Wifi in the Wilderness: Providing wifi connectivity to amateur astronomers at the Mid-Atlantic Star Party (MASP)

MASP is a week long star party held near the October new moon, at Scout Camp Durant, in central North Carolina. The site is reasonably dark and about 15km from the nearest town. We get a 100m x 100m freshly mowed field for our telescopes, tents and vehicles. Campsites further back in the woods are available for those who want to be away from the early morning sun and the lengthy exchanges of information that start shortly thereafter. Electricity is fed into the field from a panel through cables laid out each year. Home-made portable showers and a hot water heater are brought in for those who enjoy a brisk shower at near freezing air temperatures. A chuck wagon feeds those who don't like their own food and makes some money for the scouts. The admission price is low, US$15, less than a tankful of petrol/gas, to encourage people to come.

There's no reason anymore to be without your e-mail and web surfing ^[1], just because you've left the office or home, so in response to a suggestion at MASP 2001 by the organisers, I planned to provide internet access for MASP 2002.

The nearest connection to the internet was a phone 1km away at the ranger station. The chances of a reliable connection over wires run through the woods back to the ranger station seemed low.

I also needed to provide connectivity between the participants and my home brew Internet Service Provider (ISP) setup. At the time (2002), wifi was not mature with vendors offering low reliability and mutually incompatible gear. In USA new technology is offered on this basis, to allow the marketplace to choose. The rest of the world has compatible gear pushed down their throats, by that bane of informed consumers, "big government". To help consumers adopt the new technology, wifi uses marketing rather than the standard technical terminology: thus "infrastructure mode" is really "client mode", "ad-hoc" really "peer-to-peer", and "access point" is really "server" (there are other modes which can be ignored; they only work with equipment by the same manufacturer). Since wifi wasn't feasible in 2002, I planned for connectivity to users in the field by ethernet cable and hubs run alongside the power cabling. With the dirt and dust I expected the connections to soon fail. Still I was prepared to give it a go.

As it turned out, the organisers weren't interested in internet connectivity in 2002. I tried again for MASP 2003, this time volunteering through a group of people from my club the Raleigh Astronomy Club (RAC)), who, for liability reasons, were helping with MASP outside of the club. (Nowadays, if you're going to volunteer, it's all outsourced to a consortium and you have to join the consortium.) By 2003 wifi technology was more mature and largely mutually compatible (but still using marketing nomenclature). The plan was to tell everyone to bring a wifi card and give them some instructions to try out at home to increase their chances of success when they arrived.

Meanwhile, to my surprise, I found there was a phone pedestal already at the field; it just had to be activated. The phone company should have been able to turn on service at the exchange, but no, they have to come out and turn on the service at the pedestal. You just can't call up the phone company and give them the number on the pedestal and expect them to find it deep in the woods. They want a house address and expect a person to let them in. Presumably when the phone company installs a pedestal, they don't record that they did it or where it is. They must remember to send the bill, but that appears to be it. The phone was turned on in time for the camp, in both 2003 and 2004, but someone had to be there to lead the installer by the hand to the pedestal. This is a nuisance if you live several hours away.

There are only limited choices for internet connectivity in remote areas in USA. It would be nice if you could use a cell phone. In the rest of the world, cell phones all use the same protocol allowing you to make calls in different countries without special arrangements. These cell phones transmit data (they function as modems) allowing you to check your e-mail from your laptop on the train going to work. Of course consumers in USA won't stand for this sort of nonsense - we demand traffic jams and multiple cell phone standards, all different to the rest of the world. At MASP half of the people with cell phones can make calls, while the other satisfied consumers, using a different provider, can't get a dial tone. In USA, consumers demand that no-one send data over cell phones and the intimidated providers levy crippling penalties on users who do. Cell phone companies instead provide an expensive PCMCIA cell phone card (with matching service fees) for use as a modem. These cards have a proprietary interface and will not interoperate with other providers.

