Semiconductor Die Surface Inspection
Detect imperfections that can affect the quality and performance of the die
Graphical programming environment for deep learning-based industrial image analysis
In the integrated circuit manufacturing process, each die must be inspected to check the surface for cracks, chips, burnt marks, etc. as these imperfections can negatively affect the quality and performance of the die. These defects are variable and in different locations, so it is challenging for rule-based machine vision to accurately find them in a timely fashion. Since normal aberrations that do not affect the quality of the chip also occur, it is important to not spend time flagging these minor defects. Given the size and volume of chips that are processed daily, human inspection is neither efficient nor practical. In addition, minimizing human interaction reduces the chance of contaminants entering the cleanroom.
Cognex Deep Learning’s defect detection tool can find a wide range of unacceptable cosmetic defects on die surfaces that are otherwise be too complex or time-consuming for rules-based vision inspection systems. The tool examines the surface of the die to detect any combination of cracks, chips, or burn marks. The software is trained with various images that illustrate the variability in type and location of defects. After identifying potential areas of interest, the Deep Learning classification tool categorizes defects (such as cracks, chips, dust spots, etc.). Using this information, process improvements can be made to decrease defects and increase yields.