A connector signal integrity detection method and system

Abstract

The application relates to the technical field of connector signal integrity detection, and discloses a connector signal integrity detection method and system, which comprises the following steps: when a connector is made to perform a continuous reciprocating bending motion, dynamic bending angles and signal waveform data of the connector are synchronously collected, a group of multi-dimensional signal features are extracted from the dynamic bending angles and the signal waveform data, then, a normal fluctuation benchmark of the signal features under each bending angle is established, and non-periodic fluctuations deviating from the benchmark and not associated with a fixed period of the bending motion are identified, so that potential performance risks of signal transmission are determined, the application establishes a dynamic signal behavior benchmark throughout the complete motion stroke of the connector, the dynamic signal behavior benchmark can separate instant signal degradation caused by micro defects of materials and seemingly random from regular mechanical motion background changes, and thus focuses of detection are transferred from traditional conformity judgment to mechanism identification of early potential performance risks.

Description

TECHNICAL FIELD

[0001] The present application relates to a connector signal integrity detection method and system, belonging to the technical field of connector signal integrity detection. BACKGROUND

[0002] In modern electronic devices, especially in those applications that need to withstand repeated bending or layout in a compact space, the signal transmission reliability of flexible printed circuit connectors is a key technical indicator, therefore, strict performance verification of its signal integrity in the production link has become a quality control standard universally followed by the industry.

[0003] Currently, the mainstream detection method in the industry is usually to inject test signals into the connector under static or specific bending posture, and to determine whether the signal transmission quality is qualified according to whether the key performance indicators such as eye diagram template, transmission delay or crosstalk amplitude meet the preset fixed threshold, this method is effective for identifying the performance decline caused by design or manufacturing defects, however, when the connector is placed in a long-term service scene that needs to withstand tens of thousands or even hundreds of thousands of bending cycles, a risk that is easily ignored in existing static detection begins to appear, under the pressure of such a scene, local progressive interlayer peeling may occur between the insulating material and the conductor layer inside the connector on a microscopic scale, and the continuous transmission of high-frequency signals may also produce heat accumulation in the local area that cannot be accurately perceived by the external environment temperature control, the combined effect of these two effects will cause the transmission delay and waveform shape of the signal to present complex nonlinearity and instantaneous fluctuations that are not fixedly associated with bending actions when passing through these microscopic areas.

[0004] Faced with this phenomenon, existing detection logic based on fixed thresholds and linear models has limitations in its processing model and discrimination method. Its initial design intention is to evaluate signal quality under steady-state conditions. Therefore, it treats these instantaneous and non-periodic signal fluctuations as occasional and irregular background noise or measurement outliers, and tends to average or ignore them directly in data processing, rather than treating them as a key indicator containing information about the evolution of structural health. This processing method makes it difficult for the detection logic to capture and quantify those instantaneous local signal degradations with extremely short durations. Furthermore, during dynamic bending, it cannot clearly distinguish between the regular normal signal fluctuations generated by mechanical stress in the connector and the non-periodic abnormal signal fluctuations caused by internal microstructural damage. Moreover, due to the lack of in-depth correlation and cumulative analysis between signal anomalies and specific physical states, the evolution trend of any potential defects with time or stress cycle increases also becomes unobservable and unassessable. Therefore, the technical problem to be solved by this invention is how to establish a dynamic monitoring method that can penetrate conventional noise and deeply correlate with physical state in order to identify and quantify non-periodic and transient signal anomalies caused by microstructural degradation of connectors under long-term dynamic bending, thereby providing early warning of potential performance hazards and health status evolution trends of connectors. Summary of the Invention

[0005] This invention provides a connector signal integrity detection method and system. Its main purpose is to solve the problem that existing technologies, when faced with long-term dynamic bending conditions of connectors, are unable to effectively identify non-periodic instantaneous signal anomalies caused by internal microstructure deterioration due to the limitations of their detection models and discrimination logic, thus failing to provide early and accurate warnings of potential performance hazards.

[0006] To achieve the above objectives, the present invention provides a connector signal integrity detection method, which includes the following steps:

[0007] Step 1: While making the connector under test perform a preset continuous reciprocating bending motion, simultaneously acquire the real-time bending angle data of the connector during the bending process;

[0008] Step 2: Inject test signals into the connector and continuously capture the signal waveform data output by the connector, wherein each segment of signal waveform data is associated with a real-time bending angle data.

[0009] Step 3: Extract a set of multi-dimensional signal features from the signal waveform data, including the instantaneous slope change rate of the signal edge and the local opening rate of the eye diagram;

[0010] Step 4, a non-linear wave mode discrimination is performed, which first establishes a normal fluctuation range of multi-dimension signal features for each bending angle by statistically referencing sample data, and then identifies any non-periodic fluctuation in the multi-dimension signal features of the current connector to be detected, which is beyond the normal fluctuation range and has no fixed period correlation with the reciprocating bending movement.

[0011] Step 5, based on the quantified amplitude and frequency of non-periodic fluctuation, it is determined that the connector has potential performance risks of signal transmission.

[0012] Preferably, the process of identifying non-periodic fluctuation in step 4 further includes a micro-event accumulation-based analysis process, which includes: step 4.1, capturing any signal transient deviation with an amplitude or voltage rate of change beyond a preset basic threshold in the signal waveform data as a micro-event; step 4.2, binding each captured micro-event with its associated real-time bending angle data; step 4.3, binding the micro-event according to its real-time bending angle data into the corresponding preset bending angle interval, and accumulating the count in the interval; step 4.4, when the cumulative count in any bending angle interval shows a continuous growth trend in continuous bending movement, it is determined that there is a risk of structural defects in the angle interval.

