Comment by ncallaway | Hacker News Reader

Comment by ncallaway | Hacker News Reader

Parent Full thread

> And if they don't someone else will so I'm not worried... I wouldn't be surprised if something similar happens here.

I'm a little worried about it. Other than Amazon's Project Kuiper, I don't really see who else can compete.

Unlike EVs there's not an existing industry that's doing _mostly_ the same thing, that just needs to start offering a new line. Running a LEO megaconstellation is extremely different than putting up a handful of massive geostationary satellites. I just don't see the existing satellite internet players being able to compete.

Frankly, it's a massive investment of resources to get started (which is true for auto manufacturers as well), but doing it economically requires launch costs that only SpaceX is currently able to provide.

Nobody else wants to build a competitor to Starlink by paying SpaceX. Which is why I think Amazon's Kuiper is the best shot to compete—but it requires Blue Origin's New Glenn to come online to really get the economics working. I know Amazon has tapped ULA as a launch provider in the interim, so they'll be able to start getting satellites up and running soon, but the economics seem...painful if they're going to use expendable rockets for the bulk of the constellation.

It's definitely possible that we see a flourishing of offerings, but I could also see a world where in 5 years time there's really only still Starlink.

>Nobody else wants to build a competitor to Starlink by paying SpaceX

I'm not sure what you mean here. Outside of, say, military usage, every single telecom company is a competitor to Starlink. Internet is internet.

This commercial product is turning out to be what the "naysayers" thought it would be; brilliant if you're in a place with terrible internet, but noncompetitive in any kind of urban environment.

Again, imho, the huge thing here is the ability to get a quick internet set up anywhere in the world that can't easily be taken down (...military).

If you want to build a competitor to Starlink that does the same thing and is ideal for terrible internet areas, you need to put up satellites, and SpaceX is the only company that you can pay to do so economically.

> I'm not sure what you mean here. Outside of, say, military usage, every single telecom company is a competitor to Starlink. Internet is internet.

With regard to satellite Internet mega-constellations, it is true thus far. The two other major constellations (OneWeb and Kuiper) have avoided SpaceX as a launch provider, choosing competitors instead – OneWeb has gone with Arianespace Soyuz and Virgin; Kuiper has chosen ULA (and is assumably going to choose Blue Origin too once they are ready). SpaceX has said they are happy to launch competitor constellations, but it makes sense that its competitors aren't happy to fund their competition. It is very unlikely that Arianespace or ULA can actually beat SpaceX on price, so avoiding funding your competition is the only logical explanation for OneWeb and Kuiper's decisions.

SpaceX is launching other satellite telecommunications systems, that obviously also compete with Starlink to some extent, but less head-on. For example, they are contracted to launch one of the ViaSat-3 satellites. ViaSat-3 is a satellite Internet constellation, so in that sense does compete, but it only has 3 satellites in geostationary orbit, so it isn't really going after the same direct market as Starlink. (ViaSat appears to be getting scared of Starlink, as evidenced by their anti-Starlink applications to FCC and threats of a lawsuit; given that, I wonder how much longer they'll be willing to buy launch services from SpaceX.)

Is OneWeb still in the game? I saw that they had been financial issues and were sold to the UK and I kind of assumed that they had ended their constellation goals.

It’d be great if they were still working on it, though

They are still in the game, but not quite the same game.

They have done 8 launches so far, including 4 this year, with another 5 scheduled this year and then more next year. At the moment they have 248 operational satellites in their constellation. Their goal is 648 satellites, which is gong to take them (approximately) 11 more launches.

But 648, is quite small in scale compared to Starlink's plan for an initial constellation of 12,000, with an additional 30,000 currently going through the approval process. Starlink already has more working satellites in orbit than OneWeb's planned full constellation size.

Given that, it is difficult for OneWeb to compete with Starlink head-on, so they are basically ceding to Starlink the B2C space and going after B2B instead. I think with UK government and UK military they are going to have some success, since the UK will likely prefer OneWeb given that it is a UK-headquartered company with partial UK government ownership.

Well, that’s great! I’m glad they’re still competing in the space and have a good foot-hold.

I think competition in this space is important in the long run.

> I'm not sure what you mean here. Outside of, say, military usage, every single telecom company is a competitor to Starlink. Internet is internet.

