From: "Fred R. Goldstein" <goldstein@delni.enet.dec.com>
Subject: Re: What The Heck is "Fiber Optic Quality" Anyway?
Date: 24 Jun 91 17:01:08 GMT
Organization: Digital Equipment Corp., Littleton MA USA

In article <telecom11.482.8@eecs.nwu.edu>, riacs!rutgers!ttidca.tti.
com!jackson@decwrl.dec.com (Dick Jackson) writes...

> In other words fiber is no better than other MODERN digital facilities.

Yes and no.  For ordinary voice users, that's probably true: A copper
or radio digital transmission facility is usually tagged as in trouble
when its bit error rate is anywhere near the level needed to make PCM
audio audibly degraded.  And that's not common.  But it can occur:
Radio links (including satellites) are subject to rain fade.  That
raises the raw BER a bit.  Minor fade, however, is usually compensated
for by trellis coding, a form of forward error correction.  Severe
fade can lead to real errors, though.  Microwaves are not happy when
it rains _very_ hard.  System designers generally account for some
probability of this, but you can never say never.

Copper links (like T1) are subject to electrical impulse noise.  If
lightning strikes nearby, a pole line can pick up a hit.  Again this
is unusual, but it can occur.  It also matters how marginal the link
is in the first place.  Motors and other such devices can impair some
marginal copper links.  The nature of these bursts varies with the
cause and with the type of transmission system.

Fiber optic is usually transmitted in raw NRZ format, perhaps
scrambled, where each bit is represented by an on or off.  (It is not
group coded like most radio and some copper systems, where a
transmission symbol represents several bits.)  It is immune to
electrical noise or rain fade.  (Backhoe fade, of course, is harder to
avoid :-(. )  When it works well, the BER is often on the order of
10^-12, versus 10^-8 or so on many copper systems.  That's the
difference between very good and very, very, very good.  Because it's
not group coded, most errors only affect a single bit at a time,
rather than a cluster of bits as might be hit on a modem or digital
radio link.  It's gaussian noise, and the S/N-ratio to BER curve can
be taken from a simple graph in Mischa Schwartz's textbook on noise,
whose title escapes me now (I've got the new edition at home).

This all can be significant when designing error detecting and
correcting techniques.  What works best for a _pure fiber optic_
network may not be ideal for a mixed-media digital network.


Fred R. Goldstein              Digital Equipment Corp., Littleton MA
goldstein@delni.enet.dec.com   voice: +1 508 952 3274
 Do you think anyone else on the planet would share my opinions, let
 alone a multi-billion dollar corporation?