[1.3.1] Ferram Aerospace Research: v0.15.9.1 "Liepmann" 4/2/18

[Camacha]

It looks like 1.0 is delivering exactly what I was hoping for. Instead of limiting what certain mods can do, it seems they can do more than ever. Ferram, I cannot thank you enough for all your hard work, this looks absolutely amazing.

[NoPersona]

Aerodynamic updates have always happened without staging events. Deploying landing gear and dynamically sweeping wings have always worked in oldFAR, I don't know where you got the idea that they didn't.

The voxel resolution is relative to the total vehicle size. As a vehicle gets larger, the smaller parts of it really don't need to be resolved as finely, so it works out to approximately a constant memory load per vehicle. Also, I don't need to store the entire voxel for the entire time the vessel exists; I can convert it into an aero forces and orientations format that is much simpler to store and calculate from.

It uses newtonian impact theory for hypersonic drag, an implementation of slender body theory for transonic drag, and an empirical function for subsonic. That's for 0 AoA; for off-0 AoA forces I use a combination of potential flow lift (also slender body approximation) and a viscous crossflow term. It's much faster than a vortex lattice method and handles viscous effects, and in the situations where the model isn't exactly right, it tends to overpredict the forces and generally sets the requirement that people follow the assumptions of the model if they don't want it to explode in flight.

I must have remembered something silly along the way, thanks for this explanation and setting me straight.

Keep up the awesome work with this mod.

[Boamere]

Oh man, Oh man, I'm so glad you are continuing support for this, I have to say that the new aerodynamics of stock are more fun than they were before but... I have to have FAR for it to be great. Thanks for the best mod in KSP!

[Nothalogh]

Most importantly, I can implement a long-awaited feature: area-ruling.

Oh god, YES

[plausse]

Im late on the train. But this is as good as it gets. Like that unpasteurised goat cheese I had an hour ago, but better.

[Akira_R]

Will this new version solve the drag and flipping out problem some mod capsules have when they include a decoupler shroud on the bottom of the pod itself?

[DaMichel]

The voxel resolution is relative to the total vehicle size. As a vehicle gets larger, the smaller parts of it really don't need to be resolved as finely, so it works out to approximately a constant memory load per vehicle. Also, I don't need to store the entire voxel for the entire time the vessel exists; I can convert it into an aero forces and orientations format that is much simpler to store and calculate from.

And this works?! Cannot wait to break try it!

It uses newtonian impact theory for hypersonic drag, an implementation of slender body theory for transonic drag, and an empirical function for subsonic. That's for 0 AoA; for off-0 AoA forces I use a combination of potential flow lift (also slender body approximation) and a viscous crossflow term. It's much faster than a vortex lattice method and handles viscous effects, and in the situations where the model isn't exactly right, it tends to overpredict the forces and generally sets the requirement that people follow the assumptions of the model if they don't want it to explode in flight.

Hm. "Potential" rings a bell. And you can do all that without solving say a Poisson equation for the potential. That is pretty slick.

You know i just asked about the vortex lattice method because i got interested in RC aircrafts and came across programs which use it to predict aero stuff. http://tornado.redhammer.se/ http://www.flz-vortex.de/flz_vortex.html

[Instresu]

I will say that nuFAR is awesome and should be out in a day or two. Currently I'm waiting on a stripped-down parachute module from stupid_chris (who was gracious enough to offer one) to replace stock behavior. For obvious reasons, I can't release a FAR version that results in pods impacting the ground at 180 m/s. Also, I'm too sleepy to make a good judgment call about its release state.

For those wondering exactly what nuFAR / voxelFAR / vesselCenteredFAR does compared to previous versions, it basically does away with the part-centered operation of the previous FAR versions (the same behavior that powers NEAR and newStock) in favor of a vessel-centered approach. Instead of the part-centered complications of determining the interactions between parts (which are never done exactly right) it'll create a voxel model of the vehicle to work with, like these guys:

http://i.imgur.com/th6Jd6z.pnghttp://i.imgur.com/fkxx9ON.png

From this, I can get a lot of data about how the vessel will behave in flight. Most importantly, I can implement a long-awaited feature: area-ruling. Basically, it's a requirement for transonic and supersonic flight that states that for minimum drag, a vehicle's cross-section must vary smoothly. And very small changes in cross-section can have a surprisingly large effect on transonic and supersonic drag. So for ease of use, you'll get a nice graph of cross-sectional area over the vehicle so you can see where to mess with things:

http://i.imgur.com/L3ZcDkV.png

The yellow line is the 2nd derivative of cross-sectional area, which actually goes into the math behind drag. Green is obviously cross-sectional area. Transonic area ruling consists of fiddling with the cross-section curves until you smooth out the curve to get minimum drag. For obvious reasons, some people will find this unfun given the limited design choices we have for shaping things, so there are difficulty options that include various multipliers to the total drag and smoothing functions for the area to deal with the noisiness of the voxel model itself. But to give you an idea of how this makes things go, getting supersonic is an exercise in flying through a brick wall that may tear you to shreds. I've actually had to throttle down near Max Q to prevent a very nasty, very large rocket from coming apart and destroying itself.

