On the particulars of drag in 1.0.4

[Red Iron Crown]

Can I just clarify? If your engine has a node at the rear, attaching anything here will reduce drag.

I assume this is dependent on said attached item not contributing its own aerodynamic drag (ie: being fully occluded by the rest of the craft).

The attached part will contribute drag but will reduce or eliminate the drag from the open node on the engine, which is usually a net win if the attached part is suitable (i.e. a nosecone or similar).

Is it also dependent on it being positioned so as not to occlude thrust? Or does thrust magically penetrate? Does that part have to be clipped into the engine, behind the origin of thrust?

Generally speaking yes, though the Rapier is a special case because its four thrust transforms are not centerline, making it easier to avoid obstructing thrust.

[monstah]

What about the cubic octogonal strut (is that its name?)? It has a very low drag and shouldn't block the thrust significantly.

[Red Iron Crown]

It will leave a draggy open node at the end, which is exactly what we are hoping to avoid. Open nodes are very, very draggy.

[monstah]

Okay, makes sense. I thought being very, very draggy applied to the engine's node since its large and concave, but not to the cubic strut.

[panzer1b]

All i can say is that the current drag system makes rapiers the go-to engine for everything due to that rear node that allows you to add schock cones to (or other things if you prefer).

Also, shock cones and circulars are the only viable intakes as rams have a crapload of drag (makes sense as they arent pointy). Circulars make more sense if you arent going to be in atmo much though, since lighter and they arguably look better (i use them on the nose of most craft as they just look so much cooler then the cone intakes).

Also, radial anything is bad as ive learned since it creates alot of extra drag for little benefit.

My generic fighter airframe is basically a shock or circular intake, followed by a mk1-mk2 adapter, then a short or long cargo bay depending on the payload i need, then a inverted mk1-mk2 adapter, and finally cockpit, rapier, and depending on what i want teh craft to do, either a shock cone behind (looks badass and actually cuts drag+gives free intake air), a jet engine (if i want more thrust, its not the best option, but it still beats having a separate stack of parts to add the engine to), or a nuke tipped with a shock cone if im sending it far out. All other components including fuel tanks will be inside the cargo bay as it disabled crag on anything inside there (ive even shoved ion engine stacks in there before which works wonders as ions do not care about anything behind them (thrust isnt blocked by anything), and well, they create 0 drag.

All i can say is i hope squad at least takes a look at the way teh system is coded, as there are way too many completely counter intuitive things going on. For exapmple, although i know having a exposed flat thing in the rear does induce drag, why is it near the same levels as a part in the front? Drag (again, im no aero engineer so this is speculation) should be mostly based upon frontal cross section, and any radial crap protruding from teh craft. Id also like if radial stuff thats clipped inside had no drag, but we can only hope so much, and thatd be extra calcs especially if its partially clipped in and partially external to a craft. it works, but right now its way to counterintuitive especially the whole exposed rear node thing that i have no idea why it does what it does.

Anyways, a few conclusions ive learned from my own testing:

always stick everything that can go inside into a cargo or service bay, its 0 drag at that point.

always cap ends with shocks or circular intakes (circulars less mass, a bit more drag, lower heat so not the bets for spaceplanes that need to brake at interplanetary).

avoid radially attacked anything unless you have no choice (it needs wings and landing gear, although landing gear can always be stuck inside a cargo bay and clip out when deployed).

keep the fuselage as short as possible (the longer it is the more drag regardless of occlusion since you will expose some of your bottom when climbing).

as few wings as possible since they add drag and do little benefit for SSTOs (ofc you need them at lower speeds and landing, so no planes withotu any wings).

clipping pats is pointless as the drag does not change at all (unless they get clipped into a bay).

do not clip a certain part of a stack into a docking bay (such as a cockpit thats in the stack from engine to rear), as if the stack is broken somewhere along the line, you get 2 parts that are fully exposed and seem to create alot of drag.

Edited August 21, 2015 by panzer1b

[Yakuzi]

I'll make sure to update the first post so all data is available on one page... as soon as I have the time^tm?. Also, please bear with me while I generate/update the various configuration figures.

I've done some further nosecone testing based on RIC's, Foxter's, and Val's comments (as I finally had some time on my hands), and have also evaluated aerodynamic performance of various Mk2 parts. Again, use these results as a rough guide when creating a craft, depending on design/flight profile performance gains will vary greatly.

