Spectrum is different with satellites as it’s using directional antenna. In theory you can have nearly arbitrary satellite density depending on efficiency. The only effective limit is the user density you’re aiming to support.
Fixed wireless uses directional antennas, too. Towers can have sector antennae, and you can have multiple towers servicing an area, too. (Indeed, small LEO satellites' ability to spot beam is limited, but sectors on fixed wireless can provide a -lot- of densification).
Directional antennas on the ground are a much higher risk from obstructions. Multipath propagation is essential for large wireless networks without building giant towers everywhere.
> Directional antennas on the ground are a much higher risk from obstructions.
Giant towers / hillsides for the central point; small towers or rooftops for the other end.
Between two hills over my suburb, you can see 80%+ of houses. If your point is that this doesn't work well for dense cities: you're right, but it breaks down at a higher population density than satellite does.
Relying upon multipath and diffraction can work to get moderate quality LTE service everywhere, but in general this is not what WISPs are doing, because it doesn't work. (And in any case, MIMO + multipath are friends, too, which provides its own densification, and sector antennae are effective, too).
This is why WISPs are so location specific. With the right geography, population density, etc it can work but we don’t have these companies going nationwide.
We don't have these companies going nationwide because it's one of those things that's inherently a small to medium business. It doesn't work places with high population density. It doesn't work places with very low population density. It requires compliance with local codes and a relatively high amount of real estate work per subscriber. They don't benefit from legal frameworks intended to make things better for cellular carriers.
Satellite is great for very low to low population density, and fill-in coverage in medium population densities.
Fixed wireless is good for low to medium population density in most geographies, and fill-in coverage in high population densities.
Wired internet access is good for places with medium to high population density.
Cellular can provide fill-in coverage for medium to high population densities.
Satellite population density is a little different as individual Starlink satellites cover a ~580 mile radius circle. Illinois is only 232 people per square mile and near the Great Lakes so as far as the satellites are concerned, and excluding the core urban areas with high speed internet, the area has a fairly low population density.
The Northeast megalopolis is a larger issue, but again it’s next to the ocean which makes a huge difference. Further including launch costs it’s ~2 million a satellite. Depending on lifespan break even could be below 1,000 customers per satellite which could support a very dense network. Especially if they charge more for aircraft and boat internet access.
PS: Current Satellites are 20GBPs, so if they average ~1000 people per satellite that’s 20 MBps bandwidth per customer. Traditional home internet can be 10x oversubscribed, but assuming they lose a lot of that to low population areas 1-2x ~= 20-40MBPs in high density areas during peak usage periods.
My town and the town next to it have 25,000 households. If you want to provide 50 mbit/sec to 20% of households, while oversubscribing by 50x, ... you've basically used up 20-40% of a simultaneous-overhead-satellite.
At any time, you can expect there will be dozens of towns of my size in the footprint.
(And, well, most of the time, the entire SF Bay Area, and the Monterey metropolitan area).
Starlink can sell tons of subscriptions, but they can't address suburban connectivity issues.
An alternative analysis, looking at the FCC RDOF areas where funds were awarded to Starlink (completely unserved by broadband-- very low density areas) areas says that Starlink can address around 50% of these households -- not counting any capacity sold to nearby users in cities or towns -- and provide 15mbps peak hour usage-- which is expected to be typical peak hour usage in a few years. https://ecfsapi.fcc.gov/file/10208168836021/FBA_LEO_RDOF_Ass...
P.S. I see your edit and that now you're making an economic argument. I totally think Starlink can be profitable. I don't think Starlink can address anything but the least dense areas and occasional fill-in in other areas, but these are still very large potential markets. As to supporting a very dense network... there are reasonable limits on how many satellites we can expect to have overhead.
P.P.S. gbps not GBps. Currently thought to be 10gbps, but also it's thought they will not have much issue reaching to 20gbps.
Earth only has 196,900,000 square miles. A 580 square mile circle is 580^2 * 3.14 so a little over 1 million square miles. At a minimum that 12,000 satellite network means ~60 are within range at the same time. At 42,000 your at ~210 overhead at the same time. While nobody can see every part of the sky, across tens of thousands of people plenty will get at least a slice near the horizon. So, you can include satellites out over the ocean and exclude people with access to high speed internet already.
