An optical module test process credibility evaluation method and system

Abstract

The present application relates to the technical field of optical communication test, and specifically relates to a kind of optical module test process credibility evaluation method and system, comprising: obtaining optical module test process operating parameter time series and carrying out denoising preprocessing;Operating parameter time series is carried out wavelet transform abnormal fluctuation detection and obtains candidate fluctuation time, and effective fluctuation time is determined based on multidimensional joint confirmation rule;With effective fluctuation time as origin, the response characteristic parameter of each dimension is extracted;Response vector is constructed based on the response characteristic parameter of each dimension and multidimensional perturbation response description is generated;According to pre-set physical type division rule, the type norm of each physical type is calculated respectively and is synthesized into comprehensive influence index;Comprehensive influence index is compared with pre-set credibility threshold value, and test credibility evaluation result is output.The present application carries out cooperative abnormal fluctuation detection within specification threshold value and type quantitative evaluation based on physical nature, and the accuracy of optical module test result reliability judgment is improved from root.

Description

Technical Field

[0001] This invention relates to the field of optical communication testing technology, and specifically to a method and system for evaluating the reliability of optical module testing processes. Background Technology

[0002] Performance testing of optical modules is a core aspect of ensuring product quality. In R&D verification and production line testing, it is usually necessary to simultaneously monitor multiple key operating parameters of the optical module under test, including optical power, electrical signal amplitude, bit error rate, eye height, eye width, jitter, and extinction ratio. In current testing practices, quality control of the testing process mainly relies on threshold determination.

[0003] However, within the specified thresholds, all indicators of the optical module did not exceed the specified thresholds. The transient fluctuations in the measured parameters were caused by minor vibrations in the testing environment, instantaneous fluctuations in the supply voltage, short-term drifts in ambient temperature, and possibly intermittent failures of poorly soldered joints under vibration or early degradation of the laser driver circuit. Since the parameters did not exceed the specified thresholds, the system automatically determined the test results to be reliable. However, if the fluctuations originated from potential defects in the optical module itself, the test data for this batch of modules was actually collected under abnormal conditions, raising questions about its long-term reliability and potentially leading to intermittent failures during customer use. Conversely, if the fluctuations originated from external environmental disturbances and had no impact on the module's performance, retesting would result in unnecessary production waste. Summary of the Invention

[0004] To address the technical problem of quantitatively evaluating the reliability of test results based on the multi-dimensional fluctuation coupling characteristics of parameters during the testing process, when all multi-dimensional test parameters of optical modules do not exceed the specification thresholds, the present invention aims to provide a method and system for evaluating the reliability of optical module testing processes. The specific technical solution adopted is as follows:

[0005] Firstly, a method for evaluating the credibility of an optical module testing process is provided. This method includes: acquiring time series of operating parameters across multiple dimensions during the optical module testing process, and performing noise reduction preprocessing on each operating parameter time series; performing wavelet transform anomaly detection on each operating parameter time series to obtain candidate fluctuation moments for each dimension, and determining valid fluctuation moments from the candidate fluctuation moments based on a multi-dimensional joint confirmation rule; extracting response feature parameters for each dimension within a preset observation window, using the valid fluctuation moment as the origin, the response feature parameters including response delay, normalized response peak value, and recovery time; constructing a fluctuation response vector based on the response feature parameters for each dimension to generate a multi-dimensional disturbance response description; dividing the parameter dimensions into multiple physical types according to a preset physical type classification rule, calculating the type norm based on the fluctuation response vector corresponding to each physical type, and synthesizing a comprehensive impact index of the fluctuation event; comparing the comprehensive impact index with a preset credibility threshold, and outputting the test credibility evaluation result.

[0006] In one possible design, the time series of operating parameters in multiple dimensions during the optical module testing process are acquired, and noise reduction preprocessing is performed on each operating parameter time series. This includes: connecting a high-speed oscilloscope, a bit error rate analyzer, and an optical power meter through an optical module testing fixture, and synchronously acquiring multiple operating parameters in multiple dimensions at a preset sampling frequency to form a multi-dimensional operating parameter time series. The multiple dimensions include optical power, electrical signal amplitude, bit error rate count, eye height, eye width, random jitter, deterministic jitter, and extinction ratio. The time series of each operating parameter is then subjected to noise reduction preprocessing using a median filter. The filtering window length for continuous analog parameters is 5 to 9 sampling points, the filtering window length for bit error rate count is 3 sampling points, and the filtering window length for jitter parameters is 5 to 9 sampling points.

[0007] In one possible design, wavelet transform is used to detect abnormal fluctuations in the time series of various operating parameters to obtain candidate fluctuation moments for each dimension. This includes: according to the preset mapping rules between operating parameters and wavelet detection configuration, performing a first preset number of discrete wavelet transforms on continuous simulation parameters using wavelet bases with preset regularity and tight support characteristics; and performing a second preset number of discrete wavelet transforms on bit error counts and jitter parameters using wavelet bases sensitive to step signals to obtain detail coefficients for each layer. The modulus of each layer detail coefficient is calculated, and local maxima of the modulus are detected at each scale. The modulus maxima corresponding to the same event are associated in the spatial neighborhood of adjacent scales along the scale direction to form a cross-scale modulus maxima line. For each modulus maxima line, the decay slope of the modulus with the number of decomposition layers is fitted, the Lipschitz regularity index at the corresponding time position is calculated, and moments with regularity indices less than a preset singularity threshold are marked as preliminary candidate fluctuation moments for that dimension.