Comparative Nose Cone Efficiency

Several additional nosecone configurations were evaluated, including scanners, control modules and parachutes. Additionally, a number of nosecones were rotated 180° along the horizontal plane and compared to their counterparts with default orientation. A standardised LV-30 powered rocket was used as reference craft, the various nose cones/air intakes were attached as displayed below:

To be updated.

The maximum height¹ (as displayed in the F3 menu) of each configuration was recorded from three separate launches. The average maximum height and standard deviation (black error bars) of the various configurations are plotted below:

In addition to the observations made in the OP, it appears that the shielded docking port and parachutes only marginally decrease drag in comparison to the cone-less control rocket, so it might be worth stowing these in a fairing (or a cargo/service bay²). Parts with a single 0.625 node (e.g. the Small Cone, Stayputnik, CH-J3 and to some extent the M4435 Narrow Band Scanner) perform considerably better when paired with a NCS adapter (up to 114% increase over the cone-less variant). The M700 Surface Scanner performs nearly as well as the standard cone.

Several cones were rotated 180° along the horizontal plane as previously mentioned by RIC and evaluated as described above. Note: the Inverted variants were constructed from the terminal end of the LV30 engine and subsequently rotated 180° and translated upward as described by drewscriver earlier in this thread. The average maximum height and standard deviation (black error bars) of the various configurations are plotted below:

Parts rotated 180° universally performed better than their counterparts oriented in the default direction (i.e. tapered end up... you know... what makes sense in real life), an effect perfectly explained by Val:

[the nosecone] has effectively been transformed into a tailcone and the aerodynamics treats it as if it was behind a stack, not in front of one, but still treats the rest of the stacks as being behind the nosecone. Super exploity.

Exploity indeed... for instance, the standard nose cone in 180° orientation gains an additional 11km in comparison to its typical alignment. Fairings in 180° orientation are even more exploitative, since they additionally occlude parts from drag. A 180° sharp-ended fairing achieved an average max height of 104.3km, an ~145% increase in max height over the 0° aligned sharp fairing (42.5km) and a >400% increase over the cone-less rocket (20.7km). With an additional shock intake on the bottom LV30 node this rocket attains a max height of >115km.

Inverted rockets constructed from the bottom node of the LV30 engine as proposed by drewscriver did not result in any substantial performance gains over the cone-less control.

NB: Air intakes still generate intake-air if oriented 180°. Cones rotated 90° along the horizontal plane performed considerably worse (data not shown). Parachutes performed substantially better when rotated 180° (up to ~10km max height increase, data not shown) and appeared to be fully functional.

Comparative Drag Analysis of Mk2 Parts

To elucidate the variation in atmospheric performance of Mk2 craft several forum users have been experiencing, various Mk2 parts in several orientations were tested in a similar manner as the nose/tail cone tests described above. A LV-30 powered rocket was used as reference craft, the various Mk2 parts were incorporated as displayed below:

Coming soon^tm!

Craft mass was standardised at sea level as mass_wet=6,663kg and mass_dry=4,663kg for short Mk2 midsections and mass_wet=6,763kg and mass_dry=4,763kg for long Mk2 midsections. The maximum height and maximum speed (as displayed in the F3 menu) of each configuration were recorded from three separate launches. The average maximum height, speed and standard deviation (black error bars) of the various configurations are plotted below:

Substantial variations in drag exist between Mk2 parts which appear identical in shape and dimension. For instance, under the appropriate conditions the Mk2 cargobay and Mk2-1.25m adapter gain an increase in max height of up to 25% compared to their liquid and rocket fuel counterparts. In case of the Mk2 cargobay, this only applies to when the payload is attached to the bottom of bay. Similarly, the Mk2-1.25m adapter max height gains only apply when the tapered end is facing upward, or when it placed tapered end down and subsequently flipped upward 180°. The Mk2 bicoupler performs somewhat better than its liquid and rocket fuel counterparts, and may gain an even greater performance increase when all nodes are covered². Rotating Mk2 parts 90° along the vertical axis does not appear to diminish maximum height.

NB: I have observed similar behavior in Mk3 parts, though proper testing is needed...

