Starship On-orbit refueling - Options and Discussion

[edzieba]

Or maybe just 'spraying', because aerosolisation is the specific problem to avoid. A nice laminar fountain isn't really a problem, its only once that flow breaks up into small high surface area droplets that it becomes an issue (for a tank with an open inlet and outlet that assumes stratified phases).

[OTV Booster]

If you build up to much pressure differential you let the propellants slosh a bit until it decreases enough and then you continue. If you are in a hurry you might trade a bit of venting against extra RCS propellant.

That last is the detail where the devil resides. How much venting? How much settling propellant? How long to transfer?

No ullage crossover and using thermal/pressure to power the transfer is mechanically simplest and avoids risk of recirculating splatter, but is it a path to acceptable propellant consumption? I really have not a clue.

I think I know why the Poiseuille results are so non-intuitive:  they don't include dynamic pressure.  They're a solution for a flow in equilibrium, but the non-equilibrium case requires the volumetric flow to be established.  So the dynamic pressure of the flow can still cause geysering.

So, two choices for limiting geysering:

1) Limit geysering by starting the pumping operation at low power/high acceleration, then gradually increase power and decrease acceleration as the pool fills.  The pool will then provide viscous damping as the flow rate inceases.

2) Assuming that only some small amount of prop will blob up and hit the equalization inlet, let it.  Soon enough, the pool will be deep enough that viscous damping will prevent geysering.  You wind up wasting a bit of energy re-pumping prop that recirculated, but it's probably a trivial amount.

A final note on how to limit recirculation:  For ullage acceleration a and geyser velocity v_g,  a blob will fall back into the pool before hitting the equalization inlet if the height h between the top of the pool and the inlet is greater than v_g²/(2a).  This is really just a limiting condition on how much geysering you can tolerate.

If splatter ingestion is a problem something like the float check valve in a some snorkels might be he answer. Misting would get by but a glob of props would shut it for a moment.

[TheRadicalModerate]

If splatter ingestion is a problem something like the float check valve in a some snorkels might be he answer. Misting would get by but a glob of props would shut it for a moment.

Kinda hard to use a float check valve in a system where there's essentially no buoyancy.

I really don't think this is an intractable problem.  The way we fell into this rabbit hole was as a result of an analysis of what it took to use cold gas thrusters for ullage acceleration.  That's a "Doctor, Doctor, it hurts when I do this" kind of problem: don't do that. 

With adequate acceleration, you can use high transfer rates while mitigating fountaining, splash, splatter, whatever you want to call it.  High transfer rates and decent thruster Isp result in extremely modest prop losses for ullage thrust.

[mikelepage]

If splatter ingestion is a problem something like the float check valve in a some snorkels might be he answer. Misting would get by but a glob of props would shut it for a moment.

Kinda hard to use a float check valve in a system where there's essentially no buoyancy.

I really don't think this is an intractable problem.  The way we fell into this rabbit hole was as a result of an analysis of what it took to use cold gas thrusters for ullage acceleration.  That's a "Doctor, Doctor, it hurts when I do this" kind of problem: don't do that. 

With adequate acceleration, you can use high transfer rates while mitigating fountaining, splash, splatter, whatever you want to call it.  High transfer rates and decent thruster Isp result in extremely modest prop losses for ullage thrust.

To be fair, that wasn’t the only reason we fell into this rabbit hole, although I realise I was concerned about a different phenomena than the fountaining/sloshing effect we’re now talking about. That will indeed go away as the depot tanks fill up. I’m talking about the “induced swirling” effect Dan and I were thought experimenting on (that could get worse as the depot fills up).

I had to go back and find this old Dyson air multiplier video, because this shows the “entrained air” concept that I was referencing.



Depending how fast you have to pump the propellant (i.e. proportional to how much prop you expend in milli-G acceleration), you might get to a situation where you have to slow down as the depot tank gets full to avoid creating induced vortices in the depot, which could have large angular momentum and torque the whole system around.

Maybe I’m over egging this, but it seems to me that unless you’re going to fill the whole prop tank with baffles, there will be some threshold flow rate over which it’s not safe to pump.

[OTV Booster]

If splatter ingestion is a problem something like the float check valve in a some snorkels might be he answer. Misting would get by but a glob of props would shut it for a moment.

Kinda hard to use a float check valve in a system where there's essentially no buoyancy.