[0013] Preferably, the process of establishing a normal fluctuation range of multi-dimension signal features in step 4 includes: testing a plurality of reference connector samples under the same continuous reciprocating bending movement and collecting their multi-dimension signal feature data at all bending angles; statistically analyzing the collected multi-dimension signal feature data to calculate the mean and standard deviation of each signal feature at each bending angle; and defining the numerical interval determined by the mean plus or minus a preset multiple of the standard deviation at each bending angle as the normal fluctuation range of the signal feature at that angle.

[0014] Preferably, the process of extracting multi-dimension signal features in step 3 further includes extracting waveform local distortion rate and signal jitter spectrum distribution; wherein, extracting the local opening rate of eye diagram includes dividing the eye diagram display area into a pixel grid and calculating the number of pixels in each grid unit, and calculating the local contraction area where the number of pixels in the signal track decreases non-periodically; and extracting the signal jitter spectrum distribution includes performing fast Fourier transform on the time series data of signal jitter to identify newly appearing energy peaks at non-integer clock frequencies.

[0015] Preferably, after step 5, the method further comprises: recording the quantified magnitude and occurrence frequency of the identified potential performance hazard together with the corresponding real-time bending angle data to form a degradation information database; analyzing the degradation information database to calculate the occurrence frequency growth rate of the potential performance hazard in different bending angle intervals as the number of bending movements increases; and when the occurrence frequency growth rate exceeds a preset critical rate value for a plurality of consecutive analysis periods, determining that the connector has an accelerated degradation trend and outputting a warning signal.

[0016] Preferably, before calculating the occurrence frequency growth rate, the method further comprises: analyzing the degradation information database to identify bending angle intervals with a number of data points below a preset number threshold; for the identified bending angle intervals, using the historical data mean of adjacent intervals or a preset low-pass filtering algorithm to fill or smooth the data in the interval to obtain corrected occurrence frequency data; and calculating the occurrence frequency growth rate based on the corrected occurrence frequency data.

[0017] Preferably, in step 4.4, the rule for determining that the cumulative count presents a sustained growth trend is: in any specific bending angle interval, calculate the difference between the number of micro-events accumulated in the current analysis period and the number of micro-events accumulated in the previous analysis period to obtain a growth amount; and if the growth amount is greater than zero for a plurality of consecutive analysis periods, determine that the cumulative count presents a sustained growth trend.

[0018] Preferably, the step 4 of performing non-linear fluctuation pattern discrimination further comprises: pre-constructing a defect pattern library storing reference feature vectors of multi-dimensional signal features corresponding to interlayer delamination and local temperature rise defects respectively; combining the real-time extracted multi-dimensional signal features into a to-be-tested feature vector; calculating the Euclidean distance between the to-be-tested feature vector and each reference feature vector in the defect pattern library; and when any Euclidean distance is less than a preset matching distance threshold, classifying the identified non-periodic fluctuation as the defect type corresponding to the reference feature vector with the smallest Euclidean distance.

[0019] Preferably, the angle range of the continuous reciprocating bending movement is set to zero to one hundred and eighty degrees, and the frequency is set to one hertz; and the oscilloscope sampling rate used to capture the signal waveform data in step 2 is more than twenty times the test signal bit rate.

[0020] A connector signal integrity detection system, the system comprising:

[0021] a dynamic bending control and information acquisition module configured to make the connector to be detected perform a preset continuous reciprocating bending movement and synchronously acquire real-time bending angle data of the connector during the bending process;

[0022] A signal injection and capture module, which is connected to a dynamic bending control and information acquisition module, is configured to inject a test signal into the connector and continuously capture the signal waveform data output by the connector, wherein each segment of signal waveform data is associated with a real-time bending angle data.

[0023] A multi-dimensional signal feature extraction module, which is connected to a signal injection and capture module, is configured to extract a set of multi-dimensional signal features from signal waveform data, including the instantaneous slope change rate of the signal edge and the local opening rate of the eye diagram.

[0024] A nonlinear fluctuation pattern discrimination module is connected to a multi-dimensional signal feature extraction module. It is configured to establish the normal fluctuation range of multi-dimensional signal features for each bending angle by statistical reference sample data, and then identify any non-periodic fluctuations in the multi-dimensional signal features of the connector under test that exceed the normal fluctuation range and have no fixed periodicity related to the reciprocating bending motion.

[0025] A performance hazard determination module, which is connected to a nonlinear fluctuation mode discrimination module, is configured to determine the potential performance hazard of the connector in signal transmission based on the quantized amplitude and occurrence frequency of non-periodic fluctuations.

[0026] Compared with the prior art, the beneficial effects of the present invention are:

[0027] 1. By synchronizing the dynamic reciprocating bending motion of the connector with the continuous capture of signal waveforms, and by binding the extraction of multi-dimensional features such as signal edge slope and eye diagram opening status with real-time bending angle data, a set of signal behavior benchmarks spanning the entire motion stroke of the connector is established. Based on this, the method focuses on identifying non-periodic fluctuations that are not associated with a fixed periodicity in the reciprocating bending motion. This approach allows seemingly random instantaneous signal degradation caused by microscopic defects in the material to be separated from the regular background signal changes caused by mechanical motion. It shifts the focus of detection from judging whether the signal amplitude is qualified to judging whether the signal behavior conforms to its inherent laws under specific physical conditions, thereby constructing an objective and stable reference system for identifying early potential performance hazards.

