In one direction (point to point), in a sector, or in all directions (omni)?
That depends where the client is located relative to the antenna. If directly below the antenna, there is enough leakage and sidelobes to make a tolerable connection. However, it's not reliable and subject to extreme sensitivity to the client radios exact location. You would be better off with a 2nd radio acting as an access point to take care of the inside WLAN. Use a different channel (1, 6, or 11) to avoid mutual interference.
How big is your roof?
a pedestal mount ham EME dish might be a bit extreme and far too heavy. My guess is a 2 meter dish is about as large as what could be accomidated on a roof. At 2.4GHz, with an optimized feed, that will yield a gain of about 30dBi gain. gain(dB) = 20 log (7.4 * 2.4 * dia) where dia = diameter in meters 2.4 = frequency in GHz For a point to point link, you'll need to reduce your xmit power to
+21dBm to be FCC legal.
An important point is where are your clients located? Are they spread evenly around your site in all directions, or are they all located in a specific area where a directional antenna would be appropriate?
Aw, I was kinda hopeing for a big 6 ft dia dish on the roof. Oh well.
I'm philosophically opposed to using outdoor omnis (as I stare at the
8dbi omni plugged into my access point). My problems with omnis are:
They spray RF in directions where there are no users. They are a major source of interference when used outdoors.
With sufficiently high gain to be useful, they have a very narrow vertical radiation pattern that easily causes roof or tower top antennas to shoot over everyone's heads.
Ideally, a 120 sector antenna should be used outdoors. If you need
360 degree coverage, you use 3 sector antennas, and 3 access points. If high enough, use some downtilt as the vertical radiation angle is still fairly narrow.
example, the 14dbi gain will give a 7 degree vertical radiation angle.
Of course, omni's are much cheaper than sectors and multiple radios. (Remember, you asked for the best, not the most optimum or cost effective antenna).
Well, some numbers here would be nice.
What kind of range were you expecting? Figure on 6dB of gain equals double the range.
To what type of client radio? Outdoor antennas? EtherAnt? Indoors? Laptops? PDA's?
Any vertical changes in elevation? This is critical with only a 5 to 7 degree vertical beamwidth.
How many clients? How much traffic? If larger, perhaps 3 radios would not be such a bad idea.
Basically, I can't suggest a specific gain and beamwidth configuration without some clue as to the topography, technology, and use. Once the required gain and beamwidth are determined, the rest is just picking a suitable manufacturer and reputable supplier.
One, two, or three story? Hillside (like mine)? Any vertical changes in altitude to the client radios?
$190/ea is not my idea of reasonable. 15dBi gain, 5 degree vertical beamwidth, and no downtilt. Overpriced.
I recognize the antenna but I don't recall the manufacturer. I'll see if I can remember in the morning (after I recover from hauling most of my magazine pile down 50 stairs and overfilling my truck. Recycling hurts).
$120 for 15.4dBi, ?? vertical beamwidth, 3 degree downtilt. Made by Comet (NCG):
apparently discontinued. My guess(tm) is that the shipping and handling costs of these antennas killed them. They're one piece construction and over 6 ft long. Awkward is an understatement.
I've had problems with high gain (over 8dBi) omnis on 2.4Ghz in that the group delay and vswr at the band edges are awful. I can usually make them work on mid-band Channel 6, but 1 and 11 are problematic. I was asked to recommend a fix for a mountain top installation with one of these 15dBi antennas on Channel 11. I just moved it to Channel 10 and all the data errors went away. Eventually, it was replaced with a mess of sector antennas, and most of the close in coverage issues went away.
I would only use a 15dBi omni if I were absolutely stuck with using an omni, were on flat ground, and needed maximum penetration in all directions. Also, remember that a high gain antenna also hears as well as transmits. With very high gain, you're going to pickup much more interference (other users, wireless TIVO, microwave ovens, etc) than with a lower gain antenna. If you're in an area of high concentration of wireless contraptions, a very high gain omni is not a great idea. I would consider 12dBi about right.
One more problem and another reason I detest high gain omnis. With vertical colinear antennas, most of the radiation comes from the BOTTOM of the antenna. The guts are very roughly alternating 1/2 wave sections, with a 1/4 wave top section, and a sleeve matching section at the bottom. Depending upon the design, only every other half wave section radiates. Half of the RF spews from the first half wave section. 1/4th comes from the next radiating section. 1/8th from the next. 1/16th from the next and so on. By the time we get to the top section, there's very little RF left. The antenna may be 6ft tall, but bulk of the signal comes from the bottom.
While I'm bashing omnis, there's also the problem of proximity to big pieces of metal. As the gain goes up, the antenna has to be farther away from nearby tower structures. Such an antenna will work well on top of a pole, but will need to be fairly well isolated if mounted on a tower yardarm. (This is not a problem with sector antennas). Yes, you can mount the omni upside down. I've done that now 4 times and it works nicely but attacts far too much questions.
One more and I'll shut up. High gain colinear antennas have a nasty tendency to create uptilt in the pattern. With only 7 degrees of vertical beamwidth, that's a big problem. Let's pretend you end up with 3.5 degrees of uptilt. That means that *ALL* your signal is going above the horizon. That's great for talking to distant airplanes, but not very useful for talking to ground based stations.
You could also build your own from a kit:
built a similar antenna that worked nicely until I got cheap and used a too small ABS plastic pipe filled with polyurathane fence post compound. The resonant frequency went from about 2430 to to a useless
2380. Oops. It still works on Channel 1 but nothing above that. Forget the fence post compound idea.