A very nasty phantom forces bug

[Sharpy]

1.1.3, Windows, lots of parts mods but nothing affecting the physics or craft construction - seems like purely a stock behavior.

Just posting a screenshot, not .craft, because it's so trivial building it from scratch will take you 5% the time downloading the .craft, putting it in the folder and loading into VAB would take.

One quad adapter as root.

24 Mk1 LF tanks, 4x symmetry.

RUD on launchpad several seconds since launch.

 

I deactivated Kerbal Joint Reinforcement for the above screenshots - it seriously aggravates the issue (only 3 tanks tall tower enters the fandango within 3 seconds, RUD within 5. Even plain 4 tanks on quad-adapter wobble.)

Can someone confirm this in pure stock, and in 1.2?

[kiwi1960]

This is why you need struts... while they appear to be "joined" in the VAB, they really are not. You notice this as well with space planes.

 

[Sharpy]

Just now, kiwi1960 said:

This is why you need struts... while they appear to be "joined" in the VAB, they really are not. You notice this as well with space planes.

 

They make it explode even faster.

[kiwi1960]

Then take care where you place them. I've been aware of this since the old days... others have as well. It can be strutted and there should be no explosions.

 

[Sharpy]

15 minutes ago, kiwi1960 said:

Then take care where you place them. I've been aware of this since the old days... others have as well. It can be strutted and there should be no explosions.

 

Okay then, so WHERE should the struts be placed?

I placed them on the end tanks. Doesn't explode but shivers like crazy and crawls all over the launchpad. Tried 4th tanks, same but worse.

Edited October 4, 2016 by Sharpy

[Guest]

1 strut placed at top with 4x symmetry stops this. It's a weakness in the game where the tanks are not colliding with each other because they are all attached to the same part and weird stuff happens. The strut stabilizes the stack.

http://imgur.com/a/ZauGI

edit : you will have to get the craft off the launchpad with clamps also. It tends to dance around on the pad otherwise. I don't know whether making a bug report will help or not, couldn't hurt. At worst it'll get ignored or shelved until someone has time to look into it.

Edited October 4, 2016 by Otis

[Sharpy]

@Otis

1. LF fuselage

2. No launch clamps smartass

3. Try 2x phys-warp.

Strutted where you've shown.

 

 

[Guest]

17 minutes ago, Sharpy said:

@Otis

1. LF fuselage

2. No launch clamps smartass

3. Try 2x phys-warp.

Strutted where you've shown.

 

 

While I am known as one of the biggest smartasses in my circle of friends, I wasn't trying to be one in this case. If you put the stack on clamps without the struts, it will still tear itself apart. The clamps only serve to show that the stack can be stabilized once they are off the pad. Like I said, filing a bug report may help, it may not. But one doesn't know until they try.

[Sharpy]

@Otis

Primary question: does it still happen in 1.2? If not, no point filing the bug.

Launch clamps can often stabilize quite insane instabilities - they are literally "immovable", totally overriding physics when it comes to moving them from a spot, although they hold their "targets" only using standard (albeit strong) physics joints.

The problem is you can't take them into space... won't work in orbit.

[Guest]

7 hours ago, Sharpy said:

@Otis

Primary question: does it still happen in 1.2? If not, no point filing the bug.

Launch clamps can often stabilize quite insane instabilities - they are literally "immovable", totally overriding physics when it comes to moving them from a spot, although they hold their "targets" only using standard (albeit strong) physics joints.

The problem is you can't take them into space... won't work in orbit.

It does exist in latest build. Right, you cant take the clamps to space(on purpose, there have been bugs in the past where this is possible), but, I believe if you try using this configuration on a rocket, it will be ok once it gets off the pad. I believe the pad is making the problem worse, and consequently, this probably needs to be the focus of any bug report.

edit : Sharpy is right. The 4 way coupler is bugged. More attention needs to be focused on this thing if it's gonna be in the game.

Edited October 4, 2016 by Otis

[kiwi1960]

Then use the proper area and submit a bug report.

As stated, it couldn't hurt... they will either tell you what we did... or investigate.

No point going round and round in circles... we offer help, you go on and on about us being wrong...

So.... lodge a bug report. This thread should be locked.

 

[DrLicor]

@Sharpy, have been following you since our encounter yesterday, please to something about the way you post things here. 

As @kiwi1960 says, you go on and on.. On the forum we are helping each other, so please, show some respect and be gratefull when someone takes (his) time trying to help you. Don't act like only your opinion/observations are right. 

