I read in the Economist that the gen of starlink satellites will have the ability to route messages between each satellite. Would conventional routing protocols be up to such a challenge? Or would it have to be custom made for that problem? And since a lot of companies and countries are getting on that action, it seems like fertile ground for (bad) wheel reinvention?
Mike
Unlike most terrestrial links, the distances between satellites are not fixed,
and thus the latency between nodes is variable, making the concept of
“Shortest Path First” calculation a much more dynamic and challenging
one to keep current, as the latency along a path may be constantly changing
as the satellite nodes move relative to each other, without any link state actually
changing to trigger a new SPF calculation.
I suspect a form of OLSR might be more advantageous in a dynamic partial
mesh between satellites, but I haven’t given it as much deep thought as would
be necessary to form an informed opinion.
So, yes–it’s likely the routing protocol used will not be entirely “off-the-shelf”
but will instead incorporate continuous latency information in the LSDB,
and path selection will be time-bound based on the rate of increase in latency
along currently-selected edges in the graph.
An interesting problem to dive into, certainly. 
Thanks!
Matt
Mike
Satellites move constantly relative to each other and to ground stations. There is a database available which contains the parameters for calculating a satellite’s location at any instant in time. To maintain minimal link disruption, the idea is to calculate these relative relationships, and using some graph and network flow algorithms, you pre-calculate the links and then insert/remove those links and routes into the routing information base at the appropriate times. Then based upon latency, signal quality, and link availability, routing information is inserted/deleted into the forwarding information base. There are other contributors such as link saturation and overall end-to-end delays which could be applied based upon ground station state management. It becomes a multi-parameter link selection algorithm in a dynamic environment. Pretty much an interesting ‘sdn’ like scenario.
I suspect, although I have no references, that satellite to ground connectivity is probably more “circuit-based” than per-packet or frame.
Iridium has done inter satellite communication for decades. I wonder if it wouldn’t be something very similar. Although it would be totally on-brand for them to do it some “revolutionary” new way whether it actually makes any sense or not.
Yes re: Iridium. Contrary to what the Chief Huckster may say, inter-sat comms are not some revolutionary thing that he invented.
It’s also not likely to function anything like they show in marketing promos, with data magically zipping around the constellation between nodes in different inclinations. Unless they have managed to solve for the Doppler effect in a way nobody has thought of yet.
For further reading try:
-Hank
I know of a group of satellite FPGA/RF guys who have worked on this doppler thingy. It is a solved problem.
Solved years ago …
https://ieeexplore.ieee.org/ielaam/92/8502886/8412572-aam.pdf
-Jorge
I think the thing they’re calling revolutionary is the idea of those links being directional lasers.
It makes some sense… if you can basically emit the same signal you’d shoot down a strand of single mode but aim it through the mostly vacuum of space in the exact direction of your neighbor then you’ve got something… Essentially the equivalent of a fiber optic network in space.
For fun I tried plugging in some frequencies of light into a doppler calculator. Unfortunately that’s where my “would the relative speed that mere mortals could attain make enough of a difference to affect a typical optical receiver” investigation ended as I’m mobile right now and can’t do the rest of the work very easily. I’d be curious if the relative speed would be enough to cause enough shift to move it out of the pass band if a typical dwdm channel.
And, I agree that little of what musk takes credit for is revolutionary. But what I do think he deserves credit for is being insane enough to try things everyone says is unworkable and failed in the past and somehow making at least some of them work. Having more money than God helps too.
Matthew Petach wrote:
Unlike most terrestrial links, the distances between satellites are
not fixed, and thus the latency between nodes is variable, making the
concept of "Shortest Path First" calculation a much more dynamic and
challenging one to keep current, as the latency along a path may be
constantly changing as the satellite nodes move relative to each
other, without any link state actually changing to trigger a new SPF
calculation.
As LEO satellites should be leafs to a network of MEO satellites,
1 minutes of update period between MEO satellites should be enough,
which is not so dynamic.
Physical layer of MEO communications must (to save power and to
prevent broadcast storms) be point to point with known orbital
elements and link layer should be some point to point protocol
perhaps with ARQ.
As only meaningful metric between satellites is physical
distance, 16bit metric of OSPF should be enough.
The most annoying part is to have multiple ground stations,
which, as usual, makes the MEO network DFZ with more than 1M
routing table entries.
Masataka Ohta
Musk didn’t do anything revolutionary, besides launching a shload of LEO satellites.
NASA and DoD have been working for long time on optical space communications, last year LCRD was launched and preliminary tests using it as a relay showed 622Mpbs, this year NASA will include on one of the cargo missions to ISS ILLUMA-T that will be installed at ISS and it is expected to provide 1.24Gpbs or more using LCRD as a relay with the two ground stations, one in HI, and one in CA.
DoD/NRO have been working on this for some time now, but any information is in the top secret blackhole.
-J
-Jorge
Like I said, they’re calling it revolutionary. Didn’t say it was.
However the idea that you can build spaceships which are fully reusable was certainly around the industry, but the consensus was largely “we tried, it costs too much, so we’re sticking with one use rockets”. Elon for whatever reason is insane enough to dump a lot of cash in industries which everyone said was a dead end and then has been lucky enough to prove the old guard wrong.
Same for pretty much everything musk does, including starlink. So if there is anything at all “revolutionary” here it’s the insistence on ignoring conventional wisdom. I think it might be borderline insanity, but it seems to work for him.
