A bar wafer pre-scanning method based on laser scanning technology
By acquiring point cloud data of Bar strip wafers using laser scanning technology, and combining point cloud clustering and filtering, the accuracy and robustness issues of detecting high-reflectivity and dark-colored chips in Bar strip wafer pre-scanning are solved, achieving a more stable and faster pre-scanning effect, suitable for three-dimensional pose-sensitive applications.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- 恩纳基智能装备(无锡)股份有限公司
- Filing Date
- 2025-06-25
- Publication Date
- 2026-06-23
Smart Images
Figure CN120746997B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of chip technology, and in particular to a bar-scanning method for wafer pre-scanning based on laser scanning technology. Background Technology
[0002] Optical chips are a crucial chip type in 5G fiber optic communication. In the manufacturing process of optical chips, especially light-emitting chips such as laser diodes and LEDs, the wafer is first diced into multiple long strips (Bars) for easier processing. Each Bar contains multiple chips arranged sequentially, and at this stage, the chips are not yet completely separated. The Bars are then attached to a blue film to form a Bar wafer. After Bar-level testing, the optical port faces of each Bar are coated. A typical approach is to have multiple chips arranged horizontally within each Bar, facilitating coating on the vertical optical port faces and allowing unobstructed light observation in the vertical direction. Similarly, another typical approach is to have multiple chips arranged vertically within each Bar, facilitating coating on the horizontal optical port faces and allowing unobstructed light observation in the horizontal direction. After completing the coating operation on the optical port end face of the Bar bar, each Bar bar is chip-level cut to further cut the multiple chips in each Bar bar, then chip-level testing is completed, and sorting and picking operations are performed.
[0003] To ensure accurate chip sorting and pickup on bar wafers, pre-scanning is required before sorting to determine the pre-scan information of each chip within each bar. A common method for pre-scanning bar wafers, as described in patent publication CN116758076A, involves using a moving camera to acquire local images and combining this with image processing techniques to determine the pre-scan information of each chip within each bar. However, this pre-scanning method requires a customized, complex lighting system and regular calibration to ensure good image acquisition. Furthermore, the presence of scratches, water stains, or other contaminants on the blue film, or optical interference caused by highly reflective (e.g., GaAs) or dark-colored (e.g., SiC) chips within the bar, can lead to poor image processing, resulting in low accuracy and robustness in detecting the pre-scan information of the chips on the bar wafer. Summary of the Invention
[0004] This application addresses the aforementioned problems and technical requirements by proposing a Bar strip wafer pre-scanning method based on laser scanning technology. The technical solution of this application is as follows:
[0005] A bar stripe wafer pre-scanning method based on laser scanning technology, the bar stripe wafer pre-scanning method includes:
[0006] The laser scanner above the bar wafer performs a moving laser contour scan on the local area where each bar is located to obtain point cloud data of the local area where the bar is located.
[0007] Cluster the point cloud data in the local area where the bar is located to obtain multiple point cloud clusters, and filter out the point cloud clusters corresponding to each chip in the bar to obtain the chip point cloud clusters.
[0008] Pre-scan information of each chip within the Bar is obtained by using point cloud data from the point cloud clusters of each chip within the local area where the Bar is located.
[0009] A further technical solution involves selecting point cloud clusters corresponding to each chip within the Bar bar to obtain chip point cloud clusters, including:
[0010] The average depth value of the point cloud data in each point cloud cluster is calculated as the statistical depth value of the point cloud cluster, and point cloud clusters with statistical depth values in the range of [h±△h] are selected as candidate point cloud clusters; where the depth value of the point cloud data is the coordinate of the point cloud data along the vertical direction in the three-dimensional coordinate system of the laser scanner, h is the typical depth value of the chip surface in the Bar strip wafer, and △h is the depth deviation.
[0011] Candidate point cloud clusters that conform to the geometric shape of the chip within the Bar are selected as chip point cloud clusters.
[0012] A further technical solution involves selecting candidate point cloud clusters that conform to the geometric shape of the chip within the Bar as the chip point cloud clusters, including:
[0013] A reference plane is obtained by performing plane fitting on the point cloud data in each candidate point cloud cluster. The point cloud data in the candidate point cloud cluster is then projected onto the corresponding reference plane to obtain the projection mask of the candidate point cloud cluster.