I really don't think this is an intractable problem.  The way we fell into this rabbit hole was as a result of an analysis of what it took to use cold gas thrusters for ullage acceleration.  That's a "Doctor, Doctor, it hurts when I do this" kind of problem: don't do that. 

With adequate acceleration, you can use high transfer rates while mitigating fountaining, splash, splatter, whatever you want to call it.  High transfer rates and decent thruster Isp result in extremely modest prop losses for ullage thrust.

If we take 6 hours as a transfer target, and I haven't boogered the numbers, the 120t of LOX needs to move at a rate of 4.9l/sec and 30t of liquid methane at 3.3l/sec.


Do we have a pixel count of the QD inner diameters? I can't think of a reason to not expect the diameter to carry through the system. If the diameter is 10cm (WAG) we need an average flow velocity of 156cm/sec for the LOX and 105cm/sec for the methane.


That's about my limit on numbers and they really need to be verified. The target transfer time is a bit arbitrary and the plumbing diameter needs verification. Between this and the viscosity numbers, is there enough to rough in the g force needed to avoid geysering erupting starting with an empty depot and ending with completion of a 150t transfer, then at completion of each successive transfer?


From this we can probably do a sanity check on propellant needs and/or ullage pressures and volumes for the various settling schemes.

[OTV Booster]

If splatter ingestion is a problem something like the float check valve in a some snorkels might be he answer. Misting would get by but a glob of props would shut it for a moment.

Kinda hard to use a float check valve in a system where there's essentially no buoyancy.

I really don't think this is an intractable problem.  The way we fell into this rabbit hole was as a result of an analysis of what it took to use cold gas thrusters for ullage acceleration.  That's a "Doctor, Doctor, it hurts when I do this" kind of problem: don't do that. 

With adequate acceleration, you can use high transfer rates while mitigating fountaining, splash, splatter, whatever you want to call it.  High transfer rates and decent thruster Isp result in extremely modest prop losses for ullage thrust.

To be fair, that wasn’t the only reason we fell into this rabbit hole, although I realise I was concerned about a different phenomena than the fountaining/sloshing effect we’re now talking about. That will indeed go away as the depot tanks fill up. I’m talking about the “induced swirling” effect Dan and I were thought experimenting on (that could get worse as the depot fills up).

I had to go back and find this old Dyson air multiplier video, because this shows the “entrained air” concept that I was referencing.



Depending how fast you have to pump the propellant (i.e. proportional to how much prop you expend in milli-G acceleration), you might get to a situation where you have to slow down as the depot tank gets full to avoid creating induced vortices in the depot, which could have large angular momentum and torque the whole system around.

Maybe I’m over egging this, but it seems to me that unless you’re going to fill the whole prop tank with baffles, there will be some threshold flow rate over which it’s not safe to pump.

FWIW, because of the methane downcomer transfer tube the propellant inlets will be off center. This doesn't negate vortex formation but I think it would keep it from being a big concentric tank enveloping swirl. Add a 'T' fitting and there would either be two counter swirls or chaotic movement - I think. If that's not enough, a flapper valve in the outlet T could periodically reverse the output direction and nip a big vortex in the bud.


Or, as you say, some max flow rate that the above measures might raise.

[TheRadicalModerate]

To be fair, that wasn’t the only reason we fell into this rabbit hole, although I realise I was concerned about a different phenomena than the fountaining/sloshing effect we’re now talking about. That will indeed go away as the depot tanks fill up. I’m talking about the “induced swirling” effect Dan and I were thought experimenting on (that could get worse as the depot fills up)...

Maybe I’m over egging this, but it seems to me that unless you’re going to fill the whole prop tank with baffles, there will be some threshold flow rate over which it’s not safe to pump.

I don't think you have to fill the whole tank with baffles; you just need enough of them to generate turbulence before any kind of organized laminar flow builds up.  It's not often that tip vortices are your friend, but this is one time they are.

You could probably cage the outlet (inlet?  the place where stuff comes into the receiving tank) with something to break up the flow even more.  Note that I'm implicitly assuming that fill/drain is not the same as the downcomers.

[TheRadicalModerate]

If we take 6 hours as a transfer target, and I haven't boogered the numbers, the 120t of LOX needs to move at a rate of 4.9l/sec and 30t of liquid methane at 3.3l/sec.

That's the kind of time scale where prop losses from ullage thrust are gonna hurt you pretty badly.

