You need to add more latency for every time a device is encountered in the total path. As long as the amplifiers are optical that's a non issue but the satellites and base stations will have to receive and rebroadcast and the fiber optic network will likely at some point end up terminated at some router and then there will still be a number of hops to the destination.

They aren't accounting for devices in the path for land nor space, so the question becomes are there more or less devices in the average path for each case, and by how much, and what's the average penalty occurred for each device in each case?

Undersea optic cables use optical amplifiers, which is a pretty neat trick and does not introduce latency, and so should not count as a hop.

https://en.wikipedia.org/wiki/Optical_amplifier

In the end it is probably mostly a matter of how many hops that will determine which is the faster path, the two paths are mostly equal in other respects, I suspect the latency incurred for each hop is roughly the same, assuming they use error correction and re-transmission on the link layers between the satellites.

If it all gets passed through blindly then it will be a bit quicker because there is no decoding step involved but I highly doubt that because it would lead to substantial inefficiencies and loss of capacity (the difference between a switched and hub). Smarter to be a bit slower and to maximize the routing capabilities of the satellites so as to optimize for the scarce resource: inter-satellite bandwidth.

You're correct, but I don't know enough about satellites to know their retransmit delay, nor enough about fiber optics generally. I figured it all may offset because you'll never get a fiber optic cable to go exactly where you want it to go, and you may end up a mixed medium route with something like copper->fiber->copper->air->fiber->copper route, or something like that.

Starlink is supposed to be direct-to-consumer, so no hops at the end points.

Only when the endpoint nodes are both Starlink customers. Otherwise there will be a gateway at at least one endpoint and a number of hops after that.

I'm quite curious to see what their antenna setup looks like in practice that's going to be an engineering marvel.

Indeed. These satellites move which means the antenna dishes have to do so as well. Or they'll use phased arrays, and phased arrays are not consumer-cheap (yet).

From what I've heard, the consumer end gear supposed to use phased arrays "about the size of a pizza box" and cost in the range of $100-$300.

Aren’t TV satellites something like 5000-30000km away? 200km sounds easy.

Those are geostationary, these are not. So that's 'easy' as in 'a mere matter of engineering'. You will either have to track the satellite mechanically or by 'beam steering' aka phased arrays. Neither of those are trivial technology wise.

any Idea how big one of these antennas needs to be? Could you mount it on a gimbal similar to a DJI is mom and just have it aim automatically where the signal is strongest?

Is it feasible for the satellites to optically relay?

They could use laser based comms from one satellite to another:

https://en.wikipedia.org/wiki/Laser_communication_in_space

But optical signal amplification is a trick that really only works when you are on a very long run where the signal gets attenuated to the point that you need to boost it. There is no routing option in such a system, it is essentially point-to-point with in-line amplifiers.

So even if they may use laser based comms I don't think optical amplification will be used, the more likely avenue is to decode the optical signal, decide where it has to go next and then fire off another burst.

Alternatively they may use radio waves instead of lasers, I'm not sure what system SpaceX will use for this particular system, given the accuracy required for laser based communications I would assume they will use radio.

I've been reading some more about StarLink, apparently they did a test using lasers so probably that's what they will use for satellite-to-satellite communications, and radio to the ground because of atmospheric interference and sensitivity to rain and clouds.

https://www.cnet.com/news/spacex-starlink-satellite-broadban...

Are you talking satellite-to-satellite relaying? Assuming they're all moving relative to each other, then probably not, unless the laser transmitter or receiver are moving as well. And moving parts wear down and require power, so you'd want to avoid them on a satellite if possible. However, if parts of the constellation are following each other in exactly the same orbital path, then those parts might be able to optically relay without moving parts.

My first job in 1998 was working on the integration and test plans for Teledesic’s Optical Inter-Satellite Link Tracking (or OISLT, pronounced “oyselt”—probably the worst acronym I ever used). The design wasn’t done yet, but we were already planning integration tests to prove them out before anything was launched.

We were planning on tracking satellites in adjacent orbital planes (which were orbiting in the opposite direction) as well as in-plane. So, yeah, moving parts.

Whenever this topic comes up, I wonder what happened to all the work we did at Motorola in 98-99. There were several hundred engineers in Tempe generating thousands of pages of design documents for over a year. We archived all our work in spring 1999 sand were all shuffled off to other projects.

Yes. They already did some tests doing this.

Also, you need to have a cloud-free line-of-sight to the satellite.

The path between satellites would zig-zag.

Presuming a flat surface of 200 X 200, and a random line across half of that, and presuming that satellites only communicate with nearest neighbours you can guesstimate that the extra distance is significant (ignoring other factors).

If they can communicate with more distant satellites then distance could decrease (short cut segments around great circle).

The path taken by fiber optic cables is at least as bad, though. Even before you count the slower speed of light in glass.

The satellites will be meshed over 83 orbital planes starting around 1100km up. The satellites will connect to the farthest satellite it can see that makes the shortest path to the egress. So 2~3 hops verse a datacenter hops to the fiber ingress..then out the fiber across the atlantic.

I did the same math, got a similar result. I found that getting data from one side of the earth to the other should be about 17.4% faster in space.

Since the earth is 6,378 km in radius, adding another 1,150 km only makes the circumference about 18% longer.

The latency comment for repeaters someone else made is important, but it also won't be going in a circle between satellites, but in straight lines through the circle (as far as geometry allows), which will help shave a few ms.