Non-contact defect detection method and device for glass core packaging substrate

Abstract

The invention provides a non-contact defect detection method and device for a glass core packaging substrate, and relates to the technical field of electronic packaging, and the method comprises the following steps: dividing the surface area of the substrate into a metal area and a glass area, dividing the metal area into a first grid, and dividing the glass area into a second grid; applying a dual-frequency signal to the first grid and the second grid in a time-sharing manner to obtain a deformation image sequence and a dielectric frequency shift amount; performing bimodal feature extraction on the deformation image sequence and the dielectric frequency shift amount, performing defect collaborative judgment according to the extracted features, and when a collaborative judgment result meets a preset condition, generating a rechecking instruction; and in response to the re-checking instruction, adjusting division parameters of the first grid and the second grid, performing re-checking judgment until a re-checking judgment result meets a confidence threshold condition, and outputting a defect detection report. The problem of cross interference of detection signals of the metal area and the glass area of the glass core packaging substrate is solved.

Description

Technical Field

[0001] This invention relates to the field of electronic packaging technology, and in particular to a non-contact defect detection method and apparatus for glass core packaging substrates. Background Technology

[0002] Glass-core packaging substrates are widely used in high-end electronic devices due to their combination of high-density wiring and thinness. However, during manufacturing, defects such as microcracks and interlayer delamination can easily occur due to process fluctuations or material defects. Traditional contact testing methods (such as probe testing and optical microscopy) can easily cause secondary damage to precision substrates and are difficult to cover buried layers or small areas. While non-contact testing technologies (such as infrared thermography and laser ultrasound) avoid physical contact, single-mode testing has limitations: stress change detection in the metal circuit area is easily affected by the dielectric properties of the glass area, and dielectric frequency shift detection in the glass area may generate noise due to electromagnetic coupling in the metal area, resulting in low defect location accuracy and a high false positive rate.

[0003] In the existing technology, there is a lack of non-contact collaborative detection solutions that target the characteristics of composite materials in glass core packaging substrates. This makes it difficult to effectively solve the problem of cross-interference between detection signals in the metal area and the glass area, and thus difficult to meet the requirements for high-precision and high-reliability defect detection.

[0004] Therefore, it is necessary to provide a non-contact defect detection method and apparatus for glass core encapsulation substrates to solve the above-mentioned technical problems. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a non-contact defect detection method and apparatus for glass core packaged substrates. Through dual-modal non-contact detection, collaborative judgment, and adaptive verification mechanisms, it effectively solves the problem of cross-interference between detection signals in the metal and glass regions of the glass core packaged substrate, thereby improving the accuracy, reliability, and efficiency of defect identification.

[0006] This invention provides a non-contact defect detection method for glass core packaged substrates, the method comprising the following steps:

[0007] Based on the distribution of metal lines on the substrate to be tested, the surface area of ​​the substrate is divided into a metal area and a glass area, and the metal area is divided into a first grid and the glass area is divided into a second grid according to a preset initial resolution.

[0008] Dual-frequency signals are applied to the first grid and the second grid in a time-division manner according to a preset initial excitation intensity, and the deformation image sequence corresponding to the first grid and the dielectric frequency shift corresponding to the second grid are acquired simultaneously.

[0009] Dual-mode feature extraction is performed on the deformation image sequence and the dielectric frequency shift to generate the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region, respectively.

[0010] Defect co-judgment is performed based on the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region. When the co-judgment result meets the preset dual-mode conflict condition, a verification instruction is generated.

[0011] In response to the verification command, the initial excitation intensity and initial scanning accuracy of the first grid and the second grid are adjusted, the deformation image sequence and dielectric frequency shift are reacquired and verified, until the verification result meets the confidence threshold condition, and finally the defect detection report is output.

[0012] Preferably, the step of dividing the substrate surface area into a metal area and a glass area based on the metal line distribution of the substrate to be tested, and dividing the metal area into a first grid and the glass area into a second grid according to a preset initial resolution, includes:

[0013] Import the CAD file of the substrate metal circuit of the substrate to be tested, and parse it to generate the boundary coordinate set of the metal area and the boundary coordinate set of the glass area;

[0014] The preset initial resolution is loaded, and combined with the corresponding set of boundary coordinates, a first grid covering the metal area and a second grid covering the glass area are generated.

[0015] Preferably, the step of applying dual-frequency signals to the first grid and the second grid in a time-division manner according to a preset initial excitation intensity, and simultaneously acquiring the deformation image sequence corresponding to the first grid and the dielectric frequency shift corresponding to the second grid, includes:

[0016] Based on the boundary coordinate sets of the first grid and the second grid, and combined with the preset initial excitation intensity, a first frequency band excitation signal acting on the metal region and a second frequency band excitation signal acting on the glass region are respectively set, and the frequency bands of the first frequency band excitation signal and the second frequency band excitation signal do not overlap;

[0017] A time-division control strategy is adopted to sequentially apply the first frequency band excitation signal and the second frequency band excitation signal: during the period when the first frequency band excitation signal is applied, the deformation image sequence of the first grid and the dielectric frequency shift of the second grid are collected at a set sampling frequency.

