At 07:41 PM 3/22/2006, [EMAIL PROTECTED] wrote:
Dear Great Minds

Not strictly a Beowulf question but I suspect knowledge on the topic lurks herein.

I can't seem to find a straight answer on the "real" *minimum* cable length for
gigabit ethernet. I'm looking to make things on my new toy nice and tidy.

Assume Cat5e/Cat6 cable, Intel e1000 NICs pushing the data rate to 'maximum',
jumbo packets, cut through switching, titanium alloy mounting bracket screws
etc. Minimum physical target length of 200mm.


I can't see why you couldn't interconnect two physical interfaces with a zero length cable: that is, back to back line drivers on the same PC board.

Nothing special with the interconnections, they're just balanced pair transmission lines with nominal impedances of around 110 ohms.

As a practical matter, light speed is 1ns/ft in free space. Probably 75% of that in a real cable with real dielectrics. so, your 200 mm is right on 1 ns, or one bit time.


>From my limited understanding of such things, surely there would be some sort of
electrical issue with too short a cable corrupting signalling/bits etc?

nope.. can't have transmission lines too short.

The Net seems to vary between selling you a 0.5m cable v's "never use less than
1.0m" v's "there's no practical limit to how short".

Thoughts and comments appreciated. Worse case I'll prolly try a 1m v's a 0.25m
and check error rates :)

Good idea.. there might be some sort of weirdness if you happen to get just the wrong length. There is some amount of discontinuity at the connectors, and if the cable happens to be just the wrong length, then the discontinuities at each end form a resonant circuit. Back in the "good old days" of Ethernet, the cable actually had stripes on it where it was "legal" to stick in a vampire tap, just for this kind of reason.


James Lux, P.E.
Spacecraft Radio Frequency Subsystems Group
Flight Communications Systems Section
Jet Propulsion Laboratory, Mail Stop 161-213
4800 Oak Grove Drive
Pasadena CA 91109
tel: (818)354-2075
fax: (818)393-6875


_______________________________________________
Beowulf mailing list, Beowulf@beowulf.org
To change your subscription (digest mode or unsubscribe) visit 
http://www.beowulf.org/mailman/listinfo/beowulf

Reply via email to