Starship On-orbit refueling - Options and Discussion

[TheRadicalModerate]

The pipes that are already inside the cryogenic tank? 

But they're not.  None of that plumbing exists.  You'd have to build special sumps into the LOX side of the common dome (and they're not straightforward, since they'd have to be built into the convex side) and the LCH4 dome (which will be surrounded by all kinds of stuff, where it can make use of the empty space around the dome).  Then you'd have to build new downcomers to handle all of this.  It's crazy.

What do you calculate for "prop losses chilling down pipes for (and during) transfer" for the linear ullage approach?

It's the fill/drain system that's already there for the QD.  The LCH4 fill/drain goes straight into the manifold at the bottom of the downcomer, and the LOX fill/drain feeds directly into its sump.  The pipes are about as short as they can get.  The part of the run that can substantially heat up is maybe a meter or so.

Instead, you're talking about a run for the LCH4 that's more than 20m long to get past the CoM, and at least 7m for the LOX tank.  And that assumes that you're willing to feed LCH4 in partway down the downcomer, and just let LOX fall from the CoM to the sump.

[Paul451]

You'd have to build special sumps into the LOX side of the common dome (and they're not straightforward, since they'd have to be built into the convex side) and the LCH4 dome

Not only in the domes, but in the sides of the tanks. There are times (with last few tankers) where the CoM location means that the prop in the tankers are going to be pooling (or rather, spread along) the sides of the tanks. You'd need, in essence, a trough running down the sides of the tank(s), leading to a sump poking out that side. Or a similar level of redesign in the shape of the Starship mouldline, to change the shape of the tanks enough to do the equivalent. (Which, I think, we can reasonably consider to be not "just" comparable with launching 10% or even 20% more tankers.)

AG might be an option for a large, permanent depot that services many ships, but it requires too much for a 1 depot, 1 mission, 1 ship architecture that they are developing for Artemis.


[Aside: I was thinking more about the issue you raised about xferring propellant in two dimensions, causing the rotational axis to shift. It will (obviously), but it won't cause the angle of the axis to shift relative to the system (which I think is what you are picturing), rather it will cause the system itself to nutate/precess. Ie, it will still be primarily rotating in the same plane (the tanker/depot "frisbee") but the angle of that rotational plane will change (taking the "frisbee" with it.) And not by a small amount, given the ratio of moving propellant to static dry-mass. And, going back to plumbing, that nutation will add another centripetal force on the propellant, changing the location of pooling yet again, feeding back into the nutation. It's an interesting problem: stuff you can basically ignore for spinning a regular space-station becomes the dominant design issue for a rotating depot.]

[Twark_Main]

The pipes that are already inside the cryogenic tank? 

But they're not.  None of that plumbing exists.

Obviously I refer to the thermal property of being already inside the same thermal reservoir as the existing cryogenic fluids.

You'd have to build special sumps into the LOX side of the common dome (and they're not straightforward, since they'd have to be built into the convex side)

There's a single point where all the propellant will pool, so the "sump" is already integrated with the existing geometry. You could just put your LOX pickup point at the appropriate location. There's literally zero extra work here.

However, there's an easier way. You apply a small impulse before spin-up to settle the LOX at the bottom of the tank. This only adds a couple minutes of thrusting to the procedure, so it's definitely worth it.

and the LCH4 dome (which will be surrounded by all kinds of stuff, where it can make use of the empty space around the dome).

There's plenty of available space around the top dome. This is not a resource that's in short supply.

The only change in the geometry would be a "bump" in the appropriate place, so ensure a single point of pooling.This is not a substantial impingement on 3D packaging or manufacturing.

Then you'd have to build new downcomers to handle all of this.

The LOX tank only needs a small-diameter pipe (depending on the transfer rate) that's routed to the opposite corner, ~9 meters away.

For the CH4 tank, I think you have to trade between at least two options: having a comparably small-diameter pipe connecting the upper sump to the downcomer, versus running the pipe ~9 meters across and adding an additional hardpoint & connector on the top of the CH4 tank. In theory you would only need to use this for the mass-optimized tankers (not LSS etc), but that just means the depot need two sets of piping. I'm leaning toward the former option myself, but exploring the latter option is mighty tempting!

What do you calculate for "prop losses chilling down pipes for (and during) transfer" for the linear ullage approach?

It's the fill/drain system that's already there for the QD.

