which parts of KSP1 aero do we want to keep?

[OHara]

StarTheory plan for KSP2 to have aerodynamics very similar to that in KSP1
That system uses several rules---some of which may have helped to make a good game---and not all of them will carry over into a fresh implementation.

0) Each part feels aerodynamic forces as if it was moving through the air alone...
   Struts have zero drag, but all other parts have aerodynamic forces applied.

1) Except, KSP1 reduces the forces on each face that is node-attached to another part,  to the extent that the attached part covers that face.
 The need to explicitly define the 'covering area' for tube-like parts has tripped up many mods, including Making History.

2) Offsetting a part does not change the size of the aerodynamic forces on the part  (but the point where those forces are applied to the whole craft does move with the part).
 Clipping, offsetting parts inside other parts, is purely aesthetic

3) Rotating a part does change the aerodynamic forces.
 We can set angles of incidence on wings, and on SRB-fins, to give the desired effect 
 We can rotate nose-cones 180° and avoid the leading-face drag in favor of the much lower trailing-face drag 

4) Except, any part inside a closed container feels zero aerodynamic forces, (and only parts in the 'Payload' section in the VAB count as containers).
 The definition of 'inside' is reasonably sophisticated and is rechecked when the container opens or closes 
 There have always been one or two bugs where the checking goes wrong

5) Parachutes, thrusters, antennae, and landing gear do not function from inside a closed container ("cannot deploy while stowed").
 We have an override ('deploy shielded') for landing gear

6) The forces on wings as a function of angle-of-attack give a very soft stall, that behavior being defined by custom curves in physics.cfg.

7) Skin drag, from faces of a part parallel to the airflow, use the same shape factor Cd as they would have when facing the airflow.
  This gives Mk2 cargo bays slightly lower drag when open, at high speed.

8) As the side-face of a (non-wing) part is tilted into the airflow, the force proportional to tilt angle is all counted as drag,  where we expect drag proportional to sine-squared of the tilt angle from standard aerodynamics.
  An Mk2 spaceplane in KSP1 at 3° angle of attack feels sin3° = 5.2% (where one would expect only sin²3° = 0.3%) of the of bellyflop drag of its bottom surface.

9) The lift on wings as a function of Mach number, is higher at low speed, lower lift at high speed, compared to expectations from standard aerodynamics.
 A space-plane can easily have enough low-speed lift to glide near touchdown-speed the entire length of the runway.

Links to threads for reference in spoiler:

Spoiler

 

 

 

https://bugs.kerbalspaceprogram.com/issues/18105

https://bugs.kerbalspaceprogram.com/issues/13366

https://bugs.kerbalspaceprogram.com/issues/21915

 

 Now seems like a good time to discuss which of these have been good or bad for the game, in our experience with KSP1.

Edited January 17, 2021 by OHara
links didn't work, so links in the spoiler

[Voculus]

I personally dislike the stock aero a great deal, for many of the reasons you listed. I would have loved an option for a more realistic aero model, but I suppose that's where the modders will come in.

[Incarnation of Chaos]

None of it; throw it out completely and take FAR, thank the original author and the person maintaining it kindly and make it stock. It's not that much more difficult than stock (Especially if you use the !MORE BOOSTERS approch) and the voxel implementation avoids most of these issues most of the time (Bugs still crop up when things aren't defined properly). There's no reason to keep this system when a reasonably accurate mod can do the job BETTER and FASTER in most cases and already exists.

You could fudge the lift value (It's squared in stock KSP) and have a toggle for true FAR-like behavior that returns it to the real; so people can still build their kerbal-sized contraptions.

[OHara]

The mod FAR did show that simulating airflow around [analyzing the shape of] the whole craft is feasible, and I would like to see that in KSP2, but that is not the current plan :

Maybe we can argue that a rockets game should get proper aerodynamics treatment, but if that argument fails, there are some complications in the current KSP aerodynamics that would be better left out, making a simpler and easier-to-understand KSP2.   

