What if aero drag occlusion worked like heat occlusion?

[Foxster]

One of the last physics anomalies in KSP is with drag. It just works kinda weird.

If you stack same sized (and taper matching) parts then their drag is reduced. Otherwise air passes straight through leading parts and causes drag on parts behind. For instance, you can have a huge deployed heatshield out front and all the parts behind it are subject to drag as though it wasn't present. 

I get that leading parts in RL are not going to eliminate the drag of the parts behind completely but they reduce it to some appreciable extent.

Which brings us to KSP heat occlusion. It works a bit more logically. If you have a big heatshield out front then all the parts behind it in a cylinder back from its edges are protected from heat. 

So, what if the aero drag occlusion in KSP worked in the same way as heat occlusion?     

I get that this isn't a complete RL-like solution but it would surely be better than what we have now.

Knowing little about such things, I'd guess that it would be a leg-up to re-use part of the heat occlusion mechanism. 

I can imagine it would really mess with existing aircraft though as things like rear control surfaces could be occluded. So would it be a nightmare to make such a change now?   

[OHara]

Occluding airflow from anything in the shadow of another part, from the direction of the oncoming air, same as they did for shock heating, does seem an obvious thing to try.  Maybe Squad did try this, alongside the connected-face occlusion method that we have now, while working on version 1.0.5

That would avoid the complexities of shielding in closed cargo bays, because anything that fits inside the walls of the 'drag cube' (in fact almost always a rectangular prism) of the enclosing part would be occluded.  Many parts now protrude beyond the heat-occluding walls of their 'drag cube' in the part database, so those configurations would need a review, or else players will place parts that look enclosed but are really exposed.

Probably lift would need to be occluded as well, or else the mechanism encourages tucking wings behind a bulging forward airframe.  Then, as you say, the main wing on a conventional aircraft would sometimes shadow the tail, giving an interesting effect in handling. 

Occluding everything in a shadow would strongly encourage players to tuck all surface-mounted parts inside the cylinder of the rocket, or especially of the space-plane.  One nice thing (for game-play) about the connection-based drag model is that there is no incentive to tweak placement of non-aerodynamic parts for drag; so we're not tempted to tuck parachutes in just a little further; but then again we'd like to get some benefit from tucking batteries fully under a nose-cone.

You know that I think (link) that a lot of the weird aspects of the drag model are repairable.  

If there is a change, maybe you'd like to go all the way to 3D flow around the craft, like the Ferram Aerospace mod.  The goal of that mod was realism (somewhat nice for a game) but it ends up presenting the player with reasonably simple rules to get good performance (very nice for a game).  Well, area-ruling isn't a simple rule, but it is a physical rule for which we can develop an intuition (as opposed to the rule to stack 1.25-meter parts to avoid big transitions in their diameters as in PartDatabase.cfg.).

Edited July 29, 2019 by OHara

[Loskene]

Stock FAR yes please, it just feels weird designing planes without it now. How the plane's surface features and shape appear should affect how it flies, that just feels more intuitive to me and it's generally how I design my planes to look with either aero model.

[steve_v]

On 7/28/2019 at 7:48 AM, Foxster said:

So, what if the aero drag occlusion in KSP worked in the same way as heat occlusion?

I don't think fiddling with the drag-occlusion system is the answer, I expect we'd still see sleek-looking craft behaving in incongruously draggy ways and other weird behaviour no matter how it's tweaked.
The answer IMO is to analyse the overall shape of the craft and use that to calculate it's aerodynamic performance, just like FAR does. As a bonus things like occlusion and cargo bays/fairings, even cargo bays made out of random parts, "just work" without any special consideration.

I can personally attest to the fact that the shape-based aerodynamics approach works. It works amazingly well, and craft built for FAR tend to fly exactly how they look. Got yourself a replica? It'll probably perform much like the real thing... Excepting KSP's other odd deviations of course.

 

5 hours ago, OHara said:

If there is a change, maybe you'd like to go all the way to 3D flow around the craft, like the Ferram Aerospace mod. 

Ooh, me want! Gimme gimme!
 

5 hours ago, OHara said:

The goal of that mod was realism (somewhat nice for a game) but it ends up presenting the player with reasonably simple rules to get good performance (very nice for a game).

Yup . Despite FAR being more complicated under the hood, I for one find it far more intuitive to work with than stock aero. Things that look like planes fly like planes, and I don't have to think about the aerodynamics code at all. What a lovely abstraction of a complex problem.
You can find good design practices in a real-life aerodynamics textbook, without digging about in the part database, and it even has a simulator for those who dislike the Lol'Splosions trial-and-error attitude that infests the game. Wins all around.