I was initially thinking that half an hour would be about right for 150t of prop, but even that might be too leisurely.  Remember that the worst case is actually when the depot transfers all of its prop to a payload Starship.  And you can wind up with orbital constraints on how long you can take, especially if you're doing high-cadence tanker ops where you want to de-orbit for quick turnaround, or if you have a depot that's transferring its whole load in an HEEO.

If fountaining is a non-issue with any appreciable pool depth in the receiving tank, then you should start off slow but ramp up to >100kg/s as soon as possible.  Remember, in microgravity, your pump power can be tiny, even for high flow rates.

[mikelepage]

If we take 6 hours as a transfer target, and I haven't boogered the numbers, the 120t of LOX needs to move at a rate of 4.9l/sec and 30t of liquid methane at 3.3l/sec.

That's the kind of time scale where prop losses from ullage thrust are gonna hurt you pretty badly.

I was initially thinking that half an hour would be about right for 150t of prop, but even that might be too leisurely.  Remember that the worst case is actually when the depot transfers all of its prop to a payload Starship.  And you can wind up with orbital constraints on how long you can take, especially if you're doing high-cadence tanker ops where you want to de-orbit for quick turnaround, or if you have a depot that's transferring its whole load in an HEEO.

If fountaining is a non-issue with any appreciable pool depth in the receiving tank, then you should start off slow but ramp up to >100kg/s as soon as possible.  Remember, in microgravity, your pump power can be tiny, even for high flow rates.

Looks like yesterday's WDR took about 52 minutes to load Starship with 1200 ton of prop (385kg/s total - so maybe up to 200kg/s per tank). Not sure if that would be representative of an in-space fill sequence, but hard to imagine it going any faster than that.

FWIW, because of the methane downcomer transfer tube the propellant inlets will be off center. This doesn't negate vortex formation but I think it would keep it from being a big concentric tank enveloping swirl. Add a 'T' fitting and there would either be two counter swirls or chaotic movement - I think. If that's not enough, a flapper valve in the outlet T could periodically reverse the output direction and nip a big vortex in the bud.


Or, as you say, some max flow rate that the above measures might raise.

I don't think you have to fill the whole tank with baffles; you just need enough of them to generate turbulence before any kind of organized laminar flow builds up.  It's not often that tip vortices are your friend, but this is one time they are.

You could probably cage the outlet (inlet?  the place where stuff comes into the receiving tank) with something to break up the flow even more.  Note that I'm implicitly assuming that fill/drain is not the same as the downcomers.

Yeah, just been looking at this video of fire suppression sprinklers, where the flow rate is pretty high (google says 50-100 litres per second), and they create a pretty evenly distributed stream of fluid. Maybe even just having a small array of these inside the inlet to the receiver tank would be enough to make the flow sufficiently chaotic.



Still would be nice to see some CFD of liquid oxygen entering a propellant tank.

[Paul451]

Remember, in microgravity, your pump power can be tiny, even for high flow rates.

(Not really. Friction/inertia will instead dominate and create an irreducible minimum.)

But the real limit will be cavitation. That same micro-g that reduces the back-pressure from the "head" also reduces the "tail" pressure, the rate that the fluid can reach the intake. (Normally, gravity in a tank is, in effect, acting as a first-stage "pump" pushing the fluid into the actual pump.)

[TheRadicalModerate]

Looks like yesterday's WDR took about 52 minutes to load Starship with 1200 ton of prop (385kg/s total - so maybe up to 200kg/s per tank). Not sure if that would be representative of an in-space fill sequence, but hard to imagine it going any faster than that.

That brings up another interesting question:  Transfer LCH4 and LOX simultaneously, or sequentially?  Constraints on pumping simultaneously would be:

1) Pump power/energy.  You're going to be in eclipse for part of the transfer sequence, so you're reliant on batteries for 100% of the energy part of the time, and the solar power has to recharge the batteries as well as run the pumps while you're in sunlight.  Running one pump will consume less power than two.

2) Contingency complexity.  If you had a pump fail or de-prime itself while the other was running, the dynamics will be weird.  I'd be tempted not to think about that weirdness, at least in the early days.

[TheRadicalModerate]

Remember, in microgravity, your pump power can be tiny, even for high flow rates.

(Not really. Friction/inertia will instead dominate and create an irreducible minimum.)

