25 Oct
2013
25 Oct
'13
4:23 a.m.
azd30 wrote:
Hi,
I recently aquired an old Netcom system and I'd like to find out a bit more info about
this backplane.
It's an 11/23 system (based on CPU) . It has some standard DEC
cards in it (M8186 CPU, DLV11, BDV11 bootstrap board ) as well as a thrid party 256Kb
memory board and an MXV22 floppy controller.
Apparently the system ran RT-11 from floppies a few years back before it was put in
storage.
If you can specify all the boards (or at least the DEC boards) by their
board number (i.e. M8186), it would be more useful. Is the DLV11
an M8043 (actually DLV11-J)? Just in Case, please also mention
dual or quad board.
I have read through the PUPS archives, esp. a post from
Michael Sokolov back in 98, where he describes
the different types of QBus'es. Q/Q, Q/CD etc. Have also read through the info on
hampage.hu and various other sites that host PDP11 related info,
but am still unsure about this particular beast.
It looks like the backplane that I have (according to the archives and some documents
written at SLAC in late 70's)
is a serpentine or sinusoidal type configuration. The SLAC documents actually refer to the
Netcom backplane but they
describe it a simple serpentine type.
A diagram on my cardcage describes it as follows:
A B C D
------------>
<-----------
------------->
<------------
There is also a blurb about slot 2/CD being wired differently. Two slots on the diagram
are pre-printed for RL controllers. I can't rembeber the exact wording, since I
don't have the computer in front of me.
This is possible, but seems VERY unlikely with just
a 4 slot backplane. As far as I understand, in order
to be useful to support an RL controller, at least 2 slots
must be wired as ABCD. The BA23 box has the first
three slots as ABCD and the other five slots as ABAB,
so the concept is possible. Having two ABCD slots also
supports PMI memory with the memory being placed in
slot 1 and the CPU in slot 2, but that requires at least
a quad M8190 CPU card.
It gather that I insert the CPU in 1AB, then rather
than putting the next card in 2AB, I would place it in 1CD. Next would be 2CD and so on
,snaking down the line. Is this correct? While this is all fine for dual cards, but how
does this configuration
handle quad cards? I'm guessing that it has to do something with that 'blurb',
but that would mean that I can only have one quad card installed, and it would be right
towards the begining of the bus. The BDV11 should
be the last card in the chain from what I can remember.
I forgot to add, please also specify where all the current
cards are located in the back plane.
With an M8186 to start, the BDAL17-20 lines should be
either unconnected (for an 18 bit backplane) or ONLY
connected to each other (for a 22 bit backplane). If they
are unconnected, installing connections should convert to
a 22 bit backplane and you M8186 should be able to
access as much memory as you are able to place in the
backplane (up to 4 MB of course). At one point, a VT103
backplane was upgraded to 22 bits and a 4 MB memory
board installed.
Also, my CPU card is a rev. D so it can do 22bit
addressing. I'd like (if possible) to run 22bits, since this would allow
me at a later time to put in a 11/73 cpu that would give me more flexibility in what
OS's i can install on the machine.
I'll have to check if the backplane is wired for 22bits or not. I seem to remeber the
standard qbuses had the W1-W4 pins (?)
that you could wirewrap to change from 18to22 (i'm assuming that this operation
propagated the BDAL17-20 from slot 1 to the other slots) but this backplane has nothing
like that.
I'm planning on checking it with an ohmmeter, and it should be very obvious if these
signals are present across slots.
The backplane is a PCB style, but it does have std. wire wrap pins sticking out from the
PCB. If the BDAL17-20's are not present across slots, can I just simply extend them
thru by wirewrapping?
Probably. For the VT103, there were no pins, just the
solder bumps. That was sufficient to attach the required
wires. Note that all of the ABs must be connected, not
just the ABs on the left side of the backplane. For the
VT103 with a 4 slot ABAB backplane, that meant eight
connections per each of the four wires to connect the
BDAL17-20 together (each to themselves of course).