[0018] Preferably, the step of performing dual-mode feature extraction on the deformation image sequence and the dielectric frequency shift to generate the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region, respectively, includes:

[0019] Multi-frame difference operation is performed on the deformation image sequence to extract the deformation displacement field corresponding to each cell of the first grid, and a stress gradient model of the metal region is constructed based on Hooke's law to calculate the stress gradient matrix covering the metal region; wherein, the stress gradient model takes the gradient value of the deformation displacement field as input and the stress change rate of the grid cell as output.

[0020] Spectral analysis is performed on the dielectric frequency shift to extract the frequency domain features of the dielectric constant corresponding to each cell of the second grid, and the difference is calculated with the preset reference dielectric spectrum to generate a dielectric deviation matrix covering the glass region; wherein, the reference dielectric spectrum is generated based on the dielectric response of the defect-free glass region under the same excitation conditions.

[0021] Preferably, the defect collaborative determination based on the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region, when the collaborative determination result meets the preset dual-mode conflict condition, generates a review instruction, including:

[0022] A spatial alignment operation is performed on the stress gradient matrix and the dielectric deviation matrix, wherein the element values ​​of the stress gradient matrix are the stress change rates of the cells in the first grid, and the element values ​​of the dielectric deviation matrix are the frequency domain deviations of the dielectric constants of the cells in the second grid.

[0023] When the number of physical coordinates that meet the dual-mode conflict condition exceeds the threshold of the number of conflict points, the collaborative judgment result is determined to be a conflict, and a review instruction is generated; wherein the dual-mode conflict condition is defined as: under the same physical coordinate, if the element value of the stress gradient matrix exceeds the first preset threshold and the corresponding element value of the dielectric deviation matrix does not exceed the second preset threshold, or the element value of the stress gradient matrix does not exceed the first preset threshold and the corresponding element value of the dielectric deviation matrix exceeds the second preset threshold.

[0024] The threshold for the number of conflict points is set based on the background noise level when the substrate is defect-free.

[0025] Preferably, in response to the verification command, the initial excitation intensity and initial scanning accuracy of the first grid and the second grid are adjusted, the deformation image sequence and dielectric frequency shift are reacquired and verified, until the verification result meets the confidence threshold condition, and finally a defect detection report is output, including:

[0026] Based on the verification instruction, the initial excitation intensity of the first grid is increased to a first verification intensity, and the initial resolution of the second grid is decreased to a second verification accuracy; wherein, the first verification intensity is a first preset multiple of the initial excitation intensity, and the second verification accuracy is a second preset multiple of the initial resolution;

[0027] Based on the first verification strength and the second verification accuracy, a third grid covering the metal area and a fourth grid covering the glass area are regenerated; wherein, the cell size of the third grid is smaller than that of the first grid, and the cell size of the fourth grid is larger than that of the second grid;

[0028] A time-division control strategy is adopted to sequentially apply the third frequency band excitation signal and the fourth frequency band excitation signal: during the period when the third frequency band excitation signal is applied, the deformation image sequence of the third grid is acquired at a sampling frequency matching the first verification intensity; during the period when the fourth frequency band excitation signal is applied, the dielectric frequency shift of the fourth grid is acquired at an acquisition window matching the second verification accuracy.

[0029] Dual-mode feature extraction is performed on the reacquired deformation image sequence and dielectric frequency shift to generate the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region for verification, and verification is performed based on the same dual-mode conflict condition.

[0030] When the number of physical coordinates that meet the conflict conditions in the review and judgment results is less than the threshold for the number of conflict points, the review is deemed to be passed, and a detection report containing defect location and type information is finally output; wherein, the defect type is determined based on the abnormal pattern association between the stress gradient matrix and the dielectric deviation matrix.

[0031] The present invention also provides a non-contact defect detection device for glass core packaged substrates, used to perform the aforementioned non-contact defect detection method for glass core packaged substrates, the device comprising:

[0032] The grid generation module is used to divide the surface area of ​​the substrate into a metal area and a glass area according to the distribution of metal lines on the substrate to be tested, and to divide the metal area into a first grid and the glass area into a second grid according to a preset initial resolution.

[0033] The dual-frequency excitation module is used to apply dual-frequency signals to the first grid and the second grid in a time-division manner according to a preset initial excitation intensity, and simultaneously acquire the deformation image sequence corresponding to the first grid and the dielectric frequency shift corresponding to the second grid;

[0034] A dual-modal feature extraction module is used to extract dual-modal features from the deformation image sequence and the dielectric frequency shift, generating a stress gradient matrix in the metal region and a dielectric deviation matrix in the glass region, respectively.

[0035] The review triggering module is used to perform defect collaborative judgment based on the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region. When the collaborative judgment result meets the preset dual-mode conflict condition, a review instruction is generated.

[0036] The verification response module is used to respond to the verification command by adjusting the initial excitation intensity and initial scanning accuracy of the first grid and the second grid, re-acquiring the deformation image sequence and dielectric frequency shift and performing verification judgment until the verification judgment result meets the confidence threshold condition, and finally outputting a defect detection report.