But the real limit will be cavitation. That same micro-g that reduces the back-pressure from the "head" also reduces the "tail" pressure, the rate that the fluid can reach the intake. (Normally, gravity in a tank is, in effect, acting as a first-stage "pump" pushing the fluid into the actual pump.)

Friction head loss is what Poiseuille will give you.  My mistake up-thread was thinking that it also includes dynamic pressure--I'm now mostly convinced that it doesn't.

But it's an excellent point about being cavitation-limited.  Static pressures at 1mm/s² are only in double-digit Pa, even for full tanks.  The h-P diagrams I've seen don't go low enough to show what the liquid saturation curve looks like at those pressures, but I doubt that it's encouraging.

This may be the best argument I've heard for why the system has to be pressure-fed.

It still might be desirable to connect the two ullage spaces, but have a gaseous pump pulling from the receiver and pumping into the sender.  That way, if you lose your pressure differential (e.g. by uncovering the sender's side of the transfer pipe), there's a nice, simple way to restore the differential--and there's still no venting required.

[eriblo]

Pumps will not cavitate as long as the ullage pressure is large enough.

[TheRadicalModerate]

Pumps will not cavitate as long as the ullage pressure is large enough.

Good point.  So you can still have equalized pressure between the two ullage spaces, as long as the absolute pressure is a few bar, correct?

[mikelepage]

Not sure a nose by nose rotational docking is even possible. The mass center of a docked combination of full depot and an empty starship, will be somewhere in the upper third of the depot oxygen tank. How rotation in this case should help settle something is beyond my expertise. Apart from that, the rotation is poised to become extremely unstable. If SpaceX would make that work, it would be a marvel in itself.

I meant to respond to this earlier. I'm attaching a cross-section of my model, just so it's clear what I mean when I say offset nose-to-nose docking. Offsetting this way (about 5m centreline to centreline) to avoids any need to move the LOX header tank and keeps the entire docking mechanism on the leeward side of Starship, so there's no need to change the heat shield either. It also addresses any intermediate axis instability issues (the offset config increases moment of inertia of this spin axis by ~3x versus the alternative, intermediate spin axis).

You're right that in the extreme case of a 1300-1600 ton depot and empty 95 ton starship, the centre of mass will be somewhere on the LOX side of the common dome (about 25m from the centre of rotation). But in that case you can still settle the LOX, so you start transferring LOX first, until the centre of mass moves into the methane tank, then you transfer all the methane, then you finish by transferring the final bit of LOX.

Lastly, I've realised that having a depot with a "prop transfer arm", as I showed in the video in my earlier post is completely unnecessary. I'll change my animation to assume we pipe any propellant up alongside the pipe to the LOX header tank, and have the transfer occur through the 4x probe/drogue setup I depicted next to the docking port. This way any two starships can transfer prop to each other (not just to and from the depot, which was a limitation with the previous config).   

I suspect that there will be reasons for settling acceleration to be greater than 1mm/s², but if that's the case, then combusting methox thrusters are about a jillion times easier to engineer than a rotating system.  Rotating systems are terrible to make reliable, especially when you're moving their center of mass around by pumping stuff from point A to point B.

Maybe I'm missing something here, but can you give me an example of why rotating systems are so hard to make reliable?  I must admit I have a bit of a cringe whenever I see tether systems being proposed, but supposing direct docking as in this picture, spun to produce settling acceleration of 1% of G (0.45 rpm), where do you see the complexity arising? Because I'd hope the potential side benefits of this config in a crewed context would be obvious to everyone (that's not a reason to do this if it makes the prop transfer problem harder, but I'm missing how it does that).

[edzieba]

Pumps will not cavitate as long as the ullage pressure is large enough.

Good point.  So you can still have equalized pressure between the two ullage spaces, as long as the absolute pressure is a few bar, correct?

If you already have high ullage pressures, then as calculated above you can omit the pumps entirely and move fluids by pressure difference along. Can't cavitate pumps if you don't have any pumps.

[TheRadicalModerate]

Pumps will not cavitate as long as the ullage pressure is large enough.

Good point.  So you can still have equalized pressure between the two ullage spaces, as long as the absolute pressure is a few bar, correct?

If you already have high ullage pressures, then as calculated above you can omit the pumps entirely and move fluids by pressure difference along. Can't cavitate pumps if you don't have any pumps.

