3D AI AOI Device Detects Defects in Chip Pins and Dicing

In semiconductor chip production, the detection of pin coplanarity and dicing chipping is crucial. WeLinkirt's 3D AI AOI device provides an effective detection solution for this industry with its advanced technology.

User Scenario: A leading semiconductor chip manufacturer needs to strictly inspect the pin coplanarity and dicing chipping of chips in the pin soldering and dicing processes of the chip packaging production line. Its products are high-performance semiconductor chips, and the inspection objects include the pins of the chips and the dicing edges.

Pain Points: In the current trend of applying domestic GPU direct-through solutions in industrial vision quality inspection, the manufacturer faces many difficulties. The miss-detection rate of traditional detection methods is as high as 3%, which causes some defective products to flow into subsequent processes, increasing production risks and costs. At the same time, the false-alarm rate reaches 15%, resulting in a large number of good products being misjudged as defective, which increases the workload of re-inspection and labor costs. In addition, the model-changing time is as long as 30 minutes, seriously affecting production efficiency and difficult to meet the production demand for rapid model-changing.

Technical Principle

WeLinkirt's 3D AI AOI device uses a self-developed 3D camera to collect images. Through the three-dimensional topography reconstruction technology, the collected image data is converted into point-cloud data. The device uses advanced algorithms to analyze the three-dimensional topography of the chip, and can accurately identify defects such as hidden solder joints, coplanarity, micron-level topography, and air holes in the 2D optical blind area. The principle is that 3D imaging can obtain the three-dimensional information of the chip. Compared with 2D imaging, it can present the surface characteristics of the chip more comprehensively, thus effectively detecting defects that 2D imaging cannot find. For example, for the detection of pin coplanarity, 3D point-cloud data can accurately measure the height and position of each pin and compare it with the standard value to accurately judge whether the pins are coplanar; for the detection of dicing chipping, 3D imaging can clearly capture the tiny defects on the dicing edge, avoiding miss-detection.

• Self - developed 3D camera: It features high resolution and high precision, and can collect clear and accurate chip image data.
• Three - dimensional topography reconstruction: It converts 2D image data into 3D point-cloud data, providing more comprehensive information for defect detection.
• Advanced algorithm analysis: Through in - depth analysis of 3D point-cloud data, it can accurately identify various defects.
• 2D-3D fusion: It combines the texture information of 2D images and the three-dimensional information of 3D point-clouds to improve the accuracy of detection.

WeLinkirt's Solution and Product

Centered on the 3D AI AOI device, WeLinkirt provides a complete inspection solution for the manufacturer. The device can quickly and accurately detect the pin coplanarity and dicing chipping of chips. The supporting DaoAI AI AOI software system can realize the feature recognition of the visual basic model. With only one good product, it can complete 0-code automatic programming in 5 minutes. Through APDT positive-sample/few-sample learning (only 1-20 good products are needed), it effectively reduces the programming time and sample requirements. At the same time, the semantic false-alarm filtering function can reduce the false-alarm rate. In the actual implementation process, the device is installed on the chip packaging production line. Through real-time inspection of chips, the inspection results are timely fed back to the production system for timely processing of defective products.

The 3D AI AOI device provides a precise and efficient solution for semiconductor chip inspection.

Quantitative Results: After using WeLinkirt's 3D AI AOI device, the detection rate of chip pin coplanarity and dicing chipping has increased to 98.5%, and the miss-detection rate has decreased to <1.5%. The false-alarm rate has decreased by -68%, significantly reducing the re-inspection workload. The model-changing time has been shortened to 5min, greatly improving production efficiency and better adapting to the production demand for rapid model-changing.