A method and system for preventing and automatically correcting wire bonding deviation of a wire bonding machine

Abstract

This invention discloses a method and system for preventing and automatically correcting solder misalignment in wire bonding machines, belonging to the technical field of semiconductor packaging equipment. This method uses a real-time image recognition and search mechanism to calculate the displacement deviation between the real-time position and the corresponding initial scanning position. Based on the result, it performs iterative position updates and intelligent correction, improving UPH (Ultra-High Hierarchy) while achieving fully automatic judgment without relying on manual parameter settings. It introduces dynamic ROI (Region of Interest) modification technology, simulating high-to-low magnification dynamic switching recognition on a single-path camera. This eliminates the cost of upgrading hardware to a dual-path system while simultaneously improving recognition speed and positioning accuracy. Through enhanced recognition algorithms, it supports sub-pixel level (1 / 32 pixel) image resolution, significantly improving positioning accuracy while optimizing search efficiency, thereby comprehensively enhancing soldering accuracy and product yield, achieving a synergistic improvement in recognition accuracy, speed, and system economy.

Description

Technical Field

[0001] This invention belongs to the field of semiconductor packaging equipment technology, and particularly relates to a method and system for preventing and automatically correcting solder misalignment in wire bonding machines. Background Technology

[0002] As a high-precision, high-speed chip packaging equipment, wire bonding machines often adopt a pre-scanning and post-bonding operation method in actual production processes to improve overall efficiency. This means that multiple wafers arranged in front are first identified and positioned in batches, and then returned to perform wire bonding operations in a unified manner.

[0003] However, at the beginning of wire bonding, due to factors such as mechanical vibration during machine operation, insufficient clamping and positioning accuracy, or material thermal deformation and temperature, the pre-scanned wafers may shift in position, causing the previously acquired positioning coordinates to become inaccurate. If the original position data is still used for wire bonding under these circumstances, misalignment is very likely to occur, resulting in defective products and affecting product quality and yield. To solve this misalignment problem, traditional wire bonding machines generally increase the preheating delay time. This significantly affects the machine's production efficiency, and different products require different settings, making it prone to errors. It also cannot solve the problem for wide supports (support width greater than 80mm), requiring the machine to be divided into several sections, which reduces speed.

[0004] In existing technologies, although there are post-weld inspections or in-process interruption inspections based on machine vision, most of them are post-processing corrections or shutdown alarms, and cannot perform real-time, online preventive corrections before the weld deviation occurs, making it difficult to meet the production requirements of high precision and high efficiency. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for preventing and automatically correcting solder misalignment in wire bonding machines. Its core objective is to perform real-time detection and correction of the wafer position using a high-speed vision system integrated into the equipment before wire bonding. Once a positional deviation exceeds the tolerance, a coordinate update and solder joint position recalculation process is automatically triggered, achieving a fully closed-loop automated control of "detection-judgment-correction" to prevent solder misalignment at its source. Simultaneously, the application of dynamically switching ROI (Region of Interest) technology eliminates the need for additional hardware conversion dual-optical-path lenses, automatically switching ROIs during the recognition process, improving recognition accuracy and efficiency. This significantly enhances welding accuracy and product yield while maintaining high production efficiency.

[0006] The technical solution is as follows:

[0007] On the one hand, a method for preventing and automatically correcting solder misalignment in a wire bonding machine is provided, including the following steps:

[0008] Step S1, Pre-scan positioning: Control the wire bonding machine to move to the preset image capture position, perform image acquisition and recognition on the wafer under test, and obtain the initial positioning position coordinates;

[0009] Step S2, Real-time Positioning: Before the wire bonding operation, control the wire bonding machine to move back to the imaging station of the wafer to be tested, dynamically switch the ROI, collect real-time images of the wafer, perform visual positioning on the real-time images, and determine the real-time position coordinates of the wafer through image recognition algorithms;

[0010] Step S3, coordinate comparison: calculate the displacement deviation between the real-time position coordinates and the corresponding initial positioning position coordinates;

[0011] Step S4, Deviation Judgment: Determine whether the displacement deviation exceeds the preset tolerance threshold;

[0012] If it is determined that the limit has not been exceeded, then proceed to step S6;

[0013] If it is determined that the limit is exceeded, then proceed to step S5;

[0014] Step S5, position iteration update: update the initial positioning coordinates to the real-time position coordinates, and recalculate the corresponding solder joint position coordinates based on the updated coordinates; control the wire bonding machine to move to the next photo-taking station to be verified, and repeat steps S2 to S4 until the preset iteration end condition is met.