That’s certainly true, but there are real physical limitations that make it hard to compete with Starlink without a LEO constellation.

Sure, you could do ground based infrastructure build outs, but the population density means it’s not profitable or you have to charge really high rates to make up the fixed costs. Other satellite providers can compete in the rural areas, but Geo and LEO are have important differences. The first is latency (internet is internet but 60ms internet is not the same as 600ms internet).

When I made the statement I did, I was mostly referring to satellite internet providers, because land based providers have largely abandoned or drastically underserved this market segment.

I don’t see a GEO constellation being able to provide a similar internet offering to Starlink ever, due to light delay. I don’t think ground based solutions will pop up to start competing for these low density market, because they’ve already been written off as unprofitable even before there was serious competition for the market.

> This commercial product is turning out to be what the "naysayers" thought it would be; brilliant if you're in a place with terrible internet, but noncompetitive in any kind of urban environment.

Well that’s probably because the naysayers were predicting the exact same thing as the supporters. Spaced has described Starlink consistently as being designed to service rural and suburban areas that are underserved by existing providers. The goals never were to compete with providers in super high density urban markets.

I’ve been a huge supporter of Starlink since the first announcement, and it’s shaping up pretty much how I imagined which also lines up with your description.

> This commercial product is turning out to be what the "naysayers" thought it would be; brilliant if you're in a place with terrible internet, but noncompetitive in any kind of urban environment.

I mean that's what Elon Musk himself said of the service. He said they don't really have interest in competing with most cable/fiber providers and that their system would actually be complementary and they could provide network backhaul for those companies to say remote cellular antennas.

A bit of a tangent, but is there any information about how many constellations are physically possible? Obviously there's a lot of space in LEO, but it's still limited. Given that each provider would need their own very dense constellation of satellites in different orbits, things would eventually start getting crowded.

I don't think we'll get there any time soon (or ever), but kind of an interesting thought experiment.

You’re underestimating the amount of space in space (sorry, couldn’t resist). Volume increases exponentially as diameter increases.

I don’t think anyone would worry about 1800 or even 180000 things spread over the entire surface of the earth. We have billions of cars, and they mostly avoid running into each other. Even 100 miles up? Far, far more space - and a lot less ‘stuff’ to worry about too.

They could pack a million satellites up there and basic traffic management would be more than enough to avoid collisions.

But it is not the same thing, due to the speed that satelites travel.

Satellites aren't going to swerve, though.

We will see.

But it's definitely don't comparable to cars. Say you want safety margin of mere 1 second, that means reserving 8 kilometers free corridor in front of every satellite. At all times. That isn't a small volume.

Why one second and 8km? Satellites are not manually maneuvered.

Airplanes are only required to keep 1,000 to 2,000 feet vertical separation. They travel at much slower speed than satellites (550 mph vs. 17,200 mph), but they are also much bigger. They tend to travel in narrow corridors, as opposed to being spread out like satellites.

NASA and SpaceX have started working together to address the collision issues with their fleet:

https://www.nasa.gov/press-release/nasa-spacex-sign-joint-sp...

They are not manually maneuvered but there's still unpredictable inteplay of solar wind and tenuous atmosphere to correct. And some sats will fail, the environment is harsh. So I believe some safety separation will be a must, used 1 second as a rule of thumb and the orbital speed is 7.8 km/s at the altitude.

I suspect more safety time will be required. Say a micrometeoroid impact happens causing one satellite to go off track. Now the controller computer needs some time to recompute whole swarm trajectories to minimize debris impacts. Even sats that are currently on the other side of planet might need to start thrusters immediately, as it takes only 10 minutes on average to cover the distance.

Let’s do some math and see. If we assume the earth is a sphere (I know it isn’t, but it’s mostly close), and using 4(pi)r^2, it’s about 511 million km2 in area on the surface (counting oceans, etc). Double checking, it’s actually about 507 million km2, so decently close.

4 * 3.1459 * (6378^2)

Starlink is orbiting at 550km altitude (approx). At that altitude, we’re looking at…

4 * 3.1459 * (6928^2) = 603,977,365 km2

The thing to keep in mind about orbits is the defining factor is your eccentricity and velocity. Faster velocity == higher orbit, assuming it’s a round/circular orbit, that velocity will also be consistent.