In my opinon...

0) I think having drag on each part is fine.  The need to balance attachments to our rockets against their drag is good for the game.
KSP1 automatically figures drag of other parts from their models (specifically their colliders) and this is where things go wrong: new parts have models or custom drag data that does not meet the expectations of the KSP1 system. (FAR had a bug in this class that was exposed by the ReStock mod, so it happens to the best of us.)  If KSP2 documents, at least internally, the way the aero system interprets the model and configuration of parts, that should help.

1) Removing drag of covered faces is the lowest-order approximation of whole-body aerodynamics, so in absence of other offers I'll take it.
[Edit: after further thought and reading others' comments, I very much hope for a better offer, that uses the shape of the assembled craft to figure aero forces.]

2) I like offset-independence, not having to worry about exactly where I place my surface-attach parts on a spaceplane, although when I first started I did try to tuck RCS thrusters in low-flow areas, not knowing that it made no difference in KSP's model.
[Edit: after further thought, since most people expect that offsetting parts to make a smoothed craft would reduce drag, better if KSP 2 follows that expectation.]

3) KSP1 has enough realism to demonstrate how the rotation of airfoils affects stability; it shows the effect of wing dihedral, for example.  I like that.  
One can cheat the current system, by rotating nose-cones backwards, for example, but I don't think that loophole alone is enough to warrant a more complex system.

4) Players expect that tucking things inside hollow-looking parts will remove drag, and protect from heat.  I recommend applying the protection of cargo bays to all hollow-looking parts, maybe with a note in their descriptions to confirm which are hollow.

5) The rule that such protected things are forbidden from interacting with the outside world, seems an un-necessary complication, in hindsight. Drop "cannot deploy while stowed"

6) I think a more realistic stall behaviour would be fun, and make aircraft more interesting, while not affecting high-powered spaceplanes and fighter replicas very much.  KSP1 can do that with 'lifting_surfaces_curves::lifting_surface::lift,drag' and it won't affect rockets because their angle of attach remains low.
[Edit: FAR has differently unusual stall behaviour that some players would change, if it were configurable]

7) Skin drag is small enough that it can depend only on the side area of the part, without a coefficient depending on the shape of the part --- certainly it should not use the same drag coefficient Cd as the face would have if facing the airflow.

8) The configuration value 'drag_tip'  should apply to faces according cos² (dot-product squared) of their face relative to the airflow.   I do not think Mk2 spaceplanes were made draggy on purpose, or intended to force us to put an angle of incidence on our wings, nor that rockets were intended to suffer drag so severely when we deviate from prograde.

9) Having the coefficient of lift increase with low speeds was a clever idea to enable players to make light aircraft and spaceplanes with the same set of parts.   Sometimes people will say the air 'feels thick' when landing but mostly this enabling trick seems not to bother players.

Edited August 14, 2021 by OHara

[Laie]

2 hours ago, OHara said:

The mod FAR did show that simulating airflow around the whole craft is feasible, and I would like to see that in KSP2, but that is not the current plan :

On 8/31/2019 at 5:38 AM, Snark said:

• Q:  What about the aero model? Will it be like KSP 1, or something more like FAR?
• A:  We don't want to confuse players who are used to flying in KSP 1, so it'll have a very similar aero model.

 

I've understood that to mean that vessels will behave similar to old KSP, and that aerobatics like these will work as we've learned to expect them:

It may be all wrong and broken, yet I can imagine the outcry if things no longer worked that way. Just watch, it's fun!

This doesn't mean that they have to truthfully recreate the current system with all it's quirks, though. The whole complex about occlusion is obviously broken, in a bad, totally no-fun way, and won't be missed by anyone*. I think there's still hope that they take a holistic view of the vessel for drag purposes.

--

* I expect that even those people who exploit the current system will only be like "well, it was fun as long as it lasted".

