Drag Model Discussion

[Brandano]

In this line of thought, can I suggest a very rough tweak for the current drag model? Modify aerodynamic fairings and nose cones so that they subtract a certain amount of drag from the part they are attached to if they are leading them relative to the direction of flight (possibly adjusted by multiplying by the cosine of the angle between the velocity vector and the line connecting the two part centers). It's still a silly drag model, but at least it would make aerodynamic nose cones useful.

[NWM]

Dear Ferram!

Yes, it is true: those heavy engines ruin the balllast. I had straggled two weeks with stabilizing my reuseable 400/100tons tanker-SSTO (and now she - powered by those heavy engines - bings over 250tons of fuel to low orbit - realistic?). I had been forced to use those ASAS units to bring back the dry CoM forward, and keep the plane stable! While flying in the atmospere your conception of weighst is absolutely right.

But on the other hand, if we are in space, those changes undermines the ballance of the Kerbal world dV, because those changes drastically improves the full mass/ empty mass ratio. Your changes of weights mean superefficient spacesips over the athmospherically stable rockets and aircrafts!

The main problem - earth-like athmospheric conditions meet the miniature cosmos of the world of kerbal. There is no perfect "realistic" solution, just good trade-off. Your conception - no doubt - improves the conditions in the athmosphere, but there is no compensation for the orbital conditions!

[ferram4]

@Brandano: So, under this model, the best option is to build your rockets out as wide as possible, consisting of 1-tank tall liquid boosters only so that you can get as many nosecones attached as possible and maximize the drag reduction. This actually sounds like it'll encourage even less aerodynamic designs than we currently have, which means that as a quick-fix it's a bad idea since it will lead to players getting even further from proper rocket design with a good aerodynamic model.

@NWM: I think you're misunderstanding what I'm saying, and you've done it twice. Decreasing the mass of engines and increasing the dry mass of fuel tanks a proportional amount leads to the same full and empty mass overall, which means no change in dV for the proportional cases and only very small changes in dV for other cases (generally leading to less dV, since the best possible full/empty ratio is lower). As an example of what I'm saying, let's use a Mainsail + 3 Jumbos as our stack:

Current Masses:

Dry: Mainsail (6t) + 3 * Jumbo (4t) = 18t

Full: Mainsail (6t) + 3 * Jumbo (36t) = 114t

So, let's cut the mass of the Mainsail in half and add the difference to the Jumbos equally so that the dry mass stays the same. Leave the amount of fuel as it is.

Adjusted Masses:

Dry: Mainsail (3t) + 3 * Jumbo (5t) = 18t

Full: Mainsail (3t) + 3 * Jumbo (37t) = 114t

Same full and empty masses, same dV out. This is what I want, it just shifts the CoM further into the fuel tanks rather than being pulled down as much by the engine.

[NWM]

@Ferram4

I've understood. I've been thinkig of this already. I am not trolling, but I wanna show you the opposite plate of the scale.

But I assure you, the 3xJumbo+mainsail configuration will get small stability if you steal 3 tons from the bottom and dispense among the tanks. This constuction remains instable while the 3x32t=96t of fuel is consumed from the top, and the decreasing the fuel level moves down the center of mass. At 1/3rd fuel level all of 32tons of fuel is in the bottom tank, and that makes the rocket instable, not that 3 tons of extra engine weight! The salvation is in those fuel ducts or a kind of fuel ballancer! A good fuel ballancing can add far more to the stability.

And continue the math: 1 Poodle(2.5t) + 1/2 Jumbo(18/3t) = 1 Poodle(2.t) + 1/2 Jumbo(18.5/3.5t)

In this case the Mainsail looses 50%, the Poodle 20%?!

PS: have you got a FAR version without extra-engine-drag for testing and helping to find a solution?

Edited February 12, 2014 by NWM

[Nemrav]

One solution to the transition problem would be to add the new system as at first a beta, then an option in the game. And finally have a little update that simply adds a splash saying "the new aerodynamics system is now set to be fully implemented in the next update, practice and get used to it now". maybe add tutorials that explain how to tweak stuff to the new system and other transition stuff. Or simply keep it as an option in the settings

[ferram4]

@NWM: But all of that should be counteracted by the effect of the payload on top of the rocket, which will become a much larger portion of the rocket's mass and should help pull the CoM forward.