  The LCH4 fill/drain goes straight into the manifold at the bottom of the downcomer, and the LOX fill/drain feeds directly into its sump.  The pipes are about as short as they can get.  The part of the run that can substantially heat up is maybe a meter or so.[/quote]

Which part is that, specifically?

I know the downcomer is in the center, but the connector is on the perimeter. By my math that's roughly 4.5 meters.

Instead, you're talking about a run for the LCH4 that's more than 20m long to get past the CoM, and at least 7m for the LOX tank.  And that assumes that you're willing to feed LCH4 in partway down the downcomer, and just let LOX fall from the CoM to the sump.

See above.

[Twark_Main]

You'd have to build special sumps into the LOX side of the common dome (and they're not straightforward, since they'd have to be built into the convex side) and the LCH4 dome

Not only in the domes, but in the sides of the tanks. There are times (with last few tankers) where the CoM location means that the prop in the tankers are going to be pooling (or rather, spread along) the sides of the tanks. You'd need, in essence, a trough running down the sides of the tank(s), leading to a sump poking out that side. [snip]

This is not how "liquids draining from tanks" works.

I was thinking more about the issue you raised about xferring propellant in two dimensions, causing the rotational axis to shift. It will (obviously), but it won't cause the angle of the axis to shift relative to the system (which I think is what you are picturing), rather it will cause the system itself to nutate/precess. Ie, it will still be primarily rotating in the same plane (the tanker/depot "frisbee") but the angle of that rotational plane will change (taking the "frisbee" with it.) And not by a small amount, given the ratio of moving propellant to static dry-mass. And, going back to plumbing, that nutation will add another centripetal force on the propellant, changing the location of pooling yet again, feeding back into the nutation. It's an interesting problem: stuff you can basically ignore for spinning a regular space-station becomes the dominant design issue for a rotating depot.

Everything is plane symmetric, so there's no nutation. The rotation axis is always perpendicular to a plane connecting the central axis of both ships. The only thing that happens is the CoM shifts, so the axis moves the location where it intersects the plane.

[meekGee]

Please examine the following idea for a different way to get cheap, stable spin settling with ventral/ventral or bidorsal docking.

It involves “wasting” one Ship that is tethered to nose of Depot by maybe a mile of cable. The array is slowly spun. I imagine docking Tanker or Ship-taking-fuel with Depot will not be more complex than other RPOD schemes.

The two advantages I see here are
1) sustained settling acceleration for cheap
2) acceleration direction aligns with launch configuration (on both active systems) for which fuel-refuel plumbing is optimized.
3 of 2) Higher settling accelerations are available, expediting transfer ops.

One way to potentially make this work that would even be stable is if your sidereal spin rate is the same as your sidereal period of revolution. That is, from the perspective of the Earth, the depot is always down and the counterweight is always up. (Or vice versa.) The cable might need to be a hundred kilometers long or so (I keep meaning to estimate this, and I keep forgetting to), but that's not a big deal. Your settling acceleration comes from tidal forces, so you don't need to worry about the usual problems of getting a stable rotating system.

However, I couldn't figure out how to cope with the problems that occur when you add/remove propellant to/from the depot. Unless the counterweight is very heavy, the center of mass is going to move a lot and the thing is going to swing. The oscillations will damp out over some period of time (but I haven't worked that out either). Still, the sideways thrust to stabilize are probably a lot less than what's required for an hours-long ullage burn.

If we assume 0.1 mm/s² and rotating once every 90 minutes, that's a radius of ~3 km.

https://futureboy.us/fsp/frink.fsp?fromVal=%280.1+mm%2Fs%2Fs%29+%2F+%281%2F%2890+min%29%29%5E2&toVal=#calc

Right but 90 minutes qas just an example he gave.  There's a sweet spot, balancing cable length and difficulty of docking.

If you replace the ballast ship with a second depot, the cable cost is halved.

That said, I still think maybe you don't need to accelerate the whole time, but only during settling.

[OTV Booster]

Two things.


Can somebody tell me what AG means? It was probably explained somewhere but I missed it.


And, can somebody point me to a space tether deployment that worked? They did something on the shuttle but IIRC they fouled the reel.

[Paul451]

Can somebody tell me what AG means?

Artificial gravity, meaning spinning for centripetal gravity.

And, can somebody point me to a space tether deployment that worked? They did something on the shuttle but IIRC they fouled the reel.

Short tethers are used routinely by satellites to "throw away" rotational inertia left over from spin-stabilised launch. Eg, the Dawn probe use four weights on 12m tethers to de-spin.