[neistridlar]

1) Except, KSP1 reduces the forces on each face that is node-attached to another part,  

2) Offsetting a part does not change the size of the aerodynamic forces on the part

3) We can rotate nose-cones 180° and avoid the leading-face drag in favor of the much lower trailing-face drag.

4) Except, any part inside a closed container feels zero aerodynamic forces,

These would all go away if they ditched the drag cube system, and in stead adopted a voxel based system similar to what FAR uses. The drag cube system is good enough for very simple rockets and planes, but there is so much that does not work in an intuitive and realistic way due to it's limitations. Say for instance you built a cargobay out of wing sections or structural panels, put it in line with your fuselage, and blended it all nicely together. Well, all you are achieving is to create massive amounts of drag, while you would expect such a contraption to create very little drag. Switching over to a voxel based system would make this work as expected, and open up for much more interesting and kerbal designs.

7) Skin drag, from faces of a part parallel to the airflow, use the same shape factor Cd as they would have when facing the airflow.

This combined with the way drag cubes are calculated makes parts with flatish sides, such as Mk2 and Mk3, and many other profiles invented by the community a lot more draggy than they should be. The reason being that the Cd for a flat surface facing straight into the air-stream is very high, but a flat surface parallel to the airflow should not create much more drag than an equivalent rounded or slanted surface, both of which gets lower Cd. In fact with the current drag cube system you can get a squareish profile to generate a lot less drag over all if you rotate it on edge, so that it is pointy in stead of flat from the sides, before the drag cube is generated. After that you can rotate it back, so it is essentially identical to one that did not have it's drag cubes generated while sitting flat. That is just stupid as far as I'm concerned. 

6) The forces on wings as a function of angle-of-attack give a very soft stall

9) The lift on wings as a function of mach number, is higher at low speed, lower lift at high speed

I think these should be kept. It makes aircraft design more forgiving. It is hard enough for most players already, and most of the basic principles of plane design still applies. Also, changing this I think is what will alienate most players. Forcing the full complexity of FAR onto everyone is probably not going to be a good Idea.

[Motokid600]

I honestly don't think FAR is any more complex then stock. The complicated information is there for the people who want it, but you don't need it. You follow the same principles in stock as far as CoM/CoL go and still get a flyable aircraft. Only it actually feels like you're flying on a cusion of high pressure air as opposed to floating through liquid. 

Edited September 23, 2019 by Motokid600

[Gydra54]

5 hours ago, Motokid600 said:

I honestly don't think FAR is any more complex then stock. The complicated information is there for the people who want it, but you don't need it. You follow the same principles in stock as far as CoM/CoL go and still get a flyable aircraft. Only it actually feels like you're flying on a cusion of high pressure air as opposed to floating through liquid. 

Yeah I used to be so scared of mods like FAR, thinking the realism would make it less fun. Turns out all it changed practically is my rockets actually have less drag (though slightly less stable) and doing 90 degree flips with big ships annihilates them. Like, is it really more intimidating than stock aerodynamics? I doubt newcomers would be scared off by it. If anything, watching your unstable rocket flip over and consequently disintegrate would surely make for good entertainment.

[tseitsei89]

10 hours ago, Laie said:

• Q:  What about the aero model? Will it be like KSP 1, or something more like FAR?
• A:  We don't want to confuse players who are used to flying in KSP 1, so it'll have a very similar aero model.

Wow! 

That's like saying: "We have no intention to fix any glaring shortcomings of the previous parts. Instead we make the same mistakes again!"

Stock aero model is obviously lackluster and ridiculous. I really hope they improve it significantly.

Besides "confusing the old players" is a completely nonsense reason. Big reasons we would buy the new game are new challenges and improved gameplay and physics models, not just getting the same KSP1 with better graphics...

[Guest]

I don't think artificially softening the stall is necessary, or the equally artificial lift increase at low speeds. It can actually make landings harder, KSP planes tend to be prone to bouncing on landing due to that (though the shoddy wheel spring model doesn't help). Instead, wings of different kinds should be provided, for making aircraft suited to different tasks. They definitely should improve on the old aerodynamics model. Old players will be able to handle it, assuming it's adequately explained. 

