Traditional satellite internet uses satellites in geostationary orbits, or 35,786km high above the equator. This is great because satellites in that orbit are stationary relative to the ground, so you just need a parabolic antenna pointed at a specific point in the sky to get great reception. But it's also terrible for latency because the signal takes 120ms just to get up, and another 120ms to get down. Assuming you live on the equator, directly underneath the satellite. For someone living in the temperate regions, typical real rtts are in the 700ms area.

Starlink and OneWeb are different in that they intend to use a lot of satellites in low orbits to maintain constant coverage. This is technically much harder, not the least because you need thousands of satellites to get reasonably good coverage, but also because the ground station and the satellite transceiver both need to track each other. This was not technically feasible before, but modern AESA antennas can steer their signal without having to move the antenna, and can both transmit and receive multiple simultaneous beams and very rapidly move the beams around when doing time-sharing.

The minimum round-trip time to something very close to you and also close to it's own ground station will be on the order of 10ms. However, where the system really shines is long distance communication. The satellites will pass the signal between each other using lasers, and will get the signal to the other side of the earth much faster than terrestrial fiber, both because laser in vacuum travels substantially faster than one in fibre, and also because the fibres don't get to follow ideal great circle paths.

Independent researches have evaluated the likely latencies of the system, and the results are frankly shocking. For example, today on the existing fiber network, rtt between London and Singapore is ~160ms. On Starlink, the rtt will be ~90ms.

Not all satellite internet, currently, is dependent on geostationary. Read up on o3b's network and architecture, which has been operational for several years now. The cost, however, is basically "if you have to ask, you can't afford it". O3b is a market option for ISPs and telecoms in very remote or island areas that were previously dependent upon leasing Ku or C band transponder megahertz.

Or cruise ships:

https://www.ses.com/blog/behind-scenes-royal-caribbean-cruis...

> Traditional satellite internet uses satellites in geostationary orbits, or 35,786km high above the equator. This is great because satellites in that orbit are stationary relative to the ground, so you just need a parabolic antenna pointed at a specific point in the sky to get great reception. But it's also terrible for latency because the signal takes 120ms just to get up, and another 120ms to get down. Assuming you live on the equator, directly underneath the satellite. For someone living in the temperate regions, typical real rtts are in the 700ms area.

This doesn't check out.

The radius of Earth is 6,300km. Assuming a receiver on the north pole, the total distance should only be sqrt((35786 + 6300)^2 + 6300^2) = 42555km, which isn't notably larger than the 36000km at the equator.

I'm sure you're still right about the latency, but it can't be just distance doing it.

in practical real world use the absolute minimum ping time you'll see from a 1:1 dedicated capaacity (SCPC) link via geostationary, from end terminal to larger earth station, is 489 to 491 ms. This will vary up to 495ms depending on modem modulation/framing and FEC type, and actual line of sight distance to satellite (low inclination aim angle for the antenna, if you're not directly under the satellite, vs two sites that are directly under it as viewed from geostationary).

Ignoring TDMA oversubscribed VSAT networks for the moment, there are a number of different possible modulation schemes and FEC types, and FEC code rate (payload vs FEC percentages), and things that are unique to different SCPC modems, which will vary the modulation by a few ms beyond pure speed of light.

The latency from the distance light travels does not vary that much. The rest of the ~600 ms latency is caused by the same thing as latency on ground systems (routers, etc.). And the processors on spacecraft are not particularly fast.

Who builds routers that go in space? Cisco? That's got to be some crazy hardware.

Cisco has built routers for space multiple times before. The first time was way back in 2003: https://en.wikipedia.org/wiki/CLEO_(router)

I believe companies like Boeing or Space Systems Lorale do that kind of work in their satellite design groups. Same with Orbital ATK.

Apparently Elon is gonna try use consumer hardware or build their own, thats kinda part of it's cost advantage.