There have been a few examples of missions with multi-km tethers, many have failed (or only partially succeeded.) Some succeeded. For eg, NASA's Small Expendable Deployer System (SEDS) deployed 20km tethers. (Followed by a mission with a more complex 500m tether that tested electrical interaction in orbit, including generating power from orbit, and raising orbit using power.)

Many (I'd say the majority of) tether experiments have been extremely low cost, student-driven, and/or cubeasats. Often the failure mode was the spacecraft itself, not the tether.

But even when you crack the deployment, tethers aren't good for a spin-gravity system, because the short and intermediate axes are too similar, the system isn't going to be stable. (On top of that, the tether can twist, adding another entire mode of instability.) Tethers are however, brilliant for gravity-stabilised systems (using tidal effects to stay "vertical" in their orbit.) That can create a slight gravity gradient within the end-masses. Tiny, but people say you only need a tiny amount.

[TheRadicalModerate]

You'd have to build special sumps into the LOX side of the common dome (and they're not straightforward, since they'd have to be built into the convex side)

There's a single point where all the propellant will pool, so the "sump" is already integrated with the existing geometry. You could just put your LOX pickup point at the appropriate location. There's literally zero extra work here.

The entire weld where the common dome is joined to the ring segment has the same, and largest, centrifugal force being applied to it.  That's not a single point.

However, there's an easier way. You apply a small impulse before spin-up to settle the LOX at the bottom of the tank. This only adds a couple minutes of thrusting to the procedure, so it's definitely worth it.

Nope.  Even when all of the LCH4 is in the bottom of the tanks, it's still "above" (farther forward) the CoM.  When you rotate the system, it'll all fly up to the upper dome.  The same thing will happen to some of the LOX in the LOX tank if its fairly full.

and the LCH4 dome (which will be surrounded by all kinds of stuff, where it can make use of the empty space around the dome).

There's plenty of available space around the top dome. This is not a resource that's in short supply.

Probably true for a tanker or a depot, but not true for a cargo Starship or LSS.  Those want to maximize payload height, so if all this new plumbing forces a big redesign to keep the manufacturing common, it'll cause problems.

The LOX tank only needs a small-diameter pipe (depending on the transfer rate) that's routed to the opposite corner, ~9 meters away.

Is the axis of rotation the x-axis, or the y- or z-axes?  If it's the former, then things are going to wobble as the prop moves from one Starship to the other.  If it's the latter, then you need to pull all of the prop that's above the CoM off of the "roof" of the tank, and deliver it somewhere down near the "floor".  That's a lot farther than 9m.  It also has really strange hydrostatics.

What do you calculate for "prop losses chilling down pipes for (and during) transfer" for the linear ullage approach?

It's the fill/drain system that's already there for the QD.

  The LCH4 fill/drain goes straight into the manifold at the bottom of the downcomer, and the LOX fill/drain feeds directly into its sump.  The pipes are about as short as they can get.  The part of the run that can substantially heat up is maybe a meter or so.

Which part is that, specifically?

I know the downcomer is in the center, but the connector is on the perimeter. By my math that's roughly 4.5 meters.

The QD connectors themselves can get to a very high temperature.  If the pipes behind the connectors are in ullage space, then the heat will conduct along the pipe for some distance, before the cold gas puts everything in equilibrium.  1m is a guess on how far that is.  If the pipes are immersed in liquid prop, then there's no chill-down required at all.

My objection to the "pull stuff off the roof" scheme is that you either have to run piping along the outside, in which case the whole run as at a hot equilibrium, or do really complex stuff to keep it inside the tanks.  If you use only the QD plumbing, most stuff stays cold.

[Oersted]

145 pages and counting and the main conclusion seems to be that some people derive a lot of pleasure from bickering.

SpaceX is known for early and extensive use of super-computers. Elon just built a huge one for Grok, I think it was.

I think SpaceX will probably have tried to feed all the conditions and hardware possibilities and mass and force budgets into their super-computers to see which solutions they spit out when it comes to on-orbit refuelling.

Maybe we will see some quite un-intuitive, highly optimized solutions coming out.

[eriblo]

As usual the problem is that we don't know enough about the requirements so there are different ideas about what they are (and posters could always be better at including their assumptions). This combined with SpaceX always optimizing for [performance/cost/time/coolness...] makes it obvious that [preferred solution] is the only way...

