip, and processing the suppressed signal by executing a multispectral fusion interference removal algorithm to generate spectral measurement results. This step is the core data processing step of the present invention, aiming to accurately extract the background absorption spectral characteristics of the target substance from complex signals containing industrial background noise and particulate scattering interference.

[0054] For example, the neural network processor core is configured with a dedicated instruction set for quaternion processing. Traditional real-number neural networks often lose the orthogonal coupling between signal features when processing spectral interferometric signals with both amplitude and phase attributes. Quaternions, as hypercomplex numbers, can simultaneously encode the multivariate physical properties of signals in their four dimensions.

[0055] The specific execution steps of the multispectral fusion interference removal algorithm first include: mapping the suppressed signal to a quaternion space, and extracting the real and imaginary parts of the signal from the features of each input branch based on activation functions and convolution operations.

[0056] The suppressed signal feature tensor input to the neural network processor core is set as follows: The quaternion features after mapping are represented as Quaternion feature tensors contain real-part feature variables. With the imaginary part characteristic variable Feature extraction is performed using the following formula:

[0057] ;

[0058] ;

[0059] in, This represents the Gaussian error linear unit activation function, which preserves non-zero gradients in the negative region, helping to alleviate the gradient vanishing problem in deep networks when processing weak spectral features. This represents a one-dimensional convolution operation; This represents the real part of the convolution kernel weight matrix; Represents the real part bias vector; This represents the imaginary part convolution kernel weight matrix; This represents the imaginary part bias vector.

[0060] Subsequently, the extracted real feature variables With the imaginary part characteristic variable The combined data is input into a multi-scale quaternion attention module. Interference signals in industrial environments often exhibit different scale characteristics in the frequency domain. For example, baseline drift is a macroscopic, large-scale, slowly varying characteristic, while specific gas cross-absorption is a local, small-scale, sharply varying characteristic. This module extracts multi-dimensional feature information based on multiple different preset scales and concatenates the generated feature matrices to produce a concatenated multi-dimensional feature matrix.

[0061] Suppose that the preset scale set contains three different receptive field scale variables. ,variable With variables After attention-weighted extraction, feature submatrix variables of corresponding scales are generated respectively. eigenmatrix variables With eigenmatrix variables The concatenated multidimensional feature matrix variable is generated through a concatenation operation along the feature channel dimension. Its mathematical expression is:

[0062] ;

[0063] Among them, the function This indicates a tensor splicing operation.

[0064] Furthermore, the step of removing interference signals based on the stitched multidimensional feature matrix to generate the spectral measurement result includes: processing the variables of the stitched multidimensional feature matrix. Perform uniform amplitude scaling to generate band mapping matrix variables. Since the spectral energies of different wavenumber bands differ by orders of magnitude, amplitude scaling can normalize the data of all effective bands to the same dynamic range, facilitating subsequent calculations.

[0065] Next, the amplitude difference matrix and amplitude ratio matrix of the band mapping matrix of adjacent time slices are calculated to generate a direction identification matrix and an associated feature matrix indicating the direction of signal fluctuation.

[0066] Let the current time slice index variable be In its state, the band mapping matrix variables The corresponding global band index variable And the corresponding current time slice index variable The scalar magnitude element is Its previous time slice index variable The corresponding scalar magnitude element is .

[0067] Define adjacent time slice difference variables If the difference variable If the difference variable is positive, then the corresponding element in the direction identification matrix is ​​+1; If the value is -1, the corresponding element value is -1; if it remains unchanged, the value is zero. The correlation feature matrix records the ratio between the amplitude elements of the current band and the amplitude elements of the reference band.

[0068] Based on the consistency of the changing direction in the direction identification matrix and the similarity of the amplitude in the associated feature matrix, band features are aggregated to form aggregated units. In actual industrial environments, if the optical path is momentarily attenuated due to large mechanical vibrations or high-concentration dust obstruction, this physical interference is often broadband, causing a large number of continuous bands to exhibit consistent direction and highly similar amplitude attenuation ratios within the same time slice. The algorithm identifies this high degree of consistency and classifies these affected bands into an aggregated unit.

[0069] Get the preset unit weight variables Combined with the amplitude accumulation variable of the aggregation unit Average deviation variable and the largest single increment variable Calculate the interference aggregation index variable The calculation formula is as follows:

[0070] ;

[0071] in, This represents a very small positive real constant set to prevent the denominator from being zero. This interference aggregation index quantifies the probability that the current aggregation unit belongs to a non-spectral physical disturbance. When this index is large, it indicates that the aggregation change has extremely strong transient physical perturbation characteristics, rather than the slowly varying concentration absorption of the target gas.

[0072] Interference signals are eliminated based on the interference aggregation index, and the effective spectral feature matrix is ​​reconstructed, based on preset band weight variables. Calculate the spectral separation ratio variable of the effective spectral feature matrix. The effective signal purity is characterized, and the spectral measurement results are generated based on the reconstructed effective spectral feature matrix. The calculation of the spectral separation ratio is used to self-evaluate the effectiveness of the interference removal in the algorithm closed loop. If the ratio does not meet the preset purity threshold, the neural network processor core will adaptively adjust the network parameters to enter the next iteration.

[0073] like Figure 3 This is a feature tensor heatmap before and after multispectral fusion interference removal. The figure visually illustrates the data morphology transformation process before and after the neural network processor core in the system-on-a-chip executes the interference removal algorithm.

[0074] The figure contains two sub-figures, left and right. The left side (a) is the heat map before removal, and the right side (b) is the heat map after removal. The horizontal axis of both sub-figures represents the time slice index, with values ​​from 1 to 100 representing continuous sampling periods. The vertical axis represents the band index, with values ​​from 1 to 50 representing independent detection bands. Each sub-figure has a color bar on the right side representing the characteristic amplitude. The color gradient from dark blue to dark red represents the characteristic amplitude value gradually increasing from zero.

[0075] Observing the heatmap on the left before removal, it can be found that, in addition to the horizontal band of bright areas representing the true absorption characteristics of the target substance, there is an abnormal dark red bright area spanning bands 20 to 45 in the 40th to 50th time slice interval. This corresponds to the strong interference energy introduced by the sudden physical obstruction of large dust particles or the cross absorption of mixed gases in the actual measurement in the industrial field. At the same time, random high-frequency stray noise caused by the environment is scattered in the background of the entire band.

[0076] After feature extraction by the multi-scale attention module and calculation and reconstruction of the interference aggregation index, the heatmap on the right shows that the abnormal broadband burst interference block and scattered noise in the background have been effectively filtered out. The entire feature matrix has been restored to a stable low-amplitude background color, with only continuous and stable high-amplitude horizontal feature bands remaining in bands 15 to 18 and bands 35 to 38.

[0077] The striking color contrast and smooth transformation of spatial feature blocks objectively demonstrate that this scheme can effectively separate non-selective physical loss from selective chemical absorption. This ensures that even in harsh environments with external broadband physical disturbances, the system can still reconstruct and output a highly reliable effective spectral feature matrix, providing a reliable data foundation for generating high-precision spectral measurement results.

[0078] Step 4: Implement dynamic calibration of the end-to-end fault diagnosis and metrological traceability module through the security island core.

[0079] Specifically, step four includes using the security island core of the system-on-a-chip to perform full-link fault diagnosis and linking the closed-loop self-learning mechanism to optimize the optical path parameters; using the metrology traceability module of the system-on-a-chip to compare the spectral measurement results with the built-in reference spectrum, generating calibration coefficients to complete dynamic calibration, and finally forming a relative carrier envelope phase noise closed-loop suppression.

[0080] For example, in the system-on-a-chip architecture, the security island core is independent of the main computing core. It interacts with the main computing core via a secure communication bus with cyclic redundancy check (CRBC) for independent fault response. In industrial measurement and control systems, the main processor core may crash due to algorithm deadlocks or external electromagnetic interference. The security island core has independent power and clock domains, serving as the system's last line of defense, specifically responsible for real-time monitoring of the hardware and software health status. CRBC ensures that heartbeat messages or status data exchanged on the secure communication bus do not experience bit flipping.

