Semiconductor die bonding machine and control method based on intelligent visual control

Abstract

The present invention discloses a semiconductor die bonding machine and control method based on intelligent visual control, comprising: a vacuum adsorption bonding head, a chip transfer platform, a visual inspection module, a dispensing module, an angle correction module, and a die bonding module; based on the high-response and high-precision positioning and detection of the visual inspection module, angle deviation correction is performed during the chip grabbing process to achieve high-precision dispensing and improve the efficiency and accuracy of die bonding.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor packaging, and in particular to a semiconductor die bonder based on intelligent visual control and a control method thereof. Background Art

[0002] Semiconductor die bonders, crucial equipment in the packaging process, undertake the crucial task of precisely securing the chip to the substrate. With the widespread adoption of packaging technologies such as flip-chip packaging (Flip Chip), wafer-level packaging (WLP), and system-in-package (SiP), the requirements for die bonders' accuracy, speed, and intelligence are steadily increasing. To achieve efficient signal transmission, the die bonder must precisely place the chip on the substrate, and any bond deviation must be precisely controlled, otherwise it will affect signal stability and transmission speed. In the packaging of GPU chips for AI servers, the high-density chip integration requires die bonders with ultra-high positioning accuracy and fast bond speeds to meet the high-speed inter-chip communication requirements of large-scale parallel computing.

[0003] Existing technologies present the following technical challenges: 1. High-precision die pick-up stability issues: Traditional die bonders are prone to die pick-up failure or chip damage during the die pick-up process due to unstable chip suction and die head vibration. 2. High-precision positioning issues for the die bond and supply platforms: With the increasing demand for semiconductor packaging efficiency, the die bond and supply platforms must achieve precise positioning while in high-speed motion. However, factors such as inertial forces and vibration generated by high-speed motion can affect platform positioning accuracy. Balancing speed and precision is a key challenge facing this project. 3. High-response, high-precision positioning and inspection issues for the vision system: During the die bond process, the vision system must quickly and accurately identify the position, shape, and size of the chip and substrate, performing high-precision positioning. Furthermore, after die bond, the chip position and bond quality must be rapidly inspected. Traditional vision systems lack the processing speed and accuracy to meet the high-precision and high-stability requirements for die bond processing. 4. Correction issues for chip angular deviation: Chips may experience angular deviation during the gripping and transfer process. If not corrected, this can affect die bond accuracy and product quality. 5. High-precision dispensing: Precise control of the dispensing volume is crucial. Too much or too little glue can affect the bonding between the chip and substrate, as well as electrical performance. 6. Equipment intelligence and software collaboration: Achieving intelligent equipment operation requires a software system that can precisely control and monitor each device component, while also providing automatic diagnosis and optimization capabilities. Summary of the Invention

[0004] The present invention aims to at least partially address one of the technical problems encountered in the aforementioned technologies. To this end, the present invention provides a semiconductor die bonder and control method based on intelligent visual control. This method utilizes a visual inspection module for high-response, high-precision positioning and detection, corrects for angular deviations during chip capture, and achieves high-precision dispensing, thereby improving the efficiency and accuracy of die bonding.

[0005] To achieve the above objectives, the present invention proposes a semiconductor die bonder based on intelligent visual control, comprising: a vacuum adsorption bonding head, a chip transfer platform, a visual inspection module, a dispensing module, an angle correction module, and a die bonder module; wherein,

[0006] The visual inspection module is used to perform positioning inspection on the chip placed on the chip transfer platform to determine the chip's position and posture information;

[0007] The vacuum adsorption head is used to grab the chip according to its position and posture information and move the chip to the dispensing position;

[0008] The dispensing module is used to control the position of the dispensing head and the amount of glue according to the preset program, and to dispense glue at the specified position of the substrate;

[0009] The angle correction module is used to collect chip images in real time based on the flying camera at the bottom of the chip and determine the chip's angle deviation information when the vacuum adsorption bonding head moves the chip to the dispensing position. The servo motor drives the vacuum adsorption bonding head and corrects the chip's angle based on the deviation information to keep the chip in the target posture.

[0010] The die bonding module is used to place the chip in the target posture on the substrate based on multi-axis linkage control after the vacuum adsorption bonding head moves the chip to the dispensing position and perform the die bonding operation.

[0011] According to some embodiments of the present invention, the visual detection module is further configured to:

[0012] Get the image of the vacuum adsorption head grabbing the chip;

[0013] Identify the captured image to determine whether the chip is successfully captured and whether there is any damage.

[0014] According to some embodiments of the present invention, the system further includes: a quality inspection module for inspecting the die bonding quality of the chip after the die bonding operation; the die bonding quality inspection includes chip position inspection, die bonding pressure inspection, and chip appearance inspection;

[0015] If the test is qualified, the substrate will be transported to the discharge port; if the test is unqualified, the substrate will be transported to the defective product collection area and the reason for the defect will be recorded.

[0016] According to some embodiments of the present invention, the vacuum adsorption bundling head is mounted on the Z-axis lifting mechanism via a bundling head fixing base; the guide rail seat is mounted on the bundling head fixing base, and the bundling head Z-compression spring seat, spring adjustment screw, and spring are mounted on the top; the guide rail seat is elastically connected to the rotating head mounting base;

[0017] The guide rail seat has a built-in voice coil motor stator and a first cross guide rail. The voice coil motor mover and the other end of the first cross guide rail are installed on the rotating head mounting seat. The pressure control of the vacuum adsorption head is achieved through the voice coil motor.

