Station Stability

[CosmosGamer99]

So, I am building a massive station.  10 parts at least.  So, I recently added a big solar panel module and the whole thing started shaking.  The docking port in question that is the only one moving is Bluedog Design Bureau Belle - Leo docking mechanism.  For docking mods I have Mechjeb and Dock Rotate.  If there are any other mods in CKAN that you know can fix this I could you please share.  Or any other way to fix it.

Edit: A few parts have the option to hold crew but I currently have no crew on board.

Edited July 25, 2019 by CosmosGamer99

[James Kerman]

Hello @CosmosGamer99.

I've moved your question to gameplay questions. Usually this shaking in stock is caused by autostruts (usually connected to heaviest part). Do any elements of your station use autostruts?

[bewing]

This is certainly an autostrut problem. Keep the number of autostruts on your craft below 5, if at all possible. Also, turn off the torque on all reaction wheels besides your first one. Turn off RCS, too.

[Mahnarch]

Oh.. No crew on board.

 

I was gonna say, "Turn off all Latin music".

Kerbals love to mambo!

[CosmosGamer99]

Thanks, turning off some reaction wheels helped!

 

[Serenity]

Alt+F12, Visualize Autostruts, Make a tree of grandparents.Also try putting an anchor to use one heavy maybe, with testing.Careful though when docking things heavier than the anchor.

Edited July 29, 2019 by Boyster

[magnemoe]

17 hours ago, CosmosGamer99 said:

Thanks, turning off some reaction wheels helped!

 

Having reaction wheels in both end of an long station or ship who is a bit wobbly tend to cause shaking. 
Also use the 2.5 meter docking ports for anything heavy or you will dock other stuff to. For radial ports use the 1.25 meter radial connector who looks like an failed docking port then put an 2.5 m port on it. 1.25 works well for light stuff like struts with solar panels. 

[Geschosskopf]

On 7/30/2019 at 10:58 AM, magnemoe said:

Having reaction wheels in both end of an long station or ship who is a bit wobbly tend to cause shaking. 

Are you sure about this?  This has certainly always been a common belief but at least up through 1.3 that belief was actually a misconception.  Reaction wheel torque (at least until 1.3) was applied at the vessel's CoM, regardless of the location or orientation of the reaction wheel parts.  Thus, the wobbliness of long structures with SAS turned on using multiple reaction wheels was simply the result of:

• The inherent flex in the joints allowing the structure to bend to begin with
• The moment of inertia of the long structure causing it to bend any time you tried to rotate it
• This bending being exacerbated by having multiple sets of reaction wheels providing too much angular acceleration to the CoM than really needed for a slight change in attitude, thus causing an overshoot, which SAS then attempted to correct but overshot again, and;
• The inertia of the ends of the structure flexing around can actually move the CoM, and thus the point where the torque is applied, resulting in off-axis torque causing further flexing and that cascading into more wobbles.

This is why turning off most reaction wheels helps stop the wobbles.  Moving long, flexible structures slowly and gently reduces the flexing and throwing the mass around than trying to move it quickly.  But it has nothing to do with where the reaction wheels are located.

At least that's how it always used to work.  Has something changed lately?

[Claw]

On 8/1/2019 at 12:57 PM, Geschosskopf said:

Reaction wheel torque (at least until 1.3) was applied at the vessel's CoM, regardless of the location or orientation of the reaction wheel parts.

This is not quite accurate. Or, at least, maybe a misinterpretation of what's happening.

 

The reaction wheels apply the torque at their location. However...torque is a funny thing in that, in the freefall of orbit, the reaction of the vessel occurs at the vessel's CoM. If anyone is interested, I have some old posts I made in threads discussing similar questions. The threads themselves are also full of good discussion.

 

 

 

Edit: And actually, this post here addresses the statement directly.

 

 

Edited August 2, 2019 by Claw

[Geschosskopf]

1 hour ago, Claw said:

This is not quite accurate. Or, at least, maybe a misinterpretation of what's happening.

The reaction wheels apply the torque at their location. However...torque is a funny thing in that, in the freefall of orbit, the reaction of the vessel occurs at the vessel's CoM. If anyone is interested, I have some old posts I made in threads discussing similar questions. The threads themselves are also full of good discussion.

And actually, this post here addresses the statement directly.

