7 May
2014
7 May
'14
1 p.m.
Hi,
Now that everything's moved and more or less positioned in my new shop, I'm trying
to get some things running again. It's a longer term project but I figured I'd
start asking around now.
My 2065 (DECSystem-20) is missing some key pieces that I'd like to see if someone in
the community has and is willing to sell or let me borrow so that I can try to replicate
them (the prints that I have are missing critical bits of information). Here's what
I'm looking for:
KW20 "master oscillator". This is required when using MF20 memories. I've
found a block diagram but that's all I've found out about it. Any help (either in
prints or locating one) would be appreciated.
BC20C S-Bus cables. I figure if I borrow one or at least get a high res photo (both
sides) of the M9006 paddle I can replicate them since I'll need several.
I'm already planning on replacing the CPU supplies with modern switchers. That'll
do two things:
Eliminate the need for 3-phase power
Reduce the CPU power consumption down to ~2200W (from the 6000W with the linear
supplies).
Of course, I still need to do a massbus emulator for the disks but getting the CPU and
other bits running first is the priority.
Thanks.
TTFN - Guy
7 May
7 May
3:57 p.m.
The KW20 was not used on core-based systems, and was introduced with the
MF20 semiconductor memory, so the drawings for the KW20 master oscillator
are in the MF20 and MG20 print sets, which are on Bitsavers.
DEC used custom DC009 transceivers for the control signals from the M8572
to the KW20 that are used to select the oscillator frequency (25, 30, or 31
MHz, or external). These are the same transceivers used for the Xbus
(semiconductor memory bus from the KL memory bus translators to the
MF20/MG20 boxes). I've never found any DC009 specifications or other docs,
so I have no idea what the signalling is like.
The six coax clock outputs are driven in sets of three by two MC10210 ECL
dual 3-input 3-output OR gates, with near-end thevenin terminations
(voltage dividers) across outputs 2 through 6, but not output 1. It should
be trivial to build a replacement if you don't mind a fixed frequency or
manually switching the frequency. The MC10210 appears to still be
reasonably easy to find.
25 MHz is used for the KL10-PA ("Model A", no section support), 30 MHz is
used for the KL10-PV and -PW ("Model B"), and 31 MHz is used for
margin-testing the -PV and -PW.
The 2065 (or any KL10 with internal MF20/MG20 MOS memory) does not have an
Sbus. It has Xbus instead. The Sbus translators (M8519) in slots 7 and 8
of the CPU backplane are replaced by Xbus translators (M8581, unless you
have a mixed core/MOS system). The Sbus cables and paddles (slots 2 and 3)
are replaced by an Xbus cable that uses an M8572 paddle (schematic and
component placement in the MF20 print set). The Xbus cabling looks similar
to VAX SBI cabling. I'm pretty sure you got at least one M8572/cable
assembly from me, when we were scrounging for missing parts.
If you're building your own memory system instead of using an MF20/MG20, it
might be easier to use Sbus, since it uses readily available parts for the
interface, vs. the unobtanium and undocumented DC009. Sbus uses SN75453
drivers and MC10124 receivers. The Sbus cables and M9006 paddle are
nothing special, since all the electronics is on the M8519. The Xbus M8572
paddle has a small amount of electronics to drive the clock select signals
over a separate ribbon cable to the KW20, but aside from that is passive.
In either case (Sbus or Xbus), you need two translator boards
(M8519/M8580/M8581), but one cable and paddle board (M9006/M8572) supports
two chained memory boxes. You should be able to get by with a single paddle
board, unless for some reason you want to use three or four memory boxes.
Eric
4:20 p.m.
On May 7, 2014, at 3:57 PM, Eric Smith <spacewar at gmail.com> wrote:
The KW20 was not used on core-based systems, and was
introduced with the
MF20 semiconductor memory, so the drawings for the KW20 master oscillator
are in the MF20 and MG20 print sets, which are on Bitsavers.
I've been through the print sets and all I've found is the block diagrams. If you
can
point me to the specific doc then I'll look for it there.
DEC used custom DC009 transceivers for the control signals from the M8572
to the KW20 that are used to select the oscillator frequency (25, 30, or 31
MHz, or external). These are the same transceivers used for the Xbus
(semiconductor memory bus from the KL memory bus translators to the
MF20/MG20 boxes). I've never found any DC009 specifications or other docs,
so I have no idea what the signalling is like.
Right.