[0014] Candidate point cloud clusters whose projection mask area is within the range of chip area are selected as chip point cloud clusters.
[0015] A further technical solution involves obtaining the projection mask for the candidate point cloud cluster, including:
[0016] Calculate the distance deviation between each point cloud data in the candidate point cloud cluster and the reference plane in the projection direction, and filter out abnormal point cloud data whose distance deviation reaches the deviation threshold to obtain the projection mask of the candidate point cloud cluster on its reference plane after filtering out abnormal point cloud data.
[0017] A further technical solution involves obtaining the projection mask for the candidate point cloud clusters, which also includes:
[0018] The average distance deviation between all point cloud data in the candidate point cloud cluster and the reference plane in the projection direction is calculated to obtain the statistical distance deviation ΔD, and the deviation threshold Dth is determined based on the statistical distance dynamic deviation ΔD.
[0019] The further technical solution is to determine the deviation threshold Dth = D0 * (1 + η * ΔD) based on the statistical distance dynamic deviation ΔD; where D0 is the reference deviation threshold and matches the thickness of the chip on the Bar strip wafer, and η is the sensitivity coefficient with a value range of 0.1 to 0.5.
[0020] A further technical solution involves selecting point cloud clusters corresponding to each chip within the Bar to obtain chip point cloud clusters, which also includes:
[0021] When the area of the projection mask of the candidate point cloud cluster on its reference plane after filtering out abnormal point cloud data is less than the minimum area of the chip area, the candidate point cloud cluster is filtered out.
[0022] When the area of the projection mask of the candidate point cloud cluster after filtering out abnormal point cloud data on its reference plane is greater than the maximum area of the chip area, the projection of the abnormal point cloud data in the candidate point cloud cluster onto the reference plane is fitted with least squares to obtain the projection fitting line. The projection fitting line is used as the projection boundary line to divide the candidate point cloud cluster after filtering out abnormal point cloud data into multiple candidate point cloud clusters. The step of performing plane fitting on the point cloud data in each candidate point cloud cluster to obtain the reference plane is repeated for each divided candidate point cloud cluster.
[0023] The further technical solution is to obtain the pre-scan information of each chip within the Bar bar, including the point cloud cluster for each chip;
[0024] The mechanical coordinates of the chip corresponding to the point cloud cluster are taken as the position of the geometric center of the projection mask of the chip point cloud cluster on its reference plane, offset by D0 / 2 along the normal direction of the reference plane towards the laser scanner; where D0 is the thickness of the chip on the Bar strip wafer.
[0025] Based on the sequence of moving laser contour scanning of the local area where each bar bar is located on the bar bar wafer by the laser scanner, and combined with the mechanical coordinates of the corresponding chips in each chip point cloud cluster within the local area where the bar bar is located, the relative coordinates of the corresponding chips in each chip point cloud cluster within the bar bar are determined.
[0026] The further technical solution is that when the laser scanner performs a moving laser contour scan on the local area where each bar bar is located on the bar bar wafer, the moving laser contour scan is performed from the starting position of the bar bar along the first direction and the scanning length along the first direction is L+d / 2. The first direction is the chip arrangement direction in each bar bar on the bar bar wafer, L is the typical length of the bar bar on the bar bar wafer along the first direction, and d is the typical spacing length between the bar bars on the bar bar wafer along the first direction.
[0027] Determining the relative coordinates of each chip point cloud cluster within the Bar bar includes: calculating the coordinate difference between the mechanical coordinates of the chip corresponding to the chip in the chip point cloud cluster and the starting position of the Bar bar in the first direction, dividing the coordinate difference by the chip's size in the first direction to obtain the relative coordinates of the chip corresponding to the chip point cloud cluster within the Bar bar, and marking chip missing situations within the Bar bar.
[0028] A further technical solution is that obtaining the pre-scan information of each chip within the Bar bar also includes;
[0029] Calculate the angle between the projection mask of each chip point cloud cluster on its reference plane and the vertical direction to obtain the attitude tilt angle of the chip corresponding to the chip point cloud cluster.