[0018] The Z head base is fixed on the guide rail base, the grating head is fixed on the Z head base, and the grating is fixed on the rotating head mounting base. The displacement of the rotating head mounting base is controlled by the feedback of the grating head.

[0019] The lashing head rotary gland and the first motor are installed on the rotary head mounting base, and the high-head thumb screw and the air joint are installed on the lashing head rotary gland; an air guide cavity is provided inside the lashing head rotary gland, the lashing head rotating shaft and the suction nozzle clamping shaft;

[0020] The binding head rotating shaft passes through the first motor and is fixed on the first motor. The binding head rotating shaft is driven to rotate by the first motor. The binding head rotating shaft is connected to the air path of the binding head rotating cover. A suction nozzle clamping shaft is installed under the binding head rotating shaft. A rubber suction nozzle is installed under the suction nozzle clamping shaft. The rubber suction nozzle works through the air path to achieve the purpose of adsorbing the chip.

[0021] According to some embodiments of the present invention, the chip transfer platform includes a transfer adsorption cap, a platform rotating shaft, a second motor, a transfer upper seat, a transfer X mover mounting block, a transfer X mover, a transfer X stator, a transfer Y head seat, a transfer Y head, a transfer Y grating scale, a transfer middle seat, a second cross guide rail, a transfer base, a transfer Y photoelectric film, a transfer Y mover mounting block, a transfer Y mover, a transfer Y stator, a photoelectric sensor, a transfer Y origin seat, a transfer X photoelectric film, a transfer X head, a transfer X grating scale, and a transfer X head seat; wherein,

[0022] The transfer adsorption cap, the platform shaft and the transfer upper seat form a closed air guide cavity. When the solenoid valve is working, the transfer adsorption cap forms a negative pressure environment, adsorbing the chip on the chip placement position of the transfer adsorption cap;

[0023] The transfer center is mounted on the transfer base via the second cross guide rail; the transfer X mover is mounted on the transfer center via the transfer X mover mounting block; the transfer X photoelectric film is mounted on the transfer center via the transfer Y origin seat; the transfer X stator and transfer X reader seat are mounted on the transfer base; the transfer center moves in the X direction via the transfer X mover, and the X-direction displacement of the transfer platform is adjusted using feedback signals from the transfer X reader;

[0024] The transfer upper seat is mounted on the transfer middle seat via a second cross guide rail; the transfer Y reader is fixed to the transfer Y reader seat; the transfer Y mover is fixed to the transfer Y mover mounting block; the second motor, transfer Y reader seat, transfer Y mover mounting block, and transfer Y photoelectric film are mounted on the transfer upper seat; the transfer upper seat is driven to move in the Y direction by the transfer Y mover, and the Y direction displacement of the transfer platform is adjusted by the feedback signal from the transfer Y reader;

[0025] The platform shaft is fixed on the second motor; the transfer adsorption cap is installed on the platform shaft, and the rotation of the second motor drives the chip adsorbed on the transfer adsorption cap to rotate.

[0026] According to some embodiments of the present invention, the angle correction module determines chip angle deviation information, including:

[0027] The first determination module is used to extract key areas and detect feature points on the chip image using convolutional neural networks and image semantic segmentation technology to obtain coordinate information of key feature points of the chip; and generate a three-dimensional model of the chip based on the coordinate information of the key feature points;

[0028] A second determination module is used to determine the three-dimensional coordinates of the three-dimensional model in the camera coordinate system, and determine the angle information of the chip relative to the camera coordinate system according to the three-dimensional coordinates;

[0029] The third determination module is used to compare the angle information corresponding to the three-dimensional model with the preset angle information corresponding to the preset three-dimensional model, and determine the chip angle deviation information according to the comparison result.

[0030] According to some embodiments of the present invention, the quality detection module includes:

[0031] An acquisition module, used for acquiring a processed image after the die bonding operation;

[0032] a fourth determining module, configured to determine a target area based on the processed image;

[0033] The segmentation module is used to perform superpixel segmentation on the target area to obtain K subregions; calculate the gradient modulus between the central pixel and the neighboring pixels in the subregion, select the pixel with the smallest gradient modulus from each superpixel segmentation subregion, and determine it as the initial cluster center of the current superpixel segmentation subregion; iteratively calculate the initial cluster centers based on the K-means algorithm until the number of new cluster centers reaches a preset threshold, thereby obtaining several target subregions;

[0034] The recognition module is used to identify several target sub-areas respectively and perform die bonding quality detection based on the recognition results.

[0035] According to some embodiments of the present invention, the fourth determining module includes:

[0036] A grayscale processing module is used to perform grayscale processing on the processed image to obtain a grayscale image;

[0037] The generation module is used to filter out pixel points with grayscale values ​​greater than a preset grayscale value to generate a target area.

[0038] According to some embodiments of the present invention, the identification module includes:

[0039] a fifth determination module, configured to convert the target sub-region into a YUV color space, determine brightness component information based on the Y color channel value corresponding to each pixel in the target sub-region, and determine a brightness histogram based on the brightness component information of each pixel;

[0040] The sixth determination module is used to determine the recognition result according to the brightness histogram.

