Starship On-orbit refueling - Options and Discussion

[TheRadicalModerate]

In a LEO the plane will rotate around 360 degs in 25 days. So you get to all possible planes with week(s) of waiting for free.

satrot = radian/sqrt((ere+400km)^3/(G*em))
-3/2 * ere^2/(ere+400km)^2 * 1.08262E-3 * satrot * cos(23deg)  where ere is earth radius equator, 23deg is inclination, 400km is orbital height above surface, em is earth mass

EDIT: inclination is hard to change and so is argument of perihelion. So LEO circular. Lower the better for faster rotation of plane.

EDIT2: I just got -50 days for 23incl and 400km orbit

For a 200x200x28.5º orbit, I get 45.9 days for a 360º nodal precession.  But you can fiddle with this by raising and lowering the apogee/perigee.  This can increase or decrease the precession rate.

I think you were saying this up above, but this doesn't do anything for the inclination.  Inclination is fairly easy to adjust for interplanetary missions, simply by boosting to an HEEO and doing a plane change at apogee.  However, that's an Earth-centric plane change.  Whatever heliocentric plane change you need to reach an interplanetary target has to be done by brute force.  However, you can also do "broken plane" maneuvers well after interplanetary insertion, where you only do part of the plane change at interplanetary insertion, and the rest when the spacecraft is farther out, and therefore moving slower.  The orbital mechanics for this go well beyond what you can do with a spreadsheet.

[Twark_Main]

Whatever heliocentric plane change you need to reach an interplanetary target has to be done by brute force.

Venus-Jupiter flyby has entered the chat.

The orbital mechanics for "broken plane" maneuvers goes well beyond what you can do with a spreadsheet.

Funny how the most effective algorithm is still just "make a good initial guess at the DSM burn point and then 4D hill-climb with a Lambert solver." 

[rsdavis9]

In a LEO the plane will rotate around 360 degs in 25 days. So you get to all possible planes with week(s) of waiting for free.

satrot = radian/sqrt((ere+400km)^3/(G*em))
-3/2 * ere^2/(ere+400km)^2 * 1.08262E-3 * satrot * cos(23deg)  where ere is earth radius equator, 23deg is inclination, 400km is orbital height above surface, em is earth mass

EDIT: inclination is hard to change and so is argument of perihelion. So LEO circular. Lower the better for faster rotation of plane.

EDIT2: I just got -50 days for 23incl and 400km orbit

For a 200x200x28.5º orbit, I get 45.9 days for a 360º nodal precession.  But you can fiddle with this by raising and lowering the apogee/perigee.  This can increase or decrease the precession rate.

I think you were saying this up above, but this doesn't do anything for the inclination.  Inclination is fairly easy to adjust for interplanetary missions, simply by boosting to an HEEO and doing a plane change at apogee.  However, that's an Earth-centric plane change.  Whatever heliocentric plane change you need to reach an interplanetary target has to be done by brute force.  However, you can also do "broken plane" maneuvers well after interplanetary insertion, where you only do part of the plane change at interplanetary insertion, and the rest when the spacecraft is farther out, and therefore moving slower.  The orbital mechanics for this go well beyond what you can do with a spreadsheet.

I don't know if anybody else has seen this paper. Shows the min delta-v to go from just about any plane around earth to interplanetary destinations. Basically it seems to say that the incl should be > 23.5

https://selenianboondocks.com/2018/02/aas-paper-review-practical-methodologies-for-low-delta-v-penalty-on-time-departures-to-arbitrary-interplanetary-destinations-from-a-medium-inclination-low-earth-orbit-depot/

It was posted a while back.

[OTV Booster]

There was some refueling discussion in the Artemus update and discussion thread that got me thinking. The conclusion I reached was that both ships need to be doing micro acceleration during the approach and to get the propellant settled before the hookup. Everything has to happen slowly and gently. And the struts need to be active and help with the last meter. Basically small Canada arms without elbows.


This shows how it's done. The soundtrack is "Try a Little Tenderness." 

[DanClemmensen]

A couple of screenshots from the preso attached below.  Three things:

1) Still looks like the docking / berthing / stabilization mechanism is the four struts, slotted into four receptacles.  Compare that to the shots narianknight posted here, here, and here, which I presume are an early attempt at implementing the receptacles.