FAR used to change a lot, but right now, it's more of a quantitative change than a qualitative one. If anything, it makes it easier to make a well-behaved plane if you know how to use the numbers. I would be wary of voxel-based systems, though. Modern FAR has a huge performance impact, and some parts don't voxelize correctly for some inexplicable reason. However, if those issues could be overcome, one thing that could be great would be voxel-based lift, not only drag. FAR never did that and in fact, its wing code was very frustratingly limited in its ability to handle unusual wing shapes.

I'm not advocating simulating vortices and all that stuff. While neat and realistic, it would require real-time CFD, which would be extremely taxing on the CPU (one thing we really don't need more of). 

[Incarnation of Chaos]

On 9/23/2019 at 2:54 PM, Dragon01 said:

I don't think artificially softening the stall is necessary, or the equally artificial lift increase at low speeds. It can actually make landings harder, KSP planes tend to be prone to bouncing on landing due to that (though the shoddy wheel spring model doesn't help). Instead, wings of different kinds should be provided, for making aircraft suited to different tasks. They definitely should improve on the old aerodynamics model. Old players will be able to handle it, assuming it's adequately explained. 

FAR used to change a lot, but right now, it's more of a quantitative change than a qualitative one. If anything, it makes it easier to make a well-behaved plane if you know how to use the numbers. I would be wary of voxel-based systems, though. Modern FAR has a huge performance impact, and some parts don't voxelize correctly for some inexplicable reason. However, if those issues could be overcome, one thing that could be great would be voxel-based lift, not only drag. FAR never did that and in fact, its wing code was very frustratingly limited in its ability to handle unusual wing shapes.

I'm not advocating simulating vortices and all that stuff. While neat and realistic, it would require real-time CFD, which would be extremely taxing on the CPU (one thing we really don't need more of). 

Why not have a wind tunnel/simulator that performs CFD in a limited environment (like a tube) then takes the data from the run and bakes it into the flight model. This would still not simulate vortices etc. But would be far more accurate and aligned with what FAR already does (patch the aero model).

Or the alternative is we all start playing x-plane lel.

[Guest]

That wouldn't work, conditions the plane will be going through will be just too varied. KSP craft routinely go from standstill to orbital velocities. On top of that, different planets have different atmospheres, and craft routinely change their shape in flight as a result of forces acting on them. This is FAR beyond what X-plane does.  In fact, I've talked to a physicist who does somewhat related stuff about that (at one point I thought KSP could use some competition to keep the free market spirit going  ) and from what he said, it's not really possible to do it. Basically, Reynolds number varies greatly due to sheer range of speeds involved, which means it can't be assumed nearly constant, which is one of the assumptions that makes real time fluid dynamics workable. 

Accurate simulation of fluid dynamics are one of the hardest problems in modern physics computing. Attempting a real-time simulation of airflow in what is basically real-world conditions will devour CPU cycles. Even if you found a workable compromise for CFD aerodynamics, you'd be using most of the CPU just for that, leaving too little for the rest of physics calculations. Trust me, fluid dynamics are hard. A voxel-based approximation in vein of what FAR does is pretty much the best we're probably going to get.

[Incarnation of Chaos]

29 minutes ago, Dragon01 said:

That wouldn't work, conditions the plane will be going through will be just too varied. KSP craft routinely go from standstill to orbital velocities. On top of that, different planets have different atmospheres, and craft routinely change their shape in flight as a result of forces acting on them. This is FAR beyond what X-plane does.  In fact, I've talked to a physicist who does somewhat related stuff about that (at one point I thought KSP could use some competition to keep the free market spirit going  ) and from what he said, it's not really possible to do it. Basically, Reynolds number varies greatly due to sheer range of speeds involved, which means it can't be assumed nearly constant, which is one of the assumptions that makes real time fluid dynamics workable. 