Is lower latency the main benefit? Because there are other ways to combat that, like locating servers closer to where users are. Yes, there are some applications like HFT or gaming or whatever that need lower latency, but I think it would be more useful to the world at large if Starlink provides people all over the world with affordable broadband Internet access with a minimum speed of say 20mbps, say. People far from cities, people in developing countries with bad infrastructure (till a few years back, I had a 4mbps connection), people on islands with bad connectivity, trains, ships, planes, etc. Does it do the latter?

>The satellites will pass the signal between each other using lasers

I thought there were regulatory barriers to doing that

SpaceX released a video about a week ago estimating that SF-London times were be around 60 ms, IIRC.

That sounds like a realistic number, considering the measured latency for the test satellites they launched earlier this year:

https://www.reddit.com/r/spacex/comments/8mccic/musk_tintin_...

"Regular old satellite internet" is very high latency because it's few satellites in high orbit. SpaceX is doing many satellites in low orbit, resulting in much shorter paths and low latency.

Latency for medium to long distances could be better than regular fiber, both because of straigher paths and because speed of light in the relative vacuum of the satellites' orbits is higher than speed of light in fiber.

We don't know yet what kind of equipment the satellites will have, but fundamentally bandwidth is constrained by bandwidth per covered area. I guess they will sell gigabit or better to cargo ships at sea, similarly good speed to rural regions, but nothing interesting to urban areas. Providing fast internet to a city via satellite would be very challenging, providing lighning fast internet to lone people in a desert is much easier and more profitable (no competition).

I'd expect the first and easy goal to be to sell high speed access to airline passengers.

I'd rather expect it to be first sold to high frequency traders. HFT firms have paid astronomical amounts to shave milliseconds off of their connection times.

Which is why I would expect them to already have better-latency options than going to space and back. Does anyone even bother setting up an HFT firm outside fiber range of a major exchange?

Considering how hard it is to get a cell signal inside a plane on the ground, and that these will require special CPE, and the airlines can't change the equipment on a plane without a ruling by the FAA, I wouldn't bet on it.

Onboard internet provided by ground-based cellular or satellite already requires aftermarket hardware -- external antennas, routers inside the plane, and more.

It does require extensive testing and permission by the FAA (and the aircraft manufacturer, and the airline, and others) and extensive support teams in terminals but it's done semi-routinely.

not easy inside a faraday cage. In fact, how does spaceX plan to deliver the internet? Would people need access to a roof antenna? That rules out a lot of use cases

What is the last time you traveled on an airplane? In flight wifi is pretty routine now, aircraft have lots of infrastructure to enable this. It would definitely be possible to adapt that hardware to connect to the satellite system launched by SpaceX instead, all it would take is a different antenna and radio system that the Sat operator would be happy to provide.

I thought you meant some kind of direct passenger connection. I think airplanes already use satellites for the internet connection, no? In any case , i dont see the huge benefit here: they will still all be sharing the same routers and probably have the same connection speed.

The amount of bandwidth available to a single aircraft would be much larger than it is today with this system.

I had Hughes satellite service for a while. Bandwidth was good. Latency was terrible. For surfing the web and streaming content it was fine. For anything twitchy, like driving a computer desktop through terminal services or any kind of gaming, it was nigh unusable. Plus, there were fairly low data caps that limited how much streaming one could do.

The big difference with Hughes and SpaceX is that Hughes satellites are geostationary and SpaceX's will be low earth orbit, so SpaceX's transmission distances and times should be much better.

Hughes provided a decent option for very remote areas. SpaceX appears to be putting together a first class option for everyone.

SpaceX and One Web both have an unfortunately low number of ground stations per square km, so remote areas can have a receiver for everyone, and urban areas will max out the number of receivers pretty quickly.

You'd be surprised, this article describes round-trip times to be 25ms to 35ms and gigabit speeds: https://arstechnica.com/information-technology/2018/02/space...