[0037] Compared with related technologies, the non-contact defect detection method and apparatus for glass core encapsulated substrates provided by the present invention have the following advantages:

[0038] This invention employs regional gridding processing, using stress gradient detection (based on deformation image sequences and Hooke's law modeling) for the metal region and dielectric frequency shift detection (based on spectrum analysis and benchmark comparison) for the glass region. By utilizing the complementary physical characteristics of the two modes, it effectively reduces the missed detections and false detections of single-mode detection.

[0039] This invention automatically triggers a verification command by spatially aligning the stress gradient matrix and dielectric deviation matrix and determining conflict conditions (such as abnormality in the metal region while normal in the glass region or vice versa). After adjusting the excitation intensity and scanning accuracy, it re-detects, specifically solving the problem of cross-interference and ensuring high confidence in the defect determination results.

[0040] This invention employs a time-division multiplexing approach to apply dual-frequency signals in non-overlapping frequency bands, avoiding mutual interference between detection signals in the metal and glass regions. It simultaneously acquires deformation images and dielectric frequency shift data, balancing detection efficiency and data synchronization, making it suitable for non-destructive testing of precision substrates.

[0041] The initial detection of this invention uses a basic resolution grid. In the verification stage, the detection parameters are dynamically optimized by increasing the excitation intensity of the metal area (enhancing the stress signal) and decreasing the scanning accuracy of the glass area (expanding the detection range), thus balancing accuracy and efficiency and adapting to the detection needs of different defect types. Attached Figure Description

[0042] Figure 1 A flowchart of a non-contact defect detection method for a glass core encapsulation substrate provided by the present invention;

[0043] Figure 2 This invention provides a module structure diagram of a non-contact defect detection device for a glass core encapsulated substrate. Detailed Implementation

[0044] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the drawings, not all structures. Moreover, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0045] It should also be noted that, for ease of description, the accompanying drawings show only the parts relevant to the invention and not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but it may also have additional steps not included in the drawings. The process may correspond to a method, function, procedure, subroutine, subprogram, etc.

[0046] Example 1

[0047] This invention provides a non-contact defect detection method for glass core packaged substrates, referencing... Figure 1 As shown, the method includes the following steps:

[0048] S1: Based on the distribution of metal lines on the substrate to be tested, the surface area of ​​the substrate is divided into a metal area and a glass area, and the metal area is divided into a first grid and the glass area is divided into a second grid according to a preset initial resolution.

[0049] Specifically, step S1 includes the following steps:

[0050] S11: Import the CAD file of the substrate metal circuit of the substrate to be tested, and parse it to generate the boundary coordinate set of the metal area and the boundary coordinate set of the glass area.

[0051] In this embodiment, the substrate design file (usually in Gerber or ODB++ format) is first read to extract the vector graphic data of the metal circuit layers. Metal trace regions are identified using an image segmentation algorithm, generating a mask image for the metal regions, while non-metal regions are classified as glass region masks. Polygon fitting processing is performed on the mask contours to extract the vertex coordinate sets of the polygons at the metal region boundaries and the glass region boundaries. Subsequently, the micrometer units of the design coordinate system are converted to millimeter units of the detection device coordinate system, and perspective transformation correction is performed on the vertex coordinates to eliminate distortion errors caused by substrate tilt.

[0052] S12: Load the preset initial resolution and combine it with the corresponding boundary coordinate set to generate a first grid covering the metal area and a second grid covering the glass area.

[0053] In this embodiment, preset initial resolution parameters are loaded from the system configuration file: high resolution (default 1.0mm × 1.0mm) is used for the metal area, and low resolution (default 5.0mm × 5.0mm) is used for the glass area. For the metal area, a uniform rectangular grid is divided along the X / Y axis according to the resolution step size, with its boundary polygon as the constraint area, and the center coordinates of each grid cell are marked.

[0054] For the glass region, the boundary polygons are first triangulated to generate an initial mesh, and then adjacent triangular elements are merged into a regular polygon mesh with a size not exceeding the resolution. The centroid coordinates of each polygon are then calculated.

[0055] The final output consists of a rectangular mesh data structure for the metal area and a polygonal mesh data structure for the glass area, which are sent to the scanning path planning module to control the movement trajectory of the detection equipment. During mesh generation, fragmented small cells are automatically merged to ensure the integrity of the mesh cells.

[0056] S2: Apply dual-frequency signals to the first grid and the second grid in a time-division manner according to the preset initial excitation intensity, and simultaneously acquire the deformation image sequence corresponding to the first grid and the dielectric frequency shift corresponding to the second grid.

[0057] Specifically, step S2 includes the following steps:

[0058] S21: Based on the boundary coordinate set of the first grid and the second grid, and combined with the preset initial excitation intensity, a first frequency band excitation signal acting on the metal region and a second frequency band excitation signal acting on the glass region are set respectively, and the frequency bands of the first frequency band excitation signal and the second frequency band excitation signal do not overlap.

[0059] In this embodiment, based on the boundary coordinate sets of the first grid (metal region) and the second grid (glass region) generated in step S1, and combined with preset initial excitation intensity parameters (such as current intensity in the metal region and signal power in the glass region), a low-frequency excitation signal (1-10kHz) acting on the metal region and a high-frequency excitation signal (5-100GHz) acting on the glass region are respectively configured.

