A Wireless Long Shotby Rob Flickenger
There has been plenty of talk about the range of wireless LANs, so we decided to find out how far we could stretch it. This series of articles tracks our progress in trying to use the 802.11b protocol to create a link from Sebastopol to a hilltop tower 20.9 miles north, and from there on to some homes 5 miles across a valley. In this first installment, we try out the 5-mile link and create an experiment to test the loss we would encounter on the 20.9-mile link.
High-speed Internet connectivity isn't an issue here at O'Reilly & Associates, since we're in a commercial area and fairly close to Santa Rosa. However, in the hills to the north, where even reliable electricity is a recent development, things are quite different. It's home to eye-popping views of the shoreline and ocean, staggering cliffs and switchbacks, and enough redwood trees to make one think they've stepped onto the forest moon of Endor.
But bandwidth is something of a pipe dream up there. Anything other than dial-up access is a rarity. Sure, you can (and some do) use one-way or two-way satellite connections, but they are a bit pricey and not without their drawbacks (latency and the necessity for an FCC-approved installation geek at $200 per visit, to name two). Is there a better way to get high-speed bandwidth in these hills?
Running physical cable (fiber or copper) into the hills is not only impractical, but would undoubtedly meet stiff resistance from locals who don't want their land torn up by backhoes. In addition to the cost and effort of running the physical cable, recurring monthly telco line charges would be tremendous.
Is there a better way than digging a ditch 21 miles long and dragging in the local authorities and Pac Bell?
The challenge of the long shotA group of local tech heads (myself included) are working on shooting an 802.11b signal into the high hills 21 miles northwest of Sebastopol, using off-the-shelf equipment.
Figure 1. O'Reilly techies want to stretch an 802.11b wireless connection 20.9 miles from company headquarters in Sebastopol to a relay tower in the hills above town, and from there to homes 5 miles across a valley.
This is interesting because, aside from bringing free high-speed Internet access to those parts (including not only some lucky residents, but a school or two as well), we'll have a chance to measure the effects of weather over a long-distance, low-power microwave link -- and figure out how to cope with them.
Also, by using stock equipment, we'll show what can be accomplished by unlicensed operators using off-the-shelf hardware. Yes, Virginia, you can do this yourself!
In order to participate in the Internet, a network (wireless or otherwise) must have at least one computer that can route traffic to the world at large. This machine serves as the gateway for that network to the rest of the Internet, and should therefore provide a "pipe" big enough to accommodate its local users.
We have plenty of bandwidth available at O'Reilly, and adding more is fairly trivial. We are also already using 802.11b to provide access to local employees up to a mile away, if they have line of sight to our building. This makes it a good potential candidate for gateway services.
As it turns out, we have an employee who lives in those northern hills. By asking around in his local neighborhood, he found keen interest in establishing a cooperative network of users. If we could find a point in the hills that can see Sebastopol and a fair number of nearby people, we could establish a repeater site to carry all of the traffic in the area back down to the south, where it can hit the Internet and keep moving.
But how can microwaves reach a site in the middle of the redwoods?
Microwaves stop dead at trees, which look like hanging globs of stationary water to radio signals. So if you can't go through the trees, you'll have to go around them, or maybe even over them.
After asking around, we found that someone in the community has access to a 100-foot high radio tower, on a ridge high above the tree line, at an altitude of about 2000 feet. It has a clear view to Sebastopol, Santa Rosa, and the ridgeline to the southeast. But can it see O'Reilly?
Figure 2. The 100-foot relay tower has a clear shot to Sebastopol and Santa Rosa in one direction, and the coastal ridge in the other. (Click for larger image.)
To find out, we drove to the site and marked the coordinates with a GPS. Using topographical software, we could tell that the lay of the land between the tower and our building in Sebastopol looks clear. We still can't tell if there are trees or buildings in the path, but at least the land is, apparently, going to cooperate. With the latitude and longitude of both sites, we can calculate a bearing and tilt angle, so we know roughly where to point the antennae on each end.