[0028] 2、The application introduces an analysis process based on micro-event accumulation, captures any instantaneous deviation of signal waveform exceeding the preset basic threshold of amplitude or voltage change rate as a micro-event, and binds it with the corresponding real-time bending angle data, and then accumulates it in the corresponding bending angle interval. The value of this mechanism lies in that it does not make direct qualitative judgment on isolated and seemingly accidental instantaneous signal anomalies, but observes whether these micro-events show a trend of continuous growth aggregation in a specific bending angle interval in a long time and multiple bending cycles to reversely infer whether there is a structural stress concentration or micro-crack defect risk at the physical position. The analysis logic of micro-event accumulation converts the signal peaks or drops that are usually regarded as noise in the traditional method due to their non-repeatability into structural defect indications with clear physical positioning significance.

[0029] 3、The application further constructs a deterioration information database to record the quantitative amplitude, occurrence frequency and corresponding real-time bending angle of the hidden danger, and calculates the occurrence frequency growth rate of the potential performance hidden danger in different bending angle intervals with the increase of bending motion times based on this. When the growth rate continuously exceeds the preset critical value, the system determines that the connector has an accelerated deterioration trend. This extension from state recognition to trend prediction enables the detection system not only to find the current performance hidden danger of the connector, but also to evaluate the reliability evolution track of the future, which provides objective data support for preventive maintenance and whole life cycle quality management. Its significance lies in converting single quality detection into continuous monitoring of the reliability evolution process of the product. BRIEF DESCRIPTION OF DRAWINGS

[0030] Fig. 1 It is the overall logical flow diagram of the detection method of the application;

[0031] Fig. 2 It is the functional module and node architecture diagram of the detection system of the application;

[0032] Fig. 3 It is the interaction timing diagram of the benchmark establishment and real-time detection stage of the application. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0034] The disclosed connector signal integrity detection method starts with a dynamic bending control and information acquisition module. The output of the module, i.e. the signal waveform data carrying physical posture labels, is sent to a signal injection and capture module, which is then analyzed by a multi-dimensional signal feature extraction module to generate a structured feature vector. A nonlinear fluctuation mode discrimination module compares the feature vector with a set of dynamic benchmarks to identify abnormal signal events. Finally, a performance hazard determination module determines the potential performance hazards of the connector based on the quantitative statistical results of the identified abnormal events. Under the working conditions of long-term reciprocating bending of flexible printed circuit connectors, the microstructure deterioration caused by material fatigue and the local temperature rise caused by high-frequency signal transmission will introduce transient nonlinear fluctuations in the signal path that are not fixedly associated with the bending action. To effectively capture such fluctuations, the dynamic bending control and information acquisition module and the signal injection and capture module in the present scheme are configured to perform a set of synchronized dynamic information acquisition procedures. The procedures drive the connector to be detected to perform a preset continuous reciprocating bending motion from zero to one hundred and eighty degrees, with a frequency of one hertz. In this process, a high-precision rotary encoder linked to the bending mechanism outputs the real-time bending angle data of the connector at millisecond intervals. At the same time, a signal generator injects a 10Gbps pseudo-random binary sequence test signal into the connector, and the output is continuously captured by an oscilloscope with a sampling rate set to more than twenty times the bit rate of the test signal. The oscilloscope binds each segment of the captured signal waveform data with the real-time bending angle data transmitted synchronously by the rotary encoder, thereby constructing a space-time reference frame that accurately corresponds the signal instantaneous electrical behavior to the connector instantaneous physical posture point by point.

[0035] In view of signal distortion caused by microstructure deterioration, which is characterized by nonlinear changes in waveform details, the multi-dimensional signal feature extraction module is configured to execute a set of waveform microstructure-oriented feature extraction algorithms, which specifically include parallel computation of signal edge transient slope rate of change, eye diagram local opening rate, waveform local distortion rate, and signal jitter spectrum distribution; for the signal edge transient slope rate of change, the algorithm subdivides the rising and falling edge transition regions of the signal into a number of equally time-interval sampling points, calculates the transient slope of each point, and quantifies the nonlinearity degree by further calculating the second-order difference of adjacent transient slopes, a second-order difference value greater than a preset threshold corresponding to a micro-distortion of the signal edge caused by local impedance discontinuity; for the eye diagram local opening rate, the algorithm divides the eye diagram display area into a pixel grid, and identifies the local contraction area where the signal trajectory occurs non-periodic decline by counting the number of pixels of the signal trajectory in each grid element; for the signal jitter spectrum distribution, the algorithm performs fast Fourier transform on the time series data of the signal jitter to identify newly appearing energy peaks at non-integer clock frequencies; in this way, the original one-dimensional time-domain waveform data is transformed into a multi-dimensional feature vector, providing more sensitive and rich inputs for subsequent pattern discrimination; considering that the connector will also produce regular normal signal fluctuations during dynamic bending due to its own mechanical deformation, the core of the nonlinear fluctuation pattern discrimination module is to perform a nonlinear fluctuation pattern discrimination, which first collects multi-dimensional signal feature data of a plurality of reference connector samples at all bending angles, and calculates the mean and standard deviation of each signal feature at each bending angle through statistical analysis, and then defines the value interval determined by the mean plus or minus a preset multiple, such as 3 times the standard deviation, as the normal fluctuation range of the signal feature at that angle; after obtaining this dynamic reference, the module compares the multi-dimensional signal features of the current connector to be detected in real time, and identifies any fluctuations that exceed the normal fluctuation range of the corresponding bending angle and have no fixed period correlation with the reciprocating bending motion as non-periodic fluctuations; this can separate the transient signal deterioration caused by material micro-defects from the regular mechanical motion background signal changes.

