Chaos Corona Forum
Chaos Corona for 3ds Max => [Max] Bug Reporting => [Max] Resolved Bugs => Topic started by: petoboso on 2014-04-03, 10:51:04
-
Hi, I think corona needs something like vrays "subpixell mapping".
In scenes with small lights and many sharp reflections is imposible remove noise in reasonaible time.
I have tested simple scene with1 light and some glass. rendertime is more than enough for that simple scene and i didn't find the way to resolve it.
(http://i62.tinypic.com/15wf38j.gif)
(http://i58.tinypic.com/rm4dw7.jpg)
scene is in the attachment
-
GI/AA = 32?
MSI = 5?
LSM 0.2?
Try
GI/AA = 8-6
MSI = 20
LSM 12
Tint width = 2
-
Did you try the default setting before changing all these values?
-
nope, I'm not doing subpixel mapping, sorry
-
Subpixel mapping is a nonsense. Even Vlado himself suggests not to use it. Revert settings to defaults, then decrease GI/AA balance to 4 and give it another try. Also try to increase Highlight Compression.
-
Let me ask another thing.
Why are you rendering frames based on time? This will give you different noise amount in every frames, you should be rendering based on passes to achieve an uniform noise result.
Cheers.
-
Thanks, now im testing it by your suggestions, today i will put here something.
to Rawalanche: i know, subpixel mapping is "non realistic" feature and sometimes ruin quality of details, but without it a have to hugely increase AA, what mean very long rendertime.
-
Thanks, now im testing it by your suggestions, today i will put here something.
to Rawalanche: i know, subpixel mapping is "non realistic" feature and sometimes ruin quality of details, but without it a have to hugely increase AA, what mean very long rendertime.
That's how Vray works... not all renderers.
-
So finally:
Default setting:
(http://i61.tinypic.com/2rgz5lf.gif)
after some setting:
(http://i59.tinypic.com/4ftpc9.gif)
its better but still not clear, im testing this scenario because i did few animation inside dishwasher with vray and i would like use corona in future but im afraid of rendertime in this type of scenes.
(http://i61.tinypic.com/2cshvtu.jpg)
-
Vray has by default a lot lower trace depth. Like 6 or something. You could try to do the same.
-
minimal 12-16 is necessary
only way is decrease max sample intensity, but then i lose lot of reflections
-
Can you do Vray render for comparison? With same trace depth and with reflect on back side enabled? (Corona does not skip internal reflections by default) And of course with subpixel mapping disabled, because subpixel mapping is something like more aggressive form of max sample intensity.
-
vray without subpix is disaster. (that camera shake is some bug in max2014)
(http://i58.tinypic.com/2uny2pg.gif)
this is with subpix enabled- many reflection is missing, but nearly without noise
(http://i58.tinypic.com/eb68nq.gif)
if i compare without Subpix then is Corona clearly better than vray but because of rendertime i have to use vray +subpix
i will try with lower maxsample intensity today
-
In Vray with subpixel mapping, the reflections are clearly dimmer. Maybe in Corona you could:
-lower glass's reflectivity
or
-make the light less intensive and walls brighter
-
So why do you complain that with lowering MSI you lose reflection detail, when with subpixel mapping you lose a LOT more reflection detail? Simply lower max sample intensity to fairly small value (like 5-10). You can also try to raise GI/AA balance to get more refraction rays. But slightly. I think for this scene, value of 8 will provide a good balance. And you can also see that Vray render is also missing internal reflections, you probably forgot to turn them on which makes situation for Vray even easier.
-
Mmm... this is a very cool (and hard) scene. I would like to knot Ondra's opinion.
-
I agree with rawalanche on this one - lowering GI to AA ratio is the way to go. Changing MSI will probably have no effect on the fireflies. Introducing such huge face factors as subpixel mapping would go against the basic Corona philosophy: high-quality results with minimum number of algorithm-related parameters
-
Maybe you can share this simple scene?
I think we have to render this animation without this problems.
-
(that camera shake is some bug in max2014)
LOL. No. It isn't.
