PCBUILD Archives

Personal Computer Hardware discussion List

PCBUILD@LISTSERV.ICORS.ORG

Options: Use Forum View

Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Len Warner <[log in to unmask]>
Reply To:
PCBUILD - Personal Computer Hardware discussion List <[log in to unmask]>
Date:
Sun, 12 Jul 1998 22:52:21 +0100
Content-Type:
text/plain
Parts/Attachments:
text/plain (99 lines)
A SCSI interface operates as a bus,
that is, all devices are connected equally,
but the cable is designed to be "electrically long"
and transmission line effects must be allowed for.

If the cable is not terminated at its two ends
signals will be reflected and these echos may
disrupt data communication.

Termination is usually done at a device for convenience,
but separate plug-in terminators are easily available.

Termination may be "passive" using a resistor network,
"active" using a voltage regulator and series terminating
resistors, or "FPT" (forced perfect termination) using three
voltage sources, terminating resistors and catching diodes.

Example 1 - an old IBM HD with no provision for termination
is at the end of my internal SCSI cable, so it is connected
via a 50-way M-F through terminator.

Example 2 - my external SCSI cable ends at a portable CD:
if I remove it I must replace it with a terminator, or
(if its own termination is switched off) I could leave
a through terminator permanently on the end of the cable

The SCSI ID numbers are irrelevant to the termination problem:
* the host adapter is usually #7 and might offer no choice;
* BIOS controlled HDs usually have to be #0 or #1;
* higher IDs have higher service priorities and so should be
  the slower devices for maximum bus utilisation, (but I doubt
  many users need to bother about this).

>1 - Device #3 is attached to the third connector on the chain,
>it's also the last device on the chain,
>should it be terminated?

You are not clear about the cabling:
is the last device at the end of the cable?

The _cable_ needs to be terminated: if it ends at a device,
the device may be terminated, otherwise use a terminator.

>
>2- If I terminate Device #3
>then Device #0 will be in the middle of the chain.
>Will Device #0 still be shown as Disk 0?

This is a doubly confounded question:
1) terminating does not alter the position of a device
   so the explicit implication does not make sense
2) cable position is independent of SCSI ID # so
   the device ID #0 will be whatever ID #0 makes it.

>3- The device order on the cable will be
>#7-Controller (terminated),
>Disk 0 (not terminated),
>#3 Disk 1 (terminated).
>    a. Will this work correctly?

IF the terminators are at the _ends_ of the _cable_ then YES

>    b. Do I need to change the Device number order
>       to be #7, #3, #0?

Only if this makes for a better physical routing of the cable
or makes it easier to terminate correctly.

>4- The current setup is
>Device #7 - controller (terminated),
>Device #0 - Disk 0    (terminated),
>Device #3 Disk 1 (not terminated).
>
>Disk 1 is still able to be written to and read from.
>If a Device has two terminated Device's earlier in
>the chain, why is it that the Device is even usable?

One must presume: because the level of data reflections
is not sufficient to disrupt communication.

I infer from the description that this is an internal
cable so it is probably less than 18 inches long.

Consider that IDE cables are designed to work _without_
termination at a similar data rate, but are therefore
constrained to be "electrically short". The specified
maximum length is 18 inches.

Consider also that SCSI is specified to allow a stub
not exceeding 4 inches: the distance from terminated
device #0 to the unterminated end may not be much longer
than this.

It is also possible that one or both of your terminators
are FPT, which will clip the doubling of the signal voltage
otherwise caused by reflection from an open-circuit.

Len Warner <[log in to unmask]> WWW Pager http://wwp.mirabilis.com/10120933

ATOM RSS1 RSS2