Slashdot Mirror
SpaceX Wants Permission To Test Satellite Internet
An anonymous reader writes: SpaceX has filed documents with the FCC asking for permission to begin testing a project to serve internet access from space. "The plan calls for launching a constellation of 4,000 small and cheap satellites that would beam high-speed Internet signals to all parts of the globe, including its most remote regions." This follows news that Facebook and Google had stepped back their efforts in that arena. SpaceX could prove to be a better fit for the project, given that they need only rely on themselves for launching satellites into orbit. "The satellites would be deployed from one of SpaceX's rockets, the Falcon 9. Once in orbit, the satellites would connect to ground stations at three West Coast facilities. The purpose of the tests is to see whether the antenna technology used on the satellites will be able to deliver high-speed Internet to the ground without hiccups."
If at first you don't succeed... anyway no one has come close to "trying". They just played with the idea for a little. SpaceX could actually launch these satellites using their unsold cargo capacity on paid flights. SpaceX paying just the fuel cost to do the project is much different than say Google having to pay for entire launches.
Seven puppies were harmed during the making of this post.
It wouldn't take much to beat these. Both in speed and the bandwidth caps.
I doubt Musk is looking to create a satellite based Internet service. I think it more likely that he's using this network to serve as a testbed for an interplanetary network that covers Earth Mars. The global Internet service just provides him a way to monetize the project to fund its furthered development.
"Lack of speed can be overcome. In the worst case by patience." --Znork
They are planning to test the concept in a few locations in the US. The FCC license they seek covers this test.
If construction was anything like programming, an incorrectly fitted lock would bring down the entire building...
Google and facebook have realized that some problems are not (economically) surmountable.
The problems are the following: The closer you fly your satelite to the earth, the more resistance it has from the atmosphere. The density of the atmosphere reduces by a factor of 100 each 46 km of height. So at "100km", you have about 10000 times less air than at the surface. Some people call that space. At 200km the air pressure is about 100 million times less than what it is over here. That is enough to have a reasonable decay rate of weeks/months/years. "skylab" came down after a few decades, right?
The further away you fly your satellites, the longer the travel times will be for the signals. This equates to ping-times. Hmm. 200km is 0.6 ms, quite acceptable. Both ways. 1.3ms. Still fine. Double the distance to 400km for slower decay times, and you're still about 10 times faster than a normal ADSL line. Acceptable. Not a problem. (the problem here is the same for everybody. The satellites will then play "pass the hot potato" to one that's flying above the ground station and beam your packet down to earth. Assuming your halfway around the globe, that will be about 10000 km. That's with 66ms (round trip) already more than what you get with a residential ADSL line. Still not too shabby.)
The problem with putting satellites high is that the distance to the user becomes large. You want them as close as possible.
The closer you put them, the more you need. -> 4000 of them. This however is not just a one-time investment: because they are low, their orbits decay and they fall back to earth on relatively short notice. If you need 4000 of them, they are not going to be large. So they are small. If you have a cubesate (10cm cubed) weighing 1kg, its orbit will decay just like a 100kg satellite of 10x100x100cm (flying the wrong side towards the front). But a bigger satellite is likely to be 100x100x100cm and weigh not 100, but 1000kg. The extra weight helps keep it in orbit, the extra size in the flying direction does not make a big difference. So the small satellites decay fast as well!
I saw a video of the announcement in Seattle a few weeks ago, and I'm pretty sure he mentioned the number 1,100 km when asked about the altitude. But since then I've heard 6~700 km from another source. Anyway, the idea is to be high enough so that you can join any two points on the globe in only 3~5 hops. He said this would be faster than terrestrial backbone, where you typically have 15 or 20 hops between A and B, each of which adds latency in the form of processing time, not to mention that light travels almost twice as fast in vacuum as it does in fiber.
XML is like violence. If it doesn't solve your problem, you're not using enough of it. --AC
Yes, low orbit is a must.
Over 15 years ago, at the school I worked for, we were offered satellite-based Internet via DirectPC (a DirecTV subsidiary). The speed was amazing by late-1990s standards, 400Kbps sustained! But latency was a killer, at approximately one second minimum. Routing was also nightmarish, as the uplink was phone-based (thus not requiring immense power to transmit, and keeping the lag well, not acceptable, but almost).
