PDA

View Full Version : Rocket Launch Smoke



Dhovakin
10-17-2013, 03:11 AM
Hi,

since weeks now - i try to make a realistic looking rocket launch - but i donīt get it... iīve read different tutorials - about faking volumetric clouds with 2 or more layers who interact with a smoothstep node - iīve seen a really cool tutorial which use a baked uv map - and iīve tried different times to use this stuff but i think i still miss some steps - so iīve got never a good result

1. The 2+ Layer Faking Clouds look quiet good - but i donīt know how to use this for hypervoxels
http://www.scifi-meshes.com/forums/content.php?154-3D-Volumetric-Planetart-Clouds

2. This http://forums.newtek.com/showthread.php?124026-Hyper-Voxels-2-1!&p=1203492&viewfull=1#post1203492
looks fantastic... - but what the hell i have to do for this ???

Here is my poor result (using Sprite Mode)
117685

List of Problems:
- Smoke is not one single Volume (is it possible to setup them that they react like metaballs ?)
- Emitters donīt react like in the Nasa backgroundpicture (Particleweight, Gravity, Bounce - something is wrong...)
- Smoke miss a real or faked Shadow - looks to Flat - maybe itīs impossible in Spritemode
- Using a Node Setup - donīt allow me to use Particle Age as Parameter
- donīt know how to set up the emitters - that they push up the nullobject from the ground (like real engines)

117686
It would be very nice if someone can explain me how to do it better... - i have less experience with Hypervoxels, UV Maps, Baking and all that stuff - i know know nothing about light base objects or vertex colormaps and all that stuff...

3DGFXStudios
10-17-2013, 03:38 AM
With hypervoxels you'll never get a realistic result and/or movement. Try using the turbulence plugin. www.Jawset.com

stib
10-17-2013, 05:36 AM
Or Blender. www.blender.org

starbase1
10-17-2013, 05:57 AM
I managed something at the start of the year - if I don't get back in a day or so, give me a nudge! I can share some presets.

If you want to search for the images, I think I did post them here for the "earl" rocket project, a search on EARL should find them.

starbase1
10-17-2013, 06:01 AM
I just checked, you can see my results here:

http://forums.newtek.com/showthread.php?134883-My-first-proper-space-project-commision&highlight=earl

Nick

vector
10-17-2013, 01:17 PM
You can try adding a turbulence or even a bitmap gradient to the shading to getting a kind of volume using sprites.

I attach a simple scene just as example of what I say, no properly animated for what you need, only shading.

Er, I own the 9.6 version, 11 at work but now I'm at home

prometheus
10-17-2013, 05:37 PM
as mentioned by vector, try add a gradient or procedural in some channels, you can try on luminosity or color ..if you want more depth to the sprites, also ..you can raise slices to get more volume, since 1 slice is one slice of the volume, and 2 slices is 2 slices etc..so you get more volume look to it that way.

otherwise..I would use volume mode for more realism and light and thickness options, the tricky part there will be to avoid blob look, itīs a common know issues with hypervoxels for ages, that the volume mode doesnīt blend quite well..
They just recently added a blend mode for the volume mode, but I think it fails big time for that...it should have the same tension blend as the old dynamite plugin had, or the hv doubler.
missile or plane smoke or rocket smoke from a flying vehicle would be easier to do with hypervoxels, but the dynamic interaction with a ground obstacle when the smoke hits it is way harder, for that a fluid engine would be best to work with.

For the most realistic stuff, you should go and get turbulenceFD fluids..www.jawset.com

Michael

vector
10-17-2013, 10:06 PM
As stib said, Blender is fast and easy to get some results. I use it sometimes because I have no budget for Turbulence although it can become a problem when the camera and/or object moves because its recent fbx import plugin doesn't work correctly at this time

Afalk
10-19-2013, 01:54 PM
Another resource is this thread from Foundation, there are some nice looking setups in there, and some great info as well.

http://www.foundation3d.com/forums/showthread.php?t=14599&highlight=rocket+exhaust

prometheus
10-20-2013, 01:44 PM
I would think of treating the dynamics with two or more different emitters, and not trying to let the main rocket smoke hit the launch pad and have it turn and twirl like smoke, better of to simply
start another emitter that is located on the ground and push that outwards as the smoke that flows out from the launch pad, or set up several such emitters, or even use point clusters and morph it.
Using geometry that is expanded and slightly deformed could work as well too.