As long as everyone is going in the same direction and trying to maintain a circular orbit, you can stack orbits ‘on top’ of each other or put satellites quite close to each other just like planes do with flight levels. So even that number is misleading, as you could 10x that if you had decent traffic control.

Even without that, you’re talking a density of 1 satellite per 600 square kilometers of space with a million satellites for a given ‘flight level’.

The velocity matters a lot less than the relative velocity to other things you might be nearby - think of it like two cars going in lanes next to each other on the same ‘direction’ of a freeway, vs a free for all demolition derby with no traffic control, lanes, etc.

If everyone is co-ordinated you can have problem free high density. If random stuff is happening, including people potentially trying to hit you - it’s going to be a mess. If everyone just does whatever and never co-ordinates, yeah someone is eventually going to run into someone else - but unless there are a LOT of satellites, the odds are low.

With it being LEO, debris and dead satellites are also going to be a aerobraked pretty quickly.

If it's all so easy peasy and every satellite has zillions of square kilometers at disposal, why do we have collisions and required maneuvering already? And "dead satellites are also going to be a aerobraked pretty quickly" - at 550km this means months. That's definitely not quickly enough.

1) That happens INCREDIBLY rarely right now

2) Many of these incidents are from people are also putting things in Lagrange points and Geosynchronous orbits, and similar 'crowded' areas of space - unlike these constellations. Think jostling for space in downtown Manhattan, vs being fine having a place in Kansas somewhere.

3) Right now it's mostly a demolition derby out there, with only minimal traffic handling/control and the occasional 'haha, I got you' intentional satellite explosion/missile test.

As noted, even with all those factors, it is still incredibly rare. Only a tiny handful of problems have been tracked to orbital collisions with anything man made, and that is with us putting all sorts of things (including 'off the books' military top secret satellites, nuclear reactors, random bits of micro satellites, etc.) up there since the 50's.

Months is pretty quick for this kind of thing - and remember, unless it gets energy from some kind of serious-delta-v impact (or other delta-V adding type of event like a high energy explosion), it will lose velocity and DROP in altitude if the orbit remains circular. If the orbit doesn't remain circular, especially at LEO, it will have to dip significantly into the atmosphere to get to a higher altitude at any point in it's orbit, meaning even faster degradation and burn-up.

The reality is that it's huge up there, and if we do even basic thinking and planning on this to not do something obviously dumb (like tangential, perpendicular, or opposite direction orbits all over the place at the same level), it isn't likely to be a significant problem; just like it hasn't been a real (as in significant probability event) problem so far.

Cosmic rays, high potential static build-up, magnetic storms, micrometeorites, etc. are just as big or more of a problem. They rarely cancel a mission or damage anything notable, but it isn't zero risk there either.

> Why one second and 8km? Satellites are not manually maneuvered.

Satellites are actually manually maneuvered generally. Only the large constellations have somewhat automated systems (SpaceX's system is especially automated because there's too many for humans to monitor).

>Volume increases exponentially as diameter increases.

Cubically.

Thank you for the correction! That is indeed correct.

Yeah, that's certainly a good point.

The analogy with cars is a bit tricky though as we have well defined rules that help us pack them very closely together on the surface. For example: roads, lanes, controlled intersections. Those things aren't quite the same when talking about orbits because cars and satellites aren't spread out the same way.

The other commenter made the point about some safety margin, which obviously has to be a part of the equation. The speed and error of orbits would play into whatever that number comes out to be.

Regardless of there being lots of space, there's still some limit, even if it's 1,000 constellations.

Some variables I don't quite know:

1. Acceptable altitude range for constellations

2. Typical positioning error of a satellite

3. Average failure rate and failure mode of a satellite

   - ex. does a dead satellite automatically deorbit itself or do we need to wait for the orbit to decay? I'd imagine decay at that altitude is fairly quick. Would a decaying satellite in a higher orbit need to be accounted for?

4. Required satellite density in orbit and the number of unique orbits, their arrangement, etc.

I'm sure there are other factors, too...

I haven't done the math and I'm certainly not an expert, but my lay understanding is that the number of constellations is much more limited by spectrum frequency availability, rather than the space constraints in LEO.