Accurate simulation of fluid dynamics are one of the hardest problems in modern physics computing. Attempting a real-time simulation of airflow in what is basically real-world conditions will devour CPU cycles. Even if you found a workable compromise for CFD aerodynamics, you'd be using most of the CPU just for that, leaving too little for the rest of physics calculations. Trust me, fluid dynamics are hard. A voxel-based approximation in vein of what FAR does is pretty much the best we're probably going to get.

I suppose there's a good reason why we still have supercomputers then lel. I forgot how varied the conditions in ksp are.

[Superfluous J]

I really don't mind the aerodynamics we have now. It helps that I see very little utility in planes because I'm always busy making rockets.

So long as any changes they make don't affect maximum part count before lag, I'll welcome any change that people think makes it more realistic.

[Guest]

Lag is my primary concern, as well. FAR already has problems with the voxels affecting performance (that and occasional voxelization derps are why I don't use it anymore), though a built-in solution would probably be faster. The big problem with improving aerodynamics is that anything bigger than a tweak starts to have steep performance costs. Aerodynamics, in general, are hard, and in the real world, it's still very much an experimental science. No computer so far can replace wind tunnel tests, especially with unusual shapes. KSP being all about unusual shapes, it's pretty much a lost cause. I'd love to see such effects as a dissipating vortex clobbering the tail fin (actual cause of rudder failure on early Hornets), air cushions forming under a cleverly designed wing, wing interactions on multiplanes and all that funny stuff (indeed, that was part of the motivation for looking into developing a KSP-style flightsim), but the actual equations are just too darn complex, and available approximations are pretty much only good with roughly plane-shaped objects, rockets or bullets.

[EnderKid2]

I'd be fine with removing the two biggest flaws, the hollow-parts-not-acting-like-cargo-bays (I want to be able to stick stuff like avionics in nosecones), and the whole stupid way-too-much-skin-drag-on-flat-parts thing.

[Nothalogh]

On 9/22/2019 at 3:24 AM, OHara said:

StarTheory plan for KSP2 to have aerodynamics very similar to that in KSP1
That system uses several rules---some of which may have helped to make a good game---and not all of them will carry over into a fresh implementation.

0) Each part feels aerodynamic forces as if it was moving through the air alone.
   Struts have zero drag, but all other parts have aerodynamic forces applied.

1) Except, KSP1 reduces the forces on each face that is node-attached to another part,  to the extent that the attached part covers that face.
 The need to explicitly define the 'covering area' for tube-like parts has tripped up many mods, including Making History.

2) Offsetting a part does not change the size of the aerodynamic forces on the part (but the point where those forces are applied to the whole craft does move with the part).
 Clipping, offsetting parts inside other parts, is purely aesthetic

3) Rotating a part does change the aerodynamic forces.
 We can set angles of incidence on wings, and on SRB-fins, to give the desired effect.
 We can rotate nose-cones 180° and avoid the leading-face drag in favor of the much lower trailing-face drag.

4) Except, any part inside a closed container feels zero aerodynamic forces, and only parts in the 'Payload' section in the VAB count as containers.
 The definition of inside is reasonably sophisticated and is rechecked when the container opens or closes.
 There have always been one or two bugs where the checking goes wrong.

5) Parachutes, thrusters, antennae, and landing gear do not function from inside a closed container ("cannot deploy while stowed").
 We have an override ('deploy shielded') for landing gear

6) The forces on wings as a function of angle-of-attack give a very soft stall, that behavior being defined by custom curves in physics.cfg.

7) Skin drag, from faces of a part parallel to the airflow, use the same shape factor Cd as they would have when facing the airflow.
  This gives Mk2 cargo bays slightly lower drag when open, at high speed.

8) As the side-face of a (non-wing) part is tilted into the airflow, the force proportional to tilt angle is all counted as drag,  where we expect drag proportional to sine-squared of the tilt angle from standard aerodynamics.
  An Mk2 spaceplane in KSP1 at 3° angle of attack feels sin3° = 5.2% (where one would expect only sin²3° = 0.3%) of the of bellyflop drag of its bottom surface.