[0041] According to some embodiments of the present invention, the control method of the semiconductor die bonder based on intelligent visual control as described above includes:

[0042] Based on the visual inspection module, the chip placed on the chip transfer platform is positioned and inspected to determine the chip's position and posture information;

[0043] Based on the vacuum adsorption bonding head, the chip is grasped according to the chip's position information and posture information, and the chip is moved to the dispensing position;

[0044] The dispensing module controls the position of the dispensing head and the amount of glue according to the preset program, and dispenses glue at the specified position of the substrate;

[0045] The angle correction module uses a camera at the bottom of the chip to capture chip images in real time during the process of the vacuum adsorption head moving the chip to the dispensing position to determine the chip's angle deviation information. The servo motor drives the vacuum adsorption head to correct the chip's angle based on the deviation information, so that the chip is in the target posture.

[0046] After the die bonding module moves the chip to the dispensing position on the vacuum adsorption bonding head, the chip in the target posture is placed on the substrate based on multi-axis linkage control, and the die bonding operation is performed.

[0047] The present invention proposes a semiconductor die bonding machine and control method based on intelligent visual control. Based on the high-response and high-precision positioning and detection of the visual inspection module, angle deviation correction is performed during the chip grabbing process to achieve high-precision dispensing and improve the efficiency and accuracy of die bonding.

[0048] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0049] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0051] Figure 1 is a block diagram of a semiconductor die bonder based on intelligent visual control according to one embodiment of the present invention;

[0052] Figure 2 2 is a schematic diagram of a vacuum adsorption bonding head according to an embodiment of the present invention;

[0053] Figure 3 is a schematic diagram of a chip transfer platform according to an embodiment of the present invention;

[0054] Figure 4 The figure is a flow chart of a control method of a semiconductor die bonder based on intelligent visual control according to an embodiment of the present invention.

[0055] Figure numerals: binding head fixing base 1, guide rail seat 2, binding head Z compression spring seat 3, spring adjustment top screw 4, spring 5, voice coil motor stator 6, voice coil motor mover 7, first cross guide rail 8, rotating head mounting seat 9, high head hand screw 10, binding head rotating pressure cover 11, air joint 12, first motor 13, binding head rotating shaft 14, suction nozzle clamping shaft 15, rubber suction nozzle 16, Z head seat 17, grating reader 18, grating 19, transfer adsorption cap 20, platform shaft 21, second motor 22, transfer upper Seat 23, transfer X mover mounting block 24, transfer X mover 25, transfer X stator 26, transfer Y reader seat 27, transfer Y reader 28, transfer Y grating scale 29, transfer center seat 30, second cross guide rail 31, transfer base 32, transfer Y photoelectric film 33, transfer Y mover mounting block 34, transfer Y mover 35, transfer Y stator 36, photoelectric sensor 37, transfer Y origin seat 38, transfer X photoelectric film 39, transfer X reader 40, transfer X grating scale 41, transfer X reader seat 42. DETAILED DESCRIPTION

[0056] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0057] like Figure 1 As shown, the embodiment of the present invention proposes a semiconductor die bonder based on intelligent visual control, including: a vacuum adsorption bonding head, a chip transfer platform, a visual inspection module, a dispensing module, an angle correction module, and a die bond module; wherein,

[0058] The visual inspection module is used to perform positioning inspection on the chip placed on the chip transfer platform to determine the chip's position and posture information;

[0059] The vacuum adsorption head is used to grab the chip according to its position and posture information and move the chip to the dispensing position;

[0060] The dispensing module is used to control the position of the dispensing head and the amount of glue according to the preset program, and to dispense glue at the specified position of the substrate;

[0061] The angle correction module is used to collect chip images in real time based on the flying camera at the bottom of the chip and determine the chip's angle deviation information when the vacuum adsorption bonding head moves the chip to the dispensing position. The servo motor drives the vacuum adsorption bonding head and corrects the chip's angle based on the deviation information to keep the chip in the target posture.

[0062] The die bonding module is used to place the chip in the target posture on the substrate based on multi-axis linkage control after the vacuum adsorption bonding head moves the chip to the dispensing position and perform the die bonding operation.

[0063] The working principle of this technical solution is as follows: Based on the chip position and posture information provided by the visual inspection module, the vacuum bonding head accurately grasps the chip on the chip transfer platform and moves it to the glue dispensing position. The chip transfer platform provides a temporary place for chips to be placed and processed, ensuring the orderly flow of chips during the die bonding process. The visual inspection module uses intelligent vision technology to locate and detect the chip, determining its position and posture. The glue dispensing module is responsible for dispensing glue at the designated location on the substrate. The angle correction module monitors and corrects the chip's angular deviation in real time during chip movement, ensuring that the chip is positioned in the target posture when placed on the substrate. The target posture is the standard posture required for die bonding. After the vacuum bonding head moves the chip to the glue dispensing position and completes the angle correction, the die bonding module uses multi-axis linkage control technology to precisely place the chip on the substrate. A certain amount of pressure and heat are applied to securely bond the chip to the substrate, completing the die bonding process. During the die bonding process, pressure and temperature sensors monitor the die bonding pressure and temperature in real time to ensure that the die bonding quality meets the requirements. The visual detection module uses the improved YOLOv7 algorithm combined with the Transformer architecture to process the acquired images in real time.