If you want to play around with things, there are two values in the config.xml that you need to change: set incompressibleAttachDrag to 0.01 and sonicAdditionalDrag to 0.2, and then it will be realistic levels of drag.

@Nemrav: That sounds like a good way to have extra code bloating the game for no reason, especially if you have to put the work in to set up a popup to inform people that you're going to remove an option. You'll also never convince people to let you remove an option once you give it to them.

IMO, the best way to handle aerodynamics options is to put a button labeled "Disable Aerodynamics" next to "Infinite Fuel" and "Hack Gravity" in the debug menu. If people don't want to deal with aerodynamics, they don't have to; they'll just have to own up to the fact that they're using the debug menu to get past things that they can't handle.

[NWM]

@Ferram: Yes - the terran payloads ~ 10% of the full mass, the kerbal(FAR-ed ~ 4000m/s dV) payloads ~ 20%. Kerbal ones should be even less instable!

I'll count and try some inordinate vessel and make a report.

PS: 12 hours ago both of us made a mistake: the empty weight of jumbo is 8 tons!

[ferram4]

No, I read that straight from the config file; it's 4t empty. Also, all the fuel tanks (with the exception of the Oscar are 8/9 fuel and 1/9 structure, which is consistent with 4t dry, 36t full.

[Captain Sierra]

This reminds me that I really should make a modulemanager file that sets all fuel tank dry masses (and engine masses, and other stuff) to real world numbers and then release it just to let people see what crap we're really dealing with. I imagine it will be interesting.

As for the pertinent discussion, how are wings handled? in stock KSP, I don't believe they generate any lift from Bernoulli's principle, which is why you must ALWAYS maintain an angle of attack over prograde to not fall out of the sky. Now, Ferram, how does FAR handle that? I haven't flown with it in ages.

[wizzlebippi]

All wings must maintain angle of attack (AoA) to produce lift. On real aircraft, the wing is mounted at some incidence angle to the fuselage that usually corresponds to the AoA needed for cruise flight. This allows the fuselage to stay level while the wing has some AoA, minimizing drag.

KSP seems to keep the fuselage and the wing at the same angle of incidence, requiring some angle of pitch to give the wing AoA.

[Captain Sierra]

All wings must maintain angle of attack (AoA) to produce lift. On real aircraft, the wing is mounted at some incidence angle to the fuselage that usually corresponds to the AoA needed for cruise flight. This allows the fuselage to stay level while the wing has some AoA, minimizing drag.

KSP seems to keep the fuselage and the wing at the same angle of incidence, requiring some angle of pitch to give the wing AoA.

It's not entirely that, because I've tried mounting my wings at 5* AoA, and I still have to hold pitch at 5-10* to stay airborne. I've flown Cessnas and those wings are no more than 10* AoA and even fly slightly nose down if you can get some speed (which isn't easy in those planes mind you, they ain't a P51 by a longshot).

[NWM]

@Ferram4

You are again right: 4 tons... (auch! I wish knew where I'd been fobb!)

I am on the testing. Some expected(better acceleration, inline fuel ballancing keeps the rockets stable) and some unexpected results(the transonic stall starts at 250m/s instead of 280m/s)

Another question: do you want to decrease the weights of jets too, or only the rocket engines?

[majic79]

As for the pertinent discussion, how are wings handled? in stock KSP, I don't believe they generate any lift from Bernoulli's principle, which is why you must ALWAYS maintain an angle of attack over prograde to not fall out of the sky.

I think you'll find that bernoulli's principle doesn't dictate how wings produce lift, for that you always require Angle of Attack (Bernoulli's principle does explain why you get clear air turbulence behind an aircraft, as it produces a rotating flow around the wing which generates a vortex)

I'm sure that I've seen wings mounted with an inherent angle of attack on stock KSP (I watched over my mates shoulder when he was playing on his mac and he hit a key to change the angle of a single wing)

[Brandano]

Bernoulli is often misquoted. The "equal transit time" explanation of lift generation fails miserably when considering symmetrical wing profiles, and even on asymmetrical wing profiles it doesn't explain why in the real world the air on the upper surface reaches the trailing edge before the air from the lower surface does, rather than in the same time. Or why at positive AOA the stagnation point travels back from the wing leading edge on the lower surface, meaning that some air will hit the lower surface, travel in a direction opposite the relative wind to the upper surface and then back along the upper surface. Wings create a pressure difference between the upper and lower surface. And displace a volume of air in the direction opposite lift. What is the cause and what is the effect is hard to define properly, but the system as a whole does definitely both things.

