Why not four pairs of SuperDracos in the F9 S1 interstage?

[Burninate]

is wind really such a problem?
...

No, it's not.  This entire thread is an attempt to solve a problem that doesn't exist.  The F9 landing is basically a smart bomb targeting a specific set of GPS coordinates.  It's just a smart bomb that starts out higher and faster than normal and that hoverslams instead of exploding (assuming it's on target.)

For the very first attempt they discovered they need more hydraulic fluid to have precise control to the target.  The next attempt will hit the barge more directly.  Maybe even "land."  Autopilots deal with wind all the time without any external information beyond how the aircraft is actually flying.

Odd how you can say its a problem that doesn't exist, when no-one has EVER landed and recovered a first stage successfully. Close, but no cigar, to quote a phrase.Once a stage has been landed multiple times successfully, then you say its a solved problem, but not yet.

Hoverslam != explode on impact.

The stage exploded because it hit the side of the barge while doing a hard divert.  It was doing a hard divert because it could not maintain an on target trajectory.  It could not maintain an on target trajectory because it did not control over it's primary aerodynamic control surfaces for the last minute of "flight."  It didn't have control of the grid fins because it ran out of hydraulic fluid.

Wind was not the problem.  Complex systems to anticipate the wind are not the solution.  The solution is to have control authority for the entire landing process.

The hoverslam itself is not some intractable problem and the term hoverslam is overly dramatic compared to the reality.  All the stage needs is good information regarding the distance to the ground and it's current deceleration.

I do not mean to imply that the targeting and landing control systems are trivial, just that with enough simulation, testing and experimentation I am confident they can be done and that they can be done reliably.

I feel like I've repeated myself too much so I'll do my best to refrain unless I have something new to add the the conversation.

I feel like my point has been lost in the noise, and maybe I didn't do a great job explaining myself, and people elaborated on misconceptions.

I am positing that a wind limit exists (not that I know what it is, just that it exists) beyond which the stage will not be able to land, because of the mismatch between the constant velocity of the air, and the fixed position and zero velocity of the landing pad.  This limit exists for a specific reason: The only way the rocket has of holding position against a wind, is by tilting itself.
1) Vertical descent, fighting wind, tilted profile: If the rocket is substantially tilted when the first landing leg touches the ground during a vertical descent, the force of the dropping rocket will cause it to torque around and maybe topple over.
2) Diagonal descent, static with wind, straight profile: If the rocket tries to track with the wind, coming in diagonally (requiring foreknowledge of wind), but with the tube itself perfectly vertical - that's a neat trick that introduces another problem: Now your rocket is travelling sideways, which when the all four landing legs touch the ground at the same time, induces a torque and maybe topples it.
3) Vertical descent, fighting wind up until last moment, dynamic profile: If the rocket tries to land as with 1) and then rapidly pitches to get all four legs on the ground at the same time for just the moment of impact - Then your rocket is now a rotating body, with angular momentum, and it might topple of its own volition.
4) Diagonal descent, static with wind, until last moment, dynamic profile: If the rocket tries to land as with 2) and then rapidly accelerates sideways to null horizontal velocity - well, the only way to accelerate sideways (induce a horizontal acceleration of center of mass) simultaneously induces a rotation... which might topple it.

Complicating all this is the CoM vs CoP issue in flight, and the fact that large shuttlecocking effects would be observed in wind.

Above some limit of wind, the fact that the only actuator you have simultaneously induces a horizontal acceleration, and induces a rotational acceleration, at the same time, prevents you from landing.  I don't know what that limit is, but it exists in principle.  It's a problem with bringing a high profile rigid object into rendezvous with a surface that has a different velocity than the wind, while still fighting gravity & vertical momentum, on only a rear actuated thruster.

A set of SuperDracos, while a bit overkill for the task, gives very resilient control of orientation if positioned at the top of the rocket, until all four legs are on the ground and velocity & momentum have been zeroed out.  There's still the issue of whether a gale might blow a *standing* landed stage over, but I imagine this limit is well above the limit at which you encounter problems at the landing interface event.

