An Engineering Analysis of the Falcon 9 First Stage Landing Failure

schwit1 writes: AviationWeek has posted an analysis of SpaceX's latest attempt to land its Falcon 9 rocket on an ocean barge. Quoting: "SpaceX founder and chief technology officer Elon Musk tweeted that "excess lateral velocity caused it [the booster] to tip over post landing." In a later tweet that was subsequently withdrawn, Musk then indicated that "the issue was stiction in the biprop throttle valve, resulting in control system phase lag." In this statement, Musk was referring to "stiction" — or static friction — in the valve controlling the throttling of the engine. The friction appears to have momentarily slowed the response of the engine, causing the control system to command more of an extreme reaction from the propulsion system than was required. As a result, the control system entered a form of hysteresis, a condition in which the control response lags behind changes in the effect causing it.

Despite the failure of the latest attempt, SpaceX will be encouraged by the landing accuracy of the Falcon 9 and the bigger-picture success of its guidance, navigation and control (GNC) system in bringing the booster back to the drone ship. The GNC also worked as designed during the prior landing attempt in January, which ended in the destruction of the vehicle following a hard touchdown on the edge of the platform." In related news, SpaceX is hoping to attempt its next landing on solid ground.

Comments

Video from the barge

[Bruce Perens]

A video from the barge itself is here. Everything goes wrong in the last second of landing, with over-correction putting it down on one leg, and then the leg crumples.

Re:Video from the barge

[TWX]

Damn that was close. Kind of makes me wonder if using the barge as such a small target is contributing to the hard landings, simply because it's such a tiny target relative to the area that the rocket has to come down on. It's about the length of a football field; makes me wonder if they could set-down on an area ten times that large if most of these control problems simply wouldn't matter.

Re:Video from the barge

[cbhacking]

So, I asked this in the last thread but the discussion there was already mostly dead: what would it cost (presumably mostly a matter of weird) to upgrade the nose thrusters? These are cold-gas (nitrogen) thrusters, and I can't imagine they have a lot of power.

The Dragon uses hydrazine-based "Draco" thrusters for its RCS system; might it be worth adding a hydrazine thruster with a few seconds of fuel in place of the cold-gas thrusters, enabling the rocket to correct its orientation in the moment of touchdown (when it can no longer use engine gimbaling)? For that matter, how does the thrust of a hydrazine thruster (I think a Draco goes to about 90 lbf) compare to a cold-gas thruster? Wouldn't want to bend the rocket with excessive pressure at the top, after all.

Alternatively, the rocket contains a bunch of compressed gases (helium is used for pressurizing the fuel/oxidizer tanks, I believe). Would it be possible, on landing, to vent some of that pressure to provide additional attitude control? It might reduce the rocket's rigidity a bit, which could be bad, and it's a high-pressure valve (of course), but when your concern is that the rocket land upright (you'd want to use this in the moment before touchdown, to avoid excessive pressure on just one leg as appears to have happened here) it might be worth it. The obviously don't have a *lot* of excess pressure - fluids cost weight, pressurized ones moreso - but they probably have some and it might be suitable for a last-second thruster.

Thoughts?

Re:Video from the barge

[cbhacking]

Agh, bloody autocorrect and stupid failure to proofread. First paragraph:

What would it cost (presumably a matter of weight) to upgrade the nose thrusters?

Re:Video from the barge

[bledri]

So, I asked this in the last thread but the discussion there was already mostly dead: what would it cost (presumably mostly a matter of weird) to upgrade the nose thrusters? These are cold-gas (nitrogen) thrusters, and I can't imagine they have a lot of power.

The Dragon uses hydrazine-based "Draco" thrusters for its RCS system; ...

Thoughts?

You, like many people, are trying to solve the wrong problem. Fix the over-correction and there is no need for rocket powered thrusters in place of the cold gas thrusters. Fix the root cause, don't mask it with a heavy/expensive kludge that will come with a host of it's own failure modes.

Re:Video from the barge

[DerekLyons]

Kind of makes me wonder if using the barge as such a small target is contributing to the hard landings, simply because it's such a tiny target relative to the area that the rocket has to come down on

Since the first attempt hard landed because it ran out of attitude control gas, and the second hard landed because of a control valve problem... how would a larger target have helped? In case of the first attempt, you've still got to control your attitude regardless of the size of the field. In the second, the size of the field is irrelevant if you can't properly control the vehicle in the first place.

Seriously, don't be misled by the frantic activity in the final seconds of the most recent attempt. That burst of activity was the vehicle attempting to null it's horizontal velocity and then trim it's attitude before landing - something it has to do regardless of the size of the field.

The basic flaw in the landing sequence isn't the size of the target, it's the design of the vehicle. Its minimum T/W ratio is well over unity at landing, meaning it can't hover, can't ease itself down, and you have to take great care to not end up with positive vertical velocity. The only way it can land (with any reasonable sized target) is to approach at high speed, then at the last second try to null horizontal velocity without excessively reducing vertical velocity (I.E. bouncing), followed by a return to vertical and touchdown.