[0036] For certain signal transient abnormalities triggered by structural defects and with very short duration and non-repetitiveness, the nonlinear fluctuation mode discrimination module further integrates a micro-event accumulation-based analysis process that captures any signal transient deviation with amplitude or voltage rate of change exceeding a preset basic threshold in the signal waveform data as a micro-event, and binds it with the associated real-time bending angle data. After that, all the bound micro-events are classified into corresponding preset bending angle intervals according to their real-time bending angle data, and are accumulated and counted within the intervals. The rule for determining whether the accumulated count presents a continuous growth trend is: in any specific bending angle interval, if the difference between the number of accumulated micro-events in the current analysis period and the number of accumulated micro-events in the previous analysis period is greater than zero continuously in a preset number of analysis periods, it is determined that the accumulated count presents a continuous growth trend. When the accumulated count in any bending angle interval presents such a trend, it is determined that there is a risk of structural defects in that angle interval. This analysis process converts independent transient abnormalities into structural defect indications with clear physical positioning significance. After identifying non-periodic fluctuations, the performance hazard discrimination module determines whether the connector has potential performance hazards for signal transmission based on the quantitative amplitude and occurrence frequency of these fluctuations. To achieve the evaluation of the connector's health state evolution trend, the method further includes a construction and analysis process of a degradation information database. This process records the quantitative amplitude, occurrence frequency and corresponding real-time bending angle of the determined potential performance hazards together to form a degradation information database. After that, by analyzing the database, the occurrence frequency growth rate of potential performance hazards in different bending angle intervals with the increase of bending motion times is calculated. When the growth rate exceeds a preset critical rate value continuously for multiple analysis periods, the system determines that the connector has an accelerated degradation trend and outputs a warning signal, so that the detection system can not only determine the current performance hazards of the connector, but also evaluate its future reliability evolution trajectory, providing data support for preventive maintenance.

[0037] The determination of each core operation parameter in the scheme follows a standardized system deployment and calibration procedure. The value of the reference sample quantity used to establish the statistical baseline is determined based on statistical convergence criteria. Starting from the initial value, the quantity is iteratively increased until the relative change rate of the standard deviation of any core signal feature, such as the signal edge transient slope change rate, is less than 5% after the introduction of new samples for two consecutive times at the baseline bending angle. The value at this time is determined, and its typical value range is between 10 and 30. At the same time, the analysis period length used for trend judgment is defined as the number of bending cycles that can make the micro-event baseline count statistically stable. Specifically, in the continuous testing of a defect-free reference sample, the expected value of the micro-event cumulative count in any bending angle interval is stable and exceeds a statistical base of 5, and the coefficient of variation is less than 0.3. Finally, the setting of the bending angle interval division granularity is related to the physical positioning accuracy. The rule is to ensure that at the maximum bending radius of the connector, the calculated value of the corresponding physical arc length is always less than 1 millimeter, so as to realize the positioning ability matching the size of the microstructure defect. Before classifying the non-periodic fluctuation mode, the original multi-dimensional signal feature data must undergo a dimensionless processing. This processing maps each independent feature dimension to the [0, 1] interval through minimum-maximum value scaling transformation, where and are the global minimum and maximum values of the feature dimension in the full range of testing data of all calibration samples. This step aims to eliminate the inherent scale difference of different physical features and ensure the fairness of subsequent Euclidean distance calculation. In addition, to deal with the TB-level raw data stream generated by high-frequency sampling, the system engineering implementation adopts a distributed processing architecture, that is, the high-load feature extraction algorithm is executed in real time on the edge computing node close to the signal capture module, and only the structured feature vector stream generated after processing and reduced by several orders of magnitude in data quantity is transmitted to the central analysis server through the high-speed data bus for subsequent mode discrimination and trend analysis.

[0038] Example 1: In a mechanical arm application for precision surgery, the flexible printed circuit connector deployed inside the joint of the arm needs to withstand tens of thousands of high-precision, small-range reciprocating bending during the operation process. After a few weeks of continuous use, the end effector of the mechanical arm exhibits unexpected, millisecond-level micro-vibration during the execution of a specific suturing action, i.e., when the arm joint is bent to a narrow angle range near 78 degrees. This vibration introduces uncertainty into the subsequent operation process. To address this issue, the engineering team used a static test method to detect the signal integrity of the connector when the mechanical arm was stationary at multiple bending angles. However, the test results all showed that the eye diagram margin met the preset performance standards, and no stable performance degradation was found. To investigate this problem, the connector of the mechanical arm was connected to the detection system of the invention. The system began to execute a synchronous dynamic information collection procedure, i.e., while driving the mechanical arm joint to continuously reciprocate between 75 degrees and 85 degrees at a frequency of one hertz, the system synchronously acquired the dynamic bending angle and signal waveform data. In the initial stage of testing, the signal edge instantaneous slope rate and eye diagram local opening rate extracted by the multi-dimensional signal feature extraction module fell within the normal fluctuation range corresponding to each bending angle established based on the reference sample data when compared in the non-linear fluctuation mode discrimination module. The system did not determine that there was a performance risk.