[0064] The beneficial effects of this technical solution include: The visual inspection module's high-response, high-precision positioning and detection capabilities allow for angle deviation correction during chip grasping, enabling high-precision dispensing. The intelligent visual inspection and angle correction modules enable precise control of chip position and posture, significantly improving die bonding accuracy and reducing quality issues caused by chip position deviation or incorrect angle. The modules work collaboratively to enable rapid chip grasping, movement, dispensing, and bonding, improving the efficiency of the entire die bonding process.

[0065] According to some embodiments of the present invention, the visual detection module is further configured to:

[0066] Get the image of the vacuum adsorption head grabbing the chip;

[0067] Identify the captured image to determine whether the chip is successfully captured and whether there is any damage.

[0068] The working principle and beneficial effects of the above technical solution are as follows: The captured image is analyzed. Successful capture of the chip is determined by comparing features such as the change in chip position before and after capture, and the contact status between the vacuum head and the chip. If the chip is in the expected capture position of the vacuum head in the image and is in stable contact with the head, capture is considered successful. Conversely, if the chip is abnormally positioned or not in effective contact with the head, capture is considered unsuccessful. The chip's appearance is carefully inspected. By comparing it with pre-stored images of normal chips, the chip surface is identified for damage such as scratches, cracks, or defects. Furthermore, by analyzing features such as the chip's color and texture, other potential damage can be further determined. By determining in real time whether the chip has been captured successfully and whether it is damaged, defective products can be promptly identified and removed, preventing these chips from entering the subsequent dispensing and die bonding processes. This prevents product quality defects caused by chip problems and improves the final product's pass rate.

[0069] According to some embodiments of the present invention, the system further includes: a quality inspection module for inspecting the die bonding quality of the chip after the die bonding operation; the die bonding quality inspection includes chip position inspection, die bonding pressure inspection, and chip appearance inspection;

[0070] If the test is qualified, the substrate will be transported to the discharge port; if the test is unqualified, the substrate will be transported to the defective product collection area and the reason for the defect will be recorded.

[0071] The working principle of the above technical solution: Chip position detection is to accurately measure the position of the chip on the substrate after die bonding. By comparing with the preset standard position, it is determined whether the chip is in the correct position. The detection indicators are the center coordinate deviation and rotation angle deviation of the chip. The die bonding pressure detection is to monitor the pressure applied to the chip by the die bonding head in real time through the pressure sensor during the die bonding process. After the die bonding operation is completed, the pressure data is recorded and analyzed to determine whether the die bonding pressure is within the appropriate range. The detection indicators are the size and uniformity of the die bonding pressure. Chip appearance detection is to conduct a comprehensive inspection of the appearance of the chip after die bonding to identify whether there are defects on the chip surface, such as scratches, cracks, bubbles, foreign matter, etc.

[0072] The beneficial effects of the above technical solution are: through comprehensive die bonding quality testing, substandard products are promptly discovered and eliminated, ensuring that the quality of the final product meets the standards, improving product reliability and stability. The causes of defects are recorded to facilitate optimization of the production process.

[0073] like Figure 2 As shown, according to some embodiments of the present invention, the vacuum adsorption bundling head is installed on the Z-axis lifting mechanism through the bundling head fixing base 1; the guide rail base 2 is installed on the bundling head fixing base 1, and the bundling head Z compression spring base 3, spring adjustment screw 4 and spring 5 are installed on the top; the guide rail base 2 is elastically connected to the rotating head mounting base 9;

[0074] The guide rail seat 2 has a built-in voice coil motor stator 6 and a first cross guide rail 8. The voice coil motor mover 7 and the other end of the first cross guide rail 8 are mounted on the rotary head mounting seat 9. The pressure control of the vacuum adsorption head is achieved through the voice coil motor.

[0075] The Z head mount 17 is fixed on the guide rail mount 2, the grating head 18 is fixed on the Z head mount 17, and the grating 19 is fixed on the rotary head mounting base 9. The displacement of the rotary head mounting base 9 is controlled by the feedback of the grating head 18.

[0076] The binding head rotary gland 11 and the first motor 13 are mounted on the rotary head mounting base 9, and the high-head thumb screw 10 and the air joint 12 are mounted on the binding head rotary gland 11. An air guide cavity is provided inside the binding head rotary gland 11, the binding head rotating shaft 14, and the suction nozzle clamping shaft 15.

[0077] The binding head rotating shaft 14 passes through the first motor 13 and is fixed on the first motor 13. The binding head rotating shaft 14 is driven to rotate by the first motor 13. The binding head rotating shaft 14 is connected to the air path of the binding head rotating cover 11. A suction nozzle clamping shaft 15 is installed under the binding head rotating shaft 14. A rubber suction nozzle 16 is installed under the suction nozzle clamping shaft 15. The rubber suction nozzle 16 works through the air path to achieve the purpose of adsorbing the chip.

