Aerodynamics Debug [F12] not accurate

[Soviet03]

[EDIT. there is no bug other than the inaccurate debugging display, aircraft performance is not affected]

bug description: rocket fuel tanks don't produce the same body lift forces that the mK.1 (airplane) fuel tanks do when constructed using symmetry mode (In particular one side will not produce body lift at all while the other side will).

What this means for aircraft: If you use rocket fuel tanks (e.g., FL-T400s)symmetrically, one side of the aircraft will display* normal body lift, and the other appears to display* zero body lift; causing the aircraft to become unstable and pull to one side which can even lead to catastrophic failure during high altitude flight . Here we have an aircraft demonstrating this with 2 jet fuel tanks with jet engines attached and 2 rocket fuel tanks, notice how one of the rocket fuel tanks do not display* body lift as it should [Red lines=drag,Yellow lines=control surface forces, Blue lines = wing lift, Cyan lines = Body lift] http://steamcommunity.com/sharedfiles/filedetails/?id=484727876

This can be easily reproduced by creating any aircraft with 2 symmetrically placed side rocket fuel tanks and pressing F12 for the overlay while flying at a reasonably fast speed, or if the Cyan lines don't show still, pitch slightly to induce them here are some images I and others have taken showing this problem.

http://steamcommunity.com/sharedfiles/filedetails/?id=484519026

http://steamcommunity.com/sharedfiles/filedetails/?id=484562623 (The following version is the same plane, except it only has Mk.1 fuel [airplane] fuel tanks) http://steamcommunity.com/sharedfiles/filedetails/?id=484562734

http://steamcommunity.com/sharedfiles/filedetails/?id=484444858

http://steamcommunity.com/sharedfiles/filedetails/?id=484435314

Edited August 5, 2015 by Soviet03

[Soviet03]

If more people would like to test this to further confirm it, that'd be great and could possibly lead to even more information about this bug.

Edited July 20, 2015 by Soviet03

[Dunbaratu]

Why is this report tagged as intended behavior? The overlay is clearly showing cyan arrows on the left side that aren't appearing on the mirrored part on the right side. If that's meant to be correct then clearly we users aren't understanding how the game is supposed to work, in a major way. Can that be explained please rather than just claiming it's correct without any commentary whatsoever on why? Because it certainly doesn't *look* right.

[Alberto De Beni]

I discovered this bug a lot of time ago, but I thought it was an overlay bug since it doesn't seem to affect lift

[TMS]

I built a fairly massive space plane with symmetrically-placed rocket fuel tanks on the wingtips (a la Skylon) that experienced this "pulling to the right" phenomena. Ended up pulling the craft apart, thinking I'd placed a wing non-symmetrically by accident, but could never locate it. i think this was in 0.90.

Funnily enough, the frustration of this experience was one of the reasons I gave up with the game.

Thanks for posting this.

[yorshee]

Would placing each wing manually instead of auto-symmetrically be a workaround for now? I can't say I've experienced this myself.

[Soviet03]

No, I actually tested that. Even if it's built in non-symmetry mode the bug still persists. A strange bug indeed.

[Claw]

I'm moving a conversation over here from another thread, since it's off topic over there.

I've seen those posts and pictures. However, I've not seen any evidence in the pictures that actually indicates that there's some sort of problem from this. By that I mean that yes, the lines are not of equal size. However, I don't see evidence in those pictures that the airplane is having to do anything to compensate for it.

So, not to be dismissive, I went ahead and gave it a try, replicating the basic layout of so many of these planes with one exception. I used a single engine instead of multiple. That eliminates any potential for asymmetric thrust.

Here are two pictures which do indeed show asymmetric body lift lines from the mirrored FL-T800s. However, if you look at the trim state, the airplane is not compensating for this in any way. I've also turned off SAS and flown it around and it does not exhibit any rolling/yawing tendency when straight and level or turning.

Additionally, I flew it up to max speed (~1300m/s below 20km) with zero yaw or roll reactions.

Javascript is disabled. View full album

So to further my analysis, I also flew the plane without extraneous parts on the FL-T800 and took a couple pictures. Look below, and my thoughts are still that this is related to a vector display problem, rather than an actual force application problem. It might not be obvious while playing the game, but there are two sets of code. One that calculates and applies the forces, and another set of code that calculates and draws the lift lines. Because of this, it's not guaranteed that the lift lines are a direct representation of the actual forces applied.

