7 Feb
2007
7 Feb
'07
3:41 p.m.
Chuck Guzis wrote:
On 7 Feb 2007 at 11:08, jim wrote:
any other observations on WW as far as clocking? I
know that
the more expensive stuff ran mini coax around the backplane
to get much better signals, but we just used regular old 32 or
whatever guage WW wires, sometimes 26 guage for high
current.
How about ECL? Moto used to advertise that the 10,000 series was
"wire-wrappable", but the WW samples I've seen have lots of twisted
pairs on them. Looks like a real pain to do.
Cheers,
Chuck
----------------------------------------------
The Cyber 170 series from CDC used twisted pair wire-wrap. I still have a
bunch of it. The wires were color coded for length. The early back planes
were hand wired using a machine like Al described. Later machines were more
automated and wired from a reel. Proper nightmare to troubleshoot. You had
to feel a wire through the mat since all wires were the same color.
The logic was standard MECL 10K. The normal and complement outputs made it
a natural to use as a differential transmitter. So some of the signals were
single-ended, twisted with a ground wire. The longer signals and I/O lines
used the twisted pair as a differential line - no ground wire. If I
remember correctly, the standard single ended wire was 100 ohm impedance
terminated to 220 ohm resistors to -2.2v or 550 ohm resistors to -5.5v. The
differential lines used two resistors at each end.
Real pain doesn't do justice to working with these back planes. The wires
were 30 gauge single strand and brittle. When removing one, you had to
carefully trim off the bare wire end before pulling the wire out of the mat.
If you didn't, the ends would break off and fall down the vertical back
plane until they disappeared. Then you got to spend an hour trying to find
them. We used to call them "tingles".
If you didn't get them all out and applied power, they became "twinkles" as
they shorted out and vaporized. Which did wonders for the logic and gave
you a couple more hours of overtime trying to get the machine going again.
Of course this always happened late on a Saturday night, especially if you
had a date or tickets.
Putting ECOs in became a total complete nightmare. We calculated the size
of work by number of wires to change. If you were good, you could do 8-10
wires an hour.
4 years of twisted pair wire wrap experience is probably the best
explanation I can give for volunteering to move off main frames and on to
disk drives.
Billy
7 Feb
7 Feb
5:58 p.m.
Billy Pettit wrote:
Chuck Guzis wrote:
On 7 Feb 2007 at 11:08, jim wrote:
How about ECL? Moto used to advertise that the 10,000 series was
"wire-wrappable", but the WW samples I've seen have lots of twisted
pairs on them. Looks like a real pain to do.
Cheers,
Chuck
I have seen the ECL stuff too, but there was both twisted pair as billy
mentions below
as well as some with coax.
----------------------------------------------
The Cyber 170 series from CDC used twisted pair wire-wrap. I still have a
bunch of it. The wires were color coded for length. The early back planes
were hand wired using a machine like Al described. Later machines were more
automated and wired from a reel.
We never had any reliable wire cutter or trimmer.
cutting caused your tingles right from the gitgo, and
even the NoNics I gave you didn't reliably strip the
wire of insulation. We had a self stripping bit that
we used, but it required a specific wire, specific
insulation, and size for every bit, and cost several
hundred 1980's dollars, and was very exotically
machined.
but you could just insert the unstripped wire, and
it stripped and spit out the insulation.
but as it wore, it developed a tendency to break
the wire or weaken it right at the bottom of
the wrap, or didn't stip, so we went back to
pre stripped.
Proper nightmare to troubleshoot. You had
to feel a wire through the mat since all wires were the same color.
<good description of ECL snipped>
If you didn't get them all out and applied power,
they became "twinkles" as
they shorted out and vaporized. Which did wonders for the logic and gave
you a couple more hours of overtime trying to get the machine going again.
Of course this always happened late on a Saturday night, especially if you
had a date or tickets.
the other issue you had to confront was how many levels of wrap you
allowed per pin. if you had a good routing program, you could do 2
levels everywhere there weren't ECO's, or by hand running a net.
However if you put in an eco sometimes engineering practices
would allow you to add a level if there were room on the pin,
and later when you accessed it for another change, the wrap
you wanted would be the middle wrap, and you'd have to
redo two wires.
Putting ECOs in became a total complete nightmare. We
calculated the size
of work by number of wires to change. If you were good, you could do 8-10
wires an hour.
4 years of twisted pair wire wrap experience is probably the best
explanation I can give for volunteering to move off main frames and on to
disk drives.
Again we were lucky to get off w/o twisted pair for the most part.
Actually the only place we had to pay attention to it was on
the SMD and Trident Radial cable boards where the
RF came in and out, and make sure we got clean
clocks and stripped data streams to the CRC chip
we used.
The lack of a good ECC chip in TTL was one reason
we never supported ESMD since the bit rates were
too fast for whatever CRC we used (I'll have to go
look it up sometime) and it was too much redesign
to use any alternative.
We used our own format which was compatable with
only our controller, which is one of the reasons I
don't get too excited about recovering any similar
bit oriented direct device's data. I guess if you
had a reliable way to read all the bits, you could
decipher it, but without a compatable or capable
disk controller, you won't recover the data from
a disk pack or drive.
Sorry to digress off the wire wrap subject here.
Billy
6572
days inactive
6573
days old
1 comments
2 participants
participants (2)
-
Billy.Pettit＠wdc.com -
jwstephens＠msm.umr.edu