San Francisco Wireless Broadband
Or, how I came to have a 3.5 Mbps wireless link to my house
Firstly, appologies for the stark nature of this web site. I'm a network
guy and HTML has never been my strong point. This whole page was written
in vi.
I would consider myself a fairly early adopter of 802.11b. When my roommates
and I moved up to San Francisco in March 2000, we bought an Apple Airport for
$299, which was really the only affordable 802.11b access point at the time.
We also got three 3Com AirConnect PCMCIA cards for our laptops and happily
shared our PacBell DSL line wirelessly.
I was aware that 802.11b was useful for more than just wireless LAN access.
I knew that people were constructing antennas out of
Pringles cans
and setting up point-to-point wireless links. I was interested in the idea
of setting up some kind of distribution network for broadband Internet access,
but I was skeptical that 802.11b would be reliable enough for
semi-long-distance broadband links. After all, my idea of wireless Internet
access was influenced by my first experience with the Internet, which was over
2 meter amateur packet radio at 1200 baud half duplex. I went to a
BAWUG meeting to toss around the idea of
getting together a group of people to start some kind of self-supporting
wireless broadband Internet access in San Francisco. The idea was also
discussed several times on the BAWUG mailing list, but always seemed to peter
out.
There were several problems at the heart of the issue. San Franciso's
topography makes it difficult to reach all of the interested parties from one
location. The notable exception is
Sutro Tower in the center of the city, but the price to locate equipment
on the tower at a height high enough to reach everyone is probably
cost-prohibitive. Additionally, the bandwidth for the project would have to
be somehow transported to the tower, probably via T1 or multiple T1's -- also
cost-prohibitive for a community-based effort. It occurred to me that
transport costs could be greatly reduced or eliminated by placing the wireless
site at a location where good bandwidth was already available, but no one was
forthcoming with information about such a location, especially one with
line-of-sight to much of anything. I gave up on the idea for the time being
and decided to be content with my DSL.
In early 2002, both of my roommates lost their jobs. One decided to embark on
an extended road trip around the country, and the other decided to lay low for
a while and greatly reduce his living expenses. It became clear to me that my
housing situation was going to be changing. I got together with some potential
new roommates and we decided to take advantage of the depressed rental market
and look at some places a little closer to the happening parts of town.
I work for a metropolitan area network provider whose primary business is
building point-to-point Ethernet and SONET networks for enterprise customers.
While I was looking for a new place to live, two things happened. First of
all, my company established a Gigabit Ethernet link with a Tier 1 ISP so we
could offer Internet access as a value-added service. Secondly, I realized
that some of the houses we were looking at renting in Diamond Heights and
Twin Peaks had line-of-sight to one of my company's points of presence (POP)
near the intersection of 2nd St. and Bryant St. in downtown SF. Suddenly,
having line-of-sight to that location became a requirement for any of the
places we were looking at!
I was still skeptical that cheap, off-the-shelf 802.11b gear could provide a
reliable link over the distance in question -- between 3 and 4 miles for most
of the places we were looking at. As a test, I bought two
Linksys
WAP11's and built two antennas out of Safeway brand beef ravioli cans,
which have the same dimensions as the Nalley's Beef Stew cans used in the
tin can
waveguide antenna. I placed one of the WAP11's on the roof of my place
near 26th Ave. and Quintara St. in the Sunset District and aimed the ravioli
can antenna down at Ocean Beach, about 1.2 miles away. I connected the other
antenna to the Lucent PCMCIA card in my laptop and took it down to the beach.
I had little confidence that this setup was going to work, but amazingly, as
soon as I aimed the antenna back up at my house from the beach, I got a signal
almost instantly and was chatting and surfing from the beach! It was
surprisingly easy to keep the link up -- it seemed that the antenna aim didn't
even need to be all that precise.
Encouraged by that success, I re-aimed the antenna on my roof to point at the
top of San Bruno Mountain. For those familiar with the SF bay area, San Bruno
Mountain is just north of SFO. It has a large antenna farm on top that looks
kind of like the spaceship from "Close Encounters of the Third Kind" if you're
headed north on 101 past the airport on a clear night. It's also a state park
and is a great hike if you've never been. Microsoft MapPoint said the top
of the mountain was about 4.7 miles from my house. My girlfriend and I
embarked on a hike up the mountain with my laptop and antenna in hand. When
we reached the top, my girlfriend aimed the antenna back at the house and I
fired up the laptop. Sure enough, the link came right up and I fired off a
few excited e-mails to friends letting them know of my success. It was a
little more difficult to keep the link stable at this distance just by holding
the antenna in your hand, but I was confident that I could get a reliable link
working at 3-4 miles between the POP and one of the houses we were considering
once the antennas were properly mounted and the aiming fine-tuned.
