Posted by rcxiao: Tue Jul 18, 2006 3:03 pm    Post subject: test correlation

Hi all,
How do you usually do statistics study on site to site gap? I know there is repeatability can analyze individual site capability. Then can reproducibility be used to study site to site, and what is the acceptable study result?
Thanks
RC

Posted by FARel Engr: Thu Jul 20, 2006 5:04 pm    Post subject:

Hi RC,

One way to do it is to take representative test data per parameter per device type from the new site (involving at least 3 lots for each device type), and computing a theoretical guardband that each parameter must have against your datasheet limits (presumably these are also the old and new sites' QA test limits), in order to achieve your existing outgoing ppm goals.

If the computed theoretical guardbands are all the same as (or smaller) the guardbands used between the production and QA tests of the old site, then you may assume that the new site is the same as the old site.

FARel Engr

Posted by PW57 Guy: Thu Sep 28, 2006 6:51 am    Post subject:

making correlation is very tedious job...

a good correlation for both sites must have at least <5%. Acceptable range is within 5-10% but reconsider the possibility that this may or may not affect the ppm of the parts. For correlation >10%, you must investigate why the correlation factor is big.

for my experience, i'll do the correlation by doing repeatability on both sites, the do SPC on it. if the above criteria was met for <5%, then you have good correlation.
_________________
...little knowledge is very dangerous...