Performance test system and test method for high-precision optical fiber coupler

Abstract

The invention relates to the technical field of industrial data, and discloses a performance test system and test method for a high-precision optical fiber coupler, which can accurately evaluate the performance of the optical fiber coupler under a multi-dimensional coupling condition, and solve the problem of measurement deviation caused by insufficient parameter coverage and link interference in a traditional method. Through a mixed sampling strategy and an intelligent optimization algorithm, the test precision is ensured, the test amount is significantly reduced, and the method is suitable for a batch production scene. Besides, a variance decomposition and continuous calibration mechanism can identify key influence factors and maintain long-term test stability, a reliable basis is provided for device quality control, and it is ensured that a test result reflects a real working condition through construction of a multi-dimensional test condition set. The establishment of the mathematical model provides a quantifiable judgment standard for performance indexes, so that test data has clear physical significance and comparability.

Description

Technical Field

[0001] This invention relates to the field of industrial data technology, specifically to a performance testing system and method for high-precision fiber optic couplers. Background Technology

[0002] Fiber optic couplers, as key components in fiber optic communication systems, are widely used in optical signal splitting, combining, and power distribution. Their performance directly determines the loss level, signal equalization, and system stability of the communication link during transmission. Therefore, how to accurately and efficiently test the performance of fiber optic couplers under multi-dimensional conditions has become an important research direction in optical communication testing technology.

[0003] Most commonly used fiber optic coupler performance testing methods focus on single-polarization dimension testing, measuring parameters such as insertion loss, coupling ratio, and return loss using power meters or spectral analyzers. However, these methods have significant shortcomings: First, the tests are conducted only under limited polarization trajectories, failing to cover the multi-dimensional coupling conditions present in real-world applications, such as wavelength, temperature, mechanical stress, and incident power. Second, they lack adaptive optimization mechanisms, making it difficult to guarantee that the obtained results represent the true worst-case scenario. Third, the test data suffers from poor stability and insufficient repeatability, leading to discrepancies between the results and the device's performance in actual operating environments. These deficiencies limit the application value of existing methods in high-precision fiber optic coupler performance evaluation.

[0004] Therefore, we propose a high-precision fiber optic coupler performance testing system and method to address the aforementioned problems. Summary of the Invention

[0005] The purpose of this invention is to provide a high-precision fiber optic coupler performance testing system and method to solve the problem that the tests mentioned in the background art are only performed under limited polarization trajectories and fail to cover the multi-dimensional coupling conditions such as wavelength, temperature, mechanical stress and incident power in actual applications.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a performance testing method for high-precision fiber optic couplers, characterized by the following specific steps: S1. Determine the test parameter space of the fiber optic coupler. The parameter space shall include at least the operating wavelength, polarization state, ambient temperature, mechanical stress and incident power. Set the test range and stepping mode for each dimension and establish mathematical models for performance indicators such as insertion loss, coupling ratio, polarization-dependent loss and return loss. S2. Construct a test system including a tunable light source, polarization state generator, polarization state analyzer, power meter or spectrometer, temperature and humidity control device and mechanical stress loading device. Perform Jones or Mueller matrix modeling and de-embedding calibration on the system links to eliminate the influence of the test links on the performance of the device under test. S3. In the polarization state dimension, a spherical uniform distribution sampling method is adopted, and in the wavelength, temperature, mechanical stress and incident power dimensions, a Latin hypercube or orthogonal array sampling method is adopted to generate a test combination covering the entire parameter space, and the stabilization time and integral sampling time of each combination are set. S4. Under the condition of fixed external parameters, the polarization state is sampled traversally and the output power, spectral characteristics, Stokes parameters, temperature and stress data are collected simultaneously. The collected results are de-embedding processed based on the system matrix model to obtain the intrinsic performance data of the device under test under multidimensional conditions. S5. Establish a response surface model based on wide-coverage sampling data, use Bayesian optimization or confidence interval driven sampling strategy to search for extreme values ​​of performance indicators, and use multi-objective optimization method for multiple ports and multiple indicators until the performance extreme values ​​converge and the preset confidence requirements are met. S6. Perturbation sampling and simulation are performed on each parameter dimension near the determined extreme point to verify the repeatability and stability of the extreme value, and variance decomposition is performed to obtain the degree of influence of each parameter dimension and interaction term on the performance index. S7. Archive the test results and generate a test report containing performance extreme values, confidence intervals and coverage. Extract key test points based on the response surface model, generate a simplified test plan suitable for mass production, and establish a continuous calibration and drift monitoring mechanism.

[0007] Preferably, step S1 is performed in the following manner: S1.1. Based on the application requirements of the fiber optic coupler, determine the parameter space required for testing. The parameter space includes at least the operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power. Set the corresponding test range and stepping method for each dimension to form a complete set of multi-dimensional test conditions. S1.2. Based on the determined multidimensional test parameter space, establish a mathematical model of the performance indicators of the fiber optic coupler. The performance indicators include at least insertion loss, coupling ratio, polarization-dependent loss, and return loss. The model clarifies the correspondence between each indicator and the dimension of the parameter space, providing a judgment standard for subsequent testing and data processing.

[0008] Preferably, step S2 is performed in the following manner: S2.1 Construct a test system including a tunable light source, a polarization state generator, a polarization state analyzer, a power meter or a spectrum analyzer, a temperature and humidity control device, and a mechanical stress loading device to ensure that it can cover the multi-dimensional conditions required for fiber optic coupler performance testing. S2.2. Perform Jones or Mueller matrix modeling on the test system links, obtain the system matrix, and perform de-embedding calibration to eliminate the influence of the links on the polarization state and loss, and ensure that the test results reflect the intrinsic characteristics of the device under test.

[0009] Preferably, step S3 is implemented in the following manner: S3.1. A spherical uniform distribution sampling method is adopted in the polarization state dimension, and a Latin hypercube or orthogonal array sampling method is adopted in the wavelength, temperature, mechanical stress and incident power dimensions to generate a test combination covering the entire parameter space. S3.2 For the test combination, set the stabilization time and integral sampling time to ensure that each parameter remains stable and obtains valid data during the sampling process.

[0010] Preferably, step S4 is performed in the following manner: S4.1 Under the condition of fixed external parameters, the polarization state is sampled traversally or continuously scanned, and the output power, spectral characteristics, Stokes parameters, temperature and stress data are collected simultaneously. S4.2. Based on the system matrix model, the collected data is de-embedding processed to obtain the intrinsic performance data of the fiber coupler under multi-dimensional parameter conditions, ensuring the accuracy and consistency of the data.

