Installation guide

Workpiece Alignment
Advanced Dicing Technologies Ltd.
6.1-18
sharp enough, the user can teach a sub-model (clear model not on the same
street), which will be used to verify the main models’ thresholds.
Normally, the system is set up to use models, where the threshold would
not be lower than 80%. When teaching alignment for a workpiece with
corrupted models, models with threshold set between 65% and 80% are
used. These models require verification against the sub-model.
When teaching alignment for a workpiece with corrupted models, the
threshold for these models should be set to 65% - 70%. In the recipe, the
high score parameter should be set to 80% or higher. It is also
recommended to teach a sub-model for such a workpiece.
The algorithm is used in case the main model found has the threshold score
between the taught score and the high score (65% - 80%). If it succeeds, and
the found distance between the main and the low models is within
tolerance (see Main to Sub-model accuracy values below), the main
model is regarded as properly found.
The parameters are:
Use sub-model: Use the sub-model detection for low threshold main
model detection or not. If not then no change is made to the algorithm.
Low model high score (%): The high score for low models, above which
the low model is considered found, with no need to verify it through finding
a sub-model.
Main model high score (%): The high score for main models, above
which the main model is considered found, with no need to verify it through
finding a sub-model (currently not in use).
Sub-model high score (%): The high score for sub-models, above which
the sub-model is considered found.
Main to Sub-model accuracy: the maximum allowed distance (for both
X and Y coordinates) of the found sub-model from the found main model in
order for the detection to be considered successful. The recommended
values for this parameter are:
• For x240 magnification systems: 2 microns.
• For x120 magnification systems: 4 microns.
• For x60 magnification systems: 7 microns.
Note: If parameter Use sub-model = Yes, the models should be taught with a
60%-70% threshold. Otherwise, they should be taught with a regular (80%)
score