Raymond / Jorge -
Thanks for that info. Quoting from the paper, that does match my current understanding, being :
II. FEATURES OF INTER-SATELLITE COMMUNICATION LINKS AND DESIGN CONSIDERATIONS This work is aimed to design efficient ISCs links for a group of small satellites flying in cluster formation, in which 100∼150 small satellites are deployed in a limited spherical space with a diameter of 100 meters at a altitude of 500 Km above the earth.
The doppler effects in specific circumstances are solved for, yes. For example the Iridium constellation has long done ISC, but only between birds currently orbiting in the same direction, because their relative speeds were close enough that the doppler effects were manageable.
What I didn’t think was adequately solved was what Starlink shows in marketing snippets, that is birds in completely different orbital inclinations (sometimes close to 90 degrees off) shooting messages to each other. Last I had read the dopplar effects there were so much larger due to relative speed deltas it just couldn’t currently be done. If there is more out there on that solution, be glad to read up on what info anyone may have on that if they can share.
Elon for whatever reason is insane enough to dump a lot of cash in industries which everyone said was a dead end and then has been lucky enough to prove the old guard wrong.
Nobody had ‘given up’ on reusable launch vehicles. SpaceX (to their credit) just made it a core requirement in Falcon9 design from the outset, and was able to execute it.
Nobody had ‘given up’ on electric cars before Musk pushed the original founders of Tesla out.
Musk took Solarcity in the opposite direction (down) as the rest of the US solar industry grew.
Starlink still hasn’t proven any of the ‘old guard wrong’. Is Starlink operational? Yes. Has he proven it to be a viable business? No. (In fact, if you basic math on the numbers they espouse, it can’t be.)
Same for pretty much everything musk does, including starlink. So if there is anything at all “revolutionary” here it’s the insistence on ignoring conventional wisdom. I think it might be borderline insanity, but it seems to work for him.
It ‘seems to work for him’ because :
1990s Iridium was a modified version of GSM/ATM with the packetization
and routing that implies. I don't know the current constellation's
architecture but I'd be shocked if they had reverted to a bent pipe
architecture.
For those not in the know, a "bent pipe" communications satellite is
one which accepts a radio signal in one frequency and does an analog
transform to another frequency before sending it back out. Up from the
ground station on one frequency, transform, down to the customer. Up
from the customer on one frequency, transform, down to the ground
station on another.
The nice thing about a bent pipe is that you can upgrade the service
equipment to higher speeds without changing the satellite. The
satellite doesn't care. It doesn't recognize the concept of bits or
packets. The bad thing is that it's straight up and down, so when the
satellite isn't both in range of the customer and a ground station,
you can't use it.
The vast majority of satellite architectures are bent pipe.
Regards,
Bill Herrin
Worst case would be if the satellites are moving directly towards or
directly away from each other. Each satellite will be moving at a
speed of slighly under 8 km/s, and they will thus approach or depart
from each other with a relative speed of somewhat less than 16 km/s.
I get that for 1310 nm light, the doppler shift would be just under
0.07 nm, or 12.2 GHz:
l0 = 1310 nm
f0 = c / l0
f = f0 / sqrt((1 + 16 km/s / c) / (1 - 16 km/s / c))
l = c / f ≈ 1310.0699 nm
f0 - f ≈ 12.2 GHz
In the ITU C band, I get the doppler shift to be about 10.5 GHz (at
channel 72, 197200 GHz or 1520.25 nm).
(Formula from Relativistic Doppler effect - Wikipedia
first entry in the table under "Summary of major results".)
These shifts are noticably less than typical grid widths used for
DWDM (±50 GHz for the standard spacing), so it seems unlikely to me
that the doppler shift would be a problem.
/Bellman
I think it’s also likely that only modest, if any, WDM is required on those links, because the goal in most cases will only be to go far enough to get down to a ground station (excepting some low latency transatlantic use cases I have read might be in the offing), and because the satellite RF uplink/downlink capacities shouldn’t seriously challenge the bandwidth available on the optical links.
At least in the current case of general purpose internet access with dynamic IP addresses, I suspect the IP of a user-terminal is related to the ground station serving it, and there is just a parade of satellite intermediaries, but the terminal and ground station remain fixed, so the routing can be more of a connection oriented type.
Appreciate that. Definitely becoming clear to me that a lot of my knowledge here was rusty. Lots of papers on this specifically (Doppler effects on optical ISL) that I need to call in some favors to get access to.
Thanks!
And as I was bicycling home, I of course thought of another aspect
of the doppler shift: the timing between the symbols in the signal,
or in other words the baud rate. There will be something like a
phase-locked loop (PLL) in the receiver in order to know when one
symbol ends and the next one starts, and that PLL can only deal
with a certain amount of baud rate shift.
But we can use the same formula. And in general, the doppler shift
for 16 km/s is about 53 parts per million. So e.g. a 112 Gbaud signal
would be received as 6 Mbaud faster or slower than it was sent at.
And here I have to confess that I don't know how generous typical
receiver PLL:s in network equipment are.
Another potential aspect might be the decoding of phase-shift keying,
i.e. when phase modulation is used for the signal. My *very*vague*
understanding is that the typical way to decode a phase-modulated
signal, is to mix the incoming signal with a reference carrier wave,
generated locally by the receiver, and the interference between the
two gives you the actual signal. But to do that, the reference must
have the same frequency as the received wave, and, I guess, must
match very closely. Can they adapt to an incoming wave that is 53 ppm
offset from what it should be?
Or have I misunderstood this? Analogue signals is very much *NOT*
my forte...
/Bellman