[0039] As the number of bending cycles increased to more than five thousand, an analysis process based on micro-event accumulation was performed. This analysis process captured any instantaneous deviation in signal waveform amplitude or voltage rate of change that exceeded the preset basic threshold as a micro-event and bound it with the corresponding real-time bending angle data. System data showed that within the narrow bending angle range of 78.3 degrees to 78.5 degrees, the cumulative count of micro-events presented a sustained growth. Specifically, during the analysis period from the fifth thousandth to the sixth thousandth cycle, the number of micro-events accumulated in this range was 8, while in the analysis period from the sixth thousandth to the seventh thousandth cycle, this number increased to 19. In the subsequent three analysis periods, this growth was greater than zero. This phenomenon triggered the judgment rule in the non-linear fluctuation mode discrimination module about the sustained growth trend of the cumulative count.

[0040] This judgment result stems from the synergistic operation of two technical features. First, by testing multiple reference connector samples, a multi-dimensional signal characteristic normal fluctuation range was established for each bending angle, providing a dynamic benchmark for the system. This benchmark can filter out predictable background signal fluctuations caused by the regular movement of the robotic arm joints between 75 and 85 degrees. Second, it is precisely in this background signal-filtered data environment that the analysis process based on micro-event accumulation can effectively capture and accumulate those extremely weak, previously submerged, non-periodic instantaneous deviations. This operational method resolves the technical contradiction of confusing normal mechanical fluctuations and non-periodic abnormal fluctuations in dynamic testing, resulting in a clearly defined... Spatially directional structural defect risk signals were revealed. Based on the risk location information output by the system, maintenance personnel conducted high-magnification microscopic observation of the corresponding physical parts of the connector under a 78.4-degree bend, identifying a micron-sized crack located at the edge of the conductor layer that had previously failed to be detected in any static inspection. After replacing the connector, the robotic arm did not reproduce any minute jitter in subsequent tens of thousands of tests with the same movements. This detection process no longer directly determines whether the signal is qualified, but instead evaluates whether the signal behavior conforms to its inherent laws under a specific physical state. By correlating and cumulatively analyzing instantaneous signal anomalies with determined physical motion states, an objective indicator for characterizing the evolution trend of potential structural defects inside the connector was established.

[0041] Example 2: To objectively verify the effectiveness of the technical scheme of the present application in distinguishing non-periodic, transient signal anomalies caused by microstructure degradation, the following comparative test was designed and performed. The test platform consisted of an electric precision bending test bench, a Keysight N4903B high-performance serial bit error rate tester, and a Keysight Infiniium UXR series oscilloscope. The temperature and humidity of the test environment were constantly controlled. The test objects were divided into two groups. The control group A included ten flexible printed circuit connector samples confirmed to have no micro-defects. The test group B included ten connector samples with non-penetrating micro-cracks of five micrometers in width and fifty micrometers in length at the conductor-insulation layer interface at the maximum bending stress point prepared by a predetermined process. During the test, all samples were subjected to twenty thousand consecutive reciprocating bending movements at a frequency of one hertz within a bending range of zero to one hundred and eighty degrees, and a 10Gbps pseudo-random binary sequence test signal was injected synchronously. The setting of the basic threshold for micro-event capture was based on the technical consideration of distinguishing the detection sensitivity and the system inherent noise. The determination procedure for the threshold was as follows: first, the system's own noise waveform was continuously collected under the condition of open circuit at the input end of the tester, the statistical distribution of the voltage change rate was calculated, and the upper limit of the 99.7% confidence interval of the distribution was taken as the system noise baseline. Finally, the basic threshold for micro-event capture was set to one and a half times the system noise baseline. This setting aimed to make the system sensitive to signal anomalies beyond its own noise distribution range while having the ability to suppress regular noise.

[0042] If the traditional detection method of measuring the signal eye height at the ninth degree bending angle is used as the qualified threshold value of 300 mV, the determination result cannot distinguish the two groups of samples. The sample eye height of the control group A always maintains above 320 mV, and the eye height of the sample of the test group B slightly decreases with the increase of the cycle number, but the reading is still 305 mV, which is above the qualified threshold value, so the determination result based on the traditional detection method is always qualified. If the micro-event cumulative count of the two groups of samples in the bending interval of 88 degrees to 92 degrees is analyzed with the change of the bending cycle number, two different evolution trajectories can be observed. The count of the control group A is in a state of stable fluctuation with a slope close to zero and low position during the whole test cycle of 20,000 cycles, and the count value is always below 20, indicating that no progressive structural change occurs inside. In contrast, the count of the test group B is basically consistent with that of the control group A within the initial 10,000 cycles, but after the cycle number exceeds 10,000, the value presents a nonlinear and increasing trend. The count value rises to 89 at the first 15,000 cycles, and reaches 216 at the second 20,000 cycles. This growth trend triggers the built-in determination rule of the system about the cumulative count presenting a continuous growth trend, so the sample of the test group B is determined to have a risk of structural defects.

[0043] The appearance of this data trend is due to the expansion of the pre-prepared micro-cracks of the test group B sample under the cyclic stress, which increases the frequency and amplitude of the non-periodic transient abnormality of the signal when passing through the physical position. The analysis process based on the micro-event accumulation correlates and analyzes the trend of these seemingly isolated events in the time domain in the two dimensions of spatial bending angle and time cycle number, so as to translate the evolution process of the physical defects into quantifiable data indicators. The data of the test shows that when facing the signal integrity problem caused by the deterioration of the microstructure and gradually increasing, the detection method relying on a single macroscopic indicator and a fixed threshold value has the limitation of being unable to identify the potential performance risk in an early stage. The method of the present application can objectively indicate the developing physical defects inside the connector in the form of a risk of structural defects by establishing a dynamic signal behavior benchmark and accumulating and analyzing the trend of the non-periodic micro-events associated with a specific physical state.

