I would agree with this, but still concerned...
Is not it true that the new design, especially
so revolutionary as Willamette, with 3HGz pipes
and lots of domino logic, would require separate
qualification since it is (might be?) so different
from P-iii that the old data are not applicable
at all?
Some of it is, some of it's not. A good example that I can take from my experience, is that one of the biggest die yield hitters that I have experienced is particle defects. I'm mostly involved in Dry Etch, and so from process to process the tools in my area don't change too much. I etch whatever Litho sends me. What I have found, with the changing of the recipes, is changes in defect levels. If I have a product that I can run on the new process that ran on the same tool in the old process, it is invaluable to me to examine the old KLA/SEM/Tencor and any other defect related data from the old process, and compare it to the new process. Same tool, same product, different process.
By examining the wafer maps, and related Defect Metrology data, it becomes easier to determine things that are specifically related to the process, if you have an established database on the product, even if it were run on a previous process.
If I'm also using a new product, I find myself saying.... well, it COULD be the process, or it MIGHT be the product, or MAYBE the tool. On the other hand, If I find the yield hitters are in the same spot in the wafers, on the same product, as the previous process, than I can reasonably assume... A-HA, it's the tool, or the product, not the process..... Then I can put my new product on the process, find out if it IS a product specific problem, and than work the tool issue, if needed.
Granted, it doesn't always work out that easily, but it seems to me that I've had an easier time using old products, on new processes, than using new product, on new processes.
That's just my opinion, of course.
SemiconEng |
