[Coco] Multipak redesign/replacement

[Gene Heskett]

On Sunday, February 22, 2015 01:46:00 PM Andrew wrote:
> Gene,
> 
> When I was originally thinking about the idea I proposed regarding a
> "backplane" - I wasn't thinking of it having a case, and it being only
> a kit. For a low run (50 units) - something like that I think could be
> done for the prices you suggested ($60 kit, $120 assembled - and the
> end user would need to supply a case). I seem to recall something like
> this being sold in Rainbow (basically an open-frame MPI, with a cable
> to the CoCo side)...
> 
> Hmm - I guess my memory was faulty on the details - but it was called
> the "A-BUS" by ALPHA Products (242-W West Avenue, Darien, CT 06820) -
> from the advertisement on page 21 of the December 1987 Rainbow (and I
> am pretty sure they were advertising before this issue and somewhat
> afterward).
> 
> Each "motherboard" could hold 5 cards, up to five adaptors could be
> connected; they seem to have had their own line of cards for various
> purposes (all look like they were for industrial/robotics control
> purposes). Various adaptors (to connect to the motherboards) were
> available for a variety of different systems, of which the CoCo was
> one. They also mention that the adaptor could work in an MPI - so in
> theory with a CoCo (if I am reading right) - 100 slots were
> potentially possible (power req's aside)...
> 
> Anyhow - take a look at that ad if you care; that was what I was
> imagining, except with the cards (and "motherboard") turned 90 degrees
> so that they were all the same "height".
> 
> I'm ignorant about the design of the MPI, and I have not seen the
> schematic; I know it is more than just some buffered I/O data/address
> lines from the port ganged together (for one, there's the switch up
> front, plus it needed it's own power supply) - but I have to admit that
> I didn't think it was much more complex (maybe a few extra chips for
> address selection - plus other passives for filtering and such).
> 
> I do agree that the initial system for design and debug purposes should
> be kept as simple as possible, to learn from and see what works, etc. I
> also wonder if maybe (with the right design - again, being ignorant of
> the design and schematic of the MPI, this may be impossible) the number
> of slots could be configurable; that is, start with 2 or 3 slots, and
> then if you need more, extra slots could be added. That way, the final
> kit could be customizable to the needs of the end-user, and in the
> future (providing the design files were kept available to the
> community), if more slots were needed and you or whomever couldn't make
> them, they might not be to difficult to DIY. It might also lower the
> cost of the project (?).
> 
> Finally - and again being ignorant of the schematic and PCB for an MPI
> - does milling make the most sense (to you)? From what I gather, it
> seems you have the equipment and skill to do it - so that's a "sunk
> cost" - but would other methods make the turn-around faster for
> revisions (toner transfer, UV transfer, or even sending off to an
> el-cheapo chinese boardhouse)?

Any of those options involve ferric chloride or a similar chemical 
etchant, pure hell on the environment if not properly neutralized before 
disposal of spent etchant.  My machineing of the development board(s) 
makes only a bit of copper dust, and if I dig too deep, some fiberglass 
dust. Thats why I have to go to rather extreme lengths to make sure the 
top surface is dead flat.  I first machine the vacuum distribution grooves 
about 1/16" deep in a 1/2" thick sheet of micarta, stopping the ditches 
1/4" from the edge of the board, then, without moving the micarta on the 
table, machine the rectangular outline of the board blank into the 
micarta.

Then I drill a deeper hole in the center of the ditch pattern, and drill 
into the bottom of the hole with an 1/8" bit to intersect the hole, and 
drill that out to 1/4" for 3/8" ot so in from the edge of the micarta, 
then superglue 3/4" of 1/4" brass tube into the hole where I can connect 
the hose from the vacuum system.

