Anyone in AMRAD ACTIVELY messing with HSMM, possibly using Ubiquiti?

[Mike O'Dell]

Terry - you should talk to the folks at Netgate.com

Jim Thompson is one of the most experienced Wifi designers in the
business. He's designed or consulted on the design of a very
large fraction of the kit out there.

The reason you don't read much anymore about "mesh" networks is
that single channel mesh networks don't work very well at all.
if you try multichannel systems, you'll discover that the CMOS
Zero-IF Wifi chips have all the selectivity of the door of a large
airplane hangar. If you have wifi things on the same *band*, if
they can hear each other at all, put them all on the same channel.
any attemp to use more than one channel in a band (without Herculean
efforts - ie, deep cavity bandpass filters and shielding the chip)
is pointless. It has no in-band interference rejection. The good
chips out there can work on 1 channel of 2.5 and 1 channel of 5.0
at the same time, but that's it for multi-channel.

The Metricom Ricochet system was the early precursor to the 
wifi-style mesh networks, but it had the advantage of *very*
good RF design and operating on 900MHz. It was 20 years ago now,
but seeing 19.2Kbps throughput over a sustained period was ground for
considerable celebration. 

Ricochet-2 had radios that used 900 for pole-top to terminal and
then used 1.7GHz (i think) for hop-by-hop return trunking but the
big base stations had 2.3GHz radios that could run high-power for
skipping over intermediate hops in the outbound direction (toward
the terminals). That did a lot to address the traffic concentration
problem around the "fixtures in the bathtub"- all the packets had
to get out of and into the regional mesh through the hub so the traffic
concentration caused big problem.

About $400mm was raised for Ricochet-2 deployment - the 35 "NFL Cities" were
the first targets. The problem was that after the first 3 or 4 cities
were built-out, it became clear the unit economics just didn't work.
The Customer Acquisition Cost (CAC) was too high to ever make any
money on a subscriber, and at that time (timing is always a factor),
not enough things were able to take advantage of mobile connectivity
so the demand wasn't there. Even if it were there, the actual field
performance was significantly less than predicted. Part of the problem
was again the topography. Topo maps seldom have enough detail to
do high-precision propagation modelling at the "in the treetops"
level where the hardware really was. Even "small" variations in
topography produced big holes in the local field strength and 
coverage. The difference between Leaves and No Leaves was also much
bigger than predicted, so even with the new radios, the link
budgets were still only marginal.

so Metricom stopped Ricochet-2 after doing the first few city
builds, gave the remaining money back to the investors, folded their
tent and left town. Nice try, but no moko. 

I highly commend you for your enthusiasm and experimental work,
but don't get your hopes up too high lest you be disappointed.

    -mo