PCB (Printed Circuit Board) back drilling stub length measuring method based on circumference computer layered imaging

Abstract

The invention relates to the technical field of industrial nondestructive testing and precision measurement, in particular to a PCB back drilling hole stub length measuring method based on circumference computer layered imaging, which comprises the following steps: S1, acquiring RCL three-dimensional body data based on a circumference calculation layered imaging technology; s2, positioning a back drilling hole and determining an annular sampling area from the RCL three-dimensional body data; s3, carrying out gray scale statistics layer by layer in the annular sampling area along the thickness direction of the PCB, and generating a gray scale curve; s4, analyzing feature points of the curve, and determining a first feature position corresponding to a signal line layer and a second feature position corresponding to a stump tail end layer; and S5, according to the first feature position and the second feature position, the stub length of the back drilling hole is obtained through calculation. According to the method, the negative influence of aliasing artifacts generated by limited angle scanning on the measurement precision of the length of the stub of the back drilling hole is overcome, and high-precision, high-efficiency and non-destructive stub length measurement is realized.

Description

Technical Field

[0001] This invention relates to the field of industrial non-destructive testing and precision measurement technology, specifically to a method for measuring the length of residual piles in PCB back-drilled holes based on circumferential computer layered imaging. Background Technology

[0002] With the trend towards higher frequencies and higher density interconnects in printed circuit boards (PCBs), back-drilling has become a key technology for ensuring signal integrity. It suppresses signal reflection and crosstalk by removing redundant copper pillars (i.e., studs) in through-holes that are not used for electrical connections. Accurate measurement of stud length (typically between 50 and 300 micrometers) directly affects the performance stability of high-frequency circuits, thus becoming a core aspect of manufacturing quality control. However, this measurement task faces significant challenges: traditional mainstream methods involve destructive physical slicing combined with microscopic observation. While this method provides direct information, it requires cutting and destroying the sample, making it only suitable for offline sampling inspection and unable to meet the urgent needs of modern production lines for full, non-destructive, online inspection of large batches of PCBs. Among non-destructive inspection technologies, machine vision-based optical methods are limited by their penetration capabilities, only able to detect surface or near-surface features, and completely ineffective against stud structures deeply embedded within the board. In recent years, X-ray computed tomography (CT) technology has attracted attention for its three-dimensional non-destructive imaging capabilities. However, it has encountered a fundamental bottleneck when applied to PCBs, a typical high aspect ratio plate structure: when the PCB plane is parallel to the X-ray beam, the excessively long penetration path leads to severe signal attenuation and low signal-to-noise ratio of the projected data. At the same time, due to the physical space limitations of the equipment, it is difficult for the sample to be close enough to the X-ray source to obtain high geometric magnification, resulting in the spatial resolution of the reconstructed image failing to meet the precise metrological requirements of micron-level remnants.

[0003] To address the inherent limitations of CT technology in imaging plate-shaped objects, computational layered imaging (CL) technology, particularly circumferential CL employing a circular scanning trajectory, maintains a specific tilt angle between the X-ray source, sample, and detector during the scan, ensuring the X-ray beam always penetrates the PCB plane at an angle. This geometric optimization significantly shortens the effective penetration thickness, improves the signal-to-noise ratio, and allows the sample to be closer to the X-ray source, thus achieving spatial resolution far exceeding that of conventional CT, theoretically making it an ideal tool for non-destructive testing of PCB internal structures. However, this tilted scanning mode is inherently a finite-angle scan, leading to incomplete projection data. This introduces difficult-to-eliminate, structure-related aliasing artifacts into the reconstructed images. These artifacts manifest as stripes, shadows, or blurring effects, severely superimposing and interfering with the grayscale distribution of the actual PCB interlayer interfaces and the back-drill endpoint region, causing blurred edge localization and the appearance of false features. This significantly limits the accuracy of sub-pixel-level edge extraction and length measurement directly based on CL images.