If anyone can shed any light on this backplane, it
would be greatly appreciated.
thanks in advance for any replies.
--
alex
If you have any more questions and you answer the
questions, it may be possible to help some more.
Jerome Fine
25 Oct
25 Oct
7:02 a.m.
On 10/25/2013 05:23 AM, Jerome H. Fine wrote:
This is possible, but seems VERY unlikely with just
a 4 slot backplane. As far as I understand, in order
to be useful to support an RL controller, at least 2 slots
must be wired as ABCD.
I've got an original pdp11/03 in a four-slot backplane,
_all_ the slots
are Q18/Q18 serpentine. The RLV12 (quad width) lives in slot 3, the
system works great.
The BA23 box has the first
three slots as ABCD and the other five slots as ABAB,
so the concept is possible. Having two ABCD slots also
supports PMI memory with the memory being placed in
slot 1 and the CPU in slot 2, but that requires at least
a quad M8190 CPU card.
> It gather that I insert the CPU in 1AB, then rather than putting the
> next card in 2AB, I would place it in 1CD. Next would be 2CD and so
> on ,snaking down the line. Is this correct?
Assuming the bus is in fact QQ/QQ
serpentine, yes.
> While this is all fine for dual cards, but how
does this configuration
> handle quad cards?
As long as quad-width cards don't mess about on the
right-hand side
connectors (most of them don't) you can use any quad-width slot, whether
its wired QQ/CD or QQ/QQ.
With an M8186 to start, the BDAL17-20 lines should be
either unconnected (for an 18 bit backplane) or ONLY
connected to each other (for a 22 bit backplane).
Just a clarification, the
high-order four address lines are BDAL18-21...
10:27 a.m.
Ok. Apologies for not being thorough with my description.
The backplane is an 8x4.
The diagram on the card cage is as follows:
Remeber that this is a flip cardcage on hinges, so in fact when put back in the box
everything ends up as it should.
Modules are inserted upside down, right to left, bottom to top...
-------------------------------------------------------------------------
MODULE VIEW
D C B A Priority Chain
-> 8 ---- ---- ---- ---- ---> Lowest
| - - 7 ---- ---- ---- ---- <-
-> 6 ---- ---- ---- ---- --|
| - - 5 ---- ---- ---- ---- <-
-> 4 ---- ---- ---- ---- --|
| - - 3 ---- ---- ---- ---- <-
-> 2 ---- ---- ---- ---- --|
| - - 1 ---- ---- ---- ---- <--- Highest
THE PRIORITY CHAIN SKIPS SLOTS 2CD and
3CD, WHICH ARE NOT NORMAL LSI-11 BUS
-------------------------------------------------------------------------
Ok, now 1A and 1B are preprinted CPU. 2D marked RL01 2C marked M8013. 3D marked RL01, 3C
marked M8014..
CPU goes in higest slot 1AB. As it should.
Slots 2 and 3 are oviously different. Since they were meant for the 2 quad board RL01
controller RLV11 that makes sense.
On to the next question:
This were the boards in the system and where they were installed. I am assuming that's
how the previous owner had them installed
when he had the system running with RT11. I just don't see how it could have worked.
1AB - M8186 11/23 CPU (don't know if it's a YA B or C, but definitely no FP
going from memory two chips on it, CPU and MMU?, a rev D board)
1CD - Third party 256Kb (with 128K installed) parity memory board.
2CD - M7940 DLV11 (don't know if it's a YA C D or E)
2AB - MicroTechnology MXV22 RX02 compatible disk controller
3ABCD - M8012-YA BDV11
All cards duals except for for the M8012 which is a quad.
I can sort of see how the M8012 would work in slot 3 (it does match the ABCD in the chain,
and it is a slot that is different),
but if slot 2 is also different, the M7940 could not have worked. It's an AB card,
stuck in a CD slot.