[0011] Preferably, step S5 is implemented in the following manner: S5.1. Based on wide-coverage sampling data, establish a response surface model for performance indicators using Gaussian process or kernel regression methods; S5.2. Use Bayesian optimization or confidence interval-driven sampling strategies to search for extreme values ​​of performance indicators, and use multi-objective optimization methods to handle the extreme value determination of multiple ports and multiple indicators until the results converge and reach the preset confidence level.

[0012] Preferably, step S6 is implemented in the following manner: S6.1. Near the determined extreme point, perform perturbation sampling and simulation on each parameter dimension to verify the repeatability and stability of the extreme value results; S6.2. Use variance decomposition to quantify the influence of each parameter dimension and its interaction terms on performance indicators in order to identify the main factors of performance drift.

[0013] Preferably, step S7 is implemented in the following manner: S7.1 Archive the test results to form a test report that includes performance extreme values, confidence intervals, coverage and convergence curves; S7.2 Extract key test points based on response surface methodology, generate a simplified test plan suitable for mass production, and establish a continuous calibration and drift monitoring mechanism to achieve long-term quality control.

[0014] This application also includes a performance testing system for high-precision fiber optic couplers, comprising a parameter modeling module, a system calibration module, a sampling design module, a data processing module, an extreme value search module, a sensitivity analysis module, and a result solidification module; The parameter modeling module is used to determine the test parameter space of the fiber optic coupler, which includes at least the operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power; it is used to set the test range and stepping mode for each dimension; and it is used to establish mathematical models for performance indicators such as insertion loss, coupling ratio, polarization-dependent loss, and return loss. The system calibration module is used to manage and coordinate the test system hardware, including a tunable light source, polarization state generator, polarization state analyzer, power meter or spectrometer, temperature and humidity control device, and mechanical stress loading device; it is used to model the system links using Jones or Mueller matrices and perform de-embedding calibration based on the matrices to eliminate the influence of the test links on the performance of the device under test; The sampling design module is used to generate a spherically uniformly distributed set of sampling points in the polarization state dimension, and to generate sampling combinations of Latin hypercubes or orthogonal arrays in the wavelength, temperature, mechanical stress and incident power dimensions; and to set the settling time and integral sampling time for each test combination. The data processing module is used to control the PSG / PSA to perform polarization state ergodic sampling under fixed external parameters, and simultaneously acquire output power, spectral characteristics, Stokes parameters, temperature and stress data; and to perform online de-embedding processing on the acquisition results based on the Jones / Mueller matrix to obtain the intrinsic performance data of the device under test under multidimensional conditions. The extreme value search module is used to build a response surface model of performance indicators based on wide-coverage sampling data, and to perform extreme value search on performance indicators using Bayesian optimization or confidence interval driven sampling strategies; it is used to perform multi-objective optimization for multiple ports and multiple indicators until the performance extreme values ​​converge and reach the preset confidence level; The sensitivity analysis module is used to perform perturbation sampling and simulation of each parameter dimension near the determined extreme point to verify the repeatability and stability of the extreme value; and uses variance decomposition to quantify the influence of each parameter dimension and its interaction terms on the performance index. The results consolidation module is used to archive test results and generate test reports that include performance extreme values, confidence intervals, and coverage. It also extracts key test points based on the response surface model, generates a streamlined test plan suitable for mass production, and configures continuous calibration and drift monitoring strategies.

[0015] The beneficial effects of this invention are: 1. This application can accurately evaluate the performance of fiber optic couplers under multi-dimensional coupling conditions, solving the problems of insufficient parameter coverage and measurement deviations caused by link interference in traditional methods. Through a hybrid sampling strategy and intelligent optimization algorithm, it significantly reduces the amount of testing while ensuring test accuracy, making it suitable for mass production scenarios. Furthermore, variance decomposition and continuous calibration mechanisms can identify key influencing factors and maintain long-term test stability, providing a reliable basis for device quality control.

[0016] 2. This application addresses the performance evaluation bias caused by incomplete parameter coverage in traditional testing methods. By constructing a multi-dimensional set of test conditions, it ensures that the test results reflect real-world operating conditions. The establishment of the mathematical model provides quantifiable criteria for performance indicators, giving the test data clear physical meaning and comparability. The parameter stepping method improves testing efficiency, avoids redundant data collection, and lays the foundation for subsequent optimization of the test plan. Attached Figure Description

[0017] Figure 1 This is a diagram illustrating the steps of the method of the present invention.

[0018] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1: Please refer to Figure 1 A performance testing method for high-precision fiber optic couplers, the specific steps of which are as follows: S1. Determine the test parameter space of the fiber optic coupler. The parameter space should include at least the operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power. Set the test range and stepping mode for each dimension, and establish mathematical models for performance indicators such as insertion loss, coupling ratio, polarization-dependent loss, and return loss. S2. Construct a test system including a tunable light source, polarization state generator, polarization state analyzer, power meter or spectrometer, temperature and humidity control device and mechanical stress loading device. Perform Jones or Mueller matrix modeling and de-embedding calibration on the system links to eliminate the influence of the test links on the performance of the device under test. S3. In the polarization state dimension, a spherical uniform distribution sampling method is adopted, and in the wavelength, temperature, mechanical stress and incident power dimensions, a Latin hypercube or orthogonal array sampling method is adopted to generate a test combination covering the entire parameter space, and the stabilization time and integral sampling time of each combination are set. S4. Under the condition of fixed external parameters, the polarization state is sampled traversally and the output power, spectral characteristics, Stokes parameters, temperature and stress data are collected simultaneously. The collected results are de-embedding processed based on the system matrix model to obtain the intrinsic performance data of the device under test under multidimensional conditions. S5. Establish a response surface model based on wide-coverage sampling data, use Bayesian optimization or confidence interval driven sampling strategy to search for extreme values ​​of performance indicators, and use multi-objective optimization method for multiple ports and multiple indicators until the performance extreme values ​​converge and the preset confidence requirements are met. S6. Perturbation sampling and simulation are performed on each parameter dimension near the determined extreme point to verify the repeatability and stability of the extreme value, and variance decomposition is performed to obtain the degree of influence of each parameter dimension and interaction term on the performance index. S7. Archive the test results and generate a test report containing performance extreme values, confidence intervals and coverage. Extract key test points based on the response surface model, generate a simplified test plan suitable for mass production, and establish a continuous calibration and drift monitoring mechanism.

