Starship On-orbit refueling - Options and Discussion

[cuddihy]

So going to try some physics here, don't crucify me if I dropped a unit:

 if you're using N2 cold gas thrusters for settling (Isp =73 s), and you really need a =0.1 g (=0.981 m/s) of accel to get proper settling & transfer, for a Starship of 140 t (w/ landing propellant aboard), you're looking at a huge amount of thrust (F=ma=137 kN). No way the current RCS can provide that. At an ISP of only 73 sec you're looking at a mass flow rate of F/(Isp*g0) = 137,000N/(73s*9.81m/s2) = 191 kg/s or ~ 10 t over the course of a minute!

Clearly either the transfer needs to be much done faster, at a much lower settling Gs, and with a much higher ISP thruster. Or maybe my math is wildly off.

[Thunderscreech]

.1G sounds like a lot, what's the source of that?  1/10 of that should be sufficient, and then once it's settled against the inlet , a pressure differential should move the liquid with alacricity, right? 

If .01G (.098 m/s^2) is the acceleration and the LOX tank is 20 meters tall, that's just over 20 seconds to settle the tank.

[Twark_Main]

Yes, .1g is a lot. The S-IVB had a mass at stage sep of around 650 tonnes and used 2 x 15.1 kN solid ullage motors (reduced from 3x on early test flights), ie a settling acceleration of ~0.005g.

Now granted the S-IVB isn't in free-fall very long, and the tank is almost full. But if anything I think that calls for lower levels of acceleration. Remember in microgravity you only need to overcome the liquid's surface tension. You don't want your liquid to slosh around after hitting the bottom dome, so ideally the propellant will fall nice-and-slow so that surface tension effects can damp the residual sloshing. If it falls quickly (ie high settling acceleration) then the surface tension effects will be swamped.

Without the influence of surface tension it doesn't really matter what your settling acceleration is. The liquid will behave exactly the same except in "slow motion," and since fuel use is acceleration * time it becomes invariant. If you have half the acceleration and thrust then you need to burn for twice as long, so it always takes the same amount of fuel.

Surface tension breaks this mathematical invariant, since it means that propellant no longer settles in exactly the same way (except for being "time stretched") under different acceleration levels. If surface tension is relatively weak, this effect goes to zero. This is what leads me to hypothesize that the only way to "cheat the invariant" will be to use very low ullage acceleration, such that ullage forces and surface tension forces become within the same order-of-magnitude.

I'm presuming here that these non-linearities (ie those caused by operating in the low-thrust, surface tension-dominated regime) will be beneficial, but perhaps not. If these non-linearities turn out to be detrimental instead than you still only need sufficient acceleration to avoid that regime, and beyond that you again run into the same invariant where it doesn't matter how big your ullage thruster is. At that point you're just balancing thruster mass with the amount of time it takes for on-orbit refueling.

[Pueo]

A quick search shows that 1e-4 g and possibly as low as 1e-5 g is considered enough for propellant settling. Centaur apparently does 8e-5 g for longer coasts to keep the propellant settled to simplify handling and lower boil-off. I am guessing that one of the major points of interest will be what transfer rates require what settling accelerations. As has been pointed out settling of propellants is standard but large volume transfers without an running engine is not.

Why does settled propellant have a lower boil-off rate than non-settled propellant?

[TheRadicalModerate]

A quick search shows that 1e-4 g and possibly as low as 1e-5 g is considered enough for propellant settling. Centaur apparently does 8e-5 g for longer coasts to keep the propellant settled to simplify handling and lower boil-off. I am guessing that one of the major points of interest will be what transfer rates require what settling accelerations. As has been pointed out settling of propellants is standard but large volume transfers without an running engine is not.

Why does settled propellant have a lower boil-off rate than non-settled propellant?

I wondered about this too--which doesn't prevent me from making a guess:  Settled propellant has a smaller surface area.  We have basically three cases:

1) Prop in contact with a cold (non-illuminated) tank wall.
2) Prop in contact with a warm (illuminated) tank wall.
3) Prop floating around in a blob, not in contact with any walls.

You obviously want to avoid #2 as much as possible, but what about #3?  A blob is not getting any conductive heat transfer, but it is getting radiative heat transfer from warm walls.  If the blob is is in the path of that radiation, it'll absorb it, and the average extensive heat of the propellant increases, which is what you want to avoid.

