Re: 3 km wi-fi link, how?

On 27 Nov 2005 15:29:29 -0800, "Demetrio Girardi"
<dementrio@xxxxxxxxx> wrote:

>I'm wondering if it's possible to set up a decent wi-fi link between
>two points ca. 3 km (2 miles) distant from each other (I have line of
>sight).

You need MORE than line of sight. You need Fresnel zone clearance. At
2 miles, that's about 33ft radius from the center line at midpoint. Do
you have Fresnel zone clearance?
http://www.terabeam.com/support/calculations/fresnel-zone.php

>Most important, I'd like to know what gear I'd need for a
>project like this one.

Well, first you need to disclose what you plan to do with this link.
The details will determine the type of equipment. For example:
1. Do you need to pass only one MAC address or do you need a
transparent bridge (where you glue to networks together via wireless)?
2. How fast do you need to go? Mbits/sec thruput. If you're sharing
an internet connection, there's little need to go faster than the
broadband connection.
3. Need bandwidth management or QoS features? VoIP? Games?
4. How fanatical do you want to be about security? VPN? PPTP?
WPA2-AES?
5. Is there a cost limit?

>Which kind of (directional) antennae should I
>choose?

Never mind the type for now. It's the gain that's important. Once
the gain is established, then you can consider the type. For example,
if you need more than about 15dBi of antenna gain, you won't find a
yagi with that high gain. Similarly, panel antennas become huge above
14dBi. That leaves a dish which goes to 24dBi gain.

>how much should their gain be?

That needs to be calculated. See below. Incidentally, I've only done
this about a dozen times in this newsgroup but I guess it's difficult
to find with Google Groups or something. I guess I should add it to
the FAQ.

>do I need special APs to drive
>them, or external amplifiers, if anything like this is needed at all?

Nothing special. Depending on the answers to the aforementioned 5
questions, the hardware is fairly generic. For simplicity, I'll
assume a transparent bridge, where each end is identical, and passes
more than one MAC address.

>I hope that you can give me some suggestions, because I have no clue.

Hmmm... suggestion: Do the numbers first.

>I tried to search the net, but these questions must be either too stupid
>or too difficult to answer, as I found nothing (that is, nothing that
>I'd be able to understand anyway).

Yeah, that's the problem. If you don't have a vocabulary with the
necessary buzzwords, you can't find anything useful with Google.
Welcome to the wonderful world of wireless calculations.

Let's start with some assumptions.

Linksys WRT54G 802.11g with DD-WRT alternative firmware setup as a
bridge. The transmit output is about +17dBm but can be cranked up
(illegally) higher. I'll assume that you'll mount the WRT54G
somewhere near the antenna with a single 3ft cable (pigtail) using
LMR-240 coax. Loss will be about 4dB including connectors. Longer
cables will require larger LMR-400 coax plus a pigtail.

Now, I gotta do some guesswork. I'll assume that you can live with
about 6Mbits/sec thruput. That means your connection speed must be
about twice that or 12Mbits/sec. The receiver sensitivity varies with
the speed and modulation type.

The following was lifted from the DI-624 data*** but is close enough
for most similar 802.11g radios:
* 54Mbps OFDM, 10% PER, -68dBm
* 48Mbps OFDM, 10% PER, -68dBm
* 36Mbps OFDM, 10% PER, -75dBm
* 24Mbps OFDM, 10% PER, -79dBm
* 18Mbps OFDM, 10% PER, -82dBm
* 12Mbps OFDM, 10% PER, -84dBm
* 11Mbps CCK, 8% PER, -82dBm
* 9Mbps OFDM, 10% PER, -87dBm
* 6Mbps OFDM, 10% PER, -88dBm
* 5.5Mbps CCK, 8% PER, -85dBm
* 2Mbps QPSK, 8% PER, -86dBm
* 1Mbps BPSK, 8% PER, -89dBm
12Mbits/sec sets the receiver sensitivity at -84dBm.

PER means Packet Error Rate which is similar to BER (Bit Error Rate)
but is easier to measure. You would not want to operate at a 10%
error rate. 1 in 10 packets trashed is useable, but no fun. This is
*ROUGHLY* the same as a BER of 1E10^5 which is one of the common
reference levels for measuring receiver sensitivity. Again, these are
measurement reference levels, not operating points.

Next I gotta guess the fade margin. That's how much stronger the
operating level is above the receiver reference level. 20dB is
considered to be a good minimum. In other words, your operating
signal level must be 100 times stronger than the receiver sensitivity
or the link is going to be flaky and unreliable. The relationship
between fade margin and reliability is:
Reliability Fade Margin
90% 8 dB
99% 18 dB
99.9% 28 dB
99.99% 38 dB
99.999% 48 dB
99% reliability might sound great, but that means your link will be
useless for 1% of the year, or dead 3.6 days per year. Don't go below
20dB fade margin.

Plugging the above guesswork into
http://www.terabeam.com/support/calculations/som.php
and playing with the antenna gains until I get 20dB fade margin:

TX power = +15dBm
TX coax loss = 4dB (3ft LMR-240 plus a mess of connectors)
TX ant gain = unknown
Distance = 2 miles
RX ant gain = unknown
RX coax loss = 4dB (same at other end)
RX sens = -84dBm (at 12Mbits/sec)
Fade margin = 20dB
yields a minimum antenna gain of +21dBi. That makes the antenna
required to be a dish. You can get away with a lower gain antenna by
running at a slower speed, shorter coax, or more tx power (at both
ends).

In general the numbers and assumptions above can only get worse, not
better. These calculations are for ideal environmental conditions,
without interference, Fresnel zone obstructions, reflections,
multipath, atmospherics, and terrain problems. If you have another
802.11b/g user along the line of sight, the high gain of the antenna
will create a substantial interference problem. I suggest you do a
site survey to see if there are any other users around.

I have some suggestions as to specific hardware and sources, but I
would like to see the answers to my questions, and corrections to my
assumptions first.

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

Quantcast