Hacker News new | ask | show | jobs
by new_realist 2389 days ago
I think they lowered the orbit because Starlink latency was/is not competitive.
3 comments

Lowering orbit doesn't substantially reduce latency, building more satellites does.
Building more satellites doesn't reduce latency. Are you referring to the ISL?
Adding more ground stations and building more satellites does reduce latency, because there will be a satellite closer to "halfway between" you and a ground station, which defines the shortest possible path.

If there is just one (non-stationary) satellite in your sky, the geometry is probably pretty crummy, so light-speed delay is greater. Also, link budget is worse, so probably data rate is lower too.

Yes, once you have a ton of satellites, adding more doesn't improve latency much.

Having more satellites lets you communicate in a straighter line. Lower altitude lets you hug the Earth more closely. Higher altitude lets you see more neighbors. I'd wager the low altitude satellites are for last mile communication moreso than global linkage. I mean, otherwise they'd have no use for the higher altitude ones? An optimal low latency link around the planet would mostly alternate low-high-low-high-low in sequence, if there aren't enough low altitude nodes to use them exclusively. But maybe you could crush this argument by doing the actual math in this case.
You are referring to the ISL in these satellites, that are not in the first batch, nor are there any plans to change that soon from what I've seen. So without the ISL, their latency is certainly not lower by adding more satellites. The difference in latency from different altitudes LEO is negligible.
compared to geostationary sats, starlinks are 1/30 of the distance, and thus 1/30 of the latency
They didn't lower it from geostationary, so this is--while correct--irrelevant.
It makes a difference there, sure, but SpaceX lowered orbits from 1,150 to 550km. Half the latency, but only by about as much as traffic between LA to SF incurs.
I was a bit wrong in my initial estimate. More satellites can't improve latency by much once they get dense enough, and it looks like Starlink has enough such that the altitude would be on the same scale as distance. Still doesn't matter for the typical home user. Like, I get 21 ms ping to servers in my local city.
> Still doesn't matter for the typical home user.

One very important market they're going after is global arbitrageurs.

The funny part is that by providing low latency service to them, they are (hopefully) going to be able to fund SpaceX longer to build rockets to get human colonies off-planet.

It’s almost like a Gibson or Stephenson novel.

Do you have any evidence of this? I work in algorithmic trading/finance and I'm relatively certain that this isn't true.
> Still doesn't matter for the typical home user.

550km vs 1,150km might not matter, but it matters a ton vs 40,000km. Even without the bandwidth limits, I wouldn't be able to use geostationary internet for my normal usage.

Starlink has been LEO since it was first announced, so how bad geostationary internet may be is not relevant to the discussion about why SpaceX lowered the altitude.
right i didn’t know that
I was under the impression that Starlink latency is actually very competitive with terrestrial fiber. Why do you say it wasn't or isn't? Everything I find suggest Starlink will have low latency.
For most usages yes. For gaming though it's right on the edge: theoretical minimum at 350km orbit is 40ms roundtrip, which would work for most games. But many complications, including position of the satellite, latencies within the networking components, packet loss etc, can easily slow this to 100ms+, which is unsuitable for many online games.
No, the round trip for electromagnetic waves at 350km it’s about 2 ms not 40. The lowest orbit of the satellites though would be 550km and not 350km, but it would still be less than 4ms. From where did you get the 40ms figure that it’s wrong by an order of magnitude?
I don't know where they got that figure from, but it is unrealistic to not also expect a decent amount of latency to be introduced by the transmission hardware, particularly under load. I have no idea how much, but this has generally been the case with most systems of this sort.
Electromagnetic waves travel at essentially the speed of light. The speed of light is roughly 300km per millisecond. So if they are 550km up, that’s roughly 4ms round trip. Even with horrible inefficiency, it won’t likely increase that by 10x.
It can’t be both?