[883] in linux-scsi channel archive
RE: termination
daemon@ATHENA.MIT.EDU (Al Longyear)
Wed Nov 6 04:17:15 1996
From: Al Longyear <longyear@sii.com>
To: "'linux-scsi@vger.rutgers.edu'" <linux-scsi@vger.rutgers.edu>,
"'Russell Berry'" <rberry@albany.net>
Date: Tue, 5 Nov 1996 07:55:25 -0800
SCSI termination occurs at both ends of the SCSI chain. If you don't
terminate both ends of the chain then you will get reflective signals
and your bus will go to sh^t very quickly.
If you put the computer's controller in the middle of the chain then it
must not have termination.
You may have the chain which looks like:
1Gdrive(*) -----> computer -----> 1Gdrive ------> 1Gdrive(*)
(*) == terminated.
It is hard to tell from your diagram whether you do or not, especially
when you say that you don't have external devices.
However, if this is truly your configuration then it is correct. Did you
leave the terminators on the internal bus of the computer's controller?
If so, remove them.
However, if you have attached all three drives to the internal cable on
the controller then you have a condition such as the following:
computer(*) ----> 1Gdrive -----> 1Gdrive -----> 1Gdrive(*)
The one drive which does not continue on to the next item must be
terminated along with the computer's controller.
I do not know about the NCR controller. However, my Adaptec controller,
has both an internal connector and an external connector. This means
that if you have both internal and external devices on the controller
then you are putting the controller in the middle of the chain as I
diagrammed above. However, if you have _only_ internal _or_ only
external devices, then the controller is at the end of the chain and
must be terminated.
The other concept which is commonly missed is called "Termination
Power". This is the power to the terminators. There is one school of
thought that one and only one device must supply the termination power.
There is a totally different school which says that _every_ device on
the bus must supply the power. (I favor the former as the latter will
cause other problems if the voltage levels are different since the two
supplies would be run in parallel.)
However, there does seem to be agreement about one thing:
At least one device must supply the power.
The disk drives may or may not have jumpers to supply the termination
power. If they don't have jumpers for specifically enabling termination
power then it is highly probable that they just don't supply it. The
controllers such NCR53C810 as the usually do. Did you remove the jumpers
for termination power?
>----------
>From: Russell Berry[SMTP:rberry@albany.net]
>Sent: Monday, November 04, 1996 9:13 PM
>To: linux-scsi@vger.rutgers.edu
>Subject: termination
>
>I hate to sound like an idiot here, because I'm not..
>
>no, really, I'm not! I work with this crap every day, my mind
>just seems to be drawing a blank right now. I have:
>NCR53C810-----1G_fast_scsi-II terminated scsi-id0
> |------1G_fast_scsi-II not-terminated id1
> |------.5G_fast_scsi-II terminated id2
>
>all three disks internal, nothing external (yet) they work,
>and seem fast, I'm just unsure at this particular moment if
>I should have two of them terminated or not...
>
