Starlink: when can we expect Martian deployment?

[ccdengr]

You could almost certainly launch a new-design GPS constellation in a single launch.

It cost 10 billion euros to deploy Galileo, was that just incompetence and waste?

[Solarsail]

What if we go half way between Starlink/GPS block III and MRO?

I'll propose some relaxed requirements:  Atomic clocks may be required of ground terminals, ground terminals may use 2-way communication with the satellites in position determination.  Each ground terminal may add compute load to satellites and visa-versa. Location precision is 20km, and you only need contact with it on a daily basis.  To do terrain surveys, drop portable terminals in a breadcrumb trail on the ground and wait for them to get contact with the orbiting system.  Pick them up a day later.  Comms also only needs to be available at least daily.

How few satellites can you get away with now?  I'd read somewhere that you could do a GPS system on mars using 1 satellite, if you can smear your position determination across the orbital arc, and require compute work on the satellite.  Could the same satellite take on communications work in a store-forward architecture?  Do you need 1 ground station worth of calibration to reach 20km precision?  More?  Can you do it with a synthetic-aperture-radar instead?  3 ground stations and 1 satellite?  1 ground station and 3 satellites?  If you can get to useful position date with 1 satellite and a single beacon, can the system gradually improve with expansions of the satellite constellation?

[edzieba]

Lets assume for a moment that our satellites have magical super-atomic clocks that never ever drift and can ignore relativistic effects, such that we do not need to concern ourselves with satellite-to-satellite relative drift, we can synchronise all the clocks at launch and then ignore sync forever. Lets also assume that the satellites themselves have such exquisitely precise ISL steering gimbals (well above the requirements for laser communicaitons) and truly outstanding laser pulse timing (for accurate ranging) that each satellite can create an exact model of satellite relative positions independently of any ground stations. (Both of these are extremely generous assumptions that either vastly inflate satellite cost or are plain not physically possible).

The clocks do not need to be perfect. They will be adjusted to reach a consensus among all the clocks in the network. This requires software, not expensive hardware.

The ISL only needs to be accurate enough to allow intersatellite communication. The  only thing a pair of communicating satellites need on this link is an extremely precise tic embedded in the signal, similar to the IEEE 1588 hardware-supplied tic. One of these every second will be more than adequate. A single tic can be as precise as the baud rate (i.e., the symbol rate, not the bit rate.) I do no know the baud rate for a Starlink ISL but with an aggregate transmission rate of 100Gbps the baud rate is unlikely to be less than 1GHz and the tic is precise to within one nanosecond (30 cm). The two satellites determine their separation by communicating the time of receipt of the tic as measured by their own clock. This requires a very small amount of extra functionality in the transmitter's encoder electronics and the receiver's decoder electronics, plus software.

Note that this can all be done using the "GPS" omnidirectional radio signals instead of the lasers, albeit much more slowly. They already have the equivalent of an extreme precision tic. This may be needed during deployment so the satellites can find each other in the first place.

You missed (or ignored) the entire previous post. Just having an accurate clock is not sufficient for a trilateration-based GNSS. You need precise knowledge of the exact satellite positions (both relative to each other and relative to the planet's surface) and precise knowledge of timing difference - not just averages, as its the timing differences you rely on for the critical time-of-flight ranging so accurate absolute timebases are required.

It is not as easy as you want to pretend.

[sdsds]

It is not as easy as you want to pretend.

Yes, establishing a Mars GNSS system will not be easy, and there's no current use case.

In stark contrast, a UHF-to-X-band store-and-forward communications system already exists and is in active use: MaROS. It utilizes payload packages on MRO, Maven and Mars Odyssey to forward data received via UHF from Mars surface stations to Earth surface DSN X-band stations.

For the first Starship landed mission does SpaceX want to rent time on that service? I somehow doubt it.

How does that well-defined MaROS payload package compare in size and mass to the standard Starlink payload? Maybe SpaceX could provide the launch service and satellite bus in exchange for a MaROS payload and time on the DSN. Alternately maybe Starlink could leap-frog MaROS and establish a UHF-to-Laser store-and-forward communications system between (specialized) Starlinks in Mars orbit and their equivalent in Earth orbit....

[waveney]

It is not as easy as you want to pretend.

In stark contrast, a UHF-to-X-band store-and-forward communications system already exists and is in active use: MaROS. It utilizes payload packages on MRO, Maven and Mars Odyssey to forward data received via UHF from Mars surface stations to Earth surface DSN X-band stations.

It is horrendously overloaded.  The available data rate seriously limits the science done by orbiters and landers.

