Thoughts on the soup-o-sphere

[SpaceToad]

So I was curious what everyone thought about the atmosphere in the 1.X versions of the game. B9 dropped support for the stock drag model and left for FAR. That really sucks, and says something about the atmosphere for sure. Why is it so soupy? Why is FAR better?

[Alshain]

I think you have your terminology and timeline confused. The soup-o-sphere was in 0.90 and earlier, 1.x got rid of the soup-o-sphere. B9 dropped support for stock aero prior to the new aerodynamics in 1.x and haven't updated since it's release.

The new aerodynamics are great, they aren't uber-realistic but they aren't the soup-o-sphere by any means. The old FAR was closer to what stock is now, which is why it was better back then.

Edited September 26, 2015 by Alshain

[Red Iron Crown]

Yup, souposphere hasn't been a thing since 1.0. The stock model has been greatly improved so craft shape and streamlining matter, as well as craft orientation. The days of slamming a rocket over to a 45 degree angle at 10km are over.

FAR, meanwhile, has gone to an even more sophisticated model that makes a voxelized model of the craft and does its aero calculations on the resultant shape. This takes things like clipped parts and enclosures made from multiple parts into account.

My advice would be to try the new stock model and see how you like it, if you find it inadequate you can get greater fidelity through FAR.

[Kobymaru]

I really like the new atmosphere compared to the previous one. It feels much more realistic, and now I don't feel that I'm cheating if I don't have FAR installed. A few things still irk me:

- Occlusion only works for parts in a stack. Even if one part is clearly "behind" another part, it still creates a lot of drag. This is especially nasty for struts. If you want to strut your noodle-plane, no matter where, even "inline", they still create drag and heat up on reentry.

- Attaching air intakes to the *back* of engines, rotating them into the airstream and offsetting them away from the exhaust stream actually *reduces* drag. No, really.

- I still haven't gotten a single plane to stall. Flatspins - yes. But no stalling. Is that even a thing in Stock aero?

- Eve and Jool atmospheres are *super* mean. If you just so much as scratch the atmosphere (like 5km below the cutoff), you burn off all of your surface attached assets. Sad kerbal.

I kinda wish they would hire Ferram. That voxelizing thing sounds amazing.

Edited September 26, 2015 by Kobymaru

[CaptainTurbomuffin]

they would hire Ferram.

They should. They should also hire some graphics for a graphical overhaul.

[p1t1o]

- Attaching air intakes to the *back* of engines, rotating them into the airstream and offsetting them away from the exhaust stream actually *reduces* drag. No, really.

Attaching anything to the rearmost node of an engine (if it has one) reduces drag, apparently that empty node is draggy. I think its a known bug. There was a thread about it somewhere, the best thing they said to do was attach a small nosecone and use the offset tool to clip it back into the engine so the exhaust was free.

- I still haven't gotten a single plane to stall. Flatspins - yes. But no stalling. Is that even a thing in Stock aero?

A spin is just one wing stalling before the other, any roll deviation from horizontal will create asymmetric conditions on the wings so if you are going to stall, one is always more likely to stall before the other. Its probably a mark of an increase in quality of the aero model ("probably").

[Guest]

Attaching anything to the rearmost node of an engine (if it has one) reduces drag, apparently that empty node is draggy. I think its a known bug.

No, it's how FAR did things back in the day as well. One of the great oldFAR tricks was adding antennas to open nodes in order to reduce drag.

People like to think that aerodynamic calculations are trivial but the complication is in KSP itself and how you can create pretty much any shape, and how that shape can change rapidly at any moment. KSP's current simulation is pretty good and feels, and acts, a lot like pre-1.0 FAR, or NEAR maybe (never played it). It's also pretty fast.

ferram4's current system of voxelizing the vessel is a great solution to the problem but it was by no means trivial; it took him a good deal of work and pushing builds off on people like me, who just happened to be nearby, for minor testing and performance reviews to get it right. KSP could definitely go that route but Squad's dev cycle wouldn't make it easy and, besides, they've pretty much got oldFAR already.