The six coax clock outputs are driven in sets of three by two MC10210 ECL
dual 3-input 3-output OR gates, with near-end thevenin terminations
(voltage dividers) across outputs 2 through 6, but not output 1. It should
be trivial to build a replacement if you don't mind a fixed frequency or
manually switching the frequency. The MC10210 appears to still be
reasonably easy to find.
OK.
25 MHz is used for the KL10-PA ("Model A", no section support), 30 MHz is
used for the KL10-PV and -PW ("Model B"), and 31 MHz is used for
margin-testing the -PV and -PW.
Manual switching is OK though there are commands to switch it but I think
once I have it set up I'll leave it at what ever works.
The 2065 (or any KL10 with internal MF20/MG20 MOS memory) does not have an
Sbus. It has Xbus instead. The Sbus translators (M8519) in slots 7 and 8
of the CPU backplane are replaced by Xbus translators (M8581, unless you
have a mixed core/MOS system). The Sbus cables and paddles (slots 2 and 3)
are replaced by an Xbus cable that uses an M8572 paddle (schematic and
component placement in the MF20 print set). The Xbus cabling looks similar
to VAX SBI cabling. I'm pretty sure you got at least one M8572/cable
assembly from me, when we were scrounging for missing parts.
Yes, I have that assembly. That was critical too. But I think there are other
cases where the S-Bus is used and I need the cables for that. As I recall the
other uses don't involve talking to memory (I/O?).
If you're building your own memory system instead of using an MF20/MG20, it
might be easier to use Sbus, since it uses readily available parts for the
interface, vs. the unobtanium and undocumented DC009. Sbus uses SN75453
drivers and MC10124 receivers. The Sbus cables and M9006 paddle are
nothing special, since all the electronics is on the M8519. The Xbus M8572
paddle has a small amount of electronics to drive the clock select signals
over a separate ribbon cable to the KW20, but aside from that is passive.
Yes, I'm wanting to see what the M9006 paddle looks like so I can just
"clone"
it rather than having to go through the prints and figure out the exact wiring
(which I'm likely to screw up the first time or two).
If the MF20 proves too buggy, I may do my own (shouldn't be too hard).
In either case (Sbus or Xbus), you need two translator boards
(M8519/M8580/M8581), but one cable and paddle board (M9006/M8572) supports
two chained memory boxes. You should be able to get by with a single paddle
board, unless for some reason you want to use three or four memory boxes.
TTFN - Guy
5 p.m.
On May 7, 2014, at 4:20 PM, Guy Sotomayor <ggs at shiresoft.com> wrote:
I went and checked and there are 2 S-Bus cables between the CPU cabinet
and the I/O cabinet. The MF20's as far as I can tell are connected from the
I/O cabinet using the X-Bus as Eric described but there's still the requirement
to get the signals from the CPU which still requires two BC20C cables.
Is there a high quality picture or scan of the M9006 paddle? Or does someone
have BC20C's that I could either borrow (to clone) or that I could purchase?
TTFN - Guy
On May 7, 2014, at 3:57 PM, Eric Smith <spacewar at gmail.com> wrote:
I found the print for the KW20. Ugh!
Anyone have one that they're willing to part with? I'll pay $$'s. ;-)
The KW20 was not used on core-based systems, and
was introduced with the
MF20 semiconductor memory, so the drawings for the KW20 master oscillator
are in the MF20 and MG20 print sets, which are on Bitsavers.
I've been through the print sets and all I've found is the block diagrams. If
you can
point me to the specific doc then I'll look for it there.
The 2065 (or any KL10 with internal MF20/MG20 MOS memory) does not have an
Sbus. It has Xbus instead. The Sbus translators (M8519) in slots 7 and 8
of the CPU backplane are replaced by Xbus translators (M8581, unless you
have a mixed core/MOS system). The Sbus cables and paddles (slots 2 and 3)
are replaced by an Xbus cable that uses an M8572 paddle (schematic and
component placement in the MF20 print set). The Xbus cabling looks similar
to VAX SBI cabling. I'm pretty sure you got at least one M8572/cable
assembly from me, when we were scrounging for missing parts.
Yes, I have that assembly. That was critical too. But I think there are other
cases where the S-Bus is used and I need the cables for that. As I recall the
other uses don't involve talking to memory (I/O?). 
5:14 p.m.
On 5/7/14 5:00 PM, Guy Sotomayor wrote:
Anyone have one that they're willing to part with?
I'll pay $$'s. ;-)
You might want to talk to Paul Pierce. I don't know if LCM bought his 206?, or
what his plans are for it.
