On the particulars of drag in 1.2pre (measurements done on build 1553)

[Gaarst]

NB: this thread is heavily based on @Yakuzi's excellent thread describing drag of different cones in KSP 1.0.4. This is an unofficial continuation for 1.2, aiming to give an estimate of the changes for the release (it should come soon as I write this, so build 1553 should not be too far from 1.2.0 aerodynamic-wise).

 

For each different category, I built a simple launcher keeping the same structure on each test. The parts that change on each tests are the nose cones, set at different locations depending on the test (top, bottom or on radial boosters), and a ballast tank to keep constant masses for each test regardless of the mass of the cone used (if several cones are used, several ballasts are placed next to each cone and are tweaked separately to keep similar mass distributions on the craft). Note that the ballast tank mass changes with discrete increments so masses are not always identical.

The rocket is launched three times for each test, and the maximal height reached is recorded for each test. The average of the three tests is the value used on the graphs, along with error bars corresponding to 1 standard deviation (calculated over the three tests). Note that the error bars shown do not accurately represent potential statistical errors/noise arising from the reduced number of elements in each sample.

Also note that due to the square powers involved in the calculations of gravity losses and drag, the differences between each height are not linear and should not be used to calculate drag of a part as a multiple of another part's.

 

Lexicon:

Spoiler

None: ballast tank is bare

Octag: cubic octagonal strut

Small cone: small nose cone (0.625m)

NCS: NCS adapter, emptied of all fuel

Std cone: aerodynamic nose cone

Adv cone: advanced nose cone type A

Slanted adv cone: advanced nose cone type B

Tail connect: tail connector type A

Circ intake: circular intake

Shock intake: shock cone intake

Ram intake: adjustable ramp intake

Fairing sharp: 1.25m fairing, set as sharp as can be in a single section

Fairing med: 1.25m fairing, set at medium height

Fairing blunt: 1.25m fairing, set as blunt as can be in a single section

Big cone: protective rocket nose cone Mk7 (2.5m)

Small circ intake: small circular intake (0.625m)

A10 adapter: FL-A10 adapter

A5 adapter: FL-A5 adapter

 

Cones on single core rocket

Rocket used for this test (only the part in red is changed, here the cone at the top of the rocket):

Spoiler

On the graph, we can see that using cones are important to reduce drag. Comparing the results with the ones obtained by Yakuzi in 1.0.4, we notice that the spread is a lot stronger, with heights varying from 14750m to 73750m (to be compared to 21000-46000m spread in 1.0.4), so in 1.2 the drag characteristics of each part have been strengthened, with draggy parts becoming even more draggy and aerodynamic parts becoming even more aerodynamic. For the tests with the lowest values, most delta-V was lost during transonic phase, with acceleration getting reduced by several Gs; while for the least draggy tests, the transonic phase only had very limited effect on the rocket ascent.
Still comparing with the 1.0.4 values, we notice that the drag values for the intakes have been "levelled", they are now almost identical to the nose cones values and the shock cone intake does not stand out as the best way to reduce drag on a craft anymore, but is on par with the highest values of the other nose cones (shock cone intake: 73550m, NCS + small cone: 73760m, advanced nose cone: 73150m, tail connector: 73560m)
We can also notice that sharpness of the part is a good indicator of the drag produced by it, with very strong differences between each type of fairing (73140m, 71980m and 16220m average maximal heights, in the graph order) and between the FL-A10 and FL-A5 adapters (66920m and 16150m respectively); generally, blunt parts produce a great amount of drag while sharper parts only produce reduced drag (note that the intakes do not follow this observation). This statement has its limits, as shown by the values for the advanced nosecone (73150m) and aerodynamic nose cone (67250m) being rather close, and by the values of the sharpest and medium fairings in a similar fashion.
Matching part diameters also seems to be very important to reduce drag, with parts smaller than the 1.25m stack they are attached onto producing great amounts of drag (octagonal strut, and other 0.625m parts): linear occlusion does not occur for parts with different diameters. The 2.5m nosecone also produces more drag than smaller cones, but remember that rockets follow a square-cube law (drag varying with square of size, while mass varying with cube of size), therefore the greater drag (force) with only produce a small decrease in acceleration for a larger, and heavier, rocket.
Leaving empty nodes does not seem to be an issue, with values for the lone NCS (72830m) and FL-A10 (66920m) adapters being very close to values obtained for node-less cones (73760m for the NCS adapter + small nose cone). Again the square-cube law can be used to explain this, and one can predict that an open 0.625m node on a 0.625m stack will produce effects similar to an open 1.25m node here (see "None" value: 14760m).
Putting octagonal cubic struts on top of an empty 1.25m node (not surface-attached) does not reduce drag at all. The value for the octag (14750m) is even slightly lower than the value for the lone part (14760m), but this can be explained by discrete mass increments in the ballast or more simply statistical noise and one can not conclude to a drag increase Then again, linear occlusion does not apply to different diameter parts.

