Comparing Spacebelt and Server Sky

... based on very little public information about Cloud Constellation's "Spacebelt".

On Mon, Sep 05, 2016 at 05:57:56PM -0400, Darel Preble wrote:

Cloud Constellation provides no details of their satellites beyond one picture of boxes whizzing around the earth in two circular equatorial orbits. The inner orbit appears to be about 1200 km altitude, and the outer orbit about 6000 km (near the proposed server sky orbit), but who can tell from a artist depiction?

There is little information about their satellites, except that they hope to use laser links and traditional satellite manufacturing. Their product is data storage.

For comparison, the first server sky products will be van Allen belt characterization and precision debris tracking. The major goal is data services FROM and FOR the developing world.

Server sky thinsat satellites will be foil thin, 5 grams and 250 cm², and deployed in arrays of thousands. Made with photolithography in high speed roll-to-roll batches. CVD InP solar cells on one side, 5-micrometer-thinned millimeter-scale chips on the other, bleeding edge 12 nanometer hafnium gate finfets, both technologies absurdly rad-hard. That (and disappointment in the lack of progress for solar power) started me on server sky.

A 40 kg 8000-thinsat array will collect about 30 kilowatts (on average) and convert it into data-center style computation. The product will not be static data; typical applications will be deconvolving radar returns, or speech and video and image processing.

Newer generation thinsats can be lighter weight, but they cannot be "too good" as light sails or their orbits will not be stable. 2 grams per 250 cm² thinsat is about the limit, though much of that mass can be ballast (space debris, obsolete thinsats, lunar material).

I assume 6400 kilometer orbits, a tradeoff between round trip time (about 60 milliseconds) and visibility. There is room for a trillion thinsats in arrays of arrays.

I've done ultralight products before. I licensed technology that Hitachi used to make "powder RFID", chips with a few thousand transistors that were 20 by 20 by 5 micrometers. With a density of 3000 kg/m³ or 3e-12 gram/(μm)³ (average for silicon, aluminum, and silicon dioxide), those chips weighed 6 nanograms. A chip with a billion transistors in a bleeding edge process might weigh a milligram. Intel's multiCPU chips for portable devices, with graphics processors and multiband consumer radios, are bigger and thicker and vastly more power hungry, but still less than a watt and about 20 milligrams ( 10,000 x 14,000 x 50 x 3e-12, WAG ).

About 3% of US electric production feeds data centers, and that doubles every 5 years. I don't expect to displace US data centers, but to provide better services to 10 times as many people, growing at a faster clip (India cell phone deployments grow at 4% per month, for example). New economies can be built around high information, low physical resource consumption.

Assuming 1 kW of information manufacturing per person for 3 billion people, that is about a trillion thinsats, or about 2 million tonnes in orbit. Growing towards that level of service, at internet growth rates, could fund a lot of launcher development. Assuming aerospace industry learning curve rates (15% cost reduction per market size doubling) I expect that will drive a launch cost reduction of 80%.

However, if the chip industry takes over launch (learning curve rates approaching 40% cost reduction per market size doubling) then the launch cost will drop a LOT faster. See for one speculative approach.

If the information flow is powered by space energy, with the waste heat radiated directly into space, most of the world (and eventually, obsolete backwaters like the US) can reduce terrestrial power consumption, while becoming rich enough to afford expensive alternative energy. Like space solar energy, if it can be delivered in enough small streams to feed a radically decentralized world, unified at the speed of thought.

SpaceBelt (last edited 2016-09-21 05:49:36 by KeithLofstrom)