Note that at higher AoA, the body lift lines are equal. It's not obvious from the pictures, but there is no lift line on the left tank until the craft reaches a high enough AoA. Then the lift line suddenly pops into place.

Also note (in the second picture) that there are more body lift lines appearing on the right than on the left. Then look at the trim indicators in all of the pictures. In no case is there any reaction in yaw or roll, nor was there any "strangeness" in flight.

Javascript is disabled. View full album

So personally I'm not convinced there is a problem here in terms of actual gameplay, since in all the pictures I never see a reactionary response from the airplane. There are a lot of different forces that could be acting, especially at high speeds. If you have a craft that you think "pulls to the side," then the best bet is to provide craft files. Not the least of which is asymmetric thrust (which is still a problem in 1.0, though it's less obvious).

Cheers,

~Claw

[Signo]

Hi Claw, below a video of my craft and here the craft file itself (full stock).

I did a few high AoA climbs on purpose to show that the aircraft is indeed going left without any direct input.

Increasing the AoA the problem is extremely noticeable.

You will see I did a quick check of the engines that seem working properly. The video does not show the resource panel, but the craft can store 13.76U of air and they usually go out when you are around 24000m. I am not sure this is an asymmetric thrust issue.

Edited August 2, 2015 by Signo
craft

[Manwith Noname]

So personally I'm not convinced there is a problem here in terms of actual gameplay, since in all the pictures I never see a reactionary response from the airplane. There are a lot of different forces that could be acting, especially at high speeds. If you have a craft that you think "pulls to the side," then the best bet is to provide craft files. Not the least of which is asymmetric thrust (which is still a problem in 1.0, though it's less obvious).

Cheers,

~Claw

Most people who notice ill effects generally have built much larger craft than has been used to demonstrate the problem.

From my point of view, I find the behaviour of the overlay odd and I do not feel it is a problem with the overlay directly as when FAR is installed all forces appear to be shown as one would expect.

[Claw]

Increasing the AoA the problem is extremely noticeable.

You will see I did a quick check of the engines that seem working properly. The video does not show the resource panel, but the craft can store 13.76U of air and they usually go out when you are around 24000m. I am not sure this is an asymmetric thrust issue.

Sure, and that's a good video to start with. I don't think it's definitive proof that the aero overlay is showing the source of the "pulling to the side."

Most people who notice ill effects generally have built much larger craft than has been used to demonstrate the problem.

Well, that's some of my point. There are a lot of things that can explain the "pulling to the side" behavior (not just asymmetric thrust, which I didn't mean to implicate as the culprit). Large craft have their own brand of problems: bending, drag, not enough tail surface, etc.

I just think too much hat is being hung on the aero overlay. I'm not saying that it isn't indicating a problem. But just because the one line looks incorrect doesn't point to that as the source of the problem. It might be showing the problem, but there needs to be a lot more elimination of other things first.

Here's the other thing. If the lift there was actually asymmetric, then I would expect roll induced yaw (or some sort of control coupling), which is not what is happening in the video. But, in order to test that idea, this is what I would do (if I had the craft).

I would go back to the SPH and attach a single lifting surface on the right side of the plane, on top of that FL-T800. If I was really bug hunting, I would start splitting hairs on making sure the control surface lift corresponded to the body lift. ...Also, I would remove those structural intakes, so that I'm not confounding the lift lines between the FL-T800s and anything else (as I did in my own pictures).

Then go fly and see how the plane really reacts. If it wants to roll, then I know that the reaction is different than what is happening in the video, which looks to be all yaw.

It's not that I disbelive you guys. It's that I disbelive the data until proven otherwise. It's about removing possibilities until you discover what the actual cause is.

My smaller craft can be seen here, reacting to the addition of a lifting surface.

There is definite roll with this plane, and some yaw response as a result. You can see some SAS input in roll in this picture as the plane pitches up. If I remove that control surface, I can do a full loop with no issues, and I can pitch up and down with no cross coupling.

---

So perhaps post some craft files, so that others can fiddle with them. Post the things that you have removed as potential causes. Such as reducing that drag in front of the plane, and anything else you can think of. Show me that it's the body lift causing the problem, instead of assuming that the overlay is perfect.