[0021] In this embodiment: In the prior art, the performance testing of fiber optic couplers typically focuses on a single polarization dimension, measuring indicators such as insertion loss and coupling ratio using power meters or spectral analyzers. However, in practical applications, the operating environment of fiber optic couplers involves the interaction of multiple dimensions, including wavelength, temperature, mechanical stress, and incident power. Existing methods struggle to comprehensively cover these complex operating conditions. Furthermore, traditional testing lacks adaptive optimization mechanisms, failing to accurately identify the worst-case scenario. The stability and repeatability of test data are insufficient, leading to discrepancies between evaluation results and actual application performance. For example, during the deployment of optical communication systems, fiber optic couplers may face dynamic environments such as temperature fluctuations or mechanical vibrations. Existing testing methods cannot effectively simulate performance changes under such multi-dimensional coupling conditions, affecting the reliability assessment of the devices.

[0022] To address the aforementioned issues, it is first necessary to overcome the limitations of single-dimensional testing and establish a testing framework covering a multi-dimensional parameter space. This involves analyzing key influencing factors in real-world application scenarios to determine core parameter dimensions such as operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power. Secondly, to mitigate the testing complexity caused by multi-dimensional parameter coupling, an efficient sampling strategy needs to be designed to reduce the amount of testing while ensuring data coverage. Furthermore, the calibration and data processing methods of the testing system need to eliminate link interference to ensure that the measurement results reflect the intrinsic characteristics of the device. Finally, an intelligent optimization algorithm is introduced to search for performance extrema and verify the stability of the results, resulting in a reusable testing scheme.

[0023] Therefore, this application proposes a high-precision performance testing method for fiber optic couplers, with the following specific steps: First, determine the test parameter space of the fiber optic coupler, including operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power; set the test range and stepping method for each dimension; establish mathematical models for insertion loss, coupling ratio, polarization-dependent loss, and return loss; construct a test system including a tunable light source, polarization state generator, polarization state analyzer, power meter or spectrometer, temperature and humidity control device, and mechanical stress loading device; perform matrix modeling and calibration of the system link to eliminate link influence; use spherical uniform distribution sampling in the polarization state dimension, and Latin hypercube or orthogonal array sampling in other dimensions to generate test combinations, setting the stabilization time and integral sampling time; traverse the polarization state and synchronously collect output data under fixed external parameter conditions, and perform de-embedding processing based on the system matrix model; establish a response surface model using wide-coverage sampling data, and perform extreme value search using Bayesian optimization or multi-objective optimization methods; verify stability by perturbation sampling near extreme points, and quantify parameter influence through variance decomposition; finally, generate a test report and extract key test points to form a simplified test scheme and continuous calibration mechanism.

[0024] The test parameter space refers to a multi-dimensional set of parameters covering the actual operating conditions of the fiber optic coupler. Specifically, it can be achieved by defining wavelength range, temperature gradient, stress loading level, and power variation range, providing a comprehensive test scenario for performance evaluation. Matrix modeling refers to mathematically describing the optical characteristics of the test link using Jones or Mueller matrices. This is achieved by measuring the optical parameters of each component of the system and constructing a transfer function, eliminating interference from the test link on polarization states and losses. Spherical uniform distribution sampling refers to a sampling method using equal angular intervals or equal area coverage in the polarization state dimension. This can be achieved by uniformly distributing points on a Poincaré sphere using a SOP generator, ensuring comprehensive polarization state coverage. Latin hypercube sampling generates a uniform and uncorrelated combination of test points in the multi-dimensional parameter space. This can be achieved by randomly arranging the partitioned intervals of each dimension and combining the sampling points, reducing the amount of testing while maintaining parameter coverage. De-embedding refers to subtracting the systematic errors of the test link from the measurement data, achieved through matrix inversion or calibration algorithms, extracting the intrinsic performance data of the device under test. Response surface methodology (RSM) refers to fitting the relationship between multidimensional parameters and performance indicators using mathematical methods. Specifically, it can be constructed using Gaussian process regression or kernel interpolation. Its purpose is to predict the performance of untested points and guide optimization. Variance decomposition (VDO) refers to analyzing the contribution of each parameter to the performance indicator using statistical methods. Specifically, it can be implemented using ANOVA or Sobol index calculations. Its purpose is to identify key influencing factors to optimize testing strategies.

[0025] This method first defines a multi-dimensional test parameter space to construct a test scenario covering actual operating conditions, addressing the insufficient parameter coverage of traditional methods. Matrix modeling and calibration of the test system eliminate link errors, ensuring that measurement data accurately reflects device characteristics. A spherical uniform distribution and Latin hypercube sampling strategy are employed to reduce the amount of testing while ensuring comprehensive coverage of the parameter space. By synchronously acquiring multi-dimensional data and performing de-embedding processing, the intrinsic performance of the device under complex conditions is obtained. Bayesian optimization based on the response surface model quickly locates performance extrema, and multi-objective optimization methods are combined to handle the trade-offs between multi-port indicators. Perturbation sampling and variance decomposition verify the stability of extrema and identify key influencing factors, ultimately generating a simplified test scheme for efficient batch testing.

[0026] Traditional methods only test under single polarization states or fixed environmental conditions, while this method comprehensively covers dynamic operating conditions in practical applications through multi-dimensional parameter space definition and hybrid sampling strategies. Existing technologies do not perform system-level calibration of the test link, resulting in measurement results containing link errors. This method effectively separates the intrinsic performance of the device through matrix modeling and de-embedding processing. Furthermore, existing tests lack intelligent optimization mechanisms, requiring numerous repeated tests to locate extrema. This method significantly improves extrema search efficiency by utilizing response surface models and Bayesian optimization.

[0027] Through the above technical solutions, this application can accurately evaluate the performance of fiber optic couplers under multi-dimensional coupling conditions, solving the problems of insufficient parameter coverage and measurement deviations caused by link interference in traditional methods. By employing a hybrid sampling strategy and intelligent optimization algorithms, the amount of testing is significantly reduced while ensuring testing accuracy, making it suitable for mass production scenarios. Furthermore, variance decomposition and continuous calibration mechanisms can identify key influencing factors and maintain long-term testing stability, providing a reliable basis for device quality control.

[0028] Example 2: Please refer to Figure 1 The specific method for step S1 is as follows: S1.1. Based on the application requirements of the fiber optic coupler, determine the parameter space required for testing. The parameter space should include at least the operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power. Set the corresponding test range and stepping method for each dimension to form a complete set of multi-dimensional test conditions. S1.2. Based on the determined multidimensional test parameter space, establish a mathematical model of the performance indicators of the fiber optic coupler. The performance indicators include at least insertion loss, coupling ratio, polarization-dependent loss, and return loss. The model clarifies the correspondence between each indicator and the dimension of the parameter space, providing a judgment standard for subsequent testing and data processing.

