Whenever a fan of Kerbal Space Program mentions the super-efficient 'asparagus staging' in any professional context, the automatic response is 'That won't work. KSP fuel pipes are pure magic.'

The idea of 'asparagus staging' is that external boosters of the rocket feed their own fuel into the main engine, or into boosters that would separate later, so that at the moment of separation the craft's remaining tanks are still full, meanwhile, during the launch all engines are engaged - engines of later stages are not carried as dead weight, but contribute to thrust from moment one.

And yet, we had a system, where an external, detachable fuel tank would feed massive amounts of LOX bipropellant to engines in a separate stage - the shuttle's fuel tank feeding the SSME. At a glance, it looks like it's just a notch away from what would be needed for the 'asparagus'.

Can someone provide an overview of the crossfeed system - in particular the detachable joint that fed LOX to to the shuttle from the tank? Would it be efficient if reused for 'asparagus-staged boosters'? Or was it plagued with enough problems - or even just too expensive or inefficient for such use?

  • 2
    $\begingroup$ I think this question would be improved if you provide a brief discussion of what exactly you mean by "asparagus staging" (or make it clearer that's what you are talking about in the second paragraph). $\endgroup$
    – user
    Commented Oct 17, 2016 at 9:21
  • 2
    $\begingroup$ Let me add one note. As the wiki notes, using SRBs instead of liquid engine boosters may be more cost-efficient in the game - and in the real life too. But in the game, sticking small LFO tanks on top of the SRBs and feeding fuel from them to the main craft or the secondary boosters - tanks of such size that they run out when the SRBs run out - is even more efficient. Given a neat crossfeed mechanism it might be efficient in reality too. $\endgroup$
    – SF.
    Commented Oct 17, 2016 at 9:49
  • 2
    $\begingroup$ Crossfeed means from tank to tank. Since Shuttle had no tanks except the external main tank, it was not a "cross" feed, but a regular feed. $\endgroup$
    – Agent_L
    Commented Oct 17, 2016 at 14:50
  • 2
    $\begingroup$ @Andy: ...probably why I still haven't accepted one, waiting for some miracle that would help me decide ;) $\endgroup$
    – SF.
    Commented Oct 17, 2016 at 15:37
  • 5
    $\begingroup$ @FlorianCastellane: I don't have it at hand, but at least in KSP, delta-V to orbit optimalization can be a really misguiding heuristics when optimizing cost to orbit. Efficient designs tend to be expensive. SRBs are cheap. And especially in lowest stages, cost savings beat efficiency savings. $\endgroup$
    – SF.
    Commented Oct 18, 2016 at 2:18

3 Answers 3


1) Can someone provide an overview of the crossfeed system - in particular the detachable joint that fed LOX to to the shuttle from the tank?

The buzzwords to use for googling this topic are "ET Umbilical".

The system as implemented on Shuttle was complicated but caused major problems only once in the program (see below).

There were two umbilical areas towards the rear of the shuttle belly. Five fluid pathways connected through these two umbilicals: 17" liquid oxygen and liquid hydrogen lines, 3" gaseous oxygen and hydrogen repressurization lines, and a smaller hydrogen recirculation line. There were also numerous electrical and data lines. Large electrically operated doors (latched open at launch by electrically operated latches) covered the umbilical openings and also the aft structural connections between the ET and orbiter.

enter image description here

The outside of the doors was covered with shuttle TPS tiles.

At launch, obviously the doors were open and all umbilicals mated (the actual mating surface was called the ET umbilical plate). Each fluid connection had a shutoff valve on both sides of the interface.

After the SSMEs shut down, the ET separation sequence began. First the valves on each side of the fluid connections closed. Three explosive bolts on each umbilical plate fired to free the interface. Then triple redundant hydraulic actuators pulled the orbiter side umbilical plates into the Orbiter below the outer mold line. The system then checked to ensure that all valves were closed. If not, a waiting period began to let the systems blow down through the open valves and a warning was issued to the crew. Finally pyrotechnics blew the structural interface apart and the Orbiter fired down-firing jets to fly away from the ET. Then the pilot flipped switches to release the centerline latches on the doors, close the doors, and latch them closed.

The concerns with the system as designed were pretty obvious from the description above. If the fluid disconnect valves on the ET side did not close, the residual prop from the ET would vent out. If the ET was mechanically separated while this happened, it would tumble, and could hit the orbiter. Hence the waiting period. Failure of the umbilicals to retract into the orbiter could prevent the doors from closing, which would result in an imperfect heat shield for entry. This same concern existed for electrical or mechanical failures in the door closing mechanism. There was a documented EVA procedure for the crew to go out and crank the doors closed manually. This was straightforward except for accessing the area which was far aft and underneath the orbiter belly. There was also a concern that the disconnect valves could slam closed during operation of the SSMEs resulting in destruction of the vehicle. During the safety upgrades carried out after the Challenger accident, pneumatically operated latches were added to hold the valves open.