 

40 minutes ago, Loskene said:

How the plane's surface features and shape appear should affect how it flies

Damn straight it should.

This was my primary and major disappointment when the new stock errordynamics system was first introduced. I still can't stand it after being spoiled by FAR.

Moving from mass-derived drag (yes, really ) to custom drag values was a huge improvement, but aerodynamic performance is still primarily based on the properties of individual parts rather than the shape of the craft as a whole.
It's also chock-full of corner cases and brittle workarounds, like the protracted ordeal involving cargo bay shielding zones and the much loathed "cannot deploy while stowed" mechanic.

As it stands, building an aerodynamically efficient rocket or SSTO spaceplane in stock aero is all about working with the vagaries of the leaky abstraction Squad has given us - sticking to single stacks, avoiding mk2 fuselage parts, closing open-nodes, clipping wings, etc. None of this is intuitive or even remotely realistic, it's a bunch of artefacts bleeding through from the simplified aero model.

There  have been numerous investigations into the most efficient rocket nosecones to use for example, and it never seems to go the way you would expect from the shape of the parts (the shock-cone intake always wins BTW), because drag values are not calculated from shape, they're baked into the part configs.

FAR is great as a mod and all, but bring-on a stock implementation. I'm all for everyone working from the same foundation regarding aerodynamics (and therefore being able to share craft), but I'd really prefer it to be one that does it right.

[Mr. Peabody]

The present difficulty is that it is too easy to build an aircraft with the stock aero model and too difficult to build an aircraft with FAR. We need somewhat of a hybrid between the two with a better drag model. One must also keep in mind that this game is 70% launching of rockets. We really don't need that to become any more difficult.

[Superfluous J]

4 minutes ago, The Dunatian said:

The present difficulty is that it is too easy to build an aircraft with the stock aero model and too difficult to build an aircraft with FAR. We need somewhat of a hybrid between the two with a better drag model. One must also keep in mind that this game is 70% launching of rockets. We really don't need that to become any more difficult.

I vaguely recall pre-1.0 Ferram made a "FAR-lite" mod as a joke because people said they wanted an easier FAR. I don't remember the details but it was supposed to trick us into thinking we were getting an easy FAR but really we were just getting FAR. I personally loved it.

[Yakuzi]

30 minutes ago, 5thHorseman said:

I vaguely recall pre-1.0 Ferram made a "FAR-lite" mod as a joke because people said they wanted an easier FAR. I don't remember the details but it was supposed to trick us into thinking we were getting an easy FAR but really we were just getting FAR. I personally loved it.

I think it was called NEAR... gotta love Ferram!

[steve_v]

8 hours ago, The Dunatian said:

The present difficulty is that it is too easy to build an aircraft with the stock aero model and too difficult to build an aircraft with FAR.

I don't see any reason FAR couldn't be made easier. IME the main difficulty is the artificially inflated lift values stock has, and that's not a difficult thing to replicate for FAR... Whether it's a good idea is another matter though.

 

8 hours ago, The Dunatian said:

One must also keep in mind that this game is 70% launching of rockets. We really don't need that to become any more difficult.

It's not harder to launch a rocket with FAR at all. You generally need less rocket to make orbit than you do in stock aero due to reduced overall drag.

FAR only makes rocketry more difficult when you're trying to play with "pancake" (i.e. pre-1.0) designs or attempting to lift a payload that wouldn't work IRL either.

Penalising dumb designs isn't a flaw IMO.

[OHara]

There is good reason to believe that a FAR-like aerodynamics model could be adjusted for game-play balance.  Ferram pointed out (here) how KSP chose an unrealistic curve for lift-coefficient-versus-mach#, multiplied by a high pre-factor.  This choice allows a KSP aircraft to perform reasonably well through a very wide speed range, whereas in FAR you want different designs for different intended speeds.  The higher low-speed lift in stock lets us easily fly small aircraft, despite their heavier weight due largely to the unrealistically heavy engines, that seem to be intended to compensate for the smaller planet.  (You can try a more realistic lift-versus-mach, with otherwise stock aerodynamics, with a patch here.)

A 3-D-flow approach to aerodynamics would likely simulate lift-coefficient-versus-mach with an empirical curve, and in maybe KSP 2, that curve could be adjusted for game balance like it was for KSP 1.0.5.   I would hope that KSP continues to put these curves in configuration files, so we can substitute a realistic curve.

[Wjolcz]

I agree