[0078] The beneficial effects of this technical solution are: the vacuum adsorption head can move and rotate in the Z direction, automatically adjusting the position and angle of the chip, achieving a high degree of automation. This effectively reduces vibration interference during the wafer retrieval process, ensuring the stability and integrity of the chip during the grasping process, while also effectively protecting the chip from damage.

[0079] like Figure 3 As shown, according to some embodiments of the present invention, the chip transfer platform includes a transfer adsorption cap 20, a platform rotating shaft 21, a second motor 22, a transfer upper seat 23, a transfer X mover mounting block 24, a transfer X mover 25, a transfer X stator 26, a transfer Y read head seat 27, a transfer Y read head 28, a transfer Y grating scale 29, a transfer middle seat 30, a second cross guide rail 31, a transfer base 32, a transfer Y photoelectric film 33, a transfer Y mover mounting block 34, a transfer Y mover 35, a transfer Y stator 36, a photoelectric sensor 37, a transfer Y origin seat 38, a transfer X photoelectric film 39, a transfer X read head 40, a transfer X grating scale 41, and a transfer X read head seat 42; wherein,

[0080] The transfer adsorption cap 20, the platform shaft 21 and the transfer upper seat 23 form a closed air guide cavity. When the solenoid valve is working, the transfer adsorption cap 20 forms a negative pressure environment, adsorbing the chip on the chip placement position of the transfer adsorption cap 20;

[0081] The transfer center seat 30 is mounted on the transfer base 32 via the second cross guide rail 31; the transfer X-mover 25 is mounted on the transfer center seat 30 via the transfer X-mover mounting block 24; the transfer X-photoelectric film 39 is mounted on the transfer center seat 30 via the transfer Y origin seat 38; the transfer X-stator 26 and the transfer X-reader seat 42 are mounted on the transfer base 32; the transfer center seat 30 is driven to move in the X direction by the transfer X-mover 25, and the X-direction displacement of the transfer platform is adjusted by the feedback signal from the transfer X-reader 40;

[0082] The transfer upper seat 23 is mounted on the transfer middle seat 30 via a second cross guide rail 31; the transfer Y-head reader 28 is fixed to the transfer Y-head reader seat 27; the transfer Y-mover 35 is fixed to the transfer Y-mover mounting block 34; the second motor 22, the transfer Y-head reader seat 27, the transfer Y-mover mounting block 34, and the transfer Y-photoelectric film 33 are mounted on the transfer upper seat 23; the transfer upper seat 23 is driven by the transfer Y-mover 35 to move in the Y direction, and the Y-direction displacement of the transfer platform is adjusted by the feedback signal from the transfer Y-head reader 28;

[0083] The platform shaft 21 is fixed on the second motor 22 ; the transfer adsorption cap 20 is installed on the platform shaft 21 , and the rotation of the second motor 22 drives the chip adsorbed on the transfer adsorption cap 20 to rotate.

[0084] The beneficial effect of the above technical solution is that the rotation of the second motor 22 drives the chip adsorbed on the transfer adsorption cap 20 to rotate, thereby achieving the purpose of adjusting the chip angle, which facilitates the stability and reliability of chip transfer.

[0085] According to some embodiments of the present invention, the angle correction module determines chip angle deviation information, including:

[0086] The first determination module is used to extract key areas and detect feature points on the chip image using convolutional neural networks and image semantic segmentation technology to obtain coordinate information of key feature points of the chip; and generate a three-dimensional model of the chip based on the coordinate information of the key feature points;

[0087] A second determination module is used to determine the three-dimensional coordinates of the three-dimensional model in the camera coordinate system, and determine the angle information of the chip relative to the camera coordinate system according to the three-dimensional coordinates;

[0088] The third determination module is used to compare the angle information corresponding to the three-dimensional model with the preset angle information corresponding to the preset three-dimensional model, and determine the chip angle deviation information according to the comparison result.

[0089] The working principle of the above technical solution is as follows: The captured chip image is input into a trained convolutional neural network, which extracts and abstracts features layer by layer. An image semantic segmentation algorithm is used to classify and label the extracted features, identifying key regions of the chip. Feature points are detected within these regions to obtain the coordinates of the chip's key feature points. Based on the obtained key feature point coordinates, a 3D reconstruction algorithm is used to construct the chip's 3D geometric shape based on the spatial relationship of the feature points, generating a 3D model of the chip. This facilitates a visual display of the chip's morphology and structure. A second determination module maps the generated 3D model into the camera coordinate system and determines the 3D coordinates of each key feature point in the camera coordinate system. Using geometric transformation or a pose estimation algorithm, the chip's angle information, including pitch and yaw angles, is calculated based on the relationship between the 3D coordinates of the key feature points and the camera coordinate system. A 3D model corresponding to a chip with a standard pose is preset. The preset angle information is the angle of the preset 3D model relative to the camera coordinate system. The angle difference is determined based on this angle information and the preset angle information, quantifying the deviation between the chip's angle and the preset angle to determine the chip's angle deviation information.

[0090] The beneficial effects of the above technical solution are: by combining multiple advanced technologies such as convolutional neural networks, image semantic segmentation, and 3D reconstruction, it can more accurately extract chip feature information and generate a precise 3D model, thereby achieving high-precision detection of chip angle deviation. The 3D coordinates of the 3D model in the camera coordinate system are determined, and the angle information of the chip relative to the camera coordinate system is determined based on the 3D coordinates. The angle information corresponding to the 3D model is compared with the preset angle information corresponding to the preset 3D model, which provides a more comprehensive understanding of the chip's posture and facilitates angle correction.