[0081] The specific steps for end-to-end fault diagnosis in the security island core include: acquiring the real-time operating status parameters of the system and calculating the deviation variable between the real-time operating status parameters and the normal operating threshold. These real-time operating status parameters include, but are not limited to, laser temperature, photodetector bias current, chip core operating voltage, and task execution latency of each core algorithm module. Deviation variables. Multidimensional calculations are performed using Mahalanobis distance or Euclidean distance.

[0082] To achieve tiered control, a set of fault judgment thresholds is set, including a first fault threshold variable. Second fault threshold variable and the third fault threshold variable To ensure the clarity of logical progression and range, the above thresholds satisfy the following size relationship: the first fault threshold is less than the second fault threshold, and the second fault threshold is less than the third fault threshold.

[0083] When the deviation is greater than or equal to the first fault threshold and less than the second fault threshold, it is determined to be a Level 1 fault, triggering the enhanced processing mode of the multispectral fusion interference elimination algorithm and linking the closed-loop self-learning mechanism for parameter optimization. A Level 1 fault typically indicates that the system has experienced mild external environmental stress, such as a slight dispersion shift caused by a slow rise in ambient temperature, at which point the hardware structure is not damaged. The safety island core instructs the neural network processor core to increase the number of iterations, while simultaneously fine-tuning the driving parameters of the single-frequency semiconductor laser through the self-learning algorithm.

[0084] When the deviation is greater than or equal to the second fault threshold and less than the third fault threshold, it is determined to be a Level 2 fault, and the control system switches to a backup acquisition channel or backup measurement mode. A Level 2 fault indicates that the performance of a single functional module has significantly degraded or even failed, which may lead to distorted measurement data.

[0085] It is also important to note that the hardware optical path, in conjunction with the end-to-end fault diagnosis mechanism, features a physical redundancy architecture. Specifically, the output of the passive optical delay matching unit is configured with a primary and backup dual detector. The signals from the primary and backup detectors are respectively connected to the primary and backup dual acquisition channels of the system-on-a-chip (SoC). When the data deviation between the dual acquisition channels exceeds a preset deviation threshold, the SoC determines that a second-level fault has occurred and automatically switches to the backup acquisition channel. Simultaneously, the shared pump unit is configured with a primary pump module and a backup pump module. When the power fluctuation of the primary pump module exceeds a power safety threshold or a fault occurs, it automatically switches to the backup pump module within a preset switching time threshold. This physical redundancy architecture is deeply integrated with the logical judgment of the security island core, ensuring the recovery of system functionality without interrupting the measurement cycle.

[0086] When the deviation is greater than or equal to the third fault threshold, it is determined to be a level 3 fault, triggering a safety interlock output signal to disconnect the equipment power circuit and record the fault log. A level 3 fault indicates that the system has suffered severe and irreversible hardware damage, such as a broken core optical path or a short circuit in the power supply. In this case, the safety island core directly outputs a hard-wired signal through general-purpose input / output pins to cut off the system's main power supply to prevent secondary disasters.

[0087] Furthermore, to meet the requirements of legal metrology and traceability certification, the system performs dynamic calibration. The operation comparison process of the metrology traceability module includes: the metrology traceability module is equipped with a timestamp recording unit and a one-time programmable storage unit. The one-time programmable storage unit stores the chip's unique identifier and a reference spectrum database. Utilizing electronic fuse technology, the one-time programmable storage unit embeds the standard spectral characteristics certified by the National Institute of Standards and Technology into the chip silicon wafer at the time of device delivery, effectively preventing the possibility of malicious tampering at the physical level.

[0088] The spectral signals of the standard sample are periodically acquired, and the wavenumber deviation and intensity deviation of the acquired standard sample spectral signals are calculated with the corresponding spectra in the reference spectrum database to generate the calibration coefficient. The calibration coefficient is then used to correct the current spectral measurement results through multiplication.

[0089] The standard absorbance at a specific wavenumber point in the reference spectrum database is set as follows. The absorbance of the standard sample measured by the periodic acquisition module was: Then the calibration coefficient The calculation formula is:

[0090] ;

[0091] Corrected final output spectral result variables Based on the current spectral measurement results variables The calculation formula is as follows:

[0092] ;

[0093] To further ensure data traceability, the system-on-a-chip encrypts and stores calibration logs, measurement logs, fault logs, and verification logs within its internal modules. This encryption process employs advanced encryption standards (ADAS) algorithms to ensure data confidentiality. Each encrypted log entry is appended with the chip's unique identifier, a timestamp, and a cyclic redundancy check (CRC) code. Data exchange and verification with a remote platform are conducted via a network interface, facilitating remote review of the equipment's operational status and metrological legitimacy by relevant regulatory authorities.

[0094] Step 5: Coordinate the execution of the dynamic operating cycle of system closed-loop suppression.

[0095] Specifically, the relative carrier envelope phase noise closed-loop suppression of the present invention is not a single static execution, but a continuous lifecycle process. The dynamic operation cycle of the system closed-loop suppression includes: First, in the system initialization phase, after the security island core passes the self-test, the main computing core is started, and the algorithm model is loaded for the first calibration to generate initial calibration coefficients. This phase establishes the basic operating benchmark of the entire system under the current environment.

[0096] Secondly, during the real-time measurement and optimization cycle, the system simultaneously completes data acquisition, performs real-time phase noise compensation through a field-programmable gate array (FPGA) core, and eliminates interference through a neural network processor core. During data processing intervals, the system invokes the algorithm of the closed-loop self-learning mechanism to dynamically optimize the power parameters of the shared pump unit and the dispersion parameters of the differential dispersion compensation unit, forming a dynamic adaptive adjustment closed loop of measurement-compensation-evaluation-optimization.

[0097] Finally, after multi-dimensional hardware and software collaborative closed-loop suppression processing, the output beat spectrum linewidth and coherent accumulation time meet the preset accuracy conditions. In a dual-comb spectral system, a narrower beat spectrum linewidth and a longer coherent accumulation time indicate a deeper suppression of relative carrier envelope phase noise, resulting in a higher system signal-to-noise ratio and measurement sensitivity. The preset accuracy conditions set in this invention ensure that the final measured spectral resolution and gas concentration inversion accuracy meet industrial online monitoring standards.

[0098] Based on the above steps, comparing the overall technical solution of this embodiment with the prior art, it can be seen that: existing dual-comb systems generally rely on external, large discrete phase-locked loop circuits or offline software compensation algorithms from a host computer. Discrete circuits are difficult to adapt to the complex vibrations and sudden environmental changes in industrial environments; while solutions relying on host computers suffer from severe data transmission bandwidth bottlenecks and computational delays, and once the computer operating system lags, it is very easy to cause measurement interruptions or data loss.

[0099] This invention deeply integrates macroscopic passive matching of hardware optical paths with microscopic computing power from customized heterogeneous chips. In practical applications, the field-programmable gate array (FPGA) core eliminates communication latency and achieves nanosecond-level low-level phase noise offsetting; the neural network processor core utilizes quaternion space mapping and multi-scale fusion algorithms to exhibit strong robustness in handling complex background interference; and it works in conjunction with the safety island and metrological traceability module for end-to-end fault diagnosis and tamper-proof dynamic calibration closed loop.

[0100] This system enables the dual-comb spectral system to maintain long-term, stable, and legally traceable high-precision spectral measurement operation even after being removed from the laboratory constant temperature and vibration platform. It can still operate in harsh industrial environments with high dust and strong vibration, such as chemical plants and emission pipelines, effectively solving the technical pain point that traditional high-precision optical instruments are difficult to industrialize.

