I've always found USB WiFi cards more convenient because you don't need any special connectors (or soldering skills!):
EDIT: changed the link to "Poor's man wifi" because Google reports malware.
EDIT2: it's worth mentioning that the standing wave ratio using these home-made antennas can be higher than the ratio expected by the device manufacturer.
These returning waves can damage the power amplifier, but with most wireless cards with transmitter power output under 100 mW (IIRC more than that requires a special license in most EU) it's very unlikely (although possible) and it can only happen after continued usage.
Disclaimer: I'm not an expert, just used to play with wireless about ten years ago.
 Woks with usb wifi dongles are also pretty good.
Could someone here recommend a commercial wifi signal enhancer for use with a laptop that they've had good results with? (It need not have a directional antenna as in the original article.)
Whenever I've googled the commercial products, I feel like the market is rife with snake oil. I'd be much happier with a recommendation from someone on HN.
The closest thing (though it's completely different) to what is covered in the article is here: http://store.netgate.com/24-GHz-9-dBi-Linear-Feed-Antenna-N-...
The cantennas is really a waveguide. As such, it has no gain. What is being measured is the increase in SNR, due to a lowered noise floor. (Edit: to be completely fair, the antenna raises the signal and noise level.)
This one has increased gain:
I assume that the interest in this is back because of Marius Milner the Google employee who built the Netstumbler software http://www.netstumbler.com / http://stumbler.net is the suspected engineer who wrote the Google software that has been in the news about the mapping of access points by Google streets cars.
Free space loss = 32.4 + 20xLog F(MHz) + 20xLog R(Km)
At 2.4 Ghz, this formula is: 100+20xLog R(Km)
one mile = 1.61km, so your free space path loss was around 100dB.
Let's say the truck stop had about the same radio as you, maybe 17-18dBm transmit power (50-62mW) into a 2.2dBi antenna.
Signal leaves the radio at the truck stop at 17dBm, encounters about .8dB of loss in the coax and connectors, and then is raised 2.2dBi by the very short antenna, for 18.4dBm EIRP.
It then encounters 100dB of path loss, arriving at the antenna on your laptop at -81.6dBm, where it is raised 2.2dBi by the dipole in your laptop, and then encounters perhaps another 0.8dBm of losses in the coax and connectors.
The net result: the radio sees -80.2dBm, which is more than enough signal to decode 11Mbps CCK (802.11b @ 11Mbps) or 12Mbps OFDM (802.11g).
So, what you did was unremarkable, though probably fun.
In the case of a WiFi mesh network, you'd really want a set of omnidirectional antennae. An omnidirectional antenna with a 1/4 mile radius would need a lot more power to operate. Further, in a place like new york, you have to start worrying about 3 dimensions (skyscrapers are tall!) and obstructions (buildings can be difficult for radio signals to pass through).
Edit: removed redundant phrasing.
The last part (buildings) is incredibly important to remember. WiFi signals—which are typically in the 2.4GHz or 5GHz bands—are extremely susceptible to attenuation in this manner, due to the higher frequency.
Note that, in part, this is why carriers are so hungry for the sub-GHz bands (i.e. 700MHz in the US, 850-900MHz in Australia, etc) for their 3G/LTE+ deployments. The lower frequencies exhibit significantly better building penetration, offset by the need to deploy more cells in order to get similar bandwidth capacity (considered a worthwhile trade-off).
An omni that could broadcast 1/4 mile (400m) in Manhattan at 2.4GHz would need to have a VERY high transmit power. It would likely be illegal, and it would also be dangerous.
You're not going to build a mesh network in 2.4GHz in NYC.
You would use a regular omni-antennae for local client connectivity in each zone then a bunch of these to provide the backbone link to connect those zones to some centrally better connected point. Rather than trying to do a pure peer-per mesh
Anyway, Nescafe cans work very well for access points up to 600 feet away, which covers most of the city. In case Illy is not popular in your area.
But yes, the opening should be at least 1/2 the wavelength of the signal you seek. The lambda of 2.4GHz is around 0.125m, or 4.9 inches. Half this is just under 2.5 inches.
I made a double-long cantenna from some large spaghetti sauce cans years ago and it worked wonders. Then 802.11N became widespread and suddenly all the cards had multiple RPSMA connectors and antennae. Can anyone confirm how 802.11N has changed the recommended use patterns of can antennae? Should I have the same cantenna on each connector, or should they be different to provide differences in signal path for 802.11N magic to work? Could use of (well positioned, well made) can antennas actually make the performance worse than the stock 802.11N setup?
- the cable for feeding your antenna is quite long. A shorter one would have much less loss
- transmission with a directional ant can easily exceed the allowed, effective tx power (i own an Alfa 500mW hooked up to a Biquad with something like 8 to 10db gain and 20cm feed... certainly not allowed in Europe).
- a circular polarized ant is even better for WiFi (multipath interference) but more difficult to build
(oldest beer can reflector shown is from popular mechanics 1955)
there are also PDFs on the Internet that help you to create your own reflector.
always good fun to make something yourself.
You can even make a bigger hole and put an entire USB wifi dingle in the can at the correct position but it takes a bit of trial and error because you don't know where the antennea is inside the plastic
Presumably because the can is approx a wavelength you are in all sorts of complex near field regime, unlike say a parabolic sat dish?