[0008] In one possible design, the method further includes performing a second-level verification on the initial candidate fluctuation moments in the bit error count dimension. Specifically, this involves: determining a stable segment in the entire test sequence where no candidate fluctuation moments are detected as a global reference segment; dividing the sequence into windows according to a preset sampling window length; counting the cumulative bit error counts within each window; calculating the mean and standard deviation; for each initial candidate fluctuation moment, extracting the bit error count data from multiple consecutive sampling windows following it; counting the total cumulative bit error counts within each window; retaining the candidate fluctuation moment if the total cumulative bit error count within any sampling window after the candidate moment is greater than the mean plus three times the standard deviation; otherwise, discarding it.

[0009] In one possible design, based on a multi-dimensional joint confirmation rule, the effective fluctuation time is determined from the candidate fluctuation times, including: setting a time tolerance window, which is determined according to the sampling frequency and the response time difference of each dimension parameter; when at least three dimension parameter sequences simultaneously have candidate fluctuation times within the time tolerance window, and the at least three dimensions contain at least one continuous analog parameter, the time position is determined as an effective fluctuation time; wherein, the continuous analog parameter includes optical power, electrical signal amplitude, eye height, eye width, and extinction ratio.

[0010] In one possible design, taking the effective fluctuation moment as the origin, response feature parameters of each dimension are extracted within a preset observation window, including: taking a period of stable data in the initial stage of the test where no candidate fluctuation moment is detected, and using the mean of each dimension parameter in this period as the baseline level; taking data of a first preset duration starting from the effective fluctuation moment as the response observation interval, and calculating the deviation of each dimension data from the baseline level as the response waveform; taking the number of sampling points experienced from the effective fluctuation moment to the response waveform reaching the maximum absolute deviation within the observation window as the response delay; normalizing the maximum absolute value of the response waveform within the observation window to obtain the response peak value, wherein, for continuous analog parameters with a non-zero stable baseline level, baseline relative value normalization is used, and for bit error count values ​​and jitter parameters with a zero baseline level, normalization is performed using the ratio of the absolute change to the upper limit of the parameter specification; taking the number of sampling points required for the absolute value of the response waveform to drop from the peak value to within a preset tolerance range near the baseline level as the recovery time.

[0011] In one possible design, a fluctuating response vector is constructed based on the response characteristic parameters of each dimension to generate a multi-dimensional perturbation response description, including: combining the response delay, response peak, and recovery time of each dimension into a fluctuating response vector; sorting the response delay of each dimension from smallest to largest to generate a response time series; arranging the response peak of each dimension according to the response time series to form an intensity evolution sequence; and arranging the recovery time of each dimension from shortest to longest to form a recovery time difference sequence; wherein, dimensions whose recovery time exceeds the observation window are marked as persistent responses, and are uniformly treated as having the longest recovery time and placed at the end of the sequence.

[0012] In one possible design, the preset physical type classification rules include continuous power type, zero baseline jitter type, and stepped cumulative type. Continuous power type corresponds to optical power, electrical signal amplitude, eye height, eye width, and extinction ratio; zero baseline jitter type corresponds to random jitter and deterministic jitter; and stepped cumulative type corresponds to bit error count. The calculation of the type norm based on the fluctuation response vector corresponding to each physical type includes: filtering effective response dimensions, where the effective response dimension is the dimension whose candidate fluctuation moment falls within the valid fluctuation moment confirmation criteria and whose maximum absolute value of the response waveform is not less than the recovery criterion threshold of that dimension; dividing the effective response dimensions into three categories according to the preset physical type classification rules, and calculating the type norm for each category, consisting of response delay, response peak value, and recovery time. Specifically, for the zero baseline jitter type parameters, the response peak value is included in the calculation after introducing a preset small signal amplification factor; and for the stepped cumulative type parameters, the recovery time is included in the calculation after introducing a preset recovery time weighting factor.

[0013] In one possible design, the comprehensive impact index of the synthetic fluctuation event includes: summing the norms of each type after normalizing them according to their respective number of dimensions, and multiplying the sum by the proportion of the effective response dimension in the total number of parameter dimensions to obtain the comprehensive impact index; wherein, for the persistent response dimension whose recovery time exceeds the observation window, the recovery time component in the type norm is taken as the number of sampling points corresponding to the first preset duration; the credibility threshold is determined by statistical analysis of the comprehensive impact index calculated from known normal test data.

[0014] Secondly, a reliability assessment system for optical module testing processes is provided, comprising: a parameter acquisition and preprocessing unit, used to acquire time series of operating parameters in multiple dimensions during optical module testing, and to perform noise reduction preprocessing on each operating parameter time series; a fluctuation detection and confirmation unit, used to perform wavelet transform anomaly fluctuation detection on each operating parameter time series, obtain candidate fluctuation moments in each dimension, and determine the valid fluctuation moment from the candidate fluctuation moments based on multi-dimensional joint confirmation rules; a response feature extraction unit, used to extract response feature parameters in each dimension within a preset observation window, with the valid fluctuation moment as the origin, the response feature parameters including response delay, normalized response peak value, and recovery time; a disturbance response description generation unit, used to construct a fluctuation response vector based on the response feature parameters in each dimension, and generate a multi-dimensional disturbance response description; and a comprehensive impact index calculation unit, used to divide the parameter dimensions into multiple physical types according to preset physical type classification rules, calculate the type norm based on the fluctuation response vector corresponding to each physical type, and synthesize the comprehensive impact index of the fluctuation event. The credibility assessment and output unit is used to compare the comprehensive influence index with the preset credibility threshold and output the test credibility assessment result.