That particular memory was pretty late in the life of the KL, so I imagine there
never were very many of them.
6:14 p.m.
On Wed, May 7, 2014 at 6:14 PM, Al Kossow <aek at bitsavers.org> wrote:
That particular memory was pretty late in the life of
the KL, so I imagine
there
never were very many of them.
That might be true of the MG20, but there were a zillion MF20s made, and
they used the same cabling as the MF20.
Aside from using the higher density M8570 storage array in place of the
M8579, and requiring the H7131 5.0V 35A power supply rather than the H7131A
5V 20A supply used in some MF20s, the MF20 and MG20 are identical.
In any case, the BC20C cables Guy needs are not used with either the MF20
or MG20.
6:01 p.m.
On Wed, May 7, 2014 at 6:00 PM, Guy Sotomayor <ggs at shiresoft.com> wrote:
I went and checked and there are 2 S-Bus cables
between the CPU cabinet
and the I/O cabinet.
Perhaps it is nitpicking, but there is no Sbus going between the CPU
cabinet and the I/O cabinet. There are BC20C cables with M9006 paddle
cards, but it isn't Sbus, it's Cbus. That's why I couldn't figure out
why
you thought you needed Sbus.
9:47 p.m.
On Wed, May 07, 2014 at 05:00:39PM -0700, Guy Sotomayor wrote:
Is there a high quality picture or scan of the M9006 paddle? Or does someone
have BC20C's that I could either borrow (to clone) or that I could purchase?
I might be able to help you out by snapping a few pics. I am not
very familiar with these lovely orange boxes, could you describe
what you need? If I need to do a lot of dissasembling of our
machine that might be a problem.
Regards,
Pontus
11:33 p.m.
On Wed, May 7, 2014 at 10:47 PM, Pontus Pihlgren <pontus at update.uu.se>wrote:
On Wed, May 07, 2014 at 05:00:39PM -0700, Guy
Sotomayor wrote:
It's not very much disassembly. There are three cabinets that outwardly
look nearly identical, and there is some variation between KL models as to
what is installed in each cabinet. One of them is the console processor
cabinet, containing the PDP-11/40 front end, and you won't need to do
anything with that one.
The main CPU ("APR", Arithmetic Processor) is usually in the rightmost
cabinet as seen from the front. There's a 54-slot backplane full of ECL
modules. Since that dissipates a lot of power, there are air ducts above
and below it, not found in the other cabinets. The backplane is mounted to
the front side of the cabinet, so module access is from the rear side.
IIRC there's a metal cover with captive screws over the modules. If you
remove the cover, you'll see that most of the modules are hex height with
metal ejector handles, but slot 1 (rightmost) has three dual height paddle
cards with cables attached and magenta plastic handles. In slot one, the
top position (AB) should be a BC11A cable with M919 paddle cards, as would
be used for Unibus, though in this case it's not being used for Unibus. We
don't need photos of that one. In the next two positions down (CD and EF),
there should be BC20C cables with M9006 paddle cards, which are the ones
that Guy needs photos of.
There probably aren't any wiring differences, but just to be certain, I'd
recommend getting photos of both sides of both M9006 paddle cards, and also
getting photos of both M9006 paddle cards at the other end of those
cables. The other ends should be in slot 37CD and EF (leftmost) of the
RH20/DTE20 backplane in the "I/O cabinet" (the one that doesn't contain the
PDP-11/40 front end).
If you can get photos, I'd appreciate getting copies of them.
Thanks!
Eric
Is there a high quality picture or scan of the
M9006 paddle? Or does
someone
have BC20C's that I could either borrow (to
clone) or that I could
purchase?
I might be able to help you out by snapping a few pics. I am not
very familiar with these lovely orange boxes, could you describe
what you need? If I need to do a lot of dissasembling of our
machine that might be a problem.
8 May
8 May
12:30 a.m.
On Thu, May 08, 2014 at 12:33:59AM -0600, Eric Smith wrote:
It's not very much disassembly. There are three cabinets that outwardly
look nearly identical, and there is some variation between KL models as to
what is installed in each cabinet. One of them is the console processor
cabinet, containing the PDP-11/40 front end, and you won't need to do
anything with that one.
The main CPU ("APR", Arithmetic Processor) is usually in the rightmost
cabinet as seen from the front. There's a 54-slot backplane full of ECL
modules. Since that dissipates a lot of power, there are air ducts above
and below it, not found in the other cabinets. The backplane is mounted to
the front side of the cabinet, so module access is from the rear side.
