SpaceX Hits Two Milestones In Plan For Low-Latency Satellite Broadband (arstechnica.com)

SpaceX is about to launch two demonstration satellites, and it is on track to get the Federal Communications Commission's permission to offer satellite internet service in the U.S. "Neither development is surprising, but they're both necessary steps for SpaceX to enter the satellite broadband market," reports Ars Technica. "SpaceX is one of several companies planning low-Earth orbit satellite broadband networks that could offer much higher speeds and much lower latency than existing satellite internet services." From the report: Today, FCC Chairman Ajit Pai proposed approving SpaceX's application "to provide broadband services using satellite technologies in the United States and on a global basis," a commission announcement said. SpaceX would be the fourth company to receive such an approval from the FCC, after OneWeb, Space Norway, and Telesat. "These approvals are the first of their kind for a new generation of large, non-geostationary satellite orbit, fixed-satellite service systems, and the Commission continues to process other, similar requests," the FCC said today. SpaceX's application has undergone "careful review" by the FCC's satellite engineering experts, according to Pai. "If adopted, it would be the first approval given to an American-based company to provide broadband services using a new generation of low-Earth orbit satellite technologies," Pai said.

Separately, CNET reported yesterday that SpaceX's Falcon 9 launch on Saturday will include "[t]he first pair of demonstration satellites for the company's 'Starlink' service." The demonstration launch is confirmed in SpaceX's FCC filings. One SpaceX filing this month mentions that a secondary payload on Saturday's Falcon 9 launch will include "two experimental non-geostationary orbit satellites, Microsat-2a and -2b." Those are the two satellites that SpaceX previously said would be used in its first phase of broadband testing.

Comments

Looks like it's a good time

[bobstreo]

To short Comcast AT&T and Spectrum.

He thought about satellite radio

[jfdavis668]

But, since he doesn't have a car any more, he went with satellite internet instead.

Re:He thought about satellite radio

[93 Escort Wagon]

I dunno... I think the real problem with his satellite radio plans was the flagship “all Space Oddity, all the time” station.

I mean, I like Bowie as much as the next guy - but how about mixing it up a little bit? At least throw a little Modern Love into the mix on occasion.

I just hope that ...

[The Cisco Kid]

they plan to offer this on a competitive basis in all areas of the US (especially rural or suburban areas that currently have none or maybe just one existing broadband option, but even in areas that have both cable and phone options)

And that the pricing is within the reach of the average middle to low income person living in such areas.

Previously I've only seen experiments that focus on providing service to third world countries but ignore the bast under or unserved areas in the US (cough, project loon)

If this ever becomes fully available everywhere in the US, and is priced affordably, it may finally signal the start of the death of the monopolistic stranglehold the current broadband providers have on the market in the US.

That the current FCC seems to be approving of it, suggests to me that it WON'T. It will probably be priced similarly to other Musk offerings, so high as to only be affordable to people with 6 figure or higher salaries.

Because if there's one thing we know Pai protects, its the guaranteed mega profits of his corporate masters.

Re:I just hope that ...

[Gavagai80]

Project Loon was intended as a semi-charitable venture. Any purely commercial project like SpaceX's will be sure to cover the countries where the money is, like the USA. There's zero chance that the USA will not have access to this. And there's zero chance of it being priced like a Tesla, because there's obviously zero market for satellite internet that's more expensive than existing geostationary satellite internet. Also the whole design of the system is meant to make it cheaper than current satellite internet, for the purpose of competing with wired internet providers.

The FCC, despite their bias, cannot and does not simply reject projects with no reason.

Re: I just hope that ...

[Gavagai80]

That's why, according to wikipedia, "it will be linked to flat user terminals the size of a pizza box, which will have phased array antennas and track the satellites." And there are also ground stations involved.

Re:I just hope that ...

[Rei]

The SpaceX constellation is essentially global, and the intent is to undercut most global wired broadband connections on both speed and price; it'll be capable of up to 1Gbps per user, and the costs of the service will be spread around the globe. Previously this would have been unthinkable, but over the past decade there's been both a massive advance in satellite capabilities (per unit mass) and a massive reduction in launch costs (per unit payload mass). And it's all to be in LEO (nearly 12000 identical, mass-produced, mass-launched satellites), not GEO, so latencies are as low as or lower than traditional net service.

They may well pull it off. It's become so clear that such a service is now possible to implement that SpaceX's biggest problem is getting theirs in place before the competition; Samsung proposed such a constellation in 2015, and OneWeb (funded by Virgin Group and Qualcomm) is actively working toward one.