Edited October 4, 2016 by DrLicor

[Sharpy]

13 minutes ago, kiwi1960 said:

Then use the proper area and submit a bug report.

As stated, it couldn't hurt... they will either tell you what we did... or investigate.

No point going round and round in circles... we offer help, you go on and on about us being wrong...

So.... lodge a bug report. This thread should be locked.

 

Once again: before submitting a bug report I wanted to know if it makes sense. It could hurt, sending someone on a wild goose chase after a bug not present in 1.2, that's why I wanted someone to verify.

This is what I requested: someone to verify if the bug is present in 1.2. Instead of checking that and telling, you went about struts. I'd acknowledge the bug is entirely moot point if it really can be nullified by struts and go about verifying your claim. And find that hell, no, struts modify the behavior of the bug but don't fix it.

Look. I'm a software developer, and I know the value of a well-documented, well-investigated bug report, in particular one with a minimized testcase - a subassembly of 5 parts instead of a craft of 800 exhibiting the problem. Simultaneously, I know the headaches of bad bug reports. Like, ones that report bugs that are already fixed in the newest version. These really send my heart racing: "What?! So my fix didn't work? That bug is STILL present?" - and then after a flurry of preliminary work I find out the user didn't upgrade.

Yes, now that I have my question answered (bug still present in stock 1.2) I will submit it and the topic can be closed.  I know you just wanted to help me... but really, 'moar struts' is not the correct answer to the questions I asked.

 

[steve_v]

7 hours ago, Sharpy said:

Can someone confirm this in pure stock, and in 1.2?

Okay, I'll bite: Sure can confirm it, in stock 1.2.1564. probably the same as #12336. occurs with pretty much any combination of multi-coupler and tanks.
And no, struts are not the answer. Initially they appeared to fix it, but closer examination reveals that they simply constrain the phantom vibration. In physical warp this vibration tends to cause explosions.
This is clearly a physics bug, connecting tanks to a multi-coupler is very logical... having it turn into a whirligig or vibrate it's way across the launchpad is not.

[kiwi1960]

And we keep on saying the same thing... its not really a bug... its a game physics bug that has been around sine the very early days....

Its like in real life, you stack 20 barrels one on top of each other and then try to walk with them... the same thing happens in the game.

In your original photo, the strong point is the bottom... but each tank you place on top will be "lose" ... you need to strut them or work it so they are connected in some way with the strong point, in this case,t he tricoupler.

So... again.... its not a bug.... and chances are it WILL be present in 1.2 and all future versions till they decide to expend effort and fix it... but seeing as everyone knows about it and uses work arounds... its not a big deal.

Over and out.

 

[steve_v]

20 minutes ago, kiwi1960 said:

its not really a bug

If this unrealistic self-amplifying oscillation of anything attached to tricouplers/quadcouplers is intended, then sure, it's not a bug.
Otherwise it's unintended behaviour, and by definition, a bug.

 

22 minutes ago, kiwi1960 said:

Its like in real life, you stack 20 barrels one on top of each other and then try to walk with them... the same thing happens in the game.

In real life, if you make a stack of 20 barrels and leave them alone, they don't start to gyrate wildly. There's no walking involved in-game, and no external forces except gravity.
Besides, this stack is only 6 barrels high and I can reproduce it with 4.

In real life, if you make a stack of 4x3 barrels and tie them together at the top, the stack does not start to vibrate violently all by itself.
An unlashed  stack might sway, hell it might even fall over... but it certainly won't start spinning around at ever increasing speed. Unless a physics bug is involved.

18 minutes ago, kiwi1960 said:

In your original photo, the strong point is the bottom...

Uh huh... and from the top: http://bugs.kerbalspaceprogram.com/attachments/download/24024/BoosterDance.ogv

Yes, it's been around for ages, but that does not make it "not really a bug". It just makes it a really old bug.

27 minutes ago, kiwi1960 said:

its not a big deal.

To you perhaps. To me it's one of the main reasons I never use the tri/quad couplers. They almost always generate phantom forces.

[Sharpy]

@steve_v: I loaded the .mu into Blender, and it appears each of the multi-couplers has two identical collider grids (besides the standard 'draw' grid object). If that means the couplers really have two colliders each, identical and perfectly overlaid, that would likely be the source of the bug. Would it be possible to verify if that's the case?

[Green Baron]

It is present in 1.2pre stock, unless i strut the tanks on top with 4 struts.

