On Fri, 26 Sep 2008 06:30:11 -0500, Bob Bob
wrote:

Okay well the answer is relatively simple.

Nothing in antenna design is simple. Mind if I disagree with a few
details?

"Height is might". Antenna system should be as high as possible, clear
of path obstructions and provide gain in the direction of interest.

I beg to differ, having established one (failed) WISP a few years ago.
The higher you place the antenna, the more interference from other
2.4GHz users you'll receive. The license free bands are really
polluted. Once you get high enough to get adequate coverage of the
client radios, any additional height is counterproductive as it only
increases interference pickup.

Clear of path obstructions is correct, but difficult to accomplish.
The most common screwup is insufficient Fresnel Zone clearance,
especially when it hits the ground. For example, let's pretend our
WISP is running out if his house to a similar client radio, both with
rooftop antennas. That would be about 7 meters off the ground.

Quiz: At 2.4GHz how far away can these they be before an 80% Fresnel
zone hits the ground?

Answer:
http://www.terabeam.com/support/calculations/fresnel-zone.php
Plugging in various ranges until I get a 7 meter Fresnel Zone radius,
I get about 2.5km range. Any further and the antenna at one (or both)
ends will need to be elevated.

For
equal coverage in all directions this generally means a vertical
collinear design. The idea is to have a very small angle vertical
pattern as close to horizontal as possible. (ie radiates most of its
power to the horizon)

I again beg to differ. Vertical collinear antennas suck for WISP
service. Lots of problem. On omnis with sufficiently high gain, the
extremely narrow vertical radiation pattern, and the apparently
incurable uptllted pattern, results in sending the RF to various
useless directions. There are antennas with a few degrees of
downtilt, but my testing shows that they still have uptilt when
mounted on a tower outrigger.

In my never humble opinion, the proper antenna for WISP service is a
90 or 120 degree sector antenna. These offer a solution to the uptilt
problem in that they can be pointed downward without having most of
your RF go into the sky on the backside. See photo at:
http://www.qsl.net/yu1aw/3InvAmosa7.JPG
These are 3 antennas, connected to 3 different access points. However,
the 3 antennas could just as easily be run by a single access point,
by using a combiner/splitter, yielding plenty of gain with the
advantage of downtilt.

In my experience, end fed (ie at the bottom) collinear systems
(especially home made ones) tend to have current imbalances in the
elements such that the lower ones get slightly more than the upper ones.
the result is that the pattern skews undesirably upwards at a slight
angle. You also get less gain per unit length using systems that feed
from one element to the next rather than (say) a number of dipoles fed
directly in parallel. There is an optimum spacing between elements for
best gain that is difficult to get in element to element systems.

Yep. However, it's not an "imbalance" that causes the problem. End
fed vertical collinear antennas belch half of the RF from the first
1/2 wave dipole near the connector. 1/4th or half of what's left
comes out next half wave section. 1/8 of the RF from the next section
and so on ad infinitum. what that means is that there's not much RF
left by the time the you get to the tip of the antenna. It's for this
reason that I like to mount my end fed omnis upside down on the tower
outrigger. This is often vetoed by the tower aesthetics committee but
when I get my way, it works very nicely.

Incidentally, the uptilt problem is mostly solved by some commercial
omni antennas with a center fed collinear dipole arrangement inside
the fiberglass tube. This is common at VHF/UHF frequencies but I
haven't seen it at 2.4GHz.

Of course 2.4GHz antennas are harder to size accurately and test
equipment can really drain the hip pocket!

Bah. The stuff costs money but there's no need to buy state of the
art hardware. A good RF sweep generator, direction coupler, diode
detector, general purpose oscilloscope, marker generator, and a mess
of coax cables, adapters, and terminators, are usually sufficient:
http://802.11junk.com/jeffl/pics/home/slides/BL-shop5.html
The HP8620A sweep generator is fine (but my 2-4GHz plug needs work).
There are two Wiltron 610D sweep generators behind the pile. I don't
recall what I paid for all the sweepers, but it wasn't huge.

The real key to sweeping the antenna is in the directional coupler or
VSWR bridge. Good ones are expensive and easy to destroy. So, build
your own for just 2.4Ghz:
http://pe2er.nl/wifiswr/

If I was doing it for a community system I wouldnt bother doing antenna
construction as such.

Maybe. Commercial sector antennas are somewhat expensive.
http://www.superpass.com/2400-2483M.html
It's a marginal proposition to build your own for just the base
stations. However, if I were supplying client radios for the
customers, I would certainly consider rolling my own to save a few
dollars per install.

I dont know the
2.4GHz numbers off hand but the range numbers might be something like
5km, then 5.1km with an extra 6dB. Of course those users that have a
signal already will get a slightly stronger one and thus may get a
higher data rate but there will be a point where more gain and a
different system design is better. Using multiple channels and WAP
repeaters or bridges/links although likely bringing the bandwidth down
will give you greater coverage albeit sometimes at the tradeoff of speed.

Using repeaters (range extenders) or mesh networks also has the
disadvantage of hogging air time. With a single channel store and
forward repeater, it takes twice the air time to move a single packet
between link endpoints. That's not a big deal in an isolated client,
but with a heavily used and high access point, it can easily be
considered a form of interference. Wireless is a shared medium. If
there are packets flying through the air because of repeaters, they
will reduce the available air time rather rapidly. Such repeaters
also have a lousy delivery success rate, resulting in retransmissions,
which reduce the available air time even more.

Speed is also an issue, but for a non-obvious reason. In such a
shared environment, you want each transmission to occupy as little air
time as possible. That means you have to send your data as fast as
possible. Signal strength and maximum speed are directly related (all
else being equal). The last thing you need is a customer stuck at the
minimum 1Mbit/sec and monopolizing a disproportionate amount of air
time. I setup mine to do 802.11g OFDM modes only and to ignore all
802.11b speeds. That gives me a 6Mbit/sec minimum OFDM speed, which
uses approx 6 times less air time than 1Mbit/sec. It also changed
having all the beacons and broadcasts from 1Mbit/sec to 6Mbits/sec.
When possible, I like to use fixed data rates, typically 12Mbits/sec
or more.

I dont have any modeling stuff for 2.4GHz networks as such. You tend to
need a lot of obstructive rather than terrain data information. You can
however take some photos and hand enter some near field terrain info to
look like buildings. Have a look at "Radiomobile" as a prediction tool.
As far as I remember it is GPL software that you will find other users
have used for WiFi networks. You can see the effects of gain and height
very easily and it will give you a lot of go/nogo help for particular
areas you want to supply access from.

Yep. Great program. I use it often:
http://www.cplus.org/rmw/english1.html
However, it's only as good as the topographic data. Most commonly
available (for the US) are the SRTM v2 data at 1 arc second (30
meters). That's good enough for doing general topography (mountains
and valleys) but not good enough for doing urban canyons and small
neighborhood WLAN's. Data for the rest of the world varies by
country, but most of it seems to be 10 times less resolution.

Apologies for the length!

I won't apologize for my length. I'm partial to details and
explanations over the all too common one-line quips and cute remarks
devoid of any useful content.

--
Jeff Liebermann
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558