[0060] Meanwhile, the signal frequency bands are strictly non-overlapping: the metal area uses a narrowband low-frequency signal to match the metal resonance characteristics and avoid conflict with the substrate's inherent frequency; the glass area uses a broadband high-frequency signal to cover the dielectric response frequency band. The signal strength is dynamically adjusted according to the physical size of the grid (larger area grids require higher strength), and signal control commands containing frequency, strength, and duration are generated.

[0061] S22: The first frequency band excitation signal and the second frequency band excitation signal are applied sequentially using a time-division control strategy: During the time period when the first frequency band excitation signal is applied, the deformation image sequence of the first grid and the dielectric frequency shift of the second grid are collected at a set sampling frequency.

[0062] In this embodiment, the time-sharing control process includes:

[0063] Step ①: The robotic arm moves to the first grid target point for metal area excitation and data acquisition, applies a preset first frequency band excitation signal, and simultaneously triggers two types of sensors:

[0064] High-speed camera: Captures the grid area at a sampling rate of ≥10000fps to obtain a deformation image sequence (records the surface micro-deformation within the excitation cycle);

[0065] Microwave resonator: Acquires dielectric frequency shift at the same grid location (records the dielectric response of the glass region to metal excitation). Note: Although during the metal region excitation period, the dielectric data of the glass region is not interfered with because the signals do not overlap.

[0066] Step 2: Excitation and Data Acquisition in the Glass Area. The robotic arm moves to the second grid target point, applies the second frequency band excitation signal, and simultaneously acquires data.

[0067] High-speed camera: to capture the deformation of the glass area (due to the brittleness of glass, the deformation is weak, requiring a high-sensitivity camera);

[0068] Microwave resonator: Directly measures the dielectric frequency shift in the glass region (main data source).

[0069] Timing management: Excitation interval for each grid is 2ms (to avoid signal crosstalk);

[0070] Total scan cycle = number of grids × (excitation duration + interval duration).

[0071] Finally, the deformation image sequence is named according to grid coordinates + timestamp, the dielectric frequency shift is stored as a frequency domain amplitude-phase matrix, and the data packet is output to the next step, containing the grid position index and the path of the original data file.

[0072] S3: Perform dual-mode feature extraction on the deformation image sequence and the dielectric frequency shift to generate the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region, respectively.

[0073] Specifically, step S3 includes the following steps:

[0074] S31: Perform multi-frame difference operation on the deformation image sequence, extract the deformation displacement field corresponding to each cell of the first grid, and construct a stress gradient model of the metal region based on Hooke's law to calculate the stress gradient matrix covering the metal region; wherein, the stress gradient model takes the gradient value of the deformation displacement field as input and the stress change rate of the grid cell as output.

[0075] In this embodiment, the acquired deformed image sequence is processed as follows:

[0076] Multi-frame differential operation:

[0077] Select consecutive frames within the excitation period, calculate the pixel grayscale difference between adjacent frames, and generate a displacement field change map.

[0078] The displacement trajectory of feature points is traced by optical flow method, and the deformation displacement vector field (including displacement in the X / Y directions) of each first grid cell is fitted.

[0079] Stress gradient modeling: A linear stress-strain relationship is established based on Hooke's law: the spatial gradient value of the deformation displacement field (calculated by the Sobel operator) is used as input, and the stress change rate of the corresponding mesh element is used as output.

[0080] Perform the following for each grid cell:

[0081] Calculate the spatial gradient of the displacement vector;

[0082] Substitute the material's elastic modulus parameter (preset Young's modulus of the metal layer) to output the stress change rate.

[0083] Matrix construction: Based on the coordinate index of the first grid cell, the stress change rate is mapped to a two-dimensional matrix (element value = element stress change rate), and the matrix row and column numbers correspond to the grid positions.

[0084] S32: Perform spectral analysis on the dielectric frequency shift, extract the frequency domain features of the dielectric constant corresponding to each cell of the second grid, and calculate the difference with the preset reference dielectric spectrum to generate a dielectric deviation matrix covering the glass region; wherein, the reference dielectric spectrum is generated based on the dielectric response of the defect-free glass region under the same excitation conditions.

[0085] In this embodiment, the acquired dielectric frequency shift is processed:

[0086] First, spectral feature extraction is performed, specifically including:

[0087] Perform a Fast Fourier Transform (FFT) on the dielectric frequency shift of each second grid cell to extract the amplitude-phase spectrum of the characteristic frequency band (57-63 GHz).

[0088] Calculate the energy value of the resonance peak (integrate the square of the amplitude in the 57-63 GHz band).

[0089] Secondly, a reference dielectric spectrum is constructed, specifically including:

[0090] Dielectric response data of a defect-free glass region under the same excitation conditions are pre-acquired to generate a reference frequency domain template (including the average amplitude spectrum and phase spectrum).

[0091] Next, deviation quantification calculation is performed, specifically including:

[0092] For each grid cell:

[0093] Calculate the relative deviation between the measured energy value and the reference energy value in the frequency band;

[0094] If the phase spectrum shift is greater than 5°, an additional phase deviation weight is added.

[0095] Finally, matrix construction is performed, including:

[0096] The relative deviation is mapped to a two-dimensional matrix (element value = dielectric deviation rate) according to the cell coordinate index of the second grid. The output matrix shares the same physical coordinate system with the stress gradient matrix to ensure spatial alignment.