[0029] In this embodiment: This application further proposes to determine the test parameter space of the fiber optic coupler. The parameter space includes at least the operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power. For each dimension, a corresponding test range and stepping method are set to form a complete set of multi-dimensional test conditions. Based on the determined multi-dimensional test parameter space, a mathematical model of the fiber optic coupler performance indicators is established. The performance indicators include at least insertion loss, coupling ratio, polarization-dependent loss, and return loss. The model clarifies the correspondence between each indicator and the dimensions of the parameter space, providing a judgment standard for subsequent testing and data processing.

[0030] Operating wavelength refers to the operating wavelength range of the fiber optic coupler. This can be achieved using a tunable light source, such as one covering the C-band or L-band, to meet the needs of different communication systems. Polarization state refers to the direction of vibration of the electric field vector during light propagation. This can be simulated using a polarization state generator to cover polarization changes that may occur in practical applications. Ambient temperature refers to the external temperature conditions of the fiber optic coupler. This can be regulated using a temperature-controlled chamber, for example, by gradient changes within the range of -40℃ to 85℃. Mechanical stress refers to the physical deformation or pressure applied to the fiber optic coupler. This can be simulated using a stress loading device to simulate the stress states under different installation or usage scenarios. Incident power refers to the intensity of the optical signal input to the fiber optic coupler. This can be adjusted using an adjustable attenuator or power amplifier, covering the input range from low to high power. Test range refers to the upper and lower limits of each parameter dimension. This can be determined based on product specifications or application scenarios, for example, setting the wavelength range to 1520nm to 1620nm. The stepping method refers to the interval rule for parameter changes, which can be linear or nonlinear, such as temperature increasing in 5°C increments. The mathematical model describes the relationship between performance indicators and parameter dimensions through mathematical expressions, which can be polynomial fitting or empirical formulas to establish a quantitative correlation between insertion loss and parameters such as wavelength and temperature.

[0031] The test parameter space is determined by dividing it into multiple dimensions and setting their variation ranges and step rules, forming a multi-dimensional set of test conditions covering actual application scenarios. For example, in the wavelength dimension, a scanning range of 1520nm to 1620nm is set, with sampling in 0.1nm steps; in the temperature dimension, a test range of -40℃ to 85℃ is set, with step adjustments in 5℃ steps. The mathematical model is established by defining the mapping relationship between performance indicators and various parameters. For example, insertion loss can be expressed as a function of wavelength, temperature, and stress, providing a benchmark for the subsequent normalization processing of test data. Through the correlation analysis between the parameter space dimensions and performance indicators, the model clarifies the expected change trends of each indicator under different test conditions, thereby guiding the optimization and adjustment of the test scheme.

[0032] Traditional methods only perform limited testing on a single polarization state, while this application achieves full-coverage testing of wavelength, temperature, stress, and power by constructing a multi-dimensional parameter space. Existing technologies lack systematic parameter correlation models, while this application establishes a mathematical model to quantitatively correlate various performance indicators with test conditions, forming standardized judgment criteria. Existing technologies employ fixed step sizes or random sampling methods, while this application achieves a systematic division of the parameter space by setting the step size and test range.

[0033] Through the above technical solution, this application solves the performance evaluation bias problem caused by incomplete parameter coverage in traditional testing methods. The construction of a multi-dimensional test condition set ensures that the test results reflect real-world operating conditions. The establishment of the mathematical model provides quantifiable criteria for performance indicators, giving the test data clear physical meaning and comparability. The parameter stepping method improves testing efficiency, avoids redundant data collection, and lays the foundation for subsequent optimization of the test scheme.

[0034] Example 3: Please refer to Figure 1 The specific method for step S2 is as follows: S2.1 Construct a test system including a tunable light source, a polarization state generator, a polarization state analyzer, a power meter or a spectrum analyzer, a temperature and humidity control device, and a mechanical stress loading device to ensure that it can cover the multi-dimensional conditions required for fiber optic coupler performance testing. S2.2. Perform Jones or Mueller matrix modeling on the test system links, obtain the system matrix, and perform de-embedding calibration to eliminate the influence of the links on the polarization state and loss, and ensure that the test results reflect the intrinsic characteristics of the device under test.

[0035] In this embodiment, this application further proposes a specific implementation method for constructing a test system including a tunable light source, a polarization state generator, a polarization state analyzer, a power meter or a spectrum analyzer, a temperature and humidity control device, and a mechanical stress loading device, and performing Jones or Mueller matrix modeling and de-embedding calibration on the system links to eliminate the influence of the test links on the performance of the device under test.

[0036] Tunable light sources refer to laser emitting devices with adjustable wavelengths, specifically external cavity lasers or distributed feedback lasers, which cover the operating range of fiber couplers by adjusting the output wavelength. Polarization state generators are devices capable of generating light signals with different polarization states, specifically electro-optic modulators or combinations of rotating waveplates, used to simulate polarization changes in practical applications. Temperature and humidity control devices are devices that regulate ambient temperature and humidity, specifically in the form of constant temperature chambers or thermoelectric cooling modules, used to simulate device performance under different temperature conditions. Mechanical stress loading devices are devices that apply controllable mechanical stress, specifically in the form of piezoelectric ceramic actuators or pneumatic clamps, used to simulate the stress effects on fiber couplers during installation or use. Jones matrices or Mueller matrices are mathematical models describing the polarization characteristics of an optical system, specifically measured and calculated using vector network analyzers or polarization analyzers, used to characterize the polarization response characteristics of the test link. De-embedding calibration refers to eliminating the influence of the test link on the device under test through mathematical operations, specifically using error network models or inverse matrix operations, ensuring that the test results only reflect the intrinsic performance of the device under test.

[0037] The testing system integrates a tunable light source, a polarization state generator, and a temperature and humidity control device, enabling precise control of the wavelength, polarization state, and ambient temperature conditions of the input optical signal. A mechanical stress loading device applies axial tension or lateral pressure to the fiber optic coupler under test, simulating mechanical deformation under actual operating conditions. During the system link modeling phase, a Jones or Mueller matrix model, including the light source, connector, and testing instruments, is established by measuring the response data of a standard reference device. This model is then used to perform inverse operations on the original test data, eliminating the influence of the test link's own polarization response and loss on the measurement results. For example, during calibration, the system matrix without the device under test can be measured first, and then the systematic error components in subsequent test data can be removed through matrix inversion, thereby obtaining intrinsic parameters that only reflect the characteristics of the device under test.

[0038] Traditional testing methods often ignore the polarization response and loss effects of the test link itself, leading to systematic errors in the measurement results. For example, an uncalibrated polarization state generator may introduce additional polarization-dependent losses, and connector insertion loss fluctuations can interfere with the accurate measurement of insertion loss. This solution, however, by establishing a system matrix model and performing de-embedding calibration, effectively separates the performance parameters of the test link from those of the device under test, avoiding the superposition of systematic errors in the final results.

