3 Jan
2014
3 Jan
'14
2:09 p.m.
On Fri, Jan 3, 2014 at 4:17 PM, Mark J. Blair <nf6x at nf6x.net> wrote:
I'm just getting started in DEC stuff in general
and RL02 drives in particular. I'm
putting together a PDP-11/44 system, and I hope to try running both RT11 and
BSD on it.
The 11/44 has split I&D so you can go with 2.9BSD or 2.11BSD.
Generally speaking,
you'd probably want to go with 2.11BSD because you can.
Would a two-drive system be sufficient, or would I
need more storage space?
I have four RL02 drives in unknown condition, but I was only planning to rack
up two of them with the PDP-11/44 and a Kennedy 9610 9-track tape drive.
My experience is that I was able to fit the base install on the first RL02 pack,
but there wasn't room to pull the sources and rebuild the kernel without more
storage. Even so, 20MB total was kinda tight, maybe too tight. Back when
this hardware was current, it was IIRC easy to plunk 2BSD onto one or two
RK07s (28MB each). When I started playing with simh 15+ years ago, it was
easier to define an RP03 and install into one big 60MB pot (which is how I
helped find a simh bug long ago). You can give it a try now with simh by
practicing with a couple of 10MB image files and the 2BSD distro tape images.
It's an authentic experience in terms of the approach you'd have to take
and the commands you'd have to type (without bad blocks and tape errors
to pull you astray). If it fits on 2x RL02 in simh, it'll be the same on real
hardware. That reminds me - you have to have error-free packs. DEC used
to sell RL02-EF packs for use with UNIX (RT-11 can mark bad spots
on the disk and if they aren't in the first tracks or if there aren't too many,
you can mark those sectors allocated and work around them. Not so with
BSD).
For RT-11, even a single RL02 is plenty of space. I used develop applications
on an 11/23 and an RL01 was enough for the OS and all my code.
My PDP-11/44 pile came with an RL11 board and the
cable from it to the
first RL02 drive. It doesn't have a bulkhead fitting; it just goes from
the controller directly to the connector that plugs into the back of an
RL02 drive. I haven't looked at the cable carefully yet, but I can
examine, measure and photograph it if that would be helpful to anybody
else in this thread.
The bulkhead bracket was convenient but not essential. To get out of the
CPU cabinet, you almost certainly have a flat 40-pin cable, then the
transition connector, then probably the usual drive-to-drive round cable,
but just laying in a line, not bolted to your rack. It will work just fine but
there's a risk of kinking or pinching cables if you move the drives around
a lot (the bracket has a spot to screw down a tab on the shield of the
round cable as a strain relief with either a 6-32 or 4-40 machine screw).
I know I have at least 1-2 bad cables from excessive flex/pinching (I've
thought about cutting them at the point of physical damage and trying
to make one or two of them into a BC80M cable since I have the right
crimper for those Berg pins).
BTW, I am looking for drive ID lenses for my RL02
drives. I need a "1" and
a "2", and I could either buy them outright or trade "0" lenses for
them. My
RL02 pile came with three "0" lenses and a "3".
My recollection is that it was common to have 1-2 drives, but not 3-4 (for
reasons of cost), and as systems were consolidated, there were more "0"
plugs than the others combined.
I've thought about molding replacements or even 3D printing them (they'd
be suitably light-transmissive if made from natural ABS or especially natural
or clear PLA) but I haven't tried making an STL of them. There'd be a risk
of the fingers snapping off and getting stuck in the switch mechanism, but
I think an ABS drive plug that was subjected to the acetone vapor treatment
would firm up nicely and be resistant to the layers separating.
Anyone care to make up an engineering drawing of the plug with tolerances
in the 0.1mm range?
-ethan
3 Jan
3 Jan
2:42 p.m.
On Jan 3, 2014, at 14:09 , Ethan Dicks <ethan.dicks at gmail.com> wrote:
That reminds me - you have to have error-free packs.
DEC used
to sell RL02-EF packs for use with UNIX (RT-11 can mark bad spots
on the disk and if they aren't in the first tracks or if there aren't too many,
you can mark those sectors allocated and work around them. Not so with
BSD).
I guess I'm not running BSD on it any time soon, then. None of my packs are the -EF
ones.
The bulkhead bracket was convenient but not essential.
To get out of the
CPU cabinet, you almost certainly have a flat 40-pin cable, then the
transition connector, then probably the usual drive-to-drive round cable,
but just laying in a line, not bolted to your rack.
I don't recall seeing a separate ribbon cable between the RL11 and the round cable.
I'll need to take a closer look when I get back home.
Anyone care to make up an engineering drawing of the
plug with tolerances
in the 0.1mm range?
I might make a solid model of the plugs sometime. I don't know what the tabs look like
on the 1 and 2 plugs, though. It might be easiest to just make one model of the plug with
all of the tabs present, then cutting away some of them with an razor blade to get the
desired ID. I do not presently have access to a 3D printer. All of my mechanical
experience is with subtractive processes and injection molding.
--
Mark J. Blair, NF6X <nf6x at nf6x.net>
http://www.nf6x.net/
3:50 p.m.
