Hacker News Reader

Comment by myrmidon | Hacker News Reader

Very interesting article. It also illustrates how dangerous unit conversions are especially combined with thousands separators: 20000km (GNSS satellite altitude) is NOT 12.4 miles.

Does anyone know if it would be feasible nowadays to just use starlink (or other LEO satellites) as a GNSS constellation? Even without precise onboard-clocks, would it not be possible to just bounce clock signals from earth as long as latency is known?

View on HN

Yes, this is actively being explored, see for example this for Iridium: https://investor.iridium.com/2021-05-24-Iridium-Makes-Strate...

> as long as latency is known?

Ah, but how do you know latency, without accurate clocks? :) Accurate clocks facilitates measuring latency, which is used to calculate distance. :)

True, but the accurate clock doesn’t need to be aboard the satellites as long as they have continuous (low-jitter) connectivity to the timing source, and Iridium and newer Starlink satellites do.

How does NTP work? GPS is stratum 0. Starlink could be effectively stratum 1.

Elon tweeted a while back they could do positioning, but it’s just not a high priority right now

starlink is in a low enough orbit that it isn't ideal (atmospheric drag will make the position harder to accurately predict)

I’d be surprised if that were an issue.

Starlink probably could learn the position of every satellite very accurately based on latency and signal strength and angle data from all the ground stations alone if they wanted to, if they don’t already.

you would need nanosecond accurate clocks in the satellites and groundstations

Reply on news.ycombinator.com