Michael

vector
10-20-2013, 01:56 PM
It's a pity that when using two or more emitters together in a scene the render times increase exponentially. I get nice results but sometimes (or the most of the times) I can't afford those render times.

I never had used the "times" word so many times :) :) :)

starbase1
10-21-2013, 09:08 AM
It's a pity that when using two or more emitters together in a scene the render times increase exponentially. I get nice results but sometimes (or the most of the times) I can't afford those render times.


Yes, when I finally go good results for the EARL project, I really had to avoid looking along the exhaust trail or print quality renders would take literally days, (on a powerful PC).

prometheus
10-21-2013, 09:15 AM
It's a pity that when using two or more emitters together in a scene the render times increase exponentially. I get nice results but sometimes (or the most of the times) I can't afford those render times.

I never had used the "times" word so many times :) :) :)

That is probably the fact of hypervoxels treating the hv as different groups and that yields additional calculation time.
Itīs the same with using two simple nulls with large hypervoxels for clouds for example, that will become extremly slow to render, if you were to use the same hypervoxel size and settings
.but on two points or particles..it wouldnīt be the
same..rather much faster, so a method of treating several nulls as a group using the same hv setting would be nice.

Thereīs a workflow issue too if you would want to work with clouds and hypervoxels and also avoid nulls due to the render time issues those impose when using different Hv goups, the issue is that you canīt modify and direct particles and points directly in layout since we donīt have vertex manipulation(classic split application/non unified issue)
If we could manipulate points directly in layout in a decent manner, setting up clouds as you want in scene context would be improved vastly, that is a different topic though.
If it could treat different nulls or emitters by making them a choosable group for hypervoxels, it would Not impose a problem having two or more emitters...other than what the amount of particles and sizes imposes.

starbase1
10-22-2013, 06:16 AM
In my experience it's overlapping HV's that drive the render times sky high. And if you combine radiosity with it, you are probably going to think your PC has stopped dead unless you do the HV's in a separate pass.

Afalk
10-22-2013, 07:08 AM
Couldn't we comp individual HV runs atop one another? Would that net any savings in a situation like this?

prometheus
10-22-2013, 07:54 AM
In my experience it's overlapping HV's that drive the render times sky high. And if you combine radiosity with it, you are probably going to think your PC has stopped dead unless you do the HV's in a separate pass.

No I donīt think so, not exactly not the biggest issue, you can have a lot of particles overlapping as long as they are within the same hv group, as I mentioned in my explenation of two nulls with two different voxels VS two points using the same hv group. It is the issue of adding a secondary hv group to the scene. but using two nulls with two different hv settings and having them not intersecting/overlapping will not impose any slower render time that much
but when they are overlapping ..they will, so youre half part right there, but if we were to use two points with the same hv group(which we canīt do with nulls) and direct them the same way, the rendertime wouldnīt be as slow.even if they overlap.

Another thing that might help if you plan on post smoothing the voxels later, turn of volumetric AA, that is a real slowdow, but if you have volumetric lights in there then it might not work so nicely since you need the volumetric AA to get smooth
volumetric quality.

Note..with hv group, Iīm not talking about blending group, but the main hv appliance.
Michael

starbase1
10-22-2013, 10:33 AM
Couldn't we comp individual HV runs atop one another? Would that net any savings in a situation like this?

Yes, but it looks odd if they overlap - an example from the scene I linked to earlier had 5 exhausts side by side, and they start to mix a bit back from the rocket, as well as partly obscuring each other.

But in my comment I meant where there's lots of overlap within the same HV point set, so shrinking the HV radius where possible makes a HUGE difference in render times in my experience...

prometheus
10-22-2013, 11:29 AM
One thing is to keep in mind what causes Huge rendertime, and a lot of it has to do with light, and how much it needs to be calculated in the volume, a very low opacity means the light travels through the volume a lot and needs more calculation, where a high opacity setting blocks the light and thus not the need to calculate that, same goes for thickness, the lower it is, it generally renders slower too, then you got the procedural textures themself, some of them renders much slower.
Ivé heard from Kelly Myers saying that fiddling with the ray cutoff or the other settings there should be able to speed up things, but I havenīt noticed that in my tests, depends on how it is set up I guess.

And as always, you could turn of texture shadows, and even volumetric shadows for faster voxels and use gradients on different channels, but I donīt like the look of voxels that way compared to using full volume and textured shadows, depends
if you in the end would want to smooth the hvīs out.
Distance to camera is also a render clogger.