On 6/7/06, Brent Hilpert <hilpert at cs.ubc.ca> wrote:
Thanks for all the replies, Multiwire was indeed the
name I had forgotten.
Rather easy to forget... it's not a very unique name, there being lots of things
that could fittingly take on that name.
The COMBOARD for VAXBI was a Multiwire board, "Unilayer", technically,
but the same technology - they took our gerber map and created two
custom "decals" of the wire nest, then bonded them down to a 10-layer!
0.1"-center perfboard with only the VAXBI corner done in copper traces
(per DEC's VAXBI spec). The advantage was that the VAXBI corner
layout was exactly to DEC's spec, and the rest of the board was to our
spec, without the tooling costs of setting up and machining a 10-layer
PCB. We worked out the math of the costs of Unilayer and full PCB
production, and the setup-costs for 10-layer were so high that even
though Unilayer was somewhat expensive, we saved money on the initial
order of boards.
Another cost savings was ECOs in a rev 0/rev A board... it can be
tough to make a fix where the problem is on an inner layer (bad enough
with 6-layer). We did end up making one cut and one jump on the
Unilayer board, but were confident that a later board would have had
no wires.
Had there been enough demand for a second order, we had contemplated
going the full PCB route, but the by the early 1990s (when we finally
got our license to manufacture VAXBI boards), the market had moved on.
I still have a VAXBI COMBOARD in the 8300 in my basement - it runs
HASP and 3780 (we never had enough customer demand to port SNA to it),
and if I can think of something else to do with it, I might someday
redo the firmware and device driver. It is, in essence, a
really-smart dual serial port (10Mhz 68010, 512K to 2MB of local RAM,
1 x Z8530 SIO), and the PCB is rather full, so even trying to hack it
into, say, a SCSI card, would be difficult.
Still, it was fascinating to open the box and see a large board coated
with nests of fine wires... we _did_ inspect every single joint. We
found one failure on one board, but not the bonding of the wires to
the board - one pin insert was missing, leaving an IC pin hanging in
free space. Stole one from the edge of the board from an unoccupied
position and the board worked perfectly after that.
-ethan