There are limits to how much blob reduction should help you.  For example, any radiation from a warm wall that impinges on the "top" of the settled prop will be absorbed.  But some of it will just go wall-to-wall, which will heat the walls somewhat, but presumably they're pretty much in equilibrium.

Seems like minimizing surface area as much as possible and putting as much of that surface area as possible in contact with cold walls is the optimal strategy.

[TheRadicalModerate]

I haven't seen anybody discussing the mechanics of tail-to-tail docking on this thread.  IMO, the actual prop transfer is trivial compared to the RPOD.  Nobody's even come close to doing precision alignment and hard dock of two 9m-diameter surfaces before.  And then there's the not inconsiderable issue that minor errors can lean to pranging engine bells and fueling lines.

Three geometrical possibilities:

1) A NDS-like soft/hard capture system that's somehow located on the center axis and magically extends through the small gap at the center of the three SL engines on the thrust puck.  I don't think there's room to do this but, even if you could, the mechanical forces on a single-axis support column that had to deal with two Starships, one of them possibly full to the gills with prop, are likely huge.

2) Put a 5m-ish diameter extensible ring around the outside of the 3-SL engine cluster.  Looking at Rafael's CAD drawings, it looks--at least in theory--like there's room to do this, with the struts to extend the ring just outside the anchors for the gimbals, and just inside the bells of the RVacs.  If all this ring did was soft capture, that might work.  It would then be up to some kind of latch system around the outer mould to complete hard dock.  Biggest problem:  You're going to need a lot of play in a 5m-wide soft capture system unless you can substantially reduce the translation and rotation errors that are the current state of the art.

3) Instead of One Soft Capture Ring to Rule Them All, deploy, say, three separate rings, spaced 120º apart near the outer mould line, and use software to coordinate the capture and damping process.  ISTM that this would be more error-tolerant than option #2, but that's more of a hunch than something on which I've done the math.  At the very least, it reduces the soft capture problem to something IDSS-like, and maybe you could even use the existing implementations with a layer of software to get the three rings to cooperate.

One other thing relevant to this discussion:  You have to engage the fueling lines reliably.  I assume that this occurs after hard dock, but maybe there's something to be said for that being part of the hard dock process itself.

After all of that, transferring prop seems like it would be pretty boring.  Even undocking is scarier than the prop transfer.

[Redclaws]

Boring is not the same as trivial.  It’s still a real problem that has to be solved.

[wes_wilson]

I haven't seen anybody discussing the mechanics of tail-to-tail docking on this thread.  IMO, the actual prop transfer is trivial compared to the RPOD.  Nobody's even come close to doing precision alignment and hard dock of two 9m-diameter surfaces before.  And then there's the not inconsiderable issue that minor errors can lean to pranging engine bells and fueling lines.

Three geometrical possibilities:

1) A NDS-like soft/hard capture system that's somehow located on the center axis and magically extends through the small gap at the center of the three SL engines on the thrust puck.  I don't think there's room to do this but, even if you could, the mechanical forces on a single-axis support column that had to deal with two Starships, one of them possibly full to the gills with prop, are likely huge.

2) Put a 5m-ish diameter extensible ring around the outside of the 3-SL engine cluster.  Looking at Rafael's CAD drawings, it looks--at least in theory--like there's room to do this, with the struts to extend the ring just outside the anchors for the gimbals, and just inside the bells of the RVacs.  If all this ring did was soft capture, that might work.  It would then be up to some kind of latch system around the outer mould to complete hard dock.  Biggest problem:  You're going to need a lot of play in a 5m-wide soft capture system unless you can substantially reduce the translation and rotation errors that are the current state of the art.

3) Instead of One Soft Capture Ring to Rule Them All, deploy, say, three separate rings, spaced 120º apart near the outer mould line, and use software to coordinate the capture and damping process.  ISTM that this would be more error-tolerant than option #2, but that's more of a hunch than something on which I've done the math.  At the very least, it reduces the soft capture problem to something IDSS-like, and maybe you could even use the existing implementations with a layer of software to get the three rings to cooperate.

One other thing relevant to this discussion:  You have to engage the fueling lines reliably.  I assume that this occurs after hard dock, but maybe there's something to be said for that being part of the hard dock process itself.

After all of that, transferring prop seems like it would be pretty boring.  Even undocking is scarier than the prop transfer.

I always naively assumed that they'd extend the landing legs on both machines and use gear on the legs to dock "feet to feet". 

