[e2e] Bandwidth Estimation workshop

[David P. Reed]

At 10:38 AM 10/3/2003, Susan Harris wrote:
> > I believe you but I am surprised.  Is the distinction so hard to
> > explain?  If the discussion about units doesn't enlighten them, what
> > about the fact that often (all other things being equal) it's
> > desirable to minimize one and maximize the other?  (This won't explain
>
>Beginner's question - why is it desirable to minimize the Hz variety of
>bandwidth?

Well ... this is actually not so simple, because the "good" direction on 
bandwidth (in Hz) of a signal is not obvious at all.   Conventional 
(Marconi-style, early 20th century) radio architecture is based on the 
premise that a single transmitter at a time radiates in a small contiguous 
band of frequencies.  A receiver listens with a band-pass filter that shuts 
out all but that band.   In order to allow as many simultaneous users as 
possible, such systems are "better" when each transmitter uses as narrow a 
band as possible.

However, there are significant drawbacks to this conventional architecture 
for modern networking.

1) It doesn't scale.
2) It's incredibly inefficient as a "system" when there is rate variation 
and switching involved.
3) It's very susceptible to many kinds of noise and propagation problems.

Wider bandwidth radio architectures can fix all of these and more.   That's 
what spread spectrum, code-division multiple access, orthogonal frequency 
division multiplexing, and ultrawideband technologies all do, in various ways.

There's lots more to say, but perhaps you don't want to know.

In general, I would say that the wider the bandwidth of a signal, the 
better the system can be, at least these days it seems to be true.   What 
gets in the way is the complexity of the hardware, which roughly follows 
Moore's Law over time.