I would like to perform a sanity check regarding the use of 
non-functional pads on high layer count multilayer designs.  First I'd 
like to present a small scenario.

You're at your desk, doing whatever you do.  The phone rings.  Behold, 
it's your customer and the conversation goes something like this.  " I've 
got these ___(insert a large quantity) assemblies and we have opens at 
functional test."  " When I re- test, there seem to be additional 
failures."    " I was just wondering, do you know what might be 
happening?"

After the phrase "Value Added" flashes through your brain, you respond 
with the usual.  "I'll check it out and get back to you."  You reach for 
the antacid and think.   "There's another afternoon shot to hell!" Of 
course, it's either a high visibility product or something built and 
shipped six months ago, and we all know anything that old can't be 
researched.

We've all seen .187 thick panels built with little or no consideration 
for thermal abuse.  They survive just fine.  We've also seen .062 PCB's 
that look so poor, they must belong to somebody else..or do they?

Back to the subject.....

This is not a specific problem solving request and does not relate to any 
particular processing issue.  I am looking for an updated compromise 
strategy that will minimize the impact to all or most areas of 
manufacturing.  It seems that today's board  designs and manufacturing 
yield  considerations have made the addition of non-functional pads more 
of a curse than a cure.  Gone are the padmasters you just superimpose and 
be done with it!  Following are some of the mfg. issues:

1)  The addition of non functional pads increase the opportunities for 
innerlayer shorting defects, along with escapes from AOI.
2)  Drilling is impacted, you can't just stack three high with #80 drills 
and let 'er rip.

3)  Adding non functional pads to CAD data that originally did not 
incorporate them requires careful DRC and netlist retesting.  As often as 
not, the designer has encroached upon those areas and at best you can 
only add pads of a smaller diameter.  This presents a new set of problems 
such as pad spin etc.

We currently have specific criteria for incorporation.  Of course, we 
also have additional options and material upgrades  from 140 to 170 Tg 
laminate.  Like our competition, we use the best materials, chemistry 
etc. and  apply what appear to be very good process controls.  We are 
acutely aware of material constructions, aspect ratios, drill 
de-optimization etc.  and all the "good stuff".

Several questions regarding criteria  that come to mind;

1)  Is a tented via less prone to failure than an exposed one? Could non- 
functionals be omitted?
2)  Should non- functionals only be added to extremely dense areas such 
as PGA, BGA etc.
3)  Should I really compromise and just add them to the outermost 
innerlayer signals?

Your opinions and comments are appreciated.

PS:  If you never use non-functional pads, don't  worry about their use 
or just don't care, one of the following is probably true.

1)  All you manufacture is Polyimide!
2)  Your in-coming phones are out of order!