Edited August 2, 2015 by Claw

[WinkAllKerb'']

Most people who notice ill effects generally have built much larger craft than has been used to demonstrate the problem.

From my point of view, I find the behaviour of the overlay odd and I do not feel it is a problem with the overlay directly as when FAR is installed all forces appear to be shown as one would expect.

same this evening with some liquid fuel tank on a MK2 part , i solved this fixing the parts along main Mk2 axis then offset/rotate instead of final position attach (<= reproduce steps). Doing so remove the inclined light blue vectoring.

(btw the light blue vectoring appear to be linked with subsequent parts main unity axis and/or relative @ mk2 {any other part (?)} fixing point)

let me know if you need some craft file, it's easy to reproduce.

Edited August 2, 2015 by WinkAllKerb''

[Signo]

Ehm, Claw, I did upload the .craft. It is linked in my previous message.

[Claw]

Doing so remove the inclined light blue vectoring.

That's excellent! So the next question' date=' did it change the planes reaction at all?

Ehm, Claw, I did upload the .craft. It is linked in my previous message.

Yeah, sorry. It's a general "upload some craft files and try different things," and not necessarily directed straight at you.

(nice edit btw)

For example, Wink came up with a great idea to flip the tank over and the line looks fine. So the next step off of that is to see if it actually changes how the plane flies. Which is actually quite difficult if there are wings and such hanging off of it. So, in that case, I'd build a large craft which exhibits the blue line AND flies awkwardly, but doesn't have wings on it. Then flip one of the tanks and see if it flies differently at all.

[WinkAllKerb'']

Javascript is disabled. View full album

Screenshot with subtitles

craft files:

https://drive.google.com/file/d/0B5JHmZea7s91LTE2SFhneXRYVVU/view?usp=sharing

https://drive.google.com/file/d/0B5JHmZea7s91OFpEOGtUM2Q5VjA/view?usp=sharing

https://drive.google.com/file/d/0B5JHmZea7s91aFo3anBDTmhnOGc/view?usp=sharing

https://drive.google.com/file/d/0B5JHmZea7s91aHJuLUY2QVBqd1k/view?usp=sharing

Not sure what it change or not not really tested myself (just remade the 02, 03, & 04 crafts so you can have a look).

[Manwith Noname]

Show me that it's the body lift causing the problem, instead of assuming that the overlay is perfect.

I see now. You are looking at this from how it has been reported and are looking to prove/disprove it's the aero physics model causing this.

My questions would be...

1) Why do the MK1 jet fuel and structural tanks behave as expected while the MK1 rocket tanks show this odd characteristic? (This suggests the part is the problem)

2) Why, when FAR is installed, do the MK1 rocket tanks show lift in the overlay as one would expect? (This suggests the aero calculations are the problem)

3) How is the overlay actually gathering it's data for the calculations it performs? (This might point to the problem given FAR displays correctly while the stock aero does not)

That said, I am more confused than ever now Wink has posted a craft showing MK1 jet fuel tanks displaying this anomaly. Until now, I'd kinda figured it was something related to the parts and how the aero models or overlay were analysing them. Though this does swing me back to my initial though that it is something to do with how the editor creates the craft.

[Claw]

My questions would be...

1) Why do the MK1 jet fuel and structural tanks behave as expected while the MK1 rocket tanks show this odd characteristic? (This suggests the part is the problem)

I don't know, this would be one of the things to investigate. But the bug question beyond "why is the cyan line wrong" in my mind is "is this really causing instability in the craft?" The cyan line might be showing something, but (imho) it's not proven anything yet. Challenge your assumptions and everything you see.

2) Why, when FAR is installed, do the MK1 rocket tanks show lift in the overlay as one would expect? (This suggests the aero calculations are the problem)

3) How is the overlay actually gathering it's data for the calculations it performs? (This might point to the problem given FAR displays correctly while the stock aero does not)

I do not know how FAR is set up (code wise), but I do know how the stock code does it. In stock code, there are two separate sets of calculations going on. Even if one were to assume that the code calculating the lift is perfectly identical, there's still the problem that the Unity physics engine (that applies the force) is using different code than the stuff that spits out the overlay. It's not the same code, so my personal opinion is that I cannot assume it's telling me the truth all the time, until proven that it's telling me truth in this instance.

