Hacker News Reader

Comment by elteto | Hacker News Reader

This video [0] really puts into perspective what SpaceX is trying to accomplish. Incredible stuff if they pull it off!

[0] https://www.youtube.com/watch?feature=youtu.be&v=AdKNCBrkZQ4...

FYI, the constellation was revised and an updated video was made. https://www.youtube.com/watch?v=QEIUdMiColU&feature=youtu.be

That is amazing. Seeing the complexity of the grid is mind blowing - and the realization the blocks representing the satellites are the size of cities makes you realize how big space is. That said, future launches are going to have to run a gauntlet...

It seems like that doesn't it, but space is MASSIVE. Imagine if there were only 7000 people spread out on the face of the earth, would you be worried you'd accidentally run into one of them? These things are up another few hundred miles so are even more spaced out than that.

This is true but not as useful an intuition as it first appears. The major difference for satellites is that everything in low-Earth orbit is traveling at ~8 km/s, so the spatial volume they traverse in any amount of time can be surprisingly large. A cubic-meter satellite traverses 250 cubic kilometers of volume in a year. (If you want to get an order-of-magnitude estimate for collision frequency, you imagine one satellite occupies it's normal volume and the other occupies the traversal volume, and then just ask what the chance is that those volumes overlap, given the amount of space above the Earth within the altitude range of the orbits.)

And indeed, there has already been an accidental satellite-satellite collision (not just satellite-debris collision).

https://en.wikipedia.org/wiki/2009_satellite_collision

The Iridium constellation has about ~80 satellites, and one passes within 5km of another satellite about 50 times/day.

If Starlink didn't route data in space and instead used space only as a last mile, would it be significantly cheaper, or have higher bandwidth, or be feasible to provide without data caps? What are the economic and technical tradeoffs here?

I don't mind the latency of today's fiber or cable Internet, and if it's problem, it can be solved by moving servers closer to users.

That is, what if, instead of bypassing the fiber backbone, Starlink just tried to connect everyone to it? Because I'm assuming that sending a terabit of data between two major cities, like Chennai to NYC is going to be cheaper via undersea fiber.

The problem it certainly going to be uplink/downlink - even with this massive constellation, you're fairly limited by the available RF spectrum. If you had a decently high subscription rate in a reasonably densely populated area, you'd probably be down to dozens of megabits per second per user at most.

This could deliver absolutely massively better internet in less populated areas like rural, shipping, etc. but I feel like people are perhaps getting the wrong idea and thinking this could supersede terrestrial networks in urban/surburban areas - in which case there will be a lot of disappointed people.

I didn't ask about users in densely populated areas. Imagine two villagers on opposite sides of the world communicating via Starlink. Would it be cheaper per bit for Starlink to offload the data to the nearest metro area, from where it can travel around the world by fiber, till the closest metro area for the receiver, at which point it gets sent via satellite? In other words, use satellite only as the last mile on each end, not for the backbone.

If you're in the middle of the Pacific it won't work unless the satellites talk to each other and route the data to a ground station that may be thousands of miles away.

Sure but most people don't live in the middle of the Pacific, so is an optimised system possible for them?

80msec is not a good outcome considering the cost and complexity of the investment. Aren't there alternative technologies to achieve this connectivity?

I think you might be underestimating latencies in the real world.

These [0] are statistics from Verizon showing 90ms RTT for trans-atlantic connections. Trans-pacific is > 100ms. 80ms seems highly competitive for me in this context.

[0] https://enterprise.verizon.com/terms/latency/

there is a video posted in the comments that compares latencies to cables. spacex advertises latencies from 50-80 , but cable is in all cases fasters (rather obvious i think).

It is a good outcome considering the speed of light and existing technologies. I don't know what you rightfully expected if you were hoping for serious improvement on that number, other than breaking the laws of physics.

my question is if the slight improvement in latencies makes the investment worthwhile

It's still a somewhat good value, compared to GEO satellites. It's no fun to use those.

Another interesting watch is the TED talk by SpaceX COO Gwynne Shotwell earlier this year.

https://www.youtube.com/watch?v=Dar8P3r7GYA

Say what you will about Musk, but the guy is truly ambitious and willing to shoot for the stars ( or mars at the moment ).

Reply on news.ycombinator.com