Path: utzoo!mnetor!uunet!husc6!cfa!ward
From: ward@cfa.harvard.EDU (Steve Ward)
Newsgroups: comp.dcom.lans
Subject: Re: Thinwire ethernet: Question about drop cables
Message-ID: <951@cfa.cfa.harvard.EDU>
Date: 26 Mar 88 14:03:19 GMT
References: <4491@june.cs.washington.edu> <13328@uflorida.cis.ufl.EDU> <5948@cit-vax.Caltech.Edu>
Organization: Harvard-Smithsonian Ctr. for Astrophysics
Lines: 70
Summary: a problem in theory somewhere
	 Thin E-net T-connector drop cables

In article <5948@cit-vax.Caltech.Edu>, mangler@cit-vax.Caltech.Edu (Don Speck) writes:
> Simple transmission line theory.  When the wave travelling down the
> Ether hits a tee connector, the energy splits, with 4/9 of it going
> down the drop cable, 4/9 continuing on the main cable, and 1/9
> reflecting back.  The waveform loses 1/3 of its voltage.
> When the energy going down the drop cable reflects off the end and
> rejoins the main cable, the voltage goes back up.  Thus the waveform
> has a notch in it, whose duration is twice the round trip time of
> the stub.  Each time the waveform passes another stub, the notch
> gets wider and deeper.	When the width of the notch approaches half
> a bit time (i.e. 50ns for Ethernet, 500ns for Starlan), the signal
> will have reversed itself by the time the reflections add back in
> (destructive interference), and things will go downhill fast from
> there.
> 
> That means the combined length of all stubs is limited to about 3 meters.
> If you have 100 taps on the cable, that's 3cm per stub (about what you
> get from a single tee and BNC connector).  With only 10 stations, you can
> have one-foot drop cables.
> 
> It might help slightly to use high-impedance drop cables (video cable
> or IBM 3270 cable) to reduce the depth of the notch if you're going to
> cheat by hanging cables from a tee.
> 
> If you need a long drop cable someplace, make it one end of the main
> cable and terminate it at the workstation.  (Fortunately, cables have
> two ends, so even if all your drop cables need to be long, you should
> still be able to get two stations per thin ethernet segment).
> 
> Don Speck   speck@vlsi.caltech.edu  {amdahl,ames!elroy}!cit-vax!speck


There seems to be a problem in the predictions/calculations based on the
theory presented here based on my own recent experience and most
certainly if the many who told of their experiences are accurately
describing their drop cables, many seem to VASTLY EXCEEN 3 meters in
total drop cable length.

As I previously stated, I use drop cables about 12 inches in length, but
admittedly the total drop cable length on the thin segment is right
around 3 meters, not vastly exceeding it.  The segment looks fine with
TDR, E-net tester, and LAN analyzer shows that the segment is behaving
normally (no pathologic repeat packets due to noise, point-to-point
tests error free, etc).  But some of the others posting to this
discussion were talking about 3-12 foot long drop cables, with multiples
of them on a segment, tallying what must be many, many times the 3 meter
failure limit described above.

What really bothers me is that I installed the 12 inch drop cables based
on guidelines in some manufacturer-supplied documents.  I guess I am
now sufficiently motivated to go back and peruse the bookshelf and find
the information.  I will post quotations/citation in the next few days.
As memory serves me the information stated a 27 inch limit on
t-connector drop cables without mentioning any mandatory reduction in
number of connections per segment, or reduction in segment length, which
now seems incorrect in light of these discussions (the quoted message
and others).

Perhaps some of those who are using t-connector drop cables,
particularly long ones, can add the total drop cable lengths, number of
drops, and total thin E-net segment lengths and post this info.  I will
do this as well.  It would be useful to compare notes here.  Also, it
would be useful to know if you analyzed/tested the segment with a
TDR (time domain reflectometer) or other E-net cable tester and/or
performed point-to-point testing on the segment and/or monitored the
general traffic on the segment using some sort of LAN analyzer.


d