That is, if we had an very large number of companies wanting to create satellite internet constellations, we would run out of useful EM spectrum before we would run out of LEO orbits.

I'm not positive about that, though!

Yeah, that occurred to me as well, though I don't know much about how the satellites communicate with eachother. I imagine each constellation would need its own band, but not sure how wide it needs to be? Intra-constellation communication could feasibly be done with lasers, in which case, all they may need to worry about is the Space-to-Ground leg. So maybe it's not too bad? I don't really know, but certainly interested in learning more...

The limit is far more how tight you can beamform the signals and how much usable spectrum there is: this is the fundamental thing limiting the density and hence number of users starlink can have.

HughesNet and ViaSat have both been around for a while but their rates and their service are terrible. But unlike Starlink, their service is reliably and predictably crappy so there's that I guess. There's nothing stopping them from improving their own services.

> There's nothing stopping them from improving their own services.

Except the laws of physics I guess. Geosynchronous orbit is about 36,000 km out so best case scenario if you were going straight up to a satellite in such an orbit and back down even at the speed of light in a vacuum that’s a lower bound of a quarter of a second of latency. So unless these legacy satellite internet companies launch LEO constellations of their own, which there’s no indication they’re capable of doing, it seems pretty hopeless for them.

I don't know when the project started, but ViaSat has been working toward constellation service of their own for at least 6 years: https://www.viasat.com/space-innovation/satellite-fleet/vias...

ViaSat-3 is a geostationary constellation of 3 satellites, which I don't really consider to be a competitor to Starlink due to its physical inability to ever reduce the latency lower than 500ms

> There's nothing stopping them from improving their own services.

Well, they can’t improve the latency of their service due to light delay. If they want to compete on that front, they need to move to a LEO constellation.

Neither HughesNet nor ViaSat has started working on LEO constellations (publicly anyway), and these things take years. If they want to get into LEO constellations, they’ve ceded a 5 year head start to SpaceX.

I’m not optimistic that either of them will be able to remain competitive with Starlink over the next decade.

Yeah, it'll be interesting to see, although Viasat-3 will be 1/4th the cost of Starlink (on a $/Mbps/month) basis and Viasat-4 will be about 1/15th. Even launching for "free", it's gonna be pretty tough for SpaceX to meet those numbers. The GEO satellites also require simpler and cheaper ground infrastructure, both on the ISP and on the customer side.

Oh, interesting, I hadn’t seen those numbers. Where did you see the ViaSat 3/4 pricing? All I could find was the current offerings (which are more expensive), but I didn’t look hard.

I’d love to dig into the other offerings that the GEO players plan to provide. My dad was a Day 1 sign up for the Starlink beta, because it’s dramatically better than any offerings in his area (including existing Satellite), but I’d definitely like to share any better upcoming options with him.

I think latency might be a deal breaker for him, because they do a fair amount of video calling and don’t like the audio lag

Ah, those numbers are from the Viasat earnings calls and some rough estimates. Each Viasat-3 satellite will have about 1 Tbps capacity and have a lifetime of 15 years. I don't know about cost, but Viasat-2 was in the neighborhood of $300e6, so, I assumed the same for Viasat-3. Viasat-4 is claimed to have a capacity of 3-5 Tbps and I assume a similar lifetime.

Now, these numbers are all well and good, but they don't include the cost of the ground system and the terminals, and I don't know what those will be. But, I would expect them to be cheaper, most crucially because GEO gives you more flexibility in where you place ground stations (vs LEO) and once you aim the dish, you pretty much never have to move it (LEO dishes spend their whole life slewing to match the satellites; it turns out that requires some parts that aren't so cheap.)

Anyway, I don't know what Starlink or Viasat's pricing will be in a few years, but I think Viasat will have a competitive edge in terms of overall cost.

Latency is definitely an issue for some users and some applications. If you watch what Hughes has been saying for years, it's that the future is hybrid, where a terminal uses both LEO and GEO, for what they're best suited for (i.e., interactive traffic over LEO, big, cheap bulk bits over GEO.) If you Google a bit, you can see the Viasat press releases alluding to this same capability. I don't know if/when that tech will make it to residential service, but I would expect that if there's unmet market demand, they'll get it out there.

Reply on news.ycombinator.com