[Coco] SCSISYS question

Gene Heskett gene.heskett at verizon.net
Sat Jun 24 11:57:53 EDT 2006


Bob Devries wrote:
> Thanks, gentlemen!
> 
> Between Gene and George, I now have /H1 working using SCSISYS! I took 
> George's suggestion, of connecting all 50 wires between the two drives, 
> after I found that *if* I left the termination resistors in on both 
> drives, I actually got more progress than with them left out. I 
> installed a piece of 16-way cable, and VIOLA! it works.
> 
Great!  Maybe I should have tried that too.  And now you could probably 
remove the term packs on the drive in the middle if they are close 
together, say less than 4" of cable between them.  Or figure out which 
termpack pins to cut so that any one wire only has two resistors on it, 
ome on each end of the cable section involved, which would be the ideal 
situation.

SCSI has been historically treated as if one needed to have an unlimited 
   supply of virgins to sacrifice over it.  Thats not true IF the 
terminations are correct.  Termination voltage supplies are responsible 
for far more of this bitchyness.  With the usual requirement to block 
the external devices from powering the host in the event they are turned 
on but the host isn't being usually done with an SI power diode of some 
sort, the end result is that the actual supply for the termpacks is 
reduced by that nominally .7 volt drop in that diode, with the remaining 
4.3 volts then used to power the terms.

The SCSI buss is, at least at the drives, a  very high speed buss, and 
it only takes a ringing of 5 ns to be a valid pulse.  So this ringing 
MUST be terminated to reduce it below the noise margins.

The SCSI bus is done (ideally with ttl circuitry) that assumes all 
outputs are open collector, and all inputs are assumed to be ttl levels 
for logic 0 and logic 1 voltages.  All circuitry's logic 0 is easily met 
(<0.7 volts) with almost any active device as driver is able to pull the 
two terminating resistors to voltages in the .1 volt area.  This leaves 
about .6 volts for ringing and noise, so the logic 0 is going to be clean.

Conversely since there aren't supposed to be any pull-ups on this open 
collector wired or buss except the termpacks, and that the guaranteed 
logic 1 voltage required is 2.4 volts, one can do the math and see how 
much of that .6 volt noise margin between the ideal 3 volt logic 1 
resting voltage the designers envisioned is used up just by adding that 
SI power diode for host isolation.  If the psu itself is a bit low too, 
say its sagged to 4.87 volts (not uncommon at all), then the noise 
margin is all used up, and you start sacrificing chickens, then goats, 
and even virgins in attempts to make it work reliably.  Tain't gonna 
help folks, but finding some schotkey diodes (losses in the .12 volt 
range) to replace those Si diodes the friggin bean counters make the 
engineers use, and getting the psu up to a solid 5 volts will seem like 
a miracle.

> Again, thanks for the help, Gene, and George.
> -- 
> Regards, Bob Devries, Dalby, Queensland, Australia
> 
> Isaiah 50:4 The sovereign Lord has given me
> the capacity to be his spokesman,
> so that I know how to help the weary.
> 
> website: http://www.home.gil.com.au/~bdevasl
> my blog: http://bdevries.invigorated.org/
> 
> ----- Original Message ----- From: "Gene Heskett" 
> <gene.heskett at verizon.net>
> To: "CoCoList for Color Computer Enthusiasts" <coco at maltedmedia.com>
> Sent: Saturday, June 24, 2006 3:04 PM
> Subject: Re: [Coco] SCSISYS question
> 
> 
>> Bob Devries wrote:
>>> Gene,
>>> does the term power usually come from the host adaptor, or from one 
>>> of the drives?
>>> If it is the latter, could the problem be solved by simply having all 
>>> 50 wires present between the two drives?
>>>
>> ISTR I did consider that possibility, but never got a round tuit so it 
>> wasn't tried.  I bought a TC^3 instead.  Then I had to make up a 
>> longer cable, so its all there now.
>>
>> I'd add that I've got an old, slow scsi cdrom drive, and adding it to 
>> the new cable also kills the system, won't even boot.  I  haven't dug 
>> into the why nots on that either.  Need 36 hour days at times.  And at 
>> 71, sometimes even a 24 hour day seems to be too long.
>>
>> -- 
>> Cheers, Gene
>>
>>
>> -- 
>> Coco mailing list
>> Coco at maltedmedia.com
>> http://five.pairlist.net/mailman/listinfo/coco 
> 
> 


-- 
Cheers, Gene




More information about the Coco mailing list