[0015] Step S6: Based on the positioning coordinates confirmed in step S4 or updated in step S5 and the corresponding solder joint coordinates, perform the wire bonding operation.

[0016] Furthermore, the ROI range is pre-set on the host computer.

[0017] Furthermore, the image recognition algorithm is a zero-mean normalized cross-correlation (ZNCC) algorithm based on image grayscale information.

[0018] Furthermore, in the zero-mean normalized cross-correlation (ZNCC) algorithm based on image grayscale information, integral image technology is used to calculate the pixel mean of the image region.

[0019] Furthermore, the image recognition algorithm is a recognition algorithm based on the shape and contour of the wafer.

[0020] Furthermore, the wafer shape contour recognition algorithm achieves localization by calculating the matching degree between the contour point set of the wafer image's contour model and the edge point set in the real-time image. The matching degree of the wafer shape contour recognition algorithm... Represented as:

[0021] ;

[0022] in, It is the total number of contour points in the wafer model. It is the number of contour points in the real-time image. It is the first Point weight, and Representing the first Model points Its corresponding edge points on the real-time image The distance error decay function.

[0023] Furthermore, at the pixel-level positioning point To perform the sub-pixel-level positioning, a 4x4 pixel array is selected around the target point, with each pixel having a corresponding matching degree. Sub-pixel level matching degree estimation is performed using bilinear interpolation:

[0024] ;

[0025] in, Sub-pixel The degree of matching, , , , .

[0026] Furthermore, at sub-pixel points Repeat the above operation 3 times in the vicinity to obtain a precision of 1 / 32 pixel.

[0027] Furthermore, in large-scale chips, the initial positioning coordinates are determined based on the positions of at least two wafer reference points. In step S5, the position iteration update includes: acquiring the real-time position coordinates of each reference point, calculating the new initial positioning coordinates and wafer rotation angle based on the real-time position coordinates of each reference point, and updating the solder joint position coordinates. To address the potential overall translation and rotation of large-scale chips, multi-reference point positioning and rotation angle calculation achieve a more comprehensive coordinate transformation (translation + rotation), resulting in stronger correction capabilities.

[0028] Furthermore, when the object to be processed contains multiple independent process matrices, the method employs a hierarchical iterative control strategy: within the current process matrix, after completing the position verification and wire bonding operations for all wafers, it moves to the next process matrix and repeats steps S1 to S6. By adopting a hierarchical iterative control strategy, the method orderly switches and operates between multiple independent process matrices, achieving automated continuous production of large-area, multi-region substrates, thus improving equipment utilization and production efficiency.

[0029] Furthermore, the iteration termination conditions include: the displacement deviation of the current workstation is less than the tolerance threshold, or the number of iterations for position updates reaches a preset upper limit. This ensures accuracy while avoiding system dead loops caused by inaccurate positioning of individual points, thus guaranteeing a smooth production process.

[0030] On the other hand, a wire bonding machine's off-center soldering prevention and automatic correction system is provided to implement the above-mentioned method, comprising the following modules:

[0031] The vision module is used to acquire images of the wafer at the imaging station;

[0032] A motion module is used to drive the vision module and the welding head to move between the imaging stations; it includes an XY motion platform, a linear motor fixedly connected to the XY motion platform, and a motion control card that is communicatively connected to the XY motion platform and the linear motor.

[0033] The host computer processing module is communicatively connected to the vision module and the motion module, and is configured to execute the steps of the method to realize pre-welding position verification, deviation judgment, coordinate iterative update and control of welding execution;

[0034] The image processing module is used for image preprocessing, image information analysis and calculation, and image display.

[0035] The system tightly integrates a high-precision vision module, a fast motion control module, a real-time image processing module, and an intelligent decision-making host computer to form a complete "perception-decision-execution" automated correction closed loop.