One interesting theory that's been batted around is that Teslas (and presumably other cars) will quickly switch over to it for their connectivity rather than relying on 4/5G service. You can't switch phones to it because the receiver is a phased array antenna about the size of a pizza box - but you can certainly have such a receiver in a car.

Re: I just hope that ...

[Rei]

There are laws of entropy that govern how much actual data can be passed as specific frequencies

Indeed there are... in a given space. Which is why the satellites use narrow spot beams. Each beam from the lower planes targets only 52 square kilometers (a circle with a 4km radius), while the upper planes' beams are 550km^2 (a circle with a 13km radius).

While the satellites do direct satellite-to-satellite communications as well as satellite-to-ground, they're not designed to replace internet backbone services or major service provider connections; e.g. Google isn't going to be hosting its servers directly across the network. Rather, in addition to user terminals, there's also ground terminals ("gateways") that connect directly to internet backbones; the constellation is designed to provide "last leg" services. However, the distances between the user and the gateway that their data gets transmitted to can be very significant.

Re: I just hope that ...

[The Cisco Kid]

Not *everyone* would have to switch to it. And not everyone would (at least among those that had other options to start with).

But merely knowing customers had another option would hopefully put some pressure on the existing services to keep prices down and service levels up.

Re: I just hope that ...

[haruchai]

Even with close to 12,000 satellites, there isn't enough spectrum to get the service to/from the satellites to supply the broadband needs of 1/1000th the current terrestrial networks.

Much as I'm impressed with SpaceX, I don't see this working. And if they do get those thousands of satellites up into LEO, it'll just be a lot more space junk, of which there's already too much.

Re: I just hope that ...

[Rei]

Don't just inform SpaceX - inform OneWeb, Qualcomm, Samsung and Lockheed; I'm sure they'd love to hear your lecture on how you know more than them.

Lastly, junk is, by definition something that is useless. A satellite constellation providing internet services to the entire globe is pretty much the opposite of "junk". Furthermore, unlike "space junk", the constellation's satellites are all designed for deorbit procedures at end-of-life. Lastly, even if they didn't deorbit, they're LEO; "junk" doesn't persist at LEO for protracted periods of time like it does at GEO. ISS loses up to a tenth of a kilometer altitude per day (although it's an exceptional case because of its large cross sectional area)

Re: I just hope that ...

[Rei]

OneWeb, Qualcomm and Samsung have the pricing right (unaffordable for residential broadband)

What on Earth are you talking about? None of them have launched their constellations yet.

Re:I just hope that ...

[nasch]

I'll believe it when I see it, but if I had four providers offering good broadband instead of one that would be awesome.

Re: I just hope that ...

[Rei]

Here's a clue to the clueless: phased arrays are (effectively) instantly steered, steering can be up to 35 degrees, and there's always many satellites in the sky from every location (aka, the reason that there are so many satellites in the constellation)

Re:I just hope that ...

[Tailhook]

Anyone with networking knowledge

I have a working knowledge. Switching elements, particularly at peering points, add latency. Longer paths due to terrestrial geography also add latency. A large source of latency is the refractive index of fiber — about 1.47 — which means light takes 1.47 times longer to travel the same distance through fiber than through vacuum. Coaxial and twisted pair elements have similar propagation delays; anything that isn't vacuum is slower than vacuum.

A satellite network can reduce some part of all of these sources of latency. Satellites operate in near vacuum; propagation between satellites is nearly 100% light speed, and the up/down link through the atmosphere is also very fast as air adds only negligible delay. Satellites suffer no geographic detours. Satellites can reduce the number of switching elements between sites to a small number. I suspect you are way off base about the latency delta between terrestrial and LEO satellite Internet.

Re: I just hope that ...

[Memnos]

I think you might be replying to the wrong comment. You might mean the GP to you.

Re: I just hope that ...

[Memnos]

I tend to agree. It is a high-risk, high-reward venture of Musk's part. If they structure their pricing effectively to capture market share, they can deliver broadband at an affordable cost, with a profit margin.

High-risk, high-reward stuff seems to be Elon's stock-in-trade. The bet is whether he's right often enough to win overall.

Re:I just hope that ...

[Memnos]

Yes, you are correct. I work in high-throughput low-latency software tools where wave propagation times and switching element delays become gating factors. At near-light speed and with fewer switching hops to and from, the potential is for there to be a net gain. If it's well-exploited. If not, it's worth correspondingly less. Hmm... I wonder if they have engineers working on this? Oh, that's right, they do.

