0.25 KSP air intakes compared

[MiniMatt]

Which is the best air intake in Kerbal Space Program?

The answer, as with most in KSP, is, of course, it depends.

However, we can compare air intakes with a modicum of scientific rigour. Of course, science - working everything out with a pen, paper, and slide rule - is distinctly un-Kerbal; far better to strap Jeb to a test rig and launch.

So....

Methodology

Version 0.25.0642

Test rig of FL-T100 fuel tank (drained of oxidiser), four LT-1 landing legs, a single TurboJet engine, a MK1 Inline Cockpit, two MK2-R Radial parachutes and variable air intakes.

Launch procedure involved setting throttle to max, enabling SAS, and igniting the blue touch paper.

No controls (or time warp) were touched until apoapsis reached (at which point throttle position was nuked and parachutes enabled), upon landing the maximum altitude reached was recorded via the F3 results screen.

Results

Equipped with a Circular Intake the test rig reached an altitude of 128,981 metres.

Equipped with a Ram Air Intake the test rig reached an altitude of 141,239 metres.

Equipped with a Shock Cone Intake the test rig reached an altitude of 147,461 metres.

Equipped with a pair of XM-G50 radial intakes the test rig reached an altitude of 149,037 metres.

Equipped with a pair of Structural (radial) intakes the test rig reached an altitude of 103,572 metres.

Conclusion

Not sure we can actually draw conclusions from this experiment - the test rig is rather unrealistic compared to real (in KSP) spaceplane designs. Drag, mass, and speed at varying altitudes are all likely different to a typical space plane design and flight plan.

That said, of the inline intakes, we can perhaps say that the "best" intakes are - in ascending order of proficiency - the Circular Intake, the Ram Air Intake, and the Shock Cone Intake. It is important to note that the difference between the best and worst in this test is only around 10%.

The radial intake results are surprising. So much so that I re-run them, only to record results within 100 metres of the initial recordings. It is, perhaps, safe to say that if you're looking to use radial intakes then the original XM-G50 radial intake might be a better choice than the new (as of v0.25) Structural Intake. And that a pair of XM-G50 radial intakes may well be roughly equivalent to a single instance of the best inline intake.

Final thoughts

Re-iterating - the test rig is rather unrealistic. It provides a way of producing replicable results but it is very far removed from typical space plane designs and flight plans. As such, results may not be applicable to real world designs. Suggested improvements in methodology are very much encouraged.

Edited October 20, 2014 by MiniMatt

[Kulebron]

What about MechJeb orbit ascent? You may try profile 5-35-1-40 to start with. If TWR is very high, reduce the last parameter.

I compared the max air in mechjeb with their mass, and ram air intake seemed to have a little bit better ratio than the cone intake.

Edited October 20, 2014 by Kulebron

[MiniMatt]

Mech Jeb is not a bad call actually. It should provide a way of largely eliminating human variances in a more realistic flight plan and design.

I'll knock together a more plane like test rig tomorrow and figure out a methodology to get some useful results with Mech Jeb.

[Specialist290]

Keep us posted on how it works out

[MiniMatt]

As suggested by Kulebron I've put together a very simple spaceplane and set MechJeb the task of piloting to a 75km circular orbit with various air intake configurations.

Methodology

Version 0.25.0642

Test rig of FL-T400 Fuel Tank (full oxidiser & liquid fuel), a single RAPIER engine, a MK1 Inline Cockpit, a pair of Type B Structural Wings, each wing equipped with an Elevon 3 and an Elevon 2, a rudder consisting of a Type D Structural Wing and an Elevon 4, three Small Gear Bay wheels, the rearmost two affixed via structural Small Hardpoints and four fixed OX-STAT solar panels. To this, the various available air intakes were affixed for each test.

The RAPIER engine was set to automatically switch between air breathing and closed cycle mode so as to eliminate human variance. MechJeb was given an orbit ascent profile of 0.1/70/1/20 - that is, set to begin it's "gravity turn" at 100 metres (ie. just above the runway), that turn being an arc flattening out with a 20% turn shape to 1 degree at 70km. This profile resulted in an ascent which looked roughly similar to the flight path I'd use when manually piloting a very light spaceplane. A circular 75km orbit was requested of the MechJeb autopilot.