 

Cones on radially attached parts

Rocket used for this test (only the parts in red are changed, here the nosecones on the side stacks):

Spoiler

Similarly to results obtained previously, putting cones on the radial boosters helps a lot with drag, with blunt rockets being very affected by transonic drag.
Offsetting the radially attached parts inside the core to reduce the rocket span has no effect on the drag experienced by the rocket, with similar results obtained for "None" and "Clipped + none" columns (14470m and 14500m respectively). The small difference between the two values can be justified by statistical noise, again, or different radial mass/force distribution leading, in the first case, to increased chances of small trajectory deviations.
Using slanted nose cones on the radial part does not improve the drag compared to regular nose cones (68510m for slanted, 69500m for regular) but, on the contrary, tends to increase it slightly. Then again, the position of drag vectors origins (closer or further from the centre) can explain this discrepancy: the slanted cones put more stress towards the centre, making the rocket more likely to change its course by a few degrees and to therefore lose some height.
The shock cone is still an excellent alternative to regular nose cones, for spaceplane use.

 

Cones on core and radially attached parts

Rocket used for this test (only the parts in red are changed, here the nose cones on the core and the ones on the side stacks):

Spoiler

Spoiler

3 None: no nose cones on the core and radial stacks

Adv+2std: one advanced nose cone on the core and 2 aerodynamic nose cones on the radial stacks

Std+2adv: one aerodynamic nose cone on the core and 2 advanced nose cones on the radial stacks

3std: 3 aerodynamic nose cones on the core and radial stacks

3adv: 3 advanced nose cones on the core and radial stacks

None+adv: no nose cone on the core and 2 advanced nose cones on the radial stacks

None+std: no nose cone on the core and 2 aerodynamic nose cones on the radial stacks

Adv+none: one advanced nose cone on the core and no nose cone on the radial stacks

Std+none: one aerodynamic nose cone on the core and no nose cone on the radial stacks

The main objective of these test series was to test radial occlusion: putting a blunt nose cone on the top central stack to deviate airflow away from the lower radial stacks, thus producing even less drag than when using a sharp nose cone on the central stack. This was used to protect the Space Shuttle's wings from plasma flow from the cockpit (a blunt nose cone was used to avoid redirecting the plasma flow directly on the wings); it has also been used for several ICBMs as "drag-reducing spikes" (a small spike is creating a shock ahead of the missile nose, deviating air flow and allowing blunter shapes without increasing drag).
Unsurprisingly, as airflow is still not modelled in KSP, this effect does not occur and using a blunter nose cones does not reduce drag but still increases it. Should this effect have been added, we would have expected the "Std+2adv" column to be higher than the "3adv" one (or "3std" higher than "Adv+2std") but this is not true (68810m and 71740m). Only the drag of each individual nose cone matters, and the craft does not have to be considered in its entirety: reducing the drag on each stack separately is enough to reduce the drag on the entire rocket as much as possible.

