Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!evax!utacfd!merch!cpe!adaptex!adaptx1!neese
From: neese@adaptx1.UUCP
Newsgroups: comp.periphs.scsi
Subject: Re: Single-Ended vs Differential
Message-ID: <283400067@adaptx1>
Date: 11 Mar 91 17:42:04 GMT
References: <3246@legs.UUCP>
Lines: 53
Nf-ID: #R:legs.UUCP:3246:adaptx1:283400067:000:3211
Nf-From: adaptx1.UUCP!neese    Mar 10 14:51:00 1991


>/* ---------- "Single-Ended vs Differential" ---------- */
>I'm hearing a lot of differing views on single-end vs differential
>SCSI cables.  Some say you can do 10MB on SE cable, others say you
>can go 75 feet on SE cable, and some say you can go 10MB AND 75 feet
>on SE cable.  What's the real story here.  Inquiring minds want to
>know!

Never mind what people say.  The spec allows a single-ended SCSI bus
to be up to 21 feet long.  Going poast that is at your own risk.  The
initial speed ceilings and bus lengths were not chosen out of the air.
They are and are based on known technological reasons and theory.  To
violate these specs put your data at risk.
Now that the SCSI-2 spec is firm, there are a lot a folks who are anxiously
waiting for those FAST SCSI-2 devices (10MBytes/sec).  They are in for quite
a few surprises and problems.  I can see some user taking the adapter out
putting in a FAST adapter and hooking up thier external FAST drives with the
cables from the old setup.  When problems start occurring and there is a
good chance there will be problems, they will call either the drive company,
or the adapter manufacturer and complian long and loud.  Not realizing the
true potential problem is the cable.  Even if it worked fine on the slower
SCSI-1 bus, it may not be good enough for the FAST SCSI-2 bus.  The problem
lies in many of the SCSI cable manufacturers.  Cable design is a true
science and some would say an art as well.
I have seen external cables of all types of designs.  Some do not use
twisted pairs internally.  Some do, but do not properly route the pairs
internally.  Some have the cable pairs just bundled together in an arbitrary
manner inside.  Some have a shield around the bundle, and it may or may
not be properly grounded.  To say it is a bag of worms is a slight
understatement.
Anyway, I digress, in the SCSI committee there is a proposal on the table
for a new method for single-ended termination.  One that would be compatible
with the current spec, but would allow for higher noise immunity and better
impedance.  With this type of termination, one could feel comfortable
putting a FAST SCSI device on the external bus.  It costs more to implement
this type of termination, but is still cheaper than differential and does
not require the extensive board real estate that differential does.  A happy
medium as it were.
I would not put a standard single-ended SCSI device on an external cable on
a bet.  When I say FAST in this context I mean 10MBytes/sec.  FAST does not
imply 10MBytes/sec.  It does imply faster than 5MBYtes/sec.  You will see
many devices in the near future claiming to be FAST SCSi and they are right
to do so as they will be running 6.xMBYtes/sec.  This qualifies as FAST SCSI-2.
Of course, none of it means diddly.  The continous data rate is the real
trick.  Bursting data at high rates is one thing, but sustaining those rates
are something else.
Internally, you can get away with the current single-ended design.  The cable
is normally short (< 5 feet).  It is normally a flat (50 wire) ribbon cable,
which is a very good cable.  Enough for now.

			Roy Neese
			Adaptec Senior SCSI Applications Engineer
			UUCP @  neese@adaptex