[UFO Chicago] Wirelessness in Chicago

[Neil R. Ormos]

Nate Riffe wrote:

> I had a meeting [ . . . ]  We talked about various ins and
> outs of community wireless networking, including
> technicals, the regulatory environment, possible
> organizational structures, and politicization.  [ . . . ]

> The long-term goal of this network is cheap, ubiquitous,
> and wireless Internet access throughout the city. 
> [ . . . ]

> In particular, this network will not provide Internet
> access for quite some time.  Legitimate Internet access
> for a network such as this will not be cheap, which is one
> the main goals.  When the network is sufficiently large
> (and by large I mean in coverage and bandwidth, NOT in
> users), we will have the means to negotiate for cheap
> Internet access.  Until then, it is an infeasibility.

Why would coverage, as opposed to N of users, help you
negotiate cheap internet access?  For that matter, unless
you plan to charge users on a per-byte basis, how would
higher bandwidth help you negotiate cheap internet access?

> On the subject of technology, the issues are still what
> they were last summer.  802.11b (WiFi) networks operate in
> the 2.4 Ghz unlicensed band [ . . . ]

> Water reflects signals in this band, which means things
> like vegetation, weather, and large, thick flocks of
> pigeons will adversely affect network performance and even
> network availability. [ . . . ]

I believe the primary mechanism is absorption.

> The FCC regulations for the 2.4 Ghz ISM band make very low
> allowances for signal power. [ . . . ]

> The effective radiated power (ERP) of an omnidirectional
> 2.4 Ghz radiator must not exceed 1 watt (30 dBm).  The
> effective radiated power of a directional radiator must
> not exceed 1 watt for any antenna with up to 30 dB of
> gain, and for every 1 dB above 30 dB of gain, the maximum
> allowable ERP is raised by 3 dB (this is actually a very
> nice gift from the FCC).

Unless the rules have changed since October 2001, that's not
quite right.  The EIRP of a directional radiator must not
exceed 1 W for antennas having up to 6 dBi of gain.  Then,
for every 1 dBi above 6 dBi of antenna gain, the maxiumum
allowable EIRP increases by 2 dBi.  (This is a consequence
of the rule that provides "maximum peak output power of the
intentional radiator is reduced by 1 dB for every 3 dB that
the directional gain of the antenna exceeds 6 dBi".  37 CFR
15.247 (b)(i).

In addition, it's not a "gift" from the FCC either; it's
simply a common-sense regulatory quid pro quo.  The point of
the regulations is to minimize interference to other users,
including licensed users.  Directional antennas concentrate
radiated energy in the desired direction, which means that
the radiated signal is less likely to interfere with users
in any other direction, and in any case, the total radiated
power is reduced with higher antenna gain.

> So, knowing all that, how far can you make a signal go?
> [ . . . ]

> [Deletia: List of materials and corresponding attenuations]

Note that the list might be useful in back-of-the-envelope,
aggregate planning, but for any particular RF path, the
actual attenuation is likely to be quite different than
expected.  To the extent there are conductive surfaces
involved in construciton, attenuation will depend, in large
part, on the number, size, and geometry of apertures in the
surface.  Diffraction should also be considered.  One thing
that hasn't been mentioned is that older buildings often
have plaster walls built on a metal lathe (mesh) substrate,
which produces very high attenuation.  Newer buildings
sometimes have an aluminum-foil-backed sheath under the
exterior walls (it's either a vapor barrier or a radiant
energy insulator--I forget which) that also causes high
attenuation.

> [ . . . ]  Attenuation of a 2.4 Ghz signal through open
> space follows the formula 100 + 20 * log(d), with d
> measured in kilometers, so the free space loss over 200
> meters will be about 67 db.

I believe that log() function referenced in the formula is
the common logarithm (base 10), so free space path loss over
200 M would be about 86 dB, according to the formula.

I've seen this formula, too, but I'm sure it doesn't tell
the whole story.  Hams have reported exceptional 1500 KM
contacts over land and nearly 4000 KM contacts over water at
2.3 GHz, and that would correspond to -171 dB path loss.
Hams can use higher power, bigger antennas, and better
receivers, but that's still a lot of attentuation to make
up.  This would suggest that not all propagation is
consistent with the free-space model.

> [ . . . ] There are some possible solutions, however.  We
> could use higher gain antennas, but antennas with a gain
> higher than 35 dB are expensive, difficult to build, and
> VERY difficult to aim.  We could use amplifiers, but we
> would soon be in violation of FCC regulations, since the
> limit on ERP is 30dBm.

Actually, if the goal is "cheap", antennas with 27-30 dBi
gain are likely to be at the practical upper limit.
Commercial rectangular-cut-out paraboloid mesh reflector
antennas (24 by 40 inches) in the 24 dBi range are $150-200.
Commercial dish antennas (48 inch) in the 27 dBi range are
$300-400.  You could probably make your own 27-30 dBi dish
for $100-200 (maybe less if you're very handy and have
access to good tools), but it would be large, and you'd have
to figure out how mount it securely on a roof, etc.  You can
build 15-18 dBi gain antennas, in a variety of
configurations, for under $30.

You could also try to buy better receivers, but that
probably means you can't use cheap 802.11(b) adapters, as
you'll either have to buy purpose-built equipment or wait
for the performance of the cheap stuff to improve.  -85 dBm
seems pretty lousy compared to the sensitivity of
communications receivers.  I would think -100 to -110 would
be achievable without great expense.  Too bad Bob Parnass
has left the list, as he could comment on what's reasonable
here.  Another issue is, IIRC, the receiver sensitivity is
specified for 11 Mbps, but the system is operable over lower
quality links at degraded speeds.  So even if it appears
that the link is at the edge of the loss budget, you may
still have useable service.

> We could put the antennas on our roofs, which would reduce
> the number of obstructions.  If we could prop them up so
> that there was nothing but open space between the
> antennas, then the obstruction would be free space loss
> (67 dB), but with all the cable between the base stations
> and their antannas, cable loss (with varies depending on
> the quality of the cable) is also a major factor.

The obvious solution is well-placed antennas.  15-18 dBi
gain antennas can be biult pretty easily for a few bucks.
Good, low-loss transmission line is expensive, so it would
help if the equipment were located near the antenna.
Typical wireless LAN cards don't provide anywhere near 1 W
output, so you have that working against you, as well.  It
would probably be optimal to put the transceiver right at
the antenna, if you can find some way to protect it from the
weather.