Tom,
Go to the Geckodrive yahoo group and have a look at the pictures in
their file section. You can see what a "pin based" MB and cards setup
looks like... (You may have to join the list to view the
files/photos, but it'll be worth it, IMO.)
Hope this helps,
Ballendo
--- In
Homebrew_PCBs@yahoogroups.com, "Thomas P. Gootee" <tomg@f...>
wrote:
> === Roy,
>
> === THANKS for the excellent reply!
>
> === MY current responses are intermingled, BELOW, and marked
with "==="s .
>
> === - Tom Gootee
>
>
> Date: Fri, 20 Aug 2004 16:25:24 -0400
> From: "Roy J. Tellason" <rtellason@b...>
> Subject: Re: RE: making vias (was: Re: Plating thruholes.)
>
> I'm just gonna toss out a few thoughts coming to the fore, based
on me
> working on all sorts of stuff over several decades, mostly
repairs...
>
> === Great! "Repairs" usually tell the true tale...
>
> On Friday 20 August 2004 03:41 pm, Thomas P. Gootee wrote:
>
> > There are several possibilities that I have considered:
> >
> > 1) pin headers on each pcb, with ribbon cables running to similar
headers
> > on the front panel's new pcb, or
>
> Pin headers are good reliable connectors, particularly if shrouded
connectors
> are used, and if they're also using ejector levers there isn't
much stress
> on the cabling.
>
> > 2) card-edge connectors on the front panel pcb, along the bottom,
that each
> > of the three pcboards' edges could plug into, or
>
> I think less well of card-edge connectors. You'll have problems
with tarnish
> on the card, and lack of reliability eventually with the
connectors over
> time. Notice that card-edge was the way to go with the old 5.25"
and 8"
> floppy drives, and early hard drives as well. Nowadays _all_ 3.5"
floppy
> drives and all IDE and SCSI hard drives use pin connectors. That
should tell
> you something. :-)
>
> === Point taken. But all of the PCs that I've seen still use card-
edge connectors
> === for all of the CARDS, i.e. the PCBs, in the expansion slots
(e.g. ISA,
> === EISA, PCI, etc). The "slots" are just card-edge connectors. I
have
> === some that are over 15 years old that still work fine. Also, I
check out
> === and/or service or refurbish a lot of old Tektronix and other
test equipment,
> === much of which uses card-edge connectors (e.g. Tek 7000 series
scope
> === plug-ins and Tek TM500-series plug-ins). And those are heavily
used, i.e.
> === they typically undergo MANY mating/unmating cycles over their
lifetimes.
> === Occasionally there is a mechanical alignment problem, or
somesuch, that
> === causes problems. But not "too" many! And a lot of that stuff
is at least 20 or
> === 30 years old. (Then again, maybe all of the mating/unmating
cycles HELPED
> === keep the card-edge connectors in better shape (by scraping the
surfaces).
>
> > 3) right-angle headers or sockets on each pcb, that would plug
directly
> > into straight headers or sockets on the front panel pcb (pin
headers and
> > sockets, or DIN, or even D-Sub), with NO CABLING necessary, or
>
> Pin headers at both ends of the connection is workable, provided
you can
> arrange the layout of stuff to accomodate this without too much
trouble, but
> it's gonna be a real hassle to design, compared to using cables.
>
> === Well, the boards are ALREADY sitting at right angles to the
front panel,
> === with their edges just about in the right positions. And I
guess the on-board
> === part of the re-design would be about the same as if I were
using cables.
> === It seems like eliminating anything, especially %$#! cables,
would be
> === "a ∗GOOD∗ thing".
> ===
> === But HEY! Couldn't I use the backplane/buss idea, but with pin
headers and
> === sockets, instead of card-edge connectors? I just REALLY like
the idea of
> === using PCBs instead of wires and cables!
