Page 2 of 3

Re: Ifu - cloud simulator and renderer

Posted: Mon Oct 20, 2014 9:09 pm
by citadel
If it's multiple scattering dominant, the phase function should not matter that much,
and would allow using reduced parameters, according to Jensen'02, the fast BSSRDF paper.
I think the key is in its genuinely high albedo, which is particularly difficult for path
tracing to fully capture.

Re: Ifu - cloud simulator and renderer

Posted: Tue Oct 21, 2014 1:27 am
by snwy_
If real clouds scatter mostly forward but renderings of this don't come out right, it may not be the real-world physics that need redefining, for sure. But on the other hand, whatever makes the rendering look right.

With clouds, though, a good phase function won't help if the droplets aren't arranged in a realistic way, and this is where you'll really step into a pile.

Re: Ifu - cloud simulator and renderer

Posted: Tue Oct 21, 2014 6:57 am
by citadel
I agree not to redefine the physics, but using reduced parameters is more of a mathematical treatment. After all, we are just doing integration of the light transport equation.

The albedo issue is quite orthogonal to that. A close to 1 albedo either introduces serious bias for a finite-depth tracing method, or requires serious effort to address properly. I recall you using a scattering probability for the photons, which I believe is the albedo. For the correct physics, your photons should never stop scattering until they exit the medium.

Re: Ifu - cloud simulator and renderer

Posted: Tue Oct 21, 2014 12:24 pm
by snwy_
Clouds aren't much of a medium, though - quite possible for a ray to pass 'through' without hitting anything.

Rays in Ifu have a maximum depth of about 10000, but while inside a cloud mesh, they'll keep going until they find a way out.

Re: Ifu - cloud simulator and renderer

Posted: Wed Oct 22, 2014 9:14 am
by citadel
Thanks to your cloud simulator, I now have some decent renderings. I use g=0.7 HG phase function. But I made some modifications to the data. I binarized the data, so that it appears to be large scale. The key is probably that large scale cumulus clouds have surface transitional regions much thinner than the main body.

_additional_l.png
Binarized
_additional_l.png (112.6 KiB) Viewed 9213 times
_additional_s.png
Ifu output
_additional_s.png (115.29 KiB) Viewed 9213 times

Re: Ifu - cloud simulator and renderer

Posted: Wed Oct 22, 2014 12:19 pm
by snwy_
Looks more like clouds now, though I'd still say you're being shy with the forward scattering. What I've tended to notice looking at photos of cumulus clouds is that even though they look solid when far away, they get misty when you move closer.

I've two observations of the data from Ifu. First, since the clouds have no flat bottoms, they're already dissipating - I'm guessing most people associate cumulus clouds with flat bases, i.e., at the stage where they're still growing. Second, the output looks like it's from the mixing phase of the simulation, and if so, I'd recommend to either wait until the normal phase or to customize the heatmaps. And look for the beta 2 version of Ifu, should be out next weekend or so, for a bunch of improvements to the simulator.

Re: Ifu - cloud simulator and renderer

Posted: Wed Oct 22, 2014 2:12 pm
by citadel
What's the exact setting for your moisture/dust scattering type? Is that related to the sigmas?

My further tests show that the simulation never converges to beautiful results, and tends to dissipate completely. Early stage seems more like it.

Re: Ifu - cloud simulator and renderer

Posted: Wed Oct 22, 2014 5:59 pm
by snwy_
I'd guess the reason the mixing phase makes more satisfying clouds is that the heat multiplier is larger for that phase. See Settings -> Simulator and the last two options there. You can increase the multiplier for stronger cloud formation, etc.

Changes to the mixing process were made for beta 2, and that version disables mixing by default, since it still needs more work. The problem with making clouds without first mixing up the atmosphere somehow is a potential loss in realism (eg, Entrainment and Detrainment in Numerically Simulated Cumulus Congestus Clouds. Part I: General Results by Carpenter & Droegemeier).

The moisture and dust scattering is for the atmosphere only; rays will scatter every now and then on their own. Moisture uses the same phase function as the clouds, and for dust I think it's just more or less uniform in all directions.

Re: Ifu - cloud simulator and renderer

Posted: Mon Oct 27, 2014 1:10 pm
by snwy_
The beta 2 version of Ifu is out. Comes with a bunch of changes, improvements, and new features (support for wind, variable terrain, HDR environment maps...). Work in progress, etc.

Get it here: http://personal.inet.fi/muoti/eimuoti/ifu/ (Linux/Windows 64-bit)
See the changelog and release notes: http://personal.inet.fi/muoti/eimuoti/i ... ngelog.txt

For best results, I'd recommend exploring the various options in the program, especially the atmospheric ones.

I've updated the first post accordingly.

Re: Ifu - cloud simulator and renderer

Posted: Fri Oct 31, 2014 5:05 pm
by snwy_
Couple short video tutorials up:
Ifu Silent Basics - Render Setup - YouTube
Ifu Silent Basics - Atmospheric Settings - YouTube

Bring your own soundtrack, etc.

More tutorials may or may not go up on the playlist later.