Summary (i.e. tl:dr)

• Use a FL-A5, A10 or NCS (ideally) adapters between 1.25m and 0.625m node parts for reduced drag and improved atmospheric flight performance
  
• Nose cones/intakes flipped 180° along the horizontal plane greatly reduce drag
  
• Fairings flipped 180° not only have inherently reduced drag in comparison to their default oriented counterpart, they can also occlude rocket parts resulting in h4x0r atmospheric performance gains
  
• Use Mk2-1.25 adapter parts over non-taped rocket fuel parts whenever you can. Make sure to put the tapered end towards the direction of flight.
  
• Attach payloads to the bottom part of Mk2 cargobays
  

Again, please bear with me while I generate/update the various configuration figures.

¹ The max speed in the F3 menu maxes out at 750 m/s (Krakens-bane?) even though the craft will accelerate further. Accordingly, the speed graph is not included in the LV-30 nose attachment test.

² Not tested!

Edited September 11, 2015 by Yakuzi

[Red Iron Crown]

Fantastic experimentation and visualization of the results, Yakuzi, very nicely done.

One thing I'd mention is that your first test rocket is quite small, which tends to magnify the mass differences between the various nose treatments. For example, switching between a standard cone and a tail connector might mean a 10% change in mass and a 10% change in drag; do the same on a much taller stack and it might be a 2% change in mass for a 9% change in drag (numbers from the hip, but I hope you see my meaning). I wouldn't go so far as to say that mass should be equalized, as that doesn't help the player in real in-game situations, but maybe testing on the largest feasible 1.25m stack that keeps TWR above 1.5 or so might yield other interesting results.

Just a thought, and not meant to devalue your excellent experimental results in any way.

[pearldrumbum]

Do you have a screenshot of the inverted fairing craft? Did you rotate it at the fairing base making the fairing and implied payload clip inside the rest of the ship?

[GoSlash27]

Fantastic experimentation and visualization of the results, Yakuzi, very nicely done.

One thing I'd mention is that your first test rocket is quite small, which tends to magnify the mass differences between the various nose treatments. For example, switching between a standard cone and a tail connector might mean a 10% change in mass and a 10% change in drag; do the same on a much taller stack and it might be a 2% change in mass for a 9% change in drag (numbers from the hip, but I hope you see my meaning). I wouldn't go so far as to say that mass should be equalized, as that doesn't help the player in real in-game situations, but maybe testing on the largest feasible 1.25m stack that keeps TWR above 1.5 or so might yield other interesting results.

Just a thought, and not meant to devalue your excellent experimental results in any way.

Actually, I concur with this. If the comparison is meant to illustrate aerodynamic effects, you really can't get an accurate reading if differences in mass (non-aerodynamic differences) are clouding the results.

Also, it matters whether we're talking subsonic speeds or supersonic speeds since the "side" flow decreases dramatically in this region.

Going back to the tail cone vs. shock cone, the shock cone has less drag at subsonic speeds while the tail cone has less drag at supersonic speeds. How well they will perform in a test like this depends on how fast the rocket is going when it flames out.

I ran into this problem this week myself. I had built 2 different test rigs to compare these parts and found the shock cone had less drag (which I didn't expect from the math). After checking out RICs results and reengineering the rig to hit ludicrous speed, the tail cone was superior.

Best,

-Slashy

Edited September 11, 2015 by GoSlash27

[NathanKell]

It's also worth noting that the square-cube law plays into this. The heavier the rocket, the lower (surface area) / mass ratio it has (usually). That means that while the force of drag scales linearly with area, the *deceleration* imparted by drag scales with the 2/3 power of mass.

Simple example:

10 ton rocket, 1m diameter (Vanguard). Frontal area of .25pi, so let's say drag of 25pi newtons. Deceleration = 25pi / 10 = 2.5pi m/s.

2800 ton rocket, 10m diameter (Saturn V). Frontal area of 25pi, so let's say drag of 2500pi newtons. Deceleration 2500pi / 2800 = 0.8928pi m/s.

And note that's not even a fair comparison, because the Saturn V's volume was mostly LH2/LOX, at 1/3 the density of kerolox...(and was thus rather larger for its mass.)