[0004] To address the aliasing artifact problem in CL images, existing techniques generally follow a "restore first, then measure" approach, mainly falling into two categories: one is mathematical suppression methods based on projection domain filtering or iterative reconstruction algorithms. These methods are often computationally complex and time-consuming, and may lose true detail information due to over-regularization; the other is image post-processing methods based on deep learning (such as convolutional neural networks). While these methods can effectively remove some artifacts, their performance heavily depends on the diversity and representativeness of the training data, posing a risk of model generalization. Furthermore, the artifact removal process itself may smooth or distort true physical edges, introducing new uncertainties to subsequent accurate measurements. Both strategies treat artifacts as pure interference noise, failing to explore and utilize the deep physical correlation between artifact generation and the geometry of the target object, thus facing a dilemma in balancing accuracy and efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a method for measuring the length of PCB back-drilled residual piles based on circumferential computer layered imaging. In the circumferential calculation layered imaging detection of high aspect ratio PCBs, this method overcomes the negative impact of aliasing artifacts caused by limited angle scanning on the measurement accuracy of back-drilled residual pile length, and realizes a high-precision, high-efficiency, and non-destructive online measurement method for residual pile length.

[0006] To achieve the above objectives, a method for measuring the length of residual PCB back-drilled holes based on circumferential computer layer imaging is provided, including the following steps:

[0007] S1: Obtain the three-dimensional volume data of the RCL in the back drill hole area of ​​the PCB under test based on the circumferential calculation layer imaging technology;

[0008] S2: Locate the back drill hole and determine the annular sampling area from the RCL three-dimensional volume data;

[0009] S3: Along the thickness direction of the PCB, perform grayscale statistics layer by layer in the annular sampling area to generate a grayscale curve that varies with depth;

[0010] S4: Based on the correlation between the grayscale distribution of aliasing artifacts in the RCL three-dimensional volume data and the geometric position of the interface between PCB layers, analyze the feature points of the curve, and determine the first feature position corresponding to the signal line layer and the second feature position corresponding to the end layer of the residual pile.

[0011] S5: Calculate the length of the residual pile of the back drill hole based on the first feature position and the second feature position.

[0012] Further, in step S4, the feature points of the grayscale curve are analyzed based on the artifact intensity-interlayer interface transformation model; the artifact intensity-interlayer interface transformation model is represented by the AILT model, which is a quantitative model characterizing the correlation; the establishment of the AILT model includes:

[0013] Based on the tilt angle θ of the system for calculating the circumferential layered imaging, the spatial distribution characteristics of the reconstruction artifacts caused by the scanning geometry are determined; the spatial distribution characteristics of the reconstruction artifacts are manifested as a double-conical intensity field with the spatial point target as the center and a half-apex angle of (90°-θ);

[0014] Based on the distribution characteristics, a quantitative correspondence is established between the location of feature points on the grayscale curve of the annular region extracted from the RCL three-dimensional volume data and the physical location of the signal line layer and the end layer of the residual stack inside the PCB; the location of the feature points on the grayscale curve corresponds to the gradient extreme value region of the intensity field.

[0015] Furthermore, in the circumferential computational layered imaging technology, the acquisition of RCL three-dimensional volume data adopts a circular scanning trajectory, and during the scanning process, the X-ray source, PCB sample and detector maintain a fixed tilt angle θ, so that the X-ray beam and the PCB board plane are at a constant tilt angle.

[0016] Furthermore, step S2 specifically includes:

[0017] S21: Extract a central slice containing the cross-section of the back drill hole from the RCL three-dimensional volume data;

[0018] S22: Perform image preprocessing on the central slice to identify and locate the center and radius of the back drill hole;

[0019] S23: With the center of the circle as the center, define an annular region with an inner radius of R1 and an outer radius of R2 as the annular sampling region.

[0020] Furthermore, in step S22, the image preprocessing includes at least filtering and noise reduction and back-drilled hole edge detection; the back-drilled hole edge detection adopts the Hough circle detection method.

[0021] Furthermore, in step S3, the grayscale statistics are calculated as the average grayscale value of all pixels in the annular sampling area for each layer.

[0022] Furthermore, in step S4, the analysis of the feature points of the curve specifically includes: identifying the peak position of the curve as the first feature position, and identifying the trough position of the curve as the second feature position.

[0023] Furthermore, step S4 also includes sub-pixel localization of the first feature position and / or the second feature position, specifically: selecting local data points near the peak or trough, and fitting the feature position coordinates with sub-pixel accuracy through an interpolation algorithm.

[0024] Furthermore, the interpolation algorithm is a Chebyshev interpolation polynomial.

[0025] Furthermore, in step S5, the formula for calculating the length L of the remaining pile is:

[0026] L = |z1 - z2| × λ

[0027] Where z1 is the depth coordinate of the first feature position, z2 is the depth coordinate of the second feature position, and λ is the resolution of the RCL three-dimensional volume data in the depth direction.

