Hacker News Reader

Comment by supernova87a | Hacker News Reader

Does anyone have an image of what one of these ground stations looks like? Do they piggyback on some existing antenna shack? Or did they build their own from scratch?

I also wonder whether they considered partnering with AWS Ground Station? https://aws.amazon.com/ground-station/

Or maybe the Starlink constellation's requirements are not amenable to that?

View on HN

There are photos here:

https://www.reddit.com/r/Starlink/comments/m8maa9/wise_nc_st...

Across all the photos I've seen it looks pretty consistent, they build a fence and mount the antennas on small pads. It doesn't look like Starlink is building their own shelters (the ones in those photos are the carrier's). The carriers will readily offer rack space in their shelters for the right price but this does tend to suggest that the antennas are pretty self-contained and don't require a lot of external support equipment.

Those pads look like the brackets are there for for forks to move them around?

I am curious what the power consumption and bandwidth are for each of those units.

These were posted by a user on Reddit

https://m.imgur.com/a/DkfdjbU

Basically a gravel lot fenced in with a bunch of radomes mounted to what look like prefab concrete pads.

SpaceX's original plan was to build environmentally controlled pods/huts to hold equipment to aggregate ground transceivers. However they found it more economical to build un-airconditioned pods/huts and use someone's extended-temperature equipment instead.

Source: my employer is that someone.

What about huge layers of Schnow?

Or do they only use this method in areas where it's not schnowing?

> also wonder whether they considered partnering with AWS Ground Station? https://aws.amazon.com/ground-station/

Absolutely not since kuiper will be a direct competitor, and starlink's earth station design is their own proprietary equipment.

The majority of them are colocated with long haul DWDM regen huts, where they buy transport to the nearest major city. In the western US states they're adjacent to either Zayo or lumen (CenturyLink) sites.

Amazon is also planning an LEO internet constellation; being dependent upon a direct competitor for core infrastructure probably isn't a good idea, especially when that specific competitor (Amazon) has been busted before abusing its position as vendor to appropriate their customers' private information and enter their markets directly as participant.

If you are going to compete with Amazon, don't use AWS to do it.

I also doubt AWS ground station locations are numerous enough to provide the kind of service that Starlink needs.

Ah, that is a good point, of course.

What kind of ground station density does Starlink need in its eventual state?

The ground stations I saw on the map had 2.1GHz of uplink and 1.3GHz of downlink. If we assume 4b/hz (about as efficient as LTE/WiFi), thats 8.2Gbps up and 5.2Gbps down. I'm not too sure to what extent they can use MIMO techniques, but if they can, throughput could be multiplied by the number of MIMO channels.

I don't think you'd have much benefit of MIMO on a connection that's line of sight like this. MIMO is all about making the most of separate connection paths through reflections from walls, buildings etc. But in this case there'll be only one path.

Not necessarily: you could connect to several satellites and multiply your bandwidth that way.

Good point, once they have enough that could work.

Currently AWS Ground consists of a very small set of sites and antennas, not enough for any constellation. It's also very limited in terms if services.

Reply on news.ycombinator.com