MAC sublayer protocol

[Glenn Elmore <glenne@hpnmdlc1.hp.com>]

> Steve Johnson writes:
> >Are there other reasons besides spectrum usage that split frequency
> >full duplex systems aren't used on packet?  It does seem the simpler
> >approach if enough spectrum is available.

Barry VE3JF writes:
> I think the major reason they aren't more common is simply inertia, and
> perhaps some apprehension that they are difficult to build (they aren't).
> I don't believe spectrum usage is a major factor.  With various commercial
> interests encroaching on our bands in the next few years, heavy spectrum
> occupancy should have priority over maximum efficiency (within reason, of
> course! :-).

Well, as one trying to whack hardware into bit pumps over radio I have a
slightly different view.
  Full duplex is no cakewalk.. particularly for wideband/faster systems.
Not only is the spectrum coordination issue more confusing but the
hardware itself is *considerably* more difficult.
  While I know that at higher layers duplex is attractive and can offer
more than 3 dB increase in throughput, in almost any case I can think of
I would prefer to find ways to increase simplex channel capacity by a
factor of at least *10 times* before I would take on making it full
duplex.
  Not only does the radio have to be clean:  discrete spurs and
broadband noise have to be reduced and controlled to a much greater
degree in both the Tx and Rx, but even if the RF designer does an
impeccable job all it takes is a nonlinear element located close to a
shared high level site (like where this thing is being used as a
backbone) and all of a sudden you have intermodulation components in
your receive passband.  It really is almost as bad as "a rusty nail" can
put you off the air.  Things get worse as channel width increases and
offers more vulnerability in this regard.
  Basically, doubling the total channel width for FDX can far more than
double the opportunity to get clobbered and at the same time greatly
increases hardware difficulty and expense.
  Increasing channel capacity is at worst a fairly linear expense.  The
10 GHz hardware will easily run a good deal faster than the 1 Mbps that
the PCLAN Adapters could handle.  Even the cheapo NEC transceivers would
tx/rx at 8 Mbps.  I'm sure that MaComm Gunnplexers will run full
Ethernet.
  Similarly, I have slowed down the 900 MHz radios from 500 kbps to 256
kbps because of i/o hardware and spectrum availability.  They now fit
into a 1 MHz channel and we are able to double the number of clusters on
the band in a given area.  This was particularly an issue since I have a
stash of very cheap 10 MHz wide ceramic filters centered on 904 MHz and
only a single 2 MHz chunk of spectrum available there for higher speed
digital radios.

73
Glenn Elmore
de n6gn,  over.....