Detecting Edge Chipping and Burrs After Dicing
Differentiate detects from acceptable cut marks after the dicing process
Graphical programming environment for deep learning-based industrial image analysis
After the wafer undergoes various layering and etching processes, it is diced to release the individual dies. Following this procedure, a die may have chipping or burr marks along the kerf. Chipping and burrs affect IC device quality, so it is important to inspect them after dicing. A higher-than-average number of chips outside of tolerance may also suggest the dicing saw blade needs to be adjusted or replaced.
A common way to inspect the die is using rule-based machine vision, but it is often unreliable because the chips and burrs are highly variable and hard to distinguish from normal dicing marks or IC patterns. It is difficult to develop machine vision algorithms to cover all variations and distinguish unacceptable marks from marks that are within tolerance.
Cognex Deep Learning tools offer a simpler way to learn and classify chipping and burr marks and differentiate them from normal cut marks after the dicing process. The software can easily be trained to identify all chips and burrs, classify them as acceptable or unacceptable and ignore normal marks that are within tolerance.
Using this information, manufacturers can optimize the cutting process, for example, by replacing the diamond saw blade that has become too dull or too wide. Another benefit of properly detecting the difference between OK and No Good (NG) is an increased yield of good chips that otherwise might have been thrown away due to false reading.