2) In the first screenshot below, it looks kinda like the two Ships (both depot and target) are vertically aligned.

3) In the second screenshot, it looks like they're significantly offset from one another.  This could be a perspective problem, or it could be evidence that the depot's presumably male QD is separate from the female QD used with the GSE.  It also could be an indicator that the depot is longer than the target, but then you'd expect the rear flaps to be more-or-less aligned.

Have we seen any more recent renders of docking? I think these are current. They show a non-androgynous system, so one of the two participants must always be the Depot. A non-Depot can only dock to a Depot. However, in these renders the Depot is depicted with EDL hardware (TPS and control surfaces) so the renders are inconsistent, since a Depot should never EDL.

[OTV Booster]

A couple of screenshots from the preso attached below.  Three things:

1) Still looks like the docking / berthing / stabilization mechanism is the four struts, slotted into four receptacles.  Compare that to the shots narianknight posted here, here, and here, which I presume are an early attempt at implementing the receptacles.

2) In the first screenshot below, it looks kinda like the two Ships (both depot and target) are vertically aligned.

3) In the second screenshot, it looks like they're significantly offset from one another.  This could be a perspective problem, or it could be evidence that the depot's presumably male QD is separate from the female QD used with the GSE.  It also could be an indicator that the depot is longer than the target, but then you'd expect the rear flaps to be more-or-less aligned.

Have we seen any more recent renders of docking? I think these are current. They show a non-androgynous system, so one of the two participants must always be the Depot. A non-Depot can only dock to a Depot. However, in these renders the Depot is depicted with EDL hardware (TPS and control surfaces) so the renders are inconsistent, since a Depot should never EDL.

It's been said that SX intends to transfer directly from the tankers. With a maybe kinda chance at a mars shot late next year and Artemus needs, this makes sense from a minimum viable first hack PoV. I've not seen anything from SX explicitly claiming this but that means little. To avoid ambiguity, that assertion is about me, not SpaceX.


In some ways getting loiter figured out seems harder than working out the transfer itself. Each new tanker build gives an opportunity to explore this. At some point the tanker will evolve into a depot in everything but name - with tiles and fins. That's probably when tankers and depots will become different ships.

[TheRadicalModerate]

It's been said that SX intends to transfer directly from the tankers. With a maybe kinda chance at a mars shot late next year and Artemus needs, this makes sense from a minimum viable first hack PoV. I've not seen anything from SX explicitly claiming this but that means little. To avoid ambiguity, that assertion is about me, not SpaceX.

In some ways getting loiter figured out seems harder than working out the transfer itself. Each new tanker build gives an opportunity to explore this. At some point the tanker will evolve into a depot in everything but name - with tiles and fins. That's probably when tankers and depots will become different ships.

It makes lots of sense to do tanker-to-tanker transfers early in the test program, and just to rig one of them with the depot docking and transfer hardware.  It makes zero sense to do this against a valuable target Starship or HLS, where minimizing the number of RPODs reduces risk to the expensive asset.

And I know this is petty and indicative of my not-completely-latent OCD:  "Artemis", not "Artemus".  I know you pushed back against this at one point in the past on questionable linguistic grounds, so here's the Greek for it:  Ἄρτεμις.  See the iota?  See an upsilon anywhere?

(I blame you completely for the half-hour rabbit hole I went down trying to figure out--largely unsuccessfully--the leading diacritical mark(s) on the capital alpha.)

[DanClemmensen]

It makes lots of sense to do tanker-to-tanker transfers early in the test program, and just to rig one of them with the depot docking and transfer hardware.

There is a name for a tanker that is rigged with depot hardware. It's called a "Depot". Just build the depot and use it. If you are in a hurry to demonstrate propellant transfer, You can maybe defer some fancy stuff like the heat shield, and you can use a normal-sized Ship instead of a stretched version, but I see no advantage in starting with a Tanker. In particular, with those funny legs sticking out, you really don't want to bother trying to EDL, so you can skip the TPS and control surfaces.