[0036] Furthermore, the vision module includes an industrial camera, a lens module, and an independently controllable light source module comprising axial and side lighting. Equipped with independently controllable axial and side lighting modules, the illumination conditions can be optimized according to different wafer surface features and reflectivity, providing high-quality input images for image recognition algorithms and further ensuring the reliability of positioning.

[0037] The technical solution includes at least the following technical effects:

[0038] 1. This solution transforms the traditional "post-scraping remediation" of soldering deviation into "pre-scraping prevention and automatic correction" through closed-loop control of real-time visual positioning, coordinate comparison and iterative updates before soldering, significantly improving soldering accuracy and yield. It adopts a variety of image algorithms (such as ZNCC and contour matching) and adapts to strategies such as multi-reference point positioning and layered iteration, enhancing the system's adaptability to thermal deformation, incoming material errors and complex chip layouts. Ultimately, it realizes a highly stable, highly automated intelligent soldering wire deviation correction system that can effectively prevent batch defects.

[0039] 2. The entire inspection and prevention process is highly integrated into the wire bonding machine's control system, achieving a fully automated process from wafer positioning to determine the possibility of solder misalignment, and then to automatic correction and updating of the solder joints. This integrated design not only improves inspection efficiency but also reduces the possibility of human intervention, further enhancing the system's stability and reliability.

[0040] 3. Through a closed-loop process of "pre-scan positioning - real-time positioning - deviation judgment - iterative update", the position of each wafer (or key station) is verified and compensated in real time before welding, which fundamentally solves the systemic welding deviation problem caused by material thermal expansion and contraction, mechanical repetitive positioning error, and fluctuation of incoming material position.

[0041] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit the invention. Attached Figure Description

[0042] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0043] Figure 1 A schematic diagram illustrating the working principle of the host computer processing module in Embodiment 2, which is a preferred embodiment of the present invention.

[0044] Figure 2 This is a flowchart of a method for preventing and automatically correcting off-center solder joints in a wire bonding machine, provided in a preferred embodiment of the present invention.

[0045] Figure 3 This is a block diagram of a wire bonding machine's off-center soldering prevention and automatic alignment system, provided in a preferred embodiment of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0047] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0048] In order to improve overall efficiency, wire bonding machines often adopt a pre-scanning method before bonding the first two rows of wafers during production. However, due to mechanical vibration during machine operation, insufficient clamping and positioning accuracy, or material thermal deformation and temperature factors, the pre-scanned wafers are prone to positional misalignment, resulting in misalignment. The traditional solution is to increase the preheating delay time to alleviate misalignment in the first two rows of pre-scanning. However, the obvious drawback is that it reduces the efficiency of the equipment, resulting in a significant decrease in the output per unit time (UPH). Furthermore, it requires manual adjustment of parameters for different products, which is prone to human error. For wide brackets, uneven heating further limits the correction effect, requiring the machine to be divided into several sections, thus reducing the speed.

[0049] This method employs a real-time image recognition and search mechanism, comparing results to calculate the displacement deviation between the real-time position and the corresponding initial scanning position. Based on the results, it performs iterative position updates and intelligent correction, improving UPH (Uptime Per Hour) while achieving fully automatic judgment without relying on manual parameter settings. It introduces a dynamic ROI (Region of Interest) modification technique, simulating high-to-low magnification dynamic switching recognition on a single-path camera. This eliminates the cost of upgrading to a dual-path camera while simultaneously improving recognition speed and positioning accuracy. Through enhanced recognition algorithms, it supports sub-pixel level (1 / 32 pixel) image resolution, significantly improving positioning accuracy while optimizing search efficiency. This comprehensively enhances welding precision and product yield, achieving a synergistic improvement in recognition accuracy, speed, and system economy.