You could avoid this by having a circle of paved ground a quarter to half a mile in diameter - but that's not cheap to build or maintain given the need to resist a rocket's exhaust. Long term, given that the tests are essentially free*, it's cheaper and easier to figure out how to land precisely on a smaller target.

* The first stage is bought and paid for by the launch customer - and so long as the added equipment for landing poses no undue risk during ascent, they don't care what happens to it after separation.

Re:Video from the barge

[cbhacking]

I'm not trying to fix the wrong problem, I'm trying to add a backup for the fix. Shit happens. Parts will fail, valves will stick, unexpected winds or waves will occur.

I thought the fact that the primary goal was to correct the problem that caused the excessive lateral velocity was so bloody obvious that it didn't need saying, but I guess I forgot I'm on the Internet. The purpose of my idea was not "fuck it, fixing a little problem is hard, let's do something much more complicated", it's "shit happens. What can we do to survive the likely error modes?"

Re:Video from the barge

[Bruce Perens]

If there's one thing they should work on, it's not thrusters but having the capability to throttle to hover. That would potentially change the entire low approach. It is complicated by the fact that engine performance goes nonlinear in the low range.

Re:Video from the barge

[bobbied]

I'm not trying to fix the wrong problem, I'm trying to add a backup for the fix. Shit happens. Parts will fail, valves will stick, unexpected winds or waves will occur.

And when they happen, you loose the vehicle...I think you are trying to address the wrong problem. IMHO the smart money is on making what they have work though some software adjustments or if they really can't do it that way the minimum adjustments in the hardware to smooth out the valve responses. They are really close here, and from what I just was reading they are a software change away from making this work. Just remove the oscillation problem in software by tweaking the delay calculations in the feedback and I'll bet it's near perfect...

You seem to want to redesign the whole thing to increase the system's thrust and overcome it by brute force... I think a small bit of tweaking would be enough and cost a lot less.. But what do I know, I'm not a rocket scientist, just a EE software guy...

Re: Video from the barge

[Anonymous Coward]

Their ultimate goal is to land this on another planet, then take back off. That is why they are working so hard on unassisted landing.

Re:Video from the barge

[cbhacking]

Nitpick: The first attempt ran out of hydraulic fluid (for the guidance fins), not out of propellant for the RCS thrusters.

The rest of what you say is generally true, although a larger target *would* help. The advantage of a larger target is that, while you still have to zero your horizontal velocity, you don't have to zero it anywhere terribly precise. You can pick an optimal set of thrusts that results in the correct orientation and velocities (horizontal and vertical) without worrying overmuch *where* that series of thrusts has you touching down. Both attempts so far clearly demonstrate the ability to do an excellent good job of targeting a (relatively) tiny barge, but currently, if the rocket would come down even 100' (30m) to one side of its target spot, it needs to induce a horizontal momentum (which requires leaving a vertical attitude as well, it can't just translate sideways) and then null that momentum at the right moment (and fix its attitude). That's hard.

To clarify for the person who keeps misunderstanding my posts: they should, of course, plan for the barge-level of landing precision. They should aim for a precision of inches, and within a year, they may get it... 90% of the time. Stuff goes wrong, though, and (especially early in the testing of such a system) it behooves them to use a larger landing area so that there's some margin for error. I'd say their land attempt (possibly next CRS launch, in a couple months) has a very good chance of being their first success.

Re:Video from the barge

[Anonymous Coward]

Watch the vid posted in the parent, you will see the flexing of the middle once it starts tipping. Nets wont keep it from destroying itself, it would just crumple, then explode (probably).

Re:Video from the barge

[viperidaenz]

You want to add a tank of highly toxic fuel to rocket that tends to explode when it lands?

Re:Video from the barge

[complete loony]

If you look at the video posted yesterday, the rocket was coming down straight. Then it deliberately tipped itself over, shed the last of it's vertical and horizontal velocity, and tried to right itself. IMHO, if more of the horizontal velocity was shed earlier, It'd be easier to stay balanced at the end.

Re:Video from the barge

[bledri]

I'm not trying to fix the wrong problem, I'm trying to add a backup for the fix. Shit happens. Parts will fail, valves will stick, unexpected winds or waves will occur.

I thought the fact that the primary goal was to correct the problem that caused the excessive lateral velocity was so bloody obvious that it didn't need saying, but I guess I forgot I'm on the Internet. The purpose of my idea was not "fuck it, fixing a little problem is hard, let's do something much more complicated", it's "shit happens. What can we do to survive the likely error modes?"

