5 ms·

I agree. This to me is the major problem with starlink. Each sat only has 10-20gigabit/sec capacity, which really isnt a lot considering Netflix can use 15mbit/

I agree. This to me is the major problem with starlink. Each sat only has 10-20gigabit/sec capacity, which really isnt a lot considering Netflix can use 15mbit/sec per stream, and call of duty patches can be nearly 200GB. It's not going to take many users to completely saturate 20gbit. It's not far off what a single 4G+5G cell site has, capacity wise, but instead of serving a few km max we are talking 100s of km^2. I think it's highly likely there will be data caps and/or fairly serious traffic shaping on the production version at some point.

rektide 6y ago
The subscription ratio (people: satellite) is a big one. I'm also quite nervous about uplinks. According to dlgeek there are 893 satellites up right now [1]. If we grow to 10x that number, and each satellite can do 20Gbit (where is that number from, btw?), how do we begin to provide that throughout, whatever fraction of that max 2.2Tbit/s aggregate downlink for the satellite fleet? What is it going to take to literally uplink such huge volumes? I'm not well versed in satcom & how much spectrum is available, & how good ground stations are at pointing. Can one base station potentially fully saturate 10 or 50 or 250 satellites, once the sky starts to fill? Maybe we are that good, maybe we can shoot tiny narrow high-bandwidth wide-spectrum needles exactly as needed up into the sky, en masse, dozens of high bandwidth beams emmentating from each ground station. Or maybe we just need millions of ground stations, but can work that. Does even that work? I keep hearing of "constellations," trains of satellites. How close will satellites stick? Can a ground station reuse spectrum across multiple transmitters seen talking to constellation members? or are the satellites too close, lumped together? this filters back to a previous question, how much spectrum is available for uplinks? Trying to imagine the scale of this project, seeing both how we get data to in to the sky, then down to folks, it's epic. All the "zero to one, one to many" modes of growth discussion feels short of the epic planning & demands it'd take to both hoist satellites up (which SpaceX Starlink seems to be doing!) & the supply network to make the satellites useful. [1] https://news.ycombinator.com/item?id=25109020 https://news.ycombinator.com/item?id=25109020
- mschuster91 6y ago
  > What is it going to take to literally uplink such huge volumes? Put a lot of sat dishes on the top of major Internet exchanges such as DE-CIX and on all data centers of AWS, GCE and Azure.
  - rektide 6y ago
    That's kind of the idea, sort of, yeah. But what kind of limits does such an attempt run into? Can you just keep adding more and more dishes, as long as you please, each one pointing at one satellite, with zero interference from any other of your ground dishes? No, I rather suspect not. When you are uplinking, there's probably going to be other satellites also getting some of your signal, especially as we move from 1000, to 10,000, to beyond numbers of satellites. This is not my field, but I feel like there are some very interesting limits to uplinking, a lot of constraints, that the sky itself can only hold so much uplink capacity, and it'd be interesting to me at least to have some more back-of-napkin discussions on what those constraints really look like.
    - Dylan16807 6y ago
      > When you are uplinking, there's probably going to be other satellites also getting some of your signal, especially as we move from 1000, to 10,000, to beyond numbers of satellites. Only starlink is going to be using this frequency band, so it doesn't matter if it hits anything else. And of the starlink sats, you're not going to have more than a handful in view at once, so aiming is easy. And as far as base stations go, you just need one every X hundred miles, with several dishes. The problems you're imagining could be real on a sufficiently dense constellation, but there is no reason to go anywhere near that density, so it's not a problem anyone has to worry about.
      - rektide 6y ago
        Every X hundred miles feels kind of like where I was going. Even then, there's a very active question of how many uplinks each given uplink ("gateway") would need to make, to provide the massive amounts of bandwidth required. Anyhow, I found a couple numbers on spectrum that are semi-interesting to me to assess what spectrum resources Starlink has purchased. According to this site[1], Starlink has: 27.5 – 29.1 GHz and 29.5 – 30.0 GHz and 47.2 – 50.2 GHz and 50.4 – 51.4 GHz = 6.1GHz of spectrum for uplink from base-stations ("gateways") to satellites 10.7 – 12.7 GHz and 37.5 – 42.5 GHz = 7GHz of spectrum for downlink from satellite to user "terminals". According to this page, a year ago they were authorized to run 6 gateways. I'm not sure how much throughput they can push to a satellite with 6.1GHz of total spectrum. It's still interesting to me to try to consider how even the current 893 satellites get anywhere near enough throughput up to them to be well utilizing their own downlink capabilities. I'm still only inuiting, but it still feels like something doesn't add up, I still don't see how the fleet is going to get enough bandwidth. My previous image of a data-center bristling with dozens & dozens of dishes pointing & tracking upwards indeed doesn't seem quite right (and the side-lobe interference question I was asking about), but the alternative, of a massive network of ground stations each boasting colossal connectivity, seems almost equally hard to imagine. [1] https://www.elonx.net/starlink-compendium/ https://www.elonx.net/starlink-compendium/