[0030] The beneficial technical effects of this application are:
[0031] This application discloses a Bar strip wafer pre-scanning method based on laser scanning technology. This Bar strip wafer pre-scanning method utilizes the significant height difference between the chip surface and the blue film surface, and uses laser scanning technology combined with point cloud processing to perform pre-scanning to obtain pre-scanning information of each chip. Compared with traditional visual scanning technology, laser scanning technology directly measures the surface height difference and is not affected by color or lighting changes. Therefore, it has absolute height perception capability and strong anti-optical interference capability. It does not require the construction of a complex lighting system, and is more stable and robust for the detection of chips on transparent blue films, and even highly reflective chips (such as GaAs) and dark chips (such as SiC).
[0032] Furthermore, laser scanning technology can directly measure tilt, thereby detecting the chip's tilt angle and providing more attitude information for subsequent sorting, which is beneficial to sorting accuracy. This has a significant advantage in three-dimensional attitude-sensitive applications (such as FlipChip packaging). In addition, laser scanning technology acquires full-section data in a single scan, and combined with the point cloud data processing method of this application, it still maintains a high pre-scanning speed, thus improving the overall pre-scanning effect. Attached Figure Description
[0033] Figure 1 This is a diagram showing the arrangement of bar wafers.
[0034] Figure 2 This is a partial structural diagram of a single bar.
[0035] Figure 3 This is a schematic flowchart of a Bar strip wafer pre-scanning method according to an embodiment of this application. Detailed Implementation
[0036] The specific embodiments of this application will be further described below with reference to the accompanying drawings.
[0037] To address the issues of poor detection accuracy and robustness when using vision technology for bar stripe wafer pre-scanning, this application discloses a bar stripe wafer pre-scanning method based on laser scanning technology. The structure of the bar stripe wafer is first described below (please refer to the relevant section). Figure 1 A bar wafer has multiple bar strips 10 arranged in an array and bonded to a blue film 20 to form a bar wafer. That is, multiple bar strip groups are arranged at intervals along a first direction on the bar wafer. Each bar strip group includes multiple bar strips 10 arranged at intervals along a second direction, and each bar strip 10 includes multiple chips arranged sequentially along the first direction. Here, the first and second directions are perpendicular to each other and are two perpendicular directions of the plane containing the bar wafer. Figure 1 Taking the first direction as horizontal and the second direction as vertical as an example.
[0038] The lengths of the bars on the bar wafer are not exactly the same, but they are basically consistent within the tolerance range. Therefore, the typical length L of the bar along the first direction on the bar wafer can be predetermined. For ease of subsequent operations, multiple bar bars 10 arranged along the second direction in the same bar bar group are aligned, and multiple bar bars 10 arranged along the first direction in different bar bar groups are aligned to form an array. Similarly, due to the arrangement deviation when arranging the bar bars on the blue film, the orientation of the bar bars themselves may be deviated. In addition, due to the specification deviation between the bar bars, this alignment is not perfect, but it can be considered aligned within the tolerance range.
[0039] There is a gap between two bar strips spaced along the first direction. The gap between different bar strips may vary slightly, but it is generally consistent within the error range. Therefore, the typical gap length d between bar strips along the first direction on the bar strip wafer can also be predetermined. Similarly, there is a gap between two bar strips spaced along the second direction. The gap between different bar strips may vary slightly, but it is generally consistent within the error range. Therefore, the typical gap length t between bar strips along the second direction on the bar strip wafer can also be predetermined.
[0040] The bar stripe wafer pre-scanning method of this application is applicable not only to the case where the bar stripes are arranged in a rectangular array, but also to the case where the bar stripes are arranged in a circular array, that is, the number of bar stripes contained in different bar stripe groups may be equal or unequal, such as Figure 1 The diagram shows a circular array arrangement, which makes full use of the space on the bar wafer to arrange more bars, thus improving chip manufacturing efficiency.
[0041] The structural diagram of each bar bar 10 is as follows: Figure 2 As shown, Figure 2 (a) in the image is a front view of the Bar bar. Figure 2 Image (b) is a side view of the Bar 10, which includes multiple chips 11 arranged in a row. Adjacent chips 11 are separated by a cleaving channel 12. The chips in the Bar 10 are typically light-emitting chips. Therefore, according to common specifications for light-emitting chips, the transverse width W of chip 11 along the first direction is generally 0.5mm to 1.5mm, the longitudinal height H along the second direction is generally 2 to 3 times the transverse width W, and the thickness D is typically 0.1mm to 0.3mm. The cleaving channel 12 cuts to the blue film position to separate the different chips 11. The width w of the cleaving channel 12 is typically 0.04mm to 0.05mm, and the thickness of the cleaving channel 12 is consistent with the thickness D of the chip.