If I knew that FAR used the same display code and physics code, then this could be an indicator that the display is telling the truth.

That said, I am more confused than ever now Wink has posted a craft showing MK1 jet fuel tanks displaying this anomaly. Until now, I'd kinda figured it was something related to the parts and how the aero models or overlay were analysing them. Though this does swing me back to my initial though that it is something to do with how the editor creates the craft.

The body lift system uses the drag cubes. Drag cubes are based on part shape, but individual parts should not be getting treated differently. I would suspect that you could somehow start lining up parts on both sides of the plane and see this behavior. It might be related to how the parts are connected, or how they are symmetric, or a multitude of other things that might be throwing off the rotational and translational calculations. This is what the process of "narrowing down a bug" entails.

And on that note, I would also like to say that I'm still confused about what the actual problem is here. Up until now, it sounds to me like it's a "the display overlay is incorrect" problem. In which case I would say that it's unlikely that this bug would be fixed.

If, however, someone is trying to assert that the cyan overlay is the smoking gun for large plane instability, I am still unconvinced. It's a possible indicator, but (in my mind) that is far from proven.

This is why I keep pushing for people to do more analysis, more digging, post craft files, share thoughts and ideas (as you are) and prove or disprove each other. I have to do this whole process when I'm digging at the bugs for my fixes, often by myself. Sometimes the bug is in a very bizarre place, that I wouldn't have expected when I started. For example, the claw bug that I just fixed...It has nothing to do with time warp, despite the fact that time warp is when all the bad stuff happens to the ship. I'm not saying "ignore the cyan line," I'm saying "prove to me that it's causing problem X." Also, please inform me as to what problem X is.

Cheers!

~Claw

[Manwith Noname]

Also, please inform me as to what problem X is.

Heh, yes, I agree. There is no doubt something strange happening but determining what the actual cause is, from a players point of view particularly but also a programmers, rather tricky. This either needs many people trying many things or someone with a very deep understanding of how all these systems work and bind together.

On that note, I have an idea.

Edited August 2, 2015 by Manwith Noname

[WinkAllKerb'']

the strange thing is that with craft (02 ?) 03 and 04 not all the light blue vector move the same while rolling (intake stay parrallel with the body but fuel tank vector rotate)

for the vectoring inclinaison my own bet as said is that it's mostly related to the linkage from one part to another

binded unity part main axis // surface attach relative axis. <= this tend to cause weird vector move while pitch, yaw, & roll (so probably might as well be related with relative absolute system calculation and use at some point in the source)

Edited August 3, 2015 by WinkAllKerb''
note for later: do not stay awake 36h ++ typo don't like that ; ) oops

[Manwith Noname]

Uch, double posting...

I now believe FAR does indeed manipulate the overlay in some way. Prior to just now, I was relying on another users report that it worked fine with FAR installed as I have been running KSP stock. I just downloaded the latest FAR and noticed that the overlay is actually different. All lift lines appear in cyan and to my surprise, fuel tanks did not display any body lift at all, at least with the craft and speeds I was testing.

Edited August 2, 2015 by Manwith Noname

[Claw]

the strange thing is that with craft (02 ?) 03 and 04 not all the light blue vector move the same while rolling (intake stay parrallel with the body but fuel tank vector rotate)

for the vectoring inclinaison my own bet as said is that it's mostly related to the linkage from one part to another

binded unity part main axis // surface attach relative axis. <= this tend to cause weird vector move while pitch' date=' yaw, & roll (so probably might as well be related with relative absolute system calculation and use at some point in the source)[/quote']

I'd say that this is a possibility, given the history that the overlay was simply a quick debug tool to start.

I now believe FAR does indeed manipulate the overlay in some way.

Okay. So there's one piece of information that's possibly disproved. Maybe someone with a lot of FAR experience can verify.

Cheers,

~Claw

[WinkAllKerb'']

That's excellent! So the next question, did it change the planes reaction at all?

Not sure what it change or not not really tested myself (just remade the 02' date=' 03, & 04 crafts so you can have a look).[/quote']

at least it appeared neglectable when i quickly tried a few take off with thoose planes so might be just an overlay tool error matter as you said and not physics calculation.