[0101] Example 2:

[0102] In practical industrial measurement and control applications, such as chemical plant emission pipelines, high-dust workshops, or monitoring points with heavy mechanical vibrations, the open physical optical path of a dual-comb spectral system is highly susceptible to transient physical interference of non-spectral nature. When large dust particles pass through the measurement beam instantaneously, or when the base is subjected to a sudden mechanical impact causing a minute instantaneous shift in the collimated optical path, the entire dual-comb signal will experience a synchronous and significant energy attenuation across all bands within a very short time.

[0103] If the system relies solely on conventional feature aggregation algorithms, it may misjudge this full-band physical attenuation as some unknown chemical interference with extremely strong absorption characteristics, leading to severe distortion of the interference aggregation index and ultimately causing the erroneous removal of effective spectral features. To overcome this technical deficiency, the step of interference signal removal based on the spliced ​​multidimensional feature matrix also includes a compensation mechanism for transient physical disturbances. The specific execution steps include:

[0104] Step 1: Pre-setting and calibration of the reference transparent band.

[0105] Specifically, within the bands corresponding to the spliced ​​multidimensional feature matrix, a band in which the target substance does not exhibit absorption characteristics is pre-defined as a reference transparent band. In molecular spectroscopy, each specific gas molecule, due to the differences in the vibrational and rotational energy levels of its chemical bonds, exhibits characteristic absorption peaks only within a specific infrared or near-infrared wavenumber range. Between these characteristic absorption peaks, there often exists a spectral window that exhibits high transmittance for a specific target substance.

[0106] Let the total effective band index set corresponding to the spliced ​​multidimensional feature matrix be . This set contains bands from the first band to the second band. All discrete frequency points in each band, among which These are positive integer variables. Based on prior data from the high-resolution transmission molecular absorption database, continuous wavenumber intervals are selected where, under the current temperature and pressure conditions of the measurement environment, the molecular absorption cross-sections of both the target substance and the known background gas are smaller than a minimum preset absorption threshold. The band indices corresponding to these intervals are extracted to form a set of reference transparent band indices. Clearly, the reference transparent band index set It is the total set of valid band indexes The proper subset of .

[0107] In practical system-on-a-chip (SoC) processing, the reference transparent band not only serves as an anchor point for data processing, but its physical significance lies in the fact that regardless of how drastic the concentration of the target substance fluctuates in the measurement optical path, the reference transparent band index set remains constant. The amplitude of the spectral signal in the light should always remain relatively constant. If the amplitude of these reference transparent bands decreases significantly, the only reasonable explanation from a physical mechanism is that the optical path has encountered non-selective physical obstruction or beam misalignment, rather than a change in the concentration of chemical substances. Therefore, pre-setting reference transparent bands provides a reliable physical anchor point for subsequently distinguishing between chemical interference and physical disturbances.

[0108] Step 2: Real-time calculation of global energy time-difference gradient.

[0109] For example, after generating the band mapping matrix, the rate of change of the total amplitude of all bands between adjacent time slices is calculated in real time to generate a global energy transtemporal differential gradient. The band mapping matrix variable is defined as follows: This matrix not only contains the frequency domain distribution of the spectrum, but also records its time domain state as it evolves over time.

[0110] Define the current time slice index variable as This indicates the current time slice where sampling is currently in progress; the previous time slice index variable is defined as... , representing the preceding time slice immediately adjacent to the current time slice. Let the band mapping matrix variable be... The column vector variable in the current time slice is This column vector variable contains scalar amplitude elements for all valid bands; similarly, let the band mapping matrix variable... The column vector variable in the previous time slice is Let the global band index variable be... And global band index variable Belongs to the total effective band index set .

[0111] First, calculate the total amplitude of all bands in the current time slice. Its mathematical formula is the discrete summation of the amplitudes of all effective bands:

[0112] ;

[0113] Among them, variables Represents the band mapping matrix variable The corresponding global band index variable And corresponding to the current time slice Scalar magnitude elements.

[0114] Synchronously calculate the total amplitude variable of all bands in the previous time slice :

[0115] ;

[0116] Subsequently, the rate of change of the total amplitude of all bands between adjacent time slices is calculated, thereby generating the global energy transtemporal differential gradient variable. The specific formula is as follows:

[0117] ;

[0118] Global energy time-difference gradient variable The physical significance lies in quantifying the overall energy decay rate of the entire dual-comb spectral signal within an extremely short system sampling period. When the optical path is stable and only the gas concentration changes slowly, the value of this gradient variable approaches zero; however, when mechanical vibration or particulate matter obstruction occurs, the amount of photons received will decrease sharply and instantaneously, resulting in a significant positive impulse response in this gradient variable.

[0119] like Figure 4 This is a graph showing the global energy trans-temporal differential gradient response under transient physical perturbations. The graph uses a dual-vertical axis to visually demonstrate the characteristic evolution and determination process of a dual-comb spectral system when subjected to non-spectral disturbances. The horizontal axis represents the time slice index, the left vertical axis represents the total amplitude across all bands, and the right vertical axis represents the global energy trans-temporal differential gradient.

[0120] The figure contains three key waveform curves. The blue solid line represents the evolution of the total amplitude of all bands over time. The red solid line represents the global energy differential gradient fluctuation calculated in real time by the system. The horizontal black dashed line represents the physical disturbance threshold preset by the system, with a set value of 0.3.

[0121] Observing the fluctuations in the graph, it can be found that during the first to the thirty-ninth time slice, the measurement optical path is in a relatively stable state. The blue solid line shows a steady, slight fluctuation around the value of 40, while the corresponding red solid line value remains basically around zero and is below the black dashed line.

[0122] When the system reached the 40th time slice, the simulation encountered a sudden physical disturbance such as large dust particles blocking the path or a slight mechanical deviation. At this point, the blue solid line experienced a step drop, and the total amplitude of all bands instantly fell and remained around 10. Accompanying this instantaneous decay of physical energy, the red solid line simultaneously generated a positive pulse response with an amplitude of 0.75 at the 40th time slice. This red pulse curve instantly crossed and significantly exceeded the black dashed line with a value of 0.3.

[0123] This graphical transformation objectively reflects how the present invention, by calculating the rate of change of the total amplitude of all bands between adjacent time slices, can directly convert the broadband energy decay of the physical optical path into a significant gradient pulse signal. The biaxial response curve demonstrates that comparing the global energy trans-temporal differential gradient with a preset physical perturbation threshold exhibits high sensitivity, enabling accurate identification of transient broadband physical perturbations within an extremely short system sampling period. This provides a reliable basis for the system-on-a-chip to promptly trigger the interference update freeze mechanism and construct the global compensation multiplier.

[0124] Step 3: Comparison of physical disturbance thresholds and extraction of instantaneous amplitude attenuation ratio.

[0125] It should also be noted that when the global energy time-series differential gradient is greater than the preset physical disturbance threshold, the instantaneous amplitude attenuation ratio of the reference transparent band in the current time slice is extracted.

[0126] Define physical disturbance threshold variables The physical disturbance threshold variable This is an empirical constant set by performing a long-term statistical evaluation of background noise in a undisturbed environment during the system initialization phase, combined with the expected minimum physical interference intensity in the actual industrial environment. The logic comparison unit in the system-on-a-chip performs a judgment operation in real time: if the global energy time-differential gradient variable... Not greater than the physical disturbance threshold variable If the current system is in a normal spectral fluctuation state, the system continues to execute according to the conventional aggregation logic in Example 1; if the global energy transtemporal difference gradient variable Variables greater than the physical disturbance threshold This triggers the deep feature extraction mechanism for transient physical disturbances.

[0127] After triggering the deep feature extraction mechanism, the system specifically extracts data from the reference transparent band. At this point, attenuation data from the entire band cannot be used to measure the degree of disturbance because the full-band data may contain the actual changes in the concentration of the target substance. The system calculates the instantaneous amplitude attenuation ratio of the reference transparent band in the current time slice. .