[0028] Principles and advantages:

[0029] 1. In terms of measurement accuracy and resolution, this method fundamentally overcomes the inherent limitations of traditional computed tomography (CT) technology, which suffers from excessively long penetration paths, low signal-to-noise ratios, and insufficient resolution due to the high aspect ratio structure of PCBs. By optimizing the design of the circular computational layer imaging (RCL) scanning geometry, it successfully acquires volume data with ultra-high spatial resolution, capable of clearly resolving residual pile structures at the 50-300 micrometer scale. Furthermore, based on the constructed artifact intensity-interlayer interface transition (AILT) model, this method can achieve sub-pixel-level edge localization from images containing artifacts, reducing measurement uncertainty to a level better than that of a single voxel. This meets the stringent process requirements of micrometer-level precision control of residual pile length in high-frequency, high-speed PCBs, surpassing existing non-destructive testing technologies in terms of accuracy.

[0030] 2. This method successfully achieves a fundamental shift from destructive sampling to non-destructive online full inspection. Compared to traditional physical slicing methods that require destroying samples and can only be used for offline sampling, this invention completely preserves the integrity of the PCB under test, allowing it to be 100% returned to the production line. Furthermore, compared to computationally time-consuming iterative reconstruction artifact removal methods or deep learning artifact removal schemes that require large amounts of training data and have generalization risks, this method directly utilizes analytically reconstructed images to extract measurement information through an efficient algorithm, significantly shortening the single analysis cycle. This makes it possible to perform rapid, full inspection of every high-density PCB on the production line, thereby improving the reliability of quality control while ensuring production efficiency.

[0031] 3. This invention creatively overcomes the industry-wide problem of aliasing artifacts caused by limited-angle sampling in circumferential CL (Radar Clone) technology interfering with measurement accuracy, and opens up a completely new solution path. Unlike existing technologies that simply treat artifacts as harmful noise and focus on the complex approach of "removing them first and then measuring," this invention, through in-depth analysis of the physical essence of RCL imaging, discovers and quantitatively proves a strong correlation between the grayscale distribution characteristics of aliasing artifacts and the geometric position of the interlayer interfaces within the PCB. This transforms the traditional "interference signal" into a "feature beacon" that can be used for precise positioning. This paradigm shift from "artifact suppression" to "artifact utilization" is the source of the method's high precision and high efficiency.

[0032] 4. This invention provides a complete, stable, and efficient solution that seamlessly integrates high-resolution RCL data acquisition, physical model-based intelligent feature analysis, and sub-pixel computation throughout the entire process. This solution not only combines the advantages of being non-destructive, high-resolution, and adaptable to board structures, but also finds the optimal balance between accuracy and efficiency, avoiding the inherent shortcomings of traditional methods in terms of computational complexity, detail fidelity, or model generalization. Therefore, this invention provides an unprecedentedly reliable online inspection method for the mass production of high-density interconnect PCBs, and has significant industrial application value in ensuring the signal integrity and production yield of high-end electronic products. Attached Figure Description

[0033] Figure 1 This is a flowchart of a method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging, according to an embodiment of the present invention.

[0034] Figure 2 This is a schematic diagram of the RCL system.

[0035] Figure 3 Frequency domain analysis for RCL;

[0036] Figure 4 Reconstructing the intensity field for a spatial mass: (a) Oblique diaxial view; (b) Front view; (c) Top view.

[0037] Figure 5 This is the grayscale curve for the ROI region;

[0038] Figure 6 This is a schematic diagram of the sampling area;

[0039] Figure 7 This is a schematic diagram of the simulation model;

[0040] Figure 8 The RCL scan reconstruction results of the simulation phantom are shown in (a) volumetric data, (b) z-slice, and (c) x-slice.

[0041] Figure 9This is a schematic diagram showing the grayscale curve of the sampling area and the location of the key layer. Detailed Implementation

[0042] The following detailed description illustrates the specific implementation method:

[0043] Example

[0044] A method for measuring the length of residual PCB back-drilled holes based on circumferential computer layered imaging, basically as follows: Figure 1 As shown, it includes the following steps:

[0045] S1: Circular computational layered imaging (RCL) technology is used to acquire the three-dimensional volumetric data of the RCL in the back-drilled area of ​​the PCB under test. In this embodiment, the acquisition of the RCL three-dimensional volumetric data in the circular computational layered imaging technology adopts a circular scanning trajectory, and during the scanning process, the X-ray source, the PCB sample, and the detector maintain a fixed tilt angle θ, so that the X-ray beam and the PCB board plane are at a constant tilt angle. The structural schematic diagram is shown below. Figure 2 As shown.