9) The lift on wings as a function of mach number, is higher at low speed, lower lift at high speed, compared to expectations from standard aerodynamics.
 A space-plane can easily have enough low-speed lift to glide near touchdown speed the entire length of the runway.

Links to threads for reference in spoiler:

  Reveal hidden contents

 

 

 

https://bugs.kerbalspaceprogram.com/issues/18105

https://bugs.kerbalspaceprogram.com/issues/13366

https://bugs.kerbalspaceprogram.com/issues/21915

 

 Now seems like a good time to discuss which of these have been good or bad for the game, in our experience with KSP1.

>which parts of KSP1 aero do we want to keep?

 

Hopefully none of it, we need FAR style voxelization, and hopefully supersonic and hypersonic compression lift as well.

[Incarnation of Chaos]

23 minutes ago, Nothalogh said:

>which parts of KSP1 aero do we want to keep?

 

Hopefully none of it, we need FAR style voxelization, and hopefully supersonic and hypersonic compression lift as well.

Pretty sure compression lift would be a nightmare to code; it would require looking at the shape of the aircraft and then determining to what degree it would generate it. Then calculate that in real time while accounting for speed, density etc. 

[Flavio hc16]

i'm in the " let's get a far style aero model in ksp2" party

[Nothalogh]

1 hour ago, Incarnation of Chaos said:

Pretty sure compression lift would be a nightmare to code; it would require looking at the shape of the aircraft and then determining to what degree it would generate it. Then calculate that in real time while accounting for speed, density etc. 

The Trajectories mod dynamically calculates atmospheric trajectory based on vehicle orientation in real time while accounting for speed, density etc.

 

 

[Incarnation of Chaos]

4 hours ago, Nothalogh said:

The Trajectories mod dynamically calculates atmospheric trajectory based on vehicle orientation in real time while accounting for speed, density etc.

 

 

Calculating a point object and where it'll end up is far less taxing than the entire aerodynamics of an aircraft rapidly shifting while in flight; i suppose they could reduce the load by only calculating key areas and using them to approximate the entire body. But even then; it just seems like it'll end up being a massive framerate sink for very little gain. Reasonably there's only a few situations where compression lift becomes a factor; even then you have to design the aircraft to exploit it. Could it help with SSTO Aircraft? Sure, but is it really worth it for the one dude who'll design his aircraft to be a space-age XB-70?

 

[Nothalogh]

24 minutes ago, Incarnation of Chaos said:

Calculating a point object and where it'll end up is far less taxing than the entire aerodynamics of an aircraft rapidly shifting while in flight; i suppose they could reduce the load by only calculating key areas and using them to approximate the entire body. But even then; it just seems like it'll end up being a massive framerate sink for very little gain. Reasonably there's only a few situations where compression lift becomes a factor; even then you have to design the aircraft to exploit it. Could it help with SSTO Aircraft? Sure, but is it really worth it for the one dude who'll design his aircraft to be a space-age XB-70?

 

It applies to all aerodynamics in the hypersonic regime

[Voculus]

On 9/23/2019 at 12:40 PM, tseitsei89 said:

Wow! 

That's like saying: "We have no intention to fix any glaring shortcomings of the previous parts. Instead we make the same mistakes again!"

Stock aero model is obviously lackluster and ridiculous. I really hope they improve it significantly.

Besides "confusing the old players" is a completely nonsense reason. Big reasons we would buy the new game are new challenges and improved gameplay and physics models, not just getting the same KSP1 with better graphics...

Very well said, and I agree completely. It's disappointing to see them unwilling to revamp such a vital part of space flight.

[OHara]

On 9/23/2019 at 2:32 AM, Laie said:

I've understood that to mean that vessels will behave similar to old KSP, and that aerobatics like these will work as we've learned to expect them:

That video of Cupcake's Vulture inspired me to try flying it under FAR and stock KSP (version 1.3.1).