[Vultur]

It's been said that SX intends to transfer directly from the tankers. With a maybe kinda chance at a mars shot late next year and Artemus needs, this makes sense from a minimum viable first hack PoV. I've not seen anything from SX explicitly claiming this but that means little. To avoid ambiguity, that assertion is about me, not SpaceX.

In some ways getting loiter figured out seems harder than working out the transfer itself. Each new tanker build gives an opportunity to explore this. At some point the tanker will evolve into a depot in everything but name - with tiles and fins. That's probably when tankers and depots will become different ships.

It makes lots of sense to do tanker-to-tanker transfers early in the test program, and just to rig one of them with the depot docking and transfer hardware.  It makes zero sense to do this against a valuable target Starship or HLS, where minimizing the number of RPODs reduces risk to the expensive asset.

Once there's such a thing as a valuable Starship, yes. I'm not sure that's the case until actual Artemis III. Maybe HLS Demo, maybe not, depends on how much cost/work goes into the landing engines.

Everything up to - and maybe including - HLS Demo will be test articles, not really valuable.

(I also personally tend to think the risk of RPODs will drop really fast, to the point that once they have a good solution for docking-capable RCS and the software for it, there's probably no great need for optimizing for a lower number of RPODs. This may well be one of the things that takes a while to work out but once worked out becomes smooth and easy - the solution once established just needs to be applied each time.)

[TheRadicalModerate]

It makes lots of sense to do tanker-to-tanker transfers early in the test program, and just to rig one of them with the depot docking and transfer hardware.

There is a name for a tanker that is rigged with depot hardware. It's called a "Depot". Just build the depot and use it. If you are in a hurry to demonstrate propellant transfer, You can maybe defer some fancy stuff like the heat shield, and you can use a normal-sized Ship instead of a stretched version, but I see no advantage in starting with a Tanker. In particular, with those funny legs sticking out, you really don't want to bother trying to EDL, so you can skip the TPS and control surfaces.

Not necessarily.  A full-up depot has insulation, power for cryocoolers, sunshades, Whipple Shields for sensitive areas, and a whole host of prop management issues worked out.  A tanker-to-tanker test just has the docking hardware and a bare-bones prop transfer.

BTW:  I'm still unclear on whether the hard part is launching with funny legs or with the receptacles for the funny legs.  Seems like you can learn that without anything resembling a depot.  Just slap some gizmos on the dorsal side of a couple of Starships and see what happens.  The TPS and other EDL hardware won't care a bit.  And even if you can't EDL--who cares?  It's a test.  Point Nemo awaits.

[TheRadicalModerate]

Once there's such a thing as a valuable Starship, yes. I'm not sure that's the case until actual Artemis III. Maybe HLS Demo, maybe not, depends on how much cost/work goes into the landing engines.

Everything up to - and maybe including - HLS Demo will be test articles, not really valuable.

(I also personally tend to think the risk of RPODs will drop really fast, to the point that once they have a good solution for docking-capable RCS and the software for it, there's probably no great need for optimizing for a lower number of RPODs. This may well be one of the things that takes a while to work out but once worked out becomes smooth and easy - the solution once established just needs to be applied each time.)

I'll bet that the simple act of testing a purported crew-carrying Ship makes its cost about 10x the cost of a vanilla Starship.  The massive bolus of NASA paperwork will start much earlier than Arty 3.

If you want to send a Starship to Mars for an EDL test, you're going add a lot of value to it, which makes it more expensive.

Beyond all of the risk management, there's also the issue that it's likely much more prop-efficient to use a depot than it is to slowly accumulate prop in your target.  The depot is designed to manage prop in LEO for a long period.  Deep-space Starships have much less stringent requirements, because they can spend 99% of their mission with their nose pointed to the Sun.  That's not true if they're spending weeks or months in LEO.

[thespacecow]

Have we seen any more recent renders of docking? I think these are current. They show a non-androgynous system, so one of the two participants must always be the Depot. A non-Depot can only dock to a Depot. However, in these renders the Depot is depicted with EDL hardware (TPS and control surfaces) so the renders are inconsistent, since a Depot should never EDL.

It's depicting the propellant transfer demo conops, in which both the target and chaser will return to Earth after the demo. You can see it from Amit Kshatriya's presentation last year.