MechJeb's throttle feathering anti jet flameout feature was not enabled, primarily because it interfered with the RAPIER engine's automatic mode switch. As such a human pilot and/or better tweaked MechJeb settings may well get the plane to orbit with more delta-v remaining, but with the stated methodology the results should be repeatable and comparable to eachother.

Launch procedure involved staging the RAPIER engine and immediately enabling MechJeb autopilot - from runway to orbit everything was automated.

Upon attaining a circular 75km orbit, and autopilot disengaging, the closed cycle vacuum delta-v remaining, as reported by MechJeb, was recorded.

Results

Equipped with a Circular Intake the test rig had 294m/s delta-v remaining after attaining 75km orbit.

Equipped with a Ram Air Intake the test rig had 413m/s delta-v remaining after attaining 75km orbit.

Equipped with a Shock Cone Intake the test rig had 474m/s delta-v remaining after attaining 75km orbit.

Equipped with a pair of XM-G50 radial intakes the test rig had 487m/s delta-v remaining after attaining 75km orbit.

Equipped with a pair of Structural (radial) intakes the test rig had 16m/s delta-v remaining after attaining 75km orbit.

Conclusion

The results found give a very similar order and impression as those attained in the first vertical experiment. As such I'm prepared to, tentatively, be a little more firm in repeating the same conclusions found in the first experiment.

Ie. if you're looking for an inline air intake then the Ram Air Intake and the Shock Cone Intake are likely your best choices - the difference between the two appears to give a slight advantage to the Shock Cone Intake but my hunch would be that the difference is so slight as to lie within the margin of error. The Circular Intake appears to be a slightly inferior choice, however not so much so as to dictate design - if you prefer the Circular Intake for aesthetic or other reasons you will not be hobbled by your choice.

The radial intakes seem to show the same clear results in this test as in the first - a pair of XM-G50 radial intakes appears roughly equivalent to the best you can expect of a single inline intake. The new Structural Intakes appear to be a substandard choice but again not one that would cripple a design - indeed the slim form factor of the Structural Intakes perhaps opens up other design possibilities, for example one may be able to fit more of them under a wing than the XM-G50 radial intakes.

Final thoughts

A weakness of this test design lies in the flight profile chosen: it's quite possible that different intakes will perform better or worse under different flight profiles. I reached upon the 0.1/70/1/20 flight profile after three test flights to tweak settings and purposely wished to limit my fine tuning of this profile so as to avoid inadvertently tuning the test to best suit a particular intake choice. Nevertheless the first two test flights were performed with the Shock Cone Intake and a third test flight was required with the Structural Intake to tweak the flight profile such that it could achieve a 75km orbit with any delta-v to spare. Thus it's possible I've advantaged the Shock Cone Intake and the Structural Intake by tuning the flight profile to favour these intakes.

Finally, KSP devs have noted that future updates will be steering more toward balance tweaks, and include a fresh look at the aerodynamics system in use. Thus it's more wise than ever before to treat these results with a big pinch of salt if you're running a later version of KSP than 0.25.0642.

[Streetwind]

The poor performance of the structural intake is probably related to the intake area being much smaller than all the other contenders. The ingame stats seem to suggest that intake area is a major factor for performance.

In other words, if you had enough structural intakes to offer as much intake area as the other intakes, the difference would be much smaller. Just visually comparing them suggests that XM-G50 has roughly 2.5 to 3 times the intake area.

[MiniMatt]

Yep, the advantage of the new Structural Intakes is definitely going to be that their shape will allow more of them to be placed on a craft without looking too spammy.

Two of them under each delta wing, for example:

Trying to fit four XM-G50 intakes under the wings would likely look a bit air-hoggy, but should look quite natural if using the Structural Intakes.

Still having the second test rig and flight plan on a save, I just ran the plane featured in the second experiment with four Structural Intakes rather than two - this time it recorded 384m/s delta-v remaining after reaching 75km orbit (over the original 16m/s with just two intakes)

Edited October 22, 2014 by MiniMatt

[RocketscientistV]

Your results may be unfair because you are using two of each radial intake but one of each inline intake so they would have to have results under separate categories.