[0015] The present invention has the following beneficial effects:

[0016] In the reliability assessment method for optical module testing provided by this invention, the time series of operating parameters in multiple dimensions during the optical module testing process are acquired and preprocessed for noise reduction. Wavelet transform is used to detect abnormal fluctuations in each operating parameter time series to obtain candidate fluctuation moments. Based on multi-dimensional joint confirmation rules, valid fluctuation moments are determined. Response feature parameters such as response delay, response peak value, and recovery time in each dimension are extracted using the valid fluctuation moments as the origin. A fluctuation response vector is constructed based on the response feature parameters of each dimension, and a multi-dimensional disturbance response description is generated. The type norm of each physical type is calculated according to a preset physical type classification rule, and a comprehensive influence index is synthesized. Finally, the comprehensive influence index is compared with a preset reliability threshold to output the test reliability assessment result. Because this invention extracts abnormal fluctuation information from the temporal evolution of parameters in each dimension through wavelet transform singularity detection and distinguishes between single-dimensional random noise and real multi-dimensional coordinated fluctuations based on multi-dimensional joint confirmation rules, it can effectively identify abnormal disturbance events occurring during the testing process even when the parameter values ​​do not exceed the specification thresholds. Meanwhile, by classifying parameters into continuous power type, zero-baseline jitter type, and stepped cumulative type according to their physical nature, and introducing a small signal amplification factor for zero-baseline jitter type parameters and a recovery time weighting factor for stepped cumulative type parameters, this approach makes explicit the physical knowledge in the optical module field that "small jitter anomalies have high diagnostic value" and "the core evaluation dimension of bit error events is recovery time," and integrates it into quantitative calculations. This solves the problems in traditional methods where small but critical anomaly signals are easily overwhelmed by large-amplitude parameters, and the inability to fairly compare the impact of different physical types of parameters on test results. The resulting comprehensive impact index comprehensively reflects the breadth of abnormal fluctuations, the response intensity of each dimension, and recovery characteristics, achieving a quantitative assessment of the impact of abnormal disturbances during testing. This fills the technical gap in existing "specification threshold judgment methods," which can only determine whether parameters are qualified but cannot assess the reliability of the testing process itself, significantly improving the accuracy and reliability of optical module test result reliability assessment. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a reliability assessment method for an optical module testing process according to an embodiment of the present invention. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0019] This application provides a method and system for reliability evaluation in the testing process of optical modules, referring to... Figure 1 The methods include:

[0020] Step S10: Obtain the data of each operating parameter during the optical module test and form a time series of each operating parameter.

[0021] As an example, the optical module under test is connected to a high-speed oscilloscope, a bit error rate analyzer, and an optical power meter via an optical module test fixture, at a preset sampling frequency. Simultaneously collect operating parameters from multiple dimensions to form a multi-dimensional operating parameter time series. .in, Index for parameter dimensions, =1, 2, 3, 4, 5, 6, 7, 8, which correspond to optical power, electrical signal amplitude, bit error count, eye height, eye width, random jitter, deterministic jitter, and extinction ratio, respectively. For sampling time index, =1,2,3,......, , It equals the product of the duration of a single test and the sampling frequency.

[0022] Sampling frequency in this embodiment A 1MHz frequency and a 1-microsecond sampling period are used. Each parameter sequence is timestamped and stored. A median filter is used to preprocess the original sequence for noise reduction. The filter window length is set according to the parameter type: for continuous analog parameters (… =1, 2, 4, 5, 8), with a window length of 5 to 9 sampling points, corresponding to a time length of 5 μs to 9 μs; for the bit error count ( =3), the window length is 3 sampling points, corresponding to a time length of 3μs; for jitter parameters ( =6, 7), with a window length of 5 to 9 sampling points, corresponding to a time length of 5 to 9 μs.

[0023] It should be noted that the lengths of the filtering windows mentioned above are all much smaller than the rise time of typical physical disturbances. Typical physical disturbances include mechanical vibrations of the test fixture, power supply ripple fluctuations, and optical power drift caused by changes in ambient temperature. These time-domain response times are typically on the order of hundreds of microseconds to milliseconds. Therefore, median filtering only smooths occasional spike noise on the order of the sampling period, while the rise time, peak time, and other time-domain characteristics of the fluctuating signal are fully preserved. The filtered sequence is still denoted as... Sequence length and sampling time index The correspondence remains unchanged.

[0024] Step S20: Detect abnormal fluctuations in the time series of each operating parameter to obtain a set of candidate fluctuation moments.

[0025] The system calls the mapping rules between runtime parameters and wavelet detection configuration to process the time series of each runtime parameter after filtering. implement Layer discrete wavelet transform yields the detail coefficients for each layer. ,in For the decomposition level index, This is the translation position index of the wavelet transform on the time axis. Translation position index Index of original sampling time By using wavelet transform to correspond the time shift relationships one-to-one, it is ensured that the candidate fluctuation moments detected at each scale can be uniformly mapped back to the original time axis.