And I stand by my response, it's better to reduce the likelihood of shit happening. There are ways to reduce sticton and there are ways to handle it better when it happens. Trying to fight an oscillation induced by a lag/overcorrection of a main engine with a thruster is a losing battle. If the thruster is up top it will just rotate the stage around the center of mass (which is really what a thrusters job is, to orient the stage. Not move it through space laterally.) If you look at the thruster firing during the failed landing attempt, it's trying to stop the rotation of the stage. It is not trying to stop the stage's "lateral motion." The direction it's firing, a crazy powerful thruster that could have made a difference would have added to it's lateral motion right off the barge. And it would have driven the opposite leg into the deck by using the already collapsing leg as a pivot point.

To some situations the correct response is to do less, not more.

Re:Video from the barge

[Firethorn]

And when they happen, you loose the vehicle..

It stinks if you're going to lose the vehicle when 1 thing fails. Though programming wise, they might 'fix' the thruster issue the same way they did some LCD panels - change the request curve so that when ordering the valve to change a small amount - IE from shut to 10%, it very briefly orders a larger change to overcome the static friction, such as 50%, then countermands that with the 10% order a millisecond or so later.

Re:Video from the barge

[SternisheFan]

Musk has said that they're shooting for an 80% success rate by years end.

Re:Video from the barge

[catchblue22]

Is there a reason the rocket has to land upright unassisted? What about some sort of bungee-mesh around the edges to catch it if it topples?

Parachutes are too heavy for a nine engine rocket, and would either land in corrosive seawater or on land with a hard damaging landing. To be truthful, they have already accomplished the most difficult part, namely bringing a rocket from 140km altitude to a tiny location in the middle of the ocean. They have also already demonstrated the landing part several times. If you notice, those landings come down at pretty much the exact location they began at.

Re:Video from the barge

[thegarbz]

I'm not trying to fix the wrong problem, I'm trying to add a backup for the fix. Shit happens. Parts will fail, valves will stick, unexpected winds or waves will occur.

A backup is nice but from a control perspective when you have an unstable control scheme additional power (gain) is about the worst possible situation.

The problem occurred because of an overshoot of a control reaction. The only thing achieved by better thrusters would be more overshoot.

Re:Video from the barge

[cbhacking]

I thought that was him estimating an 80% chance of at least one success by year's end, which (considering that they have a lot of launches this year) is a very different thing. Do you have a source for that?

Re:Video from the barge

[cbhacking]

Explode when it crashes, there's a difference. SpaceX has demonstrated powered vertical landings on their test vehicles, there were no explosions.

Hydrazine is nasty stuff, but you don't need very much of it (a few second's worth, maybe) and the *entire point* is to avoid the crash, so what happens in the event of a crash is much less important.

Re:Video from the barge

[DerekLyons]

IMHO, if more of the horizontal velocity was shed earlier, It'd be easier to stay balanced at the end.

At the cost of increasing the difficulty of targeting the landing, and still having to null any horizontal velocity due to the wind. TAANSTAFL.

Re:Video from the barge

[viperidaenz]

Did you not see the thrusters it's already got were more than capable of manoeuvring it? In fact, they provided too much thrust and over-corrected it, apparently due to a problem with the valve or control software.

Re:Video from the barge

[Eloking]

You, like many people, are trying to solve the wrong problem. Fix the over-correction and there is no need for rocket powered thrusters in place of the cold gas thrusters. Fix the root cause, don't mask it with a heavy/expensive kludge that will come with a host of it's own failure modes.

I partially agree with this. Correcting the root of the problem seem to be the best solution, but one thing my years in precise robot precision though me is that usually, "asservissement" (french word because I don't know the exact English translation, but I would guess it's control with feedback) is more efficient while correcting the root isn't always possible.

Re:Video from the barge

[SternisheFan]

Elon Musk Verified account 12:49 PM - 13 Apr 2015

@elonmusk

Odds of rocket landing successfully today are still less than 50%. The 80% figure by end of year is only bcs many launches ahead.

https://twitter.com/elonmusk/s...

Re:Video from the barge

[Rei]

Is "The Vehicle" a euphemism for something? ;) What exactly are you talking about loosing?

(And I hardly think that suggesting a more powerful RCS as a backup (backups being critically fundamental rocket design) is "redesigning the whole thing"). I don't know why they went with cold gas thrusters, but hydrazine RCS thrusters are mature tech, one that even SpaceX themselves uses - they're reliable and have a good power to weight ratio for their size. I presume there's a reason they went with nitrogen instead, but I don't know that reason.

Re: Video from the barge

[Rei]

Not necessarily. A stronger RCS system would probably have for example saved this last rocket - even though that wasn't the fundamental problem. Again, that's the purpose of backup systems. They're not the primary - they're there for if the primary fails, which it will sooner or later even if you've ironed out the major bugs.

I'd wager that they chose nitrogen not because of the best thrust / mass combination for their needs, but to try to make regulatory approval for landing on land easier.

Re:Video from the barge

[bobbied]

So the grammar police are still out in force.... OK, what ever floats your boat.

