Q: How do I address test time when pass and fail times differ significantly?
A: In designing PRO COOL® for Wafer Sort & Final Test, we assumed that a bad device takes as long to test as a good device. Thus, the actual yield of the product being tested is irrelevant to test cost of ownership. In a previous product (COOL ONE® for Wafer Sort & Final Test), good device pass time and bad device fail time were considered separately. As a result of several inputs from our customer design review group, we did not carry that feature forward. Users commented that the more complex testing and higher yields of current devices were minimizing the pass/fail test time difference.
Of course tester related yield issues such as accuracy, mishandling and destructive testing are considered. These are what are included in equipment yield (mishandling and destructive testing) and parameteric yield (repeatability and accuracy) in the TWO COOL® models. Thus, the TWO COOL® models should be created without regard for the non-tester driven yields.
Now having stated our assumptions, what should you do if good test pass time and bad test fail time significantly differ for devices with large yield loss? We recommend that a yield weighted test time be entered into the test cell requirements screen. Yield weighted test time can be estimated using the following formula:
ave test time = (yield * pass time) + ((1 - yield) * fail time)
For example, at 75% yield, with a pass time of 60 seconds and a fail time of 10 seconds, yield weighted average test time would be 47.5 seconds. Since PRO COOL® for Sort and Test uses the life of the equipment for calculations, the average test time will give a good estimate of test COO over the long term.
Q: Are any special considerations needed when creating TWO COOL® models for subsequent use in PRO COOL®?
A: PRO COOL® process cost analysis software uses TWO COOL® cost of ownership software to estimate the cost of a sequence of process steps or a cluster of process tools. PRO COOL® reads the 13 TWO COOL® cost drivers from the COO model summary along with yield and utilization information. The use of COO summary information can lead to differences in the way PRO COOL® and TWO COOL® respond as utilization or throughput rates are changed. These differences can be minimized by taking the following precautions in constructing COO models specifically for use in PRO COOL®:
Enter material, consumable, and utility costs as a cost per unit (wafer, package, or mask) rather than as an annual cost. (For further information see PRO COOL® Technical Alert #5, "PRO COOL® vs. TWO COOL® Comparison.")
Enter software development costs, including test software, as follows:
If the development costs are in currency (per year) use "Annual Software Contract per System"
If the development costs are in hours (per week) use "Personnel/System: Engineers"
Input a typical throughput rate for the test cell as the initial module throughput rate in COO models used for PRO COOL® for Sort and Test.
In all cases differences are minimized as TWO COOL® utilization and throughput inputs approximate PRO COOL® conditions.
Call today for more information from Wright Williams & Kelly, Inc. Providing business
solutions for productivity and cost management since 1991.