[0042] Depend on Figure 2 As shown in (b), the surface of chip 11 protrudes compared to the surface of blue film 20. This application utilizes this structural characteristic to replace vision technology with laser scanning technology for Bar strip wafer pre-scanning. The laser scanner, like a traditional camera, is positioned above the Bar strip wafer and vertically downwards, with its optical axis perpendicular to the surface of the Bar strip wafer. Since the height difference between the surface of chip 11 and the surface of blue film 20 is the chip height, the selected laser scanner needs to cover the chip thickness D (0.1mm to 0.3mm) along its axial working range and meet the corresponding accuracy requirements. Many commercially available laser scanners can meet this requirement; a suitable product can be selected, such as the common LJ-X8020.
[0043] A motor drives the relative movement of the laser scanner and the bar wafer, allowing the laser scanner to move to different positions on the bar wafer to perform a moving laser contour scan of the local area where each bar is located, thereby acquiring point cloud data within that local area. After determining the arrangement of the bars on the bar wafer, the local area of each bar can be roughly determined. When performing a moving laser contour scan of the local area of each bar on the bar wafer using the laser scanner: a starting position is pre-specified. This starting position is typically located at the outermost boundary of the bar group in the first direction, or at the beginning of the bar at the outermost boundary in the second direction, usually at the position offset outwards by one distance from the first chip of that bar. Figure 1 As shown, the pre-specified starting position is the distance S offset from the top left corner of the first bar in the leftmost column of the bar wafer. This is also the commonly used starting position for pre-scanning bar wafers. A suitable laser scanner is selected so that the target width H0 of the laser scanner in the second direction is greater than the vertical height H of the chip along the second direction, and less than H+t / 2. This ensures that the target surface of the laser scanner can completely cover the entire area of the bar in the second direction without covering adjacent bar bars. For example, the target width H0 of the LJ-X8020 laser scanner in the second direction is 7mm, which can fully meet the usage requirements of chips with a vertical height H of about 5mm. The same principle applies to other laser scanner models used in other scenarios. Therefore, when performing a moving scan on the area containing a bar, it is only necessary to move the laser contour scan along the first direction from the starting position S of the bar bar. The first direction is the chip arrangement direction of each bar on the bar wafer. The scanning length of the laser scanner along the first direction is L + d / 2, where L is the typical length of the bar stripe on the bar stripe wafer along the first direction, and d is the typical spacing length between the bar stripes on the bar stripe wafer along the first direction. This ensures that the entire area of the bar stripe in the first direction is covered without covering adjacent bar stripes. Figure 1 The shaded area is the area scanned by the moving laser contour of the bar in the upper left corner.
[0044] After completing the moving laser contour scan of the local area where the current bar is located, offsetting the starting position S of the current bar allows for determining the starting position S of the next bar, and the scanning process is repeated similarly. Figure 1For example, keeping the horizontal coordinate of the current bar's starting position S unchanged, the vertical coordinate is shifted downwards by a distance H+t to obtain the starting position of the next bar. After scanning all the bars in the first column, the starting position of the last bar in the first column is shifted to the right by a distance W+d to obtain the horizontal coordinate of the starting position of each bar in the second column. Since the diameter of the bar wafer is known, the boundary position under the current horizontal coordinate can be determined, thus obtaining the starting position S of the second column. Then, by shifting in the same way, the moving laser profile of the local area where the bar is located can be scanned sequentially by row and column scanning. After the bar wafer arrangement structure is determined, the local area where each bar is located can be estimated and determined. This part can be performed according to the existing pre-scanning method, which will not be elaborated here.
