In 1981 I was invited to watch a Burroughs technician apply a patch to
the university's B6700 mainframe. The machine was entirely wire-wrapped.
He came in and pulled out the affected board and opened a sealed package
that included some drawings and text, two pieces of wire precut to the
correct length and with the ends stripped plus a manual wire-wrap tool.
I was surprised to see the latter for two reasons: we used powered
wire-wrap tools on our own boards (based on the 6800 processor and which
interfaced to the B6700 by emulating a terminal) and the idea that each
patch should pack a separate tool seemed silly since this isn't exactly
medical supply or something.
Having done a lot of wire-wrapping with both powered tools and manually,
I can do it either way without much trouble. Powered tools are faster
and more consistent, but most of the improved results I got with them
were due to using a much better wire stripper than the one built into
the manual tool.
In 1986 I had finished a 68000 computer done entirely in wire-wrap and
the number one problem had been finding a low cost board that was large
enough. The solution was to buy two boards and mount them in a wooden
picture frame, as seen on the right side of these two pictures:
http://www.smalltalk.org.br/fotos/inova6.jpg
http://www.smalltalk.org.br/fotos/inova7.jpg
Since the 68000 would take up a significant board area, I arranged it so
each row of pins was on a separate board and below the chip was the gap
between the boards. The boards had wide copper tracks which would make a
nice power distribution system, but I ended up having to do the power in
wire-wrap too and that turned out to be a significant part of the
construction effort.
One month before the big computer show in Brazil, my partner decided
that this black and white Mac-like machine was too pathetic and asked me
to design a more powerful version with color. This is the machine on the
left in those pictures. We didn't have time for a normal wire-wrap
prototype and my partner wanted something that looked closer to a
finished product. We had just bought a PCB CAD program (I think it was
called SmartWorks) but a quick test showed that though it helped with
regular designs like the memory banks, the random parts of the design
would take more time than wire-wrap.
So my friend, Ruy Cesar Pinto Neto, came up with the idea of doing the
memory as a normal PCB with the CAD (you had to print it twice with the
dot matrix to get it dark enough at four times the normal size and then
photographically reduce it to get the PCB masks) and to do the random
part as a single sided PCB with just the power distribution and normal
IC sockets (costing a fraction of what the equivalent wire-wrap sockets
cost) with each pin connected with a very short track to a wire-wrap
post on the component side. It was amazing how much faster it was to
wire-wrap without constantly flipping over the board. The wires had to
be a little longer since they had to snake around the components instead
of being point to point but that didn't prove to be a problem. We
actually were able to build three who prototypes (of which two were
working) by the start of the computer show. With the two boards
connected to each other, the wire-wrap was completely hidden and the
result looked like something that was reasonably close to shipping.
http://www.smalltalk.org.br/fotos/inova3.jpg
Though the two projects I did after this one used traditional wire-wrap,
my experience with this hybrid technique was very positive and I
recommend it. My big mistake was not designing in some spare 14 and 16
pins sockets - you can see a 14 pin wire-wrap socket glued on its side
on the top right of the bottom board in the second picture as the result
of this.
-- Jecel