There's a fundamental difference between high ullage pressures and high ullage pressure differences.  The first will eliminate cavitation.  The second will move propellant.  If you're going to pressure-feed, you need to manage the ullage pressures in both tanks, via venting (on the receiver) and heating (on the sender).

I'm still struggling a bit trying to figure out pump power requirements, so really big numbers there could change my mind.  But unless that happens, pressure-feeding sounds insanely more complicated than equalizing pressures, especially since the QD has pre-press plumbing built into it.

[edzieba]

Pumps will not cavitate as long as the ullage pressure is large enough.

Good point.  So you can still have equalized pressure between the two ullage spaces, as long as the absolute pressure is a few bar, correct?

If you already have high ullage pressures, then as calculated above you can omit the pumps entirely and move fluids by pressure difference along. Can't cavitate pumps if you don't have any pumps.

There's a fundamental difference between high ullage pressures and high ullage pressure differences.  The first will eliminate cavitation.  The second will move propellant.  If you're going to pressure-feed, you need to manage the ullage pressures in both tanks, via venting (on the receiver) and heating (on the sender).

I'm still struggling a bit trying to figure out pump power requirements, so really big numbers there could change my mind.  But unless that happens, pressure-feeding sounds insanely more complicated than equalizing pressures, especially since the QD has pre-press plumbing built into it.

Equalising pressures during pumping requires continuous ullage gas generation for the sender tank, and continuous venting for the receiver tank. This requires continuous closed-loop control of both gas generation and tank venting throughout the entire transfer process.

Pressure transfer requires pressurising the sender tank once at the start of transfer, venting the receiver tank once at the start of transfer, then opening the inter-tank valve(s) to allow fluid to flow. No additional venting or pressurisation is required during transfer. In the event of an extreme fluid volume transfer (e.g. completely full sender to completely empty receiver) then once transfer ceases (pressure equalised) the tanks can be isolated again, the sender repressurised, the receiver vented, and the inter-tank valve opened again to repeat the process. Venting and pressurisation are "run to completion" processes with no direct constraints on pressurisation/venting time or rate.

[TheRadicalModerate]

Equalising pressures during pumping requires continuous ullage gas generation for the sender tank, and continuous venting for the receiver tank. This requires continuous closed-loop control of both gas generation and tank venting throughout the entire transfer process.

No, it just requires connecting the two ullage spaces together with a small-diameter pipe.  Then, as prop is pumped out of the sending tank and into the receiving tank, that creates higher pressure in the receiver ullage and lower pressure in the sender, so ullage gas naturally flows from the receiver to the sender to equalize the difference.

I've attached a cheesy diagram of what I'm thinking about.  Note that this system will pump prop in either direction, which is a requirement for a depot, may be a requirement for a tanker (because sometimes tankers will receive prop from depots to send it to cislunar, transfer it, and then go straight back to EDL), and probably not a requirement for a target Starship.

The only requirement is that the joint ullage space be pressurized to the necessary level to avoid pump¹ cavitation.  A couple of bar?²

Pressure transfer requires pressurising the sender tank once at the start of transfer, venting the receiver tank once at the start of transfer, then opening the inter-tank valve(s) to allow fluid to flow. No additional venting or pressurisation is required during transfer. In the event of an extreme fluid volume transfer (e.g. completely full sender to completely empty receiver) then once transfer ceases (pressure equalised) the tanks can be isolated again, the sender repressurised, the receiver vented, and the inter-tank valve opened again to repeat the process. Venting and pressurisation are "run to completion" processes with no direct constraints on pressurisation/venting time or rate.

I'm not saying that pressure-fed transfers won't work; I'm saying that they're complicated, slow, wasteful, and error-prone.

Complicated:  Yes, you can do this only once at the beginning, but in a worst-case transfer (e.g. 1200t from depot to target Starship), and with the assumption that you have 5% ullage space in a full tank and you don't vent down to below half a bar, you wind up with 10.5bar in the receiving tank.  So you're venting multiple times, and you're pressurizing your sending tank to a pretty high value.

Slow:  In a blow-down tank (which is essentially what you've created here), and assuming that all of your pressure difference can be converted to dynamic pressure, then the volumetric flow rate from one tank to the other is proportional to the square root of the pressure difference.  (Δp = ½⍴v², and v = volumetricFlowRate / (π*rPipe²).  This doesn't include static drops due to viscosity, but they're tiny.)  So as the difference drops, so does your flow rate.  I'm too lazy and stupid to integrate this, but the upshot is you need to do, if not closed-loop control, then lots of vents to keep the pressures reasonable (see "complicated" above).