Again: All of this assumes a constant, predictable, prevailing wind, and has nothing to do with turbulent gusts; It has no bearing on hovering, and hovering has no bearing on it.

[Burninate]

There are likely dynamics-based solutions here, but they require extremely fine-grained (temporally and spatially) control of the throttle, and several seconds building up substantial horizontal velocity into the wind, at low vertical velocity, so that the horizontal velocity zeroes out at just the right second without shuttlecocking... but that doesn't factor in...

It's complicated.  More degrees of freedom of control, and specifically much more control authority in pitch and yaw, makes things much, much less fragile in those last few seconds.

[Jim]

I am positing that a wind limit exists (not that I know what it is, just that it exists) beyond which the stage will not be able to land,

And the difference between launch and landing wind limits are small, meaning that is not worth the additional mass, complexity and operational constraints (prelaunch and post recovery) posed by hypergolic superdracos.

[PreferToLurk]

I am positing that a wind limit exists (not that I know what it is, just that it exists) beyond which the stage will not be able to land...

I would posit that while you do NOT know what the wind limit is, that SpaceX DOES know what that limit is.  And yet, they continue attempt landings without superdracos attached to the interstage. 

Superdraco Pros:

1: solves for a hypothetical landing wind limit which we have no way to properly calculate without detailed engineering data which SpaceX will not/cannot divulge.

Superdraco Cons:

1: Weight
2: Complexity (more failure modes)
3: Cost
4: Serious thrust overkill unless you want to attempt a landing in a hurricane.
5: Might not be needed except in exceptionally strong winds: winds that would very likely (or possibly definitely, because we just don't know what the limit is) also prevent the rocket from launching in the first place.
6: SpaceX does not have any of these unknowns and has decided that the current configuration is good enough.

[JamesH]

They have made precisely one attempt to land where wind may be an issue.

Not that I think extra engines at the top are of any benefit at all. I'm more of a would it be worth being able to hover type of person.

[cscott]

It would not be worth it trying to hover.

And seriously, folks, you guys are trying to solve the problem that's been solved for 22 years, since the DC-X did a vertical landing in 1993.  Many have repeated the feat since, and has been demonstrated, the basic technology has now been thoroughly commoditized and you can buy off-the-shelf drones demonstrating all the necessary control algorithms and sensors.

There are two things which make RLVs hard which have never been done before:

1) Hypersonic reentry.  This is very hard, but we don't get pretty video and when successful isn't much to look at.  It's only when unsuccessful that you see the stage tumble/break apart/etc.  This is the hard part that CASSIOPE accomplished, and it's the part which makes the upcoming DSCOVR landing much harder than the CRS-5 attempt.  But nobody is talking about this.  Why?

2) Cost-effective reuse.  This is what F9dev2 will be investigating, but it's treated as a fait accompli somehow in this forum, with people claiming that all the New Mexico testing is worthless now.  It's not.  High-frequency operations and reuse is hard, and there is a lot of work yet to be done.

Please let's stop talking about the hoverslam.

[zodiacchris]

Amen, I can't believe we are still talking about this! 

[JamesH]

It would not be worth it trying to hover.

And seriously, folks, you guys are trying to solve the problem that's been solved for 22 years, since the DC-X did a vertical landing in 1993.  Many have repeated the feat since, and has been demonstrated, the basic technology has now been thoroughly commoditized and you can buy off-the-shelf drones demonstrating all the necessary control algorithms and sensors.

There are two things which make RLVs hard which have never been done before:

1) Hypersonic reentry.  This is very hard, but we don't get pretty video and when successful isn't much to look at.  It's only when unsuccessful that you see the stage tumble/break apart/etc.  This is the hard part that CASSIOPE accomplished, and it's the part which makes the upcoming DSCOVR landing much harder than the CRS-5 attempt.  But nobody is talking about this.  Why?

2) Cost-effective reuse.  This is what F9dev2 will be investigating, but it's treated as a fait accompli somehow in this forum, with people claiming that all the New Mexico testing is worthless now.  It's not.  High-frequency operations and reuse is hard, and there is a lot of work yet to be done.