[0046] Based on the Fourier central slice theorem, the essence of CL imaging is to acquire the Fourier domain information of an object through projection data, and then reconstruct the object's three-dimensional structure through inverse Fourier transform. RCL, as a special scanning mode of CL, directly determines the quality of the reconstructed image through its frequency domain sampling characteristics, requiring analysis based on the Fourier space sampling rules. For any three-dimensional object, its spatial domain grayscale distribution can be expressed as... ,in For a spatial position vector, the corresponding three-dimensional Fourier transform is:

[0047] MACROBUTTON MTPlaceRef * MERGEFORMATSEQ MTEqn h *MERGEFORMAT(SEQ MTEqn c * Arabic * MERGEFORMAT1)

[0048] In the formula, It is the Fourier domain spatial frequency vector. These correspond to the spatial frequencies in the x, y, and z directions, respectively. It is a scalar product.

[0049] During CL imaging, each projected image corresponds to The sampled values ​​on a certain "Fourier slice". However, although the circular scanning trajectory of RCL can achieve 360° projection acquisition, it is limited by the tilt angle θ. The actual effective sampling angle range cannot cover the entire Fourier frequency domain, resulting in some frequency components not being acquired, forming "unsampled areas". This is the root cause of aliasing artifacts in RCL images.

[0050] S2: Locate the back drill hole and determine the annular sampling area from the RCL three-dimensional volume data; step S2 specifically includes:

[0051] S21: Extract a central slice containing the cross-section of the back drill hole from the RCL three-dimensional volume data; extract a central slice along the axis of the back drill hole as a reference for the subsequent positioning of the annular sampling area. This slice should clearly show the circular cross-section of the back drill hole.

[0052] S22: Perform image preprocessing on the central slice to identify and locate the center and radius of the back drill hole; in step S22, the image preprocessing includes at least filtering and denoising, adaptive binarization and back drill hole edge detection, etc., with the aim of eliminating noise interference and highlighting the geometric features of the back drill hole; the back drill hole edge detection adopts the Hough circle detection method.

[0053] S23: Based on the results of the Hough circle detection, a ring-shaped region with an inner radius of R1 and an outer radius of R2 is defined as the ring-shaped sampling region, centered on the center of the circle. Figure 6 As shown. The average grayscale value of this annular sampling area can directly reflect the pseudo-effect strain changes at the interlayer interface.

[0054] S3: Along the PCB thickness direction, perform grayscale statistics layer by layer within the annular sampling area to generate a grayscale curve showing the grayscale variation with depth. In this embodiment, through grayscale statistics and curve analysis of the annular sampling area, establish the correlation between "grayscale features and layer positions" to locate the key signal line layer and the end layer of the residual pile required for measurement. Within the ROI volume data, along the Z-axis, i.e., the PCB thickness direction, extract the grayscale values ​​of the annular sampling area layer by layer and calculate the average grayscale value of each layer. Plot a grayscale curve with the layer number Z as the abscissa and the average grayscale value as the ordinate.

[0055] S4: Based on the correlation between the grayscale distribution of aliasing artifacts in the RCL 3D volume data and the geometric position of the interface between PCB layers, analyze the feature points of the curve to determine the first feature position corresponding to the signal line layer and the second feature position corresponding to the end layer of the residual pile. The analysis of the feature points of the curve specifically includes: identifying the peak position of the curve (the local maximum point of the grayscale curve) as the first feature position, and identifying the trough position of the curve (the local minimum point) as the second feature position. Sub-pixel localization is performed on the first feature position and / or the second feature position, specifically: selecting local data points near the peak or trough (selecting a predetermined number of continuous data points including the local maximum or local minimum point), and fitting the feature position coordinates (extreme point coordinates) with sub-pixel precision using an interpolation algorithm.

[0056] (z1, z2). The interpolation algorithm is a Chebyshev interpolation polynomial.