[TheRadicalModerate]

I always naively assumed that they'd extend the landing legs on both machines and use gear on the legs to dock "feet to feet".

That's an interesting idea.  How would the feet soak up the residual motions that weren't straight down the x-axis?

[John Santos]

I always naively assumed that they'd extend the landing legs on both machines and use gear on the legs to dock "feet to feet".

That's an interesting idea.  How would the feet soak up the residual motions that weren't straight down the x-axis?

Agreed, but I think the feet would get pretty grotty after even a single landing, and any docking mechanism on the bottom of the feet would get trashed, so they would have to be replaced or refurbished after each flight.

On the other hand, the legs would have to be pretty strong, extendible, able to handle a fair amount of off-axis force and able to quickly damp unwanted motion, so maybe the docking mechanism could be attached to the side of the legs, next to the feet.  During a fuel-transfer docking, the feet would never make contact.  The legs would extend, and then a secondary docking grapple would extend from near the bottom of each leg and mate with the same on the other Starship.

Or a second set of legs specifically for docking mounted next to the landing legs, if there is room between the vacuum Raptor nozzles.  (By mounting the docking mechanism on the same legs as the landing feet, maybe they could save some mass and space at the cost of additional complexity by sharing the motors, hydraulics or whatever is used to extend and retract them.  SpaceX often seems willing to trade mass for reduced complexity, but on the other hand, super-cooled propellant...) 

[yg1968]

On the topic of in-orbit re-fueling, see below:

At 33 minutes of [the video below], Jim Bridenstine made some interesting comments about Starship and more generally about refueling. He said that Starship heavily relies on the ability to transfer cryogenics in LEO for the purpose of taking a system all the way to the Moon. He said that SpaceX's Starship system does not require a fuel depot around the Moon, it requires a depot in orbit around the Earth.

A few minutes later, he said that NASA wants HLS to be sustainable by 2028, and by sustainable he means that it would be reusable and could be refueled. He added that the refueling could be in orbit around the earth or in orbit around the moon (that would be up to the private sector).

[TrevorMonty]

If flight rate is higher enough SpaceX may eventually move to fuel depots, these may not be much more than SS tanker with better insulation or cryocooler.

[rsdavis9]

I always naively assumed that they'd extend the landing legs on both machines and use gear on the legs to dock "feet to feet".

That's an interesting idea.  How would the feet soak up the residual motions that weren't straight down the x-axis?

This seems like a great idea.
Legs extend and have shock absorption. Line up the legs and an additional latch added to the legs would grab the other leg on the other ship and hold them in position.

Have robotic "snakes" to mate the propellant. Tesla already has snakes for charging.

[meekGee]

I always naively assumed that they'd extend the landing legs on both machines and use gear on the legs to dock "feet to feet".

That's an interesting idea.  How would the feet soak up the residual motions that weren't straight down the x-axis?

This seems like a great idea.
Legs extend and have shock absorption. Line up the legs and an additional latch added to the legs would grab the other leg on the other ship and hold them in position.

Have robotic "snakes" to mate the propellant. Tesla already has snakes for charging.

Agreed the legs are a brilliant idea.

Once docked though, no need for snakes.  The spatial relationship is known, so a straight tube will do the job, even if there's a tiny bit of misalignment left.

[libra]

Some quick calculations suggest a 250ton SS ( 100 tons propellant) would need around a 2.5kN ullage thruster for .001g acceleration.  If the thruster had a  ISP of around 280 (i.e hypergolic level ISP) it would only consume around 1kg/s.  If the thruster could maintain that .001g for 3600 seconds ( 1 hour), the mass penalty for the burn is then 3.6tons.  Seems reasonable.

At a minimum, you will be pushing 340 tonnes: two Starships (120t each) docked together, and 100 tonnes of propellant.  3.3 kN needed, using 1.2 kg/s (assuming your I_sp).

But you will also need to be able to push ~1600 tonnes: two Starships, 150 tonnes of cargo, 1200 tonnes of propellant (from a full accumulation tanker to an empty payload ship, or the last transfer from a tanker to the accumulation tanker), plus ~10 tonnes of propellant for the tanker to land.  In this case, 15.7 kN thrust is needed to achieve 9.8 mm/s², using 5.7 kg/s, or roughly 20 tonnes/hour.