On Fri, Jan 3, 2014 at 5:42 PM, Mark J. Blair <nf6x at nf6x.net> wrote:
On Jan 3, 2014, at 14:09 , Ethan Dicks <ethan.dicks
at gmail.com> wrote:
You don't need -EF packs, but you do need zero bad blocks. Check your
-DC packs for errors. If they have any, use them for RT-11.
DEC merely tested and sold -EF packs that were certified to have zero
bad blocks at the time of sale.
That reminds me - you have to have error-free
packs. DEC used
to sell RL02-EF packs for use with UNIX (RT-11 can mark bad spots
on the disk and if they aren't in the first tracks or if there aren't too many,
you can mark those sectors allocated and work around them. Not so with
BSD).
I guess I'm not running BSD on it any time soon, then. None of my packs are the -EF
ones. I don't recall seeing a separate ribbon cable
between the RL11 and the round cable. I'll need to take a closer look when I get back
home.
I'm not saying you _can't_ fit a round RL cable down through the sides
of an 11/44, but my recollection is the cable ways are kinda tight and
only suitable for a few ribbon cables to make their way out (but I
might be remembering the 11/24 if they are mechanically different).
I might make a solid model of the plugs sometime. I
don't know what the tabs
look like on the 1 and 2 plugs, though.
I would recommend starting with _any_ plug and printing one. to see
how well it works.
It might be easiest to just make one
model of the plug with all of the tabs present, then cutting away some of them
with an razor blade to get the desired ID.
I would not recommend that. Fused filament parts are not as
structurally sound as injection molded parts. You can do some light
sanding and surface work, but major work that cuts through the
"shells" produces weak parts and exposes fill layers. Imagine cutting
a round hole in a chain-link fence and how that affects the integrity
of that panel of the fence.
I do not presently have access to a 3D printer. All of
my mechanical
experience is with subtractive processes and injection molding.
I run 3D printing activites at our local hackerspace (including
monthly meetups) but I am not a 3D designer. I work on the printer
end, not the drawing end.
-ethan
4:57 p.m.
On Jan 3, 2014, at 15:50 , Ethan Dicks <ethan.dicks at gmail.com> wrote:
I'm not saying you _can't_ fit a round RL
cable down through the sides
of an 11/44, but my recollection is the cable ways are kinda tight and
only suitable for a few ribbon cables to make their way out (but I
might be remembering the 11/24 if they are mechanically different).
I took a closer look at the cable just now. It's round all the way to the controller
card, where the wires break out to a double row header with individually crimped terminals
and a strain relief clamp. The drive end has a cable clamp that looks like it might
connect to a shield. Just by eyeball, I'd call it about a 10 foot long cable. The
drive connector has a black plastic shell, rather than a metal shell like on my daisy
chain cables and terminators. I didn't see part numbers on the cable. My drives are
all missing the sheet metal piece that screws to the back panel near the connectors so
that the cables can be clamped to it, but a suitable replacement shouldn't be too hard
to kludge together. I will need to buy or kludge rack slides for at least a couple of my
drives.
I would not recommend that. Fused filament parts are
not as
structurally sound as injection molded parts. You can do some light
sanding and surface work, but major work that cuts through the
"shells" produces weak parts and exposes fill layers. Imagine cutting
a round hole in a chain-link fence and how that affects the integrity
of that panel of the fence.
Thanks for the advice. I haven't had my hands on filament printed parts yet, so I
don't know anything about their properties. Injection molded parts would sure be
nicer, but the tooling to make them would cost a few thousand dollars.
--
Mark J. Blair, NF6X <nf6x at nf6x.net>
http://www.nf6x.net/
8:05 p.m.
On Fri, Jan 3, 2014 at 7:57 PM, Mark J. Blair <nf6x at nf6x.net> wrote:
They are strong in 2 dimensions, and laminated in the third, so thin structures
like the switch cams are easy to fabricate, but not strong against
lateral stress.
The walls of any fused filament part are typically 1-3 concentric layers with a
random amount (user's choice) of fill. 25% is typical for parts that don't take
a lot of stress, 70%-90% for things like gears and extruder housings.
I'd print the drive plugs face-down so the user-visible portion is
smoothed by the heated bed (typically kapton-coated glass). The body
can be hollow, of course, and the switch cams stick up as it prints.
Put the whole thing in an acetone vapor bath to smooth the sides
(ridged from the layering) and to fuse the layers together and you'll
have a nice part. With a reasonably accurate mechanical drawing, it
shouldn't be too hard to make a solid model with OpenSCAD and tweak
the model to select the unit cams and once the shape is verified (I'm
guessing it will take a test print or two and minor dimensional
tweaks) it should be easy to print a set of 4 at once (or multiples of
the same unit if you needed them). I'd print multiples at a time to
let the cams cool (tall thin things need that extra time).
On Jan 3, 2014, at 15:50 , Ethan Dicks <ethan.dicks at gmail.com> wrote:
Sounds like you _can_ snake a round cable down into an 11/44. You
appear to be describing a BC80M or BC80J cable.
