Predictions Software TSMC DFM Compliant Logo

Defect Data

To generate an accurate yield model accurate defect data is essential. The critical area method may require defects to be classified in a different way to the current practice in your fab. The defect data required for a critical area based yield model is all of the defects present in the process that could cause a fault (if they landed in the wrong place). Therefore, it is important when collecting defect data to consider all defects irrespective of where they land on the chip. So even if the defect falls in an area with no layout it should still be "counted" as a defect. This is because we are trying to build up picture of the defect statistics independently of the layout. The critical area measurements take care of the layout interaction with defects. Of course only real defects should be counted and not nuisance or cosmetic defects.

Defect Classification

Defect should be classified according to their fault mechanism. That is shorting faults, breaks, pinhole, dielectric faults etc., so that they can be matched to the critical area measurements. All faults that cause the same kind of fault in a layer can be grouped together or if they can be put into sub-groups (e.g., black dome shorts, pancake shorts) counted in a sub-group. This way the yield loss associated with the sub-group can be identified (as shown here).
A defect shorting conductor lines causing a fault.
A defect shorting conductor lines causing a fault.

If you would like more information on how fabs have tackled the preparation of defect data for critical area based yield modeling please contact us.




© Copyright