[0045] Moving laser contour scanning of the local area containing each bar can acquire point cloud data of the surface of that local area. Therefore, it does not require a complex lighting system and has good resistance to optical interference. It is less affected by lighting and chip reflection, and is more stable in detecting highly reflective chips (such as GaAs) and dark chips (such as SiC). In addition, the moving scanning speed of the laser scanner is faster. Taking the LJ-X8020 as an example, its moving scanning speed is 500mm / s, and it only takes about 0.15s to scan the local area of a 70mm bar. In contrast, using vision technology requires capturing images of each chip one by one. At 40ms / frame, it takes 1.92s to complete the image capture of a common bar containing 48 chips. Therefore, using a laser scanner has a significant speed advantage.
[0046] Each acquired point cloud data point includes its coordinates (x, y, z) in the laser scanner's 3D coordinate system. Typically, through attitude calibration, the laser scanner's 3D coordinate system is aligned with the first direction of the bar wafer, the y-axis with the second direction of the bar wafer, and the z-axis perpendicular to the bar wafer. Therefore, the z-coordinate of the point cloud data is the distance between the point cloud data and the laser scanner. Since the laser scanner acquires surface point cloud data of the area where the bar is located, as... Figure 2 As shown, the chip surface and the blue film surface have different heights, so the distance between the chip surface area and the blue film surface area and the laser scanner is also different. The depth values of the point cloud data in these areas, i.e., the z-coordinates, are different. By clustering the point cloud data in the local area where the Bar bar is located, multiple point cloud clusters are obtained. Then, the point cloud clusters corresponding to each chip in the Bar bar can be further filtered out to obtain the chip point cloud cluster.
[0047] This application does not directly use the depth value z coordinate to filter point cloud data, but first performs clustering. This is because in practical applications, the blue film needs to be tensioned, which will cause the surface of the Bar strip wafer to not be strictly parallel to the horizontal plane. Especially if the chip thickness is small and close to the blue film, the error of directly using the depth value z coordinate for filtering is large. Therefore, this application first performs clustering, which can be implemented using various existing clustering algorithms.
[0048] After obtaining multiple point cloud clusters, the average depth value of the point cloud data in each cluster is first calculated as the statistical depth value of that cluster. Point cloud clusters with statistical depth values within the range of [h ± Δh] are then selected as candidate clusters. Here, the depth value of the point cloud data is the vertical coordinate (z-coordinate) of the point cloud data in the three-dimensional coordinate system of the laser scanner. h is the typical depth value of the chip surface in the Bar strip wafer, which can be determined after assembling the relative positions of the Bar strip wafer and the laser scanner; it represents the theoretical distance between the chip surface in the Bar strip wafer and the laser scanner. Δh is the depth deviation, which can be customized. Δh is used to compensate for actual errors, but it cannot be too large; it needs to ensure that the typical depth value of the blue film surface in the Bar strip wafer is outside the range of [h ± Δh]. This step can initially filter out the blue film area and the dicing area. Since this step only requires rapid initial screening, calculating the average value as the statistical depth value, while not very robust, is fast and can quickly complete the initial screening.
[0049] Then, candidate point cloud clusters that conform to the geometric shape of the chip within the Bar are further selected as chip point cloud clusters. Please combine this with... Figure 3 The flowchart includes:
[0050] (1) Since the candidate point cloud clusters are theoretically corresponding to the chip area, and the point cloud data on the chip surface should be located on the same plane, a reference plane is obtained by performing plane fitting on the point cloud data in each candidate point cloud cluster. In one embodiment, the SVD algorithm is used.
[0051] (2) Project the point cloud data in the candidate point cloud cluster onto the corresponding reference plane to obtain the projection mask of the candidate point cloud cluster.
[0052] One approach is to project all the point cloud data in the candidate point cloud cluster onto the corresponding reference plane to form the region, which is then used as the projection mask for that candidate point cloud cluster.
[0053] Another approach is to consider that the initial screening only roughly removes some point cloud data from the blue film region. Due to the low precision of the initial screening, candidate point cloud clusters often contain point cloud data from the blue film region in addition to the point cloud data from the chip surface. Therefore, to obtain a more accurate projection mask, after projecting the point cloud data in the candidate point cloud cluster onto the corresponding reference plane, the distance deviation between each point cloud data in the candidate point cloud cluster and the reference plane in the projection direction is calculated. Abnormal point cloud data whose distance deviation reaches the deviation threshold are filtered out. Then, the area formed by the candidate point cloud cluster after filtering out abnormal point cloud data on its reference plane is used as the projection mask of the candidate point cloud cluster. This method yields a more accurate projection mask.