I still have to figure out the 18vs22 bit issue. And correct me if I'm wrong, but
unless I have an even number of dual cards, a quad card will
not work in the system (leaves a 1/2 slot empty before it).
If I was to go with an RLV12 controller that's a single quad board in Slot2, I
don't think that I can leave slot 3 empty. Now, if it was a 22 bit
backplane I could stick TWO RLV12's in there, and that would make it all ok.
Speaking of... Anyone have a spare RLV11 set or an RLV12 they would be willing to part
with?
And thanks for the correction on the extended address lines...
--
alex
26 Oct
26 Oct
2:06 p.m.
Hi there:
On 25 October 2013 13:27, azd30 <azd30 at telus.net> wrote:
Ok, now 1A and 1B are preprinted CPU. 2D marked RL01
2C marked M8013. 3D marked RL01, 3C marked M8014..
1AB is of course a QBUS slot, 1CD, I'm going to guess was also a QBUS
slot. The wiring should then "skip" to 2AB as the next QBUS slot, then
to 3AB, then 4AB, and then start the normal serpentine fashion (4AB to
4CD, to 5CD, to 5AB...). With slots 2CD and 3CD connected to each
other (and ONLY each other) as CD interconnect.
Alternatively, 1CD may have never been anything in the original system
(see below for what I mean by that).
CPU goes in higest slot 1AB. As it should.
Slots 2 and 3 are oviously different. Since they were meant for the 2 quad board RL01
controller RLV11 that makes sense.
Yes, the CPU slot makes sense, though it could also be a Q18/CD or
Q22/CD and most CPUs wouldn't give a damn. The CD slots in this case
being reserved for the two board RLV11 set as you said.
On to the next question:
This were the boards in the system and where they were installed. I am assuming
that's how the previous owner had them installed
when he had the system running with RT11. I just don't see how it could have worked.
1AB - M8186 11/23 CPU (don't know if it's a YA B or C, but definitely no FP
going from memory two chips on it, CPU and MMU?, a rev D board)
1CD - Third party 256Kb (with 128K installed) parity memory board.
2CD - M7940 DLV11 (don't know if it's a YA C D or E)
2AB - MicroTechnology MXV22 RX02 compatible disk controller
3ABCD - M8012-YA BDV11
And this makes it sound like someone rewired the everything. Since --
quite rightly -- the M7940 should not have worked in a pure CD slot.
(It should have done absolutely nothing.)
Alternatively, if it was not rewired (you mentioned the backplane
being of PCB style design, so that sounds quite impossible, save the
update from 18-bit to 22-bit, which isn't a problem). Then the
previous owner must have misremembered it running RT-11. (Or he spent
a lot of time cutting PCB traces and rewiring the backplane).
All cards duals except for for the M8012 which is a
quad.
I can sort of see how the M8012 would work in slot 3 (it does match the ABCD in the
chain, and it is a slot that is different),
but if slot 2 is also different, the M7940 could not have worked. It's an AB card,
stuck in a CD slot.
The M8012-YA (a BDV11) is an 18-bit bus terminator and
bootstrap+diagnostic ROM card. It goes at the end of your card chain
to, as it says in its name, terminate the bus. The -YA is also 18-bit
*ONLY* so, if you do go 22-bit, you'll need to remove it.
I still have to figure out the 18vs22 bit issue. And
correct me if I'm wrong, but unless I have an even number of dual cards, a quad card
will
not work in the system (leaves a 1/2 slot empty before it).
22-bit uses some of the "unused" pins in 18-bit land; the only real
issue with the 18-bit/22-bit debate is that some cards will not work
in one or the other (see the aforementioned comment on the BDV11-YA).
If I was to go with an RLV12 controller that's a
single quad board in Slot2, I don't think that I can leave slot 3 empty. Now, if it
was a 22 bit
backplane I could stick TWO RLV12's in there, and that would make it all ok.