[Nemrav]

Idk maybe another system could be something like speed * angle * SA of the rocket (projection from side that is leaning) * 0.25

Don't know how FAR works, but its awesome and in my opinion much easier to use than stock KSP.. + more realistic

[vexx32]

Bernoulli is often misquoted. The "equal transit time" explanation of lift generation fails miserably when considering symmetrical wing profiles, and even on asymmetrical wing profiles it doesn't explain why in the real world the air on the upper surface reaches the trailing edge before the air from the lower surface does, rather than in the same time. Or why at positive AOA the stagnation point travels back from the wing leading edge on the lower surface, meaning that some air will hit the lower surface, travel in a direction opposite the relative wind to the upper surface and then back along the upper surface. Wings create a pressure difference between the upper and lower surface. And displace a volume of air in the direction opposite lift. What is the cause and what is the effect is hard to define properly, but the system as a whole does definitely both things.

Unfortunately, completely incorrect. The pressure differential is an often-quoted reason for wings producing lift, but if you take the time to research it properly, the idea has been thoroughly debunked many times. The reason wings work is that their camber (shape) deflects air downward. By the 'equal and opposite reaction' law of motion (was that the second one? I'm terribly rust on those), deflecting air downwards forces the wing upwards, thus creating lift. This is why stunt planes can fly equally well upside down, when they have a vertically symmetrical camber and there can be no possible pressure difference between the top and bottom of the wing, even if that principle was correct; they still deflect air downward.

[wizzlebippi]

Unfortunately, completely incorrect. The pressure differential is an often-quoted reason for wings producing lift, but if you take the time to research it properly, the idea has been thoroughly debunked many times. The reason wings work is that their camber (shape) deflects air downward. By the 'equal and opposite reaction' law of motion (was that the second one? I'm terribly rust on those), deflecting air downwards forces the wing upwards, thus creating lift. This is why stunt planes can fly equally well upside down, when they have a vertically symmetrical camber and there can be no possible pressure difference between the top and bottom of the wing, even if that principle was correct; they still deflect air downward.

Actually both theories are only part of how wings produce lift. If you really want to get technical, the reason (mathematically) why wings produce lift is because of circulation. Lift equals air density times forward velocity times circulation (L=ÃÂVΓ). Circulation is centered around the maximum thickness of the airfoil (Usually between 25-30% of the wing chord in subsonic flight). This shows up in Bernouli as the force causing air on top of the wing to be faster than the air below the wing. Circulation also extends far beyond the wing, causing upwash infront of the wing and downwash behind it, as vexx32 described. Finally, becuase a wing is a 3d object, higher pressure air below the wing curls up around the wingtip. This causes the center of circulation to move backwards as it approaches the wingtip resulting is wingtip vorticies and cool pictures.

Edited February 19, 2014 by wizzlebippi

[pizzaoverhead]

This page has a nice visual analogy for anyone having difficulty understanding lift.

[Dunbaratu]

Could a model halfway between "only occlude the parts directly attached" and "occlude everything in a straight line behind the part" be achieved by having parts occlude a "wedge" behind them that tapers to a point? In other words an object might occlude another object 1 meter behind it, but not, say, 3 meters behind it. Basically this simplified model would make it so that the air "comes back together" a short distance behind the leading object and starts causing drag again. Only a small wedge would be in the lee of the object.

[wizzlebippi]

Why is everyone so determined to have ksp performing CFD? What the dev's really need to be doing is reading Hoerner Fluid Dynamic Lift and Fluid Dynamic Drag for subsonic flow along with NACA TN 1135 for supersonic flow. There's more than enough information there to construct a very good and accurate drag model that doesn't eat the entire CPU. Remember, aircraft/rockets have been around for far longer than computers.

[ferram4]

Yes, and all of that data is very, very easy to find. The problem is determining the proper geometry to apply it to, and making sure that the model works for everything at every Mach number and most importantly, that it's not impossible to build a working rocket with it. It needs to work for slender rockets, fat rockets, slender planes, wide planes, rovers, flying boxes, space stations, crazy motherships, probes and anything that Whackjob might come up with.

I'd go with the USAF Stability and Control DATCOM myself; it's a lot more direct in trying to apply the physics to the shapes rather than using the flow equations directly. It reduces the calculations a lot to do things that way.