[0026] Furthermore, the modulus of each layer of detail coefficients is calculated point by point. Local maxima of the modulus are detected at each scale. Along the scale direction, modulus maxima corresponding to the same event are searched and associated in the spatial neighborhood of adjacent scales, forming a cross-scale modulus maxima line. For each modulus maxima line, the decay slope of the modulus with the number of decomposition layers is fitted, and the Lipschitz regularity index at that time position is calculated according to the wavelet transform singularity detection theory. When the regularity index is less than a preset threshold, it indicates that the signal has singularity at that moment, and that time position is marked as a preliminary candidate fluctuation moment for that operating parameter. In this embodiment, the regularity index is the singularity determination threshold, with a value ranging from 0.2 to 0.5; in this embodiment, it is 0.3. The smaller the threshold, the more stringent the singularity detection; it can be adjusted according to the noise level and detection sensitivity requirements of the optical module test.

[0027] It should be noted that the mapping rules between the operating parameters and the wavelet detection configuration are formulated based on the temporal morphological characteristics of each parameter, as follows: For continuous analog parameters, the signal changes relatively smoothly but are superimposed with mid-to-high frequency measurement noise, so the Db4 wavelet, which has good regularity and tight support characteristics, is selected, and the number of decomposition layers is 5; for step-like discrete sequences with monotonically non-decreasing bit error counts, the effective fluctuations are manifested as continuous step jumps in the count values, so the Haar wavelet, which is sensitive to step signals, is selected, and the number of decomposition layers is 4; for jitter parameters, which often exhibit step-like abrupt changes when fluctuations occur, and whose baseline level is zero under normal operating conditions, the Haar wavelet is also selected, and the number of decomposition layers is 4.

[0028] After the initial candidate fluctuation time detection is completed, a second-level verification is performed on the initial candidate fluctuation times of the bit error count value sequence. In this embodiment, the abnormal fluctuation perception of the bit error dimension adopts a hierarchical progressive confirmation strategy: median filtering is used as the first level to filter out occasional single-point spikes introduced by the sampling link; this is used as the second level to further eliminate short-term sudden abnormal fluctuations, ensuring that the finally confirmed bit error fluctuations have high reliability.

[0029] The secondary verification process is as follows: First, a stable data segment within the entire test sequence, where no candidate fluctuation moments are detected, is selected as a global reference segment with a length of at least 1000 sampling points. Within the global reference segment, windows are divided according to a preset sampling window length. The cumulative bit error rate within each window is counted, and its mean and standard deviation are calculated. Then, for each initial candidate fluctuation moment, the bit error rate count data of several consecutive sampling windows following it is extracted, and the cumulative total bit error rate within each window is counted. If the cumulative total bit error rate within any sampling window after the candidate moment is greater than the mean plus three times the standard deviation, the candidate moment is deemed valid and retained as a candidate fluctuation moment in the bit error dimension; otherwise, it is determined to be a single-bit occasional bit error or sampling noise and is discarded. The sampling window length is configured according to the sampling frequency and can be selected from 10 to 20 sampling periods.

[0030] Furthermore, the system aligns the candidate fluctuation moments detected by each dimension of the parameter set on the time axis. A time tolerance window is set, which is used when candidate fluctuation moments exist simultaneously in the parameter sequences of at least three dimensions within the tolerance window, and at least one of them contains a continuously simulated parameter. When the time position is at least one of 1, 2, 4, 5, or 8, it is determined as a valid fluctuation moment. .

[0031] The time tolerance window is determined based on the sampling frequency and the response time differences of each dimension parameter. In this embodiment, the sampling frequency is 1MHz, and the singularity localization deviation of each dimension parameter after wavelet transform is typically no more than 20 sampling periods, while the multi-instrument synchronization deviation is within 10 sampling periods. Therefore, the time tolerance window is set to 50 sampling periods, corresponding to a time length of 50 microseconds. In practical applications, this window value can range from 50 to 200 sampling periods to cover the response time differences of each dimension parameter and system synchronization deviation under different test configurations.

[0032] It should be noted that the candidate fluctuation time and the effective fluctuation time t* detected in this step may all have parameter values ​​within the specification threshold. These fluctuations cannot be identified as anomalies under the existing "specification threshold determination method". However, this step extracts the cooperative anomalous fluctuation information within the specification threshold from the parameter time series evolution process through wavelet singularity detection and multi-dimensional joint confirmation, providing input for subsequent steps to evaluate the credibility of the testing process.

[0033] Step S30: Using the effective fluctuation time as the origin, extract the response feature parameters of each dimension.

[0034] For each determined effective fluctuation moment In the time series of each operating parameter after filtering Above, cut off Then the first preset duration The data is used as the response observation interval. In this embodiment, A 30-second timeframe is used. The system takes a period of stable data at the beginning of the test where no candidate fluctuations are detected, and uses the mean of each parameter within this period as the uniform baseline level for the entire test. In this embodiment, the stable segment is taken from the first 5 seconds of data at the start of the test, and step S20 must confirm that no candidate fluctuation moments are detected in this segment to ensure that the baseline is not contaminated by abnormal fluctuations.

