Self-Feedback Matching For Vision Alignment System In Wafer-Level Flip-Chip Packaging

In the “Post-Moore Era”,miniaturization,multi-functionalization,and integration of electronic devices call for more advanced integrated chip packaging technology.As one of the advanced packaging technologies,wafer-level direct flip chip packaging technology has been widely concerned with its direct interconnection characteristics.In flip chip bonding,vision alignment system provides the position deviation information between chip and wafer,which is one of the key technologies of wafer-level direct flip chip packaging equipment.In this paper,a novel vision detection algorithm and the vision alignment system with this algorithm are proposed,the details are as follow:1.The development of chip packaging technology and the key technologies of waferlevel chip flip chip packaging are introduced,which reflects that the vision detection algorithm affects the efficiency of flip chip package.The requirements and research status of the visual alignment algorithm in wafer-level direct flip chip bonding are analyzed and there is scope for further improvement in accuracy and speed of template matching.From above,the research direction and contents of this paper are determined.2.The sequence of wafer-level direct flip package equipment is investigated and the function of vision alignment system is analyzed.From the analysis,the camera,lens,image process platform and so on are provided,and the light path and illumination scheme of dual camera & reflecting prism are designed to improve the speed of visual positioning and image quality.In addition,the imaging formation of vision alignment system are proposed,which realizes the transformation from image coordinate system to motion stage coordinate system.3.The shortcomings of traditional template matching and sub-pixel interpolation image in positioning task are analyzed and the existing template matching methods with the underutilization of matching information in the process of matching,constant matching speed and no relationship between pixel and subpixel,are pointed out.According to the analysis above,a “Pixel-Subpixel” intelligent regulation self-feedback matching algorithm with intelligent speed regulation and “Pixel-Subpixel” adaptive switching,and the reliability and robustness of the algorithm are verified theoretically to make sure the algorithm works on a stable manner.4.The experiment system is provided and the experiment which includes the designed algorithm positioning accuracy comparison with other five algorithms' s and designed algorithm positioning accuracy on different brightness and noise images are carried out.The result shows that the self-feedback matching algorithm has the minimum positioning error(0.7?m)and the matching speed(86ms)next to the speed of tiny-yolov2 algorithm.Not only that,designed algorithm performs stable on the image which is affected by-30 ? 30 linear illumination,0.1?0.25 variance Gaussian noise and 0.15?0.35 density pepper and salt noise,respectively.In addition,with C++ programing tools and UML,the wafer-level direct flip chip vision servo software based on designed algorithm is provided,and the validity of the software is verified by the concrete operation.