13 Aug
2006
13 Aug
'06
10:05 p.m.
On 8/14/06, Peter C. Wallace <pcw at mesanet.com> wrote:
I'd build one, especially if it was Negibus-compatible. I could
attach a real DF-32 to it. ;-)
Has anyone estimated the physical size? A real PDP-8/S has about 1000
transistors and a lot more passives than that, not counting external
things like a TTY interface (besides serializing the ALU, that's one
of the ways they shoehorned Straight-8 technology into such a tiny
box).
One could always start small and build something simpler from the same
technology - a simple clock or something - sort of a test-drive to see
what the scope of a full processor would entail.
-ethan
One thing that
I think *would* be awesome in surface mount,
would be to do a "scale model" based on the straight-8, with
all those little SMT diodes and transistors, and no logic
IC's at all (except possibly some SMT RAM to replace the core
planes) :-).
Vince
(It might also be cool to make some of the diodes the glowing kind.)
That has been discussed before. It would be fun to do something like the
PDP8-S on a single PCB with only resistors, capacitors, SOT-23 transistors and
dual diodes (And yes LEDs in all the FF outputs would be nice). 
13 Aug
13 Aug
10:50 p.m.
Has anyone estimated the physical size? A real
PDP-8/S has about 1000
transistors and a lot more passives than that, not counting external
things like a TTY interface (besides serializing the ALU, that's one
of the ways they shoehorned Straight-8 technology into such a tiny
box).
My PSP-8/s is one of the oddball hybrid machines, and if DEC did a
major mechanical redesign, I bet they could have matched what we could
do with SMT. The hybrid Flip Chips have acres of bare PCB real estate.
--
Will
14 Aug
14 Aug
12:45 a.m.
From: "Ethan Dicks" <ethan.dicks at gmail.com>
That would be quite a conversation piece :-)!
Naturally, it should be Negibus compatible!
> That has been discussed before. It would be fun
to do something like the
> PDP8-S on a single PCB with only resistors, capacitors, SOT-23
transistors
and
dual diodes
(And yes LEDs in all the FF outputs would be nice).
I'd build one, especially if it was Negibus-compatible. I could
attach a real DF-32 to it. ;-) Has anyone estimated the physical size? A real
PDP-8/S has about 1000
transistors and a lot more passives than that, not counting external
things like a TTY interface (besides serializing the ALU, that's one
of the ways they shoehorned Straight-8 technology into such a tiny
box).
I'd guess something like 1/4 scale would be acheivable. If you make it
too small, you'll have trouble with the user interface (as you mentioned
about the 1/12th scale IMSAI).
One could always start small and build something
simpler from the same
technology - a simple clock or something - sort of a test-drive to see
what the scope of a full processor would entail.
I'd start out by building replacements for the Rxxx and Sxxx modules, all
scaled down. (Once a few of those work, you could start in on a backplane
for them.)
I was thinking of a 9x2 header/socket arrangement for the interconnect
and soldering the header to pads on each side of the board to make little
1" wide modules. (Were there double-sided Rxxx or Sxxx modules? I don't
think so, because the connector blocks used to be one sided.) Then you
could make the modules the right shape, and route the "handles" right out
of the PCB material :-). The result would be about 1"x2.25".
For testing, you just need a standard size module with a female 9x2 right
angle connector on it, and insert the miniature card being tested. Replace
an original module in some old gear with the result, to see if you've got
things right. You could also use the technology to resurrect old gear with
missing modules :-).
That's only a little better than half-scale, though. One quarter scale
would
need finer pitched connectors, which would probably be way more expensive.
Another goofy idea I had (for inexpensive interconnect) was to try to
squeeze
the SMT modules down into the footprint of, say, an 18 or 20 pin DIP. It
wouldn't look much like a replica anymore, though.
Vince
12:51 a.m.
On 8/13/06, vrs <vrs at msn.com> wrote:
the SMT modules down into the footprint of, say, an 18
or 20 pin DIP. It
wouldn't look much like a replica anymore, though.
How about an 18 or 20 pin SIP? Tricky to pull and replace repeatedly,
but you'd get the stand-out module look you're trying to achieve.