Also i not really paid attention to it as i almost instantly found the attach middle workaround like with *01.craft.

(notice it could be related that's this plane is pretty fast and stable by nature so might require some verif with some less stable planes)

[Manwith Noname]

Given that I now know there are 2 separate systems, the actual aero physics and a separate (presumably simpler) overlay calculation, I'm coming to the conclusion that it is a problem with the overlay, though I'm struggling to understand how it would work fine with certain parts but not others. What I can say is that it seems to have issues with what would be considered low levels of force. For example, if you yaw the craft fairly hard, 2 symmetrical body lift lines appear. If you pitch hard, the missing line appears and mirrors the existing one.

[Soviet03]

Claw, I tried my best to make sure this bug report was small and easily readable. I can 100% assure you this bug DOES have something to do with how the aircraft flies, as with anything in this situation the bug and its effects are greatly increased/magnified when you start to build larger aircraft. - Such as large complex aircraft using HUGE rocket lifting bodies extended quite far onto the wings such as skylon to increase the leverage effect.

For instance, the large aircraft I mentioned in the main post will always have the right wing for no reason rip off first under load, or that at high speed/high altitude it pulls VERY harshly to the right and up. An odd pull effect that is VERY noticeable. - When I did tests with just 1 giant rocket engine and no fuel pylons the aircraft had no such issues. See the steam discussion I linked in the OP for some more in depth detail if you're interested. On the surface with a smaller aircraft I can see why you don't feel it having a real pull but I can assure you it definitely screws over larger/more complex aircraft. Such issues do not exist in FAR, or when the problem parts are removed from the aircraft (The rocket fuel tanks)

In particular my aircraft uses 2 MASSIVE rocket tanks and has quite a lot of lifting surfaces, when it pulls at high altitude its often impossible to steer out of. Its unlikely that you'll notice this effect when using such small tanks/lifting bodies on rigid aircraft. (My large aircraft is 150 parts plus)

Edited August 5, 2015 by Soviet03

[Claw]

Claw, I tried my best to make sure this bug report was small and easily readable....

And you did a fine job, although craft files would help.

In particular my aircraft uses 2 MASSIVE rocket tanks and has quite a lot of lifting surfaces, when it pulls at high altitude its often impossible to steer out of. Its unlikely that you'll notice this effect when using such small tanks/lifting bodies on rigid aircraft. (My large aircraft is 150 parts plus)

Well, my point here is that it could be a variety of reasons why it's pulling at high altitude. I'm trying to convince everyone to not immediately bite off on the display, assuming it is 100% accurate.

Hi Claw, below a video of my craft and here the craft file itself (full stock).

Okay, you are indeed correct. Your plane does not appear to suffer from asymmetric rollback or flameout (at least in the profile I climbed). I was going to write up a huge assessment of my flight with your plane, but it's getting late and there's a lot more that I want to write up. From the first post I assumed your craft would exhibit some lateral directional instability (in yaw) because you have a relatively small tail for such large intakes out front (the quad couplers). This is going to cause some "hunting" around. At high altitude, where you are likely at high speed, this hunting eventually turns into a yaw excursion and departure.

So, to test my instability theory (which may not be complete, I'll grant), I simply capped the side fuel tanks with a nosecone and moved the eight intakes onto the tail. Two fold change here: Removes drag from up front, and increases stability in back without adding more tail.

The result is that flying with the modifications made for a more directionally stable aeroplane. I did NOT fly this through the whole envelope, so I don't swear that this fixes all of the problems. Just trying to show that there's more than one thing at play here.

Javascript is disabled. View full album

There's actually a lot of stuff to point out in these pictures (a lot), but I simply don't have time. One of the interesting things is that more lines were disappearing than the one that we've been posting about.

---

---

Now, that being said, to show what I mean about two things...

1) I don't believe the aero overlay is perfectly telling the truth

2) Demonstrate what I mean by "experiment and uncover the truth"

I spent a lot of time this evening flying the above plane around. Upside down, spins, stalls, high atmosphere, dives, increasing and decreasing body lift, all kinds of things. It's a fine plane, and hunts around a bit, but I've already discussed that. My point in this post is going to be aimed at the overlay in hopes to show you how to dig at something and to also show that I don't believe you guys ought to be trusting the overlay as a smoking gun answer to your concerns.

