[Coco] Educate me about Y-Cables

[Gene Heskett]

On Friday, February 20, 2015 02:48:04 PM Joe Grubbs wrote:
> I've seen a number of references to people using Y-cables to facilitate
> the use of more than one program pak. Is there a vendor for these? If
> not, is there a description of how to make one?
> 
> What kind of paks can someone use simultaneously with one of these? I
> was always under the impression that paks could only be operated one
> at a time (hence the selector switch on the MPI), however a Y-cable
> seems to imply this isn't always the case.

The selector switch was mandated by basic's inability to handle either 
slot selection, or slot an IRQ came from gracefully.

OS9/Nitros9 OTOH, is smart enough to handle the search for an IRQ source 
nicely, and can also override the switch settings when it needs to.  That 
however IS the device drivers responsibility.

This business of the IRQ handling by way of the switches logic did however 
throw a huge monkey wrench into into os9's ability to service an IRQ in a 
timely manner and that put a huge crimp in its ability to handle the 1 per 
incoming character IRQ generation, it was unaware of the IRQ until the 
slot was switched to that slot, blocking that IRQ until the slot was 
selected.  The resulted in wholesale loss of data from frameing overruns 
in the sc6551 in the 232 and modem packs.  Even at 300 baud.

Because os9 can find the irq source in maybe 15 microseconds if the IRQ 
switching logic in the MPI is bypassed by physically tieing the #8 pin of 
all 4 sockets together, this is by far the best way to handle the problem. 
Carefull coding of a basic or assembly program when running the binary 
from rsdos can also handle this correctly, so while I have campaigned for 
something north of 25 years to be aware of possible problems with it in 
rsdos, I have yet to read on this list or its predecessors all the way 
back to Princeton days, of anyone haveing a problem using the jumpered MPI 
for RSDOS too.  It could happen I suppose, but it would probably be some 
grey haired old code from the 80's that might not behave.

===============Caveat=================

Because tieing all 4 of these lines together also ties all 4 of the pullup 
resistors together, the output gate in the individual pack is heavily 
loaded, so much so that you can see the interrupts as bars in the gime's 
video output, it shakes the supply lines that badly.

Most MPI's had 4.7k pullup resistors, and when all 4 are trying to keep 
the line high, the pack has to sink around 50 milliamps.  So the common 
practice when putting in the jumpers, is to locate those 4 resistors along 
the edge of the MPI's board, and clip one end of 3 of them, leaving only 
one, which is much easier on the pack doing the pulldown to exert the IRQ.  
Not all MPI's had 4.7k there, some had 2.7k and I ran into one way back 
that had 27k resistors there.  All are adequate to get the job done, and 
in terms of power supply loading, the higher the better. A single 220k 
might even be enough, the recovery time at the end of the IRQ would be the 
controlling factor.  It is buffered in the MPI's logic anyway so its not 
like you are driving 2 feet or wire to actually arrive at the 6x09 socket.

============y-cables==============

There is a second problem with y-cables.  When the MPI is plugged into the 
cocos side port, the spring clips on each side of the socket grab the 
'ears' on the MPI's board, and the packs are similarly equipt with ears 
and socket clips, giving a much better ground than the narrow trace from 
the one pin in the socket does.  The y-cable I have does NOT have the 
extra grounding, making "ground bounce" a much larger problem in terms of 
error free data movement.  So much so that I have not found a situation 
where I can use it with 100% dependability.  Others I understand have used 
one that way and it worked.  So all I can say is that it loads the coco's 
power supply worse than the MPI because all the power used must come from 
the coco. And it may trash a byte now and then.  Often enough I would 
never plug a disk controller into it.

A long winded way of saying YMMV I guess. :)

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>