[0054] The aforementioned deviation threshold Dth can be set as a reference deviation threshold D0, which matches the thickness of the chip on the Bar wafer and is generally slightly smaller than the chip thickness.
[0055] Alternatively, considering that different Bar strips are subjected to varying tension by the blue film, resulting in differences in surface elevation, a fixed reference deviation threshold D0 is not used. Instead, the average distance deviation between all point cloud data in the candidate point cloud cluster and the reference plane in the projection direction is calculated to obtain the statistical distance deviation ΔD. Then, the deviation threshold Dth is determined based on the statistical distance dynamic deviation ΔD. In one embodiment, the deviation threshold Dth = D0 * (1 + η * ΔD), where η is a sensitivity coefficient with a value ranging from 0.1 to 0.5. This allows for dynamic adjustment of the deviation threshold Dth to achieve more accurate point cloud filtering.
[0056] (3) Candidate point cloud clusters whose projection mask area falls within the chip area range are selected as chip point cloud clusters. Since the chip specifications are predetermined, the theoretical surface area of a single chip is known. The chip area range is defined by setting an appropriate error based on the theoretical surface area of a single chip. When the projection mask area falls within the chip area range, the filtered candidate point cloud clusters are selected as chip point cloud clusters. However, if the projection mask area exceeds the chip area range, there are typically two scenarios:
[0057] 1. When the area of the projection mask of a candidate point cloud cluster after filtering out abnormal point cloud data is smaller than the minimum area within the range of chip area, this is often because there are impurities such as water stains on the blue film. Since water stains also protrude relative to the blue film, the candidate point cloud cluster often corresponds to such impurities. However, the area of such impurities is usually small, so the area of its projection mask will be smaller than the minimum area within the range of chip area. Therefore, in this case, the candidate point cloud cluster can be directly filtered out.
[0058] 2. The area of the projection mask of the candidate point cloud cluster on its reference plane after filtering out abnormal point cloud data is greater than the maximum area range of the chip area. This situation often occurs because the surface undulation of the blue film causes some cutting paths to be unidentified, resulting in the candidate point cloud cluster actually containing point clouds from two chip regions. This leads to the processed projection mask actually covering multiple chip areas, thus exceeding the chip area range. Therefore, in this case, the candidate point cloud cluster cannot be directly filtered out. Instead, the projection of the abnormal point cloud data in the candidate point cloud cluster onto the reference plane is fitted with least squares to obtain a projection fitting line. Then, the projection fitting line is used as the projection boundary line to divide the candidate point cloud cluster after filtering out abnormal point cloud data into multiple candidate point cloud clusters. The step of fitting the plane to obtain the reference plane is then repeated for each of the divided candidate point cloud clusters. Taking the fitted projection line as an example, the point cloud data whose projections on the reference plane are located on both sides of the fitted projection line in the candidate point cloud cluster after filtering out abnormal point cloud data are divided into two candidate point cloud clusters. The same process is applied when there are multiple fitted projection lines.
[0059] The above operations can be used to filter out multiple chip point cloud clusters, and the reference plane of each chip point cloud cluster and the projection mask of the point cloud data in the chip point cloud cluster on their respective reference plane are also known. Then, the pre-scan information of each chip in the bar is obtained by using the point cloud data of each chip point cloud cluster in the local area where the bar is located.
[0060] The pre-scan information acquired for each chip includes its coordinate information, including mechanical coordinates and row and column coordinates. For the mechanical coordinates of the chip, the projection mask of the point cloud data in the chip point cloud cluster onto its respective reference plane is the projection area of the corresponding chip. Therefore, the coordinates of the geometric center of the projection mask of the chip point cloud cluster on its reference plane can be determined using the point cloud data. Then, the position after shifting this geometric center along the normal direction of the reference plane towards the laser scanner by D0 / 2 is the three-dimensional center of the chip. The three-dimensional coordinates at this position are used as the mechanical coordinates of the chip corresponding to the chip point cloud cluster.