Well, it's been argued quite strongly both ways for years, with supporting evidence from experiments and hearsay statements from the devs on both sides.  Your posts above have neither, they're just repeating your position.  Be that as it may, when such evidence has been presented, still reasonable minds can disagree, and still do.  I myself, after much experimentation and talking with the devs during a beta long ago, firmly believe that the position and orientation of reaction wheels make no difference whatsoever to the flex of a station.  What matters is the ratio of inherent structural flexibility due to the parts used to the amount of torque applied.  But that was long ago.  Perhaps things have changed in the last year or so?

In any case, nobody disagrees that building a station with a high moment of inertia combined with many longitudinal joints (especially 1.25m docking ports) is bound to get the wobbles no matter what you do.  So the answer is to NOT build ISS look-alikes.  Keep your stations dense and compact, with as low a moment of inertia and as few longitudinal joints as possible..  Basically, make a "pancake station" just like we used to make "pancake rockets" before the atmosphere started caring about frontal area..  If you MUST have longitudinal joints, only use 2.5m docking ports.  KSP simply doesn't like long, skinny structures with multiple parts.  Or use KJR and never worry about wobbly ships or stations ever again.

Edited August 2, 2019 by Geschosskopf

[Snark]

On 8/1/2019 at 10:57 AM, Geschosskopf said:

Reaction wheel torque (at least until 1.3) was applied at the vessel's CoM, regardless of the location or orientation of the reaction wheel parts.

I believe that this is not the case.

Reaction wheel torque is absolutely applied at the reaction wheel itself, now, which is easily demonstrated by inserting a free-floating hinge in between the control point and the reaction torque:

...That's completely hands-off the controls, there.  It's just SAS that's doing that.  The hinges are set to unmotorized, no damping (i.e. freely flopping), limits of motion set to plus-minus 15 degrees from center.  All I did was launch to the pad and set SAS to hold .  It freaks out because it's trying to hold the probe core perfectly upright, when it's mounted on a floppy stalk at the opposite end from the reaction wheel.

So what you see above is basically the floppy-station problem, exaggerated and in miniature.  It's because KSP's SAS is implemented in a fairly straightforward, simple way that essentially assumes a rigid system.  When you have a floppy system, especially one where the source of torque is physically separated from the control point (and therefore there's a delay between applying torque and seeing a result)... this is what you get.  The cause-effect delay causes PID tuning problems.

Speaking as a KSP modder with a fair amount of familiarity with the game's internals (at least the ones accessible to modders, which is a lot), I'm pretty sure it's worked this way (i.e. "torque is applied at the part with the reaction wheels", not at CoM) since forever-- or at least since well before 1.3.

 

[Geschosskopf]

6 minutes ago, Snark said:

So what you see above is basically the floppy-station problem, exaggerated and in miniature.  It's because KSP's SAS is implemented in a fairly straightforward, simple way that essentially assumes a rigid system.  When you have a floppy system, especially one where the source of torque is physically separated from the control point (and therefore there's a delay between applying torque and seeing a result)... this is what you get.  The cause-effect delay causes PID tuning problems.

Speaking as a KSP modder with a fair amount of familiarity with the game's internals (at least the ones accessible to modders, which is a lot), I'm pretty sure it's worked this way (i.e. "torque is applied at the part with the reaction wheels", not at CoM) since forever-- or at least since well before 1.3.

Well, I certainly respect your position.  As I said, reasonable minds can differ.  But run the equations backwards.  How are they distinguishable from the torque being applied to the CoM and the ends just flopping due to moment of inertia applied to the especially flexible joint you put in the middle?  As far as I can tell, you get the same results either way but I have been known to drink so take that with a grain of salt

The real problem with your amusing GIF (which I've saved as a classic), however, is that it's on the launchpad instead of in space..  Does the same thing happen to the same assembly in space?  And if it does, how is that different from the reverse interpretation?

I have recently tried to tie 2 reaction wheels, mounted on opposite ends of a long string of docking ports,. to a control axis and reverse the direction of 1 of them.  Had this resulted in a banana-shaped station, I'd have bowed down.  Unfortunately, you can't do that sort of thing with reaction wheels and action groups or even the KAL1000.  Thus, the jury is still out and I remain unconvinced.