Re: Video from the barge

[HuntingHades]

That may be further down the line. My understanding is that these are currently just the rocket boosters, which they want to recover simply because it would hopefully make them reusable and drive the cost of launches way down.

Re:Video from the barge

[cbhacking]

I'm not actually sure which of use that tweet backs up; I thought he meant "I estimate we have an 80% chance of at least one landing by year's end, because there are lots of launches between now and then and there's a pretty good chance at least one will succeed." It would be pretty awesome if what he meant was "By the end of the year, we'll have so much more launch practice that, even though each launch has less than 50% chance of recovery right now, by year's end it should be about 80%." I hope the latter is the case, but the two are very different statements and the latter is much more optimistic. I was hoping you could indicate which one was more accurate.

Re:Video from the barge

[SternisheFan]

By years end, odds of rocket landing successfully will be 80%. Took me awhile to understand that too.

You Can See

[The Raven]

The 'stiction' is evident when the rocket is initially coming down and swinging to the left of the video frame, before you see it (over-)correct and swing back to (and past) vertical. I watched that section wondering why the rocket went excessively to the left in the first place, and a stuck valve makes a lot of sense.

Re:You Can See

[slew]

Although I'm willing to take it on faith that it might be 'stiction' on the actuator side, I'm pretty sure extreme variations which can cause unanticipated phase lag from a measurement system (say a gyro or a flow meter) is equally disruptive to a control system and can generically explain control system oscillation events as well.

Sometimes you never really know until you set up the conditions and simulate the crap out of it in an controlled testbed as it is much harder to distinguish the difference in a "live" situation.

Re:You Can See

[Anonymous Coward]

That would require strengthening the body to withstand forces from other directions. More weight. This is a kludge and the wrong way to solve the problem.

Re:You Can See

[bobbied]

Although I'm willing to take it on faith that it might be 'stiction' on the actuator side, I'm pretty sure extreme variations which can cause unanticipated phase lag from a measurement system (say a gyro or a flow meter) is equally disruptive to a control system and can generically explain control system oscillation events as well.

Sometimes you never really know until you set up the conditions and simulate the crap out of it in an controlled testbed as it is much harder to distinguish the difference in a "live" situation.

True, but if you have good enough instrumentation and data collection built into your system it's not *that* hard to figure out what exactly happened. I'm sure they have this system well instrumented so it won't take them long to narrow down on a solution to the problems they are seeing. My guess is that they just need to tune their control systems a bit to account for the valve movement delays and all will be better. They might need to work on the actuators for the valves some if they cannot smooth things out in the software, but it looks like a small software change is all they need IMHO.

Re:You Can See

[Grog6]

I really didn't see that as a thruster fuckup. I saw the gyros overshoot, the control system forget or be confused about where "UP" was, Land, and then eventually let go. Gyros often overshoot during high delta-v operations; the bearing load changes randomly and extremely. 10g's of vibration, 1mm in length will change a bearings properties amazingly. (For an extreme example of the effects of vibration, look up "POGO effect" as it related to the Saturn V rocket. :) This may be geek sniping, so beware.) It may have been confused by the burn vibrations, and delta-v problems; there's a lot of shit going on at the end. The thrusters were doing what they could, but given the attitude it thought UP was at the end, and that it looked like they weren't taking gyro precession into account, it looked like it fell over when the gyros hit their limits. "Eh, this looks good to me, why are all you guys going apeshit?" If it doesn't use Gyros for attitude control, I'd suggest they might help. But in that case, SpaceX's Control software is even more freaking awesome. :) I hope the next one works!

Re:You Can See

[Bruce Perens]

Microminiature accelerometers are really cheap and very very light, and you don't have to wait for them to spin up or deal with their mechanical issues. I doubt you will see a gyro used as a sensor any longer.

Similarly, computers make good active stabilization possible and steering your engine to stabilize is a lot lighter than having to add a big rotating mass.

Re:You Can See

[itzly]

Microminiature accelerometers are really cheap and very very light, and you don't have to wait for them to spin up or deal with their mechanical issues. I doubt you will see a gyro used as a sensor any longer.

Gyros are still required to detect changes in rotation. However, modern MEMS gyroscopes are tiny, and they don't rotate but use a vibrating element. http://en.wikipedia.org/wiki/V...

You can get combinations of accelerometers and gyroscopes (and sometimes even magnetometers) all integrated in a single tiny device that's within reach of school kid projects: http://www.geekmomprojects.com...

Re:You Can See

[gl4ss]

sure you will, they have different uses.

thats why your modern normal smartphone has both a gyro and an acceleration detector chip. older phones tend to just have the acceleration detector. you can use the accel chip for detecting orientation, if you are not slowing or speeding up into some direction...

sure.. the gyrochips might not have an actual old school spinning device inside but they function the same.

