At one time Agilent (then HP) had a high density interconnect method for probing with their logic analyzers that required via-in-pad.
As I recall, the relevant application note recommended applying solder paste and reflowing the board twice.
On the first pass one applied paste only to the pads that incorporated vias. The amount of solder paste was sufficient to fill each via. Then one reflowed the bare board, with no components. After that, using a different stencil, one applied normal amounts of solder paste for the components, placed the components, and reflowed again.
I never saw discussion of this by third parties, nor do I know what kind of yields one might experience using this approach. Anyone?
Philip Decker
-----Original Message-----
From: Dimiter Popoff [mailto:tgi_earth@yahoo.com]
Sent: Friday, 24 November, 2000 9:25 AM
To: pci-sig@znyx.com
Subject: Re: BGA Assembly
Peter,
>I was specifically recommended not to put vias or holes under the ball, by
>someone with past experience of solder being drawn through the hole and away
>from the BGA. I guess this isn't what you wanted to hear - sorry.
I guess you just saved a board from an unsuccessful experiment :-).
Overall, though, this is no problem, because I don't expect the
device to go into automatic production quantities soon, and manually
I seem to manage it quite successfully. Moreover, from what I have
seen so far the solder drawn through the holes is not much (it
never makes a drop beneath the board, just fills the hole), unless
there is a big pad at the opposite side to be wetted; and even then
there is no problem, the connections are quite intact. Actually I
have hard times imagining what should happen to make the solder
be drawn away so there is no more a connection between the BGA
and the board (I guess we all know how do solder drops of this
size behave), but as I said earlier, I'll take easily avoidable
risks at a later stage :-) (trying if the BGA will float on
this without limiting feet, that is).
All this is valid for holes drilled with 0.3mm or less; this makes
about 0.2mm after plating. The holes had anyway to be that tiny
in order to allow 3 traces between each pair of BGA pads; which, in
turn, made routing possible on only two signal layers.
Thanks,
Dimiter