27 Dec
2008
27 Dec
'08
12:52 p.m.
On Sat, Dec 27, 2008 at 11:39 AM, Dave McGuire <mcguire at neurotica.com> wrote:
I know I've seen SMD-E drives in 5.25" form-factor, but I don't know
if SMD-E drives are happy on an older SMD controller (despite using
SMD drives since 1984, I really don't know lots about their
limitations - I only ever worked with a couple of combinations that
were known to work - I never had to match up settings and cables,
etc., on random configurations as I later did plenty of times with MFM
and ESDI).
Given the drives I do have on hand, I would, of course, love to have a
Unibus SCSI controller (or 3-4 or them, really), but they were really,
really rare back in the day. Every once in a while, I entertain the
idea of turning old COMBOARDs into some form of disk controller (IDE
or SCSI would be the easiest), but I never get past the napkin stage
of designing. What would be far more practical would be an Unibus
ESDI controller, but I don't remember seeing too many of those in the
past (what I did see was "lots" of SMD for Unibus and SMD or ESDI for
Qbus - all popular for those that were happy enough with non-DEC
controllers).
It all comes down to drivers. 2BSD I know can handle Emulex and other
controllers just fine. I'm not as certain about Ultrix-32 (it seemed
to be heavily slanted towards a DEC system disk, at least). I recall
a variety of releases of VMS drivers for SI and Emulex controllers, so
it would be handy, I'd say, to identify what OS and version targets,
seek out ancient driver install kits _then_ go looking for a suitable
controller, but I think in practice, it'll be a case of finding a $50
card and then looking for drivers.
It's this struggle that makes me start to envision what it would take
to turn a COMBOARD into a low-performance disk controller. I say "low
performance" because no matter how one would reasonably hack on the
existing hardware, and though it has a Unibus DMA engine (the 18-bits
of the Unibus are mapped into 1/4 of the memory space of the onboard
68000, so DMA cycles are as easy as a DBcc loop), it's programmed I/O
On Dec 26, 2008, at 11:04 PM, Bob Armstrong wrote:
>
> I've also got a 11/725 that's in good shape except for a few missing
> pieces of sheet metal, but like Ethan I've no working RC25 drive.
I don't know about the state of your drive, Bob, but mine is merely
missing a cartridge so I can't spin up the fixed platter. For those
that might not know, the RC25 is 25MB/25MB removable on one spindle.
I think the original idea was that you would put VMS on the internal
platter and swap out the cartridge for either backup/restore, and
perhaps installs, but in practice, VMS got large enough fast enough
that you really needed all 50MB for the OS and a few user dirs.
You _can_ fit 5.0 on it, I've seen it, but 4.x is a better fit.
A UNIBUS SMD
controller is an option - there are some small 8" SMD disks that mist
just fit inside the chassis in place of the RC25 - but a UNIBUS SCSI
controller would be a better way to go, since it'd be no problem to fit a
SCSI disk in there. Unfortunately neither SMD nor SCSI UNIBUS controllers
are easy to come by.
5.25" SMD drives do exist; Seagate (and possibly others) made them. I had
a few at one point (1990?) but no more. I know someone who has some but I
doubt he'd turn loose of them. from the 68000 side, limiting the max transfer speed to
about
200KB/sec. You _could_ stick a disk on it, but it really was designed
as a communications controller. Since you'd have to write the drivers
from scratch anyway, one might as well start with
something that's
_meant_ to talk to disks. I only keep coming back to the idea
since I
have a stack of working boards on the shelf and I have 100% of the
engineering info on them (and all the rights).
Thinking of COMBOARDs, ISTR someone (more than one?) on the list found
a COMBOARD in a machine they picked up some time back. I am still
trying to recreate our old test/development environment with simh, but
I'm lacking in a way to bootstrap a simulated 11/780 with VMS 4.x.
Once I get around the issue, I have backup saveset files made from the
last days of Software Results I can restore - all of the source and
tools and textual docs are there, and I'd be able to "cut a tape" of
the install kit for a variety of OSes and COMBOARD products. This is
assuming anyone still wants to talk HASP or 3780 from a real Unibus
box, of course. That's what the board did, and it did do it well.
The disk-controller idea is just something I kick around from time to
time.
-ethan
27 Dec
27 Dec
1:29 p.m.
Ethan Dicks wrote:
I don't know about the state of your drive, Bob,
but mine is merely
missing a cartridge so I can't spin up the fixed platter.
I've got cartridges, but I really, really, (_really_!) doubt that it'll
make your drive work. RC25s were notoriously unreliable even when they were
new, and as unsealed (even the fixed platter is open to the air) drives they
just don't age well at all. I've got two RC25 drives (well, maybe even
three but that's another story) and none will work. I spent a couple of
weeks working on them once, and all will now try to spin up and then fault
with various error conditions (I've since forgotten the error codes -
sorry).