>
> > 4) another new board, a "motherboard", in the bottom of the
enclosure, with
> > card-edge sockets, along with a total redesign of the current
boards'
> > layouts, so they could be mounted vertically, in the MB's
sockets, and a
> > new front panel pcb that would also plug into a card-edge socket
on the new
> > motherboard, or
>
> Putting aside card-edge for a minute (see above :-), what you
suggest here is
> called a backplane. It's been used in a few computers, and in
other
> equipment, both in instances where all connectors were the same,
and all
> signals were bussed (see the early Zenith Data Systems XT-class
machines for
> example) and in earlier stuff where each connector was different
and you
> could only plug one board in to one particular slot (see "Digital
Group"
> early pre-PC computers for an example of this). The biggest
problem I see
> with this is that you don't want all signals bussed because there
are likely
> to be different signals used on each board, and the other way
means that any
> changes makes for a redesign needed and difficulty in upgrades,
etc. should
> anything change in the future. Personally I'd avoid this.
>
> === Well, in general, you're probably right. But, in THIS case, I
think it would
> === be great. Basically, the one main board and the new front
panel board have
> === LOTS of things that they need to share, like multiple 2P6T
switches' wiring, and
> === a 1P12T switch's wiring, and a couple of 4P3T switches' wiring,
and various
> === other controls' wiring, plus signals in and out, etc, and, of
course, the power
> === supply rails, and all of the separate ground paths, and
probably some other stuff
> === that I'm forgetting to mention. Then there's the power
amplifier board, which
> === only shares the wiring to one side of one of the same 2P6T
switches, plus a
> === couple of i/o signal paths, and, of course, the power supply
rails and some
> === ground paths. The third and last original board is just the
power supply. And
> === all of the boards need to get power rails and separate ground
paths, etc, from
> === the power supply board. I just can't see ∗anything∗ wrong with
putting them all
> === on a nice big buss. Of course, the buss would have quite a few
spare positions,
> === for future additions. I'd probably also have at least one
spare board position,
> === for the same reason (just so I wouldn't have to change it, if
another board was
> === ever added).
> ===
> === Am I missing something?? Or is my thinking wrong, about that,
somehow??
> ===
> === I ∗DID∗ find some sources for fairly-inexpensive pcb-mount pin
headers and
> === sockets, as well as pcb-mount card-edge connectors. So the cost
of all of the
> === connectors, either pin-type or card-edge, for use on a parallel-
buss pcb with
> === 62 or more traces, and six card positions (I would only ∗need∗
5, assuming my
> === main board is split into two boards so it would fit vertically
into the same cabinet,
> === which is only about 3 inches high, and assuming I added a new
front panel pcb),
> === could be in a range as low as $10.00 (for 100 quantities),
especially if I used
> === TWO headers and sockets that were each HALF the size I need for
the main
> === board and the front
> === panel board, because then I could just use ONE of each for the
two "little" boards.
> ===
> === (Of course, even $10 is a ∗significant∗ percentage-increase in
the total parts cost
> === per unit. But, heck, I can see RIGHT OFF THE BAT that it would
save WAY more
> ===than $10, in assembly labor costs alone (and ∗maybe∗ some in
troubleshooting
> === labor, etc, too).
> ===
> === That would, technically, make it into two busses, I guess. But
the "backplane"
> === board would be made so that the "extra" sockets could be added
later, for the
> === other card positions, if it ever became necessary, without
changing the
> === backplane board's layout.
>
> > 5) card-edge sockets on all the pcbs, with small pcbs that plug
into them,
> > with discrete wires or ribbon cables soldered to the small pcbs
(i.e.
> > hand-made custom card-edge-connector cabling, hehehe...),
>
> See above comments about card-edge connectors. This is also much
more
> labor-intensive assembly and more places for things to go wrong.
>
> === Yeah. That option was just mentioned "for the sake of
completeness".