You don't *need* two RLV12s. As I personally haven't experienced any
PDP-11 software that supports multiple RL01/RL02 controllers. Plus, 4
RL01s/RL02s is big physically, and any of the operating systems that
want loads and loads of disk space probably won't fit on an RL01 or
RL02 pack as a system disk.
Cheers,
Christian
--
Christian M. Gauger-Cosgrove
STCKON08DS0
25 Oct
25 Oct
10:51 a.m.
New subject: Serpentine QQ/QQ configuration and quad cards
> <Alex D wrote:>
> While this is all fine for dual cards, but how does this configuration
> handle quad cards?
<Roe Peterson wrote:>
As long as quad-width cards don't mess about on the
right-hand side
connectors (most of them don't) you can use any quad-width slot, whether
its wired QQ/CD or QQ/QQ.
I'm confused here. All the required signals for system operation are on the AB side
(you call it QQ?). Correct? What is on the CD side then?
I always thought that the CD side was like a 'private' communication bus, as in
the M8013 talks to the M8014. I'm not getting something here.
How would a quad card not mess about on the right hand side? The only thing I can think
of, is this. If the right hand side (CD) is wired for AB there is power
on there. The quad card would just have to stay away from those pins. Does the CD part of
the quad card make use of the AB signals present?
I'm going by this:
http://www.ibiblio.org/pub/academic/computer-science/history/pdp-11/hardwar…
and my 'microcomputers and memories (1982)' handbook.
I'm feeling really 'dumb' right now...
cheers
--
alex
26 Oct
26 Oct
1:37 p.m.
New subject: Serpentine QQ/QQ configuration and quad cards
On 25 October 2013 13:51, azd30 <azd30 at telus.net> wrote:
QQ/QQ and QQ/CD can be rewritten as AB/AB and AB/CD; or as I like to
use Q18/Q18, Q18/CD, Q22/Q22, and Q22/CD. The normal QBUS signals are
on the AB side, correct. CD has power in the same positions as on the
AB side, as well as the CD interconnect "bus." In the link to ibiblio
you have in your e-mail, the CD interconnect can best be visualized as
"all the 2 pins in one slot, go to all the one pins in the next slot."
With slot 1's 1 pins ignored, and slot n (where n is the last slot)
has all its 2 pins ignored. So yes, in effect the CD interconnect is a
private communications bus, between each separate half of the a slot,
and the half slot above and below it (so slot 1 has a private link to
slot 2, slot 2 to slot 3, 3 to 4, et cetera).
The CD interconnect is also used on the PMI capable QBUS cards, so the
11/83 and 11/93 processors (and third party replacements for those),
the PMI memories (which go *BEFORE* the processor), and the
CannotRememberItsName11 QBUS-to-UNIBUS bridge used to make the 83 and
93 into an 84 and 94. CD is also used by the QBUS VAX cards (uVAX II
for examples) in a similar private memory bus design.
As long as
quad-width cards don't mess about on the right-hand side
connectors (most of them don't) you can use any quad-width slot, whether
its wired QQ/CD or QQ/QQ.
I'm confused here. All the required signals for system operation are on the AB side
(you call it QQ?). Correct? What is on the CD side then?
I always thought that the CD side was like a 'private' communication bus, as in
the M8013 talks to the M8014. I'm not getting something here.
How would a quad card not mess about on the right hand
side? The only thing I can think of, is this. If the right hand side (CD) is wired for AB
there is power
on there. The quad card would just have to stay away from those pins. Does the CD part of
the quad card make use of the AB signals present?
By not mess around, one refers to not doing things on the CD
interconnect pins, and pulling the card into a serpentine backplane
(as that would result in card, and possibly other cards releasing the
magic smoke). I.e. the KDJ11-BF CPU I have (quad-wide 11/83 processor)
would be just fine in a Q22/CD backplane (and possibly a Q18/CD
backplane), but in a QQ/QQ backplane, that could result in damage to
the card because of the PMI interface in the CD slots.