If you think a transfer needs a sizeable fraction of 1 g₀ for several hours then of course you need to spin things. But if you only need enough settling to get the propellant into contact with a cleverly designed sump you are far better of leaving the acrobatics to the professionals.

[eriblo]

To first order the cross over in propellant requirements between linear and circular acceleration would occur at a settling duration of t=2�sqrt(r/a) where r is the rotational radius at which settling acceleration a is desired.

In real life there will be some factor due to the mass distribution and thruster placement relative to r but it will likely be well within an order of magnitude.

For most of the numbers proposed here for spinning Starships this would mean that rotation would seem to win for durations more than a few minutes (unless using tethers with low accelerations), i.e. 10 m and 1 mm/s� would be ~ 3 min.

[eriblo]

I personally am squarely in the linear settling acceleration camp. I don't recall any signs that SpaceX is seriously considering rotational acceleration and I don't think it makes much sense. Sure, at first look it wastes propellant. However, it's only wasted propellant

1) if it requires more than the unavoidable losses incurred during the transfer.

2) to the extent that the thrust is not also used to maintain the orbit.

Rotational settling would not only increase the technical and operational complexity but also involve an increase in dry mass and likely residuals. It's somewhat less elegant if you after having spun up, transfered and spun down need to do a linear settling and transfer to get the last bit of propellant...


Regarding 2): Both the linear and rotational thrust can be used constructively, although it would seem to be easier in the former case. Yes, cold gas thrusters involve an I_sp penalty but if it's that critical I know what I would go for given the choice between developing rotational settling/transfer and small hot gas settling thrusters.


Regarding 1): Given the time in microgravity it is likely that most of the propellant will reach thermal equilibrium between the gas and liquid phases. That means that any lowering of the pressure will involve boiling and utilization of the heat stored in the liquid.

For example, lowering the pressure from 1 to 0.5 bar will lower the boiling point of oxygen by 6 K and vaporize 5% of the liquid. This would be ~60 t of GOX for a full V2 load and likely more than enough to both maintain ullage volume and provide settling thrust.

I would not be surprised if the challenge will be to maintain the pressure difference required for transfer while only incurring the amount of boil off needed for settling. Also consider that you if you don't have active cooling you can "store" depot boil off between transfers by allowing the temperature to increase (at least up to the expected tanker temperatures).

[Narnianknight]

I personally am squarely in the linear settling acceleration camp. I don't recall any signs that SpaceX is seriously considering rotational acceleration and I don't think it makes much sense.

I can't quite figure out why people are still saying things like this or like "Team 'pump' all the way," after Amit Kshatriya has already announced that SpaceX are in fact going to use settling thrusters and pressure differential transfer. These things are no longer unknowns.

[KilroySmith]

A quick link:
https://arstechnica.com/space/2024/04/nasa-exploration-chief-lays-out-next-steps-for-starship-development/

[TheRadicalModerate]

I personally am squarely in the linear settling acceleration camp. I don't recall any signs that SpaceX is seriously considering rotational acceleration and I don't think it makes much sense.

I can't quite figure out why people are still saying things like this or like "Team 'pump' all the way," after Amit Kshatriya has already announced that SpaceX are in fact going to use settling thrusters and pressure differential transfer. These things are no longer unknowns.

Three things:

1) It's fun to speculate, especially in an especially wide and poorly defined trade space.

2) I'll bet that both NASA and SpaceX are also aware that the trade space is especially wide and undefined.  I would take pronouncements from anybody, including Elon, with a grain of salt.

3) Thrust-settled, pressure-fed transfer is almost certainly the simplest thing they can try.  That makes it likely that it's the first thing that they try.  But I doubt it'll be the last thing.

[OTV Booster]

Can somebody tell me what AG means?

Artificial gravity, meaning spinning for centripetal gravity.

And, can somebody point me to a space tether deployment that worked? They did something on the shuttle but IIRC they fouled the reel.

Short tethers are used routinely by satellites to "throw away" rotational inertia left over from spin-stabilised launch. Eg, the Dawn probe use four weights on 12m tethers to de-spin.

There have been a few examples of missions with multi-km tethers, many have failed (or only partially succeeded.) Some succeeded. For eg, NASA's Small Expendable Deployer System (SEDS) deployed 20km tethers. (Followed by a mission with a more complex 500m tether that tested electrical interaction in orbit, including generating power from orbit, and raising orbit using power.)