Anyway, while I don't have answer for this, I wanted to mention sub-pixel cuts-out highlights only by default, you can keep sub-pixel and high highlights by introducing more samples per material (goes in conjuction with AA though). That's only on side. It was extremely helpful to me on handful of animations.
Lower trace depth (for reflection and refraction) should help with this though, I know I never had more than 4 in any sort of animation (with Vray). I don't think either MSI is contributing factor, it's simply complicated reflection, since it bounces quite few times.
-
Please, don't start to talk, again, about VRay sub-pixel mapping! It was my devil!!! ehehe... XD
If someone knows my discussion with Vlado, he can understand what I want to say.
In VRay 2.x it was near to be impossible render an interior shot with sub-pixel mapping on! Especially with glasses and strong highlight!
The rendering time became so huge (from x3 to x5 time!)
But, if we introduce sub-pixel mapping in Corona and we can speed-up the rendering (from x3 to x5 times at the expense of quality) maybe we can try to think...
Especially in animation.
-
I want this scene! :)
-
Rawalanche: internal reflections was enabled in vray version.
Yes you are right i dont say it exactly. I didnt want very degrease MSI beacuse before i tried MSI in other scene and i realized that MSI has more influence on GI/reflected GI than sharp reflections, so
Look on this comparison: with MSI 10 vs 1
GI/AA=3
Tent width = 2
Reflections are very slightly changed and also noise is the same. what is different is wall behind light (GI influence)
(http://i58.tinypic.com/9gg8cm.gif)
I dont want talk against corona or vray. I would like use corona in future so thats why i want to resolve it. (critical feedbeck is needed for achieve maximum quality of product)
This scene i attached in my first post
-
Yup... if you could share the scene, it would be a lot easier for us to help you ;)
-
This scene is attached in my first post: corona_simpe scene.max
-
My latest discovery:
Looks like something is wrong with disabled caustics for the glass material. It makes rendering time much much much longer and there are more random white dots on the walls.
-
MSI-10
glossines=0.999
big difference, who would predict that, unfortunatelly kill too much light, maybe MSI15 would be fine.
now it could be only litle less blured- 0.99999? :D
(http://i62.tinypic.com/6jldl1.jpg)
-
now it could be only litle less blured- 0.99999? :D
You can increase precision of spinners decimals in preference settings. But i doubt it would make any difference in this case :]
-
Yup, i actually intended to suggest non-perfect glossiness, but i wasn't sure so i wanted to test it first. Looks like you were faster :)
-
Mmm... Interesting! Can we cal this a "bug". Maybe Ondra can do this by default
So, when we use 1.0 (as default), internally Corona changes this parameter as 0.999
-
now it could be only litle less blured- 0.99999? :D
Isn't it funny how oddly most of this stuff scales :- ) ? I know I very much prefer the glossy transition from 0-1 in Corona compared to Vray, it's much more linear. With Vray, 0.5 is pretty much Lambertian-alike already, so why is it even full scale.
Because of this, making small glossy imperfections for glass and mirrors is quite tough, because it blurs too fast.
Did you lower the reflectance as well ? Seems so much duller, and the previous renderings were bright even with MSI as low as 1. Or did the 1.0--)0.999 transition did that alone ? Seems weird the energy conservation would apply so drastically.
-
It would seem less odd if you would see the source codes - MSI is NOT applied to ideal (glossiness 1) reflections and refractions, but it is applied to everything else.
-
So, what we need is a "simple" script that change the glossy, for every glass shaders, from 1.0 to 0.999 (if the glossy parameter is set to 1.0)
-
Why? In 99% cases there are no problems with this type of fireflies, but by using this script you would get much worse rendering quality (missing reflections - no luster & brilliance). I can instantly make you version where MIS is applied to ideal glass, but it would hurt the result in most cases.
-
To avoid this such of flickering in animation, as in the petoboso scene.
But I understand what you want to say us =)
-
sure, this can be neat trick if you dont care about quality and want to do animation fast... so put it in your bag of tricks and use it where appropriate, but only there ;)
-
I can instantly make you version where MIS is applied to ideal glass, but it would hurt the result in most cases.
How about a checkbox next to MSI "apply also to ideal reflections"? Or hidden deeper in devel settings.
-
I agree to Maru ;-)