Wasteful:  Unless you're willing to vent all of the receiver ullage into a propulsion system on both ships that delivers the ullage thrust (which coincidentally would require exactly the same plumbing you'd need for equalized ullage pressure, but with more valves), then you're dumping a few tonnes of otherwise perfectly good prop into space.

Error-prone:  Uncover the sender's side of the transfer pipe even for a moment, and you've equalized pressures.  Then you get to start all over, with more complexity and waste.  And of course you have to mitigate whatever consequences result from a high-speed gas flow.  (I hereby coin the term "depot fart" to describe this phenomenon.)

_______________
¹Here's a spreadsheet on flow rates and power, with a couple of screenshots below.  One is for what I consider the worst case:  a depot-to-target full transfer in HEEO, with a constraint of 3 hours transfer time, so that it's possible to do a one-orbit RPOD, transfer, undock, and checkout before perigee insertion.  That would require close to a 4kW pump.  (Well, two 4kW pumps, running sequentially.)  The second one is for a tanker-to-depot transfer, with a pumps limited to 1kW, which takes about half an hour.

²My assumption for cold-pressurization (i.e. ullage pressurization without the Raptors running) is that you flow liquid prop into COPVs, seal 'em up, and heat the liquid to supercritical temperatures and pressures.  Then, when you need ullage pressure, you release gas from the COPVs.  Starship is going to need sources of GCH4 and GO2 for a variety of purposes:  ullage pressurization, warm gas thrusters, combusting gas thrusters, blow-down pressure for liquid methalox-driven equipment like pressure-fed liquid thrusters or APUs, etc.  So there's some irreducible complexity involved no matter what.  But that whole system is dramatically simplified by pumping liquid, not gas, as its first step.  I suspect that you can make your transfer pumps do double duty here, but I haven't attempted writing down the plumbers' nightmare required to implement it.

[Greg Hullender]

Maybe I'm missing something here, but can you give me an example of why rotating systems are so hard to make reliable?  I must admit I have a bit of a cringe whenever I see tether systems being proposed, but supposing direct docking as in this picture, spun to produce settling acceleration of 1% of G (0.45 rpm), where do you see the complexity arising? Because I'd hope the potential side benefits of this config in a crewed context would be obvious to everyone (that's not a reason to do this if it makes the prop transfer problem harder, but I'm missing how it does that).

I started off really keen on rotating systems, but I was eventually talked out of it. I can share the reasons that changed my mind. For proof, you'll need input from actual experts.

One of the biggest issues is that they have to minimize the number of "things no one has ever done before." If we're just talking blue-sky stuff that might be in use in 2100, that's different, but we're talking about what SpaceX might do in the next three or four years. That really rules out the most interesting rotational systems. In fact, there has been so little study of rotating systems in space that merely figuring out what you have to do to make a stable system is a nontrivial task.

Another factor I didn't consider is plumbing. Apparently cryogenic plumbing is really hard to make work. Every bend in a pipe and every meter of pipe is a potential problem. So proposals that involve running a lot of cryogens through a lot of pipe are generally dead-on-arrival. Nose-to-nose proposals either require really long external pipes or entirely new piping inside the rocket.

For the configuration you propose, there's the extra challenge that the depot will have anywhere from 0 to 1500 tons of fuel in it, but the tanker will only have 150, so this thing is going to be awfully unbalanced, and the center of rotation is going to change during fueling. There might be a way to make that work, but, again, it falls into the category of "stuff no one has ever done."

By contrast, ullage burns are well-established technology for settling propellant before an engine burn. For that purpose, they work at amazingly low accelerations--the sort of thing you could do for an hour without burning much fuel. It's a bit of a leap of faith that it will work for refueling, but it seems plausible with a lot less risk. Also, refueling can tolerate more sloshing than an engine burn, since it's not the end of the world if a propellant pump occasionally sucks some gas or a gas pump occasionally sucks some fluid--assuming you designed for that.

So when you've just got a few years to make this work, and you've got a choice between a technology that you're pretty sure will work at an acceptable cost vs one that could take ten or twenty years to work the bugs out, it's obvious why only the former gets serious attention.