[0128] Define the reference transparent band index set The total number of discrete frequency points included is Set the reference transparent band index variable to... And the reference transparent band index variable Belongs to the reference transparent band index set Instantaneous amplitude decay ratio variable The calculation formula is:

[0129] ;

[0130] The above formula obtains a high signal-to-noise ratio pure physical optical path attenuation ratio parameter that eliminates chemical absorption interference by taking the ratio of scalar amplitude elements of each independent frequency point in the reference transparent band between adjacent time slices and calculating the arithmetic mean.

[0131] Step 4: Triggering and maintaining the state of the interference update freeze mechanism.

[0132] Optionally, when the instantaneous amplitude attenuation ratio exceeds the preset normal fluctuation tolerance range, an interference update freeze mechanism is triggered to pause the calculation and update the interference aggregation index of the current time slice.

[0133] Define the tolerance interval for normal fluctuations as a set of closed intervals, with its lower limit being the lower boundary variable of the tolerance interval. The upper limit is the variable at the upper boundary of the tolerance interval. Under normal circumstances, due to the presence of shot noise in the system or thermal noise in the detector, the instantaneous amplitude attenuation ratio variable... It will fluctuate within a tiny range around the value of one. The system-on-a-chip's comparison module determines the instantaneous amplitude decay rate variable of the current time slice. Does it fall within the normal fluctuation tolerance range? If the decay ratio variable falls below the lower boundary of the tolerance range... This indicates that the optical path has suffered a substantial loss of physical energy.

[0134] Once a substantial physical energy loss is confirmed, the control bus within the system-on-a-chip immediately issues a control command, triggering the interference update freeze mechanism. As mentioned in Example 1, the interference aggregation index is calculated iteratively based on the cumulative amplitude and average deviation. If such significant attenuation data caused by physical occlusion is input into the iterative formula, the historical cumulative parameters will be severely contaminated, leading to malfunctions in the algorithm over a prolonged period. Therefore, pausing the calculation and updating the interference aggregation index for the current time slice is equivalent to unlocking a protective lock in the algorithm's memory module.

[0135] Let the Boolean control variable indicating the current state of the disturbance aggregation index update be... When this mechanism is triggered, the control variable... It is set to logical truth. At this time, the historical cumulative matrix and deviation feature matrix stored in the high-speed static random access memory inside the system chip remain in the state of the previous time slice, and do not receive abnormal decay data of the current time slice, ensuring that the stability of the multispectral fusion interference elimination algorithm kernel is not compromised.

[0136] Step 5: Construction of global compensation multipliers and overall proportional compensation adjustment of the mapping matrix.

[0137] Specifically, at the same time, a global compensation multiplier is constructed based on the reciprocal of the instantaneous amplitude attenuation ratio, and the amplitude of the band mapping matrix of the current time slice is adjusted proportionally as a whole using the global compensation multiplier.

[0138] After freezing the update operation of the interference aggregation index, the system does not simply discard the measurement data of the current time slice. Despite the broadband physical attenuation, the relative depth information of the spectral absorption of the target substance is still contained within the attenuated spectral envelope. To salvage this frame of data and restore its effectiveness for subsequent analysis, the system needs to perform reverse compensation.

[0139] Define the global compensation multiplier variable as The system uses the instantaneous amplitude attenuation ratio variable obtained in the previous steps. Perform the reciprocal operation to construct the global compensation multiplier variable:

[0140] ;

[0141] Once the construction is complete, the tensor operation unit within the system-on-a-chip starts parallel computing mode, using the global compensation multiplier variable to perform scalar and tensor multiplication operations on the band mapping matrix of the current time slice, thereby achieving overall proportional compensation adjustment of the amplitude.

[0142] Let the column vector variable of the compensated band mapping matrix in the current time slice be... For any global band index variable in the band mapping matrix. (Including the absorption bands and other characteristic bands of the target detection substance), its compensated scalar amplitude element variable The mathematical operation logic is as follows:

[0143] ;

[0144] The physical mechanism of this adjustment process lies in the fact that photon attenuation caused by dust obstruction or mechanical micro-biasing typically exhibits a flat frequency response across the entire measurement band, meaning all bands experience loss at the same proportion. Therefore, by multiplying by the reciprocal of the instantaneous amplitude attenuation ratio, the reference energy level of the entire band mapping matrix can be forcibly raised and restored to its normal state before the physical disturbance. This rigid mathematical stretching based on a priori physical reference effectively separates non-selective physical loss from selective chemical absorption.

[0145] like Figure 5 This figure shows a comparison of the effective spectral absorbance recovery before and after the global compensation multiplier action. It visually illustrates the data restoration process when the system encounters non-spectral physical attenuation. The horizontal axis represents the band index, with values ​​from 1 to 50 representing the discrete effective bands processed by the system. The vertical axis represents the scalar amplitude elements.

[0146] The figure contains three waveform curves: the blue solid line represents the normal spectrum before the disturbance, the black solid line represents the attenuated spectrum after the disturbance, and the red dashed line represents the recovered spectrum after compensation.

[0147] As can be seen from the blue solid line, under normal conditions without interference, the reference scalar amplitude element of the spectral envelope is around the value of 1, and there are two significant absorption dip features in bands 15 to 18 and bands 35 to 38, which objectively reflects the chemical absorption state of the target substance.

[0148] When broadband physical disturbances such as dust obstruction or mechanical micro-polarization occur, as shown by the solid black line, the scalar amplitude elements of the entire band mapping matrix proportionally drop to around 0.25, at which point the original absorption concavity feature is severely compressed. The system extracts the instantaneous amplitude attenuation ratio of the reference transparent band and calculates its reciprocal to construct a global compensation multiplier of 4. This multiplier is then used to perform an overall reverse stretching of the damaged data corresponding to the solid black line, ultimately resulting in the dashed red line.

[0149] Comparing the waveforms reveals that the red dashed line and the blue solid line, representing the original normal state, essentially overlap in terms of scalar amplitude element scale and absorption pattern. The previously compressed relative absorption depth has been proportionally restored. This graphical transformation process demonstrates that this technical solution, by introducing a global compensation multiplier for overall proportional compensation adjustment, can effectively separate non-selective physical losses from selective chemical absorption. This allows the band mapping matrix, which experiences synchronous and significant attenuation under harsh operating conditions, to be restored to its normal dynamic range, preventing the true characteristic signal from being misjudged and eliminated by subsequent algorithms, and ensuring the usability of the reconstructed effective spectral feature matrix.

[0150] Step 6: Input for reconstructing the effective spectral feature matrix and subsequent execution.

[0151] For example, the compensated and adjusted band mapping matrix is ​​input into the subsequent processing flow to reconstruct the effective spectral feature matrix.

[0152] After completing the identification, state freezing, and reverse amplitude stretching of the aforementioned transient physical disturbances, the column vector variables of the repaired and compensated band mapping matrix are... The amplitude scale has maintained a high degree of consistency with the time slice data before the disturbance. The system-on-a-chip's data routing control module redirects the compensated and adjusted band mapping matrix back to the reconstruction logic branch described in Example 1.

[0153] Since the interference from the overall collapse of physical energy has been eliminated, if amplitude distortions still exist in certain specific bands in the compensated and adjusted band mapping matrix, these distortions can be identified as conventional spectral interference caused by cross-absorption of real external mixed gases or residual nonlinear phase noise in the system. The system continues to use the feature aggregation and extraction logic under normal conditions to reconstruct the effective spectral feature matrix, and uses this as the basis for outputting spectral measurement results and verifying the purity of the spectral separation ratio.

[0154] In existing technologies, signal processing algorithms for dual-comb spectral systems generally treat the acquired spectral data as a single black box input. These algorithms primarily focus on smoothing the data in the frequency or time domains using complex low-pass filtering or principal component analysis. However, in industrial applications, severe mechanical shocks can cause momentary misalignment of the optical path, or dusty gases containing lint, droplets, or other light-blocking substances can pass through the measurement beam, resulting in the system losing a large number of photons in a very short time. Existing purely mathematical filtering algorithms cannot distinguish between this full-spectrum physical energy attenuation and the intense absorption by broad-spectrum chemical substances. This often leads to abrupt changes in the system's output concentration of the measured substance without any physical basis, and in severe cases, can even cause the entire algorithm matrix to become singular and crash.