[numerobis]

Two radial intakes is 0.12 area. One ram is 0.1 area. So the pair of radials picks up more air -- that's why you can fly higher.

[Wanderfound]

How do the results stack up if you go for a drag-inducing equivalence (i.e., if a structural induces 1/4 the drag of a shock cone, compare 4 to 1) or mass equivalence?

[HafCoJoe]

Intriguing... Who'da thought the radial intake is deemed 'approximately' the best! Very good data. I shall keep this in mind when building.

[Clockwork_werewolf]

I'm surprise there is much discussion on the intakes efficiency, although there is one point that I would be interest to see the results of.

Usable air is directly proportional to the intake area of an intake. Circular is 0.008, Ram is 0.01, Shock cone is 0.012, New Long radial is 0.0025 x 2 = 0.005, old radial is 0.006 x 2 = 0.012. The difference between shock cone and XM-G50 Radial Air Intake can be accounted for by the difference in mass of 0.025 (cone) to 0.01 x 2 = 0.02 (radial). This does lead to the interesting question though.

Does the maximum drag of an intake (2.0) always equal the same drag or, like other parts is it related to the mass of the part multiplied by the variable drag. A radial would have 2.0 x (0.01 x 2) if true and beat 2.0 x 0.25 for the shock cone OR are the difference in results just due to the extra mass lugged up to space?

I would also like to add that the radials look the worst so I don't use them but that's another story.

Edited October 23, 2014 by Clockwork_werewolf

[MiniMatt]

I'm surprise there is much discussion on the intakes efficiency, although there is one point that I would be interest to see the results of.

Usable air is directly proportional to the intake area of an intake. Circular is 0.008, Ram is 0.01, Shock cone is 0.012, New Long radial is 0.0025 x 2 = 0.005, old radial is 0.006 x 2 = 0.012. The difference between shock cone and XM-G50 Radial Air Intake can be accounted for but the difference in mass of 0.025 (cone) to 0.01 x 2 = 0.02 (radial). This dose lead to the interesting question though.

Does the maximum drag of an intake (2.0) always equal the same drag or like more parts is it related to the mass of the part. A radial would have 2.0 x (0.01 x 2) if true and beat 2.0 x 0.25 for the shock cone OR are the difference in results just due to the extra mass lugged up to space?

I would also like to add that the radials look the worst so I don't use them but that's another story.

Short answer: I *suspect* that the mass of all intakes, for the purposes of calculating drag only, are normalised around 0.01 when they are open.

Long answer:

Yep, the drag model of air intakes is.... weird. As a rough rule of thumb I *think* it's safe to say that the right click values of an intake shown in flight are.... squiffy. At least they're squiffy when the intakes are open (which is the state we're most concerned about). When closed, radial intakes have a drag value of 0.2, inline intakes have a drag value of 0.3. When open we're actually looking at a child transform of the game object which is calculating relative velocity and, crucially, accounting for mass differently.

Why accounting for mass differently?

Drag is a function of atmospheric density, the velocity at which you're trying to push through that atmosphere, the drag coefficient (how "slippy" your shape is), and the cross sectional area of the shape you're pushing through. Adding a deflector to the roof of a truck cab doesn't affect the cross sectional area but makes that area more slippy, improving the drag coefficient. A panel van might have a similar drag coefficient to the truck, but being smaller it is pushing a smaller cross sectional area against the atmosphere.

0.25 stock KSP doesn't calculate cross sectional area but fudges it as being proportional to mass. The mass weighted average of all drag values then form the total drag value for the craft. A craft consisting entirely of 0.2 drag parts will have an overall 0.2 drag value.

Cross sectional area is calculated as being 0.008 m2/kg

This mass proportional fudge gets wonky with air intakes. Consider the new Mk1 Fuselage Intake. 0.06 intake area, just like the XM-G50 radial. Both also have a closed drag value of 0.2. But the Fuselage Intake has 9.8 times the mass of the XM-G50 radial intake - if the overall drag factor is calculated from a mass weighted average then the Mk1 Fuselage Intake will impart massively more drag on the craft than appropriate.

Imagine a craft consisting of the Mk1 Inline Cockpit (mass 1.03, drag 0.08), a full FL-T100 fuel tank (mass 0.5625, drag 0.2), and a TurboJet Engine (mass 1.2, drag 0.2).