[0097] S4: Based on the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region, a defect collaborative determination is performed. When the collaborative determination result meets the preset dual-mode conflict condition, a verification instruction is generated.

[0098] Specifically, step S4 includes the following steps:

[0099] S41: Perform a spatial alignment operation on the stress gradient matrix and the dielectric deviation matrix, wherein the element values ​​of the stress gradient matrix are the stress change rates of the cells in the first grid, and the element values ​​of the dielectric deviation matrix are the frequency domain deviations of the dielectric constants of the cells in the second grid.

[0100] In this embodiment, the generated stress gradient matrix of the metal region (element values ​​corresponding to the stress change rate of the first grid cell) and dielectric deviation matrix of the glass region (element values ​​corresponding to the dielectric deviation rate of the second grid cell) are read. A unified physical coordinate system is established with the lower left corner of the substrate as the origin, and the two matrices are mapped to this coordinate system:

[0101] Calculate the physical location of each matrix element based on the center coordinates of the first and second grid cells;

[0102] Non-overlapping regions (such as pure metal areas without glass coverage) are filled with null markers. To address the issue that the mesh resolution of the metal area (1 mm) is higher than that of the glass area (5 mm), bilinear interpolation is performed on the dielectric deviation matrix.

[0103] Map the glass area mesh to the metal area mesh coordinate system;

[0104] High-resolution dielectric deviation data is generated based on interpolation of the centroid coordinates of adjacent glass units. The output is an aligned spatial correlation matrix pair, with each physical coordinate point containing a data set of (stress change rate and dielectric deviation rate), ensuring that subsequent judgments are based on the same spatial location.

[0105] S42: When the number of physical coordinates that meet the dual-mode conflict condition exceeds the number of conflict points, the collaborative judgment result is determined to be a conflict, and a review instruction is generated; wherein the dual-mode conflict condition is defined as: under the same physical coordinate, if the element value of the stress gradient matrix exceeds the first preset threshold and the corresponding element value of the dielectric deviation matrix does not exceed the second preset threshold, or the element value of the stress gradient matrix does not exceed the first preset threshold and the corresponding element value of the dielectric deviation matrix exceeds the second preset threshold.

[0106] The threshold for the number of conflict points is set based on the background noise level when the substrate is defect-free.

[0107] In this embodiment, conflict determination is performed based on the spatially aligned matrix data:

[0108] Conflict condition definition:

[0109] Metallic anomaly: Stress change rate > first threshold (T1 = 40 MPa / μm) and dielectric deviation rate ≤ second threshold (T2 = 0.08);

[0110] Glass anomaly: Stress change rate ≤ T1 and dielectric deviation rate > T2; if either condition is met, it is marked as a conflict point.

[0111] Conflict point statistics and threshold determination:

[0112] Scan all coordinate points and count the total number of conflict points N_conflict;

[0113] Compare the threshold number of conflict points N_threshold (default is 1.5% of the total number of grid points on the substrate):

[0114] If N_conflict > N_threshold, a conflict is determined, and a review instruction is generated.

[0115] Otherwise, no conflict is determined. Threshold setting basis: 1.2 times the number of conflict points with the maximum background noise, obtained by inspecting 10 defect-free substrates.

[0116] Review instruction generation:

[0117] Record the set of physical coordinates of the conflict points;

[0118] Extract the maximum stress anomaly value (to guide the review of the excitation intensity increase);

[0119] Calculate the average dielectric deviation rate (to guide the review of frequency band optimization); output structured instructions to the next step.

[0120] S5: In response to the verification instruction, adjust the initial excitation intensity and initial scanning accuracy of the first grid and the second grid, reacquire the deformation image sequence and dielectric frequency shift and perform verification judgment until the verification judgment result meets the confidence threshold condition, and finally output the defect detection report.

[0121] Specifically, step S5 includes the following steps:

[0122] S51: Based on the verification instruction, increase the initial excitation intensity of the first grid to a first verification intensity, and decrease the initial resolution of the second grid to a second verification accuracy; wherein, the first verification intensity is a first preset multiple of the initial excitation intensity, and the second verification accuracy is a second preset multiple of the initial resolution.

[0123] In this embodiment, the detection parameters are dynamically adjusted based on the set of conflict point coordinates and abnormal parameters (such as maximum stress anomaly value and average dielectric deviation rate) in the review instruction:

[0124] Enhanced excitation intensity in the metal region: The initial excitation intensity (e.g., current intensity I0 = 2A) is increased to the first verification intensity (I1 = 1.5 × I0 = 3A) to enhance the signal-to-noise ratio of the deformation signal in the metal region and ensure that stress changes of minute defects (e.g., 0.1mm level cracks) can be accurately captured.

[0125] Reduced scanning accuracy in the glass area: The initial resolution (e.g., 5.0mm×5.0mm) is reduced to the second verification accuracy (e.g., 10.0mm×10.0mm). The total number of scanning points is reduced by increasing the grid size, thus shortening the verification cycle. At the same time, to avoid missed detections, the original resolution is maintained for scanning the area around the conflict point (e.g., within ±2mm of the conflict point coordinates), forming a hybrid accuracy strategy of "coarse scanning + fine scanning".

