Path: utzoo!attcan!uunet!lll-winken!ames!mailrus!ukma!gatech!bloom-beacon!bu-cs!kwe
From: kwe@bu-cs.BU.EDU (kwe@bu-it.bu.edu (Kent W. England))
Newsgroups: comp.dcom.lans
Subject: Re: Maximum Ethernet length
Summary: It depends on delay and attenuation
Message-ID: <27196@bu-cs.BU.EDU>
Date: 12 Jan 89 14:31:17 GMT
References: <10315@well.UUCP>
Reply-To: kwe@buit13.bu.edu (Kent England)
Followup-To: comp.dcom.lans
Organization: Boston U. Information Technology
Lines: 46

In article <10315@well.UUCP> David Vezie (dv@well.UUCP) writes:
>
>	Can someone give me the definitive word on how long Ethernet
>can be?  I've looking at the possibility of running standard thick
>yellow ethernet cable between two buildings, with a sun acting as a
>router on either side.  I've read some info which says that the max
>is 500M, while other info says that the max is 1000M, and the distance
>is between the two (probably about 550-600M).  Also, what would happen
>if I exceeded the length?  Would it not work, or would it be flakey,
>or ???
>
	The maximum round trip delay between any two transmitters on
an Ethernet is 464 bit times.  This is equivalent to about 5000 m of
coax cable or about 4.3 miles of free space (like microwave).  I don't
know off-hand how much fiber that is equivalent to, but it should be
something like 2000-5000 m.
	Repeaters and multi-port transceivers add to this delay
factor.  Multiport transceivers are equivalent to about 50 m of coax. 

	The transceiver on a coax cable is only capable of driving a
signal down about 500 m of thick coax, 185 m of thin coax.  Fiber
optic semiconductors are able to drive standard multi-mode fiber about
2000 m.  DEC recommends no more than 1000 m of total fiber path in any
repeatered network, but this is a delay limit.

	So path lengths are limited by delay or by signal attenuation,
whichever comes first.  If you violate delay, you will have undetected
collisions on the net, which will cause packets to be lost
unknowingly to the transmitter.  This will definitely reduce
throughput for some stations on your net.  If you violate the signal
specs, you will not be able to make the net work at all between certain
nodes.  In general, these problems are hard to find, since the effects
vary according to the traffic patterns.

	Another point you should keep in mind is the grounding
requirement for baseband coax cable.  Ground a thick cable in
*exactly* one location.  If you run thick coax between two buildings
whose AC power feeds result in different ground potentials, you will
have ground loops, and your Ethernet will not work.  However, many
people have blissfully installed thick between buildings (including
me) without any problem whatsoever.  It is definitely economical in
some cases.

	Kent England, Boston University

(Was that as good an answer as Jim Warner could give?  :-)