My $0.002,
Josef
1:17 a.m.
On 8/14/06, Josef Chessor <josefcub at gmail.com> wrote:
On 8/13/06, vrs <vrs at msn.com> wrote:
And it would be *substantially* cheaper than dozens of 0.1" headers and sockets.
-ethan
the SMT modules down into the footprint of, say,
an 18 or 20 pin DIP. It
wouldn't look much like a replica anymore, though.
How about an 18 or 20 pin SIP? Tricky to pull and replace repeatedly,
but you'd get the stand-out module look you're trying to achieve.
2:11 a.m.
From: "Ethan Dicks" <ethan.dicks at gmail.com>
And it would be *substantially* cheaper than dozens of 0.1" headers and
sockets.
If I'm understanding you right, it would still be 0.1" spacing, and so still
nearly two inches long. The original cards were only a little under 3.5",
so
I'm not sure the SIP would give the overall size reduction I was looking
for.
I'm also not too clear why that would be substantially cheaper. Is it
because
the header pins are gold plated?
Vince
On 8/14/06, Josef Chessor <josefcub at
gmail.com> wrote:
> On 8/13/06, vrs <vrs at msn.com> wrote:
> > the SMT modules down into the footprint of, say, an 18 or 20 pin DIP.
It
wouldn't look much like a replica anymore, though.
How about an 18 or 20 pin SIP? Tricky to pull and replace repeatedly,
but you'd get the stand-out module look you're trying to achieve.
1:15 a.m.
On 8/14/06, vrs <vrs at msn.com> wrote:
Sure... but the front panel (in this case) doesn't have anything to do
with the size of the CPU board. In the case of the -8/S, the lights
panel is one board (and plugs into the second row) and the switch
panel is another board (and plugs into the first row).
Has anyone
estimated the physical size?
I'd guess something like 1/4 scale would be acheivable. If you make it
too small, you'll have trouble with the user interface (as you mentioned
about the 1/12th scale IMSAI). I'd start out by building replacements for the
Rxxx and Sxxx modules, all
scaled down. (Once a few of those work, you could start in on a backplane
for them.)
Would it be worth the expense of a backplane? I would have thought
that the way to do it would be to lay it out on both sides of single,
largish PCB with off-board connections for I/O and front panel.
I was thinking of a 9x2 header/socket arrangement for
the interconnect
and soldering the header to pads on each side of the board to make little
1" wide modules.
While the originals R and S modules were 18-pin, I'd think that 20-pin
headers, etc, would be easier to find in quantity (not that they
couldn't be trimmed down, of course).
(Were there double-sided Rxxx or Sxxx modules? I
don't
think so, because the connector blocks used to be one sided.)
No. The module blocks were single-sided only.
For testing, you just need a standard size module with
a female 9x2 right
angle connector on it, and insert the miniature card being tested. Replace
an original module in some old gear with the result, to see if you've got
things right. You could also use the technology to resurrect old gear with
missing modules :-).
That's certainly true, but in some cases, they hand-picked modules to
get the timing right.
Another goofy idea I had (for inexpensive
interconnect) was to try to
squeeze
the SMT modules down into the footprint of, say, an 18 or 20 pin DIP. It
wouldn't look much like a replica anymore, though.
I would think the point would be a work-alike replica more than a
board-for-board form-factor-in-miniature replica, but whatever works.
One thing that would help for size is that about 20% of the backplane
size (40x4? 44x4?) is for the 4K core stack. Obviously, that would
end up being a couple of 6264s or something like it, and take up
*substantially* less space than the original. The front 3 rows (a bit
under 10%) or so are the front panel and lamp drivers, and the back
few rows are I/O drivers and cable attachments (10%? 15%?) That
leaves a bit over 50% for the CPU itself, or about 100 R and S
modules, as a rough estimate. I would think that the individual PCB
costs, even for something about 2 sq in., and the interconnect
hardware would start to make this a rather expensive project. It
would, of course, be easier to _test_ dozens of individual modules
than a monolithic CPU PCB (since there wouldn't be any interaction
issues), but quite a bit pricier. OTOH, even with a $1000 budget,
it's cheaper than the going rate for a PDP-8/S these days.