Then it depends on how long a propellant transfer takes (I'm guessing they will aim for around 30-90 minutes for a full 1200 tonne transfer), and the specific impulse of the RCS thrusters.  And "milli-g" is propably just order-of-magnitude correct, so our calculations could be off by a factor 3, or even a factor 5, in either direction...

As to leaving LEO, you don't have to burn in a prograde direction.  Do the burn 90 degrees from prograde and so as to not change perigee or apogee so that you only create slight inclination change.

My point with that remark was just that 0.1 g of settling acceleration would not be reasonable to use.  Just 10 minutes of thrust at ~100 cm/s² gives 600 m/s Δv, which would cause a 4.4° inclination change (if I'm calculating that right), or raise the orbit by several hundred kilometers.  It would also use somewhere between 72 and 340 tonnes of propellant (depending on if tanks are almost empty or almost full).

At ~1 cm/s² (0.001 g) on the other hand, a full hour of thrusting is just 36 m/s Δv.  You definitely need to take heed of it in your orbital calculations, but it's not going to give a huge change in either altitude or inclination.

just for the sake of comparison... I recently checked "On the shoulders of Titans" to see how much delta-v did the Gemini-Agena "high rides" needed.

https://history.nasa.gov/SP-4203/ch15-3.htm

In the case of Gemini 11, the  standard orbit was 300 km. They pushed the apogee to nearly 1400 km and then back to 300 km.
To achieve that: 280 m/s "up" and 280 m/s "down", total 560 m/s.

Calculator there. https://www.satsig.net/orbit-research/delta-v-geo-injection-calculator.htm

[tbellman]

I haven't seen anybody discussing the mechanics of tail-to-tail docking on this thread.  IMO, the actual prop transfer is trivial compared to the RPOD.  Nobody's even come close to doing precision alignment and hard dock of two 9m-diameter surfaces before.  And then there's the not inconsiderable issue that minor errors can lean to pranging engine bells and fueling lines.

(tbellman: Quote snipped)

I always naively assumed that they'd extend the landing legs on both machines and use gear on the legs to dock "feet to feet".

I don't remember if it has been in this thread, but there have been discussions about the propellant transfer docking on these forums.

SuperHeavy and Starship will have some way of mating and holding on to each other during ascent.  Most likely, in my opinion, is for that mechanism to be used during tail-to-tail docking of two Starships as well.  The mechanisms on SuperHeavy and Starship would be identical and symmetric, and can then be used for both purposes.  (In theory, you could then also put a SuperHeavy upside down on top of another SuperHeavy and have them latch on to each other, but why one would do that is beyond me...)  As for exactly how that mechanism will work, I don't think we have any good information.  But I don't think the feet will be involved.

And some kind of quick disconnect on the propellant tanking lines.  I suspect those will connect separately, after the docking.  I.e, they would still be separated by a few decimeters, and then lowered towards each other until the QDs connect.  The QDs are probably much more sensitive to mechanical damage than the docking mechanism, so you want to avoid them touching each other during the docking maneuver, as a slight misalignment between the ships could otherwise damage the QDs.

Likewise, I don't think we have anything concrete about the details of how those QDs will work, or how the tanking lines will be lowered towards each other.

[cuddihy]

.1G sounds like a lot, what's the source of that?  1/10 of that should be sufficient, and then once it's settled against the inlet , a pressure differential should move the liquid with alacricity, right? 

If .01G (.098 m/s^2) is the acceleration and the LOX tank is 20 meters tall, that's just over 20 seconds to settle the tank.

I actually got the 0.1G from higher up the thread.

With my numbers above, but 0.01G, that's still 14 kN for 20 seconds, which is 14,000N/(73s*9.81m/s2)*20 sec ~ 390 kg of nitrogen settling gas. That's not peanuts either, but definitely doable, and probably lighter than developing a biprop RCS thruster for the purpose.

But you still have to pressurize the tank at that point, so some kind of heater / engine is probably needed, even for the 10 ton test that NASA is paying for.

[TheRadicalModerate]

I don't remember if it has been in this thread, but there have been discussions about the propellant transfer docking on these forums.

SuperHeavy and Starship will have some way of mating and holding on to each other during ascent.  Most likely, in my opinion, is for that mechanism to be used during tail-to-tail docking of two Starships as well.  The mechanisms on SuperHeavy and Starship would be identical and symmetric, and can then be used for both purposes.  (In theory, you could then also put a SuperHeavy upside down on top of another SuperHeavy and have them latch on to each other, but why one would do that is beyond me...)  As for exactly how that mechanism will work, I don't think we have any good information.  But I don't think the feet will be involved.