[DanClemmensen]

Lets assume for a moment that our satellites have magical super-atomic clocks that never ever drift and can ignore relativistic effects, such that we do not need to concern ourselves with satellite-to-satellite relative drift, we can synchronise all the clocks at launch and then ignore sync forever. Lets also assume that the satellites themselves have such exquisitely precise ISL steering gimbals (well above the requirements for laser communicaitons) and truly outstanding laser pulse timing (for accurate ranging) that each satellite can create an exact model of satellite relative positions independently of any ground stations. (Both of these are extremely generous assumptions that either vastly inflate satellite cost or are plain not physically possible).

The clocks do not need to be perfect. They will be adjusted to reach a consensus among all the clocks in the network. This requires software, not expensive hardware.

The ISL only needs to be accurate enough to allow intersatellite communication. The  only thing a pair of communicating satellites need on this link is an extremely precise tic embedded in the signal, similar to the IEEE 1588 hardware-supplied tic. One of these every second will be more than adequate. A single tic can be as precise as the baud rate (i.e., the symbol rate, not the bit rate.) I do no know the baud rate for a Starlink ISL but with an aggregate transmission rate of 100Gbps the baud rate is unlikely to be less than 1GHz and the tic is precise to within one nanosecond (30 cm). The two satellites determine their separation by communicating the time of receipt of the tic as measured by their own clock. This requires a very small amount of extra functionality in the transmitter's encoder electronics and the receiver's decoder electronics, plus software.

Note that this can all be done using the "GPS" omnidirectional radio signals instead of the lasers, albeit much more slowly. They already have the equivalent of an extreme precision tic. This may be needed during deployment so the satellites can find each other in the first place.

You missed (or ignored) the entire previous post. Just having an accurate clock is not sufficient for a trilateration-based GNSS. You need precise knowledge of the exact satellite positions (both relative to each other and relative to the planet's surface) and precise knowledge of timing difference - not just averages, as its the timing differences you rely on for the critical time-of-flight ranging so accurate absolute timebases are required.

It is not as easy as you want to pretend.

It may in fact be a lot more complicated than I would want it to be, but I neither missed nor ignored any post on this thread. I have not responded to the "Galileo" post because I do not know enough about its funding decisions.

From a theoretical perspective, a system of multiple reasonably precise clocks that can communicate their local (X,Y,Z,T) with each other will be able to collectively refine their (X,Y,Z,T)s. When two or more of them are in "fixed" locations (e.g., on the Martian surface) This will eventually converge into a system with GPS-like functionality. Yes, I know satellites move.

To your points:

There is no such thing as "absolute" time. Here on earth, we use UTC and its kin. UTC is a consensus time based on coordination among multiple extreme-precision atomic clocks. Rubidium clocks are atomic clocks of lower quality but they still derive their accuracy from atomic transitions, and averaging enough of them will provide an adequate "absolute" time.

We used "fixed" ground reference locations that were originally surveyed by non-GPS means, because that's all we had to start with. The relative positions of those stations have since been refined using GPS, including any changes due to plate tectonics.

Relevance to Martian Starlink: Starlink depends on GPS-like positioning and will not work without it.

[ccdengr]

[MarOS] is horrendously overloaded.  The available data rate seriously limits the science done by orbiters and landers.

I'm not sure if "horrendously" is fair.  There are only two rovers operating now.  But certainly more capacity could be used.

Of course "Martian Starlink" for communication back to Earth would not only need sats to talk to the surface, but sats to talk to Earth.

As for positioning, it depends on how much accuracy you want.  https://en.wikipedia.org/wiki/Transit_%28satellite%29 got 200-meter accuracy with just a few sats.  Getting meter-scale accuracy or better is a lot harder, and I would argue is not needed any time soon.  Even if landers needed positioning information, it would be much simpler to set up a fixed network at the landing sites.  SpaceX only uses GPS now because it's an available resource they get for free.

[waveney]

[MarOS] is horrendously overloaded.  The available data rate seriously limits the science done by orbiters and landers.

I'm not sure if "horrendously" is fair.  There are only two rovers operating now.  But certainly more capacity could be used.

Of course "Martian Starlink" for communication back to Earth would not only need sats to talk to the surface, but sats to talk to Earth.

As for positioning, it depends on how much accuracy you want.  https://en.wikipedia.org/wiki/Transit_%28satellite%29 got 200-meter accuracy with just a few sats.  Getting meter-scale accuracy or better is a lot harder, and I would argue is not needed any time soon.  Even if landers needed positioning information, it would be much simpler to set up a fixed network at the landing sites.  SpaceX only uses GPS now because it's an available resource they get for free.

The main instruments on MRO can only be used for a limited amount each day.   HiRise in particular.  They could use 10 times the current bandwidth easily.

[DanClemmensen]

[MarOS] is horrendously overloaded.  The available data rate seriously limits the science done by orbiters and landers.