Re:You Can See

[Thelasko]

Yes, the controls system appears to definitely know how it wants the rocket to be oriented. It looks like it keeps overshooting it's target though. If it is PID controlled, the 'stiction' would cause the integral term to 'wind up' and overshoot once the valve becomes unstuck. This can happen every time the valve changes direction as it must come to a stop at some point to do so.

I've run into this kind of thing a few times in my career. I'm curious how they fix it. In my experience it is never an "easy fix" as Musk says. There is always some performance tradeoff.

Doesn't look close

[tomhath]

FTFA:

Falcon 9 was seconds away from what would have been the first successful landing of a used booster stage on SpaceX’s Autonomous Spaceport Drone Ship

I suppose that's one way to look at it. But actually it was seconds away from exploding in a huge ball of fire.

Re:Doesn't look close

[Wraithlyn]

That's why they used the words "WOULD HAVE BEEN".

Re:Doesn't look close

[cbhacking]

You saw the part where the reaction control thruster at the top (little white plume) is trying to keep the rocket upright, didn't you? Lasts about four seconds. The rocket had already touched down (on at least one leg, with engines shut off) at that point. If the stage had been even a *little* closer to upright, or the thruster a *bit* more powerful, the rocket would have settled onto all four legs and that would be it.

Also, the stage swings through vertical in the moment before touchdown. It's that half-second of overcorrection that doomed it. It was over the barge and in the correct orientation less than 1000 millisecond before touchdown. I'd say the only reason that "seconds away from what would have been the first successful landing" is wrong is because it was more like *one* second.

Re:Doesn't look close

[cbhacking]

The thrusters aren't even supposed to be needed there, actually. They're only supposed to fire in very short bursts, not a continuous stream like in the video. As for the legs holding up, just one of them supported the whole rocket for a few seconds; all four should have had no trouble. We know a hell of a lot more than just that it crashed. To claim otherwise is to embrace ignorance.

There is literally no point at all to living in a world where you are only concerned with the things that you absolutely know. You don't absolutely know *ANYTHING*, you could be a simulation in some advanced being's AI-run world, along with everything the program running you has ever simulated observing. The only way to achieve anything difficult is to analyze the differences between failures and successes, and a part of that analysis is to determine how close you came to success.

But then, you probably already knew that and are just a naysayer...

Input from a UI designer instead?

[Anonymous Coward]

Engineers can be useful in the beginning of a project but I think I'd like to hear from a UI designer instead, in the last couple years these guys have really been showing their stuff around the web.

Re:Input from a UI designer instead?

[Dantoo]

Thanks. I really needed that laugh.

Re:Alighting on land

[garyisabusyguy]

While watching the last landing attempt I had to wonder if the rocket experiences different amounts of lift when it is over water, as opposed to being over the barge...

Will this be a factor that is alleviated by landing on terra firma?
If they do make their next attempt on land, what are the potential take off and landing points?
Many people noted advantages to landing down range of the launch point, and not being able to launch over the continental US
Does that leave the options of
1. Launching from Brownsville Texas and landing at Cape Canaveral Florida
2. Launching from Florida and landing on some Island downrange
3. Launch from Kwaj (or at sea) and land at Vandenberg
4. Launching and landing at the same site, presumably solving whatever fuel issues there were
5. ???

I wonder why he bothers...

[Kjella]

In a later tweet that was subsequently withdrawn, Musk then indicated that "the issue was stiction in the biprop throttle valve, resulting in control system phase lag."

Anything he leaves for more than 0.5 seconds is going to be reported, retweeted, screenshotted and several articles posted. Just google "musk stiction biprop" and you get plenty hits, no real "undo" button for such a public figure.

Re:I wonder why he bothers...

[Mr D from 63]

They need to use a Segway valve... that'll keep it upright.

Re:I wonder why he bothers...

[bledri]

In a later tweet that was subsequently withdrawn, Musk then indicated that "the issue was stiction in the biprop throttle valve, resulting in control system phase lag."

Anything he leaves for more than 0.5 seconds is going to be reported, retweeted, screenshotted and several articles posted. Just google "musk stiction biprop" and you get plenty hits, no real "undo" button for such a public figure.

He's tech savvy enough to know it's not scrubbed from the internets nor the collective consciousness. This just seems to be how he uses twitter. He regularly tweets things he leaves up, but he also uses twitter to have conversations with people and then he deletes those tweets after the conversation is over. One theory is that he likes a "clean" twitter history. But who the hell knows?

Re:I wonder why he bothers...

[Solandri]

If you post something and later find out it's wrong, the responsible thing to do is to either correct it, or retract it. AFAIK Twitter does not allow you to edit a tweet, so the only responsible choice is to delete it. Leaving it up just allows the wrong info to continue spreading with the air of authority.

Re:I wonder why he bothers...

[viperidaenz]

I'm pretty sure Elon Musk doesn't want to risk falling off a cliff

Holy Stiction, Batman! WTF is hysteresis?

[__aaclcg7560]

I don't think I've ever seen such big words in a summary before. I may still have a dictionary on the bookshelf somewhere.