[Yakuzi]

One thing I'd mention is that your first test rocket is quite small, which tends to magnify the mass differences between the various nose treatments. For example, switching between a standard cone and a tail connector might mean a 10% change in mass and a 10% change in drag; do the same on a much taller stack and it might be a 2% change in mass for a 9% change in drag (numbers from the hip, but I hope you see my meaning). I wouldn't go so far as to say that mass should be equalized, as that doesn't help the player in real in-game situations, but maybe testing on the largest feasible 1.25m stack that keeps TWR above 1.5 or so might yield other interesting results.

Actually, I concur with this. If the comparison is meant to illustrate aerodynamic effects, you really can't get an accurate reading if differences in mass (non-aerodynamic differences) are clouding the results.

Actually, I also concur with this... It was something I contemplated before starting the nose/tail cone tests (the first ones were actually balanced in mass). However, from an engineering point of view I wanted to know if it was worthwhile carrying the extra mass of a heavier but potentially more streamlined part vs a lighter but aerodynamically less favorable. I reasoned at the time that this information would be more useful for the average KSP player. I fully agree that tests with standardised mass are ideal under most circumstances, hence the standardised Mk2 tests above. If I can find some time, I might repeat the cone tests with standardised mass (perchance a 1.1 project).

Also, it matters whether we're talking subsonic speeds or supersonic speeds since the "side" flow decreases dramatically in this region.

Going back to the tail cone vs. shock cone, the shock cone has less drag at subsonic speeds while the tail cone has less drag at supersonic speeds. How well they will perform in a test like this depends on how fast the rocket is going when it flames out.

It's also worth noting that the square-cube law plays into this. The heavier the rocket, the lower (surface area) / mass ratio it has (usually). That means that while the force of drag scales linearly with area, the *deceleration* imparted by drag scales with the 2/3 power of mass.

Fully agreed, the results presented here only provide data on a very specific slice of atmospheric flight. I am currently performing some preliminary tests on supersonic speeds in the lower atmosphere by launching rockets under a more shallow angle ballistic trajectory. However, as I lack any continuous data collection tools* it would be take far too much time and effort to perform any sweeping analyses.

* Do any exist for KSP, if they do please let me know!

Do you have a screenshot of the inverted fairing craft? Did you rotate it at the fairing base making the fairing and implied payload clip inside the rest of the ship?

I'm currently generating the craft configuration images, they should be done tomorrow (my sincere apologies, it's late on my side of the planet). The fairings are indeed clipped inside the ship... so no place for any other payload... unless you put another fairing on the top oriented properly, I'm positive you'd still have significant atmospheric performance gains.

Edited September 11, 2015 by Yakuzi

[GoSlash27]

Yakuzi,

+rep.

None of this is at all intended to take anything away from what you've done here. It's good data and you've presented it very well. It's just the nature of the beast that it's impossible to come up with one single set of data (or one test) to define what is best overall from an aerodynamic standpoint. Everything is constantly changing with mach number....

Best,

-Slashy

[Red Iron Crown]

However, as I lack any continuous data collection tools* it would be take far too much time and effort to perform any sweeping analyses.

* Do any exist for KSP, if they do please let me know!

Mechjeb has some data logging and visualization tools, not sure how easy it is to extract the data to analyze outside the game though. There is also Telemachus, which streams data via browser window.

[Val]

Really outstanding work, Yakuzi. Much deserved +rep. Thank you, for all the hard work.

---

... an effect perfectly explained by Val ...

I explained something perfectly on the internet? That's quite the compliment. Thanks, again.

[RocketBlam]

Wait, are you saying if I have three tanks stacked, and an engine, I should use 4 cones?

[Red Iron Crown]

Wait, are you saying if I have three tanks stacked, and an engine, I should use 4 cones?

No. Tanks in a stack are occluded by the tank above, if they are the same size their front face is ignored completely. Only the top one's front face is important and needs a cone.

[GoSlash27]

A few odds 'n' ends that I have figured out this week:

*Shrouded solar panel arrays are draggier than they look. It's much cleaner aerodynamically to surface mount single panels instead. In fact, even the Gigantor XL is less draggy than the shrouded arrays when stowed and mounted longitudinally.

*RCS blocks are draggy, while the linear ones aren't. Try to use linear ports instead.

*Landing gear are cleaner when mounted backwards.