 

Cones on the bottom of a single core rocket

Rocket used for this test (only the part in red changes, here the nose cone under the core):

Spoiler

Even though adding a nose cone on an empty bottom node still reduces the drag, the effect of this trick seems to be weaker in 1.2 than in 1.0.4: here is a spread of 28020-30440m, against a spread of 27000-30500m in 1.0.4. (Note that I used an empty node on the bottom of the fuel tank while Yakuzi used the empty node on the bottom of the Rapier engine)
The reduction of drag being minimised, cones that have different drag values overall end up giving very similar values for average maximal height reached (aerodynamic nose cone: 30310m, advanced nose cone: 30440m, and shock cone intake: 30370m). Note that here again the shock cone intake "nerf" is visible, as it is not the best choice in this case either, as opposed to 1.0.4.

 

Summary

• Nose cones are essential to overcome transonic drag
• Don't put a 0.625m nose cone on a 1.25m as it will have close to no effect: linear occlusion only happens for matching sizes
• If it looks draggy, it is draggy: avoid blunt designs
• The shock cone intake is not the magical solution to drag anymore
• Offsetting parts does not change their drag
• Slanted nose cones do not improve drag on radial boosters
• Radial occlusion is not a thing: still avoid blunt designs even with radial boosters
• Sticking a cone under an empty node is useful, but not essential

I hope this will be helpful to at least some. Feel free to discuss these results below.

 

Appendix A (few additional measurements):

Spoiler

I tested the shielded docking port and NCS adapter + SAS nose cone on a single core stack to see effects of detached shockwave. As predicted, the results suggest they do not exist in KSP: the NCS adapter + SAS nose cone assembly had an average of 73350m which is slightly lower than the value for NCS adapter + small nose cone (73760m); the shielded docking port had an average of 33770m putting it in significantly higher than the "empty" node (14760m) but also a lot lower than the most aerodynamic cones (around 72000-73000m for most, 67250m for the aerodynamic nose cone).

 

Using a small nose cone under an empty node does have an effect, but as predicted, and like the other cones, this effect is rather limited. The average maximal height was 28640m, to be compared to 28020m for the empty node and around 30400m for other nose cones.

 

Finally, I tested the new overhauled engines, in order to compare the boat-tail and bare engine versions. The rocket used was similar to the one used in the first part, but because of the engine specs change they must not be compared.

The different engines do have different drag characteristics but the overall difference in performance is not extremely significant. The boat-tail version averaged 50620m while the bare engine version averaged 48410m. The difference is of the same order of magnitude as the difference measured when adding or removing a nose cone under and empty node. Note that as the part overhaul is still a WIP, and as the released version of this mod is incomplete, the difference in drag is likely to change significantly in the future.

 

Edited October 3, 2016 by Gaarst
Added Appendix A

[I_Killed_Jeb]

Thank you for this excellent info

[fourfa]

well done, thank you

[Foxster]

Nice job. 

One conclusion I get from it...If you want a good nosecone then use a circular intake because it has the same low drag as the other best parts but weighs significantly less. This has been true for the last couple of KSP releases as well. 

Also, if you want a low drag and low mass intake then for subsonic flight, again, the best is the circular intake. For supersonic then the best is the ram intake. 

One thought...Would be interesting to see how the shielded docking port and pointy aerial do these days with their detached shockwave effect. 

Edited October 3, 2016 by Foxster

[Kertech]

I was wondering whether the detached shockwave affect would interfere with aeroheating, would be interesting to measure heating of a part at a given altitude and speed, probably beyond my time constraints though! 

[KerikBalm]

Well, you showed 0.625m nose cones on 1.25m stacks do almost nothing now... but what about for the bottom nodes? does a rapierspike do anything now? (where we define a rapier spike as using the small nose cose)

[Streetwind]

Nice overview!

On the topic of the last test, with the cones on the bottom - it would be interesting to see the effect of the boattail variant on Porkjet's new engines. Do one test with Porkjet's skipper bare and one with the boattail, perhaps?

[swjr-swis]

6 hours ago, Kertech said:

I was wondering whether the detached shockwave affect would interfere with aeroheating, would be interesting to measure heating of a part at a given altitude and speed, probably beyond my time constraints though! 

As part of creating a submission for a challenge that mandated using low heat-tolerance Mk1 parts, I found that placing a linear RCS port in front of an Mk1 cockpit allowed me to push it harder through the sky before it would explode from heating. So I would say that it still works, at least for heating, not sure about drag.