[0155] This embodiment takes a different approach, not simply increasing the mathematical complexity of the filtering algorithm, but injecting the physical absorption properties of optics as prior knowledge into the digital processing logic. By pre-setting a stable physical benchmark—the reference transparent band—and combining it with rapid early warning of the global energy transtemporal differential gradient, the system-on-a-chip can accurately identify non-spectral physical faults with extremely low computational latency.

[0156] More importantly, the solution incorporates an interference update freeze mechanism and a reverse proportional stretching operation of the global compensation multiplier. This allows the system to not only protect core historical tracking data from contamination by dirty data when subjected to transient physical damage, but also to calculate the overall energy drop ratio by the attenuation of the reference transparent band, thereby rescuing the damaged band mapping matrix and restoring it to the normal dynamic range. This technical solution, which integrates optical physical logic and underlying hardware state machine control, effectively compensates for the inherent defects of dual optical comb systems in resisting broadband physical disturbances, ensuring that the system can continuously and stably output a highly reliable effective spectral feature matrix even under harsh environmental interference.

[0157] Example 3:

[0158] In a dual-comb spectral system, the pump source is the key to the system's coherence and stability. When the system operates in industrial measurement and control environments, a main pump module and a backup pump module are typically configured to improve equipment availability. However, semiconductor lasers are extremely sensitive to changes in temperature and drive current. If a traditional hard-switching method of direct power-off and power-on is used, the backup pump module will experience a severe thermal shock when it receives its rated operating current instantaneously. This transient thermal shock can cause a significant drift in the lasing wavelength of the semiconductor laser, and may even trigger mode skipping, leading to the failure of the original phase-locked state of the dual-comb system and resulting in long-term gaps in the measurement data.

[0159] To address the practical technical problems encountered in the aforementioned industrial applications, the system-on-a-chip implements a thermal balance and crossover control mechanism for the switching process between the main pump module and the standby pump module in the shared pump unit. The specific execution steps include:

[0160] Step 1: Real-time monitoring of the main pump module's operating status and calculation of thermal load parameters.

[0161] Specifically, during normal operation of the main pump module, the system-on-a-chip (SoC) acquires the steady-state drive current and core junction temperature of the main pump module in real time, and calculates the thermal load parameters of the main pump module based on the pump source thermal impedance model. During the operation of the semiconductor laser, not all injected electrical energy is converted into optical energy output; a significant proportion is converted into heat energy due to non-radiative recombination effects and the presence of ohmic contact resistance. This heat energy directly determines the actual physical temperature of the active region inside the laser.

[0162] The steady-state drive current is set as a variable. The system-on-a-chip synchronously acquires the forward operating voltage drop across the main pump module via a high-precision analog-to-digital converter. Meanwhile, the thermistor integrated within the main pump module provides a real-time feedback signal, which the system-on-a-chip uses to determine the core junction temperature variable of the main pump module. .

[0163] By measuring the actual optical power output variable of the main pump module under the steady-state drive current. The injected electrical power variable in the current state can be calculated. The calculation formula is:

[0164] ;

[0165] Subsequently, a pump source thermal impedance model was constructed based on the law of conservation of energy, and the thermal load parameters of the main pump module were calculated. This thermal load parameter represents the actual thermal power generated inside the laser, and its calculation formula is:

[0166] ;

[0167] The above heat load parameters This reflects the continuous heat dissipation requirements of the main pump module in order to maintain the current optical power output and core junction temperature. The purpose of calculating this parameter is to establish an accurate thermal reference for the standby pump module to prevent abrupt changes in thermodynamic state during switching.

[0168] Step 2: Subthreshold thermal mirror pre-bias control of the standby pump module.

[0169] For example, after obtaining the thermal load parameters of the main pump module, the system injects a dynamic pre-bias current into the backup pump module through a digital-to-analog converter. The system controls the dynamic pre-bias current to remain below the photolasing threshold of the backup pump module to maintain a no-light output state. The system also controls the heat dissipation parameters generated by the pre-bias current to match the thermal load parameters, so that the core junction temperature difference between the backup pump module and the main pump module remains within a preset thermal balance tolerance range.

[0170] In the physical characteristics of semiconductor lasers, there exists a variable called the photolasing threshold current. When the injected current is below the laser threshold current variable At this time, spontaneous emission mainly occurs inside the laser, and no coherent stimulated emission light output is produced, i.e., it is in a state of no light output. However, spontaneous emission and nonradiative recombination of charge carriers still generate heat.

[0171] The dynamic pre-bias current injected into the standby pump module is set as a variable. And the constraint is a variable. Strictly less than the laser threshold current variable In this state, the forward voltage drop across the standby pump module is a variable. And its optical output power variable Approaching zero.

[0172] The heat dissipation parameter variable generated by the standby pump module at this time The calculation formula is:

[0173] ;

[0174] Because it is in a state of no light output, the variable Since it is considered zero, the formula for actual conversion into heat energy simplifies to:

[0175] ;

[0176] The proportional-integral-derivative (PID) controller inside the system-on-a-chip adjusts the dynamic pre-bias current variable in real time. The size of this variable causes the heat dissipation parameter of the backup pump module to change. Continuously track and equalize the thermal load parameter variables of the main pump module .

[0177] The preset thermal balance tolerance range is defined as a closed interval symmetrical about zero, with its upper limit being the upper tolerance variable. The lower limit is the tolerance lower limit variable. Under this control mechanism, the core junction temperature variable of the backup pump module... It then stabilizes. The system-on-a-chip continuously calculates the temperature difference variable. and ensure temperature difference variables It always falls within the preset thermal balance tolerance range. This process is equivalent to building a thermodynamic mirror of the main pump module in the standby pump module. Because thermal balance has been established in advance, when a switchover is required, the standby pump module is already at an equivalent operating temperature, effectively avoiding the thermal shock wave effect caused by cold start.

[0178] Step 3: Generation and execution of complementary cross-transfer control curves.

[0179] It should also be noted that when the automatic switch to the backup pump module is triggered, the system-on-a-chip synchronously outputs complementary cross-control curves. During the switchover process, the drive current of the main pump module decreases smoothly along the falling edge of the cross-control curve, while the drive current of the backup pump module increases smoothly along the rising edge of the cross-control curve and crosses the photolasing threshold until it reaches the rated operating current.

[0180] The conditions that trigger automatic switching include the output power fluctuation of the main pump module exceeding the allowable range or the occurrence of hardware anomalies such as open circuits or short circuits. If the system uses a linear decreasing and linear increasing method for current switching, the derivative of the linear function is discontinuous at its start and end nodes. This abrupt change in derivative will induce high-frequency electromagnetic transient oscillations in the parasitic capacitance and inductance of the laser, thereby interfering with the phase of the dual-comb signal. Therefore, the system uses a nonlinear smooth transition model based on the logistic function to construct complementary cross-control curves.

[0181] Set the starting physical time point of the switching process as a variable. The theoretical central physical time point of the switching process is a variable. The transient switching physical time variable is Set the rate slope coefficient of the cross transition as a variable. .

[0182] For the main pump module, the falling edge function variable of its drive current evolution over time The mathematical expression is:

[0183] ;

[0184] For the standby pump module, the rising edge function variable of its drive current evolution over time The mathematical expression is:

[0185] ;

[0186] Among them, variables This indicates that the backup pump module can provide the same rated operating current as the main pump module in terms of optical power output.

[0187] In the aforementioned complementary cross-control curves, the drive current of the main pump module decreases gradually, while the current of the standby pump module changes from the dynamic pre-bias current variable. The current rises smoothly and crosses the photolasing threshold current variable. This nonlinear control process ensures the stable transfer of the total optical power of the system, while the continuity of its derivative suppresses the generation of high-frequency current noise at the physical level.