[0050] Example 1

[0051] like Figure 2 As shown, this embodiment provides a method for preventing and automatically correcting solder misalignment in a wire bonding machine, including the following steps:

[0052] Step S1, Pre-scan positioning: Control the wire bonding machine to move to the preset image capture position, perform image acquisition and recognition on the wafer under test, and obtain the initial positioning position coordinates;

[0053] Optionally, in a large chip, the initial positioning coordinates are determined based on the positions of at least two wafer reference points; in step S5, the position iteration update includes: obtaining the real-time position coordinates of each reference point, calculating the new initial positioning coordinates and wafer rotation angle based on the real-time position coordinates of each reference point, and updating the solder joint position coordinates.

[0054] Step S2, Real-time Positioning: Before the wire bonding operation, the wire bonding machine is controlled to move again to the imaging station of the wafer to be tested, dynamically switch the ROI, acquire real-time images of the wafer, perform visual positioning on the real-time images, and determine the real-time position coordinates of the wafer through an image recognition algorithm; the image recognition algorithm is either the zero-mean normalized cross-correlation (ZNCC) algorithm based on image grayscale information or a recognition algorithm based on the wafer shape and contour.

[0055] In the zero-mean normalized cross-correlation (ZNCC) algorithm based on image grayscale information, the pixel mean of the image region is calculated using integral image technology.

[0056] The wafer shape contour recognition algorithm achieves localization by calculating the matching degree between the contour point set of the wafer image contour model and the edge point set in the real-time image. Represented as:

[0057] ;

[0058] in, It is the total number of contour points in the wafer model. It is the number of contour points in the real-time image. It is the first Point weight, and Representing the first Model points Its corresponding edge points on the real-time image The distance error decay function.

[0059] pixel-level positioning points Subpixel-level localization is performed by selecting a 4x4 pixel area around the target point, with each pixel having a corresponding matching degree. Sub-pixel level matching degree estimation is performed using bilinear interpolation:

[0060] ;

[0061] in, Sub-pixel The degree of matching, , , , .

[0062] In sub-pixel Repeat the above operation 3 times in the vicinity to obtain a precision of 1 / 32 pixel.

[0063] Step S3, coordinate comparison: calculate the displacement deviation between the real-time position coordinates and the corresponding initial positioning position coordinates;

[0064] Step S4, Deviation Judgment: Determine whether the displacement deviation exceeds the preset tolerance threshold;

[0065] If it is determined that the limit has not been exceeded, then proceed to step S6;

[0066] If it is determined that the limit is exceeded, then proceed to step S5;

[0067] Step S5, position iteration update: update the initial positioning coordinates to the real-time position coordinates, and recalculate the corresponding solder joint position coordinates based on the updated coordinates; control the wire bonding machine to move to the next photo-taking station to be verified, and repeat steps S2 to S4 until the preset iteration end condition is met; the iteration end condition includes: the displacement deviation of the current station is less than the tolerance threshold, or the number of iterations of position update reaches the preset upper limit.

[0068] Step S6: Based on the positioning coordinates confirmed in step S4 or updated in step S5 and the corresponding solder joint coordinates, perform the wire bonding operation.

[0069] When the object to be processed contains multiple independent process matrices, the method adopts a hierarchical iterative control strategy: within the current process matrix, after completing the position verification and wire bonding operations of all wafers, it moves to the next process matrix and repeats steps S1 to S6.

[0070] Example 2

[0071] like Figure 3 As shown, this embodiment provides a wire bonding machine's off-center soldering prevention and automatic correction system to implement the method in Embodiment 1, including the following modules:

[0072] The vision module is used to acquire images of the wafer at the imaging station. It includes a lens module, a GigE compact CCD industrial camera, and a lighting module. The lens module features high resolution, low distortion, and a large depth of field, allowing for flexible adjustment of focus and aperture to capture images ranging from macro details to wide-field perspectives, depending on the application scenario. The lighting module is fixed in a semi-ring shape at the front of the lens module, integrating axial and side-light sources. These sources can be controlled independently or used in combination to meet the lighting needs of different materials and surface properties. The selection of blue and red light sources not only provides a rich selection of wavelengths but also allows the vision system to optimize the lighting scheme based on the light absorption and reflection characteristics of different materials, highlighting details and improving image contrast. The industrial camera is fixed at the rear of the lens module, using a gigabit network cable for data transmission. The network cable and camera power cable are connected to the industrial control computer. The entire vision module is fixed on an XY motion platform, which moves the module to the designated position for light source switching, image capture, and image data transmission.

