STOCK solution to aerodynamic wobble?

[allmhuran]

The thick atmosphere and stock aerodynamics of KSP tend to cause a lot of aerodynamic "wobble". That is to say, after control input is halted an aircraft continues to pitch, yaw and roll under aerodynamic and inertial forces for a few seconds before finally returning to stability. Here's a 20 second demonstration of what I mean:

http://youtu.be/GMkmI5o4jFk

This effect can be mitigated fairly effectively by enabling SAS, but that also creates a very jagged and robotic response to control input.

I wonder whether anyone has found any effective means to eliminate or dampen this wobble using stock parts and with stock physics.

Edit: Answered. After much experimentation I found a design that significantly improves the flight characteristics but maintains KSP's extreme turn rates. Huge wing area (2 layers of wings), distributed mass, lots of vertical stabilizers and the secret ingredient (nosecones) does the job.

Edited August 2, 2014 by allmhuran

[MarvinKitFox]

That is perfectly normal for any vehicle.

Remember that you are inputting "joystick" commands, by slamming the joystick FULL left and FULL right.

Doesn't a car also wobble if you only steer by instantly turning MAX left or right?

Use better input controls, and the whole experience will be smoother.

[allmhuran]

Not sure what you mean by "better input controls" Marvin. Perhaps you want to take another look at the pitch/yaw/roll input markers on the bottom left.

Edited August 1, 2014 by allmhuran

[Geschosskopf]

This effect can be mitigated fairly effectively by enabling SAS, but that also creates a very jagged and robotic response to control input.

Hmm, I've never noticed "jagged and robotic responses to control input" when using SAS. Everything is all nice and smooth for me. Perhaps you're just getting low FPS?

[MarvinKitFox]

Not sure what you mean by "better input controls" Marvin. Perhaps you want to take another look at the pitch/yaw/roll input markers on the bottom left.

You are still using sharp transitions from input to non-input.

You are still nulling your controls while the plane is NOT in an aerodynamically neutral orientation.

*of course* it is going to wobble.

In your case, it is a design with strong positive stability, and the wobble rapidly tapers down to a neutral attitude.

With a bad design, that "wobble" manifests as a continuous off-straight veering, or even an accelerating deviation resulting in a flat spin.

.it.is.what.aeroplanes.do.

[allmhuran]

Perhaps you're just getting low FPS?

I just mean it's "too" stable with SAS. The aircraft sort of "locks" into place. It's hard to describe but if you compare to actual flight sim behaviour you'll see what I mean.

.it.is.what.aeroplanes.do.

LOL.

[Master Tao]

I mostly recreated that plane in my stock install. Most of the wobbles come when it's not pointing prograde, so my solution was to manually hold the attitude until the prograde caught up. That's much easier to do with analog input, but I was also able to do it using precision controls and sometimes tapping F when it was close. Both approaches seemed to minimize the jerkiness while stopping the instability.

[allmhuran]

Most of the wobbles come when it's not pointing prograde, so my solution was to manually hold the attitude until the prograde caught up.

Yep. The plane wobbles when input is zeroed, but not when input is first provided.

In other words (contrary to Marvin's opinion) this is nothing like how normal planes behave and is a result of the specific behaviour of KSP aerodynamics. The soupy atmosphere combined with the effectiveness of control surfaces at deflection means planes can turn way, way harder than they would normally be able to. In real life this would result in massive sideslip, spins, tearing off of wings, and so on, but not in KSP.

However, for whatever reason the effectiveness of control surfaces only seems to come into play when they are actually being engaged with input. When you zero the input and the control surfaces return to a "neutral" position, all of their control effectiveness appears to be lost. In other words, they'll let you turn hard when you provide input, but they won't then hold your course equally as hard when you stop providing input.

I tried adding symmetrical and balanced control surfaces to the craft with deflection and unbinding them from all inputs. For example, two small horizontal control surfaces at the rear, both on the centreline, with one pitched up a few notches and the other equally pitched down a few notches, then similarly near the front to balance out the lift and drag, and all locked to not respond to any input. But this didn't seem to make a difference. In other words, the super powerful force of control surfaces seems to come quite specifically when input is provided, but not otherwise.