The craft's drag will be:

(1.03 * 0.08) + (0.5625 * 0.2) + (1.2 * 0.2)

-------------------------------------------------

1.03 + 0.5625 + 1.2

= 0.4349 / 2.7925

= 0.156

Let's now add a single, closed, XM-G50 intake (mass 0.0125, drag 0.2)

(1.03 * 0.08) + (0.5625 * 0.2) + (1.2 * 0.2) + (0.0125 * 0.2)

---------------------------------------------------------------------

1.03 + 0.5625 + 1.2 + 0.0125

= 0.4374 / 2.805

= 0.156

Let's now open that XM-G50 intake (mass 0.0125, drag 2.0)

(1.03 * 0.08) + (0.5625 * 0.2) + (1.2 * 0.2) + (0.0125 * 2.0)

---------------------------------------------------------------------

1.03 + 0.5625 + 1.2 + 0.0125

= 0.4599 / 2.805

= 0.164

If we replace that XM-G50 with a closed Mk1 Fuselage intake (mass 0.1225, drag 0.2) we get:

(1.03 * 0.08) + (0.5625 * 0.2) + (1.2 * 0.2) + (0.1225 * 0.2)

---------------------------------------------------------------------

1.03 + 0.5625 + 1.2 + 0.1225

= 0.4594 / 2.915

= 0.158

Now let's open that Fuselage Intake (mass 0.1225, drag 2.0):

(1.03 * 0.08) + (0.5625 * 0.2) + (1.2 * 0.2) + (0.1225 * 2.0)

---------------------------------------------------------------------

1.03 + 0.5625 + 1.2 + 0.1225

= 0.6799 / 2.915

= 0.233

All this is a rather long winded way of noting that our imaginary craft with a single XM-G50 intake has a drag value of 0.156/0.164 (closed/open) but with a MK1 Fuselage Intake, which offers exactly the same intake area, has a drag value of 0.158/0.233 (closed/open). The extra mass of the MK1 Fuselage Intake is already accounted for via the Newtonian principle of the force required to accelerate a given mass, but here it's also significantly affecting the drag of the craft - a double whammy.

Thus the fudge (approximating cross sectional area through mass) requires a separate fudge - normalise mass for air intakes when calculating drag (but not when calculating F=ma).

Now, I *think* that mass gets normalised around 0.01 for intakes - to refine this we'd need someone far more experienced in modding - or a diminutive goblin experienced in pot-holing prepared to delve into the perilous cramped, dank and dark depths of the KSP code.

Given it looks like the aero system is getting a fresh lick of paint over the next few updates, I suspect the system will be more elegant and predictable in the future.

Edited October 23, 2014 by MiniMatt

[MiniMatt]

Trying to devise an experiment methodology which would go some way to explore the principles raised by Wanderfound & Clockwork_Werewolf.

If we added mass handicaps to the craft design such that in all configurations the craft had the same mass what would this tell us? I'm thinking that would make equivalent the basic F=ma effort of getting a given craft to orbit, leaving us only comparing the drag induced opposing force and air intake efficiency.

Taking values from in-game screen, in order from heaviest to lightest we see:

Single Shock Cone Intake (1 * 0.027)

XM-G50 radial intake (2 * 0.0125 = 0.025)

Structural Intake (2 * 0.0105 = 0.021)

Ram Air Intake (1 * 0.011)

Circular Intake (1 * 0.011)

The difference between the lightest and heaviest configuration is 0.016 - 16kg. Handily, the Mk1 Inline Cockpit used in the test rig includes 30kg of useless (to us) monopropellant which can be tweaked in 3kg increments.

Thus we could:

Remove all monoprop from the Shock Cone Intake configuration, as the un-handicapped baseline.

Include 3kg of monoprop in the XM-G50 radial intake configuration, making it 1kg heavier than the baseline.

Include 6kg of monoprop in the Structural Intake configuration, making it equal mass to the baseline.

Include 15kg of monoprop in the Ram Air Intake configuration, making it 1kg lighter than the baseline.

Include 15kg of monoprop in the Circular Intake configuration, making it 1kg lighter than the baseline.