>
> > 6) pcb-mount terminal blocks with discrete wires, or
>
> <shrug>
>
> Terminal blocks are okay and I'd go that way if I were dealing with
wiring
> that had to carry nontrivial levels of power (say more than a
couple of
> hundred mA or so) but for the most part you won't need it. Lots of
potential
> for error here, too.
>
> === Yes. The error potential is one of the main possible problems
that I see
> === with using any kind of single-wire connections, socketed or
not, although
> === if they're socketed/removable, the errors WOULD be much easier
to correct,
> === assuming they could be found...
> ===
> === BUT, there ARE "non-trivial" levels of current, in a few
places. The
> === power amplifier board has to be able to push up to 1.5 Amps,
through the
> === front panel connectors and on to the device under test (DUT).
And the
> === power supply board, obviously, has to supply that juice. In the
current design,
> === those high-current signals also pass through the main curve
tracer board,
> === so they can have their currents and voltages sensed and fed to
the
> === instrumentation amplifiers, etc, that then produce the x-y
outputs to the
> === scope display. But, with the new front panel PCB, ∗probably∗
only the
> === sensing-lines will need to go back to the main board.
>
> > 7) something similar to #1, but with some as-yet-unknown (to me)
connector
> > and/or cabling types (I even considered pcb-mountable modular
phone jacks
> > and cabling).
>
> Aside from a limited number of conductors these are reasonable,
and fairly
> cheap. Not terribly robust mechanically, but easy enough to crimp
those
> connectors on. Unless you need to start crossing wires, etc.
They're
> probably cheaper than pin connectors, but other than that I don't
see that
> much of an advantage to going with them.
>
> === Yes. Easy to crimp. And cheap. Might've been perfect as a
cheap, easy
> === source for short 6-wire "jumpers". But, besides being a
little "weird" in an
> === application like this one, it turns out that their connectors
would just be
> === way too BIG to have any hope of fitting onto the boards, as
they are now.
>
> > Which of those (probably out of #1 through #4) sound "good"?
>
> One thing you might consider in terms of using pin connectors is
the
> possibility of using off-the-shelf cables, such as 34 wires ("IDE
cable") or
> 50 wires ("SCSI cable"), etc.
>
> === Definitely! See farther below.
>
> > One immediate "problem" I can foresee, with just adding a second
side to
> > each pcb and then running traces (on the "new" side of each
board) to pin
> > headers: Our pcb-making process is rudimentary and does NOT
include the
> > ability to make plated-through holes. SO, to have headers that are
> > connected to the new top side traces, while the connectors are
also sitting
> > on the top sides of the boards, I wouldn't be able to solder the
top side's
> > trace directly to the headers' pins, since the headers sit right
on the
> > boards. So I'd have to make an extra row of holes next to each
header row,
> > that could have pins inserted that could be soldered on both
sides. (I was
> > thinking I might just use a single-row header, upside down with
the longer
> > pins halfway into the holes. It might be "ugly". But it would
also provide
> > handy test-points...).
>
> Test points are good. But you might also consider using a
connector that has
> way more pins than you need and using a bunch of them for grounds.
This is
> what's typically the case with a lot of PC internal interconnects,
and it
> makes for better performance in terms of crosstalk etc. between the
wires.
> Though I don't know how much of a problem that's likely to be in
the
> equipment you're talking about.
>
> === Good point. There ARE some signals for which I'm planning to
do that.
>
> > Many of the switches in the unit just happen to use groups of six
> > connections. So, for many on-board and board-to-board "jumpers",
short
> > lengths (six or eight inches max, probably) of 6-conductor ribbon
cables
> > might be handy, although, changing to double-sided boards might
("should")
> > eliminate that need.
>
> Smallest connector you'll probably find is 10 pins, I think,
unless you get
> the longer strips and cut them, but I'm not sure about the cable
connectors.
>
> === Actually, many/most of the pin headers go down to 2 pins, I
think.
