Borealis - North Light

Started by choronr, March 03, 2011, 09:25:01 PM

Previous topic - Next topic

Oshyan

Strange, those all produce similar (incorrect) results. It appears due to one of the 2 suns. I will try to test on the current release version tomorrow and see if it's an alpha issue. If so it may be a clue to a bug in recent changes, which would be good to catch!

- Oshyan

choronr

I hope you find it Oshyan; what a relief that would be.

Oshyan

I tweaked Aurora 3 a bit to get similar results and re-rendered with further adjusted render settings. Render time now down to 7 hours and 25 minutes, there's a bit more noise but not bad (less than your original version I'd say). Image is attached.

Here are the basic settings:

Atmosphere samples: 12
Cloud samples: 52
Raytraced Atmosphere: On
Antialiasing: 4
Sampling: Customized, 1/16 first samples, 0.01 Pixel noise threshold

That should get you similarly lowered render times, depending on your hardware of course. I didn't test a full render with your original settings (especially since your TGD looked very different on my system anyway), but I would guess this cut render time down by a factor of 4 or more. Note that the cloud samples and especially atmosphere are way lower than what you had. The raytraced atmosphere is doing all the work.

I'm still going to test on the public release later today for comparison. I'll report my findings.

- Oshyan

choronr

A big thanks Oshyan, as soon as my render is finished, I will begin to make the adjustments you suggested - really looking forward to this. In the image you posted here, I noted that the aurora section on the lower left appears down into the terrain - like a rainbow. Odd; but, maybe other cloud seeds will change that.

The render I'm working on has also been a very long one. It may not be finished until late Saturday.

Bob

Oshyan

Following up with testing on the Aurora 3 file under the public (free) version: it appears to give the same/similar results to what I was getting on the alpha version, so I can only conclude that the files you've been posting are *not* the same configuration/file that was used to produce the images in the beginning of this thread. Certainly the configuration is similar, but Markal must have made other changes or something, because there are some significant differences in the rendered result. It's fine though, the render time savings should hold true regardless.

I forgot to mention I also reduced main detail to 0.5 because I felt 0.8 was overkill for this scene. In particular when you're raytracing the atmosphere detail is not going to have an effect there, it's now controlled by antialiasing level, and meanwhile the terrain is fairly dark so high detail is not required. Not to mention that again with raytraced objects, the detail level is controlled by the antialiasing. So as long as you get those settings right, everything will look good and render faster with 0.5 detail.

- Oshyan

dandelO

#35
* On the side...

I've noticed extreme differences between renders when cloud layers have used higher that Q=1.
As example(and sorry, I don't even have any temp files to back this up but, I remember clearly as it's a WIP that I often open...), I have a file that uses Q=4 for cloud detail and looks lovely(this cloud layer is used as an atmosphere in itself, at ground level as a haze, and looks terrible at Q=1, the entire scene looks different at Q=1) Could this be the difference in output you're experiencing, Oshyan? I only post because I remember Bob/Markal's file used around 'Q=50' for quality so, it would then give an extremely different output than what you might have tuned it to use in an optimized version...

* All quality settings I refer to here are cloud detail only, not render detail. To clear up any confusion. *

TheBlackHole

You don't need suns to make the aurora visible. Just turn up the ambient color in the cloud layer.
They just issued a tornado warning and said to stay away from windows. Does that mean I can't use my computer?

FrankB

#37
@Dandelo: but that should only give differences in the render quality itself, and not for example change the location of the clouds.

choronr

Hey Mark; you out there? Why don't you jump in here and give us your thoughts.

Oshyan, Martin, your input in very important to us. We'll keep on with this until we reach an acceptable solution. Thanks so much.

dandelO

Frank, I don't mean 'render quality' of 1 and 4, I just mean cloud quality of '1 and 4' produces extremely different results. If I have time tonight I'll render out differences in these files, all that is changing is cloud quality, not scene detail. Hopefully back later...

Oshyan

Hmm, I must retract what I said earlier! It turns out I had been testing on the wrong version again. The current public release *does* have different results from the alpha. So I will look into this further on the development side. Good thing we caught it!

- Oshyan

Markal

#41
Some of the biggest changes I made to Bob's original file is:

lighting: envirolight - Ambient Occlusion....this lightened the image and helped the glowing effect so, I darkened the overall image under the renderers - effects - contrast and gamma correction.

atmosphere: both density fractals I increased the warp to 12 and 11 and also checked vertical warp...changed the noise type to perlin ridges and re-seeded many times....decreased the clouds to 858 and increased the depth to 5095. Sun glow amount was also increased. Dune mentioned the high cloud heights....the clouds are low. I think its the warp that has streached them to appear high and to cause the folds or curtain like appearence.
Hope this helps....thanks to all for putting up with me and your own inputs will be recognized when this file becomes available to all for free. Yeah a "Team" effort! :)
Thanks
Mark

Oshyan

Thanks for the additional details Markal. Hopefully my render setting recommendations will help you produce new versions of this scene with less headache! ;D

We have identified the issue I discovered and are working on resolving it. So it was very helpful to work with this scene actually, it identified the issue very clearly.

- Oshyan

choronr

Thanks to all of you for your inputs. Now, I need to summarize all of these suggestions and start re-working the image. My other render is near completion. Once that is done, I'll be back to: 'Borealis - North Light'.

Bob

choronr

Thanks to all - this is the result of all the great suggestions by you folks:  '6 hours, 49 minutes' render time. Much better than that 98 hour penance doer.
One last thing I'd like to do is see the bottom of the aurora; i.e., shorten it up so that one may be able to see a limited height of the aurora - bottom to top.

The adjustments I've made to the file are as follows:

•   Reduced atmosphere samples to 12.
•   Reduced the cloud samples to 52
•   Checked 'ray-traced' atmosphere to 'on'.
•   Set 'anti-aliasing' to 6
•   Adjusted sampling to 'customized' 1/16 first samples; 0.01 pixel noise threshold.
•   Although it was suggested to increase the soft shadow samples, I left it the way it was.
•   Tried the 'localization' of the clouds; but, that did not work.
•   Re-set Mark's 1st sun's elevation from 355.873 to 300.
•   For the Full Render, Detail was set at 0.8; Anti-aliasing at 6 instead of 4.

In all, there are probably many other tweaks that could be made here; but, at this point, I'm ready to move on. Should you have any thoughts, please pass them on to Mark or I.

Bob