On Sat, Mar 29, 2025 at 1:41 PM shadoooo via cctalk
<cctalk(a)classiccmp.org> wrote:
I read the "UnibusSpec1979.pdf" on
bitsavers, which reports a "Standard
Unibus" pinout in the last pages.
However in several backplanes "Small Peripheral Controller", "Modified
Unibus Device" and "Extended Unibus" are supported.
Maybe also other unlisted Unibus variants do exist (e.g VAX 11/730)?
Yes. The 11/730 is a little different because below the CPU boards,
the first 5 slots can take 1MB memory cards or peripheral cards (each
of those slots has a single memory select line that goes back to the
memory controller, so a) the cards must be 1MB each, and b) there are
only 5 select lines coming out of the PAL.
Our original COMBOARD wasn't 100% conformant to the Unibus spec - it
works in DD11 backplanes just fine but does _not_ work in an 11/730
without dozens of cuts and bodge wires.
Big doubts:
- why DEC, having defined the dual Standard pinout, had then to
implement the quad SPC backplanes?
- why DEC, having defined quad backplanes, had then to implement the hex
(standard + SPC) or (MUD + SPC) or EUB?
I mean: given that in AB all Unibus signals are present (from
specifications), what is the need for CDEF?
Provided that several signals are duplicated in hex pinout, the
backplane will connect homologue signals together,
or AB bus will always be separated from CDEF bus?
The "why" for a good portion of that was how the Unibus evolved from
1970. In the 11/20, A/B slots were the true Unibus signals, meant for
BC11 cables and M930 terminators. Many (most?) original peripherals
were implemented not as single cards, but as system units (a block of
4 or 9 slots with Unibus in and out on first and last A/B but with
unspecified wiring in between) including the console serial port
before the SPC M7800 DL11.
For system unit peripherals, there was frequently a common scheme with
an M105 and other cards to implement the Unibus parts of the
peripheral that resulted in a typical wiring arrangement of certain
signals on CDEF that were just wired over to Unibus signals from A/B.
With all the needed signals on CDEF and with Unibus in/out and power
connectors in some B and even sometimes a KM11 diagnostic board, that
left room for quad SPCs in any open slot.
Starting with the 11/05, which had to fit in a single backplane, the
amount of integration rose to where hex peripherals (and memory) were
possible.
As mentioned, EUB slots exist to support 22-bit memory addressing for
Unibus memory (11/24).
short answer - the Unibus evolved as integration density increased.
-ethan