Facebook, trying to be ever more like Google, announced last week that it was thinking of building a global ISP in the sky. Now this is something I’ve written about several times in the past and even predicted to some extent, so I’d like to look at what Facebook has said so far and predict what will and won’t work.

Longtime readers will know I’ve written twice before (here and here) about satellite Internet and twice about aerial Internet, too (here and here), so I’ve been thinking about this for over a decade and even ran some experiments back when I lived in Charleston. Oh, and of course I am building an electric airplane described here.

What Facebook CEO Mark Zuckerberg revealed are plans to work through Internet.org to implement a global network using drones and satellites. In my view drones won’t work as proposed but satellites will. I’ll explain why then offer toward the end of this column what I believe is a more plausible method of building an aerial Internet.

Drones are a bad idea for this purpose if they are expected to be solar powered and run for weeks or months without landing. Think of it this way: the best use case for solar drones is operating at the equator where there’s lots of sun and it shines precisely 12 hours each day and the worst case is operating at the poles where winter operations simply won’t work and in any case the sun (when it shines at all) isn’t as bright. If you want year-round solar-powered operations in the middle latitudes where lots of people live, then, you’ll have to design for no more than eight hours per day of good sunlight which means 16 hours per day of battery-powered flight.

Wow, this is a tough order to fill! From an engineering standpoint the challenges here look to be insurmountable with present technology.

There is very little hard information available about these overnight solar drones but it is interesting to look at Titan Aerospace, led by Microsoft and Symantec veteran Vern Raburn, to see what’s typically proposed. The smaller of the two models described on the Titan web site says the Solara 50 will have seven kilowatts of solar panels on the wings and tail surfaces with batteries in the wings. They never say but let’s guess the 50 refers to a wingspan of 50 feet (the other model is the Solara 60, which is larger but in neither case can I imagine the number refers to meters) which is a pretty big, if skinny, airframe.

If the Solara 50 generates seven kilowatts for eight hours per day with no battery losses at all (impossible) then it should be able to output for 24 hours 2333 watts or about 3.2 horsepower. Admittedly this is just an overgrown radio control glider, but it seems to me that 3.2 horsepower is too little to maintain altitude in the absence of thermal lift which is also dependent on sunshine.

Remember, too, that there’s a payload of Internet electronics that has to be operated 24/7 within that 7 kw power budget. I’m guessing if operation is at 60,000 feet that the biggest power consumer for the electronic payload will be heaters, not transmitters.

There’s no way such a vehicle could make it to 60,000 feet on its own unless they are counting on mountain wave lift, which isn’t everywhere. So I expect it will have to be carried aloft by a mother ship.  Once launched at 60,000 feet with full batteries the glider will have the advantage that parasitic drag is far less at high altitudes, though (lift) induced drag is higher. Speed doesn’t matter because beyond fighting winds to stay in one spot there’s no reason to do much more than circle.

But wait! Circling itself significantly compromises the output of those solar cells since half the time they will be facing away from the sun. So maybe it doesn’t circle at all but just drifts with the jet stream and doesn’t try to maintain a station at all. This detail isn’t covered, by the way, by Zuckerberg’s manifesto. I wonder if he has thought about it?

Our best case, then, is a free drifting glider trying to maintain altitude overnight at 60,000 feet while operating its electronics package. Can it be done with 3.2 horsepower? That depends in large part on weight. To store the net positive power output of those solar cells will require a 111 KwH battery pack weighing with present Li-Ion technology about 144 kg. If the battery comprises half the weight of the drone that gives it a gross weight of 288 kg or 636.6 lbs. Now this just happens to be very near the weight of the electric Quickie I’ve been building and I calculate the minimum power to maintain level flight of that aircraft at around 3kw. Admittedly the Solara 50 flies slower than my Quickie though it is vastly larger, but then again parasitic drag matters very little at high altitudes and low speeds, the question still being is 2333 watts enough to do the job while still powering the electronics?

I say it’s iffy and iffy isn’t what you want to count on for reliable Internet service. So forget the solar-powered drones.

In contrast, a blimp augmented with solar power for station-keeping might actually work, which is probably why Google has settled on balloons for its Internet-in-the-sky. The reason why Google opted for balloons over blimps probably comes down to the power required for station-keeping. If they are just going to let it drift then a balloon is cheaper than a blimp but just as good. Score one for Google.

Satellites I think explain themselves quite well, the only problem is getting enough of them — 1000 or more — to make a reliable network. The more satellites in the constellation the better and with space costs always coming down this is definitely the way to go, though it will take several years and cubic dollars to complete.

So is there some middle ground — some way to make a cheaper more reliable Internet-in-the-Sky that can be up and running in a year or two? I think there is and I described it back in 2004:

Now — strictly because I am twisted this way — let’s take this experiment a step further. Sveasoft supports mesh networking, though with a practical limit of three hops. Aerial WiFi links of 10+ KM ought to be possible and maybe a LOT longer. The hardware cost of a WRT54GS and antenna are on the order of $100. There are, at the moment I am writing this, more than 1,000 small aircraft flying on IFR flight plans in the U.S. So for not very much money you could have a 1,000-node aerial mesh that could serve not only airborne but also terrestrial users. Triple the money, and you could put in each plane a Locustworld mesh with two radios for each node and truly robust mesh networking.

Updating this for 2014 and taking into account the interest of Facebook, I’d advise Mark Zuckerberg to put a mesh-enabled Internet access point on every one of the more than 23,000 active airliners in the world today. Most of those aircraft are in the air at least eight hours per day so at any time there would be about 8000 access points aloft, conveniently going to and from population centers while overflying remote areas. If each access point was at 30,000 feet it could serve about 120 square miles. Figuring a 50 percent signal overlap those 8000 airplanes could offer Internet service, then, to about 480,000 square miles. That’s hardly stellar coverage, I admit, and means a hybrid system with satellites and airplanes makes more sense, but it could come at zero cost (charge passengers for Internet service) and would mean that no airliner would ever again be lost at sea without being noticed or tracked.