Re:Holy Stiction, Batman! WTF is hysteresis?

[Skidborg]

No need really, since they stop to define both of those terms in the article itself....

Re:Holy Stiction, Batman! WTF is hysteresis?

[luckymutt]

I don't think I've ever seen such big words in a summary before. I may still have a dictionary on the bookshelf somewhere.

Um...you have the Internet right in front of you.

Re:Holy Stiction, Batman! WTF is hysteresis?

[cbhacking]

A) You've mastered "dictionary", "bookshelf", and "somewhere", but an eight-letter word is too big for you?

B) Welcome to the Internet. You probably got here using "the blue e", right? Tip: highlight (whoops, big word) the scary word with your mouse, right-click on it, and click the option that will search or define the word for you! No need to go over to your bookshelf at all, and you get to avoid looking like a lazy ignoramus (sorry, is that one too long?) at the same time!

Seriously, I get that you're probably joking, but that is a really stupid thing to complain about. Definitions don't belong in a summary, especially not a summary delivered through the biggest inter-connected information network humanity has ever created.

Re:Holy Stiction, Batman! WTF is hysteresis?

[Solandri]

Stiction and hysteresis are well-known terms in engineering and physics. I think that's part of the allure of Musk. He's not some MBA CEO who has no clue about the minutia of what his company does. He's a scientist/engineer at heart who could with a little training reasonably step in at any grunt-level position at his companies.

Re:Holy Stiction, Batman! WTF is hysteresis?

[MachDelta]

It's ok, I have a car analogy for hysteresis:

So, my first car was a 92 Plymouth Laser RS Turbo. It had a (rudimentary) cruise control system that wasn't quite capable enough for a turbocharged engine. The problem was that cruise control behaved as if throttle input was linear, but a 90's era turbocharged car is the very definition of non-linear throttle input. Thus, any time cruise was on and you hit a decent slope, the system would begin to oscillate. When the car slowed from climbing a hill, cruise control would apply more throttle - but then the turbo would spool up and you'd overshoot your desired speed. This caused the system to sharply back off, causing the turbo to slow, power to drop, and thus speed to drop too rapidly. The car ended up stuck in this long loop of "too slow, too slow, more gas more gas more more mooooOOOOH CRAP TOO MUCH TOO MUCH OFF OFF OFF WAY Ooooh that's better, much better this is nice, I'm right on, wait, no no no wait too slow, too slow more gas more more MOOOOAAAAHHHH TOO FAST TOO FAST..." which could sometimes last for miles, unchecked. That's a car experiencing hysteresis. The solution was as simple as very brief and very gentle throttle input right as the system started to back off too much. You could keep the turbo spooled and nail target speed for just a second, and then cruise could simply hold the throttle for you... until the next hill. Fortunately for me, I was young at the time and not quite as interested in cruise control as I was in long 3rd gear pulls while passing people on a two-lane highway.

Re:Holy Stiction, Batman! WTF is hysteresis?

[__aaclcg7560]

The car ended up stuck in this long loop of "too slow, too slow, more gas more gas more more mooooOOOOH CRAP TOO MUCH TOO MUCH OFF OFF OFF WAY Ooooh that's better, much better this is nice, I'm right on, wait, no no no wait too slow, too slow more gas more more MOOOOAAAAHHHH TOO FAST TOO FAST..." which could sometimes last for miles, unchecked.

I had that happened to me when the fuel regulator for my car decided to go out. Line pressure leveled out once I got up to 65MPH. Slower than that it was all over the place. Drove straight to the mechanic to get it fixed.

Re:Holy Stiction, Batman! WTF is hysteresis?

[__aaclcg7560]

If you want to communicate with the public, you need to do so at the 7th grade level. Yes, ignoramus is a big word.

Stuff happens

[tomhath]

Kind of reminds me of when Neil Armstrong crashed the Lunar Lander Simulator

Re:Alighting on land

[Bruce Perens]

The booster can indeed make it back uprange to Kennedy Space Center, and they've leased a landing pad for it there. Besides the turn-around burn, they tilt the booster against the airstream and let aerodynamics push it back uprange during that 78 mile descent.

He's actually testing bombproof barges

[ankhank]

which, when perfected, will be stationed in his volcano lair ....

splashdown

[Jookey]

Why don't they just use a splashdown? It seems that corrosion resistance is relativly easier to solve than landing a rocket on a platform thats swaying in the ocean. You could even have it splashdown in a freshwater lake or a barge half filled with fresh water. You would also save the weight penalty of landing gear.

Re:splashdown

[Bruce Perens]

When you last flew a jet somewhere, why wasn't it a seaplane? Surely such things would be an easier problem to solve than building airports.

Short of giving you the starter course in rocket engineering, I can only say no, it's not easier.

Neither failure was due to target size

[DerekLyons]

The rest of what you say is generally true, although a larger target *would* help. The advantage of a larger target is that, while you still have to zero your horizontal velocity, you don't have to zero it anywhere terribly precise.