IIRC there's a metal cover with captive screws over the modules. If you
remove the cover, you'll see that most of the modules are hex height with
metal ejector handles, but slot 1 (rightmost) has three dual height paddle
cards with cables attached and magenta plastic handles. In slot one, the
top position (AB) should be a BC11A cable with M919 paddle cards, as would
be used for Unibus, though in this case it's not being used for Unibus. We
don't need photos of that one. In the next two positions down (CD and EF),
there should be BC20C cables with M9006 paddle cards, which are the ones
that Guy needs photos of.
There probably aren't any wiring differences, but just to be certain, I'd
recommend getting photos of both sides of both M9006 paddle cards, and also
getting photos of both M9006 paddle cards at the other end of those
cables. The other ends should be in slot 37CD and EF (leftmost) of the
RH20/DTE20 backplane in the "I/O cabinet" (the one that doesn't contain
the
PDP-11/40 front end).
If you can get photos, I'd appreciate getting copies of them.
I'll see what I can do on saturday. I know where the APR cabinet is and
it shouldn't be to hard to roll i out far enough to open. I know where
the I/O cabinet is too, but the cables are probably disconnected after
the previous move, but I should probably be able to find the right slot
anyway.
It's mostly a matter of time and weather :)
/P
8:11 a.m.
On May 7, 2014, at 9:47 PM, Pontus Pihlgren <pontus at Update.UU.SE> wrote:
On Wed, May 07, 2014 at 05:00:39PM -0700, Guy
Sotomayor wrote:
Thanks. I think Eric described in detail what was needed. I appreciate
the effort.
TTFN - Guy
Is there a high quality picture or scan of the M9006 paddle? Or does someone
have BC20C's that I could either borrow (to clone) or that I could purchase?
I might be able to help you out by snapping a few pics. I am not
very familiar with these lovely orange boxes, could you describe
what you need? If I need to do a lot of dissasembling of our
machine that might be a problem.
11 May
11 May
1:55 p.m.
On Thu, May 08, 2014 at 08:11:45AM -0700, Guy Sotomayor wrote:
Thanks. I think Eric described in detail what was needed. I appreciate
the effort.
I got some pictures which I hope will help. But I've not had
five mins to sit dpwn and upload them and I'm not near my camera
now.
I'll try to get them up tomorrow.
/P
8:59 p.m.
On May 11, 2014, at 1:55 PM, Pontus Pihlgren <pontus at Update.UU.SE> wrote:
On Thu, May 08, 2014 at 08:11:45AM -0700, Guy
Sotomayor wrote:
Thanks.
TTFN - Guy
Thanks. I think Eric described in detail what was needed. I appreciate
the effort.
I got some pictures which I hope will help. But I've not had
five mins to sit dpwn and upload them and I'm not near my camera
now.
I'll try to get them up tomorrow.
12 May
12 May
11:24 a.m.
On 05/12/2014 05:59 AM, Guy Sotomayor wrote:
I was unable to move the cabinets to access the rear but luckily the CPU
and I/O cabinets where not connected and one M9006B was hanging from its
cable and reachable from the front and I could take some pictures:
http://www.update.uu.se/~pontus/slask/M9006B/
I hope this help.
/P
I got some pictures which I hope will help. But I've not had
five mins to sit dpwn and upload them and I'm not near my camera
now.
I'll try to get them up tomorrow.
Thanks.
TTFN - Guy
11:50 a.m.
On May 12, 2014, at 11:24 AM, Pontus <pontus at update.uu.se> wrote:
On 05/12/2014 05:59 AM, Guy Sotomayor wrote:
I was unable to move the cabinets to access the rear but luckily the CPU and I/O cabinets
where not connected and one M9006B was hanging from its cable and reachable from the front
and I could take some pictures:
http://www.update.uu.se/~pontus/slask/M9006B/
I hope this help.
Thanks! It helps a lot. I still have to make some educated guesses but it's a lot
more feasible rather that trying to figure it out only from the schematics.
TTFN - Guy
I got some pictures which I hope will help. But I've not had
five mins to sit dpwn and upload them and I'm not near my camera
now.
I'll try to get them up tomorrow.
Thanks.
TTFN - Guy
2:28 p.m.
Found the M9006 unit assembly, which includes PCB etch, in the MA20 and
MB20 FMPS, and a diagram showing the signal assignments for the Cbus & Ebus
on CPU backplane slot 54 in the RH20 FMPS.