Vertical Integration

[mentil]

Seems Boeing is also making a swarm of LEO broadband satellites. Given they also have launch capability, they're likely to be the only company theoretically capable of competing with SpaceX. However, between Boeing and SpaceX, only one of the two companies has 'affordability' in their vocabulary. At best, Boeing will stave off antitrust complaints about SpaceX being able to undercut anyone else. From what I could find, SpaceX's swarm of >4,000 satellites will be far greater than what the competitors are planning, leading to higher max throughput, and ability to serve consumers via economies of scale. That said, SpaceX isn't really a broadband/satellite-making company, so they could screw up somewhere.

Re:Vertical Integration

[Rei]

OneWeb also exists (Virgin Group and Qualcomm). But they've hitched their horse to Blue Origin, so they better hope that Bezos pulls a rabbit out of a hat ;)

Latency

[fred911]

"SpaceX expects its own latencies to be between 25 and 35ms, similar to the latencies measured for wired Internet services. Current satellite ISPs have latencies of 600ms or more, " https://arstechnica.com/inform.... Possibly dated information. But one has to wonder, even if you've fixed a latency issue, how is packet collision handled when ground stations can't hear each other? There's only so much bandwidth allocated. Should be interesting.

Re:Latency

[Henning Rogge]

"SpaceX expects its own latencies to be between 25 and 35ms, similar to the latencies measured for wired Internet services. Current satellite ISPs have latencies of 600ms or more, " https://arstechnica.com/inform.... Possibly dated information. But one has to wonder, even if you've fixed a latency issue, how is packet collision handled when ground stations can't hear each other? There's only so much bandwidth allocated. Should be interesting.

Just the same as satellite phones and other "internet over satellite" (with uplink) providers... Time-division multiple access.

Ground stations have to allocate some time/frequency space over a "management slot" before they are allowed to transmit their normal data.

Re:Latency

[Rei]

"When ground stations can't hear each other" - what do you mean by this? Are you referring to obstructions / interference with a given satellite? The receivers are phased-array (aka virtually instantly steered) antennas and there's always multiple satellites in the sky. The satellites have both satellite-to-satellite and satellite-to-ground communications. So data can be re-routed if there's need. That said, the internet gateways ground stations (unlike typical home receivers) will be positioned and laid out specifically with the intent of minimizing obstructions and interference, for obvious reasons.

Re:Latency

[Rei]

1) Satellite networks already have inter-satellite communications. It's not new.

2) It's much easier than mesh networks on Earth. It's not an improvised network; you know exactly where every craft should be, down to incredibly fine accuracy, and they're all built specifically to operate with each other. And there's no random physical obstructions.

My understanding is that it's a "last mile" solution where the last mile can actually be several hundred. The signal goes up to the satellite and straight back down to the nearest ground station.

The last part is your error. It does not go to the "nearest ground station" to the user. It goes to the latency-weighted nearest ground station to the server which the satellite can reach. Furthermore, it's hopped directly into backbone traffic instead of filtering up through a progressive series of IPs. For example, if I traceroute anywhere out of Iceland, there's six hops within Iceland, then the traffic goes to London, then there's two hops, and then it hops onto a series of backbone routes to wherever it needs to go in the world, whether that's China, the US, elsewhere in Europe, etc. With the SpaceX constellation, the first 8 hops would disappear and be replaced by one hop through the satellite and one from the ground station to the most appropriate backbone; a single satellite could reach to North America, Europe, or North Asia from here.

For all the cases where the signal has to go via more than one satellite, you're fucked

You for some reason are assuming that the satellites have slow packet processing, or abnormal processing delays. Or perhaps you're mistakenly thinking that the physical distance traveled is longer? These are LEO satellites; for most traffic, it's shorter, as it doesn't involve snaking around the world wherever backbone lines happen to be laid.

more so when everyone is trying to do the same thing

Only a very small minority of traffic is routed satellite-to-satellite.

The reason I'm being so definite in my criticism is that we know that mesh networks don't work well (plenty of literature on the topic)

Wow, that totally makes you an expert on satellite communications, and makes you know more than SpaceX, Qualcomm, Samsung and Lockheed.

Re:Latency

[fafalone]

Is that just to the satellite? It's important for a lot of games and a few other things that the total trip is 25ms at most.

Re:Latency

[chemish]

It just can't work.

We can't make mesh networks effective here on the Earth's surface

Ummm, what do you think the internet is?

Re:Latency

[Immerman]

We make mesh networks work all the time on Earth. The Internet itself is a static mesh network. It's *ad hoc* wireless mesh networks that can have issues, where you're figuring everything out on the fly, in a decentralized system without any governing authority. And even those mostly work just fine so long as you have an acceptable node density and you're not trying to interoperate between systems from various different manufacturers. For example, my understanding is that the One Laptop Per Child ad-hoc mesh networks worked quite well in towns where they were mass-deployed.