Could be a floating point fluctuation, added up through the tank stack, thus leading to that kind of "resonance catastrophy". In terms of my understanding of game physics the rocket-tree is "upside down". Also the tanks have high mass, any dampening force is seemingly not enough. Empty tanks are stable.

For my playstyle i can ignore this behaviour because i rarely put tanks on top of multiadapters. It does not happen whith science containers on top of adapters or when the tanks are empty. I often put stacks of probes on multi-adapters and the problem never occured.

4 Struts solve/surpress the problem. Like kiwi i would say: reminds me of real life ... :-)

Edit: like the bug report suggests, it's not critical ...

Edited October 4, 2016 by Green Baron

[steve_v]

3 minutes ago, Sharpy said:

Would it be possible to verify if that's the case?

Excellent question... Dunno.

Might have a poke around tomorrow, though I'm not really familiar with part modeling for KSP, so I can't say what effect 2 colliders would have.
Should be easy enough to test though, if you can modify the model to remove one...

I'm also curious as to whether non-Squad multi-couplers do this too. Might try to find some in the wild, again, tomorrow.

[steve_v]

37 minutes ago, Green Baron said:

4 Struts solve/surpress the problem.

No, no it does not. I presume you've seen the bug report I linked, watch the second video I uploaded there.
4 struts only reduces the severity of the problem by constraining the unexpected force - wild flailing becomes strong vibration. Still a bug.

37 minutes ago, Green Baron said:

Edit: like the bug report suggests, it's not critical ...

It doesn't qualify for the "critical" priority because it doesn't render the game unplayable... this is not the same as saying it shouldn't be fixed.

Edited October 4, 2016 by steve_v

[Guest]

So, I though it was the pad that was causing most of the problem. Now, I'm not so sure. I took the strut off then cheated the thing into orbit and it started to shake around. So, I did a quick 5x time warp and it stabilized. Then it started to shake around again. Then started to spin. It gets worse and worse until it comes apart. Of course , once it comes apart, tanks collide and explode. So, there is more going on. Even if one struts the thing into submission, there are still phantom forces being generated, just not as noticeable. I'm sure it would generate other problems given enough time. I wouldn't use these adapters unless this gets sorted. I don't ever use them anyway, just not part of my bag of tricks. Anyway, yeah, I think more attention is needed if these adapters are going to be part of the game. Sharpy is right.

Edited October 4, 2016 by Otis

[Sharpy]

@steve_v: Okay, verified it, not a 'dual collider' problem. I loaded the .mu into Blender, exported .obj files for 'visual' and 'collider', then followed the fully kosher procedure of creating the part from scratch (and Blender models) in Unity. I used the same .cfg, only with names changed.

My part generates the wobble exactly as much as stock. And it does have only one collider, verified by loading its .mu into Blender. (so the multi-couplers still may have the colliders doubled, but that's merely a very mild performance issue, not the source of this bug.)

[kiwi1960]

Go to YOUtube and enter "KSP wobble" into the search... while most don't exactly relate to your definition of the problem, you can see that its been around since before...well.. for ever.

Earliest video I found was for version 0.18 and it seriously had me wanting to smash my monitor... the good old fun days... but compared to now... those days are gone THANK GOODNESS!

 

[Sharpy]

@steve_v: It's in the geometry and the mass of the binding part.

Two or more "prongs" attached to a common, light connector cause this behavior.

Same setup, except the left one has the tank and the triconnector welded together through Ubiozur's Welding. And it's perfectly stable, actively stabilizing itself against mad jerking coming from the other one.

Won't happen if the "prongs" point away from each other. Won't happen if there's just one, and is seriously weaker with just two. Won't happen if the tanks aren't attached by the ends. And doesn't happen if the connector part is heavy. I tried with a structural plate, same effect. I tried with radially attaching to another tank. Happened with that tank empty, didn't with full.

At least I've got a good set of workarounds.

[sal_vager]

This isn't particularly new unfortunately, the oldest report I've found is this one, but this issue goes back much much further.

This only affects parts with Physicsless = 0, ie they have Physics enabled (yes I am aware that is backwards).

Depending on the craft it is possible to reduce or eliminate this using struts, for the OP's craft a strut (or autostrut) from each of the first (lowest) tanks in the stack to the quadcoupler will reduce this effect to a minimum.

The worst thing about this though is that it is affected by mass, and a previously stable craft can start to oscillate after fuel is used or new parts are docked.

A fix has not been found, and as part joints are handled by PhysX it might require a modified PhysX or another physics engine.