Strange third party board in PDP-11/45

Noel Chiappa jnc at mercury.lcs.mit.edu
Tue Jul 24 15:38:56 CDT 2018


    > From: Mattis Lind 

    > Here is how it is connected:

Thanks for that - very informative!

    > The sandwiched dual boards are sitting in 27 / 26 AB. The board in 27AB
    > was empty (quick glance), while the board in 26AB has a few TTL chips
    > on it. Slot 26AB is the Unibus A slot, Slot 27 AB should be a
    > terminator on Unibus B.

I'm more interested in _what_ the two boards are doing! :-)

It seems they must be jumpering UNIBUS A and UNIBUS B together. (Which I
didn't expect, but maybe... will have to ponder.)

As to what _else_ it is doing, and why it has the cable to the main card... I
think that it must intercept MSYN from the processor and only let it pass if
there's no hit in the cache.

(To explain why it would need to do that... normally with the MS11, there's a
static partitioning between FastBus memory and UNIBUS A memory. So when the
CPU goes to do a memory cycle, it can put the address out on both the UNIBUS
and FastBus, with the certainty that it will only get a reply on one. But with
the cache, if there's a hit, it would in theory get a reply on both, which
might confuse it. Or if it takes the cache copy, and terminates the UNIBUS
cycle, that might confuse the memory.)

Or maybe I'm confused, because now that I think about it, UNIBUS A goes
straight from the CPU to the UNIBUS A out slot, so the Able board couldn't
intercept MSYN? I guess I need to understand the fine details of the UNIBUS A
and B stuff, maybe it will make sense at that point.

Oh, wait a moment: slot 26 is UNIBUS A out, slot 27 is 'UNIBUS B in', and slot
28 is UNIBUS B 'termination'. (27 is 'in' because when the M9200 is installed
in 26/27 to join the two UNIBI together, obviously one has to connect an 'out'
to an 'in'... and then 28 is not 'UNIBUS B termination', it's 'UNIBUS out' to
the rest of the system.

OK, so that works - MSYN coming out of slot 26 is intercepted by the dual
double-card, and is only allowed to pass on cache miss. Yeah, that sounds like
it should work.

    > The hex ABLE/ ACT board sits in slot 21 which is the memory controller
    > board for the MS11.

One of two; the other is slot 16.


    > From: Paul Birkel

    > I wonder whether this CACHE/45 can coexist with MS11 memory on the
    > Fastbus itself

According to that marketing thing you found, "User may optimize hit ratio by
upper/lower limit switch settings", so one would have to configure the
Cache/45 to not cache the block that the 'other' MS11 controller thinks it
owns... otherwise both might respond to requests for addresses in that
range.... :-)

	Noel


More information about the cctalk mailing list