Another thing to note is that (according to the VAB CoL indicator) vertical stabilizers will start to provide lift when the aircraft is rolled. For example, roll left in the VAB without vertical stabilizers and the CoL stays pointed along the normal out through the top of the craft. But add vertical stabilizers and do the same thing, and the vertical stabilizers will now produce an upwards force when the craft is rolled left, meaning your CoL now points not normal from the top, but angled slightly towards the right of normal (closer to vertical). I don't know if this is how they actually behave in game, but if so that would obviously contribute to instability during turns.

Edited August 1, 2014 by allmhuran

[EdFred]

This happens in real life too. My airplane doesn't zero out immediately when I stop with positive control inputs. It will oscillate for quite a while if I am not trimmed out correctly.

[allmhuran]

Yeah it happens with real aircraft and "real" flight sims as well, but nowhere near the extent it does in KSP. The wobbles that you see in KSP would kill anyone in the cockpit. As I wrote above, the reason for this is that the power of control surfaces to turn the craft as a result of input seems to be vastly higher than their ability to maintain a heading with no input.

[EdFred]

Nah, I got that when I flew the Extra 300 - and I'm still here, I think.

[Master Tao]

Yeah it happens with real aircraft and "real" flight sims as well, but nowhere near the extent it does in KSP. The wobbles that you see in KSP would kill anyone in the cockpit. As I wrote above, the reason for this is that the power of control surfaces to turn the craft as a result of input seems to be vastly higher than their ability to maintain a heading with no input.

Well, yes. Because the stock model doesn't account for drag by shape, the solution was to have the control surfaces add a force opposite their movement. They do nothing except provide lift when not deflected, though that's not much different from real life. The only way to maintain a heading is trimming until it's stable. I agree that the control surfaces provide way too much force – most planes would fall apart if you pulled 11 gees like you're doing in the video.

I'm actually surprised the stock drag model shows the inherent instability this well. You have a plane where the CoL and CoM are both close to the CoT. The only solutions are compensating with piloting or redesigning the plane. Otherwise, momentum, lift, and thrust will cause it to wobble while it's returning to a neutral position.

[allmhuran]

Really? You had wobbles on the order of 40 degrees? I very much doubt that.

Here is some extra 300L aerobatic footage, and here is a full break turn in DCS (which I also play) in an A10 for missile avoidance which shows pratically no deviation when coming out of the turn, and here is an F16 avoiding multiple SAMs, reversing high-G breaks for about 5 minutes, in which you can see the TVV is steady on the pitch ladder almost the whole time (it lags a bit when a turn is reversed of course, but it certainly doesn't wobble all over the place relative to the ladder).

They do nothing except provide lift when not deflected, though that's not much different from real life.

Sure, but when the TVV and the nose are not pointing in the same direction, they *are* deflected to the relative wind, they just aren't being actively deflected by input. For some reason the force provided by active input seems to go way beyond the force based on deflection from the relative wind. IE, they turn you hard, they don't straighten you out hard.

Now, if all the control surfaces were at the back then zeroing input would absolutely result in a sharp return of the nose towards prograde, effectively wobbling you back in the reverse direction of the turn. But if you balance the control surfaces front and rear then that shouldn't happen. Experimenting in KSP shows that it still does, which is what's so weird.

I think you have to imagine KSP "air" as more like water at low altitude. Try filling the kitchen sink and then dragging a knife through the water edge on. Now rotate the knife to initiate a turn. It will turn hard, and it *won't* sideslip much. Now straighten out the knife, and it certainly will not wobble all over the place.

Edited August 1, 2014 by allmhuran

[EdFred]

Really? You had wobbles on the order of 40 degrees? I very much doubt that.

Doubt all you want. What you saw in that 300 or F16 video, is not what you are describing in you original post.

They had controlled inputs coming out of the maneuver.

[allmhuran]

Doubt all you want.

Show me a video, from anywhere, of any aircraft, that swings this far, this rapidly, back and forth around the velocity vector while not in a complete stall. I can wait.

Hey,

is an aircarft suffering from compressor stalls on one engine. Man, look at that thing wobble! Oh... wait... nope.

What you saw in that 300 or F16 video, is not what you are describing in you original post.

They had controlled inputs coming out of the maneuver.

Hahaha, yeah, like at 3:17 to 3:25 when the pilot rolls as hard as possible directly from a full right break into a full left break, or another reverse at 5:13, or several other times in the video. You can see that the rate of turn on the compass absolutely does not switch gently from one directly to the other. It's a sharp snap from one direction to the other, and the TVV follows the compass the whole way except for some slight lag just as the reverse is initiated.

