[Coco] 4 Port MPI PCB artwork (so far)

Gene Heskett gheskett at wdtv.com
Sat Mar 7 19:40:07 EST 2015



On Saturday 07 March 2015 15:41:02 RETRO Innovations wrote:
> On 3/7/2015 1:56 PM, Gene Heskett wrote:
> > I saw that easily enough, but cannot visualize it for two main
> > boards in connectors 2 & 3 of each cable?
>
> Well, at this point, it's an idea of the type:
>
> *Don't know if it will work, but should not do anything that would
> artificially preclude it, all things being equal*.
>
> I agree the second attachment will be a bit wonky, but length is
> probably the most pressing thing, and there's nothing I can do about
> that :-(

Don't forget Jim that we CAN make our own cables, and that length limit 
is purely an artifact of the much faster (relatively) speeds of an ide 
interface.  To put that into perspective, I have a 3 drive floppy cable 
I made yonks ago that is, because of where the drives were vs the 
controller, about 6 feet from the controller to the first drive.  It has 
never dropped a single bit that I am aware of.  IOW, at our relatively 
leasurely clocking, a 24" cable or even a whole yard of it s/b fine.  If 
its troublesome, apply std scsi terminations to it.  Between getting the 
signal traces to a common termination point at both ends, and the chips, 
perhaps another $10 bill.

Properly done, that scsi-ii cable can according to the specs, be 39 
meters long. But Yogi Berra's famous quote still applies, so with 
production grade tolerances in effect , I would expect trouble at a 
fraction of that distance.  Thats the practice half of Yogi's 
quotation. :)

The generally accepted impedance of the flat ribbon 40 wire where every 
other wire is ground, is nominally 120 to 122 ohms.  The 80 wire version 
is generally said to be about 60 ohms.  Either would best be done 
in "active" terminations, but back when scsi-ii was the cadillac of 
cabling, we did not have the one chip solutions for that that we should 
have, gathering dust on the discount shelves, now.

> > On the outer edge of the board, so a slider cap could be put on the
> > switches stickup which is projecting past the edge of the board? 
> > With a small stem on the slider/knob, basically a miniature of the
> > tandy switch with the same exterior in miniature.  But I'm 200% open
> > to whatever idea makes it accessible from the outside if we make
> > cases for it while of coarse keeping costs as low as is practical.
>
> I really like the previous idea of moving the switch and LEDs to the
> main board.  I am in process of doing that now.  It allows me to
> better rationalize the need for 2 IDE cables.
>
> > jumpers or switch?
> > Flea clip jumpers per slot?  Accessible by suture clamps when that
> > cart is removed?  I certainly would not ask for anything fancier
> > than that, and certainly not when the one doing the work is doing it
> > for free.
>
> jumpers are easy, and electrically stable.  switches are not, and
> create RF (all that travel in the switch creates an antenna for non
> steady signals) so I greatly discourage them (you can use switches, if
> you put the switch on a TTL switch (like a set of buffers, or mux),
> which keeps the signal path pure, but that adds ICs.

The idea of a mux is done in SW for linuxcnc. :)  I have at least one 
such instance in the .hal files that configure the code to drive my 2 
machines.  But in practice you are right.  The flea clips are relatively 
small so I wouldn't normally point too sharp a finger at its noise 
radiation. Certainly nothing that a Holaday meter can specifically find.

> > As it is, I feel I may actually be slowing the work, when what I am
> > interested in is working hardware from the first run of boards.  The
> > many eyeballs theory.  Sometimes the minutia can escape our
> > consciousness, its happened to me on many occasions.  Fixing a
> > problem is comparatively cheaper when fixed at the line on a screen
> > stage. :)
> >
> > IOW, I can be told to shut up anytime.  Designing by committee tends
> > to get un-manageable if a halt is not called to the feature creep.
>
> As I noted before, this is not my first trek down the path, and I can
> sort the wheat from the chaff.  For instance, the ideas this morning
> are invaluable:
>
>   * The idea to move the switch and LEDS to the buffer board sounds
>     plausible, and cleans up the design.
>   * Moving the switches allows me to put all of the power on one edge,
>     which is a good idea.
>   * While moving the power around, I realized it will absolutely not
>     work where it is in a enclosure.  Since the enclosure much be
> wider than the PCB on the sides, the connectors would end up
> significantly recessed into the enclosure, which is a mess.
>   * But, moving them to the front side of the PCB is OK, because that
>     portion of the PCB can rest against the enclosure wall.
>   * The suggestion to put LEDS on individual resistors is a good
> future proof design idea, since I have an idea that will allow
> multiple slots to be *on* at once.

>   * The suggestion to move the SV connectors further from the slot
> conn is massively appreciated.
>
> On the other hand
>
>   * Symmetric mounting holes is probably not going to happen :-)
>     Something has to give.
>
> Still, a good design needs a bit of time to "season".

And breadboarded for proof of concept. If I know I'll be making 1 or two 
on the mill for that, then I should call Highland Hardware in Hotlanta 
and get another slab of micarta to make the vacuum pallet from.  
Unforch, they don't have it any more, or their search function is dumb.

It looks like ebay is about the only source but chain down your credit 
card, some of it is priced way out there. I did find enough to make 2 
pallets big enough for this, in the $100 range. 5/8" thick, s/b stiff 
enough.  I'd donate one of those pallets to this, IF I can convert your 
*.brd file to gcode. Helps keep me out of the bars dont'cha know. :)

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>


More information about the Coco mailing list