[0044] Embodiment 3: The present embodiment is combined with Figs. 1 to 3 A connector signal integrity detection method and system are described, such as Fig. 1As shown, the process begins with the handling of the connector to be tested, through a dynamic bending control and information acquisition module to make it perform continuous reciprocating bending motion, and synchronously acquire its real-time bending angle, at the same time, a signal injection and capture module injects test signals into the connector, and continuously captures the signal waveform associated with the real-time bending angle, these two information converge into a data stream named [angle + waveform] synchronous data stream, and are sent to the multi-dimensional signal feature extraction module, which is responsible for extracting signal edge instantaneous slope rate of change and eye diagram local opening rate and other features, and then generates an output named multi-dimensional feature vector, this feature vector is the core basis for subsequent analysis, it is received by a nonlinear fluctuation mode discrimination module, which identifies any non-periodic fluctuations that exceed the normal fluctuation range and are not associated with fixed cycle bending motion, to identify and output a result named non-periodic fluctuation event, it is worth noting that the discrimination module also integrates a micro-event accumulation-based analysis process, which captures and accumulates micro-events occurring at a specific bending angle to determine the risk of potential structural defects, finally, the non-periodic fluctuation event output by the nonlinear fluctuation mode discrimination module is sent to a performance hazard determination module, which determines whether the connector has potential performance hazards based on the quantitative amplitude and occurrence frequency of these events, and generates a final output named potential performance hazard determination result, in addition, the determination result is also recorded in a degradation information database, which can output an accelerated degradation warning signal when it determines that the connector has an accelerated degradation trend by calculating the occurrence frequency growth rate of the hazards.

[0045] As Fig. 2As shown, the system architecture is mainly composed of four key nodes, namely the dynamic bending test bench and signal generation and acquisition equipment responsible for physical operation, the data processing and analysis server responsible for data processing, and the engineer workstation responsible for human-computer interaction. Specifically, the dynamic bending test bench contains a motion control firmware responsible for executing bending actions and outputting angle data, while the signal generation and acquisition equipment is responsible for processing high-frequency test signals and collecting waveform data through its internal signal injection and capture program. These two field devices communicate with the core data processing and analysis server through their respective control buses such as USB / LAN or GPIB / LAN, and transmit the collected angle data and high-speed waveform data to the server through high-speed data buses such as PCIe / LAN. The server, as the center of the system, integrates a series of functional components, including data synchronization and preprocessing services for data alignment, multi-dimensional feature extraction engine for feature generation, nonlinear fluctuation discrimination service for anomaly discrimination, degradation trend analysis engine for trend prediction, and a core database for storing degradation information and defect patterns. The processing results of the server, including test results and warning information, are transmitted to the engineer workstation through standard network protocols such as TCP / IP for display and management by the test management and monitoring client software on the workstation.

[0046] As shown in Fig. 3 The upper half of the flowchart is the establishment of the normal fluctuation baseline stage. This stage begins with inputting multiple defect-free samples into the test system from the reference sample set. In a loop for each reference sample, the test system collects full-angle waveform data and passes it to the feature extractor for processing. The feature extractor extracts a set of multi-dimensional features including edge slope change rate, eye opening rate, waveform distortion rate, and jitter spectrum distribution, and transmits the feature set to the statistical analyzer in the form of a message sending feature data set. Based on the data set, the statistical analyzer calculates the mean and standard deviation of the features at each bending angle, and stores the normal fluctuation range to the baseline library accordingly, thus completing the establishment of the baseline. The lower half of the flowchart is the actual detection stage. The input at this time is the connector to be tested. The test system performs real-time feature extraction and transmits the information to the feature extractor. The feature extractor then initiates a request to the baseline library with a message querying the corresponding angle baseline. After obtaining the response returning the normal range from the baseline library, the feature extractor can identify non-periodic fluctuations through comparison, thus achieving accurate detection of potential performance risks.

[0047] In a scenario where the root cause of potential performance issues of the connector needs to be classified to guide process improvement, the technical premise is that the nonlinear fluctuation pattern discrimination module needs to have the ability to distinguish different physical failure mechanisms, which depends on a pre-constructed defect pattern library that stores multi-dimensional signal feature reference feature vectors corresponding to specific defect patterns and a set of matching discrimination thresholds. To construct this defect pattern library and calibrate the corresponding discrimination logic, the following standardization engineering procedures need to be performed. First, prepare two groups of flexible printed circuit connector samples with a known single defect type, where sample group C induces micron-level interlayer peeling between the conductor layer and the insulating layer through a specific hot pressing process, and sample group D integrates a thermistor element that generates local temperature rise under rated signal power to simulate local overheating defects. Then, place these two groups of samples together with the defect-free control group A in the detection system, perform a standard 20,000 times of continuous reciprocating bending test, and collect the complete and synchronized signal waveform data with the bending angle.

[0048] Next, the collected data is processed to generate reference feature vectors for each type of sample. For each sample in the control group A, sample group C, and sample group D, the system extracts all multi-dimensional signal features that occur during the entire test period, including signal edge instantaneous slope change rate, eye diagram local opening rate, waveform local distortion rate, and signal jitter spectrum distribution. After normalizing these feature values, a to-be-tested feature vector is constructed. Then, by calculating the average of the to-be-tested feature vectors of all samples in the same group, the reference feature vector of the defect-free pattern, the reference feature vector of the interlayer peeling pattern, and the reference feature vector of the local temperature rise pattern are obtained respectively. These reference feature vectors are stored in the defect pattern library as a reference for subsequent pattern matching.