[0091] According to some embodiments of the present invention, the quality detection module includes:

[0092] An acquisition module, used for acquiring a processed image after the die bonding operation;

[0093] a fourth determining module, configured to determine a target area based on the processed image;

[0094] The segmentation module is used to perform superpixel segmentation on the target area to obtain K subregions; calculate the gradient modulus between the central pixel and the neighboring pixels in the subregion, select the pixel with the smallest gradient modulus from each superpixel segmentation subregion, and determine it as the initial cluster center of the current superpixel segmentation subregion; iteratively calculate the initial cluster centers based on the K-means algorithm until the number of new cluster centers reaches a preset threshold, thereby obtaining several target subregions;

[0095] The recognition module is used to identify several target sub-areas respectively and perform die bonding quality detection based on the recognition results.

[0096] The working principle of the above technical solution is: the image portion containing the chip and the solid crystal area is separated from the entire processed image as the target area. Determining the target area can reduce the computational complexity of subsequent processing and improve detection efficiency, while focusing attention on key areas related to solid crystal quality and avoiding interference from irrelevant areas. Calculate the gradient modulus of the central pixel and the neighboring pixel points in each superpixel segmentation sub-region. The gradient modulus reflects the degree of change in the image brightness around the pixel point. The smaller the gradient modulus, the smoother the image change around the pixel point, and the more likely it is to represent the typical characteristics of the sub-region. Select the pixel point with the smallest gradient modulus from each superpixel segmentation sub-region and determine it as the initial clustering center of the current superpixel segmentation sub-region. Selecting the pixel point with the smallest gradient modulus as the initial clustering center can make the clustering process more stable and improve the accuracy of the clustering results. Based on the initial cluster center, pixels are assigned to the cluster with the nearest cluster center based on their distance (e.g., Euclidean distance) to the cluster center. The cluster center for each cluster is then recalculated, and this process is repeated until the number of new cluster centers reaches a preset threshold (i.e., clustering converges). This results in several target subregions. Through iterative calculations using the K-means algorithm, the superpixel segmentation subregions can be further subdivided into more representative target subregions. The target subregions obtained by the segmentation module are then individually identified. Based on the recognition module's results for each target subregion, a comprehensive assessment is made of whether the die bonding quality is acceptable.

[0097] The beneficial effects of the above technical solution are as follows: by finely segmenting the image through superpixel segmentation and K-means algorithm, and performing image recognition, it is possible to more accurately detect appearance quality problems in the die bonding process, thereby improving the accuracy and reliability of detection.

[0098] According to some embodiments of the present invention, the fourth determining module includes:

[0099] A grayscale processing module is used to perform grayscale processing on the processed image to obtain a grayscale image;

[0100] The generation module is used to filter out pixel points with grayscale values ​​greater than a preset grayscale value to generate a target area.

[0101] The above technical solution works as follows: After obtaining a grayscale image, the generation module traverses each pixel in the image and selects pixels with grayscale values ​​greater than a preset grayscale value. The preset grayscale value is a threshold set based on the image characteristics after the die bonding operation and the detection requirements. Pixels with grayscale values ​​greater than the preset grayscale value that are selected are combined to form a connected area, which is the target area.

[0102] The beneficial effects of the above technical solution are as follows: through grayscale processing and target area generation, the amount of data for subsequent processing is reduced, and only the target area related to the solid crystal quality is analyzed and detected, which greatly improves the detection efficiency and helps to eliminate interference from background areas and interference factors.

[0103] According to some embodiments of the present invention, the identification module includes:

[0104] a fifth determination module, configured to convert the target sub-region into a YUV color space, determine brightness component information based on the Y color channel value corresponding to each pixel in the target sub-region, and determine a brightness histogram based on the brightness component information of each pixel;

[0105] The sixth determination module is used to determine the recognition result according to the brightness histogram.

[0106] The working principle of the above technical solution is: after converting the target sub-region into the YUV color space, the brightness component (Y channel) is extracted separately for analysis, while ignoring the influence of chromaticity information, thereby simplifying the processing process and improving the detection ability of brightness-related features. ;in, is the brightness component; 、 、 Pixels aisle, aisle, Channel values; the brightness histogram counts the frequency of different brightness values ​​within the target subregion, providing a visual representation of the brightness distribution within the target subregion. By observing the shape, peak position, and distribution range of the brightness histogram, it is possible to determine whether there are areas of abnormal brightness within the target subregion and thus infer potential die bonding quality issues. For example, shape analysis shows that under normal circumstances, the brightness histogram exhibits a relatively smooth distribution with peaks concentrated in a certain brightness range, indicating a uniform brightness distribution within the target subregion and no obvious abnormalities. Abnormal conditions: bimodal or multimodal distributions indicate the presence of multiple brightness regions, such as areas with excessively high or low brightness, indicating a die bonding position shift. Peak position analysis shows that under normal circumstances, the peak position is within the expected brightness range. Abnormal conditions: peak shift: If the peak shifts toward higher or lower brightness levels, it indicates a poor soldering condition on the chip surface. Distribution range analysis shows that the brightness distribution range is within the preset distribution range. The preset distribution range corresponds to the brightness distribution range under normal conditions. Abnormal conditions: a wide distribution range with significant brightness differences, including distinct high and low brightness areas, indicating uneven die bonding.