[0057] In step S4, the feature points of the grayscale curve are analyzed based on the artifact intensity-interlayer interface transformation model. The artifact intensity-interlayer interface transformation model is represented by the AILT model, which is a quantitative model characterizing the correlation. The AILT model is established based on the following simple principle: the RCL system forms a pair of double cones with the origin of the frequency space as vertices in the unsampled region of the Fourier domain. The half-vertex angle (90°-θ) of these double cones is determined by the system tilt angle θ. This frequency domain deficiency causes the reconstructed response of spatial domain point targets to present a double-conical artifact intensity field with a half-vertex angle of (90°-θ). The AILT model equates the interlayer interface of the PCB to a linear combination of multiple spatial point targets. The overall artifact effect of the interlayer interface should be a linear superposition of the artifact intensity fields of each point target, and the actual position of the interlayer interface is related to the gradient extremum of the superimposed artifact intensity field. The establishment of the AILT model specifically includes the following steps:

[0058] Based on the tilt angle θ of the system for calculating the circumferential layered imaging, the spatial distribution characteristics of the reconstruction artifacts caused by the scanning geometry are determined; the spatial distribution characteristics of the reconstruction artifacts are manifested as a double-conical intensity field with the spatial point target as the center and a half-apex angle of (90°-θ);

[0059] Based on the distribution characteristics, a quantitative correspondence is established between the location of feature points on the grayscale curve of the annular region extracted from the RCL three-dimensional volume data and the physical location of the signal line layer and the end layer of the residual stack inside the PCB; the location of the feature points on the grayscale curve corresponds to the gradient extreme value region of the intensity field.

[0060] For PCB imaging, the interlayer interfaces (such as the interface between signal line layers and residual pile layers) are essentially "edge structures" with abrupt changes in grayscale, which can be equivalent to the superposition of multiple point targets. According to the principle of linear imaging, the RCL artifact of the interface is the result of the superposition of the artifacts of each point target, and the maximum value of its intensity gradient corresponds exactly to the actual position of the interface.

[0061] To verify this property, two types of regions of interest were selected along the z-axis in the intensity field: one type passes through the centroid (ROI #1), and the other type passes beside the centroid (ROI #2). Their grayscale variation curves are shown below. Figure 5As shown in the figure. The results indicate that if the ROI crosses the centroid (target edge), the z-coordinate corresponding to the centroid is the peak value of the grayscale curve; if the ROI passes beside the centroid, the z-coordinate corresponding to the centroid is the valley value of the grayscale curve. This grayscale variation pattern directly establishes a quantitative correlation between "artifact grayscale features" and "interlayer interface position," providing experimental theoretical support for "locating key layers through grayscale curve peaks / valleys" in subsequent measurement algorithms.

[0062] Peak location: Corresponds to the signal line layer. The signal line is made of metal, and its grayscale value is significantly higher than that of the surrounding substrate. Furthermore, artifacts at the interlayer interfaces cause the grayscale curve to form a peak. Figure 5 The grayscale features of ROI#1 are consistent;

[0063] The trough location corresponds to the end layer of the remaining pile. The end of the remaining pile is the boundary between the interlayer blank area and the remaining pile layer. Artifact superposition reduces the gray value to a minimum, and... Figure 5 The grayscale features of ROI#2 are consistent.

[0064] Since the sampling interval of the grayscale curve is "1 layer," that is, the z-axis resolution of 1 pixel, if the length of the remaining stake is calculated in layers, the upper limit of measurement accuracy will be determined by the device resolution. Sub-pixel-level positioning can be achieved through interpolation, meeting the requirements for high-precision measurement. Local data points near the peaks or troughs of the grayscale curve are selected, such as grayscale values ​​of 3-5 layers on each side of the peak, and Chebyshev interpolation polynomials are used for fitting. The precise z-coordinates corresponding to the maximum (peak) and minimum (trough) grayscale values ​​are solved using the interpolation polynomials, i.e., the sub-pixel position z1 of the signal line layer and the sub-pixel position z2 of the remaining stake end layer.

[0065] S5: Calculate the residual pile length of the back-drilled hole based on the first feature position and the second feature position. In step S5, the formula for calculating the residual pile length L is:

[0066] L = |z1 - z2| × λ

[0067] Where z1 is the depth coordinate of the first feature position, z2 is the depth coordinate of the second feature position, and λ is the resolution of the RCL three-dimensional volume data in the depth direction (z-axis).