Edited August 1, 2014 by allmhuran

[Tex]

The wobbles that you see in KSP would kill anyone in the cockpit.

Wut?

The g-forces experienced in the video, while they do dip into the 8-13g range, wouldn't be acting long enough on a person for the force to cause redouts or blackouts, particularly to a trained fighter pilot. It would be uncomfortable, but a racecar driver survived a fraction-of-a-second 400g+ impact.

In response to your actual question, the wobble is being caused by too sensitive of an input. If you want to remove the wobble, you would need to perform your maneuvers more slowly than what the video shows. You may say that that is as slow as you can do it, but you COULD turn on fine control. If that doesn't work either, then using SAS and fine control should work much better. If it still doesn't, there isn't much that can be done without an actual joystick to input much more gradual control than key presses do.

[EdFred]

Yes, you won't find it in those videos, because the pilots are constantly exercising positive control input.They are not "letting go" (hand off) of an aircraft in uncoordinated flight. When you cease to provide inputs in uncoordinated flight (hands off everything) the airplane will oscillate back back and forth as it attempts to return to equilibrium How many hours do you have at the controls of a real life aircraft?

[allmhuran]

Wut? The g-forces experienced in the video, while they do dip into the 8-13g range

Wut? I'm not talking about the g forces, I'm talking about the wobble when input stops. I'm saying the person would break their neck and be smashing their skull around inside the cockpit as the plane violently swings back and forth.

In response to your actual question, the wobble is being caused by too sensitive of an input. If you want to remove the wobble, you would need to perform your maneuvers more slowly than what the video shows. You may say that that is as slow as you can do it, but you COULD turn on fine control. If that doesn't work either, then using SAS and fine control should work much better. If it still doesn't, there isn't much that can be done without an actual joystick to input much more gradual control than key presses do.

This is what I use to play KSP (and DCS):

Again, I'm not asking why it happens. I know why it happens. It happens because control surfaces don't seem to be treated the same way when responding to input as when not responding to input. What I'm asking is whether anyone has figured out a way of configuring lifting surfaces, or inertial mass, or control surfaces such that post-turn-wobbles of this nature don't happen.

[allmhuran]

Yes, you won't find it in those videos, because the pilots are constantly exercising positive control input.They are not "letting go" (hand off) of an aircraft in uncoordinated flight. When you cease to provide inputs in uncoordinated flight (hands off everything) the airplane will oscillate back back and forth as it attempts to return to equilibrium How many hours do you have at the controls of a real life aircraft?

I'm not sure if there's any value in repeating myself yet again, but yes, aircraft will oscillate to equilibrium. I'm not suggesting they won't. I'm saying the degree of this oscillation is vastly exaggerated due to the nature of KSP physics, and my question is whether anyone has figured out a way of preventing this using stock parts and stock physics. In other words, providing exactly the same kind of input as one would provide in a real aircraft or realistic simulator, and getting the exact same kind of response.

[Tex]

In that case, probably not. Wobble is something that all of us space-planers contend with, and I just use SAS to quell it.

[allmhuran]

In that case, probably not. Wobble is something that all of us space-planers contend with, and I just use SAS to quell it.

Yeh, I think you're right. During the course of this thread I've tried a lot of different craft configurations to see what effect it has. Lots of wings, on the hypothesis that this would provide more "bite". Slight improvement but not much. Decentralized mass to provide more inertial resistance to pitch, again, not much improvement. Vast amounts of control surfaces both front and rear, at many different angles to see if I could get some kind of deflection stabilization, but again, nope.

[EdFred]

I didn't see anything I would consider vast or exagerrated in your video compared to what I can replicate in a real world counter-part.

[JokingHinge]

Scott Manley has a flying box, stop talking like far is stock.

[allmhuran]

*shrug* I have a flying tank. I don't think anybody in the thread is suggesting FAR is stock. That's kinda the point of the thread.

No wait, I have two flying tanks! But one is a rocket...

[Master Tao]

Try making the plane longer and shifting the CoM forward.

I don't think anybody in the thread is suggesting FAR is stock.

I took that thought and ran with it. I moved the craft over to my mod install, disabled aerodynamic failures, and dumped some fuel (rear three tanks) to move the CoM ahead of the new CoL. The problem's even worse. I see oscillations after even slight maneuvers and they can grow to over 90 degrees after large maneuvers.