At remote sites, an alternative to dial-up is satellite internet connectivity. These devices have a 1m dish antenna, exchange packets with a geostationary satellite and promise higher bandwidth than a phone line. At the last moment for 2003, a MASP organiser brought a satellite internet setup. These are designed only for home use and the company requires a 1 year contract and a US$400 installation fee to screw the dish into the side of the house (since consumers don't want to do this themselves). At their high prices, satellite providers can't compete in the home market with dial-up, DSL or cable internet connection, but you're not allowed to unscrew the dish and take it anywhere else. The connectors for the satellite equipment were flimsy indoor grade designed only to be screwed in once and never unscrewed again, and gave us intermittant connections for the whole star party. Another problem was that you need a Windows computer to authorise the connection - no other type of OS can be used. You have to bring along an internet hardened Windows computer which has no other function than to tell the satellite transceiver once a minute to keep working. This all would have been OK except that the upload direction was slow (about 1200 baud) and suffered from TCPIP resets, making impossible any sort of interactive session (e.g. retrieving your e-mail from a remote site). Non-interactive connections (e.g. websurfing or listening to music) were fine since the download bandwidth is higher. For the bandwidth I'd estimated, the satellite connection would be essential, but for MASP 2003 the satellite connection was a major hassle. For 2004 the bandwidth was handled by phone lines and we didn't use the satellite connection at all.

At home I prepared a fully redundant Linux ISP with spares (e.g. replacement disks for the servers, ready to bootup). The servers provided three networks:

Other than the satellite link (which someone else would bring), I tested it all at home in a clean controlled house sized environment.

The weekend before the star party I disassembled it all, and packed it up with my tents and food. On arrival I set my tent up at the highest point, in a corner of the field, plugged all the wires back in and connected to the internet, at first through the phoneline. With one WAP running on my table, people were on the internet within a couple of hours of my arrival. The next day I ran the ethernet cable out into the field and put the two more WAPs on tables in people's tents, where they'd be out of the sun and out of people's way. Later I connected through the satellite link. I was relieved that my house tested mockup worked fine in the field.

Figure 1. Providing wifi connectivity at MASP 2003

photo by Charles Tilley. (whoever said that wireless was wireless?)

The first problem was that the organisers had put the wifi announcement in with the children's program, where no-one had read it (except maybe the children). The only wifi-ready people were those who carry their laptops and wifi cards wherever they go. There were howls from the rest of the people (and myself) when we realised what had happened. Later in the starparty, one of the helpers went home for a night, and kindly returned with a handfull of wifi cards to sell at cost to people who didn't have them. For 2004 the announcement was moved to the main MASP web page where everyone should have seen it, but no one did. Both years I had 20-30 people ready to go (presumably the same bunch each time) and I spent the the 1st and 2nd day handing out the WEP keys and getting them connected.

I spent some time doing site surveys, walking around, squinting at a laptop screen in full sunlight to determine signal strength and connecting to each of the 3 WAPs. People would call out "can you hear me now?" (apparently a refence to a TV advertisement in USA, which was lost on me, since I don't watch TV). Connection range was about 50m, rather less than spec, but the field is convex, and the WAPs were sitting at table height rather than up in the air (there aren't a lot of places to hang a WAP in an open field). With all the metal vehicles, signals varied by 10db over distances of a few cm, indicating that the field had a lot of standing waves. Even a stationary laptop recorded large variations in signal on a time scale of seconds. This was harder to explain - something metallic was moving - belt buckles? Still people could connect from anywhere in the field, or in the lecture tent, which made me happy.

As with any group, some didn't need any help and connected to the internet without even talking to me. Another group needed hand feeding of instructions, but eventually were connected. I'm not a Microsoft person and for 2003 I had expected the people with Microsoft setups to know their machines. I was mistaken and it took an unreasonable amount of time to get these people setup. For 2004 I learnt about Windows driver installs: I took half a dozen wifi cards (both PCMCIA type I and II), 3 different ethernet cards, half a dozen different laptops, two desktop machines, running W98 or W2K and attempted to set them up for connecting by ethernet and wifi. Although the drivers and cards were all supposed to work on both W2K and W98, I found that some combinations of hardware/software didn't work at all; some required multiple deletions and reinstalls of the drivers; other combinations couldn't see the network neighborhood; some drivers appeared to load correctly but just didn't work; drivers didn't all use the same scheme for installation; not all user level programs appeared in "Programs" and not all could be deinstalled; not all configuration tools were visible in the control panel or the task bar. One set of utilities installed itself with the generic name "Wireless Utilities" so that I wouldn't know which card it worked on. Windows has an artificially imposed limit of only 6 TCPIP devices (rather than some power of 256), so that I couldn't test all combinations of cards without deleting drivers. I realised that you just can't buy a Windows compatible wifi card and expect it to work. In retrospect, I was amazed that we'd got everyone connected the year before. For 2004 I prepared a sheet of instructions for Windows users - with SSID, WEP key, IP of the server and how to configure your wifi card and handed everyone the instructions. The problem of installing drivers for the bulk of the people, just disappeared.