Would this tell us anything new or would it merely refine and remove error from the original experiment?

Second question - which mass values should we be using - values in the part.cfg files are different to that shown in game:

Single Shock Cone Intake (1 * 0.025)

XM-G50 radial intake (2 * 0.01 = 0.02)

Structural Intake (2 * 0.008 = 0.016)

Ram Air Intake (1 * 0.01)

Circular Intake (1 * 0.01)

In order to handicap these values such that all configurations mass as much as the heaviest configuration we'd have to:

Remove all monoprop from the Shock Cone Intake configuration, as the un-handicapped baseline.

Include 6kg of monoprop in the XM-G50 radial intake configuration, making it 1kg heavier than the baseline.

Include 9kg of monoprop in the Structural Intake configuration, making it equal mass to the baseline.

Include 15kg of monoprop in the Ram Air Intake configuration, making it equal mass to the baseline.

Include 15kg of monoprop in the Circular Intake configuration, making it equal mass to the baseline.

I'm tempted to think the in game screens are including a mass of intake air resource which within game calculations would be considered massless and as such the part.cfg values are possibly a better choice.

[MiniMatt]

Ok, went ahead and re-run the second (spaceplane) experiment with a mass handicap in effect. In the original experiment all configurations came equipped with 30kg of unused monoprop.

In this re-run the overall mass of the craft in each configuration was equalised as close as possible, by removing all monoprop from the heaviest configuration and adding sufficient monoprop to lighter configurations to make their mass up to equal that of the heaviest, baseline, configuration.

It is important to note that these results do not help us answer the question "Which intake should I use?" - in fact the original experiment is far better equipped to do that than this re-run; in real play your designs are already mass optimised, you're not carrying additional handicap mass and you don't get to handicap other competing designs - you pick your intakes and you live with, and consider, their mass.

What the results of this handicap run perhaps do is allow us to better investigate how intakes work.

Methodology

Taking mass values as listed within part.cfg files rather than in-game screens we see the mass of the various intakes is:

Single Shock Cone Intake (1 * 0.025)

XM-G50 radial intake (2 * 0.01 = 0.02)

Structural Intake (2 * 0.008 = 0.016)

Ram Air Intake (1 * 0.01)

Circular Intake (1 * 0.01)

As such, to equalise mass across all configurations the following was performed:

Remove all monoprop from the Shock Cone Intake configuration, as the un-handicapped baseline.

Include 6kg of monoprop in the XM-G50 radial intake configuration, making it 1kg heavier than the baseline.

Include 9kg of monoprop in the Structural Intake configuration, making it equal mass to the baseline.

Include 15kg of monoprop in the Ram Air Intake configuration, making it equal mass to the baseline.

Include 15kg of monoprop in the Circular Intake configuration, making it equal mass to the baseline.

It is assumed, thanks to the somewhat fruity way drag coefficient is calculated in stock 0.25 KSP, this handicap mass has an individual drag value of 0.08 by virtue of it residing within the Mk1 Inline Cockpit.

All further methodology exactly as described in the original experiment.

Results

Equipped with 15kg monoprop handicap & a Circular Intake the test rig had 305m/s delta-v remaining after attaining 75km orbit.

Equipped with 15kg monoprop handicap & a Ram Air Intake the test rig had 423m/s delta-v remaining after attaining 75km orbit.

Equipped with 0kg monoprop handicap & a Shock Cone Intake the test rig had 494m/s delta-v remaining after attaining 75km orbit.

Equipped with 6kg monoprop handicap & a pair of XM-G50 radial intakes the test rig had 502m/s delta-v remaining after attaining 75km orbit.

Equipped with 9kg monoprop handicap & a pair of Structural (radial) intakes the test rig had 29m/s delta-v remaining after attaining 75km orbit.

Conclusions

As highlighted above, these results don't help us answer the question "Which intake should I use?", and are indeed less helpful in answering that question than the original run.

They perhaps show us something about how intakes work. Of most interest is perhaps the results for the Shock Cone Intake and the pair of XM-G50 radial intakes. Both configurations result in 0.012m² of intake area. With the overall mass of craft within 1kg of identical the two configurations get to orbit with just an 8m/s variance in delta-v remaining.

