Pelethora of Questions

I think you mean ramble, not rable. Newbies don't annoy me. I like to take the easy questions and let everyone else try to answer the tough ones.

Is that 7km?

Ok, you got your feet wet. Next step is to jump into the swamp with both feet. Keep reading.

Basically correct. The largest practical and commonly available dish antenna has 24dBi gain. It's about 1 meter across and looks like a mangled barbeque grill.

Looks at the specification sheet for a 24dBi dish. You'll see something like "3dB beamwidth". For a 24dBi dish, that's usually about 7 degrees. That means that most of your power is within 3.5 degrees of the center line. I theory, you can be +/- 3.5 degrees off center and still have a useable link. Beyond 3.5 degrees, the signal falls off rapidly. As a general rule of thumb, you should be accurate to at least within half of the -3dB beamwidth. That's about the same accuracy as required for the typical 0.5 meter DBS pizza dish.

I would advise against building your own dish, especially if you haven't done anything like this before, and since you don't have any proper RF test equipment. Pringles and coffee can antennas are fairly non-critical. Dishes are tricky. 24dBi dishes are about $52/ea.

Ad-hoc mode does NOT talk or use a wireless router. Your best bet it to buy a pair of transparent bridge radios (i.e. WAP54G, DWL-900AP+, etc) which will allow multiple computers at each end without a router. However, if this is going to be a point to multipoint link, then use an access point (WRT54G) at the end with the broadband, and a client radio (WAP54G) at your end.

Access points do not talk to other access points unless you either use alternative (hacked) firmware, or use WDS (wireless distribution something). You can connect your access point to a switch or hub.

Many access points have external antenna connectors, usually R-SMA or R-TNC. Most client radios have either built in antennas (PCMCIA, USB, etc), that are more difficult to connect to an external antenna. Some have connectors (Proxim/Orinoco).

16ft maximum. However, repeaters are available that will go hundreds of feet. It isn't the data link that craps out. It's the 5VDC regulated power that runs the radio that dies from excessive copper loss.

I don't understand what you're trying to do with the above mess. A wireless transparent bridge is (duh!) transparent. That means it will pass anything that looks like an ethernet packet that's destined to cross the bridge. Bridges don't know anything about IP addresses, so your IP address layout is useless. With a wireless bridge, it's all one big network using the same class C IP block on both sides of the link.

Try reading white papers and articles at:

the product reviews at:

Newbie Warning: This post will rable, so get out now if newbies annoy you.

OK, I live in an area without broadband access, so I decided to look at getting a long-range wi-fi connection to some houses nearby that could get broadband (definately less than 7ks).

I've done a lot of research in the past two days before hitting this group: About.com

;-)

It seems that best performance gains come from highly-directional parabolic dish antenna, is this correct? Also, how "strict" is the directional element? The sites didn't say. At a long range, can I be off by a few degrees? Any tips for getting it right are also useful. I don't need the construct the dish to any particular specifications, do I? Not like having to calucate 3/4 wavelength were I to use a pringles-can design.

My main question, though, is about joining my home network to the network at the remote house.

Can I just have a computer at my house (connecting to my network) in ad-hoc mode with a router at the remote house, which is connected by crossover cable to that person's computer?

Or must I have two APs (or an AP at my house bridging to my network and a router at the remote house also connected to the 'net and thier computer)? If I used an AP, am I correct in thinking that I can connect it directly via ethernet to my switch (and therefore need no wireless cards at all)?

Someone told me that external antenna can only be attached to APs, and also it seems that USB cables have a maximum length of about 15 - which is too short for me (and I'm not buying an extension that converts to ethernet and back)... This would seem to suggest a network structure like:

192.168.x.x

KEY: | -- : Ethernet cable ~ ~ ~ : Wireless

0.2 INTERNET | | | | SWITCH----AP ~ ~ ~ ~ ~ WIRELESS ROUTER | | | | 0.3 | 0.2 0.1 Their Computer

Or the right-side of the diagram being alternately:

INTERNET | ~ ~ ~ AP----PC ROUTER 0.1 | 1.1 Thier Computer

Therefore the wireless stuff is but a bridge, with all the connections to computers themselves handled by ethernet.