[0061] After determining the mechanical coordinates of the chips, the relative coordinates of each chip within a bar are determined by combining the mechanical coordinates of the chips corresponding to each chip point cloud cluster within the local area of the bar strip wafer with the sequence of moving laser contour scanning of the local area of each bar strip by the laser scanner. The row and column coordinates of the chips within a bar strip indicate the bar strip to which the chip belongs and its arrangement order within the bar strip. The row coordinates of the chip's row and column coordinates represent the row coordinates of the bar strip to which the chip belongs, which can be directly determined using existing methods during the scanning process. The column coordinates of the chip's row and column coordinates represent the arrangement order of the chips within the bar strip, including: calculating the coordinate difference between the mechanical coordinates of the chip corresponding to the chip in the point cloud cluster and the starting position of the bar strip in the first direction, dividing the coordinate difference by the chip's size in the first direction to obtain the relative coordinates of the chip corresponding to the point cloud cluster within the bar strip, and marking any missing chips within the bar strip. For example... Figure 2 As shown, due to requirements such as testing gold samples, chips may be missing from the bar. The number and location of these missing chips are uncertain. For example... Figure 2 In the diagram, the three chips shown by the dashed line represent the missing chips within the Bar. This method, when faced with this situation, can detect and record the missing chips, and then... Figure 2 The row and column coordinates of the remaining chips in the Bar are marked as (1,0), (1,1), (1,2), (1,3), (1,6), (1,7)... and the missing chips at (1,4) and (1,5) can be marked.
[0062] In addition to common coordinate information, the pre-scan information of each chip includes the attitude tilt angle of the chip corresponding to the point cloud cluster by calculating the angle between the projection mask of each chip point cloud cluster on its reference plane and the vertical direction. Existing methods often do not obtain the attitude tilt angle because obtaining it through vision technology requires complex image processing calculations and is significantly affected by lens distortion. This application, however, can quickly and effectively extract the attitude tilt angle by fitting the reference plane to calculate the normal vector. Obtaining the attitude tilt angle is beneficial for subsequent grasping posture adjustment during sorting and provides richer pre-scan information.
[0063] This application also compared its method with traditional visual pre-scanning methods through experiments. The experimental results show that the traditional visual detection method has a detection rate of 89% for highly reflective chips, while the method of this application can reach 99.5%. The traditional visual detection method has a detection rate of 82% for dark chips, while the method of this application can reach 98%. The detection rate of this application for both highly reflective and dark chips is superior to that of traditional visual pre-scanning methods.
[0064] Furthermore, when the chip's tilt angle is less than 5°, the detection error using the method described in this application is approximately ±0.1°, while the detection error of traditional vision technology is approximately ±0.3°. When the chip's tilt angle is between 5° and 15°, the detection error using the method described in this application is approximately ±0.3°, while the detection error of traditional vision technology is approximately ±1.0°. When the chip's tilt angle is greater than 15°, the detection error using the method described in this application is approximately ±0.5°, while the detection error of traditional vision technology is approximately ±2.0°. It can be seen that the detection accuracy of this application is superior to that of vision detection technology, and the accuracy advantage becomes more pronounced as the chip's tilt angle increases.
[0065] The above descriptions are merely preferred embodiments of this application, and this application is not limited to the above embodiments. It is understood that other improvements and variations that can be directly derived or conceived by those skilled in the art without departing from the spirit and concept of this application should be considered to be included within the protection scope of this application.
Claims
1. A method for pre-scanning Bar stripe wafers based on laser scanning technology, characterized in that, The Bar strip wafer pre-scanning method includes: The laser scanner above the bar wafer performs a moving laser contour scan on the local area where each bar is located to obtain point cloud data of the local area where the bar is located. The point cloud data within the local area of the bar strip is clustered to obtain multiple point cloud clusters. The average depth value of the point cloud data in each cluster is calculated as the statistical depth value of the cluster, and point cloud clusters with statistical depth values within the range of [h±△h] are selected as candidate point cloud clusters. Here, the depth value of the point cloud data is the vertical coordinate of the point cloud data in the three-dimensional coordinate system of the laser scanner, h is the typical depth value of the chip surface in the bar strip wafer, and △h is the depth deviation. Plane fitting is performed on the point cloud data in each candidate point cloud cluster to obtain a reference plane. The point cloud data in the candidate point cloud cluster is projected onto the corresponding reference plane, and the distance deviation between each point cloud data in the candidate point cloud cluster and the reference plane in the projection direction is calculated. Abnormal point cloud data with distance deviations reaching the deviation threshold are filtered out to obtain the projection mask of the candidate point cloud cluster on its reference plane after filtering out abnormal point cloud data. Candidate point cloud clusters with projection mask areas within the range of chip area are selected as chip point cloud clusters. Pre-scan information of each chip within the Bar is obtained by using point cloud data from the point cloud clusters of each chip within the local area where the Bar is located.
