In the semiconductor packaging industry, detecting pin coplanarity and dicing edge chipping has always been a crucial step in ensuring product quality. However, traditional 2D detection methods have many limitations and are difficult to meet the high-precision and high-efficiency detection requirements. DaoAI provides a brand-new solution for the industry with its advanced 3D vision technology.
In the semiconductor manufacturing field, packaging is a crucial process that directly affects the performance and reliability of chips. Among them, pin coplanarity and dicing edge chipping detection are key quality control points in the packaging process. Pin coplanarity requires that the bottom surfaces of all pins should fall within the same reference plane, which is crucial for chip mounting. If the pin coplanarity does not meet the standard, it will lead to virtual soldering during mounting, affecting the electrical connection and stability of the chip. Dicing edge chipping is a defect that occurs during the chip dicing process. The depth and length of the chipping can affect the mechanical properties of the chip and subsequent processing. In severe cases, it can even lead to crack propagation in the chip, ultimately causing the chip to fail. Traditional detection methods mainly rely on 2D vision technology, which obtains image information through top-down shooting. However, this method has obvious deficiencies in detecting pin coplanarity and dicing edge chipping.
Pain Points: Why Is It Difficult?
From the perspective of quantification, the essence of pin coplanarity is the height difference, and its accuracy requirements are usually in the micron level (μm). Traditional 2D top-down shooting cannot capture the height deviation information in the Z - direction, making it difficult to accurately quantify coplanarity. For dicing edge chipping, its depth also needs to be precisely quantified. However, the 2D grayscale image flattens the depth information, making it difficult to detect small and deep chipping, resulting in a high rate of missed detections. In addition, in terms of detection efficiency, in the past, only sampling detection could be used, and the sampled products were placed on special fixtures for detection. This method is not only inefficient but also cannot achieve 100% full-inspection, posing potential quality hazards.
The root cause of these detection difficulties lies in the limitations of traditional 2D vision technology. 2D images can only provide planar information and lack three-dimensional height information, which cannot accurately reflect the true shape of pins and dicing edges. Sampling detection cannot cover all products and cannot detect potential quality problems in a timely manner. Once defective products flow into the subsequent processes, it will lead to problems such as virtual soldering and crack propagation, increasing production costs and product defect rates.
Technical Principle
DaoAI's solution is based on advanced 3D vision technology. This technology uses special imaging hardware and algorithms to reconstruct the height of the bottom surface of each pin of each device. Specifically, the 3D vision system emits specific light to the device surface, and the reflected light is captured by the camera. Based on information such as the reflection angle and time of the light, the algorithm can accurately calculate the height of the bottom surface of each pin. Compared with the traditional 2D method, 3D vision can obtain complete three-dimensional information, directly calculate the coplanarity deviation, and compare it with the pre-set benchmark. Once the height deviation of a certain pin exceeds the standard, the system can accurately locate the out - of - tolerance pin.
For dicing edge chipping detection, the same set of three-dimensional data is used to scan the dicing edge side by side. Since the 3D data contains depth information, the length and depth of the chipping can be accurately quantified, and small and deep chipping will no longer be masked by grayscale. The detected defects will be classified and graded by AI-ADC. AI-ADC uses artificial intelligence algorithms to analyze and classify the defects and grade them according to the severity of the defects. At the same time, the device edge-side completes the inference, which means that the detection process can be quickly completed locally on the device without transmitting a large amount of data to a remote server, thereby matching the high-speed rhythm of the production line and improving the detection efficiency.
Typical Application Scenarios
- Pin coplanarity detection: Before chip mounting, it is necessary to ensure that the pin coplanarity meets the requirements. Traditional 2D detection is difficult to detect small Z - direction deviations, while 3D vision can measure the height of each pin, accurately calculate the coplanarity deviation, and accurately locate the out - of - tolerance pins, avoiding virtual soldering problems during mounting.