[0068] The Artifact Intensity-Interlayer Interface Transition Model (AILT Model) is based on the measurement principle of artifact spatial characteristics. (Detailed explanation follows)

[0069] The core foundation for achieving high-precision non-destructive measurement in this method lies in the innovative understanding and utilization of the physical nature of aliasing artifacts in circular computational layered imaging (RCL). Traditional methods treat them as interference noise, while this invention, through theoretical modeling and simulation analysis, reveals and verifies the deterministic and quantitative correlation between artifacts and the structure under test. This principle forms the core of the AILT (Artifact Intensity-Interlayer Interface Transformation) model and directly guides all subsequent image processing and calculation steps. The details are as follows:

[0070] (1) Deterministic spatial origin of RCL system artifacts

[0071] Aliasing artifacts are not random noise; their generation has a definite physical root. For example... Figure 3 As shown, according to the Fourier central slice theorem, the RCL system, due to its fixed tilt angle θ for circular scanning, inherently contains an "unsampled region" in its three-dimensional frequency space sampling. This region is shaped like a pair of double cones with the origin as their vertices. This incompleteness of the frequency domain data is the fundamental reason for the structured artifacts generated in the spatial domain reconstructed image. The key technical feature is that the geometry of this unsampled double-conical region (especially its half-vertex angle) is uniquely determined by the system hardware parameter—the tilt angle θ. This means that for a given θ, the spatial distribution pattern of its imaging artifacts is deterministic and predictable, providing a prerequisite for the invention to utilize artifacts.

[0072] (2) The pseudo-effect of point targets should be: the basic model of spatial features

[0073] To quantitatively understand the spatial distribution of artifacts, a three-dimensional Dirac function was used. The ideal point target is taken as the analysis object. It is considered as the "basic building block" of the PCB interlayer interface. By analyzing its RCL reconstruction response, the transmission relationship between artifact intensity and target edge is derived. The specific derivation process is as follows:

[0074] Define the biconical mask function This function describes the frequency domain sampling characteristics of RCL and identifies the frequency region that is effectively sampled in the Fourier domain. Its expression is:

[0075]

[0076] In the formula, This indicates that the frequency point has been effectively sampled. This corresponds to the unsampled double-cone region.

[0077] Three-dimensional Dirac function The ideal Fourier transform is However, due to the limitations of RCL frequency domain sampling, the actual obtainable Fourier data is:

[0078]

[0079] Based on the Fourier imaging principle, RCL reconstructs the image. for The three-dimensional inverse Fourier transform, namely:

[0080]

[0081] To simplify the calculation of equation (4), both the Fourier domain and the spatial domain are converted to cylindrical coordinates. The axisymmetric property separates the integral variables:

[0082] 1. Fourier domain cylindrical coordinates: , where radial frequency , The azimuth angle in the plane ( Jacobi determinant is ;

[0083] 2. Spatial domain cylindrical coordinates: , where radial distance , for Azimuth angle in the xy plane .

[0084] because Only with , Related (with) (Irrelevant), the azimuth integral in equation (4) can be simplified using the integral expression of the Bessel function:

[0085]

[0086] In the formula, It is a zeroth-order Bessel function of the first kind, used to describe the spatial distribution of axisymmetric fields.

[0087] Substituting the azimuth integral result into equation (4), the reconstructed image is simplified to a two-dimensional integral:

[0088]

[0089] Based on theoretical derivation (Formulas 1-6) and its numerical simulation results, ( Figure 4 In RCL reconstructed images, a point target is not represented as a single point, but rather as a double-conical grayscale intensity field centered on the point target and with a regular spatial shape. The core technical characteristics of this intensity field can be summarized as follows:

[0090] 1. Geometric boundary: The peak position of the artifact satisfies (Right now ), corresponding to the z-axis as the central axis and the half-vertex angle The double conical surface;

[0091] 2. Intensity attenuation law: The intensity of the artifact increases with the radial distance from the center point. The intensity of artifacts increases and decreases, and the rate of decrease along the z-axis is proportional to z. Therefore, the artifact intensity is higher in the region near the edge of the target, while the artifact interference in the region far from the edge is negligible.

[0092] 3. Physical Significance: This point response function constitutes the "basic unit" for analyzing more complex structures (such as PCB interfaces). It proves that artifacts are deterministic signals with specific shapes generated after the target's geometric position passes through the system's imaging model.

