Path: utzoo!utgpu!watserv1!watmath!att!att!emory!samsung!cs.utexas.edu!hellgate.utah.edu!basset.utah.edu!haas
From: haas%basset.utah.edu@cs.utah.edu (Walt Haas)
Newsgroups: comp.dcom.lans
Subject: Re: networking on cable TV systems
Message-ID: <1990Oct23.091334.246@hellgate.utah.edu>
Date: 23 Oct 90 15:13:34 GMT
References: <6722@gaboon.UUCP> <149.2705d73f@stat.appstate.edu> <1649@syteke.be>
Organization: University of Utah CS Dept
Lines: 64

In article <1649@syteke.be> jim@syteke.be (Jim Sanchez) writes:
>I would take STRONG exception to the the proposition that a CATV
>system must be tuned to perfection to be used as a network medium.
>Since my company (formerly Sytek) started out in the broadband (CATV)
>networking business and have by far the largest installed base on this
>medium, I think I can comment...

Well since we are satisfied users of the Hughes 8200 (formerly Sytek)
token bus bridges to connect Ethernet over broadband, maybe our
experience would be useful to somebody.  We had an old cable TV system
which we upgraded to a 450 MHz high-split system for the purpose of
running data.  We do in fact successfully bridge our campus Ethernets
today over this system, and it works pretty well.

The main thing we had to do is drastically raise the engineering standards
used by the TV guys.  When I took over this system it ran TV just fine,
and ran data well enough for people to try to rely on.  Unfortunately it
did not run data well enough for users to be satisfied.  There were two
main reasons for this:

1) TV can tolerate large errors in system tuning.  The reason is that
   TV signals are on constantly, whereas data signals using most modulation
   schemes (such as the 802.4 used by the HLS 8200) have many devices on
   one channel that share the channel by turning their carriers on and off.
   TV receivers have Automatic Gain Control which is good at adjusting input
   levels to compensate for alignment errors.  AGC is impossible when signals
   are turned on and off every few microseconds.

2) TV is generally immune to ingress.  Our HLS 8200s are run at the factory
   default frequency, so the top half of their range falls on the lower half
   of the FM band.  If we pick up a strong enough FM signal on our reverse
   channel it will render the 8200s unable to communicate.

When I took over the broadband, it had large alignment errors and ingress
problems.  The only data application which worked reliably was some point-
to-point RF modems which did not rely on switching carriers on and off, and
which therefore were able to make effective use of AGC.  These modems also
were on a frequency which did not experience significant ingress.

We now have a system with no alignment errors of more than 2 or 3 dB, and
ingress at least 25 dB below nominal reverse channels.  It took a lot of
work to get to this point, but we were rewarded with a drastic improvement
in data reliability.

If you have to live with a TV system, one approach you might want to take
is to do the following:

1) Closely investigate the ingress situation.  See if there are some free
   reverse channels that don't suffer ingress.

2) If there are alignment errors, get some point-to-point RF modems that
   have effective AGC and use them instead of a technique that switches
   the carrier on and off.

The other thing that limits the reliability of a broadband system is, of
course, the large number of active electronic components that need to be
working in order to get from point A to point B on the system.  It is pretty
standard for a typical path on our system to involve 15 or 20 active
components.  We will be moving to fiber optic over the next several years
for both increased reliability and increased bandwidth.  That isn't to say
anything bad about our HLS 8200s, which work fine; if you want to understand
the reliability issue you just need to count boxes - and think about ingress.

-- Walt Haas    haas@ski.utah.edu