[0035] Furthermore, the effective fluctuation time to + Using the specified time frame as the observation window, the deviation of each dimension of data from the baseline level is calculated as the response waveform:

[0036] , ,in, Effective fluctuation time The corresponding sampling point index, The sampling frequency.

[0037] Response feature parameters are extracted based on the response waveforms of various dimensions, specifically:

[0038] ① Response delay Defined as the index of the sampling point corresponding to the effective fluctuation time. To the The number of sampling points required for the dimension parameter to reach its maximum absolute deviation within the observation window is given by the formula: ,in, Indicates that all conditions are met. Among the sampling points, take The index of the sample point with the largest value.

[0039] ② Response peak Defined as the response waveform within the observation window The maximum absolute value, after normalization, is a dimensionless value used to characterize the maximum impact of fluctuations on this parameter.

[0040] For parameters in continuous simulations, baseline relative value normalization is used: ,in It should be a very small positive number to prevent the denominator from being zero.

[0041] For bit error counts and jitter parameters, the baseline level is zero under normal operating conditions, and they are normalized using the ratio of the absolute change to the upper limit of the parameter specification: ,in The pass / fail threshold for this parameter, as defined in the optical module specification, is obtained from the specification according to the optical module model and rate class.

[0042] ③Recovery time Defined as response waveform The number of sampling points required for the absolute value to drop from the peak to near the baseline level within a preset tolerance range. The system indexes the sampling points from the peak response. Begin searching sequentially along the direction of increasing sampling point index. Find the first satisfied The sampling point location. Among them, the recovery criterion threshold. Set parameters according to their type: For continuous simulation parameters, For bit error counts, random jitter, and deterministic jitter, , As a preset recovery coefficient, the parameter value for continuous simulation type in this embodiment is 0.1, and the parameter value for bit error count, random jitter, and deterministic jitter type is 0.05.

[0043] If until If the recovery criterion is not met even after the window ends, then Marked as "outside the observation window", this dimension is treated as a persistent response, and its response characteristics are only used in subsequent calculations based on response delay and response peak value.

[0044] Step S40: Construct a fluctuation response vector based on the response feature parameters of each dimension, and generate a multi-dimensional fluctuation response description.

[0045] As an example, the first Response latency of dimensional operating parameters Response peak and recovery time Combined into a wave response vector:

[0046] ;

[0047] For effective fluctuation time The system calculates the fluctuation response vectors for each of the eight operating parameter dimensions. , , , The subscripts 1, 2, ..., 8 correspond to optical power, electrical signal amplitude, bit error count, eye height, eye width, random jitter, deterministic jitter, and extinction ratio, respectively. These eight vectors together constitute a complete response description of a wave event across all physical dimensions.

[0048] Furthermore, the system will consider response latency in various dimensions. Generate response timing by sorting the values ​​from smallest to largest. Smaller parameters respond to fluctuations first, corresponding to the order in which the fluctuations affect the system. Larger parameters exhibit relatively delayed responses. The system arranges the response peaks of each dimension according to the response time sequence, forming an intensity evolution sequence that reflects the changing trend of the influence of each parameter on the response time sequence.

[0049] At the same time, the system will include recovery time in various dimensions. Arranged from shortest to longest numerical value, a recovery time difference sequence is formed. Continuous response dimensions marked as "outside the observation window" are uniformly treated as having the longest recovery time and placed at the end of the sequence. By comparing different dimensions... It identifies parameters that remain in an abnormal state after fluctuations, with the persistent response dimension automatically identified as a parameter with persistent anomalies.

[0050] It should be noted that when the difference in the number of sampling point intervals corresponding to the response delay is less than the preset tolerance, the response timing is considered to be the same, and they are arranged in descending order of their response peak values.

[0051] The fluctuation response vector and multidimensional response description constructed in this step characterize the multidimensional coupling features of the fluctuation event from three dimensions: response timing, impact magnitude, and recovery characteristics. This provides structured input for subsequent steps to calculate the comprehensive impact index and evaluate the credibility of the testing process.

[0052] Step S50: Based on the fluctuation response vectors of each dimension, calculate the comprehensive fluctuation impact index to achieve a quantitative assessment of the credibility of the testing process.

[0053] Step S51: Determine the set of valid response dimensions to be included in the evaluation.

[0054] Based on the determination of effective fluctuation moments in step S20 and the actual performance of the response waveform in step S30, the system selects the set of effective response dimensions to participate in the calculation of this comprehensive impact index. .

[0055] The specific filtering criteria are: Dimensional parameters must satisfy both of the following conditions to be included in the set. : No. The candidate fluctuation moments of the dimensional parameter fall within the time tolerance window and are included in the confirmation criteria for valid fluctuation moments; The maximum absolute value of the response waveform of a dimension within the observation window is not less than the recovery criterion threshold of that dimension.

[0056] Let set The number of elements contained is , which represents the number of dimensions that generated a valid response in this fluctuation event.

[0057] Step S52: Calculate the comprehensive impact index based on the categorical norm.

[0058] For the current valid fluctuation moment, the system obtains the set of fluctuation response vectors for this fluctuation event from step S40. , , ..., },in Each component is a numerical result extracted and output in step S30. This represents the total number of parameter dimensions.

