SUMMARY: SCSI termination

From: Dave Mitchell (D.Mitchell@dcs.sheffield.ac.uk)
Date: Tue May 26 1992 - 16:52:54 CDT


A while ago I asked various questions about SCSI termination etc.
I didnt get quite the answers I was hoping for, but here goes anyway:

>1) is there an easy way of checking whether a SCSI cable/disk etc is
>terminated by simply putting a high-impedance multimeter on some pins
>in the connector? If so, what pins, and what resistance should be present?
>(The alternative is to spend ages dismantling a box, trying to see if
>internal terminators are present, etc).
>

noone seemed to know of an easy way. (I havent tried it, but would putting
a digital resistance meter across pins 1 & 2 when the machines are switched off
give you some idea? -eg n ohms = terminated, n/2 ohms = over-terminated,
open circuit = not terminated? Answers anyone?)

>2) What would be the general effects of having a SCSI bus unterminated
>or multiply-terminated (ie marginal working/failing, or complete refusal
>to work
>

not terminated: may work, may not, may get intermittent faults, etc - esp
                with increasing cable length
multiply terminated: a good way of blowing the internal fuses on the CPU board

>3) Are intermittent "SCSI disconnected", "SCSI reset", and
>"xxx now synchronous at 4.0Mb/s" messages symptoms of incorrect
>termination and/or too long total cable length?

quite possibly - I've shortened the cables, and havent had any problems in
the last week. ps our techicians knocked up some shortish (50cm) cables using
unshielded ribbon cable - is this safe????

>4) while I'm on the subject of disks, does anyone know a good source of
>information (manual/book/ftpable-doument) on disks (esp SCSI) that explains
>things like defect lists, disk labels, boot blocks, alt cylinders and the like
>- ie I know the basics like how to partion a disk, I'm more interested
>in how the partition label is stored on the disk etc.

noone seemed to know!
Best bet appears to be to read the disk manufacturer's manuals.

thanks to

vsh%etnibsd@NET.UU.uunet (Steve Harris)
hs@de.gwdg.ukps.demeter (Howard Schultens)
Ross A Macintyre <raz@uk.ac.hw.cs>
Roy Richter <rrichter@com.gmr.ph>
dinah@com.tivoli.rockytop (Dinah McNutt)
alastair@uk.co.eucad
guyton@org.rand
David Gunn ldavis!gunn@snowbird.central.sun.com

----- Begin Included Message -----

From: vsh%etnibsd@NET.UU.uunet (Steve Harris)

I'm no SCSI guru but ....

The SCSI bus should be terminated at each end.

Non-terminated sometimes works.

The purpose of termination, (which pulls the line to +5V), is to
prevent echos (for more specific explanation, consult your local EE).

In theory, anything beyond the termination cannot be seen. In
practice, this is not always true, but it almost always is. You should
not place an active SCSI device beyond the terminator. (I think
placing a terminator in the middle of the cable **should** divide the
cable into two separate SCSI busses, but I'm sure some signals would
trickle across the terminator and interfere with the other half of the
cable.)

Total cable length is important, as is stub connector length. The
latter is the length of the connector by which the peripheral is
attached to the SCSI bus. I don't know the spec, but I believe the
total cable length is something like 3 meters, and stub length is of
the order of 1-2 cm. Most SCSI devices simply use an IDC connector, so
the stub length is of the order of 2 to 4mm. Cheap SCSI "shoeboxes"
may be incorrectly cabled, as illustrated below:

    key:
        ___ = cable
        X = external connector (mounted on shoebox housing)
        # = connector to device

        correct: X______#_______X

        incorrect: X______X_______#

In the "incorrect" wiring, if the shoebox is in the middle of a chain
of scsi devices, the internal cable becomes the stub, which may be as
much as 20 cm. (Note that the "incorrect" wiring will work at the end
of a SCSI bus.)

Looking forward to your summary. Hope you get pointers to a couple
good texts, esp. pragmatic, how to make it work and troubleshoot
problems ones. EE oriented texts, while necessary, are less useful to
ordinary sysadmins trying to diagnose problems.

--
.. Steve Harris - Eaton Corp. - Beverly, MA - uunet!etnibsd!vsh

----- End Included Message ----- ----- Begin Included Message -----

>From rrichter@com.gmr.ph Thu May 14 13:20:30 1992

> 2) What would be the general effects of having a SCSI bus unterminated > or multiply-terminated (ie marginal working/failing, or complete refusal > to work

I've done this too often. Unterminated buses work sometimes, but mostly not at all. The best thing to do is a "probe-scsi" from the boot prompt in "new" mode, that will tell you what devices the CPU can see, and if things will work. If they don't work, you might not even get a prompt back. Anyway, my experience is the first device you lose sight of is the internal disk.

If you multiply terminate, you run the risk of blowing the SCSI fuse on the motherboard. I have a box of these (2A 125V, looks like a drum with both leads on the same side) for such an occasion. It seems that the IPC's blow easiest, the SS2's least, but multiple terminations draw more power from the bus, and eventually you'll blow. -- Roy Richter Internet: rrichter@ph.gmr.com Physics Dept, GM Research UUCP: rphroy!rrichter

----- End Included Message ----- ----- Begin Included Message -----

>From guyton%condor@org.rand Thu May 14 17:23:31 1992

> 1) is there an easy way of checking whether a SCSI cable/disk etc is > terminated by simply putting a high-impedance multimeter on some pins > in the connector? If so, what pins, and what resistance should be present? > (The alternative is to spend ages dismantling a box, trying to see if > internal terminators are present, etc).

I think it takes more than just a meter -- I have been wishing for this "meter/box/whatever" for a long time. On the other hand, a good scope would tell you what the signals looked like, and that's what gets clobbered by improper termination. Oh, there is one easy check to make with a meter -- you can test to see if termination power is being supplied. This is on pin 26 and must have juice if you're using external terminators. If you have internal terminators (and they don't use power from the bus; which is usually the case), then it doesn't matter if anyone is driving pin 26. We've just taken to always keeping a label on the disk that tells us the scsi id of the device and whether or not it's internally terminated.

> 2) What would be the general effects of having a SCSI bus unterminated > or multiply-terminated (ie marginal working/failing, or complete refusal > to work

Any of the above. You can have a bus over-terminated or under-terminated and it *might* still work. Especially if you have very short cables.

> 3) Are intermittent "SCSI disconnected", "SCSI reset", and > "xxx now synchronous at 4.0Mb/s" messages symptoms of incorrect > termination and/or too long total cable length?

A good first guess, yes.

> 4) while I'm on the subject of disks, does anyone know a good source of > information (manual/book/ftpable-doument) on disks (esp SCSI) that > explains things like defect lists, disk labels, boot blocks, alt cylinders > and the like - ie I know the basics like how to partion a disk, I'm more > interested in how the partition label is stored on the disk etc.

I've used three sources:

1) The SCSI spec from ANSI (it's ANSI X3.131-1986

2) The programming manuals from the makers of the disk/tape drive that you're dealing with.

3) SunOS source.

-- Jim Guyton guyton@rand.org

----- End Included Message -----



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:43 CDT