As for latency issues - firstly, so what? Other than multiplayer arcade-style games and a few other real-time uses, latency is mostly irrelevant unless it gets *really* bad. That said, satellites certainly introduce greater broadcast latency, though at a 500km orbit (1000km ground-to-ground, these won't be geostationary), you're introducing only about 3-1/3ms of latency per "hop". And on the plus side, you've potentially got a lot more satellite nodes servicing an area than a ground based system would have, along with the potential ability to shunt traffic to underutilized ground hubs rather than having to direct all regional traffic through the same terrestrial hub, allowing more optimal use of available terrestrial bandwidth and the ability to route around overloaded nodes that would introduce substantial logistical latency.

Re:Latency

[Immerman]

I would assume so as well - it makes wonderful sense to have at least one major wired hub every, say, thousand miles or so, then all satellites will always have a hub within 500 miles, and can do a single-hop customer-to-hub link to the nearest hub.

Where you get a lot of extra flexibility though, is when you realize that 500 miles or horizontal displacement is barely a stretch for a satellite - power is still 1/2 of what it is in the pure vertical case. And even with that limitation you can create single-hop point-to-point links spanning 1000 miles on the surface. So the reality will generally be that each satellite will be in range of several different different hubs at any given moment, as well as having several different satellites capable of covering any particular customer region.

Apply a little whole-system analysis, and it should be relatively easy to arrange for the "hops" to direct traffic to hubs in close geographic/network proximity to the destination. Especially if you assume each satellite can target several sites simultaneously - 2-3 hub links, plus several customer zones, could potentially be routed extremely efficiently. Maintain one link to a hub next door to a Netflix distribution hub, and you can avoid sending all that video across the wired internet entirely, while still utilizing only a single satellite hop.

Re:Why would they need approval from FCC ?

[FormOfActionBanana]

Fine them. The people of the united states own all of certain resources, such as the communication capability in the country's airspace.
A lot of smaller countries rubber-stamp approval for things approved by the USA. So, for a US based company who wants to do business globally it is absolutely plain that an early prerequisite for them is to obtain approval from the US authority.

How much lower latency? What speed?

[sjbe]

SpaceX is one of several companies planning low-Earth orbit satellite broadband networks that could offer much higher speeds and much lower latency than existing satellite internet services.

How much lower latency? Any satellite service necessarily is going to have significant latency just because of the physics involved. Always nice to have options but what sort of speeds and how much latency are we talking about compared with existing wire line and wireless terrestrial options?

Re:How much lower latency? What speed?

[wbr1]

Did you read? These will be (extremely) LEO satellites as opposed to geo-sync ones. That means not 32000 KM up, but much closer. The biggest contributer to latency is the distance, so instead of 250-300ms up and another 250-300ms back to ground, you get 5-15ms one way. Total bandwidth is of more interest/concern to me.

Re:Give me numbers

[tsqr]

In the coming years, the company hopes to launch 4,425 interlinked broadband-internet satellites into orbit some 700 to 800 miles above Earth, plus another 7,500 spacecraft into lower orbits.

Source

Re:Give me numbers

[wbr1]

TFA, not TFS states 25-30ms. If you have questions, maybe take time to actually read rather than shitpost. But, this is /.

Re:How much lower latency? What speed?

[b0bby]

The most interesting part of the article was towards the bottom:

SpaceX has said it will offer speeds of up to a gigabit per second, with latencies between 25ms and 35ms. Those latencies would make SpaceX's service comparable to cable and fiber. Today's satellite broadband services use satellites in much higher orbits and thus have latencies of 600ms or more, according to FCC measurements.

The demonstration satellites will orbit at 511km, although the operational satellites are planned to orbit at altitudes ranging from 1,110km to 1,325km. By contrast, the existing HughesNet satellite network has an altitude of about 35,400km, making for a much longer round-trip time than ground-based networks.

Re:Give me numbers

[Rei]

Be careful about applying numbers from any one network to any other, particularly older networks in comparison to new ones. Satellite communications technology is anything but static, and specific implementation details matter greatly.

Internet Serivce Anywhere On Earth

[DickBreath]

This could mean good internet service at any point on the earth's surface. From the middle of the ocean to the most rustic remote unabomber cabin.

On the highest mountain. In Antarctica. Even the most inhospitable places like New Jersey.

Re:Internet Serivce Anywhere On Earth

[Rei]

Obviously, the service is only meaningful in places that are remotely livable. But it'll at least be nice to have service on mountains, oceans, cabins and Antarctica.