[DanClemmensen]

It makes lots of sense to do tanker-to-tanker transfers early in the test program, and just to rig one of them with the depot docking and transfer hardware.

There is a name for a tanker that is rigged with depot hardware. It's called a "Depot". Just build the depot and use it. If you are in a hurry to demonstrate propellant transfer, You can maybe defer some fancy stuff like the heat shield, and you can use a normal-sized Ship instead of a stretched version, but I see no advantage in starting with a Tanker. In particular, with those funny legs sticking out, you really don't want to bother trying to EDL, so you can skip the TPS and control surfaces.

Not necessarily.  A full-up depot has insulation, power for cryocoolers, sunshades, Whipple Shields for sensitive areas, and a whole host of prop management issues worked out.  A tanker-to-tanker test just has the docking hardware and a bare-bones prop transfer.

BTW:  I'm still unclear on whether the hard part is launching with funny legs or with the receptacles for the funny legs.  Seems like you can learn that without anything resembling a depot.  Just slap some gizmos on the dorsal side of a couple of Starships and see what happens.  The TPS and other EDL hardware won't care a bit.  And even if you can't EDL--who cares?  It's a test.  Point Nemo awaits.

Any ship that needs to connect to Depot needs those four socket thingees, and we have seen pictures of them on the V3 Ships in the Megabay.

As I said earlier, a prototype Depot for the transfer demo can be minimal, with none of the extra stuff you mentioned. You want to call it a modified Tanker. Fine, call it what you like. I want to call it a prototype Depot: a ship capable of transferring propellant.

[Greg Hullender]

And I know this is petty and indicative of my not-completely-latent OCD:  "Artemis", not "Artemus".  I know you pushed back against this at one point in the past on questionable linguistic grounds, so here's the Greek for it:  Ἄρτεμις.  See the iota?  See an upsilon anywhere?

He's thinking of Artemus Gordon, from The Wild, Wild West!

[OTV Booster]

It's been said that SX intends to transfer directly from the tankers. With a maybe kinda chance at a mars shot late next year and Artemus needs, this makes sense from a minimum viable first hack PoV. I've not seen anything from SX explicitly claiming this but that means little. To avoid ambiguity, that assertion is about me, not SpaceX.

In some ways getting loiter figured out seems harder than working out the transfer itself. Each new tanker build gives an opportunity to explore this. At some point the tanker will evolve into a depot in everything but name - with tiles and fins. That's probably when tankers and depots will become different ships.

It makes lots of sense to do tanker-to-tanker transfers early in the test program, and just to rig one of them with the depot docking and transfer hardware.  It makes zero sense to do this against a valuable target Starship or HLS, where minimizing the number of RPODs reduces risk to the expensive asset.

And I know this is petty and indicative of my not-completely-latent OCD:  "Artemis", not "Artemus".  I know you pushed back against this at one point in the past on questionable linguistic grounds, so here's the Greek for it:  Ἄρτεμις.  See the iota?  See an upsilon anywhere?

(I blame you completely for the half-hour rabbit hole I went down trying to figure out--largely unsuccessfully--the leading diacritical mark(s) on the capital alpha.)

I yam sorry for your discomfort. The fault, dear RadMod, is not in the cranium, but in the apple spellchecker itself. If you saw my native spelling you'd think I was talking about Arty Johnson. 

Methinks doing early tests between a completed starlink delivery and a ship with the appropriate plumbing would be easier on scarce launch infrastructure - if the pad can be turned around fast enough. Working up to that, it would make some sense for the first non payload orbital flight to pack on as much propellant a possible and spend extended time in orbit to characterize boiloff and power and RCS consumption.

As long as they're committed to not leaving a hulk on orbit, the interplay between pad turnaround and loiter time dictates when they can give it a shot. A second pad sure would help. Q3 2026?

On reflection, I guess a mis isn't worth the mus. I'll try to do better. 🙄

[OTV Booster]

It makes lots of sense to do tanker-to-tanker transfers early in the test program, and just to rig one of them with the depot docking and transfer hardware.