MA20 FPMS, pages 58-62:
http://bitsaversorg/pdf/dec/pdp10/memory/MP00010_MA20_Schematic_Jul76.pdf
MB20 FMPS, pages 70-74:
http://bitsavers.org/pdf/dec/pdp10/memory/MP00179_MB20_Schematic_Aug76.pdf
RH20 FMPS, page 64-65:
http://bitsavers.org/pdf/dec/pdp10/KL10/MP00056_RH20_Apr81.pdf
AB54 is the M929, and CD54 and EF54 are the M9006. Note typo in
designation of F54 on page 65.
2:53 p.m.
On May 12, 2014, at 2:28 PM, Eric Smith <spacewar at gmail.com> wrote:
Found the M9006 unit assembly, which includes PCB
etch, in the MA20 and
MB20 FMPS, and a diagram showing the signal assignments for the Cbus & Ebus
on CPU backplane slot 54 in the RH20 FMPS.
MA20 FPMS, pages 58-62:
http://bitsaversorg/pdf/dec/pdp10/memory/MP00010_MA20_Schematic_Jul76.pdf
MB20 FMPS, pages 70-74:
http://bitsavers.org/pdf/dec/pdp10/memory/MP00179_MB20_Schematic_Aug76.pdf
RH20 FMPS, page 64-65:
http://bitsavers.org/pdf/dec/pdp10/KL10/MP00056_RH20_Apr81.pdf
AB54 is the M929, and CD54 and EF54 are the M9006. Note typo in
designation of F54 on page 65.
Cool!
Thanks Eric. I hadn't thought to look through those docs. I expected that
they'd
have been with the CPU.
TTFN - Guy
7 May
7 May
5:44 p.m.
On May 7, 2014, at 3:57 PM, Eric Smith <spacewar at gmail.com> wrote:
> The KW20 was not used on core-based systems, and
was introduced with the
> MF20 semiconductor memory, so the drawings for the KW20 master oscillator
> are in the MF20 and MG20 print sets, which are on Bitsavers.
On Wed, May 7, 2014 at 5:20 PM, Guy Sotomayor <ggs at shiresoft.com> wrote:
I've been through the print sets and all I've
found is the block diagrams.
If you can
point me to the specific doc then I'll look for it there.
From either MP00622_MF20_Oct82.pdf or
MP01904_MG20_Jun84.pdf, pages 9
through 12 of PDF files:
D-DD-5412851-0 Rev B: Drawing Directory, Master Oscillator
D-UA-5412851-0-0 Rev D: Unit Assembly, Master Oscillator
D-CS-5412851-0-OSC1 Rev A: Circuit Schematic, Master Oscillator Select Logic
D-CS-5412851-0-OSC2 Rev B: Circuit Schematic, Master Oscillator Power Supply
> The Xbus cabling looks similar
> to VAX SBI cabling. I'm pretty sure you got at least one M8572/cable
> assembly from me, when we were scrounging for missing parts.
Yes, I have that assembly. That was critical too. But I think there are
other
cases where the S-Bus is used and I need the cables for that. As I recall
the
other uses don't involve talking to memory (I/O?).
Nope. Sbus and Xbus don't do anything other than talk to to memory, Sbus
to MA20/MB20 core or DMA20 external KI memory bus, and Xbus to MF20/MG20
semiconductor. The only thing that they do when talking to memory boxes
that isn't actually a memory access is read/write of the memory controller
diagnostic registers. In a system with external (Ibus) memory, which is
never used on a 20xx, one Sbus (slot 3) is jumpered to slot 3 of the DMA20
memory interface.
Everything else is over the Cbus (channel) and Ebus (internal I/O), from
slot 1 of CPU backplane to slot 37 of RH20/DTE20 backplane. In a system
with a DIA20/DIB20 external I/O adapter adapter, the Cbus and Ebus are
jumpered from slot 1 of the RH20/DTE20 backplane to the DIB20 or
DIA20/DMA20 backplane.
The Cbus (slot 1CD and 1EF) does use BC20C cables to the RH20/DTE20, but
it's not Sbus.
There's a drawing of the BC20C in MP00300_KL10ErevM_Feb83.pdf on pages
182-183, but it doesn't give a schematic.
3879
days inactive
3884
days old
17 comments
4 participants
participants (4)
-
aek＠bitsavers.org -
ggs＠shiresoft.com -
pontus＠Update.UU.SE -
spacewar＠gmail.com