In March 2002, we signed a lease on a house on Twin Peaks above Market St.
with a spectacular view of downtown SF, including line-of-sight to 2nd and
Bryant where the POP is located. I got permission from my employer and the
building owner to mount a small antenna mast on the roof of the POP. I used
a length 1" electrical conduit as the mast and secured it to an existing pipe
on the roof. To make the installation weather-proof, I removed the guts of
the WAP11 from its case and placed them inside a waterproof enclosure from
Home Depot which I mounted to the mast. To keep the ravioli can antennas from
rusting, I spray-painted them with some white Rustoleum. This made them look
not unlike the Winstar microwave antennas already mounted elsewhere on the
roof. I constucted a homebrew power-over-Ethernet cable and ran this from
the antenna mounting location, down the side of the building and into the
window of our POP room. At my new place, I made a similar installation and
mounted the mast to the railing of our deck.
I had some initial difficulties with the power-over-Ethernet setup for the
access point at the POP side. The cable run was rather long, and the voltage
drop over that length of cable was too much to allow the access point to
operate. I ended up purchasing a 2 amp variable-voltage DC power supply at
Fry's with a little dial allowing voltage increments of 1.5V. The power supply
that comes with the WAP11 is rated for 5V, and after turning the new power
supply up to 7.5V the access point powered up. After a bit of configuration
and antenna aiming, I had myself a 3.5 mile wireless link! Some pictures of
the set-up are
here.
Once I'd done some initial usability tests for general Internet access at the
house, I set up a ping monitor to test packet loss across the link. Amazingly,
in the two weeks that had the monitor running, not a single packet was lost.
I also used several of the publicly available "bandwidth meter" sites to test
the link speed. All of the sites reported about 1 Mbps downstream and
300-400 Kbps upstream. The downstream was at least as good as DSL at my old
house, and the upstream was quite a bit better. I was confused about the
assymetrical speed of the link. The only explanation I could come up with was
that there was more interference receiving packets at the POP side due to a
higher density of 2.4 GHz devices.
After a couple of weeks, I decided I would get some commercially made 2.4 GHz
antennas to see if I could improve the speed of the link. I decided on two
24 dBi parabolic grid
antennas made by Pacific Wireless.
These are quite a bit larger than the ravioli can antennas, but surprisingly
lightweight and very reasonably priced at $75 each. Once these were installed
and aimed, the link speed was significantly better -- about 3.5 Mbps each way.
It's been running at this speed without downtime for nearly three months now.
Security is handled via IPSec running between two routers -- one at each end
of the link.
While having a 3.5 Mbps link at your house is pretty cool, it soon became
apparent that we would never come close to utilizing it to its fullest. None
of us are big into downloading music, movies, porn, etc. I began to think
again about the possibility of a community-based group of people getting
together to provide wireless broadband Internet access in San Francisco, and
realized that I was now in a position to offer my resources to get that effort
off the ground. My employer is receptive to the idea, as long as the
additional bandwidth consumed does not expose the company to additional charges
from our ISP. At current levels of utilization, I'm in the clear even if my
wireless link runs at full 3.5 Mbps capacity 24/7. Additionally, even if the
project does eventually consume more bandwidth, I am fully confident that the
participating members will be able to support the cost since there are no
transport costs involved.
At this point, there is no facility for public access to the network at my
house -- only the point-to-point link. It would be wonderful for someone to
donate or loan an additional access point, pigtail and omni or sector patch
antenna so that we could get the first public access point off the ground. If
you are in a position to provide said items, please contact me. I'd also like
to hear from people who have similar good locations for wireless access
sites. If you have line-of-sight to my house, it would be easy to get
bandwidth from my house to your location for distribution to local users.
Finally, I'd like to hear from end users who are interested in participating
who might have line-of-sight to my house. Some pictures of my view are
here so that you might be able to determine if you are
a good candidate for line-of-sight.
Jim Meehan
jmeehan (a t) vpizza (d o t) org