Apologies again for the long post, but this stuff wasn't obvious - even from the About.com articles, which were very helpful. Also, it helps to have some people who know what they're doing directly look at the case.

Thanks for any replies :-)

"Learn by Destroying(tm)"

original photos were dishes made from flattened paint and soda cans.

It's not that difficult. The problem is that many highly directional antennas suffer from boresight mis-alignment. The RF peak does not coincide with the mechanical boresight axis. If you build your dish with a center tube support, you'll have a built in boresight for aiming. If not, then clamp a cheap telescopic sight onto a right angle carpenters square, and place the square across the width of the dish. I've also used CCD TV cameras, inclineometers, bubble levels, lasers, and brute force. At 7km, a +/-1 degree error is: tan(1) * 7000m = +/-122 meters so you have some leeway.

Nope. Let's play with the numbers.

You have a 24dBi dish at one end, and a 2dBi piece of junk on the AP at the other. The typical access point belches +15dBm. The range is

7km or 4.35 miles. You'll need an absolute minimum of 10dB fade margin for anything to work. 20dB would be a good target value. I don't have a clue as to how much coax cable you're gonna use so I'll assume that you place the radio on the roof along with the dish at one end, and directly to the access point (in a window) on the other end.

Using receiver sensitivity numbers from a DI-624 (close enough). * 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) I can now arbitrarily decide that you'll want to run 12Mbits/sec which will yield a thruput of about 6Mbits/sec. That gives a receive sensitivity of -84dBm. Distance = 4.35 miles TX power = +15dBm RX sens = -84dBm (at 12Mbits/sec OFDM) TX ant gain = 24dBi RX ant gain = 2dBi TX coax loss = 3dB RX coax loss = 0dB Plugging into:

a 5dB fade margin. You'll barely get a connection, but the speed and reliability will suck. The slightest bit of interference will kill the connection. You'll need about 20dB more antenna gain to make this work. That means a dish antenna at both ends.

Are you planning to do this legally? With too big a dish, you could go over the FCC power limit. It's fairly hard to do that, but it can be done.

Big dishes have even narrower beamwidth than the typical 24dBi dish. however, that's not the major construction problem. It's building a suitable feed that properly illuminates the dish. See:

Optimism, the first symptom of impending disaster. The "depth" of a dish is measured by the f/D (focal length / Diameter) ratio. A deep dish (f/D < 0.35) requires a scalar ring feed for proper illumination. It also doesn't scale well as reflection from the dish near the edge, doesn't contribute as much gain as reflections from near the center feed.

On the other foot, relatively flat shallow dishes are more efficient and require a less complex feed, but have rediculously long center boom. If you don't mind a long ugly boom, that's probably the way to go. Suggestion: Figure out the characteristics of your feed and size the dish to fit, not the other way around.

I can supply URL's for dish construction articles, but dinner comes first.

Nope. Those are the people that live near a free hot spot and want to get free broadband internet from the hot spot. The problem with this arrangement for a point to point link is that you'll pickup far too much junk and reflections at the omni antenna access point. Antenna gain isn't just for getting a good signal. It's also for excluding interference from the sides and back of the antenna.

Incidentally, the gain of the typical coffee can antenna is about 8dBi at best. That's not enough gain for a reliable 7km link.

IPMasq is NAT. Ad-hoc will deliver exactly one MAC address. So, if you have more than one computah on your LAN, you'll need to use a router and NAT at your end to shovel them through the one IP address. If you built a transparent bridge, you could pass multiple MAC addresses and live without the router and NAT.

No. One MAC address passed allow only one computah to go through the link.

A wireless router is a wireless access point (bridge) and a router in one box.

In that case, you'll need two boxes that can play transparent bridge. These are also called workgroup bridges or game adapters. The game adapters are somewhat of a crap shoot as some will pass multiple MAC addresses, while others will pass only one. It's difficult to tell from the data sheets how many MAC addresses will be passed. Good transparent bridges are Linksys WAP54G and DLink DWL-900AP+. I'm not thrilled with the DLink DWL-2100AP. There are others, but make sure you're buying a wireless bridge, not an access point or wireless router. The cheapo wireless bridges will do 32 MAC addresses (30 machines). The newer boxes will do up to 256. The better boxes (Proxim, Alvarion) will do 2048 addresses.