[0073] The motion module is used to drive the vision acquisition and the welding head to move between the photo-taking stations; it includes an XY motion platform, a linear motor fixedly connected to the XY motion platform, and a motion control card that is communicatively connected to the XY motion platform and the linear motor; the control card sends instructions to the linear motor, and the linear motor drives the XY motion platform to move after receiving the instructions.

[0074] The host computer processing module is communicatively connected to the vision module and the motion module, and is configured to execute the steps of the method to achieve pre-welding position verification, deviation judgment, coordinate iterative update, and control of welding execution. It includes an industrial computer and a host computer. The host computer sends motion signals to the motion module to drive the XY motion platform to a designated position, and then sends a trigger signal to the vision module. Upon receiving the trigger signal, the industrial camera takes an exposure picture and transmits the image data back to the host computer, which then analyzes the image information.

[0075] The image processing module is responsible for image preprocessing, image information analysis and calculation, and image display. It includes an image preprocessing unit, an image analysis and calculation unit, and an image display unit. All functions of the image processing module are real-time online, enabling detection and information feedback to be completed at extremely high speeds.

[0076] The image preprocessing unit is used for image grayscale conversion, region of interest cropping, smoothing and noise reduction, etc.

[0077] The image analysis and calculation unit is involved in the processing of image grayscale information and the calculation of the normalized similarity between the image and the target.

[0078] The image display unit is used to save the processed results in the host computer data, and the host computer performs real-time image display or annotation of the processing results.

[0079] Figure 1 In the process, the pre-scan sequence locates n initial positions P. n At each photo booth D x Image recognition will be performed to obtain the latest location P. xT Determine P xT With P x If the deviation exceeds the threshold T, then update the position P. x If a loop detection process is executed, the normal operation process begins; otherwise, it begins. Figure 1 As shown, during actual operation, the wire bonding machine drives the XY motion platform to move to the preset photo-taking positions (D1, D2, D3...D...). n The wafers within the range are pre-scanned and positioned to obtain the initial positioning coordinates (P1, P2, P3...P...). n Then, it returns to the initial positioning coordinates P1 to perform the wire bonding operation. Subsequently, the host computer sends a command to the motion module, driving the XY motion platform to move to the preset image capture station D1. At the same time, the vision module automatically switches to the preset lighting conditions set when teaching the wafer template image, and the lighting module switches to the specified light source. The host computer further sends a trigger signal to the vision module to start the industrial camera for exposure and image acquisition, and transmits the image data back to the host computer.

[0080] The host computer uses a visual image matching algorithm to accurately locate the wafer target in the D1 station image and calculate its actual position coordinates P. 1T .

[0081] Among them, the zero-mean normalized cross-correlation (ZNCC) algorithm based on image grayscale information is the most widely used in the field of recognition and is more resistant to the influence of illumination changes. Its core principle is:

[0082] ,

[0083] in, It is the standard for judging the similarity of wafers, and is used to determine the position of wafers on the image. The coordinates on the wafer image are pixel values, The coordinates of the currently captured image are The pixel value.

[0084] ,

[0085] ,

[0086] Where M and N are the length and width dimensions of the wafer image.

[0087] To achieve high-speed, real-time performance, this embodiment replaces the mean calculation in the ZNCC algorithm with integral image technology; that is, integral image technology is used to calculate the pixel mean of the image region.

[0088] ,

[0089] in, It corresponds to the integral graph. The value represents the distance from the top left corner of the image to the zero point. The sum of all pixel values ​​at a location, using an integral image, allows for the calculation of the average pixel value of an image region in the shortest possible time.

[0090] In another embodiment, in addition to the ZNCC algorithm, users can also choose a wafer shape contour-based recognition algorithm. Compared to the ZNCC algorithm, it is more robust to target occlusion and complex textures, requires less data to compute, and is faster. The wafer shape contour-based recognition algorithm achieves localization by calculating the matching degree between the contour point set of the wafer image's contour model and the edge point set in the real-time image.