[0093] 3) From point targets to inter-layer interfaces: Establishing association rules

[0094] The interlayer interfaces of a PCB (such as the boundary between signal lines and dielectric) can be modeled as a set of continuously distributed point targets. According to linear system theory, the overall artifact stress of this interface is the linear superposition of the artifact intensity fields of all point targets constituting the interface.

[0095] Based on this principle, by analyzing the grayscale changes along different paths through the interface, core association rules that can be directly used for measurement can be derived:

[0096] When the measurement path (such as a ring-shaped sampling area) passes through a high-contrast interface (such as a signal layer), artifacts with the same phase characteristics generated by all point targets at that interface will constructively superimpose, thus forming a significant local maximum (peak) on the grayscale curve along the depth direction (Z-axis). Figure 5 (The case of ROI #1).

[0097] When the measurement path passes beside a low-grayscale area and approaches an interface (such as the end of a pile remnant), the change in the artifact intensity field gradient experienced by points along the path will cause a local minimum (trough) to appear on the grayscale curve (corresponding to...). Figure 5 (The case of ROI #2).

[0098] (4) Summary of the essence and measurement rules of the AILT model

[0099] In summary, the AILT model constructed in this invention is essentially a computational model that encapsulates a complete knowledge chain encompassing "system geometric parameters (θ) → artifact spatial distribution characteristics (double conical field) → interface grayscale response rules (peaks / troughs)". It transforms the physical information implicit in complex images into explicit, programmable judgment logic.

[0100] Signal layer positioning rule: On the depth-grayscale curve of the annular sampling area, find the peak position, which is the precise coordinate of the signal layer on the Z-axis.

[0101] The positioning rule for the end layer of the residual pile is as follows: On the same curve, find the specific trough position that follows immediately. This position corresponds to the precise coordinates of the end of the residual pile (i.e., the end point of back drilling) on ​​the Z-axis.

[0102] The "grayscale curve extraction," "feature point analysis," and "subpixel localization" steps in this embodiment are all strictly performed based on the above physical principles and quantitative rules. This ensures that while avoiding complex and time-consuming artifact removal processes, this method can directly extract subpixel-precision dimensional information from RCL images containing artifacts, achieving a balance between measurement accuracy and processing efficiency.

[0103] To verify the correctness and accuracy of the measurement method proposed in this patent, a simulated back-drilled hole model experiment was designed. The performance indicators of the method were quantified through RCL scan reconstruction, algorithm flow verification, and comparison with the ground truth (GT). A simulation model was constructed based on the layer structure of an actual high-density PCB to simulate the real geometry and material properties of back-drilled holes, residual spikes, signal lines, and vias. Figure 7 As shown.

[0104] The experiment used the TIGRE (Tomographic Iterative GPU-based Reconstruction Engine) open-source library for tomographic reconstruction to complete the scanning and reconstruction process of the simulation phantom. First, a scanning physical model conforming to the characteristics of the RCL system was built based on the library, accurately simulating the circular scanning trajectory of the X-ray source and detector moving synchronously at a fixed tilt angle, generating a projection dataset containing back-drilled holes and attenuation information of each structural layer. Then, the dedicated RCL analytical reconstruction module within the library was called to reconstruct the acquired projections. The reconstruction results... Figure 8 As shown.

[0105] against Figure 8 Based on the reconstruction results, this method is used to extract grayscale curves and locate key positions, such as... Figure 9 As shown, the peak position of the grayscale curve of the CL_FDK reconstructed image (layer 81.1) perfectly matches the signal line layer defined by the simulation model, and the trough position (layer 30) is consistent with the starting layer of the residual pile layer, proving the effectiveness of this method in locating key layers through "artifact grayscale features".

[0106] This paper proposes an innovative measurement scheme based on Residual Cavity (RCL) technology and the grayscale characteristics of artifacts, addressing the problems of destructiveness, insufficient resolution, and artifact processing deficiencies in traditional methods for measuring the length of residual stubs in high-density PCB back-drilled holes. Breaking away from the traditional approach of "removing artifacts before measurement," this scheme deeply analyzes the Fourier domain principle of RCL imaging and the causes of artifacts, discovering and establishing a strong correlation between the intensity of CL aliasing artifacts and the geometric features of PCB interlayer interfaces. An AILT model is constructed, enabling the reverse utilization of artifact information rather than its removal. Experiments using the TIGRE reconstruction library to perform RCL scanning and analytical reconstruction of a simulation model verify the accuracy of key layer localization, demonstrating that this method can achieve high-precision measurement of residual stub length without complex artifact removal. It overcomes the destructiveness of physical slicing methods and the resolution limitations of CT technology, while avoiding the computational redundancy and detail loss problems of traditional artifact processing methods. Overall, this measurement solution combines the advantages of non-destructive testing, high resolution, and high efficiency, and has good engineering adaptability. It can effectively meet the online full inspection requirements of high-density PCBs and provide reliable technical support for ensuring signal integrity and mass production quality control of high-frequency PCBs.