The most obvious difference by far is stock KSP's greater the lift and drag coefficients at low speeds, making the air feel much more dense (point 9 above). 

Stock KSP uses a lift coefficient 5× that in FAR, making aerodynamic forces 5× stronger at any given speed.  Stated in other ways, the Vulture can pull a 5-G maneuver at √5 times lower speed, or make turns 5× tighter, in stock compared to FAR.

The next obvious difference follows from the same difference in physics: controls feel more effective at low-speed in stock, less effective at high speed -- where in FAR coarse action on the controls at high speed can rip the wings off.

The Vulture has landing gear that retract into the nacelles, leaving nothing exposed.  With FAR, when we try to extend them, FAR tells KSP that they are enclosed, so KSP says "cannot deploy while stowed".

KSPs configuration files have become complex over the years, but they do let us change things pretty easily.  It is easy to substitute more realistic lift and drag forces as a function of airspeed, and/or make the stall behavior less mild, if we ever want : 

Spoiler

@PHYSICSGLOBALS {
	@LIFTING_SURFACE_CURVES {
	  @LIFTING_SURFACE[Default] {

	// Give wings a sharp stall
	    @lift {  // max Cl was 0.9 now 0.45, unchanged initial slope vs AoA
		!key = delete
		//sin(AoA) Cl
		key = 0    0      0    2.0
		key = 0.20 0.40   2.0  2.0
		key = 0.25 0.45   0   -1
		key = 1    0     -1    0
	    }

	// Remove the low-speed thickness of the atmosphere
	    @liftMach { // low-mach lift was 1.0 now 0.3
		!key = delete
		//  Mach# Cl-multiplier
		key = 0.0  0.30  0    0
		key = 0.5  0.33  0.2  0.2
		key = 1.0  0.50  0    0
		key = 1.5  0.30 -0.2 -0.2
		key = 5.0  0.10  0    0
		}
	    @dragMach { // low-mach drag was 0.125 now 0.02
		!key = delete
		//  Mach# Cd-multplier
		key = 0    0.02  0    0
		key = 0.80 0.02  0    0
		key = 1.10 0.40  0    0
		key = 1.40 0.30 -0.5 -0.5
		key =  2   0.20 -0.1 -0.1
		key =  5   0.10  0    0
	    }}}}

 

Edited October 21, 2019 by OHara

[OHara]

On 9/22/2019 at 11:18 PM, OHara said:

The mod FAR did show that simulating airflow around the whole craft is feasible,

That is not quite true.  I had the mistaken impression that nuFAR divided parts into voxels, and used some computationally expensive technique to figure the airflow around the craft (maybe something like the vortex lattice method, I thought) making the CPU do the hard work.

Looking into the code, I see that @ferram4 did the hard work of coding the engineering methods from the middle of the last century to figure aerodynamic forces from various measures of the aircraft shape.  Body drag, lift from wings, interactions between wings, are all handled explicitly.   It seems that each simple computation-heavy method misses some important aspects for some regimes of flight.

What FAR shows us is that, with intricate code, simulating realistic aerodynamic forces, real-time in a game, is feasible.
Star Theory might consider whether FAR's general approach for body drag will be simpler and easier to maintain than the rather complicated drag-cube method of KSP1, which is difficult to maintain (as KSP 1.8.0 shows).

On 9/24/2019 at 3:23 PM, 5thHorseman said:

I see very little utility in planes because I'm always busy making rockets.

If you ever find you've figured everything out with rockets, airplanes in KSP are surprisingly fun, given that it was meant to be a rockets game.

==
Although some people say "what should we keep? Nothing!", I think we want to keep several things KSP1 that gets right:
+ Fins make a rocket fly straighter
+ Pointy things have less drag
+ Wing dihedral gives an airplane roll stability
etc.

Edited October 21, 2019 by OHara
caught the grocers' apostrophe