2. The Bar strip wafer pre-scanning method according to claim 1, characterized in that, The Bar strip wafer pre-scanning method further includes: The average distance deviation between all point cloud data in the candidate point cloud cluster and the reference plane in the projection direction is calculated to obtain the statistical distance dynamic deviation ΔD, and the deviation threshold Dth is determined based on the statistical distance dynamic deviation ΔD.
3. The Bar strip wafer pre-scanning method according to claim 2, characterized in that, The deviation threshold Dth=D0 is determined based on the statistical distance dynamic deviation ΔD. (1+η) ΔD), where D0 is the reference deviation threshold and matches the thickness of the chip on the Bar strip wafer, and η is the sensitivity coefficient with a value range of 0.1 to 0.
5.
4. The Bar strip wafer pre-scanning method according to claim 1, characterized in that, The Bar strip wafer pre-scanning method further includes: When the area of the projection mask of the candidate point cloud cluster after filtering out abnormal point cloud data on its reference plane is less than the minimum area of the chip area range, the candidate point cloud cluster is filtered out. When the area of the projection mask of the candidate point cloud cluster after filtering out abnormal point cloud data on its reference plane is greater than the maximum area of the chip area, the projection of the abnormal point cloud data in the candidate point cloud cluster onto the reference plane is fitted with least squares to obtain the projection fitting line. The projection fitting line is used as the projection boundary line to divide the candidate point cloud cluster after filtering out abnormal point cloud data into multiple candidate point cloud clusters. The step of performing plane fitting on the point cloud data in each candidate point cloud cluster to obtain the reference plane is repeated for each divided candidate point cloud cluster.
5. The Bar strip wafer pre-scanning method according to claim 1, characterized in that, The pre-scan information of each chip within the Bar bar is obtained, including the point cloud cluster for each chip; The mechanical coordinates of the chip corresponding to the point cloud cluster are taken as the position of the geometric center of the projection mask of the chip point cloud cluster on its reference plane, offset by D0 / 2 along the normal direction of the reference plane towards the laser scanner; where D0 is the thickness of the chip on the Bar strip wafer. Based on the sequence of moving laser contour scanning of the local area where each bar bar is located on the bar bar wafer by the laser scanner, and combined with the mechanical coordinates of the corresponding chips in each chip point cloud cluster within the local area where the bar bar is located, the relative coordinates of the corresponding chips in each chip point cloud cluster within the bar bar are determined.
6. The Bar strip wafer pre-scanning method according to claim 5, characterized in that, When performing a moving laser contour scan on the local area where each bar bar is located on the bar bar wafer using a laser scanner, the moving laser contour scan is performed from the starting position of the bar bar along a first direction and the scanning length along the first direction is L+d / 2. The first direction is the chip arrangement direction in each bar bar on the bar bar wafer, L is the typical length of the bar bar on the bar bar wafer along the first direction, and d is the typical spacing length between the bar bars on the bar bar wafer along the first direction. Determining the relative coordinates of each chip point cloud cluster within the Bar bar includes: calculating the coordinate difference between the mechanical coordinates of the chip corresponding to the chip in the chip point cloud cluster and the starting position of the Bar bar in the first direction, dividing the coordinate difference by the chip's size in the first direction to obtain the relative coordinates of the chip corresponding to the chip point cloud cluster within the Bar bar, and marking chip missing situations within the Bar bar.
7. The Bar strip wafer pre-scanning method according to claim 1, characterized in that, Obtaining the pre-scan information of each chip within the bar also includes; Calculate the angle between the projection mask of each chip point cloud cluster on its reference plane and the vertical direction to obtain the attitude tilt angle of the chip corresponding to the chip point cloud cluster.