[0107] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, based on the guidance provided in this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for measuring the length of residual PCB back-drilled holes based on circumferential computer layered imaging, characterized in that, Includes the following steps: S1: Obtain the three-dimensional volume data of the RCL in the back drill hole area of ​​the PCB under test based on the circumferential calculation layer imaging technology; S2: Locate the back drill hole and determine the annular sampling area from the RCL three-dimensional volume data; S3: Along the thickness direction of the PCB, perform grayscale statistics layer by layer in the annular sampling area to generate a grayscale curve that varies with depth; S4: Based on the correlation between the grayscale distribution of aliasing artifacts in the RCL three-dimensional volume data and the geometric position of the interface between PCB layers, analyze the feature points of the curve, and determine the first feature position corresponding to the signal line layer and the second feature position corresponding to the end layer of the residual pile. S5: Calculate the length of the residual pile of the back drill hole based on the first feature position and the second feature position.

2. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 1, characterized in that: In step S4, the feature points of the grayscale curve are analyzed based on the artifact intensity-interlayer interface conversion model; the artifact intensity-interlayer interface conversion model is represented by the AILT model, and the AILT model is a quantitative model characterizing the correlation. The establishment of the AILT model includes: Based on the system tilt angle θ of the circumferentially calculated layered imaging, the spatial distribution characteristics of the reconstruction artifacts caused by the scanning geometry are determined. The spatial distribution characteristics of the reconstructed artifacts are manifested as a double-conical intensity field centered on the spatial point target with a semi-vertical angle of (90°-θ); Based on the distribution characteristics, a quantitative correspondence is established between the location of feature points on the grayscale curve of the annular region extracted from the RCL three-dimensional volume data and the physical location of the signal line layer and the end layer of the residual stack inside the PCB; the location of the feature points on the grayscale curve corresponds to the gradient extreme value region of the intensity field.

3. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 2, characterized in that: In the circular computational layered imaging technology, the acquisition of RCL three-dimensional volume data adopts a circular scanning trajectory, and during the scanning process, the X-ray source, PCB sample and detector maintain a fixed tilt angle θ, so that the X-ray beam and the PCB board plane are at a constant tilt angle.

4. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 1, characterized in that: Step S2 specifically includes: S21: Extract a central slice containing the cross-section of the back drill hole from the RCL three-dimensional volume data; S22: Perform image preprocessing on the central slice to identify and locate the center and radius of the back drill hole; S23: With the center of the circle as the center, define an annular region with an inner radius of R1 and an outer radius of R2 as the annular sampling region.

5. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 4, characterized in that: In step S22, the image preprocessing includes at least filtering and noise reduction and back-drilled hole edge detection; the back-drilled hole edge detection adopts the Hough circle detection method.

6. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 5, characterized in that: In step S3, the grayscale statistics are calculated by taking the average grayscale value of all pixels in the annular sampling area for each layer.

7. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 6, characterized in that: In step S4, the analysis of the feature points of the curve specifically includes: identifying the peak position of the curve as the first feature position, and identifying the trough position of the curve as the second feature position.

8. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 7, characterized in that: Step S4 further includes sub-pixel localization of the first feature position and / or the second feature position, specifically: selecting local data points near the peak or trough, and fitting the feature position coordinates with sub-pixel accuracy through an interpolation algorithm.

9. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 8, characterized in that: The interpolation algorithm is the Chebyshev interpolation polynomial.

10. The method for measuring the length of PCB back-drilled hole residual piles based on circumferential computer layered imaging according to claim 8, characterized in that: In step S5, the formula for calculating the length L of the residual pile is: L = |z1 - z2| × λ, where z1 is the depth coordinate of the first feature position, z2 is the depth coordinate of the second feature position, and λ is the resolution of the RCL three-dimensional volume data in the depth direction.

Images