Of course "Martian Starlink" for communication back to Earth would not only need sats to talk to the surface, but sats to talk to Earth.

Agreed. "Martian Starlink" is for on-Mars and near-Mars communication. Starlink satellites are not optimized for Mars-to-Earth comms, which need different hardware and different protocols. The planet-to-planet links can and should be on different satellites. These satellites could connect to their Mars users via ISL links to the Starlink satellites.

[waveney]

[MarOS] is horrendously overloaded.  The available data rate seriously limits the science done by orbiters and landers.

Of course "Martian Starlink" for communication back to Earth would not only need sats to talk to the surface, but sats to talk to Earth.

Agreed. "Martian Starlink" is for on-Mars and near-Mars communication. Starlink satellites are not optimized for Mars-to-Earth comms, which need different hardware and different protocols. The planet-to-planet links can and should be on different satellites. These satellites could connect to their Mars users via ISL links to the Starlink satellites.

I thought that originally, but now think it would be simpler to equip them all with the planet to planet capability.  They can then be used in parallel to provide a much larger inter planet capability.  Different Sats would use different wavelengths to communicate to Earth at the same time. 

I originally thought a Martian GPS (MPS) would use separate sats to Starlink, but I now accept your arguments and think one design that can do everything is better.

One design that does all three functions is I think the best.  it provides redundancy, needs less sats in total.  One starship could deliver enough to provide all the comms and GPS needed for a long time,  It would have spare space for delivering other payloads to Martian orbit as well.

[edzieba]

From a theoretical perspective, a system of multiple reasonably precise clocks that can communicate their local (X,Y,Z,T) with each other will be able to collectively refine their (X,Y,Z,T)s. When two or more of them are in "fixed" locations (e.g., on the Martian surface) This will eventually converge into a system with GPS-like functionality. Yes, I know satellites move.

The entire mechanism of action of current GNSS trilateration systems is to use the timing differences between satellite clocks and the actual time of arrival. You cannot bootstrap clock offsets using the clock offsets, it's like trying to determine the absolute length of a standard metre by averaging three pieces of string of unknown length: averaging does not work that way.

There is no such thing as "absolute" time. Here on earth, we use UTC and its kin. UTC is a consensus time based on coordination among multiple extreme-precision atomic clocks. Rubidium clocks are atomic clocks of lower quality but they still derive their accuracy from atomic transitions, and averaging enough of them will provide an adequate "absolute" time.

The 'absolute' time used for GNSS is by ground reference clock, and the key is that every satellite is referenced to the same clock, giving each satellite an absolute time reference rather than just references to neighbouring satellites. Combined with accurate positioning for each satellite, that is what allows trilateration GNSS to work at all - it is a fundamental assumption that if you receive a timestamp of 1200 + 1us from one satellite and 1200 + 2us from a second satellite, that 1200 will be the same for both satellites. If it is not, then you have no way to tell if the 1us differential time of arrival is actual a differential time of arrival, or if the first satellites clock is running 3ms slow and the second running 1ms fast and the actual DToA is -3us.

If you goal is to measure the distance between two satellites by using the relative difference in time of arrival of their reference clocks, then you cannot also use the relative difference in time of arrival of their clocks to both derive a pseudo-reference-clock and to also measure the distance. That's just throwing unknown error terms together and then ignoring them and hoping they go away.

We used "fixed" ground reference locations that were originally surveyed by non-GPS means, because that's all we had to start with.

We use ground reference locations derived from alternative geodetic measurements because that is what is required to produce a baseline measurement against which to calibrate GNSS.
In order to calibrate a system you either need an external refence standard to calibrate to, or to start from absolute fundamental physical parameters in order to create the standard standard to calibrate to. Traceability of measurement standards is not a case of just taking some clocks and averaging them together, that is fundamentally not how metrology works.

Relevance to Martian Starlink: Starlink depends on GPS-like positioning and will not work without it.

Starlink requires positioning for self-avoidance and for avoidance of other satellites in nearby orbits. That is not going to be an issue encountered around Mars for quite some time, much more gross and irregularly updated orbital parameters will suffice in the near to mid term.

[DanClemmensen]

In order to calibrate a system you either need an external refence standard to calibrate to, or to start from absolute fundamental physical parameters in order to create the standard standard to calibrate to. Traceability of measurement standards is not a case of just taking some clocks and averaging them together, that is fundamentally not how metrology works.

Like other atomic clocks, Rubidium clocks "start from some absolute fundamental physical parameters." These parameters are expressed in the transition frequencies of Rubidium atoms. All Rubidium clocks tic at the same rate, with tiny implementation errors. Both general and special relativity have non-trivial effects when comparing two clocks and must be taken into account.