This is based on things I remember, I'd have to reread some manuals to
give you the 100% correct information. If I remember te 1983 QBUS
Interfaces book (
http://bitsavers.trailing-edge.com/pdf/dec/qbus/EB-23144-18_QbusIntrfs_1983…
) has sections on three of the available QBUS backplanes then (H9273
Q18/CD of the BA11-N box, H9275 Q22/Q22 OEM backplane, which can fit
in a BA11-N box (I have an H92275-A but no box...), and the H9276
Q22/CD of the BA11-S box); which also has diagrams of the
interconnection of the slots in the quad width, nine long backplanes.
Cheers,
Christian
--
Christian M. Gauger-Cosgrove
STCKON08DS0
7:59 p.m.
New subject: Serpentine QQ/QQ configuration and quad cards
Christian Gauger-Cosgrove wrote:
On 25 October
2013 13:51, azd30 <azd30 at telus.net> wrote:
As long as
quad-width cards don't mess about on the right-hand side
connectors (most of them don't) you can use any quad-width slot, whether
its wired QQ/CD or QQ/QQ.
I'm confused here. All the required signals for system operation are on the AB
side (you call it QQ?). Correct? What is on the CD side then?
I always thought that the CD side was like a 'private' communication bus, as in
the M8013 talks to the M8014. I'm not getting something here.
the PMI memories (which go *BEFORE* the processor), and
the
CannotRememberItsName11 QBUS-to-UNIBUS bridge used to make the 83 and
93 into an 84 and 94. CD is also used by the QBUS VAX cards (uVAX II
for examples) in a similar private memory bus design.
However, with the uVAX II, there is also a cable over the top
connecting the CPU to the memory, so perhaps the CD portion
of the ABCD slot is not used in the same manner as with a
PDP-11/83.
How would a
quad card not mess about on the right hand side? The only thing I can think of, is this.
If the right hand side (CD) is wired for AB there is power
on there. The quad card would just have to stay away from those pins. Does the CD part of
the quad card make use of the AB signals present?
By not mess around, one refers to not doing things on the CD
interconnect pins, and pulling the card into a serpentine backplane
(as that would result in card, and possibly other cards releasing the
magic smoke). I.e. the KDJ11-BF CPU I have (quad-wide 11/83 processor)
would be just fine in a Q22/CD backplane (and possibly a Q18/CD
backplane), but in a QQ/QQ backplane, that could result in damage to
the card because of the PMI interface in the CD slots.
This is based on things I remember, I'd have to
reread some manuals to
give you the 100% correct information. If I remember te 1983 QBUS
Interfaces book (
http://bitsavers.trailing-edge.com/pdf/dec/qbus/EB-23144-18_QbusIntrfs_1983…
) has sections on three of the available QBUS backplanes then (H9273
Q18/CD of the BA11-N box, H9275 Q22/Q22 OEM backplane, which can fit
in a BA11-N box (I have an H92275-A but no box...), and the H9276
Q22/CD of the BA11-S box); which also has diagrams of the
interconnection of the slots in the quad width, nine long backplanes.
My one KDJ11-BF CPU is in a BA123 in the third slot with
2 DEC PMI 2 MB boards in slots 1 and 2. Slot 4 has a quad
RQD11-EC (third party MSCP controller and maybe not the
correct board number - quad ESDI controller which supports
4 EDSI hard drives).
Also, with all of this discussion of ABCD slots and PMI memory,
most, maybe all, M8190 boards also support the use of PMI
memory if placed in ABCD slots with the CPU below the memory.
They will run a bit slower than an actual KDJ11-BF with the
18 MHz crystal (as opposed to a 15 MHz crystal), but the PMI
memory provides most of the increased speed. So if anyone
has actual PMI memory (at one point maybe over 10 years ago,
PMI memory was no more expensive than regular memory maybe
due to lack of demand and understanding) and with an M8190-BB
board and a BA23, they will work well together.