[0039] By employing the above technical solution, this application resolves the measurement bias problem caused by the failure to consider the polarization characteristics of the system link in traditional testing methods. Through precise modeling and calibration, it ensures that the test results only reflect the intrinsic performance parameters of the fiber coupler under test, providing an accurate data foundation for subsequent performance analysis under multi-dimensional conditions.

[0040] Example 4: Please refer to Figure 1 The specific method for step S3 is as follows: S3.1. A spherical uniform distribution sampling method is adopted in the polarization state dimension, and a Latin hypercube or orthogonal array sampling method is adopted in the wavelength, temperature, mechanical stress and incident power dimensions to generate a test combination covering the entire parameter space. S3.2 For the test combination, set the stabilization time and integral sampling time to ensure that each parameter remains stable and obtains valid data during the sampling process.

[0041] In this embodiment: This application further proposes to adopt a spherical uniform distribution sampling method in the polarization state dimension, and a Latin hypercube or orthogonal array sampling method in the wavelength, temperature, mechanical stress and incident power dimensions to generate test combinations covering the entire parameter space, and to set the stabilization time and integral sampling time of each combination.

[0042] The spherical uniform distribution sampling method refers to generating uniformly distributed polarization state points on the surface of a Poincaré sphere using mathematical methods. Specifically, it can be achieved by using Fibonacci spherical sampling or equal-angle interval sampling. Its purpose is to ensure the spatial coverage integrity and isotropy of polarization state test points, and to avoid the test blind zone caused by the limitation of polarization trajectory in traditional methods.

[0043] Latin hypercube sampling refers to the use of stratified random sampling to ensure that each parameter dimension is uniformly covered. Specifically, random permutation or optimization algorithms can be used to generate sample points. Its purpose is to achieve effective coverage of the high-dimensional parameter space with fewer test points, thereby improving testing efficiency.

[0044] Orthogonal array sampling refers to constructing test combinations with balanced distribution characteristics using orthogonal arrays. Specifically, standard or extended orthogonal arrays can be used to generate samples. Its purpose is to ensure the orthogonality between different parameter dimensions, which facilitates the separation of the influence of various factors in subsequent data analysis.

[0045] Settling time refers to the time it takes for the system to reach a steady state after the test parameters have been adjusted. Specifically, the response time of different parameter combinations can be determined through preliminary experiments. Its purpose is to eliminate the influence of transient interference on the measurement results.

[0046] The integral sampling time refers to the length of time for signal integration during data acquisition. It can be dynamically adjusted according to the noise level and signal strength. Its function is to improve the signal-to-noise ratio and suppress random errors.

[0047] In the polarization state dimension, a spherically uniform sampling distribution is employed. For example, a Fibonacci algorithm is used to generate polarization state points covering the surface of a Poincaré sphere, ensuring that each Stokes parameter direction is tested with equal probability. In the wavelength, temperature, mechanical stress, and incident power dimensions, Latin hypercube or orthogonal array sampling is used to generate multidimensional test combinations. For example, 10 equally spaced points are selected in the wavelength range, and 5 gradient points are selected in the temperature range. Combinations are generated using orthogonal arrays to cover all parameter interactions. Subsequently, a settling time is set for each test combination, for example, waiting 30 seconds after a temperature change to allow thermal equilibrium to stabilize, and the integration sampling time is set to 100 milliseconds to collect valid data.

[0048] Traditional methods typically perform polarization state testing only at fixed wavelengths and temperatures, and the use of linear scanning or random sampling leads to insufficient parameter coverage. This scheme combines spherical uniform distribution with orthogonal sampling, which significantly reduces the number of high-dimensional parameter combinations while ensuring full spatial coverage of polarization states. Furthermore, by setting the stabilization time and integral sampling time, it solves the data inconsistency problem caused by parameter fluctuations in traditional methods.

[0049] Through the above technical solutions, this application achieves efficient coverage of multi-dimensional parameters and improves data reliability in fiber optic coupler performance testing, avoiding test deviations caused by unbalanced parameter sampling or system instability, and providing a high-quality data foundation for subsequent extreme value search and performance modeling.

[0050] Example 5: Please refer to Figure 1 The specific method for step S4 is as follows: S4.1 Under the condition of fixed external parameters, the polarization state is sampled traversally or continuously scanned, and the output power, spectral characteristics, Stokes parameters, temperature and stress data are collected simultaneously. S4.2. Based on the system matrix model, the collected data is de-embedding processed to obtain the intrinsic performance data of the fiber coupler under multi-dimensional parameter conditions, ensuring the accuracy and consistency of the data.

[0051] In this embodiment: This application further proposes the following specific method for step S4: Under the condition of fixed external parameters, the polarization state is traversed and sampled or continuously scanned, and the output power, spectral characteristics, Stokes parameters, temperature and stress data are collected simultaneously; the collected data is de-embedding processed based on the system matrix model to obtain the intrinsic performance data of the fiber coupler under multi-dimensional parameter conditions, so as to ensure the accuracy and consistency of the data.

[0052] External parameter fixing refers to maintaining constant parameter values ​​in wavelength, temperature, mechanical stress, and incident power. This can be achieved by locking the temperature and humidity control device, the mechanical stress loading device, and the light source power. Its purpose is to eliminate interference from other parameters on polarization state testing. Traversal sampling or continuous trajectory scanning refers to a sampling method that covers the entire polarization state. This can be achieved by using a polarization state generator to output different polarization states at equal angular intervals or through continuous rotation. Its purpose is to fully acquire the influence of polarization state on performance indicators. Synchronous acquisition refers to recording output power, spectral characteristics, Stokes parameters, and environmental parameters at the same time reference. This can be achieved using a multi-channel data acquisition card. Its purpose is to ensure the temporal correlation of the data. De-embedding processing refers to using a system matrix model to eliminate the influence of the test link on polarization state and loss. Specifically, this can be achieved by converting the original measured values ​​into device intrinsic parameters through matrix inverse operations. Its purpose is to eliminate test system errors.

[0053] After external parameters such as wavelength, temperature, mechanical stress, and incident power are fixed, the polarization state generator generates a uniformly distributed polarization state covering the spherical surface using discrete point or continuous scanning. Simultaneously, a power meter or spectrometer records the output power and spectral characteristics of the fiber coupler in real time, while the polarization state analyzer measures Stokes parameters. Temperature and stress data are acquired via sensors. All data, after being timestamped, is input into a pre-established system matrix model for de-embedding operations, converting the raw data, which includes the influence of the test link, into a parameter set reflecting only the intrinsic performance of the fiber coupler. This process eliminates environmental fluctuations and systematic errors, ensuring the comparability of data from different batches and under different test environments.

