I've spent from
I think that these data support a particular root cause for
the wire bond ruptures.
The design, glob-top and all, is OK, on the evidence of more
than two dozen full-up boards made with either zero wire bond breaks, or a very
few.
Then, on a particular production day late last spring,
boards for the first time began to show hundreds of open wire bonds. After
than, there were some days on which boards with zero defective wire bonds were made,
and other days on which the boards had many dozens of broken wire bonds. This strongly suggests a production-lot
related problem.
A number of observations (Diane Kolos's
images, observations by several different people at different times and
locations) show that the glob-top resin is debonded
and separated from the pc board material at (and near) the wire bonding
site. This places huge stresses onto the
bond heel, which Diane's images show has (in two distinct cases) ruptured unter tensile stress --- just as we would expect to happen when
this separation happens.
Inspection of the assembly process has shown that a Kapton
tape was applied to precisely the region that was later used for wire bonding, and
on the ends that are rupturing. Vapor
degreasing and plasma etching were used to prepare this region for later wire
bonding.
Kapton tape is supplied by MMM with two alternative
adhesives: acrylic-based or silicone-based.
We know that the manufacturer had silicone-based adhesive in its
assembly areas: Jim cound not say whether we know
whether this silicone-based Kapton tape was applied to the MCMs, or not. Now, all we can say is that this is possible.
It would certainly explain the lack of adhesion of the
glob-top to the polyimide pc board: silicone contaminants cannot be removed by
normal vapor degreasing or by normal plasma etching or combinations of these. And
silicone contamination is effective at limiting adhesion of glob-top resins
(including resins) in monolayer thicknesses.
Diane removed the glob-top resin from a specimen she was
working on, and noted that the adhesion was sharply limited or not present. Examination
of the exposed surface using EDS did not show any contaminating materials ---
but it could not have seen anything thinner than about 1 um, and silicone is
effective as an adhesion-limiting layer at a thickness of less then 1 nm --- a
thousand times thinner than EDS can "see".
Jin and I sketched some other tests that can be carried out
to gather evidence bearing on this possible root cause. We hope these may help in deciding what risk
there is in using the several hundred already made assemblies.
Regards,
Henning Leidecker