Re:Internet Serivce Anywhere On Earth

[Fluffymuffin Cocobut]

Check current monthly prices (and available bandwidth) for boat & ship-based internet service. 4-5 figures monthly for paltry Mbps... a pretty standard rate is $1 per MB for 4mb down/1mb up service - plus thousands in hardware and presumably a static monthly account fee...

Re:Internet Serivce Anywhere On Earth

[eth1]

This could mean good internet service at any point on the earth's surface. From the middle of the ocean to the most rustic remote unabomber cabin.

On the highest mountain. In Antarctica. Even the most inhospitable places like New Jersey.

Or, more seriously, unfiltered Internet in North Korea, China, etc. (although it does involve radio transmission, so would be vulnerable to easy detection by authorities)

Re:Why would they need approval from FCC ?

[DickBreath]

Sat phone companies, such as Iridium must do this. So why do you consider that it is an insurmountable problem for SpaceX?

Prices will come down

[cmseagle]

so high as to only be affordable to people with 6 figure or higher salaries

To start, maybe. Musk realized that an electric car wasn't going to be cost competitive right off the bat. He had launch a luxury brand so that consumers would be willing to pay the premium until prices could be brought down. The base price of the Model 3 ($35,000) is 60% lower than the base price of the 2008 Roadster (~$90,000), and you get a much more practical car for your money.

Re:where's the spectrum coming from?

[Immerman]

Perhaps they don't need it? "Satellite-ready" bands are special because they represent offer an extremely "quiet" piece of spectrum suitable for antennas that broadcast and/or receive over very wide areas - potentially the entire cross section of the Earth, at ~13,000 km across.

If it's true, as someone mentioned above, that these would use tightbeam antennas that only cover an area a few km across, then you're talking pretty low broadcast power needed per antenna - your typical cell phone has 10x that range. Shouldn't take many solar panels to power a broadcast station at "terrestrial spectrum" levels over such a small region.

The distance through space is irrelevant (aside from latency and implementation details) - all that matters is the amount of broadcast power, and the size of the "spotlight" it makes on the Earth. Well, and what percentage of the signal is "off target" so that it doesn't hit the "spotlit" region - but modern tightbeam antennas can be very impressively directional.

Well - not quite irrelevant I suppose - you also have the customer antennas sending a signal back. The real limiting factor on acceptable "noisiness" of the spectrum might actually be the directionality and associated power consumption of the customer's phased-array antennas - those have to be mass-produced, and thus incur far larger economic constraints.

Re:Give me numbers

[torkus]

I mean, completely ignoring the article and referring to basic definitions of GEO and LEO

GEO: 36,000km (72,000km round trip minmum)
LEO: 1,000km (2,000km round trip minimum)

Light flitters about at 300,000km/s

Basic math here says GEO requires 240ms just to bounce a signal to GEO and 6ms for LEO.

So THERE. It's two orders of magnitude better and I've fed a troll today to help prevent their extinction.

Re:Give me numbers

[Immerman]

Yes, there's numbers in the article: 500km orbit. Meaning ~1000km ground-to-ground. Meaning roughly 3-1/3 ms of broadcast latency. Up to twice that for a link between points ~1,700km apart.

I'll admit, it would have been nice if the writers had included such numbers themselves.

Re:Why would they need approval from FCC ?

[DickBreath]

The FCC gives a satellite phone operator a certain frequency range to operate in. Their signals should be within this range even considering Doppler shift .

As the satellite is coming towards your phone, the frequency shifts significantly due to satellite speed. The sat phone handset knows this, fully expects it, and is able to tune to the correct frequency for the satellite coming into view. Any single satellite is only in range for a few minutes. So handoff is constant. Doppler shift is part of design.

Re: Give me numbers

[oobayly]

Maybe if you read the article, instead of complaining that there's not enough detail in the summary.

Missed Naming Opportunity

[zenbi]

Starlink?! There will never be a more opportune time to name a service "Skynet"!

Re:Missed Naming Opportunity

[jfdavis668]

How about the Outernet?

Re:Jammable service Anywhere On Earth

[Hal_Porter]

North Korea punishes people to talking to the South

https://www.theguardian.com/wo...

The nature of the revised punishments provides a stark reflection of the regime's anxiety at the nature and scale of cross-border activities, the source explained. A minimum of five years "re-education" or the death penalty can be decreed for those caught communicating with the outside world, a minimum of 10 years re-education is the maximum punishment for simply watching South Korean media or listening to foreign radio, and a minimum of five years reeducation is possible for drug smuggling.

How broad of a band are we talking about?

[Visarga]

What speeds can we expect from this sat network?