Yeah, you do. Given the narrow footprint and the low CG of the vehicle, if the horizontal velocity wasn't as close to zero as you can get at touchdown - it's very likely to tip over. (Even if you don't damage the landing legs in the process.) The upper part of the vehicle isn't heavy, but it has a very long lever arm.
 

You can pick an optimal set of thrusts that results in the correct orientation and velocities (horizontal and vertical) without worrying overmuch *where* that series of thrusts has you touching down.

In the end, that makes far less difference than you think because while you can reduce the amount of horizontal velocity that needs to be nulled you cannot eliminate it. (Not without launch criteria that include "near zero wind at the recovery site", which is beyond impractical.) The result is, with the current vehicle, you still have to null horizontal velocity at the last second before touch down. The basic problem is that the vehicle is badly designed for what it's being asked to do.
 

. Both attempts so far clearly demonstrate the ability to do an excellent good job of targeting a (relatively) tiny barge, but currently, if the rocket would come down even 100' (30m) to one side of its target spot, it needs to induce a horizontal momentum (which requires leaving a vertical attitude as well, it can't just translate sideways) and then null that momentum at the right moment (and fix its attitude). That's hard.

Both times they've hit the barge almost dead center - I fail to see how that's an arguement for a larger landing area since neither failure was caused by the landing area being too small. Both vehicles would have crashed regardless of the size of the landing area due to control system failures. (Attitude control on the first, throttle control on the second.) That's what neither you nor the OP seem to grasp.

Re:Neither failure was due to target size

[Firethorn]

Yeah, you do. Given the narrow footprint and the low CG of the vehicle, if the horizontal velocity wasn't as close to zero as you can get at touchdown - it's very likely to tip over. (Even if you don't damage the landing legs in the process.) The upper part of the vehicle isn't heavy, but it has a very long lever arm.

You're not getting what he was saying. To put it in car terms, he's saying that 'stop the car in front of that building' is a much easier task than 'stop the car on that postage stamp', which is what they're trying to do with the barge.

In the former scenario, if you under or overshoot the location of your stop, you're still fine. You'd miss the barge, but with a large piece of flat land, you're still landing on flat terrain. You can tune your control jets to always use 'small' corrections, and wait for things to steady before thrusting again.

Both times they've hit the barge almost dead center - I fail to see how that's an arguement for a larger landing area since neither failure was caused by the landing area being too small.

That's how you see it. I'm seeing that in order to maintain that small of a target the thrusters are thrusting a lot harder than they would have to otherwise, which has resulted in failures.

Though if they're that good at targeting, maybe programming up an adaptive 'catcher' robot would work? I'm thinking of something along the lines of 3 arms that have a range of motion, and when the rocket's within a few feet, they gently 'grab' the rocket using shaped and padded interfaces(I'm picturing a semi-circle matching that of the rocket) and provide stability.

Hysteresis

[Whiteox]

"entered a form of hysteresis, a condition in which the control response lags behind changes in the effect causing it."
I had a girlfriend with that condition.

Re:Hysteresis

[electrosoccertux]

"entered a form of hysteresis, a condition in which the control response lags behind changes in the effect causing it."
I had a girlfriend with that condition.

haha

Understatement of the year...

[Macman408]

"Rocket landed on droneship, but too hard for survival."
https://twitter.com/elonmusk/s...

Too hard for survival? By which you really mean 'it went SPLOOIE in an impressive fireball'? ;-)

In other news, a small amount of smoke was reported aboard the Hindenburg.

Re:Understatement of the year...

[cbhacking]

Not clear whether or not you realize this, so to give you the benefit of a doubt:

A) In orbital launch booster rocketry terms, that's a pretty damn small fireball.
B) At the time he sent that tweet, nobody (Musk included) had seen the video from the barge. I doubt he even had the video from the airplane, probably just the telemetry data that showed touchdown followed by a loss of telemetry.

Re:Understatement of the year...

[Macman408]

No, I agree on both points. But when I watched the video from the airplane a few hours after seeing his tweet, with it still ringing in my head, I literally LOLed.

It doesn't work that way.

[DerekLyons]

You're not getting what he was saying.

I do get what he's saying. What neither of you seem to grasp is that the size of the target isn't as relevant as you think, because you have to null your horizontal velocity regardless of the size of the target. It doesn't matter whether you're stopping on a postage stamp or anywhere in a given block - either way you still have to stop. It's the stopping that's problem, not the deciding where to stop. Stopping is very difficult for the Falcon 9 because it's T/W ratio is so far out of the optimal range and a larger target area won't make it all that much easier.
 

Though if they're that good at targeting, maybe programming up an adaptive 'catcher' robot would work? I'm thinking of something along the lines of 3 arms that have a range of motion, and when the rocket's within a few feet, they gently 'grab' the rocket using shaped and padded interfaces(I'm picturing a semi-circle matching that of the rocket) and provide stability.