[0059] Based on the physical source and response characteristics of each parameter within the optical module, The dimensions are divided into the following three physical types, as shown in Table 1:

[0060] Table 1: Parameter Type Classification

[0061]

[0062] set According to the types in the table above, separate and form subsets. , , The number of its elements is denoted as . , , .

[0063] for For continuous power parameters, their norm components are directly taken. , and These three characteristics together constitute it. Because... This is already the normalized value divided by the baseline level in step S30. The three components are comparable in dimensions and can be directly calculated as the sum of squares and the square root. The specific formula is:

[0064] ;

[0065] for For zero-baseline jitter parameters, a small signal amplification factor is introduced when calculating their norm components. Its value range is preferably 3 to 10. Multiply It is then included in the calculation. It should be noted that in optical module testing practice, the absolute values ​​of random jitter and deterministic jitter are usually much smaller than parameters such as optical power and eye height, but their minute anomalies are often the only observable signs of early faults such as laser degradation and clock recovery circuit lockout. If the norm is calculated without processing and mixed with other parameters, the anomalies in the jitter dimension will be completely drowned out by the numerically dominant continuous power parameters. Amplification factor The introduction of this parameter is based on the physical knowledge in the field of optical modules that "minor jitter anomalies have high diagnostic value," ensuring that anomalies in this type of parameter are reflected in the comprehensive impact index in a way that matches their physical importance. The specific formula is:

[0066] ;

[0067] for The tiered cumulative parameter is essentially the cumulative number of transmission errors in the digital link. In optical module testing, the core evaluation dimension for bit error events is "whether the digital link recovers to normal," rather than the instantaneous magnitude of the bit error count. Therefore, a recovery time weighting factor is introduced when calculating the norm component of this parameter. , will be multiplied by This value is then used in the calculation to reasonably amplify the contribution of recovery time to the norm, in order to match the physical characteristic that "recovery time is the most important evaluation indicator" in the bit error rate dimension. The preferred value range is 2 to 5. The specific formula is:

[0068] ;

[0069] in, Recovery time is the dimension of the bit error count. Option 3 is preferred.

[0070] The system normalizes the type norms of the three physical types according to their respective dimensions, sums them, and then calculates the comprehensive impact index of this fluctuation event by combining the proportion of effective response dimensions. :

[0071] ;

[0072] in, This represents the total number of parameter dimensions; in this embodiment, it is set to 8. The preset normalized baseline value, independent of specific fluctuation events, is determined through statistical analysis of the comprehensive response vector magnitude in historical test data of the same type of optical module under known severe fluctuation conditions; It is a very small positive number.

[0073] It should be noted that the above-mentioned normalization based on the number of dimensions after summing the categorical norms ensures that the contribution of each physical type to the comprehensive impact index is independent of the number of effective response dimensions it contains, and depends only on the degree of anomaly within that type. This avoids the problem of a single type having a large number of dimensions dominating the overall assessment, ensuring that each of the three types of physical signals participates in the final decision equally according to its physical importance. Independently reflects the breadth of the impact of fluctuations. Regarding the recovery time in step S30... The persistent response dimension marked as "out of observation window" has its Δi_mr component in each type norm taking values ​​of... .

[0074] Step S53: Output the test credibility evaluation results.

[0075] The system compares the calculated comprehensive impact index with the credibility threshold and outputs the credibility assessment of this test based on the comparison result: if the comprehensive impact index is less than or equal to the credibility threshold, it is determined that the comprehensive impact of the fluctuation on the test result during this test is within an acceptable range, the test result is credible, and it is marked as "passed"; if the comprehensive impact index is greater than the credibility threshold, it is determined that the comprehensive impact of the fluctuation on the test result during this test has exceeded the acceptable range, the test result is unreliable, and it is marked as "re-tested".

[0076] The evaluation results simultaneously output a detailed breakdown of the response types for each dimension during this fluctuation. For persistent response dimensions marked as "out of observation window" in step S30, their parameter names and "persistent abnormality" status are separately noted in the evaluation report, indicating that this parameter failed to recover to normal levels within the observation window after the fluctuation. Amplification factor in class Significantly amplifying the contribution of jitter anomalies, and In the class For high-confidence burst error events with a probability of 3, the assessment report will label them as "Minor Signal Anomaly Warning" and "High-Confidence Error Event," respectively, to provide interpretable supplementary information for subsequent manual review. These labels are supplementary information to the assessment report and do not affect the calculation results of the comprehensive impact index or the overall credibility judgment.

[0077] The credibility threshold is determined by statistical analysis of the comprehensive impact index calculated from known normal test data. For example, the mean of the comprehensive impact index under normal conditions is taken plus three times the standard deviation. The specific value can be configured according to the optical module model, rate level and test accuracy requirements.

[0078] This application provides a reliability assessment system for optical module testing processes, including a parameter acquisition and preprocessing unit, a fluctuation detection and confirmation unit, a response feature extraction unit, a fluctuation response description generation unit, a comprehensive impact index calculation unit, and a reliability assessment and output unit.

[0079] The parameter acquisition and preprocessing unit is used to acquire multiple dimensions of operating parameter data during the optical module testing process, form time series of each operating parameter, and perform median filtering and noise reduction preprocessing on each time series; the multiple dimensions include optical power, electrical signal amplitude, bit error count, eye height, eye width, random jitter, deterministic jitter, and extinction ratio.