Clock synchronization depends on another fundamental physical parameter: The speed of light. Distances are measured based on speed-of-light delay, which is measured using using a technique that does not depend on prior knowledge of positions or absolute time. I send you a message at my time M1. You receive it at your time Y2. You respond at your time Y3. I receive it at my time M4. Your message contains your values of Y2 and Y3. I compute the round-trip time as M4-M1 - (Y3-Y2).   This is all pretty much exactly like IEEE 1588. If we have agreed to use "my" clock as the base clock, you will now offset your absolute clock by this computed difference.  All of these measurements and computations have uncertainties. The combined and repeated measurements reduce the total uncertainty.

Yes this is how metrology works, at least for creating the practical UTC time standard.

[sdsds]

Yes, deriving and updating ephemerides for each station is both difficult and necessary. Note in the inertial frame, fixed surface stations are moving too, at one rotation every ~24.6 hours (sol). And orbital stations are moving on non-Keplerian trajectories due to any number of factors (moons, non-uniform Mars mass distribution, solar wind, etc.) Some truth can be derived from accelerometers, star-trackers and other navigation aids to assure the consensus ephemerides remain "grounded."

Lots of equations; lots of unknowns; lots of uncertainties. Protocols like NTP manage decent clock synchronization despite uncertain transmission latency.

(FWIW I'm liberally interpreting the topic of this thread to include both "when" Marslink service is first established and also "what" that service will include and "how" it will be established. Apologies to @whvholst if that wasn't the intent of creating the topic.)

[Solarsail]

Apparently the current mars satellites are able to do radio location determination to 100m accuracy already. (Even with just 1 satellite)  Far better than my guess of 20km...  100 meters needs the satellite and ground terminal to do 2-way communication during line-of-sight events, and then use doppler shift measurements on the signals.  Doing such measurements on discreet satellite passes allows for gradual triangulation of ground position of an object.  I wonder how valuable it is to improve upon that when Mars has 1 town / city on it, 1 satellite in orbit for positioning and data relay...  Can an algorithm like that support gradual improvement as new satellites are added?

When comparing that to 20-60m tall radio masts on nearby hills, we get a trade space of solutions for different kinds of navigational problems.  If you want to locate which rock an astronaut tripped and fell behind, you're better off with live location data to <5m accuracy.  This is only necessary near base.  If you have a mars rover traveling 50m per day, visual navigation is probably far better than radio solutions.  Surveys of terrain probably don't kneed fast resolution of location, but probably aren't helped by location precision as large as 20km...  (Are you on top of Olympus Mons?  Or somewhere dramatically lower?)  If you're flying an airplane thousands of miles between airports, you probably want live location knowledge through your trip...  But you don't need the accuracy to find the runway.  The airport's own beacons can handle local navigation.  What if you're driving a land train (or rail train) thousands of miles through terrain between two cities?  What trade space do you find between speed of update, precision requirements, continuity of location knowledge, and how much of those things can be accomplished with local visual navigation?

Then we've got in-space navigation for vehicles attempting Mars EDL.  They will want a triangulated location in a direction that GPS satellites don't support.

[DanClemmensen]

Yes, deriving and updating ephemerides for each station is both difficult and necessary. Note in the inertial frame, fixed surface stations are moving too, at one rotation every ~24.6 hours (sol). And orbital stations are moving on non-Keplerian trajectories due to any number of factors (moons, non-uniform Mars mass distribution, solar wind, etc.) Some truth can be derived from accelerometers, star-trackers and other navigation aids to assure the consensus ephemerides remain "grounded."

Lots of equations; lots of unknowns; lots of uncertainties. Protocols like NTP manage decent clock synchronization despite uncertain transmission latency.

(FWIW I'm liberally interpreting the topic of this thread to include both "when" Marslink service is first established and also "what" that service will include and "how" it will be established. Apologies to @whvholst if that wasn't the intent of creating the topic.)

NTP is insufficiently precise. That's  why I keep referring to the hardware-assisted version of IEEE 1588, which is also called PTP -- Precision Time Protocol.

[StraumliBlight]

When asked "what markets are particularly exciting for your company?" at today's keynote session. Gwynne Shotwell replied "I'm really looking forward to having communications to bases on the Moon and Starlink around Mars".

[StraumliBlight]

SpaceX has been awarded a ~$300,000 contract by NASA to investigate adapting Starlink for Mars. The study will conclude in August and results released later this year.

[Emmettvonbrown]

And now NASA has public-private partnerships for a) ISS cargo (ex- COTS)  b) ISS crews (CCDEV)  c) ISS replacement (CLD) d) lunar cargo and science (CLPS)  and e) lunar crews (HLS). With f) Mars orbit cargo and science : incoming.

They are building a bridge all the way to Mars.