[0188] like Figure 6 This graph shows the evolution of drive current and core junction temperature during the switching process between the primary and backup pump modules. It visually illustrates the dynamic physical process of the system-on-a-chip (SoC) executing thermal equilibrium and crossover control mechanisms. The horizontal axis represents the transient switching time in milliseconds, the left vertical axis represents the drive current in milliamperes, and the right vertical axis represents the core junction temperature difference in degrees Celsius.

[0189] The figure contains four key curves. The blue solid line represents the evolution of the main pump drive current over time, the green solid line represents the evolution of the backup pump drive current over time, the red solid line represents the core junction temperature difference between the backup pump module and the main pump module, and the two horizontal black dashed lines represent the preset thermal balance tolerance range, which is set to ±0.1 degrees Celsius.

[0190] Observing the current waveform corresponding to the left vertical axis, it can be seen that at the beginning of the switching cycle, the backup pump module has received a dynamic pre-bias current of 50 mA, putting it in a subthreshold thermal mirror state with no light output. As the switching process progresses, the blue solid line shows a smooth decreasing trend based on the logistic function, gradually decreasing from the rated operating current of 500 mA to close to 0.

[0191] Simultaneously, the green solid line smoothly rises from the 50 mA pre-bias starting point and eventually crosses to the rated operating current of 500 mA. This complementary nonlinear crossover ensures the continuity of the current derivative change. Meanwhile, observing the red solid line corresponding to the right vertical axis reveals that throughout the crossover switching, the core junction temperature difference between the primary and backup pump modules fluctuates slightly around 0 degrees Celsius and is strictly limited to the range of ±0.1 degrees Celsius within the black dashed line.

[0192] This graphical data objectively confirms that by injecting dynamic pre-bias current and cooperating with smooth crossover transition control, the backup pump module can maintain the same heat dissipation state as the main pump module, effectively suppressing the transient thermal shock phenomenon that is easy to occur under the traditional direct on / off hard switching method, thereby ensuring the smooth transition of hardware physical state and output optical power of the dual optical comb system during light source switching.

[0193] Step 4: Freezing the phase feedback control loop and loading the preset chirp compensation matrix.

[0194] Optionally, during the switching of the cross control curve, the system-on-a-chip pauses the dynamic update of the phase feedback control loop and synchronously loads the preset chirp compensation matrix associated with the backup pump module.

[0195] Under normal conditions, the dual-comb spectral system relies on the beat frequency signals acquired by the primary and backup detectors to extract the carrier envelope phase shift. An error correction is then generated using a proportional-integral-derivative feedback control algorithm and applied to the differential dispersion compensation unit to stabilize the phase. However, when the backup pump module crosses the photolasing threshold current variable... During the process of increasing the current to the rated operating current, the carrier concentration in the active region inside the device undergoes a drastic nonlinear change. This change in carrier density, through the plasma effect, causes an instantaneous change in the effective refractive index of the semiconductor medium, which in turn causes a rapid drift in the optical length of the output beam. This physical phenomenon is called the transient wavelength chirp effect.

[0196] Because the transient chirp effect caused by the injection current change occurs extremely quickly, exceeding the response bandwidth of conventional closed-loop feedback control algorithms, if the phase feedback control loop remains active during this period, the feedback system will overshoot due to the detection of a large phase change, leading to loop instability and divergence. Therefore, the system-on-a-chip (SoC) cuts off the error feedback path and suspends the dynamic update of the phase feedback control loop during this period.

[0197] To maintain phase locking, the system employs a feedforward compensation mechanism. A preset chirp compensation matrix, bound to the individual physical characteristics of a specific backup pump module, is pre-programmed into the non-volatile memory within the system-on-chip. This preset chirp compensation matrix is ​​set as a variable. This matrix records the phase shift mapping relationship caused by the backup pump module under different current gradient change rates.

[0198] Step 5: Calculation of feedforward phase correction and implementation of feedforward compensation.

[0199] Specifically, the feedforward phase correction is calculated using the preset chirp compensation matrix, and the feedforward phase correction is output to the differential dispersion compensation unit to counteract the transient phase jump generated at the moment of switching.

[0200] Transiently switch physical time variables at any point during the switching cycle. The rising edge function variable of the system-on-a-chip for the backup pump module Perform first-order time derivative to calculate the transient current rate of change at the current moment. :

[0201] ;

[0202] Subsequently, the system is based on the transient current rate of change variable at the current moment. And the current base value, through table lookup and linear interpolation algorithm, calls the preset chirp compensation matrix variable. Calculate the estimated value of the transient phase jump that is expected to occur. :

[0203] ;

[0204] Based on the estimated variable of this transient phase jump The system-on-a-chip generates feedforward phase correction variables with opposite polarities. The calculation formula is:

[0205] ;

[0206] The high-speed digital output interface of the system-on-a-chip will change the feedforward phase correction variable. This is converted into a corresponding driving voltage and applied to the miniature piezoelectric ceramic device within the differential dispersion compensation unit. Since the feedforward control does not rely on the hysteresis feedback from the sensor, the deformation response of the miniature piezoelectric ceramic device and the transient wavelength chirp process of the backup pump module maintain a high degree of synchronization on the time axis, thereby precisely offsetting and canceling the transient phase jump generated at the moment of switching in the physical optical path.

[0207] like Figure 7 This graph shows the offsetting curves between the standby pump's transient chirped phase jump across the threshold and the feedforward correction. It visually illustrates the dynamic physical process of feedforward compensation performed by the system-on-a-chip within the master / standby light source switching dead zone of a dual-comb spectral system. The horizontal axis represents the transient switching time in milliseconds, and the vertical axis represents the phase shift in radians.

[0208] The figure contains three waveform curves: the blue solid line represents the estimated transient phase jump, the red dashed line represents the feedforward phase correction, and the green solid line represents the system's residual phase error. Observing the blue solid line, it can be seen that around the 5th millisecond after the backup pump module crosses the photolasing threshold, a transient wavelength chirp effect is triggered by a drastic change in the active region carrier density, resulting in a nonlinear phase jump surge of up to +3 radians in the output beam. This value clearly exceeds the effective capture band of conventional closed-loop feedback control.

[0209] To suppress this phase abrupt change, the system-on-a-chip synchronously loads a preset chirp compensation matrix and calculates the output control command shown by the red dashed line. This red dashed line is highly synchronized with the blue solid line on the time axis and exhibits a -3 radian response with opposite polarity in amplitude, representing the feedforward physical correction action applied to the micro piezoelectric ceramic device of the differential dispersion compensation unit.

[0210] Superimposing the blue solid line (representing interference) and the red dashed line (representing compensation) on the physical optical path yields the final result shown by the green solid line. Observing the green solid line, it can be seen that the drastic phase jump at the moment of switching is effectively canceled out, and the residual phase error of the system throughout the entire switching cycle is strictly limited to a small range of ±0.1 radians, fluctuating smoothly. This timing graph objectively proves that during the transient maintenance period of pausing the dynamic update of the phase feedback control loop, the pure feedforward phase correction mechanism introduced in this invention can offset the unavoidable carrier refractive index disturbance, ensuring the stable and continuous phase-locked state of the dual-comb system during the light source takeover process.

[0211] In general, in existing redundant designs of dual-comb spectrometers, the switching of light sources mostly employs electrical relays or solid-state switches for hard cut-off and closure on a microsecond or even millisecond scale. This simple and crude physical switching mode is destructive to the low-tolerance optical comb phase-locked loop. The cold standby laser is subjected to a large current instantaneously, inevitably causing severe thermal shock and carrier oscillations. The resulting mode jumps and uncontrollable phase drifts often prevent the system from outputting effective spectral coherence data for several minutes after the switch, which defeats the purpose of the redundancy design.