[0049] Finally, the Euclidean distance matching threshold for defect categorization is calibrated. The system calculates the Euclidean distance between the feature vector of each unknown defect sample and the three reference feature vectors in the defect pattern library. Through this process, two types of distance data are obtained. One is the distance between all defect samples and their corresponding correct defect pattern reference feature vectors, i.e., intra-class distance. The other is the distance between these samples and other irrelevant defect pattern reference feature vectors, i.e., inter-class distance. The determination rule for the matching distance threshold is that its value is greater than the maximum value of all calculated intra-class distances, and at the same time, it is less than the minimum value of all calculated inter-class distances. The threshold can take the midpoint of these two boundary values. This rule aims to ensure that when the Euclidean distance between an unknown sample and a reference feature vector is less than the threshold, the unknown sample is classified as the defect type corresponding to the reference feature vector. Through the above procedure, the defect pattern library and the matching distance threshold are established, so that the nonlinear fluctuation pattern discrimination module can not only identify the non-periodic fluctuations, but also preliminarily classify their physical origins.

[0050] In order to ensure the adaptability of the deterioration trend analysis function before applying the detection system of the present application to batch detection of a new type of flexible printed circuit connector, a preliminary data processing procedure and critical value calibration process need to be performed. This process first analyzes the deterioration information database collected in the preliminary test stage, identifies the bending angle interval with the number of data points less than a preset number threshold. For the identified bending angle interval, the system is configured to use a moving average low-pass filter algorithm to smooth the occurrence frequency data in this interval to obtain corrected data for subsequent growth rate calculation.

[0051] Next, the system performs accelerated deterioration testing on a set of reference connector samples of this new type, i.e., continuously performing reciprocating bending motion until its signal transmission performance is below a preset failure standard. In this process, the system continuously records the frequency growth rate of each reference sample in different bending angle intervals, and finally forms a statistical distribution containing all growth rates of all samples throughout their life cycle. The warning critical rate value of this specific type of connector is determined according to this statistical distribution. The determination procedure is as follows: extract the occurrence frequency growth rate of each sample in the last three analysis periods before failure from the distribution, and set the 75th percentile value of the collection of these rate values as the preset critical rate value of this type of connector. Through this calibration, the trigger threshold of the warning signal is associated with the actual deterioration behavior of this specific type of connector near the end of its service life.

[0052] In order to ensure the accuracy of subsequent batch detection, when the detection system needs to adapt to a new connector to be detected which is different from previous models in both physical structure and material composition, a pre-calibration procedure of system deployment needs to be performed on an automated production line. The procedure first performs a baseline quality verification of the test signal source by a loopback test, i.e. connecting the output of the signal generator directly to the input of the oscilloscope, to measure and record the jitter spectrum and noise floor of the test signal itself. Only when these indicators are below the system preset quality standard, the subsequent calibration process is started, which can exclude interference introduced by the test signal source itself. Then, the procedure performs a bend characteristic scanning test on a reference sample of the new connector model, which drives the connector to bend in a wide frequency range from 0.1 Hz to 10 Hz and synchronously analyzes the frequency spectrum of its signal response to identify the mechanical resonance frequency points of the connector structure. Finally, the frequency of the reciprocating bending motion for formal testing is set to a value lower than the lowest identified resonance frequency to avoid the influence of mechanical resonance on signal transmission stability. At the same time, during this scanning test, the system measures the average rise and fall time of the signal of the connector model, and sets the pixel grid density in the eye diagram local opening rate analysis according to this time, with the setting rule being to ensure that the transition region of the signal edge can span at least ten grid units on the time axis, so that the shape of the signal edge can be analyzed.

[0053] Finally, the transient anomaly identification rule in non-periodic fluctuation discrimination is calibrated. The system uses signal waveform data from multiple defect-free reference samples of the same model to statistically analyze the probability distribution of the occurrence of m data points (where m is 0, 1, 2,...) that exceed the normal fluctuation range of their corresponding bending angles in a sliding sampling point window of length N (where N can be 5). Then, according to a preset false positive rate of less than 0.001%, a combination of m out of N, such as 3 out of 5, is determined from the probability distribution as the final transient anomaly determination rule. Through this series of procedures, the detection parameters for the specific connector model, including test frequency, grid density, and transient anomaly determination rule, are finally determined.

[0054] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.