Besides that fun, the voxel approach also allows better resolution of body lift, which can give bodies their proper forces, rather than a simpler approximation from parts. Another benefit of the voxel approach is that cargo bays and payload fairings are emergent from the system. Rather than being defined top-down like in FAR, NEAR, or newStock, any collection of parts in a hollow arrangement will be tried as shielded from the airflow. Although the current wing model is the legacy oldFAR wing model, I also expect that the voxel approach can be used to calculate wing shapes and provide a much more accurate model of the wings as well as the fuselage.

So, yeah. Enjoy newStock for now or stay on 0.90 and keep oldFAR. nuFAR will be out soon, when I will battle the demons of what version to give it.

Thanks, you are awesome. I seriously love you.

[futrtrubl]

Now we need a fuselage part mode that allows us to create custom, smooth shapes. Perhaps by setting cross-sections at certain points and interpolating between those. I keep considering what some actual smooth-shaping tools might work like, because even if you go back to WWII, the fuselages weren't straight cylinders.

Makes me wish I could find enough time to mod KSP. I consistently can't even find time to develop the games I want to make on my own.

Procedural Parts allows you to create tanks and fuselages with changing cross-section.

[StevenRS11]

I get the feeling that ferram is a little excited about NuFAR, heh. How long has this been in the works?

Personally, I can't wait. Parts won't need special far configs, and I never have to worry if my clever arrangment of parts is actually behaving the way I think it will. SO EXCITED AAAA!!!!

Ok calm now.

[tater]

It looks like 1.0 is delivering exactly what I was hoping for. Instead of limiting what certain mods can do, it seems they can do more than ever. Ferram, I cannot thank you enough for all your hard work, this looks absolutely amazing.

Was going to type something similar, but this post makes it redundant.

Thanks, ferram!

[WoollyMittens]

Looks like Ferram is mathsing his arse off. Good on him. I hope he'll save us from the sideways re-entry problem with Squad's heat shields.

[ss8913]

So now I'm kind of stuck; I want to play 1.0 but I know about nuFAR now. I need my proper aero and I don't think I'll be able to play 1.0 until nuFAR releases FAR was always my "most important" mod on any version of KSP that I played, and it looks like that isn't going to change.

Is DRE still needed too or is stock heating realistic?

[Drew Kerman]

Is DRE still needed too or is stock heating realistic?

Plenty of relevant discussion on this already in the DRE thread

[ZalgoWaits]

Just want to say thanks for continuing with this even with the new aero model in place. Stock is way improved over where it was, but I miss skip reentries in particular that I just can't seem to be able to do even with the new model.

[Autochton]

Procedural Parts allows you to create tanks and fuselages with changing cross-section.

Yes - but changing how? Can I squeeze it into a rounded-corner rectangle? A triangular shape like the one on the SR-71? Can I do this? Last I looked I only had circular cross-sections available.

[futrtrubl]

Yes - but changing how? Can I squeeze it into a rounded-corner rectangle? A triangular shape like the one on the SR-71? Can I do this? Last I looked I only had circular cross-sections available.

Oh yes, only circular cross sections. Changing in that cross sectional area changes along the part, this allows you to try match the trans-sonic area rule better at least. I just put in a pull request for waist type shape for exactly that purpose.

Edited April 28, 2015 by futrtrubl

[Temeter]

Always a step ahead, huh? That is some impressive stuff.

[BudgetHedgehog]

What the effect on framerate with nuFAR as compared to oldFAR? After months of <30FPS due to the amount of mods I had, a stock install plays at a nice 60 FPS - I'm assuming you're turning the stock aero off and then replacing it, does it result in less of a framerate impact?

Also, nuFAR looks awesome - can't wait!

[Wanderfound]

Adding to the choir: I just had a play with the new stock aero, and while it is a large improvement for stock, FAR is in no danger of becoming obsolete. I'll probably be leaving KSP largely unplayed until FAR updates.

[Clythoris]

I will say that nuFAR is awesome and should be out in a day or two. Currently I'm waiting on a stripped-down parachute module from stupid_chris (who was gracious enough to offer one) to replace stock behavior. For obvious reasons, I can't release a FAR version that results in pods impacting the ground at 180 m/s. Also, I'm too sleepy to make a good judgment call about its release state.

Um, does this mean realchutes will become obsolete if I have nuFAR installed?

[yafeshan]

The new stock model is good but not adequate. It is still far too easy to reach space than orbit. I played very rarely without FAR and I do not want to start game without it again. I am reading very good stuff about this new nuFAR. I have read Ferrams explanation on new features and my mind blown out

[CaptainBlue]

I had a taste of voxelFAR in 0.90. Now I have ksp 1.0 just sitting here, waiting to be launch. I can't do that without nuFAR.

[EMKAY]

I'm not very familiar with aerothermodynamics, nevertheless my idea is: voxel based shapes to calculate reentry heat-flow (and connect it somehow to DR or to the new stock system). Maybe even shock-shock interactions can be considered that way (antennae or even engine nacelles ripped off during reentry!!).

[AndreyATGB]

Um, does this mean realchutes will become obsolete if I have nuFAR installed?

No I think he's only going to use the physics aspects of real chute (the way chutes open gradually). You'll still need RC if you want to make custom chutes.