[0091] The contour model of a wafer image can be represented as: ,in, Represents the set of contour points. A set representing the gradient directions of contour points;

[0092] ;

[0093] ;

[0094] The matching degree based on the wafer shape and contour recognition algorithm is expressed as:

[0095] ,

[0096] A higher matching degree indicates that there exists a set of contour points and a set of wafer contours in the captured image. The more similar they are. Among them, It is the total number of contour points in the wafer model. It is the number of outline points in the photographed image. It is the first i Point weights are determined during the teaching process on the host computer; contour points closer to the center of the wafer image have higher weights. and Representing the first Model points Its corresponding edge point on the photographed image Distance error attenuation function,

[0097] The gradient direction error decay function, expressed using a Gaussian model, is as follows:

[0098] ,

[0099] ,

[0100] Among them, parameters and The value affects positioning accuracy and success rate; the smaller the value, the more accurate the positioning but the lower the success rate, and vice versa. Parameter settings are changed via a host computer, and users select appropriate parameter settings based on the characteristics of the wafer material.

[0101] The latest target position P is obtained by using either the zero-mean normalized cross-correlation (ZNCC) algorithm based on image grayscale information or the ZNCC algorithm based on image grayscale information. 1T The host computer immediately performs a position comparison and judgment. If the target position P is detected... 1T The deviation from the system's preset position P1 exceeds the set threshold T (P1 refers to the pre-scan positioning position of station D1, P...). 1T (Referring to the new location), the system immediately triggers the location update mechanism to correct the coordinate reference P1=P in real time. 1T And recalculate the solder joint location information.

[0102] Solder joint location P x P yDuring teaching, it is determined that there is a fixed offset from the preset position P1 of the reference chip:

[0103] Offset x =P 1 -P x ，

[0104] Offset y =P 1 -P y 。

[0105] When P1 is updated to P 1T At that time, the solder joint position P x P y Updated accordingly.

[0106] ,

[0107] ,

[0108] in, It is the rotation angle of the wafer; the actual position of the solder joint will deflect with the wafer rotation angle.

[0109] In some embodiments, for large chips, a two-point wafer template is used to determine the preset position. Two-point wafer template positions , The position of the wafer is determined during the teaching process by the host computer. When updating the solder joint position, the latest position of the wafer is first calculated at the imaging station according to the above positioning algorithm. and ,but:

[0110] ,

[0111] Rotation angle :

[0112] ,

[0113] The solder joint location information is also updated using the method described above.

[0114] It should be noted that a large chip refers to a chip that cannot be fully seen within the field of view of the camera lens. Generally, a chip with a size of 80*200um or larger can be considered a large chip.

[0115] If the system determines that the current state does not require triggering the position update mechanism, the equipment will directly switch to the preset normal wire bonding operation process; otherwise, if the system confirms that the position update mechanism needs to be activated, the control system will drive the XY motion platform to precisely move the onboard vision module to the next predetermined photo-taking station D2.

[0116] Upon reaching the photo-taking station D2, the system will re-execute the aforementioned position update algorithm: image data of the current position will be acquired through the vision module, and after image processing and coordinate analysis, new reference coordinate information P2 will be obtained. At this time, the host computer will synchronously perform real-time analysis and error calculation on the updated position data, continuously judging the deviation between the current coordinates and the target position.

[0117] This iterative correction process will continue until one of the following conditions is met: the system detects the coordinates P after the nth update. n The error value between the target position and the target position falls within the preset tolerance range (i.e.) If the system determines that the position update process has reached the preset iteration limit, the system will exit the position update loop and officially enter a stable and accurate normal wire bonding operation process.

[0118] In some embodiments, for process handling of composite matrix types, the system employs a hierarchical iterative control strategy. Within each independent matrix unit, the system must fully execute a global update algorithm for position coordinates—that is, for all wafer coordinate points contained in the matrix, calculate and update their corresponding target positions one by one. P nT Specifically, the system will repeatedly execute a series of precise calculation processes within the current matrix range, including visual positioning, coordinate comparison, deviation judgment, and position iterative update.