Who knows if they'll actually reach these speeds/latencies in practice though, that remains to be seen.

Yeah, a significant portion of the latency is just the distance to the existing satellites in geosynchronous orbit. For light to get out and back to the geo belt is around 240 ms, excluding any other factors. Getting up to the promised gigabit speed will be significant challenge. The original demo satellites that SpaceX launched only had data rates of ~15 Mbps.

These satellites are planned below 2000km altitude, where the radio signal delay is more like 3 milliseconds. It should be comparable to accessing AWS us-east datacenters from St. Louis MO

Perhaps their true market will be for industrial IOT devices that need to connect from all over the world.

If I'm selling large HVAC units with cloud monitoring functionality, it'd be attractive to ship units that are pre-connected to SpaceX's satellite network. I wouldn't have to worry about setting up agreements with multiple cellular providers or having to depend on site-provided Wifi/ethernet.

The point is, those devices could very lucrative for SpaceX without requiring as much bandwidth (no need to stream Netflix).

There is definitely a market for this. Orbcomm already does this for a lot of machine to machine communication. However, their data rates are quite slow on the order of kbps.

Same with Iridium Certus/Next - their devices are rumoured to be tiny chips (even with current constellation the hotspot is just slightly larger than modern 4G ones).

You don't really need much bandwidth for most IoT apps.

Presumably those gigabit speeds are shared by the thousands/millions of people on your node. Wireless is a shared medium.

You're conflating bandwidth and latency.

Not much needs both high bandwidth and low latency.

The latency has the potential to be better than the current internet backbone when going intercontinental due to straighter paths, but it'll be higher when going to your local AWS, probably.

Bandwidth is anyone's guess.

The amount of BW is based on how much spectrum you have (see Shannon's Law). I am not sure how much they have. Also, I am not sure they will have enough power to provide in building service.

> Also, I am not sure they will have enough power to provide in building service.

The issue for that is the wavelengths they are using. Their properties basically completely rule out any kind of in-building service, there needs to be a clear uninterrupted path between the receiver and the transmitter. (But they penetrate water well, so should not have a problem with clouds.)

FCC-18-38A1.pdf say 10.7Ghz - 12.7Ghz, that's a lot of spectrum. Depending on the SINR they get, which depends on the power, antenna sizes etc. you could get some pretty good speeds, probably above 100Mbps. But at those frequencies not likely to have any in car or in building. You can penetrate a building with higher frequencies if you have high enough power levels, but not from a satellite. So to use it you will have to install an antenna on the roof of your house.

I was actually just curious about bandwidth, I've only ever used satellite internet through the context of being on an airplane, and was curious if satellite internet was good at supporting things like netflix.

Yes, there is plenty of bandwidth to watch Netflix on satellite.

Over long distances, their latency is going to be significant better, as light speed in fiber is over 30% less than through vacuum.

There were recently two simulations done on possible latencies based on the FCC filings. You can find the video simulations in the links at the bottom of this post.

You can see international links that have latencies lower than the existing internet by a significant margin.

Unfortunately neither of those simulations take into account that the first iteration has been changed to only use radio and no laser links.

[0]: https://www.youtube.com/watch?v=AdKNCBrkZQ4

[1]: https://www.youtube.com/watch?v=QEIUdMiColU

The first video mentions at 6m55s[1] that as part of the third phase they're deploying satellites in high inclination orbits, among other reasons to provide coverage for Alaska, since this is an FCC requirement.

I wonder if it's exclusively because of that reason, and if so that the reason international customers will enjoy coverage at high northern and southern latitudes is because of the accident of history that the US ended up acquiring Alaska, and that there's some FCC regulation that says you need to provide coverage for all 50 states.

1. https://www.youtube.com/watch?v=AdKNCBrkZQ4#t=6m55s

This might be true of geo stationary satellites, but not of LEO satellites, which are close enough to even consider video or voice conferencing, albeit with latency and jitter even worse than bluetooth.