Folks,

I have been reviewing all the issues/problems that were faced by my host
company and its supplier ? before I arrived. I would like to share some for
comment as well as my observations, if you would be so kind.

We have one board design using seven .8 mm ceramic CSP?s made by TI. The
first issue facing my company was cracking or separation of balls from the
device. I have reports from Rockwell, who did the F/A and from TI both
pointing to the failure mechanism as possible excessive IMC formation as in
the following without the photos, of course:

Attached are four photos:

CSP-1 shows multiple insertions into sockets.
CSP-2 shows x-section of stock unit
CSP-3 shows magnified x-section of stock unit. Ni/Sn intermetallic layer
appears thick.
CSP-4 shows x-section of device which broke off the PCB - clean break at the
intermetallic layer.

Because of the multiple insertions into sockets, as evidenced by
indentations on the balls, it is considered that multiple thermal excursions
contributed to IMC growth. As the clean break occurred at the IMC interface,
again, this is thought to be the failure mechanism. This may be so, but
other things must be considered as well.

This issue possibly was resolved with TI?s "prototype" parts and we?ve seen
no repeat of this occurrence. No reason is available to me, at least,
concerning the part?s improvement. No matter, there now exist other issues.

I believe, as in a few comment trades, with others (Dave Fish for one), that
our assembler?s inexperience profiling and soldering these devices has been
responsible for poor quality/reliability solder joints. In their defense,
there aren?t/weren?t many suppliers with this experience. Simply, cold
solder joints probably contributed to unacceptability and complete failure
as received ? evidenced by the infamous "C" clamp. The ceramic portion of
the device is extremely massive compared to ball size and the ability to get
required heat to the solder medium and balls to effect acceptable solder
joints. This means, I believe based on past experience, the reflow profile
must be "maxed" out in terms of solder paste performance and joint
formation. I don?t think this was done.

To add to the situation, the CTE mis-match of ceramic and PCB material
certainly has been known to ruin even good solder joints. This may account
to the failure at the IMC interface as well as at the board level.
Therefore, a "too thick" IMC layer probably is/was not the failure
mechanism, or was it?

All solder pads have been re-evaluated and now definitely meet alignment and
size requirements. Stencil apertures are as specified. Solder paste now is
Kester?s 562R, solder volume is acceptable, surface solderability is as
required. What remains is developing a solder reflow profile that will
effect acceptable solder joints for the CSP?s as well as all other devices
on the board.

I realize this should not be a big issue, but having had to go way back to
visit initial failures at the device ball interface, I have some concerns
about whoever builds our next lot of boards. Steve, are you listening.

I would appreciate any comments concerning this small part of the picture. I
would appreciate anyone sharing experiences with what should be another BGA
success story but for what I have presented here.

Thank you all much,

Earl