Strut-Connectors to Boosters take Rocket off-balance, why ?

[Dieter_G]

Hi there,

I found something strange to which I have no explanation :

I build a (Test-)Rocket using two SRBs mounted on radial decouplers. To prevent the Boosters from skewing I added one strut-connector for each booster (see images). The problem now is, that the version WITH strut-connectors will lean to one side during flight and wants to tilt while the version WITHOUT strut-conns won't.

I now try to figure out and understand why this is the case. Please note that my primary interest ist NOT to find a solution (like using two struts on each booster), but to understand why it is the case.

Before answering, please consider the following :

-The struts I use are perfectly aligned. They are exactly perpendicular to the fuel-Tank and do point exactly to the center of the each booster. I did this by adjusting the .craft file.

Thought I could have made a mistake symetrising the struts, the behaivior was the same with normally placed struts (as good as can be done by eye).

-The two Rockets are identical (apart from the struts) . I "generated" the Rocket without struts by removing them from the one with struts, so everything else should be unchanged.

-The struts work by tension, not by pressure, so they should self-align.

Previously I had the struts at the top, where they work by pressure. The effect was the same but there was an explanation for that : Because they are just a point-to-point connection they cannot

absorb torque. By "pressing" at the fuel-tank from both sides, they made the tank to elude to one side (you know what I mean?)

Here are the images (3 per Rocket) : The upper ones is with struts, tilting to the left during flight; The lower ones is without struts, going straight up

Edited September 14, 2014 by Dieter_G

[Dieter_G]

I could up my .craft file if needed, if someone tells me how

[EdFred]

I have had the same exact rocket tilt on separate launches - it happens.

[Dieter_G]

So you mean it was just a coincidence ?

No, I tried several times: always the same.

[MarvinKitFox]

The struts have a teensy bit of "give" on them, before they kick in and hold position.

Using *any* sort of strutted booster-on-radial connector will lead to either leaning, as your does, or a tendency to spiral.

What you can do though is the minimize the effect a LOT.

Use two struts for each booster.

Don't have the struts run straight to the tank, rather have then splayed out. At least 30 degrees, more is better.

This will vastly reduce the wiggle room.

As for myself, I typically use 3 struts per booster. 2 at bottom, angled to the side

1 at top, angled close to vertical.

p.s.

You can get away with just the one strut.

Move your radial mount position waaaaay up on the booster. right at the top!

Then your single strut at the bottom will be 3-4 times as effective, due to increased moment arm of leverage

[SRV Ron]

To prevent that issue, strut in pairs top and, or bottom. Or, for a ring of SRBs strut in a ring top and bottom.

Single strut from the top will also work if centered to the SRB.

Edited September 14, 2014 by SRV Ron

[Dieter_G]

Ok, thanks for your replies

p.s.

You can get away with just the one strut.

Move your radial mount position waaaaay up on the booster. right at the top!

Then your single strut at the bottom will be 3-4 times as effective, due to increased moment arm of leverage

I'll try that

The struts have a teensy bit of "give" on them, before they kick in and hold position.

Using *any* sort of strutted booster-on-radial connector will lead to either leaning, as your does, or a tendency to spiral.

Yes I got that. At least I'm not the only one having this problem. Its logical for a strut at the top but...

however I still do not understand the reason of it as far as a strut >>>at the bottom<<< is concerned. As I said, it works on tension. It could be a rope with a lot of "give", thought. Still it shouldn't induce a tendency to "lean" to one side. For me this opposes common logic.

Also, there is no effect without a strut. So the Booster itself doesn't want to "lean" to one side. The behaviour is induced by the strut.

So I'm looking for a better understanding of the (simulated) physics. Are there people from the team on the board ? Maybe one of them knows ?

Sorry for beeing so sticky.

- - - Updated - - -

@SRV Ron

Your "interplanetary lander" is looking interesting, especially it's price

I'll keep the thing with the nearly vertical struts at the top in mind. Thank you both for that

Edited September 14, 2014 by Dieter_G

[Dieter_G]

update :

I've put the whole rocket on a giantic lifter, capable of delivering >50 tons to space, and voila : the problem persists. It also wants to flip in space. So it has nothing to do with aerodynamics.

[SRV Ron]

@SRV Ron

Your "interplanetary lander" is looking interesting, especially it's price

I'll keep the thing with the nearly vertical struts at the top in mind. Thank you both for that

The ship was designed for the Duna, Ike, Eve exploration contract.The SRBs are turned down some for best performance during launch. If you fly this for direct intercept of Duna or Eve and use aerobraking to get into orbit, you will have more then enough fuel left for a controlled landing using parachutes.

The smaller cheaper design works well for the Mun, Minmus contract and can return to orbit to do additional data request contracts.

My probe tipped over on Mun due to landing on a slope. However, it remained undamaged and was able to fulfill its contract.