And some kind of quick disconnect on the propellant tanking lines.  I suspect those will connect separately, after the docking.  I.e, they would still be separated by a few decimeters, and then lowered towards each other until the QDs connect.  The QDs are probably much more sensitive to mechanical damage than the docking mechanism, so you want to avoid them touching each other during the docking maneuver, as a slight misalignment between the ships could otherwise damage the QDs.

Likewise, I don't think we have anything concrete about the details of how those QDs will work, or how the tanking lines will be lowered towards each other.

I tend to agree that the feet are going to be dicey to use for this.  I also agree that repurposing the QDs for use as the hard-dock mechanism makes sense, especially since the hard-dock doesn't need to hold pressure (except in the fueling lines themselves, of course).

The real issue is soft capture.  That's not only a question of 6 degree-of-freedom shock absorption but also of 6 degree-of-freedom position correction.  We have two general classes of systems that have been developed: non-androgynous probe-and-drogue-like systems (pretty much everything up through Apollo-Soyuz), and androgynous systems with a standoff soft-capture ring (APAS-xx, IDSS).  There is of course no reason why SpaceX can't invent a new way of doing things, but this is a hard problem with a lot of fairly well-developed technology behind it.  I suspect that they'll want to repurpose some of that tech.

Probe-and-drogue systems aren't androgynous, which would be a real problem for a docking between two generic tankers (which is likely to be much more common during prop accumulation than docking between a tanker and a payload Starship).  So we're left with something IDSS-like.

I don't think that Starship needs 1m-wide soft capture systems if it's using 3 of them spaced out along the circumference of the skirt, but it likely will need the standoff capability (perhaps with an even longer standoff to ensure that all 3 pairs are soft-latched before they begin the damping process).

After soft-capture, you have to damp out the residual motions.  We're dealing with much larger vehicles, with even larger moments of inertia, so the loads are going to be quite a bit higher.  On the other hand, if you're got 3 soft capture systems, they can split the load to some extent.

Retraction should be straightforward.  I'd expect to see the use of IDSS-like guide pins to do fine alignment.  Only then will the quick disconnect system become a quick reconnect.  Once all of that is done, then the fueling connectors can extend and engage.

[TheRadicalModerate]

.1G sounds like a lot, what's the source of that?  1/10 of that should be sufficient, and then once it's settled against the inlet , a pressure differential should move the liquid with alacricity, right? 

If .01G (.098 m/s^2) is the acceleration and the LOX tank is 20 meters tall, that's just over 20 seconds to settle the tank.

I actually got the 0.1G from higher up the thread.

With my numbers above, but 0.01G, that's still 14 kN for 20 seconds, which is 14,000N/(73s*9.81m/s2)*20 sec ~ 390 kg of nitrogen settling gas. That's not peanuts either, but definitely doable, and probably lighter than developing a biprop RCS thruster for the purpose.

But you still have to pressurize the tank at that point, so some kind of heater / engine is probably needed, even for the 10 ton test that NASA is paying for.

The 1E-4 to 1E-5 m/s² range, per the Centaur cites above, sounds more reasonable. Settling time isn't only a function of how long it takes for the "highest" blob to "land", but also how much of a splash it makes.  Then you need to wait for the sloshing to abate as well.  Lower accelerations = lower impact velocities = less energy dissipation.  For a 20m tank, the free-flight of the highest blob will take 10.5 minutes, but it'll only impact at 64 mm/s.

For a top-off with a max payload, that's 240t + 100t + 1200t = 1540t of mass, so 1E-4 m/s² needs no more than 154N of thrust, so the cold-gas system has to provide 0.22kg/s

Note that you have to maintain the settling thrust throughout the entire prop transfer.  I have no clue what the fill rate should be.  As a SWAG, I'm going to use 500kg/s, so a worst-case 1200t takes 40 minutes.  So:

1) Settling time is small compared to prop transfer time.
2) You need roughly 700kg of cold gas.

[Nevyn72]

For the capture and docking process you need somewhere to put the mechanism.

You're going to have 6 dirty great cargo pods inside the skirt area.....
Why not use part of that volume for the docking mechanism?

With six equally spaced around the circumference you can have 3 'male' and 3 'female' mechanisms.
When 2 SS's are tail to tail this gives you the 180^o relative position we've seen in all renders for fueling so far.