Jerome Fine
27 Oct
27 Oct
7:44 a.m.
New subject: Serpentine QQ/QQ configuration and quad cards
On Saturday, 26 October 2013, Jerome H. Fine <jhfinedp3k at compsys.to> wrote:
Just a suggestion, but you might mean 11/84 instead of
11/93.
Both the 11/93 and 11/94 support ONLY memory on the CPU
board. The 11/83 is all Qbus and uses Qbus PMI memory. The
11/84 is very special and uses special PMI memory that is ONLY
for an 11/84 system.
Hmm, it was my understanding that the M8190-A* J11 CPU boards were usable
as an 83 or 84. The difference between an 83 or 84 being that the 84 has a
KTJ11 (M8191) Q-U bridge. Same with the 93 and 94 also only differing in
the 94 having the KTJ11 board. That information is based on the module
field guides on can find across the net.
In terms of 93/94 PMI memory usage, I assumed they would support such,
since not all 9* processors had the full 4MB on card, and they needed to
speak with the KTJ11 card.
However, with the uVAX II, there is also a cable over
the top
connecting the CPU to the memory, so perhaps the CD portion
of the ABCD slot is not used in the same manner as with a
PDP-11/83.
Correct uVAX II had an over the top cable, hut it still used the CD
interconnect, I'm pretty sure. (I.e. For its own uVAX II specific uses.)
As far as I remember, all quad CPU boards which use
the J11
chip (11/53, 11/73, 11/83 and 11/93) may be inserted into
an ABAB slot and the memory below it will then be used as
normal memory. I do not know about the use of PMI memory
in an ABAB slot, but I definitely have placed a KDJ11-BF in a
VT103 which is definitely an ABAB slot with a 4 MB quad memory
board below it. RT-11 reported the CPU as PDP-11/73 in that
case.
Hmm, I was informed that a KDJ11-BF (M8190-AE, 11/83 or 11/84 processor,
clock at 18 MHz, and FPJ11A FPU installed) would not work in a Q22/Q22
serpentine backplane, and that it would release the magic smoke (unless one
were to cut the traces to the PMI fingers, or something equally radical
involving changing the CPU board or the backplane). That's the primary
reason I've been searching for a KDJ11-AC (M8192-YC, LSI-11/73 dual wide
processor, with FPJ11A FPU) and 22-bit boot ROM board, since I specifically
cannot use the M8190 in my H9275, while the M8192 would work fine.
Based on what I was told at least.
Also, with all of this discussion of ABCD slots and
PMI memory,
most, maybe all, M8190 boards also support the use of PMI
memory if placed in ABCD slots with the CPU below the memory.
They will run a bit slower than an actual KDJ11-BF with the
18 MHz crystal (as opposed to a 15 MHz crystal), but the PMI
memory provides most of the increased speed. So if anyone
has actual PMI memory (at one point maybe over 10 years ago,
PMI memory was no more expensive than regular memory maybe
due to lack of demand and understanding) and with an M8190-BB
board and a BA23, they will work well together.
Isn't that configuration (PMI processor, non-PMI RAM) the so called
"11/73B"? In either case, I personally don't have any capability for PMI,
and OP's machine also sounds like PMI would not work on it.
Plus, haven't PMI prices skyrocketed? Then again I'm looking for an
MSV11-QC (i.e. 4MB QBUS memory) board, so I shouldn't mention large and
expensive prices.
Cheers,
Christian
...Who is still looking for an M8192-YC (it must be an M8192-YC (you know,
the one with the FPJ11 installed), M8192-YB and M8192-YA not accepted), and
some 22-bit boot ROM board (M8012 is fine, if its a 22-bit revision) to
trade for n M8190-AE.
--
Christian M. Gauger-Cosgrove
STCKON08DS0
5:03 p.m.
New subject: Serpentine QQ/QQ configuration and quad cards
On 27/10/2013 02:59, Jerome H. Fine wrote:
That's not right.
