Page 1 of 2
pic not clearing...
Posted: Wed Feb 22, 2012 12:37 pm
by Rogurt
Hi all
after some time testing with simple geometry I´d like to get a more complex scene to work with octane. But I am confused since after 7,5h rendering the picture is far from beeing cleared even in uniform non specular areas. Maybe someone could have a look at my specs and at the picture and give some impression whether the times are plausible.
I did the pic on pathtracing kernel and let it render to 64.000 s/px
Cheers,
Rogurt
Re: pic not clearing...
Posted: Wed Feb 22, 2012 12:46 pm
by t_3
hm, this is a problematic setup in general; lots of small light sources + environmental light, lots of glossy mats, only one entry for sunlight/env.light. you should have a look at the sample_rate parameter in the emitter, to boost the samples of the smaller lights; pmc and playing with the direct_light_importance may give better results; maybe a reduced maxdepth for better speed, and the use of a portal (normals inwards!) covering the door could also help...
Re: pic not clearing...
Posted: Wed Feb 22, 2012 1:08 pm
by thiago
I'm not sure if that's the case, but try to decrease the hotpixel_removal setting a little.
Re: pic not clearing...
Posted: Thu Feb 23, 2012 12:27 pm
by Rogurt
thank you guys for your advice. I made some testing and maybe you´re interested in the results.
I tried to tweak the kernel settings as to give indirect lighting more processing time and to have more paths evaluated than going too deep into single paths. Next I tried to add portal and further tweak the sampling settings of the spot emitters (IES) on the ceiling.
In short: portal does a really good job - noise of environmental lighting gets reduced quite well
tweaking sampling time of bright spots does the rest - shadows and lighting in according zones are way better (less grainy).
So most of the time I think one can spare the fumbling in the kernel except for the bounce depths..
But see yourself

- OO_205_PMC_10k-spx_7.41min_001.png (181.64 KiB) Viewed 5163 times
1 all settings default except for depth=6 alphashadows=off

- OO_205_PMC_10k-spx_8.38min_002.png (198.62 KiB) Viewed 5163 times
2 exploration strength from 0.7 to 0.15 -> grainier less fireflies

- OO_205_PMC_10k-spx_7.02min_003.png (191.66 KiB) Viewed 5163 times
3 exploration strength from to 1.0 -> more noise and more fireflies than initial setting (0.7)

- OO_205_PMC_10k-spx_6.49min_004.png (190.57 KiB) Viewed 5163 times
4 direct light importance set from 0.1 to 0.02 -> again more noise than initial setting and more fireflies

- OO_205_PMC_10k-spx_8.09min_005-Portal.png (177.3 KiB) Viewed 5163 times
5 simple plane portal used in front of windows -> way less noise

- OO_205_PMC_10k-spx_8.42min_006-Portal_+Spot-sampling.png (166.21 KiB) Viewed 5163 times
6 tweaked sampling of main Spots in ceiling -> less noise in spot lit areas/shadows
Re: pic not clearing...
Posted: Thu Feb 23, 2012 10:41 pm
by ribrahomedesign
try to render with a 3x higher resolution and then downscale to the size you want ,your picture
will clear much faster
Rico
Re: pic not clearing...
Posted: Fri Feb 24, 2012 10:23 pm
by Rogurt
thanks for the advice. it really shows less noise when calculating for the same time with 3 times resolution than downsampling to original size (PSD standard filter). In addition the picture is remarkably sharper with more details where small dark features stand in front of a very bright background.
So that seems to be a really good thing to do...
I must correct my remarks on test #5. The portal does not seem to give remarkable noise reduction in that scene compared to standard setting (#1). Must have compared it to other pic than #1...
Edit: sample rendering attached

- OO_205_PMC_10k-spx_8.40min_001x3+downsized.png (156.12 KiB) Viewed 4891 times
Re: pic not clearing...
Posted: Sun Feb 26, 2012 1:44 pm
by Rogurt
Edit: Pathtracing was the kernel here.
Thought Pathtracing would be a lot faster. But as it seems after 8.40min the picture is not as consistent as PMC version (compared to test #6). In some areas less noise (shadows on floor) but the fireflies give me some extra time in Photoshop. I already had the hotpixel removal reduced to the lowest value that wont let small picture parts vanish...
Is there some rule of thumb as: in interiour scenes with much indirect lighting pmc is superiour in terms of quality AND speed or the like?
What unbiased kernel do you guy prefer?

- OO_205_Pathtracing_8.40min_007.png (156.68 KiB) Viewed 4862 times
Re: pic not clearing...
Posted: Sun Feb 26, 2012 5:31 pm
by Rogurt
For completeness rendered again with direct lighting kernel. Result surprises me. While there´s less noise in indirect lit/shadow areas there are lots of fireflies in the render. Also I thought taht the overall appearance would differ from the two pictures due to completely different kernels. But it´s quite the same.
So the big final conclusion. Drop everthing but PMC? I remain confused.
Any help for orientation appreciated...
Cheers,
Rogurt
Addendum: kept the bounces throughout all different tests/kernels at 6

- OO_205_directLighting_8.45min_008.png (159.22 KiB) Viewed 4865 times
Re: pic not clearing...
Posted: Sun Feb 26, 2012 9:56 pm
by roeland
Rogurt wrote:For completeness rendered again with direct lighting kernel. Result surprises me. While there´s less noise in indirect lit/shadow areas there are lots of fireflies in the render. Also I thought taht the overall appearance would differ from the two pictures due to completely different kernels. But it´s quite the same.
So the big final conclusion. Drop everthing but PMC? I remain confused.
Any help for orientation appreciated...
Cheers,
Rogurt
It depends on the scene. For scenes with a lot of glossy and specular surfaces or with lots of light sources PMC will work the best. For scenes with simpler lighting path tracing may work better.
--
Roeland
Re: pic not clearing...
Posted: Tue Feb 28, 2012 7:19 pm
by Draydin_r
Hmmm, I'm wondering what kind lights your using. I was having the same problems with my scene until I scaled down the IES emitters. sharpened the lights and dramatically reduced the noise emitted.