[0212] This embodiment proposes a system-level control mechanism for subthreshold thermal mirror pre-biasing and feedforward chirp compensation by deeply analyzing the thermodynamic mechanism and electro-optic modulation characteristics of semiconductor lasers. In practical applications, subthreshold thermal mirror control ensures that the backup light source maintains the same thermal dissipation state as the main light source at all times, eliminating the risk of thermal shock at its source. Furthermore, the nonlinear smooth cross-transition of the logistic function suppresses radio frequency interference caused by abrupt changes in the current derivative. Finally, by introducing a pure feedforward phase correction based on matrix mapping within the dead zone of closed-loop failure, unavoidable carrier refractive index perturbations are constrained.

[0213] This integrated technical solution makes the takeover process of the main and backup light sources highly smooth and consistent, ensuring that the relative carrier envelope phase of the dual optical comb system remains stably constrained within the effective capture band of the phase-locked loop during switching, thus guaranteeing the data continuity and measurement reliability of the industrial online monitoring system when light source anomalies occur.

[0214] Example 4:

[0215] In the background technology, existing dual-comb spectral systems typically rely on externally constructed, large, and complex discrete optical phase-locked loop devices to suppress relative carrier envelope phase noise, or transmit the acquired interference data to an external general-purpose personal computer (PC) via data lines for offline algorithm compensation using host computer software. However, in actual industrial measurement and control scenarios, such as monitoring chemical plant emission pipelines and online analysis of gas concentrations in high-dust workshops, severe environmental vibrations and sudden temperature changes can introduce high-frequency random phase abrupt changes. Discrete optical devices struggle to maintain mechanical stability over long periods, while pure software algorithm solutions relying on external computers are limited by the task scheduling mechanism and communication bus bandwidth of general-purpose operating systems, resulting in significant data transmission and processing delays that cannot meet the real-time requirements of high-frequency closed-loop control. Furthermore, traditional architectures lack hardware-level fault diagnosis and online legal metrological calibration mechanisms at the device level, making the measurement accuracy of the system highly susceptible to drift in complex dynamic environments.

[0216] like Figure 8 As shown, in order to resolve the conflict between the high real-time compensation requirements and the interference of complex dynamic environments, this embodiment provides a passive suppression system for relative carrier envelope phase noise in a dual-comb spectral system. This system deeply integrates passive matching at the optical level with heterogeneous acceleration of underlying hardware at the microelectronic level.

[0217] Specifically, the passive suppression system for relative carrier envelope phase noise in a dual-comb spectral system described in this embodiment includes: a hardware optical path module and a custom system-on-a-chip. In the physical structure of this device, the optical components and the electronic control board are highly integrated into the same industrial-grade chassis with electromagnetic shielding and constant temperature control functions.

[0218] The hardware optical path module is used to generate a dual-comb signal through a shared pump unit and a differential dispersion compensation unit, and to perform physical-level phase matching using a passive optical delay matching unit. In actual equipment assembly and operation, the hardware optical path module is the physical source for generating coherent detection signals. The shared pump unit typically employs a single-frequency continuous-wave semiconductor laser with narrow linewidth characteristics, such as a distributed feedback semiconductor laser (DFB) and erbium-doped or ytterbium-doped fiber amplifier components. By physically splitting the same pump source, uncorrelated common-mode noise between different pump sources is reduced at the source. The differential dispersion compensation unit in the actual device is physically fused together with multiple segments of precisely calculated length dispersion-shifted fiber (DSF) and chirped fiber gratings (CFG) with specific periods, used for dispersion broadening of the convergent pulse during transmission on the physical optical path. The passive optical delay matching unit includes a miniature mechanical optical delay line driven by a stepper motor and a piezoelectric ceramic (PZT) mirror with a response bandwidth in the kilohertz range. This physical combination of optical components achieves basic phase difference compensation and timing alignment at the physical level when photons travel across free space or fiber optic media, providing high-quality analog signal input for subsequent digital signal processing.

[0219] The customized system-on-a-chip (SoC) is communicatively connected to the hardware optical path module. The SoC includes at least a field-programmable gate array (FPGA) core, a neural network processor core, a security island core, and a metrology traceability module. In the actual hardware circuit board design, the SoC is connected to an external high-speed analog-to-digital converter (ADC) and a high-precision digital-to-analog converter (DAC) via high-frequency microwave radio frequency traces, thereby achieving communication with the hardware optical path module. The photodetector converts the interference-generated optical signal into an analog electrical signal, which is then converted into a digital code stream by the ADC and input to the chip. This customized SoC employs a multi-core heterogeneous architecture, designed to allocate tasks with different computational characteristics to the most suitable hardware logic units.

[0220] Inside this custom system-on-a-chip (SoC): the field-programmable gate array (FPGA) core receives the dual-comb signal and performs synchronous acquisition and passive phase noise suppression processing of the dual-comb signal to obtain the suppressed signal. In industrial measurement equipment, the beat frequency signal generated by dual-comb interference often has high-frequency and massive data throughput characteristics. The FPGA core utilizes its rich internal digital signal processing (DSP) slices and high-concurrency lookup table logic to process dual-channel high-frequency sampling data with nanosecond-level extremely low deterministic latency. Compared to the traditional serial processing method based on the central processing unit (CPU) instruction set, the FPGA core directly implements digital quadrature demodulation and mixing subtraction at the hardware logic gate level. Its ability to eliminate high-frequency random phase jitter in real time avoids the phase slip phenomenon caused by operating system interrupt response, thereby outputting a phase-stable suppressed signal.

[0221] Subsequently, the data stream flows on the high-speed bus inside the chip. The neural network processor core receives the suppressed signal, processes it by executing a multispectral fusion interference removal algorithm, and generates spectral measurement results. In practical applications facing complex gas mixture cross-interference or dust obstruction, traditional linear filtering often fails to accurately extract target features.

[0222] The neural network processor (NPU) core integrates a pulsating array computing unit specifically designed for tensor operations and matrix multiplication and addition. When a suppressed signal is received, the NPU core can perform multidimensional feature matrix mapping, difference calculation, attention weight allocation, and iterative calculation of interference aggregation index with extremely high energy efficiency. This hardware-level AI computing acceleration enables the device to complete pattern recognition of complex spectral morphologies and nonlinear inversion of target gas absorbance within milliseconds, generating accurate spectral measurement results and significantly improving the device's anti-interference capability under harsh operating conditions.

[0223] To address the stringent reliability requirements of industrial environments, the Safety Island core is used for end-to-end fault diagnosis and to optimize optical path parameters through a closed-loop self-learning mechanism. In actual equipment, the Safety Island core is designed as a physically isolated microcontroller subsystem with an independent phase-locked loop clock source and power management module, meeting industrial functional safety requirements such as those of the IEC 61508 standard. It polls low-level hardware parameters in real time, including laser operating temperature, thermoelectric cooler (TEC) drive current, and detector bias voltage, via a proprietary monitoring bus within the chip. Upon detecting physical anomalies deviating from the normal operating range, such as broadband physical obstruction or sudden changes in pump source output, the Safety Island core can independently trigger the switching of backup hardware channels, initiate the self-learning mechanism to adjust laser drive parameters, or directly output a hardware interlock signal to cut off the system power circuit in the event of severe hardware damage, ensuring the inherent safety of the instrument's operation.

[0224] Finally, to meet the requirements of environmental monitoring and industrial metrology for data legality and traceability, the metrological traceability module is used to compare the spectral measurement results with the built-in reference spectrum, generate calibration coefficients to complete dynamic calibration, and form a closed-loop suppression of relative carrier envelope phase noise. In the physical implementation of the device, the metrological traceability module includes a single-use programmable storage area composed of an electronic fuse (eFuse). During factory calibration, the reference spectrum data calibrated under the standard gas cell is fixed in this physical area to ensure its immutability throughout the entire life cycle of the device. During the regular measurement cycle, the system periodically switches to the standard sample optical path, and the metrological traceability module quickly calculates the characteristic deviation between the current measured spectrum and the reference spectrum through built-in dedicated comparison logic, generates the corresponding calibration coefficients, and applies them in reverse to the final data output terminal.