I hope "low orbit" is close enough to the Earth to dillute all that latency. Say, if the distance is comparable to what we get in a transatlantic connection, it might just be usable for everything-but-gamers :)
Using your number of 40,000km and the aforementioned 1,100km that SpaceX is considering, would a system with just 2.75% of the latency be enough for you? 3ms for every 100ms that you previously experienced due to the distance alone.
Just re-watching the video: He gets into the specs at around 3:30, citing a goal of 1Gbps @20~30ms latency a couple of minutes later. At around 9:30 he specifically mentions 1100 km for the altitude.
XML is like violence. If it doesn't solve your problem, you're not using enough of it. --AC
They will clean themselves up. Low earth orbit has a very small amount of atmosphere, but enough to slow the satellites over time and cause them to re-enter. Most likely they would burn up pretty much completely, though a few pieces might reach the ground.
You're right and wrong. He wants a system that covers both Earth and Mars, but he also wants to run a satellite internet service. As he said at the Seattle announcement a few weeks ago, we don't know exactly what we'll need to build a city on Mars, "But one thing's for sure, it'll take a whole lot of money." So he intends to use the ISP satellite network to fund the overall Mars mission.
XML is like violence. If it doesn't solve your problem, you're not using enough of it. --AC
The latency would be ridiculous for most use cases.
Are you sure? A round-trip latency of 13ms to the base station(s) seems fairly reasonable to me. These are Low Earth Orbit satellites with an altitude between 99 and 1,200 miles, not geostationary ones at 22,236 miles; that's 1/18th the distance, and thus latency, of existing satellite Internet providers like WildBlue or HughesNet. At the minimum LEO altitude the latency would be another order of magnitude lower still (around 1ms). Even the high-LEO delay is significantly less than the 20-40ms time to the first router reported by traceroute for my Qwest DSL connection.
The trade-offs of LEO include a requirement for many more satellites for the same coverage, the necessity of hand-offs as the satellites pass overhead, and lower orbital lifetimes / higher fuel consumption due to increased atmospheric drag.
"The state is that great fiction by which everyone tries to live at the expense of everyone else." - Bastiat
With spaceX's new proposal you are looking at 2.2 ms as the minimum earth to ground delay + presumably something up to 15-16,000 km (15-16 ms) if your packets had to travel to the exact opposite side of the globe. Add in a 1-2 ms delay for each hop between satellites due to the actual switching and he could be much much much faster for intercontinental packets.
Plus I'm assuming under this scenario that there will be hundreds of terrestrial transmittal points to use versus just a few base stations to make the terrestrial hops even less.
I'd wager that financial market trading traffic alone could pay for a significant portion of this bill at super premium rates, especially overseas traders. Not to mention traffic from ships, planes, rural 1st world locations all paying a premium. They can implement zone pricing pretty easily because they will always be able to able to triangulate a transmission down to the inch. With a network that dense it would greatly surpass the accuracy of the existing GPS constellation.
Many years ago (mid 2Ks or so), I found that a highway rest area had wifi available. Suspecting that it was geosat-based, I pinged my home server. Yep, just over 1000ms. GEO is about 250ms away as the photon flies, times two for the outbound round trip, and times two again for the response round trip.
The phone uplink for DirectPC was basically a cheat to get rid of one of those round trip delays. Not all satellite internet systems did that, and I can recall seeing pairs of junked interface boxes where the uplink and downlink units were stacked together.
#naabhaprzrag, #sverubfr-000, #agi-fcbafberq, negvpyr[pynff*=' negvpyr-ary-'] { qvfcynl: abar !vzcbegnag; }
I'd wager that financial market trading traffic alone could pay for a significant portion of this bill at super premium rates, especially overseas traders. Not to mention traffic from ships, planes, rural 1st world locations all paying a premium. They can implement zone pricing pretty easily because they will always be able to able to triangulate a transmission down to the inch. With a network that dense it would greatly surpass the accuracy of the existing GPS constellation.