[Gaarst]

9 hours ago, Foxster said:

Nice job. 

One conclusion I get from it...If you want a good nosecone then use a circular intake because it has the same low drag as the other best parts but weighs significantly less. This has been true for the last couple of KSP releases as well. 

Also, if you want a low drag and low mass intake then for subsonic flight, again, the best is the circular intake. For supersonic then the best is the ram intake. 

One thought...Would be interesting to see how the shielded docking port and pointy aerial do these days with their detached shockwave effect. 

I'll try but I don't think it will do anything particular. I'd say the SAS cone thing has the same (or slightly higher) drag than the small nose cone, and the shielded docking port has low-average drag (lower than any nosecone but higher than the tank alone.

7 hours ago, Kertech said:

I was wondering whether the detached shockwave affect would interfere with aeroheating, would be interesting to measure heating of a part at a given altitude and speed, probably beyond my time constraints though! 

Same as above, I don't think that detached shockwaves exist at all in KSP: if they don't have any effects on aero, they shouldn't have any effect on heat.

6 hours ago, KerikBalm said:

Well, you showed 0.625m nose cones on 1.25m stacks do almost nothing now... but what about for the bottom nodes? does a rapierspike do anything now? (where we define a rapier spike as using the small nose cose)

Didn't think about this. Since the difference between each bottom-attached node is rather small, I don't expect to see any obvious difference between the small nose cone and no attachement, but I'll definitely give it a try (I'll also try the rapierspike with different cones to see if there is any difference with cones under a tank, if I find time).

6 hours ago, Streetwind said:

Nice overview!

On the topic of the last test, with the cones on the bottom - it would be interesting to see the effect of the boattail variant on Porkjet's new engines. Do one test with Porkjet's skipper bare and one with the boattail, perhaps?

Interesting. I somehow completely forgot these even existed when I had them on my screen during the entire tests. I'll fire a few launches (probably with a T45 since that is the engine I have already been using).

[Sresk]

I've recently been poking under the hood of some mods and i've noticed that the dragCoeff value on parts is quite high compared to real life. Game values tend to be 0.4-0.6 where as real values for streamlined bodies tend to me much lower with values as low as :

0.001

laminar flat plate parallel to the flow

0.021

F-4 Phantom II(subsonic)

0.095

X-15 (Not confirmed)

I was wondering if these valeus are off from real life by an order of magnitude do to the screwy in game physics or if kerbals just suck at building streamlined bodies. 

[fourfa]

My feedback on looking at Rapier tail attachments for SSTOs - you will want to do a standardized test of an SSTO in mostly horizontal flight. The test in this thread is partially a test of simple mass, unless balanced with (for instance) dead weight in a cargo bay to make all tests have strictly the exact same mass. When I did some quick flight tests on a spaceplane of mine, the drag changes dominated over mass changes. Unfortunately these weren't scientific or well documented. 

Test a small spaceplane with minimal external drag, use mechjeb or similar pilot mod to hold a fixed AoA like 5-10 degrees from wheels-up, prevent the Rapier from switching mode, and look at max speed and altitude gained - should work as it did in my tests. Any pilot mods working in 1.2 yet?

[Gaarst]

48 minutes ago, fourfa said:

The test in this thread is partially a test of simple mass, unless balanced with (for instance) dead weight in a cargo bay to make all tests have strictly the exact same mass.

As I explained somewhere in this wall of text, the FL-T100 tanks in the rockets are exclusively ballast tanks. Their fuel is not used by the engine and their only use is to balance the masses and their distribution when adding different nose cones. Even though I can't balance the rocket mass to a 1kg accuracy, the masses have never deviated of over 10kg from the "empty" craft (over a wet total of 5-10t depending on the tests), that makes a variation of under 0.2% of the mass of the whole rocket (bring that up to 0.5% max if you consider dry masses). I'd say this is not a test of mass. 

[Gaarst]

So I did a few more tests that were suggested to complete the ones I already had:

I tested the shielded docking port and NCS adapter + SAS nose cone on a single core stack to see effects of detached shockwave. As predicted, the results suggest they do not exist in KSP: the NCS adapter + SAS nose cone assembly had an average of 73350m which is slightly lower than the value for NCS adapter + small nose cone (73760m); the shielded docking port had an average of 33770m putting it in significantly higher than the "empty" node (14760m) but also a lot lower than the most aerodynamic cones (around 72000-73000m for most, 67250m for the aerodynamic nose cone).