I'm not saying you _can't_ fit a round RL
cable down through the sides
of an 11/44...
I took a closer look at the cable just now. It's round all the way to the
controller card, where the wires break out to a double row header with
individually crimped terminals and a strain relief clamp. The drive end
has a cable clamp that looks like it might connect to a shield. Just by eyeball, I'd call it about a 10 foot long
cable.
Probably a good guess. It's likely to be either 6' or 10'. Total length
from controller to furthest drive has to be under
100'. Never seen
anyone get anywhere close.
The drive connector has a black plastic shell, rather
than a metal
shell like on my daisy chain cables and terminators.
That's common. Probably pre-FCC.
I didn't see part numbers on the cable.
It might be by the controller end or a sticker on the Berg connector.
My drives are all missing the sheet metal piece that
screws to the back
panel near the connectors so that the cables can be clamped to it, but
a suitable replacement shouldn't be too hard to kludge together.
It's just a Z-shaped bracket with some screw holes for the drive and
some rivnuts for the cables. Easy to kludge.
I will need to buy or kludge rack slides for at least
a couple of my drives.
Those are a little more odd. You should have runners riveted to the sides
of the drives, and those fit onto slide rails for the RL01 and RL02. Those
routinely get tossed with racks when nobody wants to figure out how to
unbolt them (it takes a few minutes). You are probably better off fabricating
shelves than rails if you can't find inexpensive surplus rails.
I would not
recommend that. Fused filament parts are not as
structurally sound as injection molded parts...
Thanks for the advice. I haven't had my hands on filament printed parts
yet, so I don't know anything about their properties. Injection molded parts would sure be nicer, but the
tooling to make them would cost a few thousand dollars.
Yep. It's not worth spinning up a full run, but at ~$10/hr for the machine or a
few dollars per cc of plastic, it might be worth fiddling some 3D shapes.
-ethan
9:02 p.m.
On Jan 3, 2014, at 20:05 , Ethan Dicks <ethan.dicks at gmail.com> wrote:
Sounds like you _can_ snake a round cable down into an
11/44. You
appear to be describing a BC80M or BC80J cable.
I put some pictures of my heap of RL02 drives up on my web page. The last four pictures
show the cable:
http://www.nf6x.net/2014/01/dec-rl02-drives/
--
Mark J. Blair, NF6X <nf6x at nf6x.net>
http://www.nf6x.net/
4:59 p.m.
On Jan 3, 2014, at 15:50 , Ethan Dicks <ethan.dicks at gmail.com> wrote:
You don't need -EF packs, but you do need zero bad
blocks. Check your
-DC packs for errors. If they have any, use them for RT-11.
How likely am I to find packs nowadays with no bad blocks? Maybe I should leave room in
the rack for some other kind of drive in addition to the RL02 units.
--
Mark J. Blair, NF6X <nf6x at nf6x.net>
http://www.nf6x.net/
4 Jan
4 Jan
5:56 a.m.
On Jan 4, 2014, at 5:31 AM, "Mark J. Blair"
<nf6x at nf6x.net> wrote:
I have a dozen RL02-DC packs that I recently initialized under RT-11. They all came up
with no bad blocks at all, both with FORMAT /VERIFY:ONLY and DIR/BAD.
So, I would venture to say that the likelihood of finding error free -DC packs is high.
On Jan 3, 2014, at 15:50 , Ethan Dicks
<ethan.dicks at gmail.com> wrote:
You don't need -EF packs, but you do need zero bad blocks. Check your
-DC packs for errors. If they have any, use them for RT-11.
How likely am I to find packs nowadays with no bad blocks? Maybe I should leave room in
the rack for some other kind of drive in addition to the RL02 units.
--
Mark J. Blair, NF6X <nf6x at nf6x.net>
http://www.nf6x.net/
7:06 a.m.
On Fri, 3 Jan 2014 14:42:53 -0800
"Mark J. Blair" <nf6x at nf6x.net> wrote:
On Jan 3, 2014, at 14:09 , Ethan Dicks <ethan.dicks at gmail.com> wrote:
--snip--
Of the 35 or so RL02 Packs I have, only a couple are EF qualified.
I cleaned and tested every pack when I obtained them - and only two or
three have ANY bad sectors. Those that do, all have "correctable"
re-assignments (both RT and RSX can support this capability).
I have had no problems running BSD using error free "non-EF" (DC)
packs...
Lyle
--
Bickley Consulting West Inc.
http://bickleywest.com
"Black holes are where God is dividing by zero"
That reminds me - you have to have error-free
packs. DEC used
to sell RL02-EF packs for use with UNIX (RT-11 can mark bad spots
on the disk and if they aren't in the first tracks or if there
aren't too many, you can mark those sectors allocated and work
around them. Not so with BSD).
I guess I'm not running BSD on it any time soon, then. None of my
packs are the -EF ones.
4007
days inactive
4008
days old
8 comments
4 participants
participants (4)
-
ethan.dicks＠gmail.com -
lbickley＠bickleywest.com -
nf6x＠nf6x.net -
roe＠liveblockauctions.com