3D AI Inspection for Hidden Solder Joints of BGA/QFN: Clearly Detect Voids, Bridging, and Non - wetting Defects at Once

As electronic devices continue to develop towards miniaturization and high-performance, bottom-terminal packages such as BGA and QFN are widely used in communication and high-density modules. However, the hidden solder joints of these packages have always been a difficult problem in inspection. WeLinkirt's DaoAI 3D AI-AOI inspection equipment provides an effective solution to this problem.

Industry background and user scenarios: In today's electronics manufacturing field, the pursuit of miniaturization and high-performance in products such as communication and high-density modules has promoted the wide application of bottom-terminal packaging technologies such as BGA (Ball Grid Array) and QFN (Quad Flat No - leads). These packaging forms can achieve more electrical connections in a limited space, improving the integration and performance of products. However, this packaging method also brings a thorny problem, that is, the solder joints are naturally blocked by the package. In the actual production process, factories need to strictly detect the quality of these solder joints to ensure the reliability and stability of products. But traditional planar optical detection methods cannot see the real state of the bottom solder balls, which poses a huge challenge to the detection work.

Pain points: Why is it difficult?

From the quantitative dimension, in terms of detection coverage, the traditional X-ray sampling inspection method can only cover part of the products. The sampling ratio is usually about 10% -20%, which means that the quality of hidden solder joints of most products cannot be comprehensively detected. In terms of detection rhythm, the X-ray detection speed is slow, and its rhythm often cannot keep up with the production speed of the SMT line. The SMT line can complete the placement of 50-100 components per minute, while the X-ray detection can only process 10-20 products per minute. In terms of defect detection rate, the detection rate of conventional 2D AOI for defects such as voids, bridging, and non-wetting of hidden solder joints is relatively low, only reaching 60% -70%.

The root cause of these difficult - to - solve problems lies in the hidden characteristics of the solder joints. The solder joints of BGA and QFN are located below the components and are tightly blocked by the package. Planar optical detection can only see the outer edge of the components and cannot obtain detailed information about the bottom solder balls. Although X-ray sampling inspection can penetrate the package to see the solder joints, due to its sampling inspection method, it cannot guarantee the detection of all products. In addition, defects such as voids, bridging, and non-wetting have similar performances in grayscale images. Relying on traditional grayscale analysis, it is difficult to stably distinguish these defects, resulting in frequent misjudgments and missed judgments. For new package models, due to the lack of sufficient defect samples, it is difficult for traditional detection models to quickly and accurately build models, further affecting the detection efficiency and accuracy.

Technical principle

The DaoAI 3D AI-AOI inspection equipment adopts advanced technical mechanisms. In terms of imaging, it performs 3D height imaging on the solder joint area. Through a special optical system and algorithm, it can accurately obtain the three-dimensional height information of the solder balls and clearly restore the bottom shape of the solder balls hidden under the package. In terms of algorithm, it combines an AI model to interpret the shape of the solder balls. The AI model has been trained with a large amount of data and can learn the characteristics of different defect types. It uses three-dimensional features to stably distinguish defects such as voids, bridging, and non-wetting.

Compared with traditional methods, traditional 2D AOI only relies on grayscale information for detection and has difficulty distinguishing defects with similar grayscales. The three-dimensional information of DaoAI 3D AI-AOI provides an additional height-dimension basis for defect interpretation, greatly improving the detection accuracy. Although the X-ray sampling inspection method can see the solder joints, it has the problems of incomplete coverage and slow rhythm. DaoAI 3D AI-AOI can operate online according to the SMT rhythm to achieve full inspection and can promptly detect batch-type welding abnormalities. In addition, the APDT few-shot learning algorithm adopted by the device can quickly build models even when there are few defect samples for new package models, improving the adaptability and flexibility of the device.

Typical application scenarios

• Void detection: When detecting void defects, DaoAI 3D AI-AOI first performs 3D imaging on the solder joint area to obtain the height information of the solder balls. The existence of voids will cause local height abnormalities in the solder balls. The AI model can accurately identify void defects by analyzing these height changes. The difficulty lies in the fact that the sizes and positions of voids vary, and some tiny voids show insignificant height changes, which requires high-precision imaging and complex algorithms for judgment.
• Bridging detection: For bridging defects, the device judges by detecting the height and connection between adjacent solder balls. When there is an abnormal connection between adjacent solder balls, their height and shape will change, and the AI model identifies bridging based on these characteristics. The difficulty lies in the diverse forms of bridging. Sometimes, it may only be a slight connection, which is difficult to distinguish from the normal shape of solder balls, and requires accurate three-dimensional feature extraction and analysis.
• Non - wetting detection: Non - wetting defects are manifested as poor bonding between the solder balls and the pads. In 3D imaging, the height and shape of the non-wetting area will be significantly different from the normal area. The AI model identifies non-wetting defects by comparing these differences. The difficulty lies in the different degrees and ranges of non-wetting, and it may be affected by factors such as surface impurities, which requires careful image processing and analysis.
• Collapse detection: Collapse defects will cause the height of the solder balls to decrease. DaoAI 3D AI-AOI judges whether there is collapse by detecting the height change of the solder balls. The difficulty lies in accurately distinguishing normal height fluctuations from collapse defects, which requires establishing reasonable height thresholds and judgment rules.
• Poor - wetting detection: Poor wetting will affect the connection quality between the solder balls and the pads. The device detects poor wetting by analyzing the three-dimensional shape of the solder balls and their contact with the pads. The difficulty lies in the complex manifestation of poor wetting, which requires comprehensive consideration of multiple factors, such as the shape, angle, and contact area of the solder balls with the pads.

Implementation case

There is a medium-sized electronics manufacturing factory that mainly produces communication and high-density module products. Before introducing the DaoAI 3D AI-AOI inspection equipment, it had been using a combination of X-ray sampling inspection and conventional 2D AOI for detection. The implementation process was relatively smooth. WeLinkirt's technical team provided professional training to the factory employees to ensure that they could operate the equipment proficiently. Before the implementation, the comprehensive detection rate of hidden solder joint defects was only about 70%. The defective rate of welding-related problems found in the subsequent functional testing reached 5% -8%, and X-ray detection bore most of the detection work. After the implementation, the situation improved significantly.

There's no need to rely on sampling inspection luck for invisible solder balls. DaoAI 3D AI-AOI inspects every single one.

WeLinkirt's solution and product

WeLinkirt's DaoAI 3D AI-AOI inspection equipment is specially designed to solve the difficult problem of detecting hidden solder joints of BGA and QFN. The device integrates advanced 3D imaging technology and AI algorithms, enabling online full inspection of hidden solder joints. Its APDT few-shot learning function allows new package models to quickly build models even when there are few defect samples, greatly shortening the model-changing time. The operation interface of the device is simple and intuitive, easy for employees to operate. At the same time, it has efficient data analysis and management functions, which can promptly feedback detection results and provide strong support for production decision-making.

Quantitative results: After the implementation, the comprehensive detection rate of hidden solder joint defects such as voids, bridging, and non-wetting reached over 99%, nearly 30 percentage points higher than before. The defective rate of welding-related problems found in the subsequent functional testing decreased significantly, dropping to below 1%, a reduction of over 80%. X-ray detection is now only used for confirming difficult cases instead of bearing all the detection work, and its usage frequency has decreased significantly, and the detection cost has also been correspondingly reduced. This fully demonstrates the remarkable results of the DaoAI 3D AI-AOI inspection equipment in improving detection efficiency and product quality.

This article was generated by AI. Customer cases are simulated scenarios based on real product capabilities and figures are illustrative; see product pages for official benchmarks.