[0126] S52: Based on the first verification strength and the second verification accuracy, regenerate a third grid covering the metal area and a fourth grid covering the glass area; wherein the cell size of the third grid is smaller than that of the first grid, and the cell size of the fourth grid is larger than that of the second grid.

[0127] In this embodiment, the detection grid is re-divided according to the adjusted parameters:

[0128] The third grid generation for the metal area: Based on the original boundary coordinates of the metal area, a rectangular grid is divided along the X / Y axis at a new resolution (e.g., 0.8mm × 0.8mm) to ensure that the grid cells completely cover the metal trace area; for fragmented boundary areas (e.g., trace endpoints), an irregular grid cell is automatically generated using the Voronoi diagram algorithm to ensure that all metal areas are completely covered. The size of the third grid cell is smaller than the initial first grid (1.0mm × 1.0mm).

[0129] The fourth mesh generation of the glass area: Based on the original glass area boundary coordinates, triangulation is performed at a new resolution (10.0mm×10.0mm), and adjacent triangular units are merged into a regular polygonal mesh with a size not exceeding the resolution; for the area around the conflict point within 2mm, the original resolution (5.0mm×5.0mm) mesh is forcibly retained, forming a hybrid mesh structure of "coarse mesh (10mm) + fine mesh (5mm)", and the size of the fourth mesh unit is larger than the initial second mesh (5.0mm×5.0mm).

[0130] S53: The third frequency band excitation signal and the fourth frequency band excitation signal are applied sequentially using a time-division control strategy: During the period when the third frequency band excitation signal is applied, the deformation image sequence of the third grid is acquired at a sampling frequency that matches the first verification intensity; During the period when the fourth frequency band excitation signal is applied, the dielectric frequency shift of the fourth grid is acquired at an acquisition window that matches the second verification accuracy.

[0131] In this embodiment, an improved time-sharing control strategy is used to re-collect data:

[0132] Excitation signal configuration: Apply a third-band excitation signal (frequency f3 = 1.2 × f1, such as 12 kHz) to the metal region, and increase the signal strength to I1 = 3 A; apply a fourth-band excitation signal (frequency f4 = 0.8 × f2, such as 40 GHz) to the glass region, and maintain the original power of the glass region (to avoid over-excitation and damage to the glass).

[0133] Data acquisition sequence:

[0134] Metal region scanning: The robotic arm moves along the third grid path, with the dwell time at each grid extended to 5ms (originally 2ms). The high-speed camera (sampling rate 20000fps) and microwave resonator are triggered simultaneously to acquire deformation image sequences (recording the deformation process within 0-5ms) and dielectric frequency shift (recording the weak response of the glass region to metal excitation).

[0135] Glass area scanning: The robotic arm moves along the fourth grid path, with the dwell time at each grid shortened to 3ms (due to reduced resolution). A high-speed camera (sampling rate 5000fps) captures the micro-deformation of the glass area, and a microwave resonator directly measures the dielectric frequency shift of the glass area in the main frequency band (40-45GHz).

[0136] Timing control: The scanning of the metal area and the glass area are performed alternately, with an interval of 2ms. The total scanning cycle = number of third grids × (5ms + 2ms) + number of fourth grids × (3ms + 2ms).

[0137] S54: Perform bimodal feature extraction on the reacquired deformation image sequence and dielectric frequency shift to generate the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region for verification, and perform verification judgment based on the same bimodal conflict condition.

[0138] In this embodiment, bimodal analysis is performed on the data collected for review:

[0139] Stress gradient matrix verification in the metallic region:

[0140] An improved multi-frame difference operation is performed on the deformed image sequence: 10 consecutive frames of images within 0-5ms are selected, the gray-level difference between adjacent frames is calculated, and a high-precision displacement field (resolution 0.1μm) is generated; feature points are tracked by optical flow method, and the X / Y direction displacement vector of each third grid cell is fitted.

[0141] The stress model is recalibrated based on Hooke's law: the actual elastic modulus of the metal layer (e.g., E = 110 GPa for copper substrate) is substituted, the spatial gradient of the displacement field is calculated (Sobel operator kernel size 3×3), and the stress change rate matrix is ​​output (element value = element stress / μm).

[0142] Verification of the dielectric deviation matrix of the glass region:

[0143] Perform wideband FFT analysis on dielectric frequency shift: extend the frequency band to 35-65GHz, extract the amplitude-phase spectrum of the main resonance peak (40-45GHz), and calculate the frequency band energy integral value.

[0144] Reconstructing the reference dielectric spectrum: Based on the dielectric response of the defect-free glass region under the complex excitation condition (f4 = 40 GHz, power as before), a new reference energy value (E_ref = 1200 mV) is generated. 2 ).

[0145] Calculate dielectric deviation rate: the relative deviation between the measured energy value (E_meas) and E_ref, with an additional phase weight (ΔE+5%) when the phase offset is >3°.

[0146] S55: When the number of physical coordinates that meet the conflict conditions in the review judgment result is lower than the threshold of the number of conflict points, the review is deemed to be passed, and a detection report containing defect location and type information is finally output; wherein, the defect type is determined according to the abnormal mode association between the stress gradient matrix and the dielectric deviation matrix.