Please let's stop talking about the hoverslam.

No, because nothing you say in the post means a hoverslam is guaranteed to be successful which is what you are saying (I actually think it will be, but I have no evidence to support that and no-one has supplied any)

A huge first stage rocket, with a T:W > 1 is not a three axis controllable electric propeller driven drone, that's a strawman. It's also not a DC-X (which could hover...). This 'hoverslam' (and I dislike the term but there seems no other option) has never been tested successfully. You can quote as many other technologies as you like, or extrapolate from what we have seen, but it has NEVER been tested successfully.

I'm all for optimism (I think they may well get it within the next couple of flights), but try and keep it pragmatic optimism.

I agree with your points 1 and 2, that's optimism gone too far.

[cscott]

A hoverslam is no different from all the other extremely complex control problems that computers accomplish effortlessly, every single day.  Forget C++ and python---go find a *MATLAB* programmer and ask them about it.

 (Oh, and the grasshopper/F9dev1 already demonstrated hoverslam at least once. Go search the forums for the analysis.)

Note that I'm not saying that executing a hoverslam is *easy*---it's still an engineering challenge to build a reliable machine which can execute your control program with extremely tight mass margins and after having survived hypersonic reentry, and do so in a way which is rapidly reusable, etc --- but see: those are the parts of the task which are *not a hoverslam*.

[JamesH]

Saying that a hoverslam is somehow different from all the other extremely complex control problems that computers accomplish effortlessly, every single day, reveals nothing but a profound misunderstanding of the task.  Forget C++ and python---go find a *MATLAB* programmer and ask her about it.

 (Oh, and the grasshopper program already demonstrated hoverslam at least once. Go search the forums for the analysis.)

Note that I'm not saying that executing a hoverslam is *easy*---it's still an engineering challenge to build a reliable machine which can execute your control program with extremely tight mass margins and after having survived hypersonic reentry, and do so in a way which is rapidly reusable, etc --- but see: those are the parts of the task which are *not a hoverslam*.

Cool. I wasn't aware that the Grasshopper turned its engines off, dropped until it reached two hundred mile an hour whist travelling at an angle, then reignited its engines, deployed it legs and landed successfully.

Mainly because it didn't. As far am I know, that is the sort of sequence required to hoverslam and land. Grasshopper had rigid legs, and I don't think ever turned it's engine off in flight.

I not disagreeing that a lot of complex stuff has already been achieved,  that is clear by the results so far. What I don't understand is why people think it's a completely solved problem when it's never been successfully demonstrated. Put it like this, if it's a solved problem you would be more that happy to bet considerable sums on money on working next time. Would you willing to bet considerable sums of money on the next flight being successful?

Please explain why you are so sure this is a solved problem when even Musk only gives it a 60% chance of landing next time.

(I used to work with the wife, a mathematician, of one of the matlab developers, so I am aware of it. Not sure what referencing it has to do with the discussion though)

[sublimemarsupial]

Saying that a hoverslam is somehow different from all the other extremely complex control problems that computers accomplish effortlessly, every single day, reveals nothing but a profound misunderstanding of the task.  Forget C++ and python---go find a *MATLAB* programmer and ask her about it.

 (Oh, and the grasshopper program already demonstrated hoverslam at least once. Go search the forums for the analysis.)

Note that I'm not saying that executing a hoverslam is *easy*---it's still an engineering challenge to build a reliable machine which can execute your control program with extremely tight mass margins and after having survived hypersonic reentry, and do so in a way which is rapidly reusable, etc --- but see: those are the parts of the task which are *not a hoverslam*.

Cool. I wasn't aware that the Grasshopper turned its engines off, dropped until it reached two hundred mile an hour whist travelling at an angle, then reignited its engines, deployed it legs and landed successfully.

Mainly because it didn't. As far am I know, that is the sort of sequence required to hoverslam and land. Grasshopper had rigid legs, and I don't think ever turned it's engine off in flight.