2) Would it be efficient if reused for 'asparagus-staged boosters'?

I cannot answer this, but my opinion is that the shuttle system design was unique to its application.

3) Or was it plagued with enough problems - or even just too expensive or inefficient for such use?

Despite the risks involved, the only actual major issue I am aware of with the system occurred during the STS-35 launch preparation, when the umbilicals experienced massive hydrogen leaks that delayed the flight significantly and resulted in the stack being rolled back off the pad into the assembly building.


1988 Shuttle News Reference Manual, p. 142 fwd and p. 92 fwd



Personal experiences

  • 4
    $\begingroup$ Good answer, one niggle: You say "When the SSMEs shut down in orbit" but presumably the shuttle would actually be suborbital at this point. Otherwise the tank would not re-enter. $\endgroup$
    – Jack B
    Commented Oct 17, 2016 at 14:59

The crossfeed seen in KSP works by pumping fuel from one fuel tank to another, i.e. against tank pressure. To do this, you need pumps (independent of the engine turbopumps). Pumping large amounts of fuel quickly is not easy.

You need to stop pumping cleanly when the tank is empty (or risk a pump explosion when the pump runs dry), this is difficult: you have to shut down a massive flow without causing a water hammer.

The Shuttle had a much easier setup: there was only one tank feeding propellants to the engines, no flow from one tank to another. The only pumps involved were the engine turbopumps. Shutdown was done by shutting down the engines, no need to close valves against a huge flow. The only thing the Shuttle's ET has in common with asparagus staging is the in-flight disconnection.

  • 1
    $\begingroup$ It's not THAT bad - the same turbopump that feeds the engine can feed the crossfeed - not pumping into the second tank, but into a secondary, smaller 'buffer' tank in the other booster, that the other engine is fed from. When fuel is about to run out, throttle the turbopump so that it only feeds own engine, shut the valves of the umbilical ET, open a valve of the main tank of the second engine, to start feeding into the buffer tank. Perform booster shutdown and separation normally. $\endgroup$
    – SF.
    Commented Oct 17, 2016 at 10:14
  • $\begingroup$ but then you move the problem: you have to switch the turbopump feed between the buffer tank and the main tank without causing the turbopump to hiccup. $\endgroup$
    – Hobbes
    Commented Oct 17, 2016 at 17:12
  • $\begingroup$ I have just to switch the input of the buffer tank; it does act precisely as a buffer to smooth out the flux, fed either from second booster or main tank, or even as independent tank with no input at all, during the phase between the crossfeed closure and opening main tank valve. The turbopump picks fuel from the buffer tank at all times; its output though is either split (to yet another tank and the engine) or not, which might cause a hiccup - but we're at the phase of shutting the engine down anyway. $\endgroup$
    – SF.
    Commented Oct 17, 2016 at 18:39
  • 1
    $\begingroup$ @Joshua: I care about explosions while the rocket is still within the blast radius of the spent stage. $\endgroup$
    – Hobbes
    Commented Oct 18, 2016 at 7:26
  • 1
    $\begingroup$ @SF. Feeding the engine and crossfeeding another tank requires very different pressures. What is needed for the engine would destroy the tank. $\endgroup$
    – Uwe
    Commented Oct 31, 2023 at 0:47

There are lots of factors in KSP's simulation model which make "asparagus" staging viable and beneficial. The big one is the weak drag model (at least in early versions): the penalty for making the rocket wider isn't too bad. Conversely, the weak coupling between rigid components in KSP causes very tall rockets to "wobble", and the overpowered and under-damped reaction wheel control systems can exacerbate that. So KSP rockets tend to have a much lower "aspect ratio" than real rockets.

The fuel crossfeed model is unrealistic in that it allows you to connect tanks A and B in any orientation and will always drain A first, regardless of having to pump against G forces. It's also possible to detach the stages in KSP while the engines are running - the Space Shuttle only detached the tank when the engines had shut down. There's no consideration of the risks of leaking drips of fuel and oxidiser from the crossfeed. Detaching a real side tank while under acceleration involves complex forces and may be hard to do cleanly; it quite often goes wrong in KSP!

  • $\begingroup$ While KSP allows many unrealistic configurations, it doesn't disallow realistic ones. Side boosters are not uncommon, especially in Russian designs (Soyuz, Energia). As long as the feed goes into a tank, and not directly into an engine, the engine is never undergoing any fuel supply hiccups. The "main" side can be secured against leaks with a rather simple unidirectional valve (allowing fuels "in" but never "out"; the booster - there are many ways to force separation rapid enough that by the time any risks occur, they can't endanger the main craft. $\endgroup$
    – SF.
    Commented Oct 18, 2016 at 13:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.