Every once in a while, I entertain the
idea of turning old COMBOARDs into some form of disk controller (IDE
or SCSI would be the easiest),
Another option would be to build something that plugs into the LESI (aka
Aztec) controller and pretends to be a disk. That'd be way cool. But I've
never seen any documentation on the LESI interface, either electrical or the
protocol, so I've no idea how hard that would be. As long as the interface
remains undocumented, I'd guess "really hard". And of course the LESI
interface wasn't very popular (it was only ever used for the TU81 and the
RC25) so the potential market, outside of you and me, is limited :-)
The important thing is that whatever you come up with has to be close
enough to a standard MSCP controller so that you can boot it with the
standard DU bootstrap and so that VMB can talk to it. Unfortunately some of
the third party SMD controllers weren't really MSCP compatible and needed
custom bootstraps and VMS drivers (we used to have an SI controller on a 780
that fell into this category) - those would be a problem unless a) you can
also recover the software (difficult), and b) they supported a 11/730 (even
more unlikely!). The LESI approach at least avoids this problem.
Of course if you only want to run Un*x then you have more flexibility :-)
Bob
2:10 p.m.
On Sat, Dec 27, 2008 at 1:29 PM, Bob Armstrong <bob at jfcl.com> wrote:
I've owned two 11/725s over the years. As I've posted before, I've
never had the problems with the RC25 that others report. I know they
are notorious, but _I_ never had one fail. That being said, of
course, the chances of mine working are now diminished merely because
I've said something. ;-)
Yes... that would be an interesting way to go, but as you point out,
lack of documentation makes that an unlikely path.
Ethan Dicks
wrote:
I don't know about the state of your drive,
Bob, but mine is merely
missing a cartridge so I can't spin up the fixed platter.
I've got cartridges, but I really, really, (_really_!) doubt that it'll
make your drive work. RC25s were notoriously unreliable even when they were
new, and as unsealed (even the fixed platter is open to the air) drives they
just don't age well at all. I've got two RC25 drives (well, maybe even
three but that's another story) and none will work. I spent a couple of
weeks working on them once, and all will now try to spin up and then fault
with various error conditions (I've since forgotten the error codes -
sorry). Every once in a
while, I entertain the
idea of turning old COMBOARDs into some form of disk controller (IDE
or SCSI would be the easiest),
Another option would be to build something that plugs into the LESI (aka
Aztec) controller and pretends to be a disk. That'd be way cool. But I've
never seen any documentation on the LESI interface, either electrical or the
protocol, so I've no idea how hard that would be. As long as the interface
remains undocumented, I'd guess "really hard". And of course the LESI
interface wasn't very popular (it was only ever used for the TU81 and the
RC25) so the potential market, outside of you and me, is limited :-) The important thing is that whatever you come up with
has to be close
enough to a standard MSCP controller so that you can boot it with the
standard DU bootstrap and so that VMB can talk to it. Unfortunately some of
the third party SMD controllers weren't really MSCP compatible and needed
custom bootstraps and VMS drivers (we used to have an SI controller on a 780
that fell into this category) - those would be a problem unless a) you can
also recover the software (difficult), and b) they supported a 11/730 (even
more unlikely!). The LESI approach at least avoids this problem.
Agreed. OS driver support is always foremost in my mind when fiddling
with 3rd party disks. We always had to wait for SI to release driver
patches for our SI9900 since a Fuji Eagle is not the same size as any
DEC disk (in our case, they would patch the geometry table in
DRDRIVER.EXE to "oversize" the RM05 entry since we didn't have any
real RM05s on the system).
It sure would be nice to find a Unibus SCSI card that looked to the
system like a UDA50 - i.e. - true and proper MSCP emulation. I don't
know if there ever was such a product, but the VAXBI ones I saw years
later were $10,000 new. :-(
Of course if you only want to run Un*x then you have
more flexibility :-)
True, and I _do_ care about Unix (2BSD for PDP-11 and Ultrix-32 for
VAX), but I also care about RT-11 and pre-6.0 VMS.
Just thinking back to when I used to do this every day for a living, I
don't recall there ever being an ideal solution, just solutions that
fit enough criteria to be acceptable (Re: price-compatibility-capacity
matrix). If you had money, you paid for DEC disk. If you had a no
budget, you bought 3rd party and decided what features to give up (the
ability to seamlessly install the OS and upgrade at will was almost
always the first thing to go).
Not all of the "good old days" were as good as we'd like to remember.
-ethan
5:30 p.m.
On Saturday 27 December 2008, Ethan Dicks wrote:
It sure would be nice to find a Unibus SCSI card that
looked to the
system like a UDA50 - i.e. - true and proper MSCP emulation. I don't
know if there ever was such a product, but the VAXBI ones I saw years
later were $10,000 new. :-(
My UC18 seems to work like that, and I've had it up and running on my
11/780 with both VMS (a few different releases) and 4BSD.
