http://www.cucug.org/ar/ar404_Sections/feature4.HTML (Amiga Plus Extra No. 5/97, 05/1997)
SCSI Bus Termination
===========================================================================
SCSI Bus Termination
Robert Davis -- Amateur Radio K0FPC rdavis@nyx.net
===========================================================================
In the Usenet newsgroup, comp.sys.amiga.hardware, there seems to be a
continuing discussion about "termination of the SCSI bus" especially
concerning the Amiga 3000.
This article will attempt to explain SCSI bus termination. The discussion
in csa.hardware will continue, no matter what.
And this article will be about the type of SCSI found in Amiga computers.
Newer implementations, especially differential SCSI, have not been used in
host adaptors on the Amiga, so there is no reason (yet) to cover them.
First of all, here is the standard way to install SCSI termination on the
bus:
----- ----- ----- ----- ----- ----- ----- -----
| | | | | | | | | | | | | | | |
T| D |===| D |===| D |===| D |===| D |===| D |===| D |===| D |T
| | | | | | | | | | | | | | | |
----- ----- ----- ----- ----- ----- ----- -----
In the diagram, D is any SCSI device, and T is a resistor termination pack.
=== is the connecting cable between the SCSI devices. One of D must be
"hard drive controller" or more properly, a SCSI host adapter. Other D
could be a SCSI hard drive, a SCSI CDROM drive, a SCSI document scanner, or
other SCSI device for which driver software is available.
In the vast majority of Amiga applications, there will be two or more (but
no more than eight) of D, and there will be two and exactly two T.
Attaching the various SCSI devices together are ten control lines and eight
data lines. I'll use the Amiga 3000 SCSI system as an example, mostly
because the A3000 is the topic of discussion most of the time in
csa.hardware.
Nine of the control lines and all of the data lines are tied to the Western
Digital SCSI IC. The tenth control line comes from the RESET circuitry and
is not controlled by or routed through the host adaptor.
Each of those control and data lines is terminated at each end of the SCSI
bus like this:
o --> +5 volts (termination power)
|
\
/ 220 ohm resistor
\
/
|
o ------> one SCSI control or data line (eighteen total)
|
\
/ 330 ohm resistor
\
/
|
----- --> connection to ground (or common)
/ / /
This is not the only way to properly terminate the SCSI bus, and it is not
even the best way, but it is the termination nearly always found in Amiga
implementations.
Now it gets exciting. The SCSI bus is more than just a bunch of wires
connecting stuff together. Because of the speed at which it operates, we
have to consider the bus to be a set of "transmission lines" and we must
consider the "characteristic impedance" of the lines to under- stand why
termination is needed.
Signals are sent out onto the bus by the host adapter and by the other SCSI
devices millions of times per second. The system acts more like a
collection of small radio transmitters connected to receivers which are
very picky about the quality of the signals they receive.
Once we recognize that situation, we can move to transmission line theory,
which has been well understood for many years. Amateur Radio operators
should be familiar with the American Radio Relay League's Antenna Book,
which devotes several pages to matched and mismatched transmission lines.
When signals are sent by one of the devices onto the SCSI bus, those
signals must not be interfered with by other signals on the bus. The
potential interference comes from a property of transmission lines. A
signal sent onto an infinitely long transmission line wlll never come back
to interfere with other signals on that line. But real transmission lines
are never infinitely long. Each wire has two ends.
And there is the reason for terminating the SCSI bus. At each physical end
of the cable, we want the cable to appear to the devices attached on the
SCSI bus to be infinitely long. On a real transmission line, not
infinitely long, signals sent by those devices onto the cable will be
"reflected" back onto the bus, and will conflict with the signals sent out
anew by the devices.
You may see evidence of this problem on your television screen. One result
of a mismatched cable between your TV and your antenna or between your TV
and your cable system is "ghosting" on the TV screen. In that situation,
the reflections are visible as less strong images offset from the primary
images on your TV.
The way to make a short transmission line appear to be infinitely long is
to terminate that line with a resistance equal to the line's characteristic
impedance. That is called a "matched" transmission line. A perfectly
matched line has no reflections. But perfection is hard to achieve.
Not all SCSI cables are alike. The differences are slight, but they
matter. Sometimes we use cables not really suited for SCSI. Any use of a
25-conductor cable disturbs the characteristic impedance of the 18 signal
lines on the bus, because each line is supposed to be in a twisted pair
arrangement with its own individual ground line.
Sometimes the cables are not properly affixed to the connectors which plug
into the various devices on the bus. Sometimes a good cable is used in
conjunction with a bad cable.
In almost every installation, the characteristic impedance of the SCSI bus
is different from the impedance presented by the termination resistors.
Another type of termination, called active termination, does much to fix
this problem. But few devices with active termination are used in Amiga
systems.
So here are ways to cause trouble on your Amiga SCSI bus.
A. Don't have termination at the physical ends of the bus.
(Termination resistors must be installed and TERM POWER applied
to those resistors).
B. Have termination somewhere other than at the ends of the bus,
(If your SCSI host adapter is not at one end of the bus,
remove those termination resistors from the host adapter).
C. Use cables which are not appropriate for SCSI.
(Cheap RS-232 cables may have all 25 lines, but may not be physically
twisted so that the line impedance is something near correct).
D. Install your termination resistors the wrong way 'round.
(Pin one of the resistor socket is usually marked by a square
pad around the trace. Pin one of the resistor pack itself
is marked by a dot or line at one end of the pack).
E. Don't apply power to the termination resistors.
(The IOmega ZIP drive seems not to be able to supply term power
to its built in termination. Term power must come through the
SCSI cable).
Some Amiga owners report that their systems violate one or more of the
above precepts, and everything still works OK.
I say, "If it ain't broke, don't fix it." If your SCSI system works, don't
change anything.
But if you are having trouble, getting the termination set correctly may
solve your problems.
======================================================================
Much more information about the SCSI bus is available on the Internet.
The SCSI-FAQ can be grabbed by anonymous FTP from ...
ftp.wang.com:
garyf/scsi/scsi-faq.part*
OR
rtfm.mit.edu:
pub/usenet-by-group/comp.periphs.scsi/comp.periphs.scsi_FAQ_part*
where the asterisk indicates the part number of the FAQ. (1 or 2).
OR
on the World Wide Web at:
http://www.cis.ohio-state.edu/hypertext/faq/usenet/scsi-faq/
OR
http://www.paranoia.com/~filipq/HTML/LINK/F_SCSI.html
======================================================================
[Contents] Amiga Report Main Page
Amiga Web Directory
HTML Conversion by AG2HTML.pl V2.951201, perl $RCSfile: perl.c,v $$Revision: 4.0.1.8 $$Date: 1993/02/05 19:39:30 $
Patch level: 36
& witbrock@cs.cmu.edu