I went to the Project Loon event today (Sunday June 16) at the Air Force Museum in Wigram, Christchurch, New Zealand. Here’s what I learned. (Some of this may be misunderstood or misremembered, so don’t take it as absolute truth, and this is all early prototype stuff anyway.)
It was quite a big museum-style exhibit with lots of people (it had been on the news), and several members of the team were there answering questions.
Currently there are four balloons up. Several have recently been successfully been brought down and recovered from the ocean (land landings are easier to recover but there is no nearby land east of here). The main reasons for testing here are that the stratospheric winds are “boring” (basically all west-east) and this latitude has few countries to coordinate with (they’re working with Chile and Argentina).
The project involves a “couple of dozen” people and has taken two years to go from “ideas on a chalkboard” to this.
One of the balloons was there, inflated to its maximum size (launch size is smaller because of pressure differences, but it’s a superpressure balloon with a maximum size). It was roughly 8m high and 12m in diameter. It has an upper portion with helium and a lower portion with air, with an impeller that can change the pressure in the air part, and thus control the height. The height control is used to navigate, by moving to layers with different wind directions and speeds, and can take it right down to the ground (though that sounds a bit untested). They are aiming for flights of “hundreds of days”.
There are multiple control and networking systems connected together with a CAN bus (for controller area network, used in cars). It sounds like 4-5 ARM cores total: two Linux systems (“unfortunately” two) and one realtime controller in the main payload, and a failsafe controller up on top of the balloon (which controls helium release and parachute deployment). The controllers are designed to recover from being reset very quickly, since resetting is used as a kind of universal problem solver (sensor issues, cosmic rays flipping bits, etc.).
The stratospheric daily temperature swing (30°C down to -80°C) causes difficulties, so the electronics are in styrofoam boxes about 3cm thick, and a heating system keeps the electronics warm and the batteries warmer. The rough altitude is 20km, but it sounds like they cover about a 5km height range, I think connected to the daily cycle somehow too. There is a standard aircraft transponder (yellow and black cables in first picture), and the corners of the solar panel have strobe lights to meet air codes.
The ~1.7m square solar panel provides 100W in full sunlight. It is mounted directly under the transparent balloon (which reduces the amount of energy by 20-25%). The batteries hold about “ten laptop” batteries' worth.
There are three vertically mounted omnidirectional antennas for balloon-to-balloon communication, and one downward facing antenna. The downward one has a 90° cone angle, and is designed so that the signal strength is even across the 40km diameter ground area. 2.4GHz and 5.8GHz, one for balloon-to-balloon and one for ground (I think 2.4GHz is ground).
The communication protocol is custom, to account for the large distances and to coordinate the ground systems so they don’t transmit at the same time (since they can’t see each other). The system basically acts as a VPN between end users and the ground station (upstream ISP), and traffic inside the VPN is encrypted separately as well. Currently upload and download speeds are symmetric (they don’t really know what the speed will end up being, but roughly the same as 3G).
The eventual goal is commercial internet access to parts of the world that can’t get it other ways, but there are no concrete plans for how that will work yet, since they expect to go through many more iterations of prototypes first. Commercial use will necessarily involve large fleets of balloons to provide continuous coverage (even if they end up covering more are each), since they move quite a lot. The balloons measure atmospheric conditions themselves and are coordinated from the ground.