Grasshopper didn't, but the first stages of both F9-9 and F9-10 successfully did the hoverslam on the ocean surface before being destroyed by the subsequent tip over and impact.

[cscott]

Musk gives it a low probability because *the hypersonic reentry will be hard*.

[Misha Vargas]

[....]
Grasshopper had rigid legs, and I don't think ever turned it's engine off in flight.
[...]

Grasshopper II (actually F9R Dev-1) had legs just like those seen on the orbital flights. You can see them working just fine in three videos on SpaceX's YouTube channel. Anyways legs aren't seemingly an issue at any point, looking at their successful deployment in multiple flights. And hoverslams are already proven, as cscott said. It's really nothing more than slowing down right up to touchdown.

[mme]

...
(I used to work with the wife, a mathematician, of one of the matlab developers, so I am aware of it. Not sure what referencing it has to do with the discussion though)

The fact that it's the primer tool in designing and modeling real time control systems with complex feedback mechanisms is highly relevant:

http://en.wikibooks.org/wiki/Control_Systems/MATLAB

http://www.mathworks.com/products/control/index.html

http://www.mathworks.com/help/control/examples.html

[mme]

Actual real-world facts

March 7, 2013 Grasshopper lands with a TW > 1 using closed loop thrust vector and throttle control.

September 29, 2013 CASIOPE survives hypersonic reentry, restarts engine for the landing burn and decelerates until engine shutdown do to spinning caused by aerodynamic forces.  SpaceX reinforces the internal baffles and beefs up the cold gas thrusters.  Every subsequent stage that attempts reentry survives all the way to the water.

April 18, 2014 CRS-3, first stage with legs.  Survives hypersonic reentry, restarts it's center engine, deploys legs and touches down gently.  Garbled video is restored by a massive effort of volunteers so you can watch it yourself.

July 14, 2014 OG2, second flight with legs.  Survives hypersonic reentry, restarts engine for the landing burn, deploys legs and touches down gently.  You can watch the video yourself.

January 10, 2015, third flight with legs and the first with grid fins. Deploys the grid fins, survives hypersonic reentry and controls flight with grid fins until it runs out of hydraulic fluid and they go "hard over", restarts the engine, deploys legs, decelerates and tries to compensate for missing the target with a hard divert.  Hits the barge from the side and explodes.

So the following problems are already solved and are irrelevant to the complexity and difficulty of the task of landing the stage moving forward:

Hypersonic reentry (at least for certain mission profiles.)
Deployment of the gird fins and use from hypersonic down to subsonic.
Multiple engine restarts, including the landing burn.
Deploying the landing legs.

All that's left is improving the targeting so no hard divert is required at the end and the final touchdown.  Most of us believe that having control of the gird fins for the final minute makes the problem of targeting the landing area solvable.

The stage is decelerating from terminal velocity during the landing burn.  The landing burn is over 20 seconds long.  That is an eternity for a control system using LIDAR feedback to touch down gently.  The final seconds of the landing the rocket is moving quite slowly, the control system has an excellent map of the Merlin's actual real world responsiveness during the current flight.  And Grasshopper has demonstrated landing with a TW > 1.  The problem is clearly solvable.

Conclusion

I don't know how many attempts or refinements it will take, but the problem of returning an F9 from a mission and landing it on a barge is solvable without adding the ability to hover.

[The Amazing Catstronaut]

Actual real-world facts

I don't know how many attempts or refinements it will take, but the problem of returning an F9 from a mission and landing it on a barge is solvable without adding the ability to hover.

...Just to add another detail here to the excellent point above; if you try to fix something that isn't damaged, you risk creating problems where no problem is present - it happens all the time in engineering, programming, psychology, you name it.

In this case, you add a significant amount of mass to the top of the stage. This, in turn, will affect various other factors; the amount of RC thrust required to flip the stage before boostback/generally manoeuvre the stage, the stability of the F9R bottom stage whilst returned onto the pad (You're adding mass to the top of a very thin, otherwise stable object), potential plumbing problems, payload cuts, reducing the effectiveness of the grid fins, quite possibly having to reprogram the guidance software...

It's going to work as-is.