Another group had severely misconfigured machines that in some cases took hours to get going. You wound up putting 99% of your effort into 1% of the people. Other people undid their working settings and manually set their IP to the IP of the server, bringing the network to a halt. One person appeared to be running a WAP on the same channel as one of my WAPs. Two of us tracked him down with a handheld computer. He had no idea that he was running his wifi client card as a WAP, but when he ejected his wifi card the interference stopped and when he put it back in, the interference started again. We asked him to stop, which he did, but he had it up and running again the next morning. On my return, a wifi professional in the club explained that he was running his wifi card in peer-to-peer mode. It had never occured to me that anyone would use any setting except client mode, but in Windows, peer-to-peer is the default setting, to maximise chaos. I'll put this in next year's instructions. I didn't notice the problem in 2003 as that club member had scurried around shutting these people down without bothering to tell me about it.

The WAPs I was using for 2003, LinkSys WAP11's, have a known narcolepsy problem that isn't cured by firmware upgrades. All connections through that WAP would hang and then would start working again when power cycled. This occurred at least twice a day at the starfest. Initially I kept track of each WAP, but this took a lot of effort and I just wound up pre-emptively power cycling a WAP whenever I passed by. In 2004, I used LinkSys wrt54g's, which didn't have the narcolepsy problem, but the range was reduced and people couldn't connect from the presentation tent. I didn't realise the range problem till late in the star party and never got a chance to find out the reason. The LinkSys WAPs are administered via a webform with gratuitous graphics that isn't rendered correctly by netscape or mozilla. I couldn't change settings on some of the WAPs (e.g. the channel) while the others were changeable. I assumed, along with the narcolepsy problem, that the WAPs were junk. On setting up the Windows boxes in 2004, I found that the admin forms were rendered correctly by IE, showing that it was the admin forms that are junk. You can store 4 WEP keys in the WAP but they are cleared when you run without a WEP key. The keys have to be entered 2 digits at a time into separate windows in the form. You can't swipe the key with a mouse and transfer it to the form. I don't recommend either of these devices.

While using a computer indoors is simple, outdoors there are two times of the day when it's hard to use a computer: when the sun's up and when the sun's down. In the morning and afternoon, the sun streams in the sides of the tent and you can't read the screen. Surprisingly a laptop screen is easier to see if you have the sun behind you shining directly onto the screen. In the reverse direction, when the background is illuminated by sun, you can't see anything on the screen. In 2004 I saw people using large plastic laundry tote bins, turned on their side, as laptop light shields for night running. Presumably these would work in the day too. As it turns out, at night you're no better off: now you can't see the keyboard. Even if you're a perfect touch typist, the commonly used keys (control, Esc, backsp) are in different places on each keyboard, while the rarely used keys (the alphabet) are in the same place. The fiducidal marks on the 'f' and 'j' keys are only of limited use for a typist of my abilities. You can't see black keyboards at night - they may look cool, but I'd rather be able to type. At MASP 2004 I saw people walking around with headlamps with a single red LED running on 3 AAA batteries. On my return, I bought one and found that I can type in darkness at home using the red LED, so I'll try it next year. However there isn't much to do after nightfall other than checking occassionally to make sure everything is still up - most people are observing and aren't connecting to the internet anymore.

In the planning stages, to estimate the amount of connectivity needed I'd assumed that I was a typical computer user (I have about 20 computers at home and I spend all day at work and half of my evenings on a computer, which seems pretty normal to me). Multiplying my connectivity requirements by 300 people started to look like a lot of connectivity. As it turns out, I'd overestimated by a factor of 100: only 10-15\% of the people have laptops and they only connect to retrieve their e-mail. It took 3 yrs to get a more accurate estimate, but in the meantime I worried about providing adequate service and arrived with a setup that was 100x overkill. I'm used bidding on contracts where "it" has to work the first (and usually only) time you do it and penalties are large. This wasn't really an appropriate model for a volunteer effort. For 2003, I had asked for volunteers to be available to help, but it turned out that showing them what to do took longer than doing it myself. For 2004 I decided it was a single person job and did it myself. In 2004 I just plugged everything in, turned it on and handed out the info sheets. For 2003 I'd brought a lot more gear than I needed, it was heavy and filled my car. For 2004, except for the redundant pair of servers, I brought laptops instead of desktops and I left the heavy UPS's behind - I figured we could just be down for a while if the power went out. I still brought spare disks and lots of cables but I never used the redundant server. For 2005 I'll bring less gear: I'll still have redundant computers, but I'll inventory my setup before I pack it and will just bring a few extra of everything, but no more.