*Folding ladders are draggy and they don't get any better by clipping. Try to avoid using them.

*The tail cone is draggier than the shock cone intake in the subsonic regime, but much cleaner in supersonic flight. It's worth using on larger spaceplanes where it's mass isn't such a penalty.

*Some parts look cleaner than they actually are. For example, I see this a lot:

This *looks* like it should be clean, but it's really not. KSP treats the entire face of a part as having the same uniform drag coefficient. The bicoupler has an area of 2.92m^2 and drag coefficient of .843 when arranged this way. The shock cones occlude 2.24m^2 of this, but leave .68m^2 exposed at a drag coefficient that's little better than a flat plate.

It'd be better to radial mount mk1 parts and attach the intakes/engines to that, where the intake occludes the entire stack.

*On the subject, the structural intakes are much cleaner than initially reported. They are nearly as clean as shock cones.

Best,

-Slashy

Edited September 13, 2015 by GoSlash27

[Starhawk]

A few odds 'n' ends that I have figured out this week:

*Shrouded solar panel arrays are draggier than they look. It's much cleaner aerodynamically to surface mount single panels instead. In fact, even the Gigantor XL is less draggy than the shrouded arrays when stowed and mounted longitudinally.

That's extremely interesting and unexpected.

*RCS blocks are draggy, while the linear ones aren't. Try to use linear ports instead.

Absolutely. Also the linear ports have double the heat tolerance.

*On the subject, the structural intakes are much cleaner than initially reported. They are nearly as clean as shock cones.

That is very surprising. I've tried them a couple of times and had absolutely no luck. I'll have to give them another look.

Thanks Slashy.

Happy landings!

[Red Iron Crown]

Interesting results, Slashy. I've been using linear RCS ports on spaceplanes because they're easier to balance, good to hear they're less draggy, too.

I'm also surprised at the structural intake being good dragwise, I've been avoiding radially mounted parts when possible due to drag concerns. Guess I can start using them more often.

[GoSlash27]

All,

Please ignore my previous about the intakes. I just tested it out and it doesn't seem to work that way in practice.

At the same mass, the bicoupler intake setup actually has better performance.

Upon further testing, the occlusion is more than I had thought and the math works out.

2 shock cones with a bicoupler represents a flat plate of 1.14m^2

a sharp nose with 2 parallel shock cones is 1.70m^2; way worse.

a single shock cone with 3 structural intakes (adequate) is 1.16m^2; still not quite as good.

For parallel pods, it's better to go with pairs of Mk 1s than the bicoupler. That gets the flat plate area down to .728 vs. 1.14 for the bicoupler.

Sorry for the confusion!

-Slashy

Edited September 13, 2015 by GoSlash27

[Cairol]

Good to hear. Because I love my bi-coupler shock-cone twin-intakes. They look cool and it seems like their drag is not so bad.

What I didn't expect was, that placing a cone at the back of for example a R.A.P.I.E.R will reduce drag. Problem with this is: I don't like to clip parts in an unrealistic manner. :/

Really great information here Yakuzi. *thumbsup*

[GoSlash27]

Interesting!

After digging and experimenting, I've managed to come up with an intake combo that's less draggy than the shock cone :

Tail connector A occluding a precooler with 1 structural scoop attached. This adds a little mass for a complete engine assy (2.4t vs. 2.1t), but it dramatically reduces the supersonic drag.

Best,

-Slashy

[Starhawk]

Interesting!

After digging and experimenting, I've managed to come up with an intake combo that's less draggy than the shock cone :

Tail connector A occluding a precooler with 1 structural scoop attached. This adds a little mass for a complete engine assy (2.4t vs. 2.1t), but it dramatically reduces the supersonic drag.

That should make for some interesting aesthetics.

Thanks again for sharing, Slashy.

Happy landings!

[GoSlash27]

That should make for some interesting aesthetics.

Thanks again for sharing, Slashy.

Happy landings!

Playing with the concept, a 60t+ payload tanker

Not sure if I'm gonna go this way in the final version...

Best,

-Slashy

[Starhawk]

Playing with the concept, a 60t+ payload tanker

Not sure if I'm gonna go this way in the final version...

Nice!

Hmm, I don't see a docking port. Does it have a belly-mounted port or something?

Happy landings!