rektide 6y ago

The subscription ratio (people: satellite) is a big one. I'm also quite nervous about uplinks. According to dlgeek there are 893 satellites up right now [1]. If we grow to 10x that number, and each satellite can do 20Gbit (where is that number from, btw?), how do we begin to provide that throughout, whatever fraction of that max 2.2Tbit/s aggregate downlink for the satellite fleet? What is it going to take to literally uplink such huge volumes? I'm not well versed in satcom & how much spectrum is available, & how good ground stations are at pointing. Can one base station potentially fully saturate 10 or 50 or 250 satellites, once the sky starts to fill? Maybe we are that good, maybe we can shoot tiny narrow high-bandwidth wide-spectrum needles exactly as needed up into the sky, en masse, dozens of high bandwidth beams emmentating from each ground station. Or maybe we just need millions of ground stations, but can work that. Does even that work? I keep hearing of "constellations," trains of satellites. How close will satellites stick? Can a ground station reuse spectrum across multiple transmitters seen talking to constellation members? or are the satellites too close, lumped together? this filters back to a previous question, how much spectrum is available for uplinks? Trying to imagine the scale of this project, seeing both how we get data to in to the sky, then down to folks, it's epic. All the "zero to one, one to many" modes of growth discussion feels short of the epic planning & demands it'd take to both hoist satellites up (which SpaceX Starlink seems to be doing!) & the supply network to make the satellites useful. [1] https://news.ycombinator.com/item?id=25109020 https://news.ycombinator.com/item?id=25109020

mschuster91 6y ago
> What is it going to take to literally uplink such huge volumes? Put a lot of sat dishes on the top of major Internet exchanges such as DE-CIX and on all data centers of AWS, GCE and Azure.
- rektide 6y ago
  That's kind of the idea, sort of, yeah. But what kind of limits does such an attempt run into? Can you just keep adding more and more dishes, as long as you please, each one pointing at one satellite, with zero interference from any other of your ground dishes? No, I rather suspect not. When you are uplinking, there's probably going to be other satellites also getting some of your signal, especially as we move from 1000, to 10,000, to beyond numbers of satellites. This is not my field, but I feel like there are some very interesting limits to uplinking, a lot of constraints, that the sky itself can only hold so much uplink capacity, and it'd be interesting to me at least to have some more back-of-napkin discussions on what those constraints really look like.
  - Dylan16807 6y ago
    > When you are uplinking, there's probably going to be other satellites also getting some of your signal, especially as we move from 1000, to 10,000, to beyond numbers of satellites. Only starlink is going to be using this frequency band, so it doesn't matter if it hits anything else. And of the starlink sats, you're not going to have more than a handful in view at once, so aiming is easy. And as far as base stations go, you just need one every X hundred miles, with several dishes. The problems you're imagining could be real on a sufficiently dense constellation, but there is no reason to go anywhere near that density, so it's not a problem anyone has to worry about.
    - rektide 6y ago
      Every X hundred miles feels kind of like where I was going. Even then, there's a very active question of how many uplinks each given uplink ("gateway") would need to make, to provide the massive amounts of bandwidth required. Anyhow, I found a couple numbers on spectrum that are semi-interesting to me to assess what spectrum resources Starlink has purchased. According to this site[1], Starlink has: 27.5 – 29.1 GHz and 29.5 – 30.0 GHz and 47.2 – 50.2 GHz and 50.4 – 51.4 GHz = 6.1GHz of spectrum for uplink from base-stations ("gateways") to satellites 10.7 – 12.7 GHz and 37.5 – 42.5 GHz = 7GHz of spectrum for downlink from satellite to user "terminals". According to this page, a year ago they were authorized to run 6 gateways. I'm not sure how much throughput they can push to a satellite with 6.1GHz of total spectrum. It's still interesting to me to try to consider how even the current 893 satellites get anywhere near enough throughput up to them to be well utilizing their own downlink capabilities. I'm still only inuiting, but it still feels like something doesn't add up, I still don't see how the fleet is going to get enough bandwidth. My previous image of a data-center bristling with dozens & dozens of dishes pointing & tracking upwards indeed doesn't seem quite right (and the side-lobe interference question I was asking about), but the alternative, of a massive network of ground stations each boasting colossal connectivity, seems almost equally hard to imagine. [1] https://www.elonx.net/starlink-compendium/ https://www.elonx.net/starlink-compendium/