[Snark]

1 hour ago, Geschosskopf said:

How are they distinguishable from the torque being applied to the CoM and the ends just flopping due to moment of inertia applied to the especially flexible joint you put in the middle?  As far as I can tell, you get the same results either way but I have been known to drink so take that with a grain of salt

It's easily observable.  Don't take my word for it, set it up yourself.  There are plenty of ways you can set up a reaction wheel at the end of a loose-jointed contraption and it's easily visible that it's the reaction wheel itself that's twisting around.  For example, build a mobile with various reaction wheels dangling on "ropes" of floppy joints, and only the one that's turned on will flop around when you give it control input.

1 hour ago, Geschosskopf said:

The real problem with your amusing GIF (which I've saved as a classic), however, is that it's on the launchpad instead of in space..  Does the same thing happen to the same assembly in space?

Not really a problem-- the laws of physics are the same in both cases.  Yes, it does exactly the same thing in orbit, as one would expect.  And if you make it easier to observe by slowing it waaaaay down (e.g. by reducing the control authority on the wheel to 1%) and change its mode to "Pilot Only" and play with the pitch/yaw controls a bit, it's very visibly obvious that it's just the wheel itself that has any torque on it.

1 hour ago, Geschosskopf said:

I have recently tried to tie 2 reaction wheels, mounted on opposite ends of a long string of docking ports,. to a control axis and reverse the direction of 1 of them.  Had this resulted in a banana-shaped station, I'd have bowed down.  Unfortunately, you can't do that sort of thing with reaction wheels and action groups or even the KAL1000.

Yup, that's not gonna work, because the input to reaction wheels is global (whether from SAS or from player input); it's just their torque output that's applied at the part itself.  So that test's not doable in KSP.

However, you can build strings of floppy hinges to make flexible "ropes", which makes it pretty easy to see where the torque is.  My example is just a very simple one-- there are more complex constructions you can build that make it even more obvious.  (Example:  Have your probe core rigidly mounted to launch clamps, a fair distance above the launchpad.  Have a loose/floppy string of hinges hanging down, with a reaction wheel on the end.  Play around with the pitch/yaw controls while SAS is off.  It becomes very obvious that it's just the reaction wheel that's torquing itself around.  You can have multiple reaction wheels dangling like Christmas ornaments from multiple ropes, and try turning them off and on, and only the ones that are turned on will be affected.)

 

[Claw]

4 hours ago, Geschosskopf said:

supporting evidence from experiments and hearsay statements from the devs on both sides

What I said about where the torque is applied is not hearsay and not derived from experimentation. It comes from first hand knowledge about the code and how the unity engine works. Torque is applied at the reaction wheel, but (due to physics) the equal-and-opposite reaction of the vehicle occurs about the CoM when the vehicle is unconstrained (for example, in freefall).

 

I said nothing about flex in my statement, though it's referenced in the links.

 

4 hours ago, Geschosskopf said:

the position and orientation of reaction wheels make no difference whatsoever to the flex of a station

Orientation will matter if the torque is not symmetric about all three axes. If it's the same, then I agree that orientation doesn't matter.

Position does matter as to how the flex behaves, but is mostly masked within KSP's mechanics. So in essence, the position is much less of a factor than is the actual design of the station. 

 

4 hours ago, Geschosskopf said:

What matters is the ratio of inherent structural flexibility due to the parts used to the amount of torque applied.

Yes, we can agree on this even if you do not believe my assertions about where the torque is applied within KSP's mechanics.

Also, I didn't intent that my statement of "torque is not applied at the CoM" as me claiming it as the causal factor to the OP's problem with station wobble/flex. Aside from the torque@CoM statement, I agree with your other points about how the vessel behaves and ways to help minimize/prevent the issue.

 

Snark already explained why the "reverse reaction wheel" won't work. So here's another series of example pictures.

 

 

The first picture shows a basic setup and the location of the CoM.

Spoiler

 

The second image shows the result of a "turn left" input. The beam is clearly flexing, which it can't possibly do if the CoM is where it's at, and behind the launch clamps.

Spoiler

 

The third image shows another setup, which clamps down the CoM even further and has some markers on one side of the beam (to show twisting in the fourth picture).

Spoiler

 

And the fourth picture shows the beam being twisted. I'm not sure how this would be possible if the torque was being applied at the CoM.

Spoiler

 

This is not an experiment to hypothesize where the torque is applied, it's a model to demonstrate it being applied at the end of the beam.

 

Edited August 3, 2019 by Claw