There is a name for a tanker that is rigged with depot hardware. It's called a "Depot". Just build the depot and use it. If you are in a hurry to demonstrate propellant transfer, You can maybe defer some fancy stuff like the heat shield, and you can use a normal-sized Ship instead of a stretched version, but I see no advantage in starting with a Tanker. In particular, with those funny legs sticking out, you really don't want to bother trying to EDL, so you can skip the TPS and control surfaces.

Not necessarily.  A full-up depot has insulation, power for cryocoolers, sunshades, Whipple Shields for sensitive areas, and a whole host of prop management issues worked out.  A tanker-to-tanker test just has the docking hardware and a bare-bones prop transfer.

BTW:  I'm still unclear on whether the hard part is launching with funny legs or with the receptacles for the funny legs.  Seems like you can learn that without anything resembling a depot.  Just slap some gizmos on the dorsal side of a couple of Starships and see what happens.  The TPS and other EDL hardware won't care a bit.  And even if you can't EDL--who cares?  It's a test.  Point Nemo awaits.

The first test will be RPOD for sure. And maybe several following tests if it doesn't work right out of the gate. BUT, do you think SpaceX will ignore all the other depot kit while they get RPOD to work? That's not the way they do things.


They'll add kit one or more bits at a time until they have a kludge minimally viable depot in all but name. That is when they will strip off the tiles and fins, call it a depot and keep it up there for multiple transfers. If the timing works out and NASA accepts it as a necessary test on the path to Artimis (woo hoo, I did it but spell check ain't happy), tank up a ship and send it on to Mars to test EDL. Maybe put a car on board for old times sake.

[OTV Booster]

Some detail thoughts on propellant transfer involving micro acceleration. The various spin schemes (IMO) may be great in the future but need too much R&D for a time sensitive program.

It takes higher G to settle props than it takes to keep them settled. Settling has to overcome viscosity along the tank walls and other impedimenta. Viscosity aids in keeping settled that propellant which is already settled. Settling props will have at least some small amount of slosh that must be damped while settled propellants, by definition, have all slosh damped out.

It has been experimentally determined that 0.0001g will settle props. To the best of my knowledge (which isn't all that great) it is unknown what it takes to keep it settled. It would be great if it was another order of magnitude but it will be lower than 0.0001g.

A side issue that may impact settling g is that foamy propellants that would be unacceptable for transfer or an engine burn would have little impact on ROPD and might even be beneficial. Not big gobs of bubbles. More like beer foam. I think a layer of foam would help damp small sloshes.

ISTM props should already be settled during final approach so as to avoid disruptive movement, the correction of which could cascade into more disruptive movement.

An idealized RPOD would have both ships in the same plane but at different heights, both under micro acceleration to keep the already settled props in that state. The low ship in a circular orbit, the high ship very slightly elliptical such that its perigee kisses the low ships orbit as the low ship is at that point. As the arms reach out to mate the low ship kicks in a bit of acceleration to match velocity.

At the moment the struts lock into the sockets there will be a slight jerk as the two ships will most likely not have exactly the same velocity vector. If care is taken the magnitude will not be great. If the arms act as dampers and are strong enough, the momentum will transfer with minimum propellant disruption. As long as any resultant slosh is not energetic enough to overpower and break the struts, all is good.

Once the two ships are locked together that ticklish bit is done with. Increase acceleration a bit to resettle props and convince any bubbles that they need to get with the program. Once everything is settled again thrust can be reduced.

If pumps are used for the transfer they should be in counter rotating sets of two and placed to avoid torque, although some clever person might find some way to use a bit of torque to good advantage. Hmmm. A combination turbopump/CMG? Nah.

[TheRadicalModerate]

The first test will be RPOD for sure. And maybe several following tests if it doesn't work right out of the gate. BUT, do you think SpaceX will ignore all the other depot kit while they get RPOD to work? That's not the way they do things.

They'll add kit one or more bits at a time until they have a kludge minimally viable depot in all but name. That is when they will strip off the tiles and fins, call it a depot and keep it up there for multiple transfers. If the timing works out and NASA accepts it as a necessary test on the path to Artimis (woo hoo, I did it but spell check ain't happy), tank up a ship and send it on to Mars to test EDL. Maybe put a car on board for old times sake.