>
> > The primary GOAL is still to make the unit easier, faster, and
cheaper to
> > assemble. SO, I really don't want to use cabling schemes that
require a lot
> > of time or expensive equipment, to assemble (IDC?). Socketed
connections
> > are preferred, so that units can be easily disassembled for
repairs or
> > board-level upgrades. If ribbon cables are used, I would prefer
having
> > pre-assembled cables available.
>
> See above comment on that. :-)
>
> === Yup!
>
> > I did buy a couple-hundred new IDE 40-pin cables (with three 2x20
sockets
> > each), for $5! And I got several hundred 40-pin breakable gold-
plated
> > single-row pin headers for about $10. And I got 300 2.5-inch-long
36-wire
> > ribbon cables that have 36-pin single-row sockets on both ends,
for $30
> > including shipping, and 200 36-pin single-row gold-plated headers
for about
> > $5 or $10. (Actually, I got the ∗300∗ of the new IDE cables, for
$5.99 plus
> > $21.42 s/h.)
>
> There you go...
>
> === Ebay can be good for that, as can the military surplus
auctions. But I
> === ALWAYS run into TWO HUGE problems, doing it that way: 1) It
takes
> === WAY too much time, just to FIND exactly the stuff I need, and
even
> === more time to find a large lot that's "a steal". And, 2) It's a
nightmare,
> === as far as having a reliable, i.e. repeatable source of supply.
> ===
> === I've also seen some stuff that looks like it would be PERFECT,
that was
> === in fairly large qtys, and was VERY cheap. But it's stuff that
I've
> === NEVER seen anywhere else. So I definitely don't want to design
it in...
>
> > SO, I may end up cutting the IDE cables so they have just two
sockets with
> > about 8 inches of cable between them, and using those. Or I may
use the
> > short 36-pin cables and single-row headers.
> >
> > However, I am ALREADY running into the problem of ∗possibly∗ not
having
> > enough ROOM on the new front-panel PCB, for that many large
connectors. SO
> > I ∗STILL∗ probably need to find something smaller (fewer
conductors), for
> > the two smaller boards to use, to connect to the front panel PCB,
and/or to
> > the other boards.
>
> Hm.
>
> > ANOTHER IDEA: If I could find a very large (and very cheap)
surplus
> > stockpile of ISA "RISER BOARDS" (or even 8-bit passive backplane
boards, or
> > somesuch) that have 5 or more slots, those could make PERFECT
ready-made
> > motherboards, to mount in the bottom of the case, with the slots
parallel
> > to the front panel. Then ALL of my boards could plug into the
slots and all
> > be connected together, including the front panel. Of course, I
could make
> > my own similar motherboard-type boards, fairly easily, with
available
> > card-edge connectors. (But the large, new card-edge-connectors
are usually
> > quite expensive. Maybe I can find a large surplus lot of them...)
But MAYBE
> > there's a really low-cost stockpile of something similar,
somewhere, which
> > would certainly make things MUCH easier and faster and cheaper.
>
> I have a few of those riser cards on hand that I was hoping to find
a use for
> some time, and you're welcome to those, but I don't know about
several
> hundred. :-) I had a guy come by with a load of "computer junk"
the other
> day and declined to take the one Packard Bell machine that he had
that was
> one of those boxes, probably could've snagged another one, but...
>
> === Thanks for the offer. I just don't want to "design in"
something that I
> === am not certain of having a large, fairly-reliable supply of. I
could probably
> === very-easily buy a few thousand identical PCs, from a military
> === surplus auction or two, probably for about $1000 to $1500 for
each semi
> === trailer load, which is what I used to see them go for, all the
time. But,
> === just having to go through the buying and transporting and
storing and
> === removing what I wanted and disposing of the rest would almost
certainly
> === make it uneconomical, not to mention "WAY too much work". (It
might
> === be different if I could find a decent way to use the whole
case, and the
> === power supply, AND the motherboard slots, ALL left INTACT, where
I
> === could just plug in my boards and mount my front panel stuff
> === somewhere... Hmm.... Whacky. But maybe for some other
product!)