 

Using a small nose cone under an empty node does have an effect, but as predicted, and like the other cones, this effect is rather limited. The average maximal height was 28640m, to be compared to 28020m for the empty node and around 30400m for other nose cones.

 

Finally, I tested the new overhauled engines, in order to compare the boat-tail and bare engine versions. The rocket used was similar to the one used in the first part, but because of the engine specs change they must not be compared.

The different engines do have different drag characteristics but the overall difference in performance is not extremely significant. The boat-tail version averaged 50620m while the bare engine version averaged 48410m. The difference is of the same order of magnitude as the difference measured when adding or removing a nose cone under and empty node. Note that as the part overhaul is still a WIP, and as the released version of this mod is incomplete, the difference in drag is likely to change significantly in the future.

[Foxster]

19 minutes ago, Gaarst said:

... to see effects of detached shockwave. As predicted, the results suggest they do not exist in KSP...

Oh, they do exist. Or at least they do in 1.1 and I wouldn't have thought they would have been removed for 1.2. I've tested it and it has also been confirmed by the dev team. It's just likely their effect didn't kick in during your testing. It seems most pronounced with supersonic flight on, say, an SSTO. 

If I might suggest one more thing to test?...Rotate a nosecone by 180° so the blunt end is uppermost. In 1.1 this reduced the drag of a stack, maybe by reducing the drag at the top and at the bottom? *shrugs*

[Gaarst]

Just now, Foxster said:

Oh, they do exist. Or at least they do in 1.1 and I wouldn't have thought they would have been removed for 1.2. I've tested it and it has also been confirmed by the dev team. It's just likely their effect didn't kick in during your testing. It seems most pronounced with supersonic flight on, say, an SSTO. 

So what effects would it have? If I could design something that exaggerates this effect I could measure it more accurately, but using simple rockets I haven't noticed anything.

Quote

If I might suggest one more thing to test?...Rotate a nosecone by 180° so the blunt end is uppermost. In 1.1 this reduced the drag of a stack, maybe by reducing the drag at the top and at the bottom? *shrugs*

I will try it.

[Norcalplanner]

Great data. Any chance of seeing results for parachutes instead of nosecones?

I remember NathanKell's thread about the detached shockwave - mainly it reduced aero heating at high speeds. Probably not going fast enough with these rockets to make a difference.

[fourfa]

2 hours ago, Gaarst said:

As I explained somewhere in this wall of text, the FL-T100 tanks in the rockets are exclusively ballast tanks. 

Ah deepest apologies - I missed that first time.  Carry on then.

Edited October 3, 2016 by fourfa

[Gaarst]

2 hours ago, Norcalplanner said:

Great data. Any chance of seeing results for parachutes instead of nosecones?

I remember NathanKell's thread about the detached shockwave - mainly it reduced aero heating at high speeds. Probably not going fast enough with these rockets to make a difference.

Parachutes and proper Rapierspike tests will depend on my free time and willingness to repetitively launch dozens of rockets and watch them go up.

I'm interested in the detached shockwave mechanics since I appear to never have noticed it when playing. Pinging @NathanKell hoping for a link to the thread you mentioned.

[swjr-swis]

17 minutes ago, Gaarst said:

I'm interested in the detached shockwave mechanics since I appear to never have noticed it when playing.

Like I said, in my message earlier in this thread: I have, while designing craft for the express purpose of surviving the highest possible speeds for a forum challenge. It works for heat at least.

I can't say if it definitively reduces drag, because I had to solve the heat problem before drag became an issue: the mandatory Mk1 cockpit was exploding from heat well before drag set the speed limit. So for all I know, it affects both heat *and* drag, or maybe reducing drag is how it solves the heat problem. All I know is it keeps an Mk1 cockpit safe from heat for longer and up to significantly higher speeds, allowing me to pursue the design towards higher speed records.