[0119] The system's operational logic within a single matrix cell exhibits significant timing optimization characteristics: after completing the verification and updating of all chip positions within a matrix, the system does not immediately jump to the next matrix cell. Instead, based on the latest calibrated position data, it immediately initiates and executes a complete wire bonding operation within the current matrix. Only after all wire bonding tasks within the matrix are completed will the control system instruct the XY motion platform to perform coordinate transformation, precisely positioning the vision module and processing head to the reference position of the next matrix to be processed.

[0120] In this subsequent matrix, the system will completely reproduce the aforementioned complete algorithm process, including the re-establishment of the coordinate system, the iterative update of the position parameters, and the sequential execution of the wire bonding operation, thereby ensuring that each matrix unit achieves the same processing accuracy.

[0121] Through the above methods, the system achieves online and dynamic compensation for positional deviations, effectively avoiding welding misalignment problems caused by cumulative errors. This technology significantly reduces the frequency of manual intervention in the production process, ensures the stability of continuous equipment operation, thereby improving welding consistency and precision while maintaining high production efficiency and effectively extending the mean time between failures (MTBA).

[0122] The entire inspection and prevention process is highly integrated into the wire bonding machine's control system, achieving a fully automated process from determining the possibility of misalignment by locating the wafer to automatically correcting and updating the solder joints. This integrated design not only improves inspection efficiency but also reduces the possibility of human intervention, further enhancing the system's stability and reliability.

[0123] This invention provides a method or system for preventing solder misalignment based on real-time visual positioning and automatic correction. Its core innovation lies in constructing a fully closed-loop, high-response pre-soldering position detection and compensation system. This system verifies and calibrates the wafer position in real time before soldering, eliminating solder misalignment defects caused by positional deviations at the source, significantly improving the quality control level and product yield of the packaging process.

[0124] The key to this technical solution lies in establishing a real-time pre-welding detection and automatic feedback mechanism. The system acquires real-time images of the wafer using a high-speed industrial camera integrated into the wire bonder, and uses advanced image recognition algorithms to accurately locate the wafer's current position. The acquired real-time position data is quickly compared with the reference coordinates from the pre-scanning stage. Once the system detects a displacement deviation exceeding a preset threshold, it immediately determines that the original positioning information is invalid and automatically triggers a position update process.

[0125] Upon detecting a positional deviation, the system recalculates the solder joint position in real time based on the latest acquired wafer coordinates and automatically updates it to the motion control system. The entire correction process requires no manual intervention or system pause, completing the refresh and synchronization of wafer and solder joint position data within milliseconds. This effectively avoids production interruptions caused by line stoppages for confirmation and prevents the continued generation of defective products. This mechanism not only ensures equipment production efficiency and mean time between failures (MTBA) but also significantly reduces defect rates and production costs.

[0126] At the technical implementation level, this invention fully leverages the adaptability and robustness of software algorithms under complex working conditions. Even in challenging environments where the wafer undergoes translation, rotation, or changes in ambient lighting conditions, the system can still stably and accurately complete target recognition and positioning tasks. This algorithm-driven solution fully reuses the existing hardware resources of the wire bonding machine, achieving high-precision position compensation without the need for additional dedicated components. While ensuring performance improvement, it also boasts excellent engineering applicability and economy.

[0127] In specific implementation scenarios, the wire bonding machine of this invention can maintain extremely high welding position accuracy in high-speed production environments, effectively improving the process stability and market competitiveness of the equipment in advanced packaging applications.

[0128] During production, the pre-scanned sequence wafers are continuously corrected until a certain result falls within the deviation range, at which point the iteration is terminated early; this process is not performed on all wafers. The detection timing is optimized based on material properties and the type of production matrix. Iterative comparisons are performed within a single matrix, rather than across all matrices, minimizing the impact of material deformation caused by prolonged heating.

[0129] Unlike similar existing methods, this method incorporates early termination of iteration and optimizes the detection timing, which can calculate the minimum number of detections based on material properties and matrix type, reducing meaningless bias comparisons and improving machine production efficiency.