Edit: Apologies if somebody has already mentioned all of these points; I didn't read the full thread before posting.

[JamesH]

I salute your optimism. In fact I agree with you.

Everything is solved.

We just need to wait until the solution can be implemented. Rocket science is easy, rocket engineering is hard. I'm not arguing against the software or the science behind the technique. That is clearly sound. I'm less optimistic that the engineering can ensure the landing are reliable. A lot needs to work just right, guidance, hydraulics, engine, environmental conditions and there are no second chances with the hoverslam approach.

WIth regard to one point above, the F9R-dev cannot have done a test flight with an engine TW >1 at all times. That basically the definition of going up, not down. It carried a fuel load to ensure the engine could be throttled back enough to give TW<1.The returning stage does not have that much fuel, so at no point has a TW<1 except when it's engine off, so that is an invalid argument (and the F9R-dev can hover, as we all know, so really, it's flights are not entirely relevant wrt this discussion)

[Lars-J]

I salute your optimism. In fact I agree with you.

Everything is solved.

We just need to wait until the solution can be implemented. Rocket science is easy, rocket engineering is hard. I'm not arguing against the software or the science behind the technique. That is clearly sound. I'm less optimistic that the engineering can ensure the landing are reliable. A lot needs to work just right, guidance, hydraulics, engine, environmental conditions and there are no second chances with the hoverslam approach.

And what second chances do you think you get with hovering capability? "Oh wait, I came down in the wrong spot, let me move over and try again?" Is that what you think is needed to make this reliable enough for your engineering stamp of approval?

How much spare propellant must F9R carry - and extra throttling range to the M1D must be added - before you deem it advisable to even start testing?

This is a TEST program. You make things reliable by testing them. Sometimes it works, sometimes it doesn't. But that's OK. SpaceX is not going to go under if they can't reuse the next landed first stage. And no one is being forced to sit on it as it lands. It is an UNMANNED test.

WIth regard to one point above, the F9R-dev cannot have done a test flight with an engine TW >1 at all times. That basically the definition of going up, not down. It carried a fuel load to ensure the engine could be throttled back enough to give TW<1.The returning stage does not have that much fuel, so at no point has a TW<1 except when it's engine off, so that is an invalid argument (and the F9R-dev can hover, as we all know, so really, it's flights are not entirely relevant wrt this discussion)

Are you intentionally being obtuse? When was it claimed the whole flight used a TW > 1?

[cambrianera]

A lot needs to work just right, guidance, hydraulics, engine, environmental conditions and there are no second chances with the hoverslam approach.

If the things you described don‘t work right, you are toasted also if you can hover.
And hovering with a rocket is utterly inefficient (remember, only racing requires more performance than rockets!)

[nadreck]

Is it worth my mentioning again that the mechanics of bringing a hovering F9 first stage down onto the deck of a barge (or to an eventual land based location) are far more demanding than the direct in, reduce to zero V_x + zero V_y + zero V_z @ zero altitude and zero x, y and z offset from the centre of the X on the barge.  Once you are hovering, you now are being exposed continuously to the local winds, when you were coming in you could adjust your engine angle by a much smaller amount to compensate for wind (10s before landing your velocity is 100m/s a 5m/s gust/shear requires 1/10th the correction that it would at 10m/s 1s before landing and 1/20th the correction that you would need hovering 5 meters above the deck (presuming that you descend from hover at .5 g to gain speed and then decelerate with 1.5g thrust the remainder of the time).  As well, your grid fins had some authority down to a handfull of seconds before touch down and the longer you hover the longer you have to put up with buffeting at the level where it is most gusty/sheary.

Either the controls are accurate enough to allow the computer to manage engine thrust, reduce to zero altitude zero V at the target at 1G deceleration straight on to the deck, or it is not good enough to manage to go from hovering anywhere and get to the deck and you need a human remotely piloting it to land it.  If the engine responds variably to gymballing and thrust level control inputs the piloting program must respond immediately to that adjusting the control signals, but it already had to vary all of these things continuously as the weight changed and air density changed to match its course profile.