What follows is a shortened and condensed version of my testing, again because it's late and I'm sleepy.

So, in order to attack the overlay I did I simply built a 4x symmetric rocket around a central core in the VAB.

Then I flew it east, south, and west to see if the lines showed any difference (due to craft directional orientation). I didn't bother with North, since the other three directions matched. For completeness, I should have...

Javascript is disabled. View full album

Then I tried rolling the craft into different orientations, and in those movements you can see that the line blinks into and out of existence. There's only two pics here showing the change.

Javascript is disabled. View full album

So, roll orientation matters!

Then I flipped the tanks around in the VAB, so that the bottom of the tank is at the top of the rocket.

Javascript is disabled. View full album

Well, that seemed to matter quite a bit. Everything basically flipped with the tank orientation.

Hmm, so what about FL-T400s?

Javascript is disabled. View full album

So that's even more strange, but notably, the lines are only flipping around on the surface mounted parts...and not the stack mounted ones.

So, lets try the Mk1 fuselage, since someone said that doesn't suffer from the same problem.

Javascript is disabled. View full album

So it actually DOES suffer from the same problem, except that it's happening in a different place...under the belly, instead of the right wing...

So what's the same between the FL-T800 and FL-T400, but is different than the Mk1 Fuel tank?

Lets look at the part files, shall we? Turns out there's a lot of differences, but lets cut to the chase and look at the node_attach, which controls surface attachment.

FL-T800

	scale = 0.1
	node_stack_top = 0.0, 15, 0.0, 0.0, 1.0, 0.0
	node_stack_bottom = 0.0, -15.1, 0.0, 0.0, -1.0, 0.0
	node_attach = 5.01, 0.0, 0.0, 1.0, 0.0, 0.0, 1

FL-T400

	scale = 0.1
	node_stack_top = 0.0, 7.72552, 0.0, 0.0, 1.0, 0.0
	node_stack_bottom = 0.0, -7.3, 0.0, 0.0, -1.0, 0.0
	node_attach = 5.01, 0.0, 0.0, 1.0, 0.0, 0.0, 1

Mk1 Fuel Tank

	scale = 0.0125
	rescaleFactor = 1
	node_stack_top = 0.0, 75.0, 0.0, 0.0, 1.0, 0.0
	node_stack_bottom = 0.0, -75.0, 0.0, 0.0, -1.0, 0.0
	node_attach = 0.0, 0.0, -51.0, 0.0, 0.0, 1.0, 1

Turns out Mk1 fuel tank attaches via the Z coordinate and rotation, vice the X coordinate and rotation. X is the "left/right" of the plane, and Y is the "belly/top" of the plane.

Also, to add the the wackiness, turns out that the +/- (sign) value is important too.

So what other part is similar to the Mk1 Fuel tank? Ah, the Mk1 structural fuselage.

	rescaleFactor = 1
	node_stack_top = 0.0, 75.0, 0.0, 0.0, 1.0, 0.0
	node_stack_bottom = 0.0, -75.0, 0.0, 0.0, -1.0, 0.0
	node_attach = 0.0, 0.0, -51.0, 0.0, 0.0, 1.0, 1

So I also tried the Mk1 structural fuselage, and it was the same as the Mk1 fuel tank. I'll spare you the pictures of that one...

Lets take all that together then, and see if we can build a craft that shows zero body lift on the booster parts.

Javascript is disabled. View full album

Almost, but not quite...

Anyway, hopefully you can see that there's something else going on here, and it's not quite a smoking gun. Otherwise all of these wacky contraptions would probably be going out of control at some point. But if you roll one of them even slightly, all the lines show up. Not to mention that if I send the things out of control, the lines start pointing everywhere.

Javascript is disabled. View full album

So, all of this work and time spent, and really all I've (hopefully) shown is that the overlay isn't necessarily trustworthy. I think the craft would have flow much differently from each other if the lines were accurately showing what's going on under the hood. So, this doesn't (in any way) discount that something else might be going on that's causing yaw problems. Especially since I can show you yaw drift with rovers, planes, rockets, and unpowered gyros. There's a bias built into KSP somewhere that causes things to bend in their flight path, not to mention all the other physics things going on.

Cheers,

~Claw