[0225] In summary, the passive suppression system for relative carrier envelope phase noise in the dual-comb spectral system disclosed in this embodiment provides a highly integrated, real-time, and self-diagnostic industrial-grade intelligent optical measurement device by deeply integrating the passive matching of the hardware optical path with the heterogeneous computing power of the customized system-on-a-chip.

[0226] In practical applications, this system not only effectively eliminates the dependence of traditional high-precision dual-comb instruments on anti-vibration optical platforms and external computing control servers, but also effectively solves the problem of spectral distortion caused by dynamic environmental interference. Through the low-level high-speed phase noise suppression of the FPGA core, the intelligent anti-interference processing of the NPU core, the hardware-level security control of the safety island, and the online dynamic calibration of the metrology traceability module, this system can achieve full-link closed-loop suppression of relative carrier envelope phase noise in complex industrial environments with high dust, strong vibration, and alternating temperatures, ensuring the high precision and high stability of dual-comb spectroscopy technology in the field of industrial online monitoring over a long period of time.

[0227] Example 5:

[0228] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0229] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0230] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0231] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0232] The memory 103 stores a computer program corresponding to the passive suppression method for relative carrier envelope phase noise in the dual-comb spectral system of the above embodiments of the present invention. This computer program is executed under the control of the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0233] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0234] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A passive suppression method for relative carrier envelope phase noise in a dual-comb spectral system, characterized in that, Includes the following steps: A hardware optical path is constructed, and a dual optical comb signal is generated by sharing a pump unit and a differential dispersion compensation unit. A passive optical delay matching unit is used to perform phase matching at the physical level. The dual optical comb signal is input into the field-programmable gate array core of the custom system-on-a-chip to perform synchronous acquisition and passive phase noise suppression processing of the dual optical comb signal, and obtain the suppressed signal. The suppressed signal is input into the neural network processor core of the system-on-a-chip, and the suppressed signal is processed by executing a multispectral fusion interference elimination algorithm to generate spectral measurement results; The specific execution steps of the multispectral fusion interference removal algorithm include: mapping the suppressed signal to a quaternion space; extracting the real and imaginary parts of the signal from the features of each input branch based on activation functions and convolution operations; inputting the extracted real and imaginary feature data into a multi-scale quaternion attention module; extracting multi-dimensional feature information based on multiple different preset scales; concatenating the generated multiple feature matrices to generate a concatenated multi-dimensional feature matrix; and removing interference signals based on the concatenated multi-dimensional feature matrix to generate the spectral measurement result. The system-on-a-chip's security island core is used for full-link fault diagnosis, and a closed-loop self-learning mechanism is used to optimize optical path parameters. The system-on-a-chip's metrology traceability module is used to compare the spectral measurement results with the built-in reference spectrum to generate calibration coefficients to complete dynamic calibration and form a closed-loop suppression of relative carrier envelope phase noise.

2. The method of claim 1, wherein, The hardware optical path is configured as follows: the shared pump unit uses a single-frequency semiconductor laser, the pump power of which is adjusted in steps by a digital-to-analog converter, and power monitoring data is collected by an analog-to-digital converter; the differential dispersion compensation unit uses a combination of dispersion compensation fiber and chirped fiber grating to control the dispersion difference between the two optical combs within a preset dispersion threshold; the passive optical delay matching unit includes a fiber stretcher and a piezoelectric ceramic driven mirror group to compensate for differential optical path delay.

3. The method of claim 1, wherein, The architectural features of the system-on-a-chip include: The field-programmable gate array core is equipped with a synchronous data acquisition module and a real-time phase noise compensation module for dual-channel synchronous sampling and timing jitter control. The neural network processor core is configured with a dedicated instruction set for quaternion processing; The security island core is independent of the main computing core and interacts with the main computing core through a secure communication bus with cyclic redundancy check to perform independent fault response.

4. The method of claim 1, wherein, The step of removing interference signals based on the stitched multidimensional feature matrix to generate the spectral measurement results includes: Perform a uniform amplitude scaling mapping on the spliced ​​multidimensional feature matrix to generate a band mapping matrix; Calculate the amplitude difference matrix and amplitude ratio matrix of the band mapping matrix of adjacent time slices to generate a direction identification matrix and an associated feature matrix indicating the direction of signal fluctuation; Based on the consistency of the changing direction in the direction identification matrix and the similarity of the amplitude in the associated feature matrix, the band features are aggregated to form aggregated units, and the preset unit weights are obtained. The interference aggregation index is calculated by combining the cumulative amplitude, average deviation and maximum single increment of the aggregated units. Interference signals are eliminated based on the interference aggregation index and the effective spectral feature matrix is ​​reconstructed. The spectral separation ratio of the effective spectral feature matrix is ​​calculated based on the preset band weights to characterize the purity of the effective signal. The spectral measurement results are generated based on the reconstructed effective spectral feature matrix.

5. The method of claim 1, wherein, The specific steps for the security island core to perform end-to-end fault diagnosis include: Obtain the real-time operating status parameters of the system, and calculate the deviation between the real-time operating status parameters and the normal operating threshold; When the deviation is greater than or equal to the first fault threshold and less than the second fault threshold, it is determined to be a first-level fault, triggering the enhanced processing mode of the multispectral fusion interference elimination algorithm, and linking the closed-loop self-learning mechanism to optimize parameters. When the deviation is greater than or equal to the second fault threshold and less than the third fault threshold, it is determined to be a level 2 fault, and the control system switches to the backup acquisition channel or backup measurement mode. When the deviation is greater than or equal to the third fault threshold, it is determined to be a third-level fault, triggering a safety interlock output signal to disconnect the equipment power circuit and record the fault log. Wherein, the first fault threshold is less than the second fault threshold, and the second fault threshold is less than the third fault threshold.

6. The method of claim 5, wherein, The hardware optical path, in conjunction with the end-to-end fault diagnosis mechanism, features a physical redundancy architecture. The output of the passive optical delay matching unit is configured with a primary and backup dual detector. The signals of the primary detector and the backup detector are respectively connected to the primary and backup dual acquisition channels of the system-on-a-chip. When the data deviation between the system-on-a-chip and the dual acquisition channels exceeds a preset deviation threshold, the second-level fault is determined to have occurred and the system automatically switches to the backup acquisition channel. The shared pump unit is configured with a main pump module and a backup pump module. When the power fluctuation of the main pump module exceeds the power safety threshold or a fault occurs, it automatically switches to the backup pump module within a preset switching time threshold.

7. The method of claim 1, wherein, The operation and comparison process of the measurement traceability module includes: The metrology traceability module is equipped with a timestamp recording unit and a one-time programmable storage unit. The one-time programmable storage unit stores the chip's unique identifier and a reference spectrum database. The spectral signals of the standard sample are periodically acquired, and the wavenumber deviation and intensity deviation of the acquired standard sample spectral signals are calculated with the corresponding spectra in the reference spectrum database to generate the calibration coefficient. The calibration coefficient is then used to correct the current spectral measurement results through multiplication.

8. The method of claim 7, wherein, The system-on-a-chip encrypts the calibration log, measurement log, fault log, and verification log and stores them in its internal modules. Each encrypted log entry is appended with the chip's unique identifier, a timestamp, and a cyclic redundancy check code, and is then exchanged and verified with a remote platform via a network interface.

9. The method of claim 1, wherein, The dynamic operating period of the relative carrier envelope phase noise closed-loop suppression includes: During the system initialization phase, after the security island core passes the self-test, the main computing core is started, and the algorithm model is loaded for the first calibration to generate the initial calibration coefficients. During the real-time measurement and optimization cycle, data acquisition, real-time phase noise compensation and interference removal are completed simultaneously, and the algorithm of the closed-loop self-learning mechanism is called to dynamically optimize the power parameters of the common pump unit and the dispersion parameters of the differential dispersion compensation unit. After closed-loop suppression processing, the system output beat spectrum linewidth and coherent accumulation time meet the preset accuracy conditions.

Images