Some people have unrealistic expectations of a volunteer's role and don't understand the difference between acceptable performance and the performance they deserve. Even though the phone line connecting to the internet was inactive half the time, some people think that we needed a higher bandwidth line, presumably so it would be inactive even more of the time. After 2 years of attempting to rationally discuss these matters, I've found it best to ignore these people.

I provided functionality that no-one used: web pages with the chuck wagon menu, speaker schedule, weather pages and IRC (in case people wanted to send each other messages). I found that people would rather socialise face-to-face than through IRC; if they were hungry or just wanted to hang out they'd go to the chuckwagon and find something they liked; if they wanted the speaker schedule they'd look at the schedule posted all over the campsite. People were using their computers only when they had to and didn't sit on a connection any longer than they needed. You don't need to provide commercial level service - anything that works is good enough. Still you don't want to debug a failure on-site.

Only about 10% of the people used the internet connectivity (more than use the showers from what I saw). Some people like having internet connectivity, mostly for e-mail and the rest of the people couldn't care less. If you have someone who's prepared to setup wifi (like me), then it's an added feature that some people will enjoy. Presumably wifi doesn't make the difference between someone coming or not coming, but for people with things happening back in the real world that they can attend to by e-mail, this allows them to have a better time at the star party.

Announce that you're providing internet connectivity in event flyers, registration package, and webpage. Tell people to bring their computers and a wifi card (and the accompanying CD with the drivers). Don't expect people to read any of this: they only read the announcements for dates, the cost and where to send the registration. They've been coming to star parties for long enough that they've handled everything else themselves. No-one is expecting anything new, like internet connectivity, at a star party. Even though you're sitting prominently in a tent surrounded by lots of computer hardware, it will take several years for people to realise that you're there.

You need a connection to the internet. The more remote you are, the more expensive and difficult the connection setup will be. You don't need to provide commercial level throughput or uptime. People just seem to check their e-mail and a few webpages and then go back to the star party. Dial-up (28-56kbps) is sufficient for a star party of 300 people where only 10\% of people will be bringing wifi cards - you'll find the link inactive half the time.

You'll need a machine(s) to act as the router and firewall and to serve DHCP, DNS, outgoing mail, webpages, NTP, files and to be a webcache. People with digital cameras will upload a total of about 0.5G of images each star party to the fileserver area, so have some disk space available. Although most of the world is Windows based, remember there are other platforms out there, so make your setup as OS agnostic as possible. Have table space so you can help someone sitting beside you. Your setup must survive bad weather (rain, wind, sand) and operate through the night without you to help it. Although almost no-one will use the connectivity at night, there will be people using the internet at sunrise, even if you're still asleep, so know how long before your dial-up connection times out (one of my providers times out in 8hrs and another in 12hrs). If weight or volume of gear is a problem, use laptops wherever you can. Displays are heavy - you can use a laptop as a serial console to debug boot problems.

You have to cover the field with wireless access points (WAPs). Plan on everyone being within 50m of a WAP. One year I got good coverage with 3 WAPs and then the next year got poor coverage with 5 WAPs at table height on a curved hillside of 100m x 100m. Access points are linked by ethernet cable back to your server(s). Remember that the maximum length for cat5 ethernet cable is about 100m.

You want some spare time at the star party for yourself, so you don't want to spend all your time helping users. Bring a hand-out sheet with network parameters (SSID, WEP keys, workgroup), the URL/location of the various web and file servers and and information on loading drivers, since many people don't know how to do this. Test your setup at home - don't expect to debug it in the field. Bring duplicates of everything that can fail.

You're doing this as a volunteer and any functionality you provide will be better than none. Initially plan to provide what you can, rather than what you (or anyone else) thinks you should have (my initial approach).

Astronomy was my first hobby, starting when I was 9yrs old, and was followed shortly by my other lifetime hobby of ham radio (I started off in 1965 as VK2ZJM and am now NA3T). My interest in naked eye astronomy came from doing a boy scout astronomy merit badge. I like watching the ecliptic and the orientation of the horns of the moon's cusps move with the seasons, watching planetary conjunctions, being able to navigate by the stars and I like astrolabes. In college I planned on becoming a radio astronomer, but finding my math skills marginal, I instead chose a career in biochemistry. 35yrs later, after establishing a career, I bought my first telescope prompted by the appearance of the comets Hyakutaki and Hale-Bopp. I'm a member of the Raleigh Astronomy Club (RAC)), My job is a systems engineer at a government supercomputer center.