[0080] The fluctuation detection and confirmation unit, connected to the parameter acquisition and preprocessing unit, is used to perform abnormal fluctuation detection on the time series of each operating parameter, obtain candidate fluctuation moments in each dimension, and determine the valid fluctuation moments from the candidate fluctuation moments based on the multi-dimensional joint confirmation rules. The fluctuation detection and confirmation unit performs discrete wavelet transform on each type of parameter sequence according to the mapping rules between the operating parameters and the wavelet detection configuration, and detects the preliminary candidate fluctuation moments in each dimension through the modulus maxima line and the Lipschitz regularity index. It also performs a second-level verification on the bit error count dimension to eliminate invalid candidate moments.

[0081] The response feature extraction unit, connected to the fluctuation detection and confirmation unit and the parameter acquisition and preprocessing unit, is used to extract response feature parameters of various dimensions within a preset observation window, with each effective fluctuation moment as the origin; the response feature parameters include response delay, normalized response peak value, and recovery time.

[0082] The fluctuation response description generation unit, connected to the response feature extraction unit, is used to combine the response delay, response peak and recovery time of each dimension into a fluctuation response vector, and generate a multi-dimensional fluctuation response description containing response time sequence, intensity evolution sequence and recovery time difference sequence.

[0083] The comprehensive impact index calculation unit, connected to the response feature extraction unit, is used to filter effective response dimensions based on the fluctuation response vectors of each dimension and the pre-configured physical type classification rules. After calculating the type norms for continuous power type, zero baseline jitter type and stepped cumulative type respectively, the comprehensive impact index of this fluctuation event is calculated. Among them, a small signal amplification factor is introduced for the zero baseline jitter type parameter and a recovery time weight factor is introduced for the stepped cumulative type parameter.

[0084] The credibility assessment and output unit is connected to the comprehensive impact index calculation unit. It is used to compare the comprehensive impact index with a preset credibility threshold, output the test credibility assessment result, and mark at least one of the following in the assessment result: continuous abnormal dimension, small signal abnormality warning and high confidence error event.

Claims

1. A method for evaluating the reliability of an optical module testing process, characterized in that, The method includes: The time series of operating parameters in multiple dimensions during the optical module testing process are obtained, and noise reduction preprocessing is performed on each time series of operating parameters. Wavelet transform is used to detect abnormal fluctuations in the time series of various operating parameters to obtain candidate fluctuation moments in each dimension. Based on the multi-dimensional joint confirmation rule, the effective fluctuation moments are determined from the candidate fluctuation moments. Using the effective fluctuation time as the origin, response feature parameters of each dimension are extracted within a preset observation window. The response feature parameters include response delay, normalized response peak value, and recovery time. A fluctuation response vector is constructed based on the response feature parameters of each dimension to generate a multi-dimensional fluctuation response description; The parameter dimension is divided into multiple physical types according to the preset physical type classification rules. The type norm is calculated based on the wave response vector corresponding to each physical type, and the comprehensive impact index of the wave event is synthesized. The comprehensive impact index is compared with a preset credibility threshold, and the test credibility assessment result is output.

2. The reliability assessment method for the optical module testing process according to claim 1, characterized in that, The process of acquiring time series of operating parameters in multiple dimensions during the optical module testing process, and performing noise reduction preprocessing on each operating parameter time series, includes: A high-speed oscilloscope, a bit error rate analyzer, and an optical power meter are connected via an optical module test fixture. Multiple operating parameters are collected synchronously at a preset sampling frequency to form a multi-dimensional operating parameter time series. The multiple dimensions include optical power, electrical signal amplitude, bit error rate count, eye height, eye width, random jitter, deterministic jitter, and extinction ratio. The time series of each operating parameter are preprocessed for noise reduction using a median filter. The filtering window length for continuous analog parameters is 5 to 9 sampling points, the filtering window length for bit error count is 3 sampling points, and the filtering window length for jitter parameters is 5 to 9 sampling points.

3. The reliability assessment method for the optical module testing process according to claim 1, characterized in that, The wavelet transform anomaly detection of the time series of each operating parameter yields candidate fluctuation moments in each dimension, including: Based on the preset mapping rules between the operating parameters and the wavelet detection configuration, the Db4 wavelet basis is used to perform a 5-level discrete wavelet transform on the continuous simulation parameters, and the Haar wavelet basis is used to perform a 4-level discrete wavelet transform on the bit error count and jitter parameters to obtain the detail coefficients of each level. Calculate the modulus of the detail coefficients at each level, detect the local maxima of the modulus at each scale, and associate the modulus maxima corresponding to the same event in the spatial neighborhood of adjacent scales along the scale direction to form a cross-scale modulus maxima line. For each modulus maximum line, the slope of the modulus decay with the number of decomposition layers is fitted, the Lipschitz regularity index at the corresponding time position is calculated, and the time when the regularity index is less than the preset singularity judgment threshold is marked as the initial candidate fluctuation time of that dimension.