-ethan
2:23 a.m.
From: "Ethan Dicks" <ethan.dicks at gmail.com>
That would be a lot cheaper, but would sacrifice some of the "vintage"
aesthetics :-).
Sure. One could use keyed 20 pin receptacles, and 18 actual pins.
Sure, and depending on how closely one followed the old schematics, that
might
still be needed :-).
One would certainly have to pay extra for the look-alike over the
work-alike.
On 8/14/06, vrs <vrs at msn.com> wrote:
> I'd start out by building replacements for the Rxxx and Sxxx modules,
all
> scaled down. (Once a few of those work, you
could start in on a
backplane
for them.)
Would it be worth the expense of a backplane? I would have thought
that the way to do it would be to lay it out on both sides of single,
largish PCB with off-board connections for I/O and front panel. > I was thinking of a 9x2 header/socket arrangement
for the interconnect
> and soldering the header to pads on each side of the board to make
little
1" wide
modules.
While the originals R and S modules were 18-pin, I'd think that 20-pin
headers, etc, would be easier to find in quantity (not that they
couldn't be trimmed down, of course). > For testing, you just need a standard size module
with a female 9x2
right
> angle connector on it, and insert the miniature
card being tested.
Replace
> an original module in some old gear with the
result, to see if you've
got
> things right. You could also use the technology
to resurrect old gear
with
missing
modules :-).
That's certainly true, but in some cases, they hand-picked modules to
get the timing right. > Another goofy idea I had (for inexpensive
interconnect) was to try to
> squeeze
> the SMT modules down into the footprint of, say, an 18 or 20 pin DIP.
It
wouldn't
look much like a replica anymore, though.
I would think the point would be a work-alike replica more than a
board-for-board form-factor-in-miniature replica, but whatever works. I would think that the individual PCB
costs, even for something about 2 sq in., and the interconnect
hardware would start to make this a rather expensive project.
I know a couple of places that will do panels inexpensively without
quibbling
about the number of designs on them. So I'd recommend panelizing about
8"x10"
of the things and produce them 30 at a time or so :-).
It
would, of course, be easier to _test_ dozens of individual modules
than a monolithic CPU PCB (since there wouldn't be any interaction
issues), but quite a bit pricier. OTOH, even with a $1000 budget,
it's cheaper than the going rate for a PDP-8/S these days.
Agreed.
Vince
1:24 a.m.
On Sun, 13 Aug 2006, vrs wrote:
I'm guestimating that once these miniaturized Rxxx and Sxxx modules are
designed, manufacturing them in the quantities required for even a single
Straight-8 would be very cheap. Now about the backplane: Is it possible
to make it with a PCB, or is wire-wrap absolutely required?
I'd guess something like 1/4 scale would be
acheivable. If you make it
too small, you'll have trouble with the user interface (as you mentioned
about the 1/12th scale IMSAI).
I agree with the quarter scale. Beyond that and one would need tweezers
or hemostats to handle the flipchips.
One could
always start small and build something simpler from the same
technology - a simple clock or something - sort of a test-drive to see
what the scope of a full processor would entail.
I'd start out by building replacements for the Rxxx and Sxxx modules, all
scaled down. (Once a few of those work, you could start in on a backplane
for them.) For testing, you just need a standard size module with
a female 9x2 right
angle connector on it, and insert the miniature card being tested. Replace
an original module in some old gear with the result, to see if you've got
things right. You could also use the technology to resurrect old gear with
missing modules :-).
That's only a little better than half-scale, though. One quarter scale
would need finer pitched connectors, which would probably be way more
expensive.
It would cost more, but I don't think it would be by very much. How about
having three rows of .1" connectors?
Another goofy idea I had (for inexpensive
interconnect) was to try to
squeeze the SMT modules down into the footprint of, say, an 18 or 20 pin
DIP. It wouldn't look much like a replica anymore, though.
Nah. From a casual browse of my catalogs, it'll cost you around 75 cents
per card/backplane connection. Doing the DIP thing would probably end up
being more than that.
--
David Griffith
dgriffi at cs.csubak.edu
A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
1:31 a.m.