Of course, all of that fancy map work won't help if we can't get a signal to carry for 21 miles. It's akin to trying to have a conversation at the beach. At close range, communication is possible if you're talking loud enough. As you move away, the sounds of human voices quickly fade into the rolling of the waves, until everything sounds like background noise. The only ways that humans have come up with to keep the conversation going are: talk louder, focus the sound (like using an old-time megaphone), talk slower, or some combination of the above.
Due to FCC restrictions on using unlicensed equipment in the 2.4-Ghz band, we can't legally talk any louder (that is, use an amplifier). We can, however, use a megaphone: highly directional dish antennae, one on each end. If we can get two dishes pointed at each other, accounting for altitude, distance, and curvature of the earth, with nothing in the way between them, we might be able to carry on a conversation.
The test shot
A high gain dish is more difficult to aim the further out you go. (See the sidebar for an experiment to learn how tricky this is!) This is where a spectrum analyzer helps tremendously. It's a piece of equipment that gives you a visual representation of what's going on over a range of radio frequencies, shown as a line that sweeps across a screen (much like an oscilloscope).
If you can get the two dishes approximately correct, you can see a spike in the range of the channel your card is tuned to, in relation to the surrounding background noise. By slowly sweeping the antenna on each end, you can lock down the antenna at the point where the spike is highest (and the signal is strongest).
This is precisely what we did, at a distance of 5 miles as a first try. We were very pleased with the performance of the Lucent WaveLAN (now renamed Orinoco) cards we tried; we had no trouble finding the other end of the link. Fine-tuning dish placement was a piece of cake with the spectrum analyzer. By slowly sweeping right and left, and then up and down, we quickly found the point at which our signal was maximized. Of course, we were extremely lucky to have a spectrum analyzer at our disposal, since they run anywhere from $5,000 to $30,000 new.
In a pinch, you can use either the Linux or Windows signal strength meter tool, watching the bars rise and fall as you slowly move the antenna. Having a decent tripod, and slowly moving one end at a time, helps considerably.
Running 5 miles is all well and good, but what about 21? How do we know that we're putting out enough power to cover that distance, without having to go to the effort of climbing towers and hoping for the best?
The formula for calculating the path loss (that is, the amount of signal lost in transit along a clear line of sight) is as follows:
L = 20 log(d) + 20 log(f) + 36.6
(where L=loss in db, d=distance in miles, f=freq in Mhz)
For 5 miles at 2.437 Ghz (channel 6):
L = 20 log(5) + 20 log(2437) + 36.6 L = (20 * 0.69) + (20 * 3.38) + 36.6 L = 13.8 + 67.6 + 36.6 L = 118
For 21 miles at the same frequency:
L = 20 log(21) + 20 log(2437) + 36.6 L = (20 * 1.32) + (20 * 3.38) + 36.6 L = 26.4 + 67.6 + 36.6 L = 130.6
So the difference in path loss between 5 miles and 21 miles at 2.437 Ghz is:
130.6 - 118 = 12.6db
To simulate the signal loss, we added a 12-decibel attenuator (or "pad") between the radio and antenna on our 5-mile test. This let us approximate what the signal would look like if we had a 21-mile line-of-sight shot.
Our radio expert prepared a pad that would give us the required 12 db. He set an adjustable attenuator for 10 db, and measured it to being closer to 12. Then the cables connecting the antenna to the pad and on to the spectrum analyzer further reduced the loss to somewhere between 18 db and 20 db. Never forget to factor in cable length, and keep your antenna runs as short as possible. Patch cables eat up your signal like there's no tomorrow!
As it turns out, the greater loss was in fact a bonus. With about 60 percent more loss than we are anticipating in our real world run, the radios performed flawlessly. We could ping, transfer files, and even flood ping back and forth with minimal packet loss, all synced at 11 Mbps!
To be continued...
Now that we've achieved Ethernet-like speeds over a 5-mile wide valley, and simulated a 20+-mile link over the same, we're preparing for our 21-mile shot. Our experiment proved that it is theoretically possible to drive 802.11b signals well over 20 miles, using stock equipment.
It has certainly given me confidence that, barring intervening ground clutter, we stand a good chance of getting bandwidth to go where no bandwidth has gone before.
With any luck, we'll be able to report our results in the next few weeks. Stay tuned!
Return to the Wireless DevCenter.