Overall a 42,000 satellite network could connect something like 1/2 the “1.3 million people in California without access to a wired connection capable of 25 Mbps download speeds.” As I doubt 1/2 those households are going to want to pay 100$/month for internet the 42,000 satellite network should cover California just fine.
PS: Of note their ground receiver can only point to a slice of the sky, but I suspect plenty of people near the ocean will focus on that slice of the sky for better bandwidth.
> At a minimum that 12,000 satellite network means ~60 are within range at the same time.
It doesn't work like that, unfortunately. It's not uniform and satellites in highly inclined orbits spend more time around the poles.
580 miles radius is also far too much. This may -barely- work for the outer shells, but the link budget and rates get worse there, and of course, these outer shell satellites need to be shared over an area enclosing more population.
Of course, link budget, data rates, and spectral efficiency also get worse as the satellites are lower in the sky, too.
Of course, receiving from an arbitrary subset of 60 transmitters occupying the same frequency band with an electrically steered phase array gets pretty hard when there's a big difference in the received power levels between the transmitters, too.
Sure, it’s also worth nothing that satellites over the ocean may not have a direct connection to a ground station or eventually be relaying long distance connections.
That said, a ground station on the coast connecting to a satellite over the ocean has a huge impact on density calculations as so many people live along coastlines. Especially so when looking at Hawaii and other islands.
Seriously, go read the analysis I linked. I was pleased to see it aligned well with my back of the envelope numbers, and it seems to say that Starlink has a bit of a potential challenge in reaching the underserved populations that it's receiving FCC subsidy to address... They've got a fairly decent model of number of satellites overhead vs. number of subscribers in that area (and assume no one outside of the subsidy customers will buy, which is pessimistic).
I did, but I will more directly address where they are being overly pessimistic. For example, they are likely to prioritizing US coverage which could reach desired goals before completing the global network.
Allocate 12,000 satellites equally spaced across 72 planes to approximate the Starlink fleet That’s a reasonable criticism for 2028, but not their actual goal.
Assuming a 70% broadband uptake rate of assigned locations Clearly not the goal at 12k total satellites otherwise they would not be aiming for a larger network.
500-km coverage radius Doesn’t seem to be accurate based on other sources, but I would accept an actual source such as a current user.
For RDOF locations, we have uplifted these estimates of peak usage to establish a minimum capacity required of 3.6 Mbps per subscriber That’s not how users behave, normally people use zero, ~maximum, or stream a specific amount of bandwidth. Which responds to the maximum available bandwidth.
Anyway, they control the rate and geographic location of new users and can therefore maintain minimum bandwidth standards by slowing adoption based both their current network and user behavior.
> For example, they are likely to prioritizing US coverage
This is difficult, because the earth spins under the orbital planes.
> Assuming a 70% broadband uptake rate of assigned locations Clearly not the goal at 12k total satellites otherwise they would not be aiming for a larger network.
This is the locations that Starlink has bid and received FCC subsidy for to provide connectivity for / receive subsidy for that are completely unserved by broadband.
> 500-km coverage radius Doesn’t seem to be accurate based on other sources, but I would accept an actual source such as a current user.
It's somewhat pessimistic, going all the way down to the elevation limit. tangent(55 degrees) * 550 kilometers = 785.481404 kilometers
At the same time, it's not likely you'll often want to be talking to a satellite at the elevation limit, as it'll be 6-7 dB+ further down even before taking into account the phased array will offer less gain.
> For RDOF locations, we have uplifted these estimates of peak usage to establish a minimum capacity required of 3.6 Mbps per subscriber That’s not how users behave, normally people use zero, ~maximum, or stream a specific amount of bandwidth. Which responds to the maximum available bandwidth.
Forecast average peak hour demand was what was used to size these numbers, which is reasonable. Assuming stochastic demand with the same average makes this worse rather than better.
> Anyway, they control the rate and geographic location of new users and can therefore maintain minimum bandwidth standards by slowing adoption based both their current network and user behavior.
Starlink has committed to provide service to these users as a term for receiving these FCC subsidies.
> Starlink has committed to provide service to these users as a term for receiving these FCC subsidies.
And they can prioritize them over other US customers. But making the service available doesn’t mean 70% adoption in 6 years. Nor does it mean they need the same satellite density globally to cover their US customers. They can very much prioritize deployment of a larger network before finishing their 12,000 customer network just as some North American customers receive service before global coverage was available.