>
> > There are also several connections to the rear panel, usually
with only one
> > or two wires, that I need to worry about. I am thinking of using
either
> > one- and two-wire pin headers and sockets, for those, or small
terminal
> > blocks of some type. However, I still would LIKE to have all pre-
assembled
> > cables (i.e. sockets already on both ends of appropriate-length
cabling.
>
> For something like that I've seen some commercial gear that used
different
> styles of connectors (all 2-pin) so you couldn't mix them up and
plug
> something into the wrong place. You could also handle that issue
by setting
> the wire length to be only appropriate for where it's supposed to
go and
> similar tricks. Something to worry about, anyway.
>
> === Another very good idea. Noted!
>
> > Peak currents in some of the signal conductors could reach 1.5
Amps.
> > However, most of those waveforms are triangular or sawtooth,
making the
> > average (DC-equivalent) current only HALF of the peak value. But
the main
> > DC power supply rails MAY have to be connected from the separate
power
> > supply board to the other boards using discrete wiring that's
screwed into
> > terminal blocks, for that reason (max current-carrying
capability).
>
> Or at least something that's heavier-duty than pin connectors,
though it's of
> course possible to use several of those pins for this purpose.
>
> === Even the low-cost pin headers from jameco.com are rated at 1
amp per
> === pin. In this case, the 1.5 amps is the PEAK. But it's for ramp-
type
> === waveforms. So the average "DC-equivalent" current would only be
HALF
> === of the 1.5 Amps. And yes, the power supply rails (and grounds)
for
> === each board were going to be kept separate, anyway. And, as you
said,
> === I could always run them doubled (or more), for any high-current
ones.
>
> The Osborne 1 computer had a weird custom board on their floppy
drives, which
> was A and which was B was determined only by where the terminator
position.
> This was before twists in cables and similar nonsense. They also
ran the
> power for those drives up through the ribbon cable, to a card-edge
> connector. There were reliability problems with those over time...
>
> === Is it possible that the newer card-edge connectors are better,
now?
> === Well, never mind. Wherever I would have wanted to use card-
edge, I
> === can use pin headers and sockets, instead.
>
> > Sorry to have blathered-on for so long, here! If there's
anything you can
> > offer, I'm all ears!
>
> Hopefully some of what I've kicked out here will be of some help.
I see a
> lot of different things being done in commercial gear manufactured
over a
> long period of time, from no connectors at all to a bunch of
different
> alternatives.
>
> === YES. ∗∗QUITE∗∗ helpful. Some great ideas, and stored wisdom!
And
> === usually it also helps just to hash through it, with someone
else, anyway.
> ===
> === I DEEPLY appreciate your taking the time and energy to respond,
> === so well, and your willingness to share your knowledge and
experience.
>
> If I were building something I think I'd probably tend to favor pin
connectors
> for signals and something a little heavier for handling any kind of
power,
> maybe a "pin connector" that's larger (0.156 spacing?) and that
uses the
> square pins for a better contact.
>
> ===
> === Sounds right. I might even use the PC disk-drive-type power
cables,
> === for the heavier stuff. (I just missed a chance to get a lot of
something like
> === 500 brand new "Y"/splitter PC disk-drive power supply cables,
on ebay,
> === that went for something like $10, total...! Those could have
connected
> === all three of the boards... :-o )
> ===
> === Thanks again, ∗so∗ much! You, and the others on this wonderful
group,
> === are truly great. If there's EVER anything that ∗I∗ can do, to
help (any
> === of) YOU, please, just ASK!!
> ===
> === Highest regards,
> ===
> === Tom
> ===
> === Thomas P. (Tom) Gootee
> === tomg(AT)fullnet.com
> === http://www.fullnet.com/u/tomg
> === Jasper, Indiana, USA
> ===
> ===-------------------------------------
>
>
>
> [Non-text portions of this message have been removed]