I think that's mostly right, but you want to test things that allow you to recover and reuse as many Ships as possible.  You don't have an MVP depot until you strip off the TPS and apply insulation / coatings / sunshades.  So I'd expect lots of individual components to be tested and retrieved, and try to minimize the number of cycles on full-up depots.

The other thing about the insulation / coatings / sunshades / etc. is that a poorly performing version may be adequate for a real refueling campaign.  That may be another reason to expect longer cycle times between depot versions.

Another related question:  are they gonna rejigger the dome positions to have more prop for tankers and depots sooner, or later?  Tankers don't technically need to be rejiggered--it's just a performance tweak.  But if boiloff rates are a problem, more tankage might be a requirement for the depot to maintain enough prop through the rollout and launch cycle for an HLS or Mars mission.

[TheRadicalModerate]

Some detail thoughts on propellant transfer involving micro acceleration. The various spin schemes (IMO) may be great in the future but need too much R&D for a time sensitive program.

It takes higher G to settle props than it takes to keep them settled. Settling has to overcome viscosity along the tank walls and other impedimenta. Viscosity aids in keeping settled that propellant which is already settled. Settling props will have at least some small amount of slosh that must be damped while settled propellants, by definition, have all slosh damped out.

It has been experimentally determined that 0.0001g will settle props. To the best of my knowledge (which isn't all that great) it is unknown what it takes to keep it settled. It would be great if it was another order of magnitude but it will be lower than 0.0001g.

A side issue that may impact settling g is that foamy propellants that would be unacceptable for transfer or an engine burn would have little impact on ROPD and might even be beneficial. Not big gobs of bubbles. More like beer foam. I think a layer of foam would help damp small sloshes.

Some dynamic viscosities and surface tensions, for comparison (and mention of foam):

  Substance    Dyn Visc (Pa-s)    Surf Tens (mN/m)
Boiling LCH4:  1.1E-04                   14.0
Boiling LOX:    1.9E-04                   13.2
NTO:              4.7E-04                 ~21
MMH:             8.8E-04                 ~54.3
Water:            1.0E-03                   72.0
Beer:              2.5E-03                   ??   (my surface is less tense after consumption)

I don't think you're going to have much problem with foam or even bubbles.

I'm still kinda hoping that propellant management devices can eliminate continuous settling acceleration completely.  Surface tensions of LOX and LCH4 are significantly lower than MMH and NTO, which is what most PMDs have been designed for.

[JIS]

   imagine a half full vehicle being thrusted in one direction, setting the fluid to one side.  In this condition, let's say the Force is 10, the Mass is 1000 (10 struct+990 prop), and then the accel is 0.01.  Now you went to far and need to thrust a little in the opposite direction.  Now the fluid is just floating in the middle.  Your Force is still 10, but your mass is 10, so your accel is 100 times higher at 1.  This is obviously extreme and might be completely physically wrong.  If it's not wrong, then your GNC is really going to struggle with the non-linear behavior.

It's never been tested before, and can't be tested on earth.  No space docking has ever had so much fluid mass, unattached to the walls.

After reading this, my perception of the difficulty of docking any Starship has gone from "it's trivial" to "it's really, really hard".  When some of your mass is fluid that can slosh, the acceleration due to your thrusters will have variable and complex delays, and these delays are far more than minor nuisances. The smaller the needed velocity corrections, the more the delay and I suspect the uncertainty in the delay also  increases as the velocity change decreases. This makes the effect on the final stages of docking disproportionately severe.

This affects all Starship docking in zero g. The more propellant, the bigger the problem, but even actively docking Starship to Gateway is likely to be an issue. Docking a Tanker to a Depot will be "interesting".

Don't think it is a big deal as long you know the total spacecraft mass. If you want to change the velocity by some dV, you just calculate impulse your thrusters need to give your spacecraft and you perform the manoeuvre. At the beginning, when the propellants are sloshing the dV of the spacecraft can be all over the place, but after some time the impulse has as time to propagate through the propellants and the final dV should match your prediction.

So the problem is the transient time. With more propellants (nearly full tanks) the effect of sloshing is bigger but transient time is shorter. With less propellants the effect is smaller but longer.

Doesn't look to me like something unsolvable. Chaotic systems could actually have some "simple" solutions.