[Hodo]

something like a more simple version of far would be nice.

Sorry but FAR is pretty simple. If you have seen a real aircraft, glider, or even made a paper airplane that goes more than 10ft you can make a plane that works in FAR.

[Bessy]

yeah its nice and easy to make a paper air plain fly in far just like in real life... why not try a replicate a real air craft in far

far isnt perfect but its a 1000x better than the current model and i find things that work in rl work similar in far

the devs should think about doing a new drag model before adding any more parts otherwise they most likely going to be doubling the work load having to re write all the parts to work with the new model when they finally go with it if they ever do

its weird most devs would be getting all the background basic under the hud things like a GUi and physics inc drag first before thinking about making a game out of it

they wasting time on tutorials that will not work when they finally do get round to what they should have done first

they rushing a game along because of steam they havent even got a good GUI yet for most things .. yet they plowing on with a mission game that has brought parts out that i will never use as they are pointless like the lab iv deleted it out of my folder because i see no point in it ... this is the sort of crap they said they would never do rush it out

steam is the cause i wasnt bothered to much with what people were saying at first but it turns out they were right steam was a bad idea

all them things they worked on and showed off last year in kerbalsp the old twitch for all you new people ...including that cyan gas giant and mining equipment and dif fuels hmm out with a puff of smoke it seems

can we have a dev team that will stick to a plan and work in a proper order eg base engine stuff like the drag model down first then make a game around it

farram shouldn't be needing to update the FAR mod for version 24 i think ... i dont think iv played with the stock model since he released FAR like many many others that play KSP and for all those people crying but i wont ever get use to it .... well i give my friend jamie a go with the stock game 3 months ago he hated that his planes acted so stupid and said the game was crap ... i told him to wait and dragged the FAR mod in and started it back up ... he loved that he could mess with all the settings and once he had a few crashes and realized where he was going wrong in his plane design he loved the game as the planes pulled Gs and turned correctly etc

if that isnt a gd msg to squad about who they are aiming the game at i donno what is ... yeah some people like when things act how you expect them too makes a game feel much more immersive and enjoyable

rant over

far isnt perfect but id take it over the current stock any day

[sal_vager]

KSP is still a work in progress, so just because things haven't been done yet doesn't mean they won't be, and if the devs did what you suggest none of us would have a game at all in a playable state.

Enjoy what you have so far (pun?) it'll just get better

[wizzlebippi]

Yes, and all of that data is very, very easy to find. The problem is determining the proper geometry to apply it to, and making sure that the model works for everything at every Mach number and most importantly, that it's not impossible to build a working rocket with it. It needs to work for slender rockets, fat rockets, slender planes, wide planes, rovers, flying boxes, space stations, crazy motherships, probes and anything that Whackjob might come up with.

I'd go with the USAF Stability and Control DATCOM myself; it's a lot more direct in trying to apply the physics to the shapes rather than using the flow equations directly. It reduces the calculations a lot to do things that way.

For stability, yes. For lift and drag, there's a far simpler way. At it's simplest, a rocket is a cylinder with a cone on top. Two basic shapes that have a very well understood drag profile, even in combination. There's data for boxes, bowls, struts, and everything else someone felt the need or was bored enough to stick in a wind tunnel. The overall shape isn't terribly important, as long as you can construct it using basic shapes and know their orientation and dimensions. By using known drag data with known dimensions, a drag coefficient can be calculated using a basic drag buildup. From there's it's a basic sophomore year homework problem to calculate drag force. For adding boosters, you simply add a interference drag multiplier based on the strut height and diameter. Wings don't add much complexity, though there's an induced drag factor to consider as well, and they should assume a reasonable craft lift coefficient of something like 1.5.

To implement this in game, the ideal place would be after construction and storing the calculated coefficient for later. Simply take known dimensions, look for fayed surfaces, and calculate a drag coefficient based on known profiles. From there's it shouldn't be much more complex than what KSP's already doing. Wave drag is even easier since you could assume it's a constant multiplier for everything. We don't need absolute realism, just realistic tendencies.

The real catch is all this discussion is trivial unless KSP implements some sort of indicated airspeed system. Currently it's groundspeed only, which is fine for orbital calculations, but makes for poor aerodynamics. This isn't even that hard, because you could assume there's no wind in the Kerbal universe and use ground speed as true airspeed to convert to equivalent airspeed, and with a few assumptions later that's good enough.