24 comments

[ 3.6 ms ] story [ 59.9 ms ] thread
Ahem! ...could be as low as 8ms latency.
edited to make quote more accurate
Yeah I'm struggling to get too excited here. Theoretically latency with current geostationary satellites "could be as low as" 477ms, but they're usually 600ms+.

Plus this is for v2, which Elon has said will be launching on Starship ... so some years away.

To be fair, geostationary orbit is 36000km, whereas Starlink is at 550km. Geo hits, yes, roughly 400ms under perfect conditions whereas 550km would be expected to hit 9ms.
These are _not_ geostationary orbits. Starlink's network is of a fast/low/hand-off variety rather than a high/stationary/single-satellite->communications variety.

Their satellites are MUCH closer and are constantly whizzing across the horizon (and so require a motorized antenna) than the transitional geostationary communications satellites. This would require a huge number of satellites to get full-sky-coverage, but that's exactly what Starlink is doing.

1) Speed of light in space is faster than speed of light in a fiber optic cable...

2) And the number of hops in the void of space could be MANY fewer than from city to city to ocean to city.

I'm pretty sure the antenna is not motorized but is a phased array instead.
Perfectly straight upward, to the lowest-flying tier of satellites, just pinging the satellite itself ... maybe?
If they are able to keep the orbit at 550km and get the latency down to 3.66 ms, they will have changed physics forever.
I hope Elon has a special pricing tier and high bandwidth limits for primarily solar-powered devices that spend most of their duty cycle airborne --that truly would be transformative as existing cell towers are typically only available near highways and commerce centers.
FCC review of their preliminary application (I believe for access to the subsidized US rural market) said that they had a hard time believing the total network roundtrip would have better than 100ms latency, and now it's 8ms?
Indeed. This smells like a company I worked for about 20 years ago. The engineering team had an idea and were checking feasibility. The sales team overheard a meeting about it and proceeded to prepare a marketing pack for it which leaked out into the IT press. Analyst said it was rubbish (they were correct). About a month later the engineers had given up. Another month after that they were told their product had the first customer and nearly fell of their chairs.

Incidentally they did deliver after slowly lowering expectations over the period of 6 months.

Why would you believe the FCC over Elon Musk?

Then.... they have not even commented specifically on Starlink.

Then.... it was just one page in 141, hardly a report or anything of note.

https://docs.fcc.gov/public/attachments/FCC-20-77A1.pdf (Page 42)

Elon Musk has put billions into this, and it's not going to really work if it's above 100.

Clearly the 8ms is a optimistic hope, no one expects it to be that for all people all the time. But it'll be below 100ms.

Because the FCC doesn't care, but Elon Musk has billions on the line dependent on getting below 100 latency.

Musk has a reason to misrepresent the facts... And a history of doing so.

musk also has a history of delivering the products he spouts off on twitter about. they are often delayed but his track record is pretty good on actually delivering working products that do what he says.

I wouldn't be surprised if starlink hits around 20ms.

"Because the FCC doesn't care"

Not at all true. The head of the FCC, Ajit Pai, is very much in the pocket of the telecom industry.

Musk needs that latency below 100ms for phase 1 or he could get shut out of the rural auction. The auction is next month so the PR blitz is on. He's also running early trials so there's hard data to give to the skeptics at the FCC.

https://www.teslarati.com/spacex-starlink-user-terminal-beta...

This is a very roundabout PR blitz. Just show a speedtest!
My hunch is that version 1 isn’t quite at the 100msec mark or he would be showing ping tests.

Version 2 is the real starlink but it’s not going to be ready for this crucial window. So it’s going to take some FCC finesse to move ahead.

That's the obvious inference, but how is Starlink 1 over 100 ms when their goal is 20 ms?
The second version will use laser links between satellites.
That should make latency worse, not better. Right now all traffic goes through just one satellite so it doesn't make sense that latency is so much worse than projected.
End to end latency to things on the internet of 100ms isn't a deal breaker. There are a number of things out there on the internet that are far more than 100ms away from me, because they're on the other side of the atlantic or pacific oceans. Exactly what distance and topology this 100ms is being measured over needs to be specified.

There are already a number of residential and small business last mile access technologies that introduce 20-25ms latency. It can be 20-25ms latency from an ADSL2+ modem to the DSLAM. Or on a moderately loaded DOCSIS3 network latency from modem to CMTS can also be 20-25ms.

Starlink is expected to be 20-25ms for RTT latency from the CPE to the earth station. From the earth station (those of which I know about are colocated with long haul fiber sites), latency to destinations on the internet would be the same as terrestrial internet customers in the same location.

It's important to define two different things:

a) space segment latency, which will be 20-25ms

b) terrestrial latency, which will be exactly the same as any other properly-peered and interconnected large domestic ISP (such as a Wave, Charter, Comcast, Centurylink, etc).

The satellites will be serving as bent pipe relays.