I had not thought of that idea before in terms of a potential customer for this set-up. That is an excellent point. Iridium could have been used for something like this (which also has a digital data component), but given the technology capabilities available at the time Iridium was being built, they could only get about 4800 baud for individual customers... something that makes the bandwidth latency sort of irrelevant. High bandwidth and low latency combined with global coverage would indeed be a good customer.
The major competitor to this concept in that regard is an even older technology though, mainly the 19th Century concept (updated to using 21st Century materials) of the cable laying ship. An awful lot of fiber cable has been laid down across all of the oceans of the world between major cities. It is only when you can't access that fixed terrestrial network that something of this nature really becomes useful (as you've mentioned).
As a means to deliver that last mile architecture, it really opens up possibilities.
Did you miss reading the bit about "hit with latency issues?" A 1000ms ping time is no fun.
#naabhaprzrag, #sverubfr-000, #agi-fcbafberq, negvpyr[pynff*=' negvpyr-ary-'] { qvfcynl: abar !vzcbegnag; }
I bet you believe that oil reserves are a hard number, too, and they don't vary with oil price. By reducing the cost of launches, it becomes affordable to launch things that weren't previously affordable, thus increasing the potential business by lowering the price point.
#naabhaprzrag, #sverubfr-000, #agi-fcbafberq, negvpyr[pynff*=' negvpyr-ary-'] { qvfcynl: abar !vzcbegnag; }
Your math is wrong. [satsig.net] It's 240ms round trip straight-on from the equator, directly below the bird, up to 280ms with both ends at extreme angles. (Damn, I thought it was 250-ish each way, not round trip.)
It depends on your definitons of "each way" and "round trip". In particular we don't tend to have servers collocated on the sattelites. so the typical satelite internet scenario is client->sattelite->base station->server->base station->sattelite->client.
So it's a minimum of
120ms client->sat
120ms sat->base
120ms base->sat
120ms sat->client
Assuming delays on the ground are negligable that is a minimum of 480ms round trip time to a server on the internet for a two way geostationary system. Add medium access control protocols that require another round trip to request permission to send a non-trvial ammount of data or significant latency on the ground and that can easilly get much worse. Afaict round trip times of over a second are quite common in practical systems.
In the easly days of sattlite internet it was common to see systems that used sattelite for downstream and dialup for upstream. This significantly reduces the total round trip time (no need for medium access control, only go via the sattelite once but runs into the problem that even assuming asymetric traffic patterns the upstream bandwidth provided by dialup is inadequate by modern standards.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
step 1: build company that launches sats.
step 2: build another company that requires launch capability to be cost effective (give it time.)
step 3: build another company that can generate revenue off of capabilities of 1 and 2 -- repeat
Building your own vertical integration; even if all this basically requires government handouts/subsidies -- i'd argue he's doing far more with them than most companies and contractors do.
Space is really, *really* fucking big. Even low earth orbit at any given altitude is vast; it's literally larger than the surface of the planet. Add altitude shells to that - go up a few KM and you're now a few KM away from anything in the lower shell, even at closest point of approach - and there's an astonishing amount of room in space.
You wouldn't ask if there's room for 4000 more ships on the ocean, despite the fact that there's a lot less ocean and a lot more things crossing it vs. what we have in LEO. You wouldn't even ask if there's room for 4000 more cars on the road in the continental US, despite there being many orders of magnitude less space on US roads than there is in any given LEO altitude. Satellites, functional or not (including debris), move in predictable patterns, and functional satellites have thrusters that allow them to alter or maintain their course.
Agreed, of course, that the satellites should be capable of de-orbiting. But seriously, this "is there enough space in LEO?!?" meme is kind of dumb, at least right now. Let's assume you put each satellite in the middle of 20x20 KM non-overlapping exclusionary zone (omitting the third dimension for now). 400 KM^2 per bird. 1,600,000 KM^2 total. Sounds big, right? You could fit that entire collection, with a hundred thousand square KM left over, into Alaska. Don't get me wrong, Alaska is a big place, but it's not *that* big on the world scale. That's all in one orbital shell.
There's no place I could be, since I've found Serenity...