Edited September 15, 2014 by SRV Ron

[DrMonte]

Putting struts uncentered and added engine torgue force will rotate the craft. That's quite clear.

Perfectly centered connection points does not exist in reality and even in KSP. "Perfect" might get the other meaning in computer language - like division by zero. I'm not saying that I know what is really happening but I think that you will not be able to put a perfect snap point even by editing the save file.

Following idea of imperfectness we will always need some additional force to compensate ship imbalance. Like SAS, winglets or engine gimbal.

[Dieter_G]

@DrMonte

Of course, even my symetrised struts aren't totally perfect. But the effekt I'm speaking about is strong, and it manifests on a lot of different rockets, some build normally, others symetrised by hand, its always the same and it always vanishes, as soon as the struts are removed. Of course, even without struts the rocket won't go perfectly straight up, the SAS always needs to correct a little bit because of imperfections of the construction. But there is a big difference in 1.) the SAS doing some corrections 2.) the SAS running constantly and still not beeing able to hold the rocket straight.

So, I want to stress again that it is not a question of some minor imperfections but instead it is related to the struts. Just remove the struts (leave the rest the same) and it is gone.

Damn, I'd like to up my Rocket so you can try yourself.

I also did a tri-symetrical design. Its even worse then : at the end of the burning-phase of the boosters the rocket flips over.

[MarvinKitFox]

however I still do not understand the reason of it as far as a strut >>>at the bottom<<< is concerned. As I said, it works on tension. It could be a rope with a lot of "give", thought. Still it shouldn't induce a tendency to "lean" to one side. For me this opposes common logic.

Its not the strut that wobbles, but the radial connector.

The connector resists compression very well, but is weak against torsion.

With no strut, the torsion is severe, but all as a bending moment upwards. i.e. the booster leans inwards, swiveling about the radial connector. It stops bending in only when it leans against the main tank.

*with* a strut at the bottom, it is completely prevented from leaning in, as the strut refuses to stretch and radial connector refuses to compress. BUT the radial connector allows a wobble to the side, satisfying the booster's craving for a lean.

It thus leans to the left or right of its decoupler attachment. Causing your whole rocket to either angle off by about 1-2 degrees, or if opposing boosters lean in the same way(seen radially), it will spiral.

I do hope this description is graphic enough. You do **NOT** want me to start using my MsPaint skills to illustrate. Really, you do not!

[Kasuha]

Both the decoupler and the strut are sticks to the physics engine. Don't get fooled by their model, even the decoupler has just one point by which it is connected to the central tank, and one point (not necessarily directly against the first point, depends how exactly you attach it) by which the SRB is connected to it. 3D geometry of parts is only applied for collisions between different ships, not between parts of single ship. So the ship itself behaves like if it was built from rigid sticks with springs at each connection point between two sticks.

So the central tank, the decoupler, the SRB and the strut together form kind of a pantograph device and it definitely has room to move even in up/down direction.

If you look into the .craft file, you can find exact coordinates of each of these connection points. And you may notice that they're not integer numbers, they have a long row of decimals each. However well aligned they look on the ship, they're not aligned perfectly. When the thrust is applied, initially small deviation further increases by the applied force (kind of positive feedback until it is dampened by stiffness of the connection).

[Dieter_G]

@Marvin, Kashua

Ok, I got you. I had the same thoughts already so I perfectly understand, you don't need to apply your MSpint-skills, Marvin

I also tried to approve it visually but couldn't see it (I watched the Rocket from straight above, cam in "chase" mode but couldn't see the tank move)

Also, I tried your suggestions for improvements of the design, but unfortunately none of them worked really well. I tried two angled struts (see Image), I tried two nearly vertical struts (thought I don't understand how they should improve things), I tried putting the decoupler to the very top and have a single strut at the bottom (suggestion from Marvins first post I think) and also having the boosters connected to each other, at the top with a single strut (no connection to central part)

because the improvement -using two angled struts- also didn't work I discarded the idea you are describing (that the reason of the observed bahaviour is that the decouplers are modeled as sticks by the physics engine and have a lot of "give" to the side). Ah, as you can see I've put the angled struts at the top, where they work on pressure an stiffness against [moving to the side] is neccessary, so that the decouplers (which are sticks) only have tension applied. But as I said : no success.

So, I'm still sceptical, especially since I discouvered another -definite- bug, which may be related. More on that tomorow... I'm very tired now and writing to this forum puts a heavy load on my brain since english isn't my native language.

cu

[Dieter_G]

It stops bending in only when it leans against the main tank.

Thats not correct, the radial decouplers do absorb torque. In my example (without struts) the boosters don't touch the tank, so bending is stopped due to the torque beeing absorbed by the decouplers.

@Kasuha : I think the decouplers are modeled as a >>set of<< sticks, otherwise they couln't absorb torque, right ?