- Dicing edge chipping detection: After the chip dicing process, there may be chipping defects on the dicing edge. 3D vision scans the dicing edge side by side, which can accurately quantify the length and depth of the chipping, effectively detect small and deep chipping, and prevent crack propagation in subsequent processes.
- Pin deformation detection: During the packaging process, the pins may be deformed. 3D vision can obtain the three-dimensional shape information of the pins and compare it with the standard model to detect the deformation of the pins in time and ensure the normal function of the pins.
- Chip surface flatness detection: The flatness of the chip surface also affects its performance. 3D vision can comprehensively scan the chip surface to detect small undulations on the surface and ensure that the chip surface meets the flatness requirements.
Implementation Case
A medium-sized semiconductor packaging factory produces a variety of chip packages. Before introducing DaoAI's solution, the factory used the traditional 2D vision sampling detection method. Pin coplanarity detection could only be carried out by sampling, which was inefficient, and the missed detection rate was as high as 15%. In the dicing edge chipping detection, the missed detection rate of small and deep chipping reached 20%, resulting in a defect rate of about 10% in the subsequent processes due to virtual soldering and crack propagation. During the implementation of the DaoAI solution, the technical team conducted a detailed investigation and evaluation of the production line and customized the system deployment according to the actual situation of the production line. After a period of debugging and optimization, the system was successfully launched and put into operation.
3D vision technology has brought a qualitative leap to semiconductor packaging defect detection, effectively solving the problems of traditional detection methods.
DaoAI's Solution and Product
DaoAI's solution is centered around 3D vision, combined with AI-ADC technology and edge-side inference capabilities, providing a complete defect detection solution for the semiconductor packaging industry. Its product can achieve 100% full-inspection of pin coplanarity and dicing edge chipping, and the accuracy of measuring the height of each pin reaches the micron level. By reconstructing the bottom surface height of each pin of each device and scanning the dicing edge side by side, the coplanarity deviation and the length and depth of the chipping can be accurately quantified. At the same time, AI-ADC classifies and grades the detected defects, and the edge-side inference ensures that the detection results can be quickly fed back, matching the high-speed rhythm of the production line and improving production efficiency.
Quantitative Results: After introducing DaoAI's solution, the packaging factory achieved online full-inspection of out - of - tolerance pin coplanarity, and the missed detection rate decreased from 15% to less than 1%. The missed detection rate of small and deep dicing edge chipping decreased significantly from 20% to less than 2%. The two types of defects that previously relied on sampling were successfully converted to pin - by - pin and edge - by - edge quantitative detection. The defect rate in the subsequent processes due to virtual soldering and crack propagation decreased from 10% to less than 3%, significantly improving product quality. At the same time, the quality data is more traceable, providing strong support for the optimization of the production process and quality control.
FAQ
What problems exist in the detection of pin coplanarity and dicing edge chipping?
The essence of pin coplanarity is the height difference. 2D top-down shooting cannot capture the Z - direction deviation. In the past, sampling detection was inefficient and could not achieve full-inspection. The depth of dicing edge chipping is flattened by the 2D grayscale image, and small and deep chipping is often missed. These problems make it difficult to guarantee product quality and increase the defect rate in subsequent processes.
How does DaoAI solve the problems of pin coplanarity and dicing edge chipping detection?
DaoAI uses 3D vision to reconstruct the bottom surface height of each pin of each device, calculates the coplanarity deviation and locates the out - of - tolerance pins. It scans the dicing edge side by side, quantifies the length and depth of the chipping, classifies and grades the defects through AI-ADC, and the edge-side inference matches the rhythm of the production line to achieve efficient and accurate detection.
What are the effects after the implementation of the DaoAI solution?
After implementation, online full-inspection of pin coplanarity can be achieved and the specific pins can be accurately identified. The missed detection of small and deep dicing edge chipping is significantly reduced. The two types of defects are converted to pin - by - pin and edge - by - edge quantification, reducing defects in subsequent processes. The quality data is more traceable, improving the overall 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.