[0107] The beneficial effects of the above technical solution are as follows: based on the brightness histogram, it is convenient to accurately determine the quality inspection of the chip after the die bonding operation, especially the chip appearance inspection, thereby improving the accuracy of the quality inspection result.

[0108] like Figure 4 As shown, according to some embodiments of the present invention, the control method of the semiconductor die bonder based on intelligent visual control as described above includes steps S1-S5:

[0109] S1. Based on the visual inspection module, the chip placed on the chip transfer platform is positioned and detected to determine the chip's position and posture information;

[0110] S2, based on the vacuum adsorption bonding head, the chip is grasped according to the chip's position information and posture information, and the chip is moved to the dispensing position;

[0111] S3, the dispensing module controls the position of the dispensing head and the amount of glue according to the preset program, and dispenses glue at the specified position of the substrate;

[0112] S4: When the vacuum adsorption head moves the chip to the dispensing position, the angle correction module collects chip images in real time based on the flying camera at the bottom of the chip to determine the chip angle deviation information; the servo motor drives the vacuum adsorption head to correct the chip angle according to the deviation information to make the chip in the target posture;

[0113] S5. After the die bonding module moves the chip to the dispensing position on the vacuum adsorption bonding head, the chip in the target posture is placed on the substrate based on multi-axis linkage control, and the die bonding operation is performed.

[0114] The beneficial effects of this technical solution include: The visual inspection module's high-response, high-precision positioning and detection capabilities allow for angle deviation correction during chip grasping, enabling high-precision dispensing. The intelligent visual inspection and angle correction modules enable precise control of chip position and posture, significantly improving die bonding accuracy and reducing quality issues caused by chip position deviation or incorrect angle. The modules work collaboratively to enable rapid chip grasping, movement, dispensing, and bonding, improving the efficiency of the entire die bonding process.

[0115] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A semiconductor die bonder based on intelligent visual control, characterized in that: include: Vacuum adsorption bonding head, chip transfer platform, visual inspection module, dispensing module, angle correction module, and die bonding module; among them, The visual inspection module is used to perform positioning inspection on the chip placed on the chip transfer platform to determine the chip's position and posture information; The vacuum adsorption head is used to grab the chip according to its position and posture information and move the chip to the dispensing position; The dispensing module is used to control the position of the dispensing head and the amount of glue according to the preset program, and to dispense glue at the specified position of the substrate; The angle correction module is used to collect chip images in real time based on the flying camera at the bottom of the chip and determine the chip's angle deviation information when the vacuum adsorption bonding head moves the chip to the dispensing position. The servo motor drives the vacuum adsorption bonding head and corrects the chip's angle based on the deviation information to keep the chip in the target posture. The die bonding module is used to place the chip in the target posture on the substrate based on multi-axis linkage control after the vacuum adsorption bonding head moves the chip to the dispensing position and perform the die bonding operation; The vacuum adsorption head is installed on the Z-axis lifting mechanism through the head fixing base; the guide rail seat is installed on the head fixing base, and the head Z compression spring seat, spring adjustment screw and spring are installed on the top; the guide rail seat is elastically connected to the rotating head mounting seat; The guide rail seat has a built-in voice coil motor stator and a first cross guide rail. The voice coil motor mover and the other end of the first cross guide rail are installed on the rotating head mounting seat. The pressure control of the vacuum adsorption head is achieved through the voice coil motor. The Z head base is fixed on the guide rail base, the grating head is fixed on the Z head base, and the grating is fixed on the rotating head mounting base. The displacement of the rotating head mounting base is controlled by the feedback of the grating head. The lashing head rotary gland and the first motor are installed on the rotary head mounting base, and the high-head thumb screw and the air joint are installed on the lashing head rotary gland; an air guide cavity is provided inside the lashing head rotary gland, the lashing head rotating shaft and the suction nozzle clamping shaft; The binding head rotating shaft passes through the first motor and is fixed on the first motor. The binding head rotating shaft is driven to rotate by the first motor. The binding head rotating shaft is connected to the air path of the binding head rotating pressure cover. A suction nozzle clamping shaft is installed under the binding head rotating shaft. A rubber suction nozzle is installed under the suction nozzle clamping shaft. The rubber suction nozzle works through the air path to achieve the purpose of adsorbing the chip.

2. The semiconductor die bonder based on intelligent visual control according to claim 1, characterized in that: The visual detection module is further used to: Get the image of the vacuum adsorption head grabbing the chip; Identify the captured image to determine whether the chip is successfully captured and whether there is any damage.

3. The semiconductor die bonder based on intelligent visual control according to claim 1, characterized in that: Also includes: The quality inspection module is used to perform die bonding quality inspection on the chip after the die bonding operation; the die bonding quality inspection includes chip position inspection, die bonding pressure inspection and chip appearance inspection; If the test is qualified, the substrate will be transported to the discharge port; if the test is unqualified, the substrate will be transported to the defective product collection area and the reason for the defect will be recorded.