[0147] In this embodiment, the conflict point verification determination is as follows: the stress gradient matrix generated by the verification is spatially aligned with the dielectric deviation matrix (method as in S41), and the number of physical coordinates N_conflict_re that satisfy the dual-mode conflict condition (T1 = 35 MPa / μm, T2 = 0.06) is counted. If N_conflict_re < N_threshold (0.8 times the original conflict point threshold, such as 1.2% of the total number of grids), the verification is deemed successful; otherwise, S51-S54 are executed repeatedly until the condition is met or the maximum number of verifications (preset 3 times) is reached.

[0148] Defect type association: Determine the defect type based on the anomaly pattern in the final matrix:

[0149] High stress in the metal region (>T1) and normal dielectric (<T2): judged as "microcracks in the metal layer";

[0150] High dielectric deviation in the glass region (>T2) and normal stress (<T1): judged as "glass layer peeling or voids";

[0151] Both modes are abnormal (>T1 and >T2): determined as "metal-glass interface delamination".

[0152] Report output: Generates an inspection report containing the defect location (physical coordinates), type, and severity (stress / dielectric deviation value), and marks it on the substrate design drawing in a visual form. It also outputs the parameter adjustment records and data traceability information during the review process.

[0153] Example 2

[0154] This invention also provides a non-contact defect detection device for glass core packaged substrates, used to perform the aforementioned non-contact defect detection method for glass core packaged substrates, with reference to... Figure 2 As shown, the device includes:

[0155] The mesh generation module 100 is used to divide the surface area of ​​the substrate into a metal area and a glass area according to the metal line distribution of the substrate to be tested, and to divide the metal area into a first mesh and the glass area into a second mesh according to a preset initial resolution.

[0156] The dual-frequency excitation module 200 is used to apply dual-frequency signals to the first grid and the second grid in a time-division manner according to a preset initial excitation intensity, and simultaneously acquire the deformation image sequence corresponding to the first grid and the dielectric frequency shift corresponding to the second grid.

[0157] The dual-modal feature extraction module 300 is used to extract dual-modal features from the deformation image sequence and the dielectric frequency shift, respectively generating the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region.

[0158] The review trigger module 400 is used to perform defect collaborative judgment based on the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region. When the collaborative judgment result meets the preset dual-mode conflict condition, a review instruction is generated.

[0159] The verification response module 500 is used to respond to the verification command, adjust the initial excitation intensity and initial scanning accuracy of the first grid and the second grid, reacquire the deformation image sequence and dielectric frequency shift and perform verification judgment until the verification judgment result meets the confidence threshold condition, and finally output the defect detection report.

[0160] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0161] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0162] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A non-contact defect detection method for glass core packaged substrates, characterized in that, The method includes the following steps: Based on the distribution of metal lines on the substrate to be tested, the surface area of ​​the substrate is divided into a metal area and a glass area, and the metal area is divided into a first grid and the glass area is divided into a second grid according to a preset initial resolution. Dual-frequency signals are applied to the first grid and the second grid in a time-division manner according to a preset initial excitation intensity, and the deformation image sequence corresponding to the first grid and the dielectric frequency shift corresponding to the second grid are acquired simultaneously. Dual-mode feature extraction is performed on the deformation image sequence and the dielectric frequency shift to generate the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region, respectively. Defect co-judgment is performed based on the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region. When the co-judgment result meets the preset dual-mode conflict condition, a verification instruction is generated. In response to the verification command, the initial excitation intensity and initial scanning accuracy of the first grid and the second grid are adjusted, the deformation image sequence and dielectric frequency shift are reacquired and verified, until the verification result meets the confidence threshold condition, and finally the defect detection report is output.

2. The non-contact defect detection method for glass core packaged substrates according to claim 1, characterized in that, The step of dividing the substrate surface area into a metal area and a glass area based on the metal line distribution of the substrate to be tested, and further dividing the metal area into a first grid and the glass area into a second grid according to a preset initial resolution, includes: Import the CAD file of the substrate metal circuit of the substrate to be tested, and parse it to generate the boundary coordinate set of the metal area and the boundary coordinate set of the glass area; The preset initial resolution is loaded, and combined with the corresponding set of boundary coordinates, a first grid covering the metal area and a second grid covering the glass area are generated.

3. The non-contact defect detection method for a glass core encapsulated substrate according to claim 2, characterized in that, The step of applying dual-frequency signals to the first and second grids in a time-division manner according to a preset initial excitation intensity, and simultaneously acquiring the deformation image sequence corresponding to the first grid and the dielectric frequency shift corresponding to the second grid, includes: Based on the boundary coordinate sets of the first grid and the second grid, and combined with the preset initial excitation intensity, a first frequency band excitation signal acting on the metal region and a second frequency band excitation signal acting on the glass region are respectively set, and the frequency bands of the first frequency band excitation signal and the second frequency band excitation signal do not overlap; A time-division control strategy is adopted to sequentially apply the first frequency band excitation signal and the second frequency band excitation signal: during the period when the first frequency band excitation signal is applied, the deformation image sequence of the first grid and the dielectric frequency shift of the second grid are collected at a set sampling frequency.