Proper statistical analysis is really required but I think it hints toward the mass of an intake being relevant only in regard to the force being required to accelerate to orbital velocities and it's effect is minimised on the slightly fruity drag coefficient calculations.

[Wooks]

What about the Mk1 Fuselage - Intake, the Engine Nacelle and the Radial Engine Body? Any data on those against air intakes and between themselves?

[davidpsummers]

I'm not seeing the Shock Cone Intake and the Structural Radial Inakes in my parts list? They are stock?

[MiniMatt]

I'm not seeing the Shock Cone Intake and the Structural Radial Inakes in my parts list? They are stock?

They are, as of version 0.25. Their official wiki pages are here & here.

Are you running an earlier version? Or are you perhaps looking at a Career Mode save which has yet to unlock these parts?

What about the Mk1 Fuselage - Intake, the Engine Nacelle and the Radial Engine Body? Any data on those against air intakes and between themselves?

Hadn't tested those ones - truth be told I forgot about them initially. They perhaps complicate the issue due to their mass being significantly greater than the regular intakes, plus two of them having the added functionality of carrying fuel.

From the stats I would presume them to perform similar to a single XM-G50 intake in terms of the air they can provide, but their extra mass proving a significant hindrance such that they'd likely perform very poorly on these tests.

[davidpsummers]

They are, as of version 0.25. Their official wiki pages are here & here.

Are you running an earlier version? Or are you perhaps looking at a Career Mode save which has yet to unlock these parts?

I'm running 0.25. (BTW, when you upgrade to the next version, can you just replace the program file? At the moment I make a new KSP folder and then transfer all the saves, craft, and mod over.)

I am running a Career Mode (now a "Science" mode) where I have unlocked all the tech levels. I was pretty sure I had looked for these in my Tech Tree, in case it came in unlocked, but I didn't see it.

Your link (looking at the shock cone) says it is in hypersonic flight. But I don't see it listed (locked or unlocked). Neither do I see the structural intake under high altitude flight.

[zekes]

why are those radial intakes so effective? It's strange...

[Wanderfound]

why are those radial intakes so effective? It's strange...

There is a complicating factor. As I understood it, different types of intake are supposed to have different efficiencies at various speeds and altitudes. It's possible that a flatter, faster ascent may return different results.

[numerobis]

why are those radial intakes so effective? It's strange...

At some point (0.23 maybe?) the radials' mass got reduced to be the same as the circular and ram intake. So now they have 60% the airflow:mass ratio of the ram, rather than the 6% they had in olden days.

Edit: oh right, they also had 0.04 area, as opposed to 0.06 now. And they have slightly reduced drag now. So back in the day the ram air intake was 25x better, whereas now it's merely 50% better.

Edited October 30, 2014 by numerobis

[numerobis]

I've updated the intakes for 0.25 in the KSP-scripts

The underlying code doesn't appear to have changed since 0.18 when Amram snuck off with the code snippet.

[m1k3ol]

I've updated the intakes for 0.25 in the KSP-scripts

The underlying code doesn't appear to have changed since 0.18 when Amram snuck off with the code snippet.

I guess this is the most important part?

# Updated for 0.24
# name, mass, area, options
circularIntake = intake("Circular Intake", 0.01, 0.008, drag=0.3)
mk1Intake = intake("Mk1 Fuselage - Intake", 0.12, 0.006, capacity=1, intakeSpeed=12)
nacelleIntake = intake("Engine Nacelle", 0.15, 0.005)
radialBodyIntake = intake("Radial Engine Body", 0.15, 0.005)
radialIntake = intake("XM-G50 Radial Air Intake", 0.01, 0.006, capacity=1)
ramAirIntake = intake("Ram Air Intake", 0.01, 0.01, drag=0.3)
shockConeIntake = intake("Shock Cone Intake", 0.025, 0.012, capacity=0.8, intakeSpeed=12, drag=0.3)
structuralIntake = intake("Structural Intake", 0.008, 0.0025, capacity=1)

[numerobis]

I guess this is the most important part?

The jets module has the intakes, yes. The snippet you copied has an obvious typo (0.2*5*) that I just fixed, along with pushing some other stuff I had on my laptop.