4. The reliability assessment method for the optical module testing process according to claim 3, characterized in that, The method further includes performing a second-level verification on the preliminary candidate fluctuation times in the bit error count dimension, specifically: In the entire test sequence, a stable segment where no candidate fluctuations were detected was selected as the global reference segment. The window was divided according to the preset sampling window length, and the cumulative bit error rate in each window was counted. The mean and standard deviation of the error rate were calculated. For each initial candidate fluctuation moment, extract the bit error count data of multiple consecutive sampling windows after it, and count the cumulative total number of bit errors in each window; If the total number of cumulative bit errors in any sampling window after a candidate time point is greater than the mean plus three times the standard deviation, the candidate fluctuation time point is retained; otherwise, it is discarded.

5. The reliability assessment method for the optical module testing process according to claim 1, characterized in that, The determination of valid fluctuation moments from the candidate fluctuation moments based on multi-dimensional joint confirmation rules includes: A time tolerance window is set, which is determined based on the sampling frequency and the response time differences of each dimension parameter; When candidate fluctuation moments exist simultaneously in parameter sequences of at least three dimensions within the time tolerance window, and the at least three dimensions contain at least one continuous simulation-type parameter, the time position is determined as a valid fluctuation moment. The continuous simulation parameters include optical power, electrical signal amplitude, eye height, eye width, and extinction ratio.

6. The reliability assessment method for the optical module testing process according to claim 1, characterized in that, Using the effective fluctuation time as the origin, response feature parameters of each dimension are extracted within a preset observation window, including: Take a period of stable data at the beginning of the test where no candidate fluctuations were detected, and use the mean of each dimension parameter within this period as the baseline level; Starting from the effective fluctuation moment, the data of the first preset duration is extracted as the response observation interval, and the deviation of each dimension of the data from the baseline level is calculated as the response waveform. The number of sampling points from the effective fluctuation moment to the response waveform reaching its maximum absolute deviation within the observation window is taken as the response delay; The response peak value is obtained by normalizing the maximum absolute value of the response waveform within the observation window. For continuous analog parameters with a non-zero stable baseline level, the baseline relative value is used for normalization. For bit error counts and jitter parameters with a zero baseline level, the absolute change is normalized relative to the upper limit of the parameter specification. The recovery time is defined as the number of sampling points required for the absolute value of the response waveform to drop from its peak to near the baseline level within a preset tolerance range.

7. The reliability assessment method for optical module testing process according to claim 1, characterized in that, The process of constructing a fluctuation response vector based on response feature parameters of each dimension to generate a multi-dimensional fluctuation response description includes: The response latency, peak response, and recovery time of each dimension are combined into a fluctuating response vector; The response delays of each dimension are sorted from smallest to largest to generate a response time sequence. The response peaks of each dimension are arranged according to the response time sequence to form an intensity evolution sequence. The recovery times of each dimension are arranged from shortest to longest to form a recovery time difference sequence. Among them, dimensions whose recovery time exceeds the observation window are marked as persistent responses, and are uniformly processed according to the longest recovery time and placed at the end of the sequence.

8. The reliability assessment method for optical module testing process according to claim 1, characterized in that, The preset physical type classification rules include continuous power type, zero baseline jitter type and stepped accumulation type, wherein continuous power type corresponds to optical power, electrical signal amplitude, eye height, eye width and extinction ratio, zero baseline jitter type corresponds to random jitter and deterministic jitter, and stepped accumulation type corresponds to bit error count value; The calculation of the type norm based on the wave response vector corresponding to each physical type includes: Filter effective response dimensions, wherein the effective response dimension is the dimension whose candidate fluctuation moment falls into the valid fluctuation moment confirmation criterion and whose maximum absolute value of the response waveform is not less than the recovery criterion threshold of that dimension; The effective response dimension is divided into three categories according to the preset physical type classification rules. For each category, the type norm consisting of response delay, response peak value and recovery time is calculated. For the response peak value of the zero baseline jitter type parameter, a preset small signal amplification factor is introduced before it is included in the calculation. For the recovery time of the step accumulation type parameter, a preset recovery time weighting factor is introduced before it is included in the calculation.

9. The reliability assessment method for optical module testing process according to claim 8, characterized in that, The comprehensive impact index of the synthetic volatility event includes: The comprehensive influence index is obtained by summing the norms of each type after normalizing them according to their respective number of dimensions, and then multiplying them by the proportion of the number of effective response dimensions in the total number of parameter dimensions. Among them, the recovery time exceeds the continuous response dimension of the observation window, and its recovery time component in the type norm is the number of sampling points corresponding to the first preset duration.

10. A system for evaluating the reliability of an optical module testing process according to claim 1, characterized in that, include: The parameter acquisition and preprocessing unit acquires the time series of various operating parameters and performs noise reduction on the time series. The fluctuation detection and confirmation unit uses wavelet transform to detect abnormal fluctuations in the time series, obtains candidate fluctuation moments in various dimensions, and determines the valid fluctuation moments from the candidate fluctuation moments through multi-dimensional joint verification. The response feature extraction unit extracts the response delay, the normalized response peak value, and the recovery time within a preset observation window, taking the effective fluctuation time as the origin. The fluctuation response description generation unit constructs a fluctuation response vector based on the response features of each dimension, generating a multi-dimensional fluctuation response description. The comprehensive impact index calculation unit divides the parameter dimensions according to the preset physical type, calculates the type norm based on the response vector of each type of wave, and synthesizes the comprehensive impact index of the wave event. The credibility assessment and output unit compares the comprehensive influence index with the credibility threshold and outputs the test credibility assessment results.

Images