[0130] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

Claims

1. A method for preventing and automatically correcting off-center soldering in a wire bonding machine, characterized in that, include: Step S1, Pre-scan positioning: Control the wire bonding machine to move to the preset image capture position, perform image acquisition and recognition on the wafer under test, and obtain the initial positioning position coordinates; Step S2, Real-time Positioning: Before the wire bonding operation, control the wire bonding machine to move back to the imaging station of the wafer to be tested, dynamically modify the region of interest (ROI), collect real-time images of the wafer, and determine the real-time position coordinates of the wafer through an image recognition algorithm; Step S3, coordinate comparison: calculate the displacement deviation between the real-time position coordinates and the corresponding initial positioning position coordinates; Step S4, Deviation Judgment: Determine whether the displacement deviation exceeds the preset tolerance threshold; If it is determined that the limit has not been exceeded, then proceed to step S6; If it is determined that the limit is exceeded, then proceed to step S5; Step S5, position iteration update: update the initial positioning coordinates to the real-time position coordinates, and recalculate the corresponding solder joint position coordinates based on the updated coordinates; control the wire bonding machine to move to the next photo-taking station to be verified, and repeat steps S2 to S4 until the preset iteration end condition is met. Step S6: Based on the positioning coordinates confirmed in step S4 or updated in step S5 and the corresponding solder joint coordinates, perform the wire bonding operation.

2. The method for preventing and automatically correcting off-center soldering in a wire bonding machine according to claim 1, characterized in that, The image recognition algorithm is the zero-mean normalized cross-correlation (ZNCC) algorithm based on image grayscale information.

3. The method for preventing and automatically correcting off-center soldering in a wire bonding machine according to claim 2, characterized in that, In the zero-mean normalized cross-correlation (ZNCC) algorithm based on image grayscale information, the pixel mean of the image region is calculated using integral image technology.

4. The method for preventing and automatically correcting off-center soldering in a wire bonding machine according to claim 1, characterized in that, The image recognition algorithm is a recognition algorithm based on the shape and contour of the wafer.

5. The method for preventing and automatically correcting off-center soldering in a wire bonding machine according to claim 4, characterized in that, The wafer shape contour recognition algorithm achieves localization by calculating the matching degree between the contour point set of the wafer image contour model and the edge point set in the real-time image. Represented as: ； in, It is the total number of contour points in the wafer model. It is the number of contour points in the real-time image. It is the first Point weight, and Representing the first Model points Its corresponding edge points on the real-time image The distance error decay function.

6. The method for preventing and automatically correcting off-center soldering in a wire bonding machine according to claim 1, characterized in that, In large-scale chips, the initial positioning coordinates are determined based on the positions of at least two wafer reference points; in step S5, the position iteration update includes: obtaining the real-time position coordinates of each reference point, calculating the new initial positioning coordinates and wafer rotation angle based on the real-time position coordinates of each reference point, and updating the solder joint position coordinates.

7. The method for preventing and automatically correcting off-center soldering in a wire bonding machine according to claim 1, characterized in that, When the object to be processed contains multiple independent process matrices, the method adopts a hierarchical iterative control strategy: within the current process matrix, after completing the position verification and wire bonding operations of all wafers, it moves to the next process matrix and repeats steps S1 to S6.

8. The method for preventing and automatically correcting off-center soldering in a wire bonding machine according to claim 1, characterized in that, The iteration termination conditions include: the displacement deviation of the current workstation is less than the tolerance threshold, or the number of iterations for position updates reaches a preset upper limit.

9. A wire bonding machine's anti-offset soldering and automatic alignment system, used to implement the method as described in any one of claims 1 to 8, characterized in that, include: The vision module is used to acquire images of the wafer at the imaging station; A motion module is used to drive the vision module and the welding head to move between the imaging stations; it includes an XY motion platform, a linear motor fixedly connected to the XY motion platform, and a motion control card that is communicatively connected to the XY motion platform and the linear motor. The host computer processing module is communicatively connected to the vision module and the motion module, and is configured to execute the steps of the method to realize pre-welding position verification, deviation judgment, coordinate iterative update and control of welding execution; The image processing module is used for image preprocessing, image information analysis and calculation, and image display.

10. The wire bonding machine's off-center soldering prevention and automatic alignment system according to claim 9, characterized in that, The vision module includes an industrial camera, a lens module, and an independently controllable light source module containing axial and side lighting.

Images