A quick and easy test setup that shows a clear difference:

Spoiler

Mk1 cockpit, probe core, reaction wheel, battery, TVR-2160C 1-to-4 adapter, 4x precooler, 4x RAPIER, 4x tail fins. Set SAS to stabilize, throttle 100%, stage, let it fly without input.

Between 15-16km, the cockpit will blow up from heat. I only did a handful tests, but it blew up every single time.

Only difference: a linear RCS port radially attached and centered on the tip of the cockpit.

Again only a handful of tests, but it survived every single time, and with still a significant margin.

Conclusion: the detached shockwave of the RCS port causes the cockpit behind it to suffer less heat (and perhaps, also less drag, but this test cannot prove that).

 

[problemecium]

Perhaps these changes in 1.2 will finally cure the plague of asparagus rockets with flat-topped cylindrical tanks bulldozing their way through the atmosphere. ;P

[CdrFuzz]

Cool science project! Thanks for the info.

[NeverCast]

Just confirming that by no occlusion you mean I cannot hide a part behind another part and expect the front part to take the drag? i.e. a radially attached engine hiding behind a slanted body part?  If not, that's rather unfortunate. Also interesting that offsets have no effect. Since that's what I've been trying to do. 

[Foxster]

4 hours ago, NeverCast said:

Just confirming that by no occlusion you mean I cannot hide a part behind another part and expect the front part to take the drag? i.e. a radially attached engine hiding behind a slanted body part?  If not, that's rather unfortunate. Also interesting that offsets have no effect. Since that's what I've been trying to do. 

Correct. A kind of umbrella effect isn't modelled. The wind sails straight through parts unless adjacent parts are the same size or joined with a cone-shaped tapering part. 

[NeverCast]

14 hours ago, Foxster said:

Correct. A kind of umbrella effect isn't modelled. The wind sails straight through parts unless adjacent parts are the same size or joined with a cone-shaped tapering part. 

Thanks for that clarification. Will have to redesign a few of my craft. Do you know if FAR/NEAR model the umbrella effect? (Though neither are updated yet).

[magnemoe]

On 3.10.2016 at 11:18 PM, swjr-swis said:

Like I said, in my message earlier in this thread: I have, while designing craft for the express purpose of surviving the highest possible speeds for a forum challenge. It works for heat at least.

I can't say if it definitively reduces drag, because I had to solve the heat problem before drag became an issue: the mandatory Mk1 cockpit was exploding from heat well before drag set the speed limit. So for all I know, it affects both heat *and* drag, or maybe reducing drag is how it solves the heat problem. All I know is it keeps an Mk1 cockpit safe from heat for longer and up to significantly higher speeds, allowing me to pursue the design towards higher speed records.

A quick and easy test setup that shows a clear difference:

  Hide contents

 

Mk1 cockpit, probe core, reaction wheel, battery, TVR-2160C 1-to-4 adapter, 4x precooler, 4x RAPIER, 4x tail fins. Set SAS to stabilize, throttle 100%, stage, let it fly without input.

 

Between 15-16km, the cockpit will blow up from heat. I only did a handful tests, but it blew up every single time.

Only difference: a linear RCS port radially attached and centered on the tip of the cockpit.

 

Again only a handful of tests, but it survived every single time, and with still a significant margin.

 

Conclusion: the detached shockwave of the RCS port causes the cockpit behind it to suffer less heat (and perhaps, also less drag, but this test cannot prove that).

 

Did some test of this. MK1 cocpit, 1.25 meter probe core 90 liter fuel tank and T45 engine. Full trust.
Without the linear port max speed was 538 m/s and attitude 16133 meter, with it was 531 m/s and 15651 meter, removed monoprop for the linear port launch to compensate for weight.

The linear rsc port handle 2600 degree who is 600 more than the cocpit, it looks like it shield the cockpit from much of the heating, my guess is that with its small mass it manages to dump lots of its heat to the cockpit, keeping it colder. 
Replicated your test with the four rapier engines with the same result, it works, you can just as well point it forward so you can use is as an rcs port too  
So an small part who can handle lots of heating protects attached part from overheating with an slightly increase in drag.