On 8/14/06, David Griffith <dgriffi at cs.csubak.edu> wrote:
I'm guestimating that once these miniaturized Rxxx
and Sxxx modules are
designed, manufacturing them in the quantities required for even a single
Straight-8 would be very cheap. Now about the backplane: Is it possible
to make it with a PCB, or is wire-wrap absolutely required?
On a related note, but a separate question, does anyone actually have
the wiring schedule for a PDP-8/S backplane? A long time ago, I was
willing to start constructing flip chip modules that made up a PDP-8,
but I could never find the wiring schedule for the backplane online so
that I could truly build the computer.
(I'm great at assembly[sic], horrible at design).
Josef
1:48 a.m.
Josef Chessor wrote:
On a related note, but a separate question, does
anyone actually have
the wiring schedule for a PDP-8/S backplane? A long time ago, I was
willing to start constructing flip chip modules that made up a PDP-8,
but I could never find the wiring schedule for the backplane online so
that I could truly build the computer.
I think if you are building a replica is to design your own Flip/Chips
and build a clean design from scratch. A PDP-8/S does not look that
impressive, I know I seen one in the flesh once,as compared to a
PDP-8/e. If you go with 20 I/O pins, (and sneaky pcb pinout )a simple
GAL (SMT) could make up a generic card and be re-programmed for
what ever logic you want.
2:04 a.m.
On 8/14/06, woodelf <bfranchuk at jetnet.ab.ca> wrote:
I think if you are building a replica is to design
your own Flip/Chips
and build a clean design from scratch. A PDP-8/S does not look that
impressive, I know I seen one in the flesh once,as compared to a
PDP-8/e. If you go with 20 I/O pins, (and sneaky pcb pinout )a simple
GAL (SMT) could make up a generic card and be re-programmed for
what ever logic you want.
Yes, but in the case of a Straight-8 or PDP-8/S, the R-series and
S-series logic is utterly incompatible with TTL. The supply voltages
are +10V and -15V, and asserted logic is +3V or -3V, thus the reason
for staying with discrete SMT transistors.
The SMT GAL design could be useful for building an M-series machine
(-8/L, -8/i), but with that era of machine, the FLIP CHIP modules are
double sided, with 36 pins (32 signal pins and up to 4 power/ground
pins).
-ethan
2:49 a.m.
From: "woodelf" <bfranchuk at jetnet.ab.ca>
I think if you are building a replica is to design
your own Flip/Chips
and build a clean design from scratch. A PDP-8/S does not look that
impressive, I know I seen one in the flesh once,as compared to a
PDP-8/e. If you go with 20 I/O pins, (and sneaky pcb pinout )a simple
GAL (SMT) could make up a generic card and be re-programmed for
what ever logic you want.
Well, that would be a cool project, but I'd say you were designing a
new PDP-8, and would be troubled by calling it a replica.
I guess if the big-board version (that shares only the schematic) can
be a called "replica", then so can one that shares the aesthetics but
not the schematic. To my mind, a replica should really have both (as
much as is feasible).
Vince
2:42 a.m.
From: "Josef Chessor" <josefcub at gmail.com>
On a related note, but a separate question, does
anyone actually have
the wiring schedule for a PDP-8/S backplane? A long time ago, I was
willing to start constructing flip chip modules that made up a PDP-8,
but I could never find the wiring schedule for the backplane online so
that I could truly build the computer.
(I'm great at assembly[sic], horrible at design).
AFAIK, no. The easiest way I know to recreate it is to simply redraw the
schematic in question. I have libraries for a lot of M-series modules,
the relevant connectors, etc. It wouldn't be that hard to create an R/S
series library, and then use that to redraw the 8/S or straight-8. It
is many hours of somewhat tedious work, though. (I did it for the TC08.)
Once it is redrawn, one can export the netlist, and I even have some Perl
to convert that into wire lists suitable for wire-wrap :-).
Vince
2:35 a.m.
From: "David Griffith" <dgriffi at cs.csubak.edu>
I have done layout for a board that implements the wire list for the TC08.
(That's a lot simpler than a straight-8, but might compare with the 8/S.)
It required 4 layers to route -- wasn't even close in 2 layers. The wire
wrap allowed an essentially unlimited number of wire crossings :-).