Two edges of the pocket get a decorator hold down screw installed, tapped 
for about an 0-80 thread, and a flathead 0-80 screw then pushes the board 
to wedge it into the opposite corner.  And to simplify the mirroring math 
for the machine, a 3rd screw is installed to push it to the opposite side 
when the board has been turned over to do the other side.  The micarta 
also has a small pipe installed vertically just off one corner of the 
pocket, and this serves as the alignment zero by measureing the electrical 
contact with a routine I wrote that finds the exact (+- a thou or less) 
center of that pipe.  The pipe is the contact and the toll ding the 
touching is ground for this.

Then cut out a piece of double-sided board, and carefully file the edges 
until it fits in the machined pocket with  maybe 5 thou of clearance.

Tighten which ever screw(s) push it in the correct direction gently and 
bring up the vacuum.

I do not drill holes all the way thru when doing a drill hole, just a hair 
over half-way as that would destroy the vacuum, but once the board has 
been turned over and the traces cleared, the last op is to drill the holes 
from that side of the board.  Usually you cannot find a step in the middle 
of the hole where they meet.  So I have vacuum hold down until the last 
drill operation is started.  At that point its moot if the board warps 
upward from the loss of vacuum.

I have 2 problems however, first being that I know squat about programming 
a cpld or fpga.  I don't have it on screen atm, so I don't even know what 
cpld little John had in mind.  Nor do I have a programmer.  So, my costs 
would obviously be connected to whichever one of you that does do that 
sort of thing wants for a pre-programmed cpld I can just set on the board 
& reflow in place.  Or set into a socket if that would be preferred.

I also don't, at this stage, want to get heavily involved with actually 
producing it once it is in an operational state, so when it works, it all 
gets posted to my web page, and Ed or one of you can take over the 
logistics of farming the board out, collecting the kits parts and 
collecting the sheckles as you ship.  I don't want to be in the hole like 
Roger got himself into, if I should fall over (again).  I've had my 10 
minute warning buzzer once already.
> Maybe not for the very first one (which may be all you are after
> initially) - but for revisions afterward...? I'm not questioning your
> skills or experience here; I'm just trying to gain some understanding
> myself about the process, as I said that I don't have any experience in
> design and development of a PCB. It also may be the size of the PCB
> involved (which I wonder if a smaller but modular design like I
> suggested above might not help in the initial design and debug stage as
> well).
> 
Without a license for eagle, little Johns board uses all the real estate 
on the pcb that the free eagle allows. So would a 4 slot daughterboard, 
but that can be a step & repeat process, therefore not a real problem at 
all. And there is zero reason not to use an xt or atx psu's drive cables 
to give each 4 slot board its own direct psu including the supplies + 12 
volts intended to spin hard drives. seriel packs would need an inverting 
charge pump, or better yet, put that on each slot carrier.  Then the teeny 
little tin plated smoke generator with the aztec label on it can be 
removed and the pack jumpered to use the +- on the slot socket.  Thats a 
genuine win-win there.

I would also be ecstatic if in this process, we could figure out a way to 
reach back into the coco and disable the internal /CS except for the 2 
pia's in it, addressed at $FF00-03 and at $FF20-23, as just that seemingly 
simple oeration would open up 14 each, 4 byte wide addresses worth of I/O 
space BELOW the floppy at $FF40.  That would go miles toward alleviating 
the CoCo3's loss of all the I/O space above $FF7F which is usable on the 
level 1 coco's.  That could then be decoded in the mpi, making a 16 slot 
mpi feasible IF we can find the system ram to run all those drivers in.  
That is just one of the reasons we need level 3.

But I do not know if that reaching back into the CoCo3 to kill the /CS 
from the far side of it's side port socket is even possible. I have 
contemplated hacking into it with 2 more gates and fixing mine.  But where 
do I put it?  Except for under the keyboard, mine is best described as 
stuffed.

> Well - that's enough for now...oh, and I would definitely be willing to
> buy one of these, even if it came to more than $120.00 for a kit. Also,
> I have a few different cartridges that could be tried out (not just rom
> paks - though I have plenty of those - but also others; RS-232, floppy,
> speech-sound, Orchestra-90, even a DS-68B video digitizer somewhere).

Tasty sounding, thats for sure.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>