The flower basket test is the test installed between the cleaning before inspection and the sorting of silicon wafers. Remove defective silicon wafers to reduce the chance of sorting machine downtime, thereby increasing the overall production capacity.
Share
1). Project background:
Project requirements:
Due to the residual damaged silicon in the silicon sorting machine, resulting in downtime, it is necessary to manually remove the broken silicon in the machine, which takes a long time and directly affects the production capacity. The customer needs to take out the broken silicon wafer outside the sorting machine, and reserve the discharge space for multiple groups of flower baskets, so that a worker can take into account multiple lines, achieve productivity improvement while avoiding a large amount of personnel investment.
Technical requirements:
The silicon wafer in the container is tested as a whole to detect whether the silicon wafer in the flower basket has many pieces, few pieces, thick pieces, broken pieces, and whether the silicon wafer has wrong teeth. The defective products are selected to improve the yield. The CT time required is within 3s.
2). Solution architecture:
The scheme uses 5 million small area array camera and VM algorithm development platform to realize the detection of flower basket silicon wafer.
(1)Process idea:
(2) Selection idea: The detection target is the silicon wafer in the flower basket. In order to maintain the stability of the feature and avoid the height difference missed detection caused by the excessive incidence Angle, the 4-hole surface light is selected to cover the imaging field of view, and the incidence Angle of the camera is reduced and the risk of height difference missed detection is reduced by successive detection by multiple cameras and multiple regions.
(3) Debugging ideas: Adjust the exposure imaging, so that the silicon wafer in each camera imaging gray level is roughly unified, to ensure the silicon wafer edge and background 1 to 2 pixel transition zone imaging quality, adjust the sensor position and check with the electrical, as far as possible to ensure that the flower basket in the camera imaging center.
3). Program advantages:
1. image quality: uniform brightness, no stretching deformation, edge transition zone clear.
2. algorithm time: industrial computer 1 drag 8, double conveyor belt detection and stability takes 2s.
3. scene adaptability: miss rate <1%, false detection rate <1%.
4. customer value: the action of taking out the broken silicon wafer is implemented outside the sorting machine, and the layout space of multiple groups of flower baskets is reserved, so that a worker can take into account multiple lines, achieve productivity improvement while avoiding a large amount of personnel investment.
EN