[0054] Traditional methods, by failing to fix other external parameters when testing polarization states, allow factors such as temperature drift and changes in mechanical stress to cross-interfere with the polarization state, resulting in test results that cannot accurately reflect the influence of a single variable. This proposed solution, however, avoids multi-dimensional coupling errors at the data source through parameter isolation and synchronous acquisition mechanisms. Furthermore, existing technologies lack systematic calibration for test link errors; this solution, through matrix model embedding, removes inherent system biases at the algorithmic level, leading to more accurate extraction of intrinsic performance data.

[0055] Through the above technical solution, this application solves the problem of insufficient accuracy caused by data interference under multidimensional testing conditions, and realizes performance testing with independent polarization state dimensions and repeatability. The synergistic effect of synchronous acquisition and de-embedding processing effectively improves data consistency, providing a reliable basic dataset for subsequent extreme value search and stability verification.

[0056] Example 6: Please refer to Figure 1 The specific method for step S5 is as follows: S5.1. Based on wide-coverage sampling data, establish a response surface model for performance indicators using Gaussian process or kernel regression methods; S5.2. Use Bayesian optimization or confidence interval-driven sampling strategies to search for extreme values ​​of performance indicators, and use multi-objective optimization methods to handle the extreme value determination of multiple ports and multiple indicators until the results converge and reach the preset confidence level.

[0057] In this embodiment: This application further proposes a performance testing method for high-precision fiber optic couplers, including establishing a response surface model of performance indicators based on wide-coverage sampling data using Gaussian process or kernel regression method, performing extreme value search on performance indicators using Bayesian optimization or confidence interval driven sampling strategy, and handling the extreme value determination of multiple ports and multiple indicators through multi-objective optimization method until the results converge and reach the preset confidence level.

[0058] Gaussian process or kernel regression method refers to nonparametric regression techniques based on statistical principles. Specifically, radial basis function kernels or polynomial kernels can be used for model training to construct the mapping relationship between performance indicators and test conditions in a high-dimensional parameter space, thereby realizing the modeling of complex nonlinear relationships.

[0059] Bayesian optimization or confidence interval driven sampling strategies refer to global optimization methods based on probabilistic models. Specifically, they can employ expected improvement or confidence boundary algorithms to quickly locate extreme value regions by dynamically adjusting the sampling point positions, effectively reducing the number of invalid tests.

[0060] Multi-objective optimization methods are mathematical tools for handling trade-offs between multiple conflicting performance indicators. Specifically, Pareto front analysis can be used to select the optimal solution set by defining dominance relationships, thereby solving the trade-off problem in the performance evaluation of multi-port couplers.

[0061] Convergence and reaching the preset reliability means that the estimation error of the performance extreme value is less than the set threshold through iterative calculation. Specifically, the standard deviation or the confidence interval width can be used as the convergence criterion to ensure that the test results meet the reliability requirements.

[0062] After completing wide-coverage sampling, Gaussian process regression is used to fit the multidimensional test data to generate a response surface model of the performance index. This model can characterize the predicted performance value and its uncertainty at any point in the parameter space. Subsequently, new sampling points are generated based on the Bayesian optimization algorithm, prioritizing the region with the greatest expected improvement for supplementary testing, and gradually narrowing the extreme value search range. For multi-port devices, a multi-objective optimization method is used to transform the performance index of each port into a Pareto optimal solution set, and redundant test combinations are eliminated through iterative updates. During the testing process, the confidence interval of the extreme value estimation is continuously monitored. When the interval width is less than a preset threshold, convergence is determined, thereby significantly improving testing efficiency while ensuring the reliability of the results.

[0063] Traditional methods typically employ fixed-step traversal searches, which cannot effectively handle the combinatorial explosion problem caused by high-dimensional parameter spaces and lack quantitative analysis of conflicting relationships among multiple indicators. In contrast, this solution combines response surface modeling with an active sampling strategy to quickly identify key test areas. Simultaneously, it utilizes multi-objective optimization methods to achieve collaborative evaluation of multi-port performance, thus solving the problems of low testing efficiency and high extreme value false negative rates in existing technologies.

[0064] Through the above technical solution, this application achieves precise positioning of the performance extremes of fiber optic couplers, effectively reducing test resource consumption, and especially avoiding repetitive experiments in multi-port device testing. By dynamically adjusting the sampling strategy, it ensures that the test data can cover extreme conditions under actual operating conditions, significantly improving the reliability and applicability of the test results.

[0065] Example 7: Please refer to Figure 1 The specific method for step S6 is as follows: S6.1. Near the determined extreme point, perform perturbation sampling and simulation on each parameter dimension to verify the repeatability and stability of the extreme value results; S6.2. Use variance decomposition to quantify the influence of each parameter dimension and its interaction terms on performance indicators in order to identify the main factors of performance drift.

[0066] In this embodiment: This application further proposes to archive the test results to form a test report including performance extreme values, confidence intervals, coverage and convergence curves; extract key test points based on the response surface model to generate a simplified test plan suitable for mass production, and establish a continuous calibration and drift monitoring mechanism to achieve long-term quality control.

[0067] Test result archiving refers to storing performance data under multi-dimensional parameter conditions in a standardized format. This can be achieved using a database management system, establishing an index relationship through timestamps and parameter tags to ensure data traceability. Response surface methodology involves constructing a mapping relationship between performance indicators and multi-dimensional parameters using mathematical methods. This can be implemented using Gaussian process regression or kernel function methods to identify key regions affecting performance in the parameter space. Continuous calibration and drift monitoring mechanisms involve periodically verifying the accuracy of the test system. This can be achieved through reference device comparison or standard signal injection, correcting system deviations by establishing an error compensation model.

[0068] After completing the multidimensional parameter space testing, all test data are categorized and stored in a structured database. Simultaneously, a comprehensive report is automatically generated, including the statistical distribution of performance extreme values, confidence interval ranges, and parameter coverage. Based on the established response surface model, sensitivity analysis is used to identify key parameter combinations that significantly impact performance indicators, resulting in a streamlined scheme containing only necessary test points. During the calibration phase, system status verification is periodically performed using a standard light source and reference coupler. When a power meter reading deviation exceeds a set threshold, an automatic compensation procedure is triggered to maintain the long-term stability of the test system.

[0069] Traditional methods record only single-dimensional test data and lack structured storage, resulting in inefficient data retrieval and an inability to trace historical states. Existing testing solutions lack parameter sensitivity analysis mechanisms and still employ full parameter traversal in mass production, leading to wasted testing resources. Current calibration processes are mostly performed manually on a periodic basis, making it difficult to monitor system drift in real time.

[0070] Through the above technical solutions, this application achieves efficient management and rapid retrieval of test data, reduces testing time and equipment wear in mass production, effectively identifies the main factors of test system performance drift, and ensures the reliability and consistency of fiber optic coupler quality evaluation results in long-term use.