Anyway, you're right that the decouplers leave a lot of room for movement (up/down and left/right, I studied it today) and this could explain why the tanks gets forced to one side even with struts at the bottom.

But as I said, even thought your explanations make perfect sense, I'm not sure if this is the (only) reason of the oberved behaviour, as the angled struts didn't alleviate it.

Btw.: I forgot to say, thanks for reading my post carefully and picking up the point with the ">>>struts at the bottom<<<", both marvin and kasuha.

[Kasuha]

@Kasuha : I think the decouplers are modeled as a >>set of<< sticks, otherwise they couln't absorb torque, right ?

I think the most correct image of how parts of the same ship behave is this:

Each part has a center (CoM) where all mass of the part is located

Each part has a set of points by which it is attached to other parts. Some of these points are predefined (axial points), some are ad-hoc set up during build (radial points). These points are rigid relative to the part center. You can imagine these points floating in space around CoM, or connected to the CoM via rigid sticks. Most important point is that this structure is 100% rigid and does not shear, bend or twist under any kind of stress.

Parts are connected by their attach points. Initially, attach point of one part exactly touches attach point of the other part. These two points are held together by a "spring" that allows some displacement, bending, or twist in the connection, and applies counteracting force proportional to amount of displacement, bending, or twist already in place. That makes the connection flexible while parts are rigid.

A decoupler is exactly this case - a part that has two attach points, one touching corresponding ad-hoc radial attach point of the center tank, the other touching corresponding ad-hoc radial attach point of the booster.

A strut may be (and probably is) a different case. Here, the strut itself may be the "spring". As well as connection between two normal parts is trying to hold attach points of two parts on top of each other, the strut is trying to hold two distant attach points in their exact relative distance and position. And the distance may have some effect on behavior and "springiness" of the connection.

Edited September 16, 2014 by Kasuha

[Dieter_G]

But now, as I "promised" yesterday, another one.

Look at that :

Got it ?

The engines are OFF after the second image, speed is negligible. Where does the energy come from, causing the wobbling of the tanks to increase steadily until the rocket decomposes (long since the engines where shut down) ? Same happens outside the atmosphere btw.

Try yourself :

capsule

|

tri-coupler

| | |

800 800 800

| | |

800 800 800

| | |

LV-T30 LV-T30 LV-T30

where "800" = FLT-800 fuel-tank.

The answer to my question (where does energy come from) is likely : negative damping, maybe due to a sign-error. Thought it looks funny, these are things that are totally ruining my confidence that the physics are modeled correctly. I mean, this is not a bloated fancy design with hundrets of components, it is very basic instead. I don't understand how one can miss that (as a developer).

I'm now at a point where I cannot happily play around with different designs, examining how they perform, but instead are constantly afraid that e.g. adding one more strut to a working design might lead to unforseeable changes in its behavior. That is annoying and kills most of the fun.

Ok, complained enought for the time beeing...

what do you think of this ?

Edited September 16, 2014 by Dieter_G

[Dieter_G]

I just realise that my "ASCII-arts" illustrating how the rocket should be build is constantly corrupted by the -oh so smart- editor, removing these unneccessary leading blanks in front of each line. Really clever !

I hope you can nevertheless figure out how it is meant...

[Kasuha]

Oh, these are peculiarities of the physics engine.

You could try the self-destructive ship, all you need is to deploy it on launchpad:

- go to the VAB and start with the standard command pod

- attach the long I-beam on top of it

- switch to 4x symmetry and attach four horizontal long I-beams to the top of that first I-beam

- add small fuel tank at the end of each I-beam

Edit: I would also suggest you to start using Imgur albums. They have nice support here on forums, take less space and lower the page loading time.

Edited September 16, 2014 by Kasuha

[Dieter_G]

"peculiarities" is a very nice word for a bug, I must admit

which are the "long I-beams" ? I think I don't have researched them up to now. But I'll try it, thanks.

[Dieter_G]

Hmmm, maybe there is some bug enabling a ship that doesn't consume fuel ?

Just kidding...

[Dieter_G]

ok, I saw your update with the image. I'll try it

I'll also check out these thing with the albums, thanks

[Kasuha]

"peculiarities" is a very nice word for a bug, I must admit

which are the "long I-beams" ? I think I don't have researched them up to now. But I'll try it, thanks.

For experiments and research, it is good to have a sandbox save. Then you don't have to be concerned about your cosmic program even if you completely break something.

[Dieter_G]

Important : i forgot to say, that this "peculiarity" was/is not present in the demo !

There this design behaves normally

[Kasuha]

Demo has other "peculiarities", and the part connection system was completely rewritten since when the demo was released.

Seriously, these "issues" or "bugs" are absolutely negligible in 99% of cases. The remaining 1% are cases when your space station starts resonating and either self-destructs or at least destroys its solar arrays. I don't see them fixed in the game anytime soon, there are other, much more important things to do and fix.

Edited September 16, 2014 by Kasuha