Orbit decaying for no reason?

[Kobymaru]

Hi! I have recently taken to the task of following my debris, placed in a graveyard orbit to a spiral to the ground.

One piece, consisting of solely a TR-18D stack separator, however doesn't even seem to need any friction to spiral downwards. At an altitude of 76.000m, the apoapsis seems to decay 1 m per 5 s.

I can post screenshots if you want, but they'd be pretty boring.

So what's going on here? Are TR-18D Stack Separators the new Quantum Drives?

[LostOblivion]

During physics (that is, not on rails) most spacecraft seem to have some form of changing orbits.

[Alshain]

During physics (that is, not on rails) most spacecraft seem to have some form of changing orbits.

No, not really. Not with patched conics. It would take thrust to lose a meter every 5 seconds.

Now, sometimes due to floating point math the game can't decide on an Apoapsis, but that would be inconsistent rather than a steady decline.

[asheft]

I think I've seen this before also. I haven't played in a while though, and am not at my KSP computer right now to check so I can't be sure.

[marce]

Phantom forces, a common problem. Kind of an unintended kraken drive with various levels of thrust.

The more complex (part intensive) a craft is the worse it gets. It even happened to me that the craft accelerates so fast it leaves the Sun SOI.

[Crown]

[...]The more complex (part intensive) a craft is the worse it gets.[...]

But a single TR-18D can't be considered 'complex'.

Either Squad implemented some kind of 'high atmosphere' or it's a problem with the orbit calculation.

[Alshain]

But a single TR-18D can't be considered 'complex'.

Either Squad implemented some kind of 'high atmosphere' or it's a problem with the orbit calculation.

It's more common with more parts, that doesn't mean it can't happen with 1. I doubt it is a problem with the orbital calculation, and I'm certain they didn't implement N-body physics.

[wossname]

Hmm, you didn't have Swigert Kerman stir the O₂ tanks did you?

[Kobymaru]

Ok, apparently this is reproducible. I used a completely stock install, except for hyperedit.

Hyperedit a TR-18D into Orbit of 75.000x75.000 km, and now I can watch the periapsis/apoapsis move, the periapsis lowering and the apoapsis rising. Sometimes it's the other way around, sometimes both are lowering.

I have not yet figured out the direction of the acceleration or the magnitude. Can somebody point me to simple-enough experiments to figure that out?

[Vegetal]

That happened to my ship once orbiting jool. The game didn't let me time warp, saying the craft was under acceleration. Had to physical time warp until the force disappeared, and corrected the orbit.

Had the impression it's more likely to happen when low in the orbit, this time I was near the periapsis, and it was quite low (not inside the atmo by the way, ships get annihilated anyway).

[Kamuchi]

Phantom torque, something that should have never passed testing.

Go to the Mun with a 2 part craft, undock and watch the 2 crafts start to drift and rotate without any forces applied to it (engine or rcs).

This happens everywhere, more or less depending the body but the Mun is like kraken central with this feature.

[MalfunctionM1Ke]

I can post screenshots if you want, but they'd be pretty boring.

Footage is never boring. Please post them

[sal_vager]

Moving to Support.

Floating point errors add up over time, the further you are from the center of the game world the worse they get which the krakensbane sought to resolve by resetting your position and speed, and moving the universe around the player.

SAS was added to KSP to deal with floating point induced rotation as well as to provide flight stability, and improvements have been made to the accuracy of speed and trajectory calculations, see the readme/change notes.

Things are as good as they are going to be until Unity5, sorry :/

[Kerbart]

I get the impression that as long as you're in an atmospheric physics bubble, atmospheric drag continues to exist. I've seen the Kerbin atmosphere create drag while at an 85km altitude (very little drag but still). I suspect that the formula that calculates atmospheric pressure (based on an exponential curve) doesn't have a cutoff in place. Once you warp and put the craft on rails it disappears.

[sal_vager]

Kerbins atmosphere ends at 69,100m, though the music doesn't start till 70,000m you're in vacuum before then.

[LostOblivion]

Now, sometimes due to floating point math the game can't decide on an Apoapsis, but that would be inconsistent rather than a steady decline.

I guess that is what I am seeing.

- - - Updated - - -

Phantom torque, something that should have never passed testing.

This might not actually be a bug. It's called gravity-gradient torque and is something that is simulated explicitly in Orbiter, but is probably passively simulated in KSP because of its part system.

What happens is that in KSP each part has a mass and is individually influenced by gravity. Parts farther away from the major body experience less gravity, causing them to want to float away from parts that are closer to it. Also, parts in a spacecraft are obviously connected with a joint, so these movements all combine to cause an induced torque and eventually rotation of the overall spacecraft.

A test to demonstrate this is to connect two spacecraft together with a long truss and decoupler in between, ideally pointing away from the major body so there is a notable difference in orbital height between the two spacecraft. Set the decoupler to have no ejection force. After a little while, the spacecraft should begin to rotate according to gravity-gradient torque. Now kill the rotation (by time-warp, for example) and decouple the two spacecraft. The two spacecraft should now float completely still right next to each other, but after a little while, because of their slightly diffferent orbits, will eventually start floating away from eachother. This is the same thing that causes the rotation, only that they are no longer connected to each other.

[Red Iron Crown]

This might not actually be a bug. It's called gravity-gradient torque and is something that is simulated explicitly in Orbiter, but is probably passively simulated in KSP because of its part system.

What happens is that in KSP each part has a mass and is individually influenced by gravity. Parts farther away from the major body experience less gravity, causing them to want to float away from parts that are closer to it. Also, parts in a spacecraft are obviously connected with a joint, so these movements all combine to cause an induced torque and eventually rotation of the overall spacecraft.

A test to demonstrate this is to connect two spacecraft together with a long truss and decoupler in between, ideally pointing away from the major body so there is a notable difference in orbital height between the two spacecraft. Set the decoupler to have no ejection force. After a little while, the spacecraft should begin to rotate according to gravity-gradient torque. Now kill the rotation (by time-warp, for example) and decouple the two spacecraft. The two spacecraft should now float completely still right next to each other, but after a little while, because of their slightly diffferent orbits, will eventually start floating away from eachother. This is the same thing that causes the rotation, only that they are no longer connected to each other.

There are no gravity gradients in KSP, all parts of a vessel experience the exact same gravity (I believe calculated from the root part's altitude/position).

[sal_vager]

KSP still does some neat things though.

{clicky}

[LostOblivion]

Cool!

[Kobymaru]

Moving to Support.

Floating point errors add up over time, the further you are from the center of the game world the worse they get which the krakensbane sought to resolve by resetting your position and speed, and moving the universe around the player.

SAS was added to KSP to deal with floating point induced rotation as well as to provide flight stability, and improvements have been made to the accuracy of speed and trajectory calculations, see the readme/change notes.

Oh, that's kinda sad, because it makes me wonder how many trajectories in my KSP life have been messed up by this "behaviour".

Things are as good as they are going to be until Unity5, sorry :/

Well that's totally fine, if it actually will be fixed by Unity 5. But are there any indications it will be? Does it have better floating point math?

Oh btw, does the 64-Bit version actually have bigger floating point numbers? Bigger numbers could potentially increase accuracy.

[sal_vager]

Fixed? Unlikely.

Improved? Possible.

64bit does allow for more numbers after the decimal point, but whether or not those numbers are more precise depends on a lot of things.

[Claw]

I've also noticed there is a physics discontinuity at 100km altitude over Kerbin. Above that, the orbit and attitude doesn't seem to drift. So if you would like more stable orbits, I would aim to be above that altitude.

Cheers,

-Claw