Unfortunately for you, I'm not willing to give it up (and I only have
one). A potentially easier (and definately less expensive) path you
might try is a Qniverter and QBUS SCSI card, which could be had for
less than $100 each, last I was looking at this stuff. After I found
the UC18, though (in the 11/780 I'm using it in even), I gave up on
that path. 2U or so for a small 4-quad slot QBUS box shouldn't be too
hard to come up with. :) With some trickery, you could probably mount
the SCSI disk in there too, since there will be plenty of left-over
space.
Not all of the "good old days" were as good
as we'd like to remember.
Oh, now that's not something that I'd expect someone in this group to
say. :)
Pat
--
Purdue University Research Computing --- http://www.rcac.purdue.edu/
The Computer Refuge --- http://computer-refuge.org
7:26 p.m.
On Sat, Dec 27, 2008 at 5:30 PM, Patrick Finnegan
<pat at computer-refuge.org> wrote:
I see the smiley there, but as someone who has been playing with DEC
equipment since 1982, some stuff is brilliant, and some stuff drove us
downright nuts. Fortunately, there were many more brilliant moments.
-ethan
On Saturday 27 December 2008, Ethan Dicks wrote:
Nice.
It sure would be nice to find a Unibus SCSI card
that looked to the
system like a UDA50...
My UC18 seems to work like that, and I've had it up and running on my
11/780 with both VMS (a few different releases) and 4BSD. Unfortunately for you, I'm not willing to give it
up (and I only have
one)\.
Fair enough. I wouldn't mind eventually finding a UC17 or UC18 myself, though.
A potentially easier (and definately less expensive)
path you
might try is a Qniverter and QBUS SCSI card, which could be had for
less than $100 each, last I was looking at this stuff.
I already have a couple of Qbus SCSI cards, but I'll have to check
which ones might be TMSCP only.
After I found
the UC18, though (in the 11/780 I'm using it in even), I gave up on
that path. 2U or so for a small 4-quad slot QBUS box shouldn't be too
hard to come up with. :) With some trickery, you could probably mount
the SCSI disk in there too, since there will be plenty of left-over
space.
Well... in a BA-11, there's plenty of room. I even have a DEC 18-bit
"Qniverter" for putting an IBV11 on an 11/34 (it came with a DDV11C,
if you can imagine such a thing). I should dig that out and see if I
can get it working.
It won't solve the problem of how to put a SCSI disk _in_ an 11/750
(using the internal DD11D backplane) or inside my 11/725. It would
probably work fine in my 11/44 or an 11/70, though.
Not all of the
"good old days" were as good as we'd like to remember.
Oh, now that's not something that I'd expect someone in this group to
say. :) 
10:04 p.m.
Patrick Finnegan wrote:
A potentially easier (and definately less expensive) path you
might try is a Qniverter and QBUS SCSI card, which could be had for
less than $100 each, last I was looking at this stuff.
Interesting -- I actually hadn't run across that scenario somehow. But I
think I have more things claiming to be S-100 SCSI interfaces than I do
for the Qbus...
Hey Brad, what about that universal disk controller for UNIBUS you were
working on? http://www.heeltoe.com/retro/udisk/index.html (Apologies if
I missed an update.)
--S.
28 Dec
28 Dec
8:35 p.m.
Steven M Jones wrote:
Hey Brad, what about that universal disk controller for UNIBUS you were
working on? http://www.heeltoe.com/retro/udisk/index.html (Apologies if
I missed an update.)
ugg. It works, but is currently collecting dust waiting for the economy
to force me into another idle period (which I predict will occur soon).
I programmed it to look like an RL02 and got it to boot rt-11 on my
11/44. But I didn't like some of the hacks I had to do and so re-did
the cpld and then ran out of space. Always the way.
I plan to have another go at it this spring/summer. I think the idea
has merit. I think I should have used an fpga instead of a cpld however
but I was being cheap. It's nice to put the entire bus transaction
engine with dma support inside the cpld. Originally I had the cpu doing
everything but found it was hard to keep bus timeouts from happening.
The last working version has several state machines which minimize the
cpu work.
It can look like rom and can be a bus master so I use it to goose a copy
of the boot loader into ram which makes life easier.
like everyone, I took on some huge projects over the last 1.5 years and
all my fun work was on hold. But those days appear to be over for a
while :-| so my back room will be a little warmer.
-brad
5684
days inactive
5686
days old
6 comments
5 participants
participants (5)
-
bob＠jfcl.com -
brad＠heeltoe.com -
classiccmp＠crash.com -
ethan.dicks＠gmail.com -
pat＠computer-refuge.org