Hacker News Reader

Comment by NeutronStar | Hacker News Reader

Considering Starlink et al., thoses cables won't be as needed in the future hopefully.

any additional satellite capacity is good, but that's comparing apples to oranges. the capacity of any satellite/RF based system will be significantly less than submarine fiber. the capacity of one singlemode strand of fiber is measured in the hundreds of terahertz.

Just one strand can support a ridiculous number (40+) of 100/200/400 GbE, coherent modulated signals in a DWDM bandplan.

I found it interesting to read that Starlink will use optical links between satellites. You don't need fibers in space. Aim well and your bandwidth per watt will be comparable to fiber on earth. I don't know about the capacity through the atmosphere which is what you're getting at I assume.

No. The point is that you can increase capacity of a fiber link by adding more fibers. Each fiber is an independent link. With radio, you have just a single link between two points, and you can‘t just put two dishes next to another because of diffraction.

I think lolc was referring to optical links in space specifically — no radio involved. No fiber necessary (or possible) because of line of sight.

An earth-to-space optical link would still suffer from diffraction, right? As I understand it, this is also the same reason why it's dangerous to point laser pointers at airplanes.

It's dangerous to point lasers at airplanes because laser light doesn't decrease as 1-over-r-squared. Which means the human pilot can be temporarily blinded by the beam, even many miles from the source. It has nothing to do with diffraction.

Thanks for that explanation!

This is incorrect. Laser beams are only "straight" at short range. At long range, laser beam intensity does decrease as 1/r^2. The reason for that is that the diameter of the laser beam increases due to diffraction. If you want to know why, I recommend reading about Gaussian beams and Rayleigh Range.

It’s dangerous to point lasers at planes because if you hit the window the light scatters and illuminates the complete window, blinding the pilot.

There are videos on YouTube if you want to see what it looks like. It isn’t like you’re pointing directly into the pilots eye, the chances of that are rather slim.

What exactly does it mean to have an optical link in space that's not a radio? Radiowaves are light, aren't they? Is this just semantics in regards to light near the visible spectrum vs not?

Frequency matters; masers are much more challenging than lasers, for example. But in any case, a direct laser link can carry enormous amounts of data. Even indirect LED transmission has been shown to viably carry many Gb/s. The challenge, of course, is the aim. It's not that easy to hit a pea that's several thousand of km away. I imagine feedback aiming systems where the vessels steer each other's transmitters, but that requires some amount of surface area for reception and that part doesn't scale as easily as fibers do.

Diffraction affects laser beams just like all other radiation. You can't just focus a laser beam on a pea over a distance of 1000km. (well, theoretically you could, but you need a 20m parabolic mirror to focus it)

You can probably get 100gbit/s between satellites, maybe more. A single fiber link can already support several tbit/s, some submarine cables can transport 40tbit/s or more. And the Atlantic and Pacific are full of submarine cables, 20k satellites won't be enough to replace any of that. Just the amounts that Google/FB/AWS etc push between continents is more than a satellite link could sustain.

For latency sensitive applications (e.g. finance, video links), starlink can indeed be a good alternative but it won't replace backbones, e.g. distribution of youtube videos to edge locations around the world.

I can't imagine being able to establish multiple spatially separated beams between satellites that are hundreds or thousands of miles apart. If the beam were small enough to not 'paint' the entire receiving satellite, it would be ridiculously sensitive to alignment. Even the tiniest movement would break the link.

Realistically, using optics, I can only see one spatial path to another satellite. You can use various wavelengths of light to establish multiple carriers, and probably get upwards of 1-2 Tb/s of speed. But nothing near what is possible in the same space with fiber optics.

A 288 strand optical fiber cable, with jacket and armor is about 2cm in diameter. And each one of those 288 strands can carry terabits per second of data. It will be a very long time before free-space optics can match that kind of density.

I realize that space is big but those optical links also become susceptible to obstruction. This is not a problem with cables.

Agreed, but I'd say the likelihood of something in space disrupting an optical path between two satellites for any significant duration is quite small, much smaller than the potential for a cable to be physically damaged.

But I'm still quite certain that satellites nor free-space optics will not be coming anywhere near the capacity of fiber in the foreseeable future.

Doesn't really matter (yet). Satellites presently still need to downlink any transferred data and that's going to be the rate-limiting factor.

I agree that alignment and interference are problems if one wants multiple beams. If you're using lasers one way I see is to split each signal beam into a grid so it's many rays in a grid pattern where a few rays will hit the target sensor array. Multiple grids can then be overlaid without interference. The sensor grid at the target must be able to detect which sensor is receiving which beam and adapt over time.

Overall that doesn't sound infeasible to me. But then I don't do optics, so maybe I'm ignorant of a basic limitation.

a 288 strand cable wouldn't really be the best comparison vs satellite, in the context of submarine fiber. long distance submarine fiber is usually 2 to 8 strands maximum. the huge amount of throughput possible is achieved through a combination of high-powered, dense channel spacing, DWDM systems on the terrestrial ends of the fibers (google "ITU 25 GHz DWDM grid"), and modern high density modulation systems like 100/200/400GbE coherent (QPSK, 8PSK, 16QAM + FEC) links.

But even a submarine cable with <8 strands can have 40tbit/s. And they are still considerably cheaper to install than satellites (esp including long-term maintenance).

EDIT: As an example, the new Atlantic cable "MAREA", partly sponsored by Facebook and Microsoft, has a capacity of 160tbit/s.

http://www.datacenterknowledge.com/networks/facebook-and-mic...

I do not think that the total bandwidth will be comparable to fiber. Even if every starlink satellite has a 200 Gbps full duplex link to the satellite "ahead" of it in orbital plane, the satellite behind it, and the two satellites to the east and west of it in adjacent orbital planes, it'll still be vastly less than the capacity of two strands of singlemode. 200 Gbps x 4 is about 4/76th of the capacity with a modern DWDM system (assuming a maxed out point to point topology such as 80-channel bandplan, QPSK modulated coherent signals in the fiber C-band spectrum).

Reply on news.ycombinator.com