That was also with DEC connector block spacing. If you make the connections
closer together, you worsen the congestion.
So the backplane (which also wants to be physically large) ends up being a
really expensive (and nearly impossible to fix) PCB. Might be better to use
wire-wrap (or maybe space the connectors way apart and try for 2 layers) for
the prototype :-).
That would get messy to connect to the board edge, and I'm not sure you'd
find the receptacles affordable?
Vince
On Sun, 13 Aug 2006, vrs wrote:
> I'd start out by building replacements for the Rxxx and Sxxx modules,
all
> scaled down. (Once a few of those work, you
could start in on a
backplane
for them.)
I'm guestimating that once these miniaturized Rxxx and Sxxx modules are
designed, manufacturing them in the quantities required for even a single
Straight-8 would be very cheap. Now about the backplane: Is it possible
to make it with a PCB, or is wire-wrap absolutely required? That's
only a little better than half-scale, though. One quarter scale
would need finer pitched connectors, which would probably be way more
expensive.
It would cost more, but I don't think it would be by very much. How about
having three rows of .1" connectors?
6:12 p.m.
On 8/14/06, vrs <vrs at msn.com> wrote:
So the backplane (which also wants to be physically
large) ends up being a
really expensive (and nearly impossible to fix) PCB. Might be better to use
wire-wrap (or maybe space the connectors way apart and try for 2 layers) for
the prototype :-).
How about what DEC did for later backplanes... 2-layer board plus
wire-wrap? You could do the easy stuff, power connections, etc., in
copper, then do some of the congestion relieving stuff in the
wire-wrap layer. That many wire-wrap sockets (of whatever style)
wouldn't be cheap, so I doubt it would save much money over a 4-layer
backplane, but you'd make part of the backplane circuit accessible.
-ethan
7:11 p.m.
From: "Ethan Dicks" <ethan.dicks at gmail.com>
That'd probably work pretty well. One could also bias things so that the
parts one was sure of went into the etch, and anything that looked one
wasn't sure they got right could go in the wrap.
Vince
On 8/14/06, vrs <vrs at msn.com> wrote:
> So the backplane (which also wants to be physically large) ends up being
a
> really expensive (and nearly impossible to fix)
PCB. Might be better to
use
> wire-wrap (or maybe space the connectors way
apart and try for 2 layers)
for
the prototype
:-).
How about what DEC did for later backplanes... 2-layer board plus
wire-wrap? You could do the easy stuff, power connections, etc., in
copper, then do some of the congestion relieving stuff in the
wire-wrap layer. That many wire-wrap sockets (of whatever style)
wouldn't be cheap, so I doubt it would save much money over a 4-layer
backplane, but you'd make part of the backplane circuit accessible. 
9:51 a.m.
vrs wrote, in part:
I'd guess something like 1/4 scale would be
acheivable. If you make it
too small, you'll have trouble with the user interface (as you mentioned
about the 1/12th scale IMSAI).
I'd suggest going slightly smaller: 1/6 scale. Then G.I. Joe could use it,
not to mention Barbie. :)
(It'd be the Ultimate Collectible on eBay... <g>)
-O.-
11:28 a.m.
From: "O. Sharp" <ohh at drizzle.com>
vrs wrote, in part:
My wife would be just the collector for it :-). She's not too big on Barbie
anymore, though. (Seems to be into "Fashion Royalty", which are
approximately
the same size.)
Vince
I'd guess something like 1/4 scale would be
acheivable. If you make it
too small, you'll have trouble with the user interface (as you mentioned
about the 1/12th scale IMSAI).
I'd suggest going slightly smaller: 1/6 scale. Then G.I. Joe could use it,
not to mention Barbie. :)
(It'd be the Ultimate Collectible on eBay... <g>)
6558
days inactive
6558
days old
18 comments
7 participants
participants (7)
-
bfranchuk＠jetnet.ab.ca -
dgriffi＠cs.csubak.edu -
ethan.dicks＠gmail.com -
josefcub＠gmail.com -
ohh＠drizzle.com -
vrs＠msn.com -
wdonzelli＠gmail.com