4. The semiconductor die bonder based on intelligent visual control according to claim 1, characterized in that: The chip transfer platform includes a transfer adsorption cap, a platform rotating shaft, a second motor, a transfer upper seat, a transfer X mover mounting block, a transfer X mover, a transfer X stator, a transfer Y head seat, a transfer Y head, a transfer Y grating ruler, a transfer middle seat, a second cross guide rail, a transfer base, a transfer Y photoelectric film, a transfer Y mover mounting block, a transfer Y mover, a transfer Y stator, a photoelectric sensor, a transfer Y origin seat, a transfer X photoelectric film, a transfer X head, a transfer X grating ruler, and a transfer X head seat; wherein, The transfer adsorption cap, the platform shaft and the transfer upper seat form a closed air guide cavity. When the solenoid valve is working, the transfer adsorption cap forms a negative pressure environment, adsorbing the chip on the chip placement position of the transfer adsorption cap; The transfer center is mounted on the transfer base via the second cross guide rail; the transfer X mover is mounted on the transfer center via the transfer X mover mounting block; the transfer X photoelectric film is mounted on the transfer center via the transfer Y origin seat; the transfer X stator and transfer X reader seat are mounted on the transfer base; the transfer center moves in the X direction via the transfer X mover, and the X-direction displacement of the transfer platform is adjusted using feedback signals from the transfer X reader; The transfer upper seat is mounted on the transfer middle seat via a second cross guide rail; the transfer Y reader is fixed to the transfer Y reader seat; the transfer Y mover is fixed to the transfer Y mover mounting block; the second motor, transfer Y reader seat, transfer Y mover mounting block, and transfer Y photoelectric film are mounted on the transfer upper seat; the transfer upper seat is driven to move in the Y direction by the transfer Y mover, and the Y direction displacement of the transfer platform is adjusted by the feedback signal from the transfer Y reader; The platform shaft is fixed on the second motor; the transfer adsorption cap is installed on the platform shaft, and the rotation of the second motor drives the chip adsorbed on the transfer adsorption cap to rotate.

5. The semiconductor die bonder based on intelligent visual control according to claim 1, characterized in that: The angle correction module determines the chip angle deviation information, including: The first determination module is used to extract key areas and detect feature points on the chip image using convolutional neural networks and image semantic segmentation technology to obtain coordinate information of key feature points of the chip; and generate a three-dimensional model of the chip based on the coordinate information of the key feature points; A second determination module is used to determine the three-dimensional coordinates of the three-dimensional model in the camera coordinate system, and determine the angle information of the chip relative to the camera coordinate system according to the three-dimensional coordinates; The third determination module is used to compare the angle information corresponding to the three-dimensional model with the preset angle information corresponding to the preset three-dimensional model, and determine the chip angle deviation information according to the comparison result.

6. The semiconductor die bonder based on intelligent visual control according to claim 3, characterized in that: Quality inspection module, including: An acquisition module, used for acquiring a processed image after the die bonding operation; a fourth determining module, configured to determine a target area based on the processed image; The segmentation module is used to perform superpixel segmentation on the target area to obtain K subregions; calculate the gradient modulus between the central pixel and the neighboring pixels in the subregion, select the pixel with the smallest gradient modulus from each superpixel segmentation subregion, and determine it as the initial cluster center of the current superpixel segmentation subregion; iteratively calculate the initial cluster centers based on the K-means algorithm until the number of new cluster centers reaches a preset threshold, thereby obtaining several target subregions; The recognition module is used to identify several target sub-areas respectively and perform die bonding quality detection based on the recognition results.

7. The semiconductor die bonder based on intelligent visual control according to claim 6, characterized in that: The fourth determination module includes: A grayscale processing module is used to perform grayscale processing on the processed image to obtain a grayscale image; The generation module is used to filter out pixel points with grayscale values ​​greater than a preset grayscale value to generate a target area.

8. The semiconductor die bonder based on intelligent visual control according to claim 6, characterized in that: Identification module, including: a fifth determination module, configured to convert the target sub-region into a YUV color space, determine brightness component information based on the Y color channel value corresponding to each pixel in the target sub-region, and determine a brightness histogram based on the brightness component information of each pixel; The sixth determination module is used to determine the recognition result according to the brightness histogram.

9. The control method of a semiconductor die bonder based on intelligent visual control according to any one of claims 1 to 8, characterized in that: include: Based on the visual inspection module, the chip placed on the chip transfer platform is positioned and inspected to determine the chip's position and posture information; Based on the vacuum adsorption bonding head, the chip is grabbed according to the chip's position information and posture information, and the chip is moved to the dispensing position; The dispensing module controls the position of the dispensing head and the amount of glue according to the preset program, and dispenses glue at the specified position of the substrate; The angle correction module uses a camera at the bottom of the chip to capture chip images in real time during the process of the vacuum adsorption head moving the chip to the dispensing position to determine the chip's angle deviation information. The servo motor drives the vacuum adsorption head to correct the chip's angle based on the deviation information, so that the chip is in the target posture. After the die bonding module moves the chip to the dispensing position on the vacuum adsorption bonding head, the chip in the target posture is placed on the substrate based on multi-axis linkage control, and the die bonding operation is performed.

Images