That's "good at targeting" a couple of orders of magnitude better than what they've demonstrated to date (which is, pardon my french, already pretty fucking amazing). You're talking about some kind of articulated arm (which can survive being essentially inside rocket exhaust)... Which is, quite frankly, makes things much harder and more complicated and introduces a metric buttload of additional possible points of failure. Much easier to simply re-engineer the throttle valve.

Re:It doesn't work that way.

[itzly]

What neither of you seem to grasp is that the size of the target isn't as relevant as you think, because you have to null your horizontal velocity regardless of the size of the target. It doesn't matter whether you're stopping on a postage stamp or anywhere in a given block - either way you still have to stop.

The problem of having to null both velocity and position at the same time is harder than just having to null velocity. The latter gives you a bigger solution space.

Re:It doesn't work that way.

[Firethorn]

What neither of you seem to grasp is that the size of the target isn't as relevant as you think, because you have to null your horizontal velocity regardless of the size of the target.

As itzly, says, you can say 'null your horizontal velocity' all you like, but we agree with you there - what we're saying is that arranging for velocity AND position to be 'null' at the same time is harder than simply arranging for velocity to be null and position to be +/- 100m(or so).

You're talking about some kind of articulated arm (which can survive being essentially inside rocket exhaust)

I think you're picturing something different. I'm picturing something pretty big that comes in from the sides, staying well away from the exhaust.

Re:It doesn't work that way.

[bbn]

The engine can throttle between 50% and 100%. It is true that even 50% of just one of the nine engines is enough to shoot the rocket back towards space. Which is why it can not hover.

But hover is not relevant. Hitting zero velocity at height zero is.

Let the rocket drop freely while continuously calculating needed thrust to hit the target velocity zero at zero height. Keep dropping until your algorithm says you need 75% thrust. At this moment you relight your engine at 75%.

Now you can do a control loop with feedback to vary the throttle between 50% and 100% with the perfect descent hitting 75%. This should be plenty to do the job.

In addition you have at least two other mechanism to help a successful landing. One is the braking by aerodynamic mechanisms that can be varied. Those might be more precise than engine throttling. Another is the landing legs. They will be designed to absorb hitting the ground at a small non zero velocity.

Re:It doesn't work that way.

[DerekLyons]

what we're saying is that arranging for velocity AND position to be 'null' at the same time is harder than simply arranging for velocity to be null and position to be +/- 100m(or so).

*Sigh*
 
I understand what you're saying - but as with my previously reply, you don't grasp the problem.
 
The appearance of the vehicle "working hard at the last second" during the first attempt was a consequence of running out of hydraulic fluid - and would have occurred regardless of the size of the target. The appearance of the vehicle "working hard at the last second" during the last attempt was a consequence of the throttle valve not operating to spec - and would have occurred regardless of the size of the target.
 
From the point of view of the final landing sequence it's not all that much easier to arrange for velocity to be null and position to be +/- 100m than it is to arrange the same +/- 1m. Selecting a landing point occurs at a relatively high altitude (and on a, relatively speaking, relaxed timeline) and final trim starts around a kilometer or so up (AIUI). From there, jittering the variables (burn time and timing, gimbal angles, and throttle settings) a tiny amount one way or another to maintain targeting isn't a substantial burden (on the software or the hardware) compared to the much larger problem of nulling your velocities.

You're talking about some kind of articulated arm (which can survive being essentially inside rocket exhaust)

I think you're picturing something different. I'm picturing something pretty big that comes in from the sides, staying well away from the exhaust.

That just makes an already heavy, complex, and expensive system even heavier, more complex, and more expensive than I envisioned.

Video shows the thruster control lag ?

[GlowingCat]

It looks like the video shows the lag of the rocket thrusts reaction. First the thrust is vectored right and it stays right even after the tipping point. And after being several seconds late it vectors to left, causing the excess force.

Almost there!

[msevior]

That video is really impressive. It's damn hard to hard a rocket on Earth and it looks like SpaceX has almost done it in a just a few iterations of their design. My *guess* is that the next attempt will succeed.

Once they do, the cost to put stuff in orbit will drop by an order of magnitude.

Re:Alighting on land

[Rei]

An empty stage with no payload gets about 1 1/2 orders of magnitude more delta-V for its last kilogram of fuel than it got for its first kilogram of fuel when the countdown hit zero at the pad. And on top of that it's already got altitude, and can use the atmosphere to shed unwanted lateral momentum or aerodynamically redirect it to change direction, with little consumption of fuel. Its these things that make flyback a lot easier than it sounds at first glance.

Still not "easy", but a lot easier.

Re:Too easy man

[weilawei]

You've got to slow the top half down too. Inertia is still a thing. If you suddenly stop the bottom but the top keeps going, you get half a stage. I'm not sure the body is built to take a load from that direction.