4. The non-contact defect detection method for a glass core encapsulated substrate according to claim 3, characterized in that, The step of performing dual-mode feature extraction on the deformation image sequence and the dielectric frequency shift to generate the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region, respectively, includes: Multi-frame difference operation is performed on the deformation image sequence to extract the deformation displacement field corresponding to each cell of the first grid, and a stress gradient model of the metal region is constructed based on Hooke's law to calculate the stress gradient matrix covering the metal region; wherein, the stress gradient model takes the gradient value of the deformation displacement field as input and the stress change rate of the grid cell as output. Spectral analysis is performed on the dielectric frequency shift to extract the frequency domain features of the dielectric constant corresponding to each cell of the second grid, and the difference is calculated with the preset reference dielectric spectrum to generate a dielectric deviation matrix covering the glass region; wherein, the reference dielectric spectrum is generated based on the dielectric response of the defect-free glass region under the same excitation conditions.

5. The non-contact defect detection method for a glass core encapsulated substrate according to claim 4, characterized in that, The defect collaborative determination based on the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region, when the collaborative determination result meets the preset dual-modal conflict condition, generates a review instruction, including: A spatial alignment operation is performed on the stress gradient matrix and the dielectric deviation matrix, wherein the element values ​​of the stress gradient matrix are the stress change rates of the cells in the first grid, and the element values ​​of the dielectric deviation matrix are the frequency domain deviations of the dielectric constants of the cells in the second grid. When the number of physical coordinates that meet the dual-mode conflict condition exceeds the threshold of the number of conflict points, the collaborative judgment result is determined to be a conflict, and a review instruction is generated; wherein the dual-mode conflict condition is defined as: under the same physical coordinate, if the element value of the stress gradient matrix exceeds the first preset threshold and the corresponding element value of the dielectric deviation matrix does not exceed the second preset threshold, or the element value of the stress gradient matrix does not exceed the first preset threshold and the corresponding element value of the dielectric deviation matrix exceeds the second preset threshold. The threshold for the number of conflict points is set based on the background noise level when the substrate is defect-free.

6. The non-contact defect detection method for a glass core encapsulated substrate according to claim 5, characterized in that, In response to the verification command, the initial excitation intensity and initial scanning accuracy of the first and second grids are adjusted, the deformation image sequence and dielectric frequency shift are reacquired and verified, until the verification result meets the confidence threshold condition, and finally a defect detection report is output, including: Based on the verification instruction, the initial excitation intensity of the first grid is increased to a first verification intensity, and the initial resolution of the second grid is decreased to a second verification accuracy; wherein, the first verification intensity is a first preset multiple of the initial excitation intensity, and the second verification accuracy is a second preset multiple of the initial resolution; Based on the first verification strength and the second verification accuracy, a third grid covering the metal area and a fourth grid covering the glass area are regenerated; wherein, the cell size of the third grid is smaller than that of the first grid, and the cell size of the fourth grid is larger than that of the second grid; A time-division control strategy is adopted to sequentially apply the third frequency band excitation signal and the fourth frequency band excitation signal: during the period when the third frequency band excitation signal is applied, the deformation image sequence of the third grid is acquired at a sampling frequency matching the first verification intensity; during the period when the fourth frequency band excitation signal is applied, the dielectric frequency shift of the fourth grid is acquired at an acquisition window matching the second verification accuracy. Dual-mode feature extraction is performed on the reacquired deformation image sequence and dielectric frequency shift to generate the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region for verification, and verification is performed based on the same dual-mode conflict condition. When the number of physical coordinates that meet the conflict conditions in the review and judgment results is less than the threshold for the number of conflict points, the review is deemed to be passed, and a detection report containing defect location and type information is finally output; wherein, the defect type is determined based on the abnormal pattern association between the stress gradient matrix and the dielectric deviation matrix.

7. A non-contact defect detection device for glass core packaged substrates, used to perform a non-contact defect detection method for glass core packaged substrates as described in any one of claims 1 to 6, characterized in that, The device includes: The grid generation module is used to divide the surface area of ​​the substrate into a metal area and a glass area according to the distribution of metal lines on the substrate to be tested, and to divide the metal area into a first grid and the glass area into a second grid according to a preset initial resolution. The dual-frequency excitation module is used to apply dual-frequency signals to the first grid and the second grid in a time-division manner according to a preset initial excitation intensity, and simultaneously acquire the deformation image sequence corresponding to the first grid and the dielectric frequency shift corresponding to the second grid; A dual-modal feature extraction module is used to extract dual-modal features from the deformation image sequence and the dielectric frequency shift, generating a stress gradient matrix in the metal region and a dielectric deviation matrix in the glass region, respectively. The review triggering module is used to perform defect collaborative judgment based on the stress gradient matrix of the metal region and the dielectric deviation matrix of the glass region. When the collaborative judgment result meets the preset dual-mode conflict condition, a review instruction is generated. The verification response module is used to respond to the verification command by adjusting the initial excitation intensity and initial scanning accuracy of the first grid and the second grid, re-acquiring the deformation image sequence and dielectric frequency shift and performing verification judgment until the verification judgment result meets the confidence threshold condition, and finally outputting a defect detection report.

Images