The 11/93 and 11/94 were not supported with memory other than on the CPU
board, but some revisions did allow that.
The 11/84 uses exactly the same processor boards as the 11/83 and also
uses exactly the same PMI memory boards. The difference is that the
11/84 has also a Qbus to Unibus converter for the rest of the system and
does not support QBus peripherals.
Actually the very earliest MSV11-J modules would work in an 11/84 but
were not supported in an 11/73 or /83 because of a bug in an ASIC, but
later revisions were fixed and worked in all 11/73, 11/83, and 11/84.
Christian
Gauger-Cosgrove wrote:
The CD interconnect is also used on the PMI capable QBUS cards, so the
11/83 and 11/93 processors (and third party replacements for those),
Just a suggestion, but you might mean 11/84 instead of 11/93.
Both the 11/93 and 11/94 support ONLY memory on the CPU
board. The 11/83 is all Qbus and uses Qbus PMI memory. The
11/84 is very special and uses special PMI memory that is ONLY
for an 11/84 system. However, with the uVAX II, there is also a cable over
the top
connecting the CPU to the memory, so perhaps the CD portion
of the ABCD slot is not used in the same manner as with a
PDP-11/83.
Indeed, it's not, because it doesn't (can't) use PMI nor normal QBus
memory. The over-the-top cabling carries the signals for wider larger
memory, along with some of C/D for memory access. But the C/D signals
are used differently to PMI.
As far as I remember, all quad CPU boards which use
the J11
chip (11/53, 11/73, 11/83 and 11/93) may be inserted into
an ABAB slot and the memory below it will then be used as
normal memory. I do not know about the use of PMI memory
in an ABAB slot, but I definitely have placed a KDJ11-BF in a
VT103 which is definitely an ABAB slot with a 4 MB quad memory
board below it. RT-11 reported the CPU as PDP-11/73 in that
case.
Are you sure you're remembering that correctly? All the manuals say
not, and Micronote 039 explicitly states "KDJ11-B cannot be used in a
backplane (or that part of the backplane) where there is Q-bus in both
AB and CD slots" and that's because it will get some signals on pins
that expect either nothing or PMI.
PMI memory will definitely not work in Q/Q backplanes because it needs
the C/D Interconnect. Moreover, PMI memory manuals take great trouble
to point out "Do not install MSV11-J in a Q-Q backplane since this
backplane shorts Q-bus signals to PMI bus signals and the system will
not operate. This can also result in damage to the MSV11-J and other
modules in a Q-Q backplane."
For example, some of the PMI signals will be shorted to ground, one to
+5V, one will be shorted to the QBus HALT line, and several others
shorted to incompatible QBus signals.
Also, with all of this discussion of ABCD slots and
PMI memory,
most, maybe all, M8190 boards also support the use of PMI
memory if placed in ABCD slots with the CPU below the memory.
True, all M8190 can use QBus memory if it's below the processor. But
then it's not operating as PMI memory; a PMI board below the processor
operates as ordinary QBus memory, and that's why it's slower.
--
Pete Peter Turnbull
Network Manager
University of York
26 Oct
26 Oct
4:52 p.m.
New subject: Serpentine QQ/QQ configuration and quad cards
On Oct 25, 2013, at 13:51, azd30 <azd30 at telus.net> wrote:
How would a quad card not mess about on the right hand
side? The only thing I can think of, is this. If the right hand side (CD) is wired for AB
there is power
on there. The quad card would just have to stay away from those pins. Does the CD part of
the quad card make use of the AB signals present?
By and large, quad cards (except ones specifically meant to use
the CD pins, like KDJ11-B, PMI memory and RLV11) don't
touch much on the CD slots. Usually they take the power, and they
may have jumpers to optionally provide grant continuity for
serpentine configurations (you would pull the jumper out for CD backplanes;
the KDF11-B does that, among others).