Yep. It's a problem with both high gain dishes and panels. When the beamwidth is I thought I was following along, that the RF isn't squarely off the

Just getting it close. You'll find that it's amazingly difficult to align a dish antenna to within a few degrees without some kind of optical alignment aids. It's really bad on a tower where you can't back off away from the antenna to take an optical sighting. The carpenters square gets you in the ballpark. The rest is tweaking (and guesswork).

Nope. Those are even more difficult to align. A carpenters square will only work for horizontal alignment. One of my dumber ideas as shown in:

into a real challenge to align. So, I cheated. I setup a nearby 2.4Ghz X10 tv link, jiggled the antenna for maximum, attached a rifle scope to the edge, and aligned it with the X10 xmitter. That determined the exact RF peak direction. Then, I pointed it at the distant mountaintop. It worked the first time.

There are other tricks. One of my favorite only works with Pacfic Wireless dishes. They use an extension tube for the 24dBi dish. By removing the locking bolt from the extension tube, it can be moved inward about 2 inches. When moved about an inch, the antenna is defocused into a much wider pattern (and lower gain). The much wider pattern makes it easier to aim.

With higher gain antennas, there are even weirder antenna pointing problems. A center mounted feed presents a "shadow" on the dish, and therefore has reduced gain along the boresight. The result is a dip in gain along the boresight with the peak gain forming a donut around the boresight line. Aiming one these abominations is no fun. The good news is that you probably won't see this with a dipole or yagi fed 2.4Ghz 24dBi dish. You will see it with a much larger, cantenna fed dish.

You've said this before, and I've missed the methodology. Are you saying that the RF pattern isn't aligned where one would think it is? I thought I was following along, that the RF isn't squarely off the visual bore of the dish, until you suggested putting a carpenters square on it. If the RF isn't squarely aligned, why do you want a visual aid that is squarely aligned?

I thought you were referring to something like a DirecTV dish, which has a feed that is off center.

My USB-dongle sees a 3dB received signal increase by placing a flat cookie sheet about an inch behind the dongle. About 9dB with a double cofffee can. I don't think I'll be trying for a 7km link. I only wanted about two city blocks to hit a "UPS Store" hotspot.

Haha, I do that myself on all the Linux groups :-)

Yep.

Heh, I plan to use lots of fine mesh for mine.

Hmmm... methinks this will take a few hours of aiming... Do you think I could have an AP at one end and a big dish at the other... and just aim the dish at the AP... would that cover 7km?

I plan to build quite a massive dish, I've got some sturdy trees to string it up between.

Aahhh, I do plan to build one myself. I'm a good mathematician (but a poor speller). Parabola: Ax^2 + Bx + c I can handle that. Fine mesh wire, deeper parabola for better transmission, larger dish for better reception.

I'll use a relatively deep dish like 0.7*(x^2), and have that equaltion to create 16 "spokes". Then, at regular intervals along the spokes, there will be a "rim", which can have it's cirumference calculated by:

r=x circumference = 2*pi*r

Easy peasy.

OK, I've been reading more, and it seems that most people using LONG links buy the little USB "nub", and place that at the focal point of the dish. This would suggest to me either ad-hoc point-to-point, or an AP at one end, and a truely gigantic dish at the other.

If I had two computers conencted via ad-hoc, am I correct in thinking that the computers connected TO the computers on the wireless link will NOT have to masquerade. This is hard to explain.

Can the computer at each end of the ad-hoc link be connected to other computers via a regular switch, and have them all on the same network; sort of like a daisy-chain? Also, when I said router; I didn't mean a wireless one - just another PC on the network.

Yes... but those USB > Ethernet > USB connectors (150m ethernet max length is good) are very expensive.

Yes, this is what I want! Everyone to be in the same network - but I was unsure of whether network packets would be forewarded through the wireless link or not; and whether chaining the computers together would mess this up.

Cheers for all your help!