Don't quit on that one so fast, i learned it right here from other
'mericans in the latter half of the 20th century.
I have another idea. Instead of a quad pad array, place them in an
"Indian tee-pee" <sp> configuration. You stand them up, and solder a big
blob node on the end of all four. Make a jig so they are all exactly the
same.
What you get is more real estate, so a bigger array fits into the same
areal space. It also has the benefit of making those nodes a compact,
singular alloy, with low intermetallics issues, and the devices are
closer together,(the corners would touch) so it would be closer to the
ideal. It also allows easier probing on the blob. You could make two
"sheets" of interconnected tee-pees and then flip one completed "sheet"
over to make another set of connections from sheet to sheet.
Darn, if I had my old CadKey loaded, I could draw it up real quick. I
already have macro drawings of an SMD resistor and mlcc SMD cap.
I have cap arrays potted in polyurethane and multiplier banks, etc.
One is worth like $2k, and that was ten years ago. It would be $5K now.
I made it puke one day (CadKey) by trying to array the bottom layer of
9.5 ton stones that make up the Cheops Pyramid. I wonder if my current
machines could manage such a huge array. The part that makes it hard is
the 5 mil space between stones that span several feet on each side. The
package doesn't like having to manage two sets of scale precision
together, OR it simply gobbled up all the memory it had and simply
expired. Still, I have yet to sit down and get that drawing project
started again on a modern package. Not an outline, like most today, but
a full, all stones included rendering. Then, a duplicate rendering that
includes the missing facade stones and capstone.