[0071] Example 8: Please refer to Figure 1 The specific method for step S7 is as follows: S7.1 Archive the test results to form a test report that includes performance extreme values, confidence intervals, coverage and convergence curves; S7.2 Extract key test points based on response surface methodology, generate a simplified test plan suitable for mass production, and establish a continuous calibration and drift monitoring mechanism to achieve long-term quality control.

[0072] In this embodiment: This application further proposes to archive the test results to form a test report including performance extreme values, confidence intervals, coverage and convergence curves; extract key test points based on the response surface model to generate a simplified test plan suitable for mass production, and establish a continuous calibration and drift monitoring mechanism to achieve long-term quality control.

[0073] Test result archiving refers to storing performance data obtained under multi-dimensional testing conditions in a standardized format to a database or cloud platform. This can be achieved using a distributed storage architecture combined with data encryption technology to ensure the integrity and traceability of the test data. Response surface methodology for extracting key test points involves analyzing the sensitivity of multi-dimensional parameters to performance indicators and identifying combinations of characteristic parameters that significantly impact the results. This can be achieved using principal component analysis or gradient descent methods, thereby reducing the number of testing dimensions. Continuous calibration and drift monitoring mechanisms involve establishing benchmark reference values ​​based on historical test data and tracking equipment status through periodic retesting and comparison. This can be achieved using time series analysis combined with control chart methods to promptly detect system deviations.

[0074] After de-embedding, the test data is automatically uploaded to the database according to a preset format. The system generates a unique identifier based on the timestamp and device number, and integrates the parameter combinations, confidence intervals, and convergence curve characteristics corresponding to performance extreme points into a structured report. Based on the response surface model established by Gaussian processes, the system identifies operating wavelength and temperature combinations with coupling ratio sensitivity exceeding the threshold by calculating the partial derivative matrices in each parameter direction, defining them as key test points. For mass production scenarios, the system automatically generates test sequences containing only key test points and extreme value verification points, and deploys a periodic self-check program. By comparing the standard deviation of historical calibration data with real-time measurements, the system initiates the device recalibration process when a warning threshold is triggered.

[0075] Traditional methods typically store test data in fixed-format paper reports or isolated spreadsheets, lacking unified archiving and correlation analysis capabilities. Furthermore, production test plans rely on manual experience for setup, making dynamic optimization of the number of test points impossible. This solution, through the combined application of a structured database and a response surface methodology, achieves efficient management of test data and intelligent optimization of test plans. Simultaneously, a continuous calibration mechanism effectively suppresses measurement drift caused by equipment aging.

[0076] Through the above technical solution, this application solves the problems of scattered data management, low production testing efficiency and insufficient long-term stability of traditional testing methods, realizes the systematic storage and intelligent analysis of test data, significantly improves batch testing efficiency, and ensures the consistency of measurement accuracy of the test system throughout its entire life cycle through a dynamic calibration mechanism.

[0077] This application also includes a performance testing system for high-precision fiber optic couplers, comprising a parameter modeling module, a system calibration module, a sampling design module, a data processing module, an extreme value search module, a sensitivity analysis module, and a result solidification module; The parameter modeling module is used to determine the test parameter space of the fiber optic coupler. The parameter space includes at least the operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power; it is used to set the test range and stepping mode for each dimension; and it is used to establish mathematical models for performance indicators such as insertion loss, coupling ratio, polarization-dependent loss, and return loss. The system calibration module is used to manage and coordinate the test system hardware, including tunable light source, polarization state generator, polarization state analyzer, power meter or spectrometer, temperature and humidity control device and mechanical stress loading device; it is used to perform Jones or Mueller matrix modeling on the system links and perform de-embedding calibration based on the matrix to eliminate the influence of the test links on the performance of the device under test; The sampling design module is used to generate a spherically uniformly distributed set of sampling points in the polarization state dimension, and to generate sampling combinations of Latin hypercubes or orthogonal arrays in the wavelength, temperature, mechanical stress and incident power dimensions; and to set the settling time and integral sampling time for each test combination. The data processing module is used to control the PSG / PSA to perform polarization state ergodic sampling under fixed external parameters, and simultaneously acquire output power, spectral characteristics, Stokes parameters, temperature and stress data; and to perform online de-embedding processing on the acquisition results based on the Jones / Mueller matrix to obtain the intrinsic performance data of the device under test under multidimensional conditions. The extreme value search module is used to build a response surface model of performance indicators based on wide-coverage sampling data, and to perform extreme value search on performance indicators using Bayesian optimization or confidence interval driven sampling strategies; it is used to perform multi-objective optimization for multiple ports and multiple indicators until the performance extreme values ​​converge and reach the preset confidence level; The sensitivity analysis module is used to perform perturbation sampling and simulation of each parameter dimension near the determined extreme point to verify the repeatability and stability of the extreme value; and uses variance decomposition to quantify the influence of each parameter dimension and its interaction terms on the performance index. The results consolidation module is used to archive test results and generate test reports that include performance extreme values, confidence intervals, and coverage. It also extracts key test points based on the response surface model, generates a streamlined test plan suitable for mass production, and configures continuous calibration and drift monitoring strategies.

[0078] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0079] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A performance testing method for a high-precision fiber optic coupler, characterized in that: The specific steps are as follows: S1. Determine the test parameter space of the fiber optic coupler. The parameter space shall include at least the operating wavelength, polarization state, ambient temperature, mechanical stress and incident power. Set the test range and stepping mode for each dimension and establish mathematical models for performance indicators such as insertion loss, coupling ratio, polarization-dependent loss and return loss. S2. Construct a test system including a tunable light source, polarization state generator, polarization state analyzer, power meter or spectrometer, temperature and humidity control device and mechanical stress loading device. Perform Jones or Mueller matrix modeling and de-embedding calibration on the system links to eliminate the influence of the test links on the performance of the device under test. S3. In the polarization state dimension, a spherical uniform distribution sampling method is adopted, and in the wavelength, temperature, mechanical stress and incident power dimensions, a Latin hypercube or orthogonal array sampling method is adopted to generate a test combination covering the entire parameter space, and the stabilization time and integral sampling time of each combination are set. S4. Under the condition of fixed external parameters, the polarization state is sampled traversally and the output power, spectral characteristics, Stokes parameters, temperature and stress data are collected simultaneously. The collected results are de-embedding processed based on the system matrix model to obtain the intrinsic performance data of the device under test under multidimensional conditions. S5. Establish a response surface model based on wide-coverage sampling data, use Bayesian optimization or confidence interval driven sampling strategy to search for extreme values ​​of performance indicators, and use multi-objective optimization method for multiple ports and multiple indicators until the performance extreme values ​​converge and the preset confidence requirements are met. S6. Perturbation sampling and simulation are performed on each parameter dimension near the determined extreme point to verify the repeatability and stability of the extreme value, and variance decomposition is performed to obtain the degree of influence of each parameter dimension and interaction term on the performance index. S7. Archive the test results and generate a test report containing performance extreme values, confidence intervals and coverage. Extract key test points based on the response surface model, generate a simplified test plan suitable for mass production, and establish a continuous calibration and drift monitoring mechanism.