- Dave
I'm going by this:
http://www.ibiblio.org/pub/academic/computer-science/history/pdp-11/hardwar…
and my 'microcomputers and memories (1982)' handbook.
I'm feeling really 'dumb' right now...
cheers
--
alex
7:30 p.m.
Roe Peterson wrote:
On 10/25/2013
05:23 AM, Jerome H. Fine wrote:
This is possible, but seems VERY unlikely with
just
a 4 slot backplane. As far as I understand, in order
to be useful to support an RL controller, at least 2 slots
must be wired as ABCD.
I've got an original pdp11/03 in a four-slot backplane, _all_ the
slots are Q18/Q18 serpentine. The RLV12 (quad width) lives in slot 3,
the system works great. The BA23 box
has the first
three slots as ABCD and the other five slots as ABAB,
so the concept is possible. Having two ABCD slots also
supports PMI memory with the memory being placed in
slot 1 and the CPU in slot 2, but that requires at least
a quad M8190 CPU card.
Assuming the bus is in fact
QQ/QQ serpentine, yes. It gather that I insert the CPU in 1AB, then
rather than putting the
next card in 2AB, I would place it in 1CD. Next would be 2CD and so
on ,snaking down the line. Is this correct?
While this is all fine for dual cards, but how does
this configuration
handle quad cards?
With an M8186
to start, the BDAL17-20 lines should be
either unconnected (for an 18 bit backplane) or ONLY
connected to each other (for a 22 bit backplane).
Just a clarification, the high-order four address lines are BDAL18-21... from another post (or maybe an early version of this
thread)
and forgot to verify the numbers.
Jerome Fine
27 Oct
27 Oct
5:03 p.m.
On 27/10/2013 02:30, Jerome H. Fine wrote:
As for the PDP-11/03 you mention, it is my impression
that
BDAL18-21 must be wired - maybe with power on one or
more of those lines whereas with the dual PDP-11/23 CPU
(M8186 board), BDAL18-21 must be either empty (which
is a Q18/Q18 in that case) or wired ONLY to each other
(Q22/Q22). Specifically, a PDP-11/23 (M8186 board)
must not be inserted into a backplane (and power turned
on) wired for the PDP-11/03 (or LSI-11/03).
Not so. The 11/03 backplanes don't do anything dangerous with the lines
that would otherwise be BDAL18-21 (BC1,BD1,BE1,BF1). They're simply not
connected together. The 11/03 processors, both dual and quad versions,
put some odd signals there, and the earliest (Rev.A) KDF11-A dual cards
(M8186) also did (but not the same odd signals). It was very common,
for example, to replace an 11/03 CPU with a KDF11-A, particularly in
BA11-N boxes, and no other change was necessary. Even the 4x4 H9270
backplane in a BA11M is OK; DEC merely list it as "restricted
compatibility" because it is "18-bit addressing only" -- DEC even listed
a special version called H9270-Q that was 22-bit and fully compatible,
with the extra lines bussed.
I forget who asked that question, but my
understanding
differs. With the VT103 backplane which is a 4 slot
ABCD (QQ/QQ) serpentine, if a dual M8196 is in
1 AB, then 2 AB must also be used - either with a
useful dual board or a dual bus grant board. Can
someone else please verify this point as to which is
correct?
Yes, you are correct -- if you have a dual card in AB of the top slot in
any serpentine backplane, and you need to preserve the grant chain, then
you must have either a dual card or a grant card in CD of that slot. If
you don't care about the grant chain (and some systems don't) then it
doesn't matter. In this respect the top slot is no different to any other.
--
Pete Peter Turnbull
Network Manager
University of York
4073
days inactive
4076
days old
11 comments
6 participants
participants (6)
-
azd30＠telus.net -
captainkirk359＠gmail.com -
fraveydank＠gmail.com -
jhfinedp3k＠compsys.to -
pete＠dunnington.plus.com -
roe＠liveblockauctions.com