The numbers assume 70% uptake among these customers -- which may or may not be pessimistic.
The numbers also assume 0% uptake among anyone near these people that are not one of these customers. This is probably a little optimistic from a capacity planning perspective.
> just as some North American customers receive service before global coverage was available.
This was a feature of where initial ground stations were located, not of orbital dynamics. A given high inclination short-orbit-period satellite serves all longitudes equally, and the distribution of latitudes it serves is purely a function of its inclination.
Their are several ways to can synchronize the latitudes and longitudes such that you get a repeating pattern. This lets you cover the continental US without having global coverage.
> Their are several ways to can synchronize the latitudes and longitudes such that you get a repeating pattern. This lets you cover the continental US without having global coverage.
This is seriously confused. The Earth turns under that repeating pattern, and the turning of the Earth is asynchronous unless the period of rotation is an integer multiple of the orbital period (which it isn't).
We're in the peak hour right now. This is Starlink. There's no magical favoring of North America (well, except we have a non-operational "train" of new launches overhead). https://imgur.com/a/h94WRsQ
The only thing you can really do is pick your inclination wisely.
Using integer multiples are the obvious, but not only choice. If you look at this visualization you see a higher density coverage over Europe and parts of North America. That’s what let’s them roll out early as the gaps at the equator eventually get filled in. http://www.circleid.com/images/uploads/12233d.jpg
It’s also a significant cost saving measure, though a lot of time is wasted in the band over the Southern Hemisphere it’s still a net win due to global population density from Europe, North America, and Asia. Note, the actual effect is more pronounced as lines of longitude vary their width. But I couldn’t find a better projection.
> If you look at this visualization you see a higher density coverage over Europe and parts of North America. That’s what let’s them roll out early as the gaps at the equator eventually get filled in.
So, you've chosen to just show me an illustration of the sole effect in play, that I've been mentionng to you all day long?
9 hours ago "satellites in highly inclined orbits spend more time around the poles."
5 hours ago "A given high inclination short-orbit-period satellite serves all longitudes equally, and the distribution of latitudes it serves is purely a function of its inclination."
1 hour ago "The only thing you can really do is pick your inclination wisely."
This is not favoring the US. It is favoring things a bit less than 53 degrees North and South (the inclination is 53 degrees). Each satellite spends time up there (down there?) apparently reversing its motion from slightly northwards to slightly southwards, painting a big 'U' in the sky.
And you think this somehow invalidates the data in the study? The study assumed SpaceX's actual constellation dynamics. There's no way to favor North America "more" than the study does.
Edit: Whatever it’s 2:30 am here I am just beating a dead horse.
Not quite I am saying the effect demonstrated is more significant than that was suggesting.
The distance between lines of latitude deceases as you move north. That’s what allows for 24/7/365 coverage in the continental US before the Equator.
Anyway it’s also not just about time spent near though not reaching the poles. The effect is also significant at 30 degrees North vs the Equator and thus the entire continual US benefits.
PS: As to 52 degrees North, much like over the ocean satellites can reach further south than just the area beneath them. Which is why the constitution reaches further north than seems optimal at first glance.
You said "North America" vs. global coverage. I've been mentioning inclination all day long and that yes, inclined orbits spend less time near the equator than at the northern and southern extrema of their orbits. You've insisted that something else in play. Stop trying to save face-- it's OK to not fully understand something.
> PS: As to 52 degrees North, much like over the ocean satellites can reach further south than just the area beneath them. Which is why the constitution reaches further north than seems optimal at first glance.
FFS, broski. Do you read? "It is favoring things a bit less than 53 degrees North and South (the inclination is 53 degrees)"
> The distance between lines of latitude deceases as you move north. That’s what allows for 24/7/365 coverage in the continental US before the Equator.
How does that work? Do you mean longitude? You seemingly use latitude and longitude interchangeably in our conversation. Satellites spend more time at their northern/southern extrema, because ... sine and cosine spend more time near 1 than near 0. And, of course, yes, there's less total land area far north than near the equator.
Your diagram has the actual footprints drawn and they're appropriately elongated when the satellite is north or south. There's no further magical effect based on map projection. The study I cited appropriately adjusts for satellite coverage density but you somehow suggested SpaceX could further concentrate satellites over CONUS beyond the study's analysis, which is how we got onto this entire track. They can't, can they?