2. The performance testing method for a high-precision fiber optic coupler according to claim 1, characterized in that: The specific method of step S1 is as follows: S1.

1. Based on the application requirements of the fiber optic coupler, determine the parameter space required for testing. The parameter space includes at least the operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power. Set the corresponding test range and stepping method for each dimension to form a complete set of multi-dimensional test conditions. S1.

2. Based on the determined multidimensional test parameter space, establish a mathematical model of the performance indicators of the fiber optic coupler. The performance indicators include at least insertion loss, coupling ratio, polarization-dependent loss, and return loss. The model clarifies the correspondence between each indicator and the dimension of the parameter space, providing a judgment standard for subsequent testing and data processing.

3. The performance testing method for a high-precision fiber optic coupler according to claim 2, characterized in that: The specific method of step S2 is as follows: S2.1 Construct a test system including a tunable light source, a polarization state generator, a polarization state analyzer, a power meter or a spectrum analyzer, a temperature and humidity control device, and a mechanical stress loading device to ensure that it can cover the multi-dimensional conditions required for fiber optic coupler performance testing. S2.

2. Perform Jones or Mueller matrix modeling on the test system links, obtain the system matrix, and perform de-embedding calibration to eliminate the influence of the links on the polarization state and loss, and ensure that the test results reflect the intrinsic characteristics of the device under test.

4. The performance testing method for a high-precision fiber optic coupler according to claim 3, characterized in that: The specific method of step S3 is as follows: S3.

1. A spherical uniform distribution sampling method is adopted in the polarization state dimension, and a Latin hypercube or orthogonal array sampling method is adopted in the wavelength, temperature, mechanical stress and incident power dimensions to generate a test combination covering the entire parameter space. S3.2 For the test combination, set the stabilization time and integral sampling time to ensure that each parameter remains stable and obtains valid data during the sampling process.

5. The performance testing method for a high-precision fiber optic coupler according to claim 4, characterized in that: The specific method of step S4 is as follows: S4.1 Under the condition of fixed external parameters, the polarization state is sampled traversally or continuously scanned, and the output power, spectral characteristics, Stokes parameters, temperature and stress data are collected simultaneously. S4.

2. Based on the system matrix model, the collected data is de-embedding processed to obtain the intrinsic performance data of the fiber coupler under multi-dimensional parameter conditions, ensuring the accuracy and consistency of the data.

6. The performance testing method for a high-precision fiber optic coupler according to claim 5, characterized in that: The specific method of step S5 is as follows: S5.

1. Based on wide-coverage sampling data, establish a response surface model for performance indicators using Gaussian process or kernel regression methods; S5.

2. Use Bayesian optimization or confidence interval-driven sampling strategies to search for extreme values ​​of performance indicators, and use multi-objective optimization methods to handle the extreme value determination of multiple ports and multiple indicators until the results converge and reach the preset confidence level.

7. The performance testing method for a high-precision fiber optic coupler according to claim 6, characterized in that: The specific method of step S6 is as follows: S6.

1. Near the determined extreme point, perform perturbation sampling and simulation on each parameter dimension to verify the repeatability and stability of the extreme value results; S6.

2. Use variance decomposition to quantify the influence of each parameter dimension and its interaction terms on performance indicators in order to identify the main factors of performance drift.

8. The performance testing method for a high-precision fiber optic coupler according to claim 7, characterized in that: The specific method of step S7 is as follows: S7.1 Archive the test results to form a test report that includes performance extreme values, confidence intervals, coverage and convergence curves; S7.2 Extract key test points based on response surface methodology, generate a simplified test plan suitable for mass production, and establish a continuous calibration and drift monitoring mechanism to achieve long-term quality control.

9. A performance testing system for high-precision fiber optic couplers, characterized in that: The high-precision fiber optic coupler performance testing system is used to execute the high-precision fiber optic coupler performance testing method according to any one of claims 1 to 8, including a parameter modeling module, a system calibration module, a sampling design module, a data processing module, an extreme value search module, a sensitivity analysis module, and a result solidification module. The parameter modeling module is used to determine the test parameter space of the fiber optic coupler, which includes at least the operating wavelength, polarization state, ambient temperature, mechanical stress, and incident power; it is used to set the test range and stepping mode for each dimension; and it is used to establish mathematical models for performance indicators such as insertion loss, coupling ratio, polarization-dependent loss, and return loss. The system calibration module is used to manage and coordinate the test system hardware, including a tunable light source, polarization state generator, polarization state analyzer, power meter or spectrometer, temperature and humidity control device, and mechanical stress loading device; it is used to model the system links using Jones or Mueller matrices and perform de-embedding calibration based on the matrices to eliminate the influence of the test links on the performance of the device under test; The sampling design module is used to generate a spherically uniformly distributed set of sampling points in the polarization state dimension, and to generate sampling combinations of Latin hypercubes or orthogonal arrays in the wavelength, temperature, mechanical stress and incident power dimensions; and to set the settling time and integral sampling time for each test combination. The data processing module is used to control the PSG / PSA to perform polarization state traversal sampling under fixed external parameters, and to simultaneously acquire output power, spectral characteristics, Stokes parameters, temperature and stress data; The acquired results are then subjected to online de-embedding processing based on the Jones / Mueller matrix to obtain intrinsic performance data of the device under test under multidimensional conditions. The extreme value search module is used to build a response surface model of performance indicators based on wide-coverage sampling data, and to perform extreme value search on performance indicators using Bayesian optimization or confidence interval driven sampling strategies; it is used to perform multi-objective optimization for multiple ports and multiple indicators until the performance extreme values ​​converge and reach the preset confidence level; The sensitivity analysis module is used to perform perturbation sampling and simulation of each parameter dimension near the determined extreme point to verify the repeatability and stability of the extreme value; Furthermore, variance decomposition was employed to quantify the impact of each parameter dimension and its interaction terms on performance indicators. The results archiving module is used to archive test results and generate test reports that include performance extreme values, confidence intervals, and coverage. Based on the response surface model, key test points are extracted to generate a simplified test plan suitable for mass production, while configuring continuous calibration and drift monitoring strategies.

Images