Claims

1. A connector signal integrity detection method, characterized in that, The method includes the following steps: Step 1: While making the connector under test perform a preset continuous reciprocating bending motion, simultaneously acquire the real-time bending angle data of the connector during the bending process; Step 2: Inject test signals into the connector and continuously capture the signal waveform data output by the connector, wherein each segment of signal waveform data is associated with a real-time bending angle data. Step 3: Extract a set of multi-dimensional signal features from the signal waveform data, including the instantaneous slope change rate of the signal edge and the local opening rate of the eye diagram; Step 4: Perform a nonlinear fluctuation mode discrimination. This discrimination first establishes the normal fluctuation range of multi-dimensional signal features for each bending angle by statistically referencing sample data, and then identifies any non-periodic fluctuations in the multi-dimensional signal features of the connector under test that exceed the normal fluctuation range and have no fixed periodicity related to the reciprocating bending motion. Step 5: Based on the quantized amplitude and frequency of non-periodic fluctuations, determine whether the connector has potential performance issues in signal transmission. Step 3, which involves extracting multi-dimensional signal features, further includes extracting the waveform local distortion rate and the signal jitter spectrum distribution. Specifically, extracting the eye diagram local aperture ratio involves dividing the eye diagram display area into a pixel grid and calculating the number of pixels in the signal trajectory within each grid cell, as well as calculating local contraction regions used to identify non-periodic decreases in the number of pixels in the signal trajectory. Extracting the signal jitter spectrum distribution involves performing a fast Fourier transform on the time-series data of the signal jitter to identify newly emerging energy peaks at non-integer multiples of the clock frequency. Furthermore, after step 5, the method further includes: recording the quantized amplitude and occurrence frequency of the identified potential performance hazards together with their corresponding real-time bending angle data to form a degradation information database; analyzing the degradation information database to calculate the rate of increase in occurrence frequency of potential performance hazards with the number of bending movements in different bending angle ranges; when the rate of increase in occurrence frequency exceeds a preset critical rate value for multiple consecutive analysis cycles, it is determined that the connector has an accelerated degradation trend, and an early warning signal is output. Furthermore, before calculating the frequency growth rate, the method further includes: analyzing the degradation information database to identify bending angle intervals where the number of data points is lower than a preset threshold; for the identified bending angle intervals, using the historical data average of its adjacent intervals or a preset low-pass filtering algorithm to fill in or smooth the data in the interval to obtain corrected frequency data; and calculating the frequency growth rate based on the corrected frequency data.

2. The connector signal integrity detection method according to claim 1, characterized in that, Step 4, which identifies non-periodic fluctuations, further includes an analysis process based on micro-event accumulation. This analysis process includes: Step 4.1, capturing any instantaneous deviation of the signal waveform data whose amplitude or voltage change rate exceeds a preset basic threshold as a micro-event; Step 4.2, binding each captured micro-event to its associated real-time bending angle data; Step 4.3, assigning the bound micro-events to the corresponding preset bending angle intervals based on their real-time bending angle data, and accumulating the count within those intervals; Step 4.4, when the accumulated count within any bending angle interval shows a continuous increasing trend during continuous bending motion, it is determined that there is a structural defect risk within that angle interval.

3. The connector signal integrity detection method according to claim 1, characterized in that, Step 4, which establishes the normal fluctuation range of multi-dimensional signal characteristics, includes: testing multiple reference connector samples under the same continuous reciprocating bending motion and collecting multi-dimensional signal characteristic data at all bending angles; performing statistical analysis on the collected multi-dimensional signal characteristic data to calculate the mean and standard deviation of each signal characteristic at each bending angle; and defining the numerical range determined by the standard deviation plus or minus a preset multiple at each bending angle as the normal fluctuation range of the signal characteristic at that angle.

4. The connector signal integrity detection method according to claim 2, characterized in that, In step 4.4, the rule for determining that the cumulative count shows a continuous growth trend is as follows: within any specific bending angle interval, calculate the difference between the number of micro-events accumulated in the current analysis period and the number of micro-events accumulated in the previous analysis period to obtain a growth amount; if the growth amount is greater than zero continuously in multiple preset analysis periods, then the cumulative count is determined to show a continuous growth trend.

5. The connector signal integrity detection method according to claim 1, characterized in that, Step 4 involves discriminating nonlinear fluctuation modes, which further includes: pre-constructing a defect mode library, which stores reference feature vectors of multi-dimensional signal features corresponding to interlayer peeling and local temperature rise defects, respectively; combining the multi-dimensional signal features extracted in real time into a test feature vector; calculating the Euclidean distance between the test feature vector and each reference feature vector in the defect mode library; and classifying the identified non-periodic fluctuation as the defect type corresponding to the reference feature vector with the smallest Euclidean distance when any Euclidean distance is less than a preset matching distance threshold.

6. The connector signal integrity detection method according to claim 1, characterized in that, The angle range of the continuous reciprocating bending motion is set to 0 degrees to 180 degrees, and the frequency is set to 1 Hz; furthermore, the oscilloscope sampling rate used to capture the signal waveform data in step 2 is more than 20 times the bit rate of the test signal.

7. A connector signal integrity detection system, the system being used to run the connector signal integrity detection method of claim 1, characterized in that, The system includes: A dynamic bending control and information acquisition module is configured to enable the connector under test to perform a preset continuous reciprocating bending motion and simultaneously acquire the real-time bending angle data of the connector during the bending process; A signal injection and capture module, which is connected to a dynamic bending control and information acquisition module, is configured to inject a test signal into the connector and continuously capture the signal waveform data output by the connector, wherein each segment of signal waveform data is associated with a real-time bending angle data. A multi-dimensional signal feature extraction module, which is connected to a signal injection and capture module, is configured to extract a set of multi-dimensional signal features from signal waveform data, including the instantaneous slope change rate of the signal edge and the local opening rate of the eye diagram. A nonlinear fluctuation pattern discrimination module is connected to a multi-dimensional signal feature extraction module. It is configured to establish the normal fluctuation range of multi-dimensional signal features for each bending angle by statistical reference sample data, and then identify any non-periodic fluctuations in the multi-dimensional signal features of the connector under test that exceed the normal fluctuation range and have no fixed periodicity related to the reciprocating bending motion. A performance hazard determination module, which is connected to a nonlinear fluctuation mode discrimination module, is configured to determine the potential performance hazard of the connector in signal transmission based on the quantized amplitude and occurrence frequency of non-periodic fluctuations.

Images