11 Feb
2003
11 Feb
'03
3:10 a.m.
--- "Jeffrey S. Worley" <Technoid(a)30below.com> wrote:
No one knew just what they or other folks were going
to hang in these
backplanes, or what functions those cards would perform. If they had
known, it would have been pretty simple to provide a set of jumpers or
some other (easier) means of moving signals around, but they didn't so
they went for total flexibility in this department and so required
jumper wiring.
I think you need to go back in time a bit and consider the original
Unibus processors, like the 11/20. It doesn't have single-card
periperals, it has peripherals with their own custom-wired backplanes.
The Unibus-in and Unibus-out slots were standardized, but in between,
it was up to the designer. There's lots of consistency in DEC
peripherals - they tended to use M105s and other "standard" Unibus
modules, but that wasn't an absolute requirement.
In those days, to initiate an NPR cycle (i.e., start a DMA operation),
the peripheral broke the connection of the NPR signal between the
Unibus-in and Unibus-out slots. When boards began to get more dense
and it was finally possible to put an entire peripheral on a single
card, these peripherals couldn't operate the same way. The new method
was to break the signal between a pair of pins on the same slot.
It wasn't possible to ship all new backplanes (like the DD11DK) without
the NPR jumper removed because there were some early non-DMA periperals
(like the line-printer interface) that may or may not jumper that
signal on their own.
So... going back at least as far as 1981 (the earliest Unibus COMBOARDs),
makers of NPR devices had to provide instructions for the customer
on how to locate, remove and replace the NPR jumper wire. We also
shipped a wider grant card (the "Grantasaurus Rex" - GC747) so that our
customers didn't have to rewire the backplane when our board was not
installed.
I was told that the original scheme existed because nobody planned for
a DMA peripheral that occupied only a single slot - too much hardware
to fit in such a small space. It did take a number of years, so
the expectation wasn't that unreasonable.
-ethan
Some card slots don't carry certain signals over to the next slot down
the line if there is no card installed in that slot.
For example. If you have an hdd controller in slot A, a blank slot B,
and a serial terminal controller in slot C, my machine wouldn't work at
all. In this scenario, I would have to bridge the blank slot B by
jumping two signals (IRQ and XSTL?) from A to C in order for those two
critical signals to propagate past the blank slot.
You might wonder why they would omit two of the most important lines
from the backplane like that. Well, it is flexibility in mind here.
With this method, you could have card's B and C connected via ribbon
cable at their leading edges to carry the signal from B to C for use and
then have card C propagate the signals over the backplane after they've
served their purpose. You could even bring those signals outside the
machine entirely to some specialized equipment before returning them to
the backplane via a card. In this way, you could largely replace the
backplane with your own custom wire job.
Regards,
Jeff
-----Original Message-----
From: cctalk-admin(a)classiccmp.org [mailto:cctalk-admin@classiccmp.org]
On Behalf Of Jay West
Sent: Tuesday, February 11, 2003 10:04 AM
To: cctech(a)classiccmp.org
Subject: More 11/44 basic newbie help request
Eric wrote...
Get the manuals from Al's site.
You know... I think I've just been told "RTFM" *grin*
S'ok, was well deserved :)
After perusing the manuals for the 11/44 and the RL02's, and the RL11
(looking for docs on the 8 port mux now).... Most of my questions are
answered there (big suprise). Not trying to start any holy wars or
anything,
but to my "newbie with unibus" mind, the design of having to change
backplane wiring for certain cards seems to be rather... ummmm silly
(euphamism{tm}), at least as compared to other systems of the period.
Probably there is something I'm missing as to why this is a good thing.
Jay West