29 Apr
2015
29 Apr
'15
4:24 p.m.
On Wed, Apr 29, 2015 at 6:36 PM, Noel Chiappa <jnc at mercury.lcs.mit.edu> wrote:
*Mostly* on DEC equipment, reversing non-keyed cables is safe. Mostly.
From: Josh
Dersch
it never occurred to me to just try flipping it around since usually
that's a pretty good way to let the magic smoke out of things.
Exactly. Except for cases where I _know_ the connector was designed to be
able to withstand being plugged in backwards (e.g. IDE), I would _never_ try
reversing a cable 'on spec'. The chances are just too big one will kill
something. (Amusing factoid: QBUS cards have the same thing, they
are designed so that
you can plug them in backwards, and not kill things. I am somewhat ashamed to
admit that one one occasion, I actually did this! Luckily the engineer who
did the QBUS saved my behind.)
You can plug an RL02 interface cable backwards into an RL-11 and
nothing bad happens, and many, many other such examples exist, to the
quantity that back in the day, customers assumed nothing would go
wrong if it was reversed, and, again, in nearly every case, they were
correct - it wouldn't work, but it shouldn't smoke, either. In that
environment, when we were moving from making Qbus and Unibus products
to add a BI product, our design engineer had a dilemma - formerly,
there was enough room on our products to put the core functionality
and the traffic lights and EIA line drivers and 40-pin plugs right on
the edge of the board. No problems. The BI is quite different in
that respect (amount of square inches, no edge connectors allowed,
etc). His solution was for us to have a I/O plate board with two
30-pin cables that plugged into the non-BI segments of the backplane
and give the user serial indicators and do EIA level shifting off the
main board. I understand the limitations we were facing and why he
did that, but the part that got me was I told him to ensure the cables
could be reversed safely. He scoffed and said that nobody would ever
get the keyed cables upside down and backwards...
So we make the product, we ship the product, and... 100% of our
customer installs failed because each and every customer and/or 3d
party installer did just that. The swap ran +12V where it shouldn't
go and took out the low-voltage side of the receivers on the I/O
bulkhead board.
After the engineer left, I got to support that product. I ended up
socketing each and every adapter board so we could quickly repair the
damage, and for the remaining installs, between extra labels on the
cables, and several dire warnings in the docs, *and* numerous verbal
warnings on installation support calls, we got the rate down to 60%
damage.
At least the damage was less than $5 in parts, and the damage to our
reputation was minimized by the scant sales of that product because by
1990, who was buying new BI machines anyway?
I still have several of those boards - MC68010, 512K to 2MB of DRAM,
64KB of ROM, Z8530 SIO working in async or sync mode... Not sure what
one could do with that besides a comms adapter, but as long as you
don't plug in the EIA driver cable in backwards, it's a nice product.
-ethan