Many (I'd say the majority of) tether experiments have been extremely low cost, student-driven, and/or cubeasats. Often the failure mode was the spacecraft itself, not the tether.

But even when you crack the deployment, tethers aren't good for a spin-gravity system, because the short and intermediate axes are too similar, the system isn't going to be stable. (On top of that, the tether can twist, adding another entire mode of instability.) Tethers are however, brilliant for gravity-stabilised systems (using tidal effects to stay "vertical" in their orbit.) That can create a slight gravity gradient within the end-masses. Tiny, but people say you only need a tiny amount.

Thank you. Exactly what I was looking for.


The shuttle experiments caught my eye. Since then I've been passively looking for tether attempts but apparently was looking in the wrong places.


So, a flexible dumbell is stable?  A rigid dumbell rocks and rolls with passing mascons AIUI.

[Paul451]

[tethers]
So, a flexible dumbell is stable?

No. With a flexible tether, the end-weights can twist back-and-forth on the end of the tether. You need something other than a cable to prevent twist. And the longer the tether, the worse it is and the harder to counter.

Additionally, any AG tether system is likely going to have very similar rotational-inertia in both the short axes. That risks intermediate-axis-instability.

Essentially, tethered systems are likely to be inherently unstable around the long axis.

[DanClemmensen]

[tethers]
So, a flexible dumbell is stable?

No. With a flexible tether, the end-weights can twist back-and-forth on the end of the tether. You need something other than a cable to prevent twist. And the longer the tether, the worse it is and the harder to counter.

Additionally, any AG tether system is likely going to have very similar rotational-inertia in both the short axes. That risks intermediate-axis-instability.

Essentially, tethered systems are likely to be inherently unstable around the long axis.

A tether can be converted into a truss by using pressurized hoses instead of wires. It does not need to be very strong in compression to eliminate twisting.

Is there a physically-possible way to conserve the angular momentum so this system can be spun up and down without much use of thrusters? That would be a really serious reaction wheel, inside the Depot or counterweight. Maybe not feasible.

[Narnianknight]

https://twitter.com/INiallAnderson/status/1872613197347512828

Low Earth Orbit. SpaceX will conduct a range of Starship operations in low-Earth orbit (�LEO�). Each fully reusable Starship spacecraft is capable of carrying up to 150 metric tons to Earth orbit, and this authorization will enable SpaceX to reliably launch and deploy satellites to support broadband, mobile connectivity, earth observation, science, and other use cases that will benefit humanity. Missions beyond LEO will also require a tanker version of Starship for propellant aggregation. During these missions, SpaceX will launch one or more propellant tanker versions of Starship. Some of these tanker variants will remain in LEO as �depots,� and will be filled with propellant by subsequent tanker launches. LEO operations will occur in a circular orbitat 281 km altitude (+/- 100 km) and an inclination ranging from equatorial (0 degrees) to polar.

Medium-Earth Orbit/High-Earth Orbit/Final Tanking Orbit. Missions beyond LEO will also require space station operations in medium-Earth orbit (�MEO�) to high-Earth orbit (�HEO�). For example, crewed lunar missions will include a secondary propellant transfer in MEO/HEO, the Final Tanking Orbit (�FTO�). Operations in MEO/HEO will occur in an elliptical orbit of 281 km x 34,534 km and an altitude tolerance of +116,000/-24,000 km apogee and +/- 100 km perigee, with inclination between 28 and 33 degrees (+/- 2 degrees).

[InterestedEngineer]

Medium-Earth Orbit/High-Earth Orbit/Final Tanking Orbit. Missions beyond LEO will also require space station operations in medium-Earth orbit (�MEO�) to high-Earth orbit (�HEO�). For example, crewed lunar missions will include a secondary propellant transfer in MEO/HEO, the Final Tanking Orbit (�FTO�). Operations in MEO/HEO will occur in an elliptical orbit of 281 km x 34,534 km and an altitude tolerance of +116,000/-24,000 km apogee and +/- 100 km perigee, with inclination between 28 and 33 degrees (+/- 2 degrees).

It looks like we can confirm HEO tanking for Luna.  Interesting!  I didn't think it would be needed.

Now I see why they think they might need 20 launches.  10 for two full starships that go up to FTO, combine fuel, and one boosts for Luna at perigee and one EDLs.

That's an awfully high FTO, generally I've found a half tank for both craft gets you to about 15,000km.  Guess I'd better go check the numbers.  That is basically GTO btw (minus about 1000km)