On Wednesday 02 July 2003 09:47, Russell Adams wrote:
>> >This sounds suspiciously like either bad hardware or a bad SCSI
>> > chain. Are you sure everything is terminated properly? How's
>> > your cable length? Does it work on another machine or with a
>> > different SCSI card? What color was the goat?
>
>The tape drive was recently replaced, but I'd hate to replace it
> again...
>
>> That driver has been undergoing some work from what I see in the
>> kernels ChangeLog (I don't use it myself, my card is an advansys,
>> bulletproof) and if he has the aic7xxx.old driver available, maybe
>> he should give that one a try. My kernel is 2.4.21, and I don't
>> recall if both drivers are still available or not. That would be
>> my next suggestion, that he move on up to a newer kernel, thereby
>> getting the latest code.
>
>Actually, I've narrowed it down to just having to rm/insmod the sg
>driver only, between every call. :P
Which is indeed weird. I can't do that, haveing long ago built it
into the kernel.
>> And since this is scsi we're talking about here, the goat must be
>> pure white and a virgin. Not to mention that in a scsi setup, the
>> last device on the cable must be on the last plug on the cable. A
>> few inches of open cable past the device will make nearly all bets
>> instant losers. Then of course it doesn't matter what color the
>> goat was, the spell is broken... :)
>
>Damn. I've got a funky polkadotted goat, and I don't think its a
>virgin. Must I really obtain a new one?
Chuckle... You did get the point I see :)
>SCSI setup is very basic. One aic7000 card, one tape drive with
>termination. ;]
If you have a voltmeter, check to see if the term power available on
the scsi cable is high enough, I consider anything below 4.7 volts to
be pretty borderline, and you'd be amazed to see how many card makers
will use a cheap si diode for the isolation. At nominally .65 volts
drop acorss the si diode, that leaves about 4.3 to 4.4 for term
power. They really should be useing a schotky diode, with its .2
volts drop there but they cost more.
If the drive has a jumper that will allow it to supply its own term
power, enabling that will sometimes help.
The bottom line is that the TTL circuitry involved has a grey area
where the outputs are not guaranteed, from about .7 volts to 2.4
volts. While MOST chips will switch at about .7 to .85 volts, some
logic families will use the full range of the spec. As this, for
regular scsi, not lvd, is an 'open collector' circuit, this means
that a logic zero is done by turning on the output stage, which will
pull the data line down to maybe .05 to .1 volts, so there is an
adequate 'noise margin' of about .6 volts for the logic zero state.
When using resistive terms, the ideal would be to have the cable both
terminated in its characteristic impedance, which for this ribbon
stuff is about 120 ohms, and maintained when at rest at a voltage
high enough to give a decent noise margin for the turned off state.
2.4 + .6 = 3.0 (nominally) This would all normally be accomplished
by connecting a 220 ohm resistor to the 5 volt line and the data
line, and a 330 ohm resistor from the data line to ground. This
would establish an at rest voltage of 3.0 volts even, giving that
logic one state a noise margin of about .6 volts. And the parallel
combo about matches the cable impedance, all done with standard off
the shelf parts values, a huge cost advantage.
Unforch, there must be isolation diodes to prevent a scene where a
scsi device in an external box is powered up, but the main box is
not. Without the isolation diodes the external box will try to
supply power via the cable to the main box. The problem gets real
hairy when the isolation diodes are silicon because they'll drop the
voltage at the top ot the 330 ohm resistor to about 4.4 volts, and
you wind up with only 2.5 to 2.6 volts at the 220-330 junction. At
that point much of the noise margin has been used up, and any
residual ringing on the cable will give false signals as there is
nothing there to absorb them but the terminating resistors. Things
head for the toilet pretty quickly in that event.
In your case, I think I might be tempted to try a different scsi card,
based as much on the fact that the aic7xxx cards seem to be
generating more than their fair share of problems as anything else.
Humm, nother thought. These things have a bios of their own, and
sometimes they can become missconfigured. Reboot, and when the cards
own bios comes up, hit the ctrl+a or whatever to get into it and
check its config. Maybe its own terms have become disabled or
something. They should be on if it and the tape drive are the only
things on the bus.
Maybe we can keep from sacrificing that polky-dotted goat yet :)
>Russell
--
Cheers, Gene
AMD K6-III@500mhz 320M
Athlon1600XP@1400mhz 512M
99.26% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attornies please note, additions to this message
by Gene Heskett are:
Copyright 2003 by Maurice Eugene Heskett, all rights reserved.
|
