Hey Folks!
Making a personal project using Arnold, but I'm getting some noisy specular's.
My goals is to reduce my render time and fix the noise issue.
I know I can check my AOVs to see what is causing the problem, and it IS the Specular / Specular Direct. Arnold advises to raise the sample levels of the render and the lights, or raise the roughness of the shader, which I did and doesn't seem to work unless I raise Camera AA samples, which I don't want to do too much because it'll increase render time dramatically.
First question is: To fix specular noise do you raise the specular sample levels in just "Sampling" or "Ray Depth" or both? (I have done this and checked my specular AOV's and it doesn't work.)
The main offenders seem to be the cable bundle wire and the wire connecting the circuit board (SEE IMAGE)
The main cable bundle was made using ramps, using the same ramp greyscale for the bump, and a displacement map ramp to bulge out the cable as if its tied every 8 inches.
Maya Version - 2019.3.1
Arnold - 4.0.2.1
I am using aiAtmosphereVolume @ .004 Density / 8 samples
8 lights in my scene - Light Samples - 6 (7 area lights, 1 spot)
2 of them are projecting an animated texture - the green oscilloscope readout and the television
I think that may cause the noisy specular?
450 frames @ 2000x2000
Current settings
Light Samples - 6
Camera (AA) - 6
Diffuse - 3
Specular - 4
Transmission - 8
SSS - 2
Volume - 5
-
RAY DEPTH
Total - 10
Diffuse - 3
Specular - 4
Transmission - 8
Volume - 5
Trans Depth - 10
Please help!
Enable the filter section, beneath the sample section.
Set the indirect which defaults to 10.0 down to 0.01 and rerender.
This kills off all those nasty very expensive highlights at very low samples, and thus greatly speeds up your render, note that you will lose a bit of DR, but you can compensate with that very easily with grading the highlights in the Imager CC node, and you can raise the indirect in the filter section to something like...1.0 you can also go over, but the higher you go, the more time it takes to kill the sparks and time increases.
Dont worry, this is a classic case, and easy to treat without introducing long render times.
Hey!! thank you for your help!
May I ask if you're referring to the "adaptive sampling" area? or "clamping" ? (see image)
Set my "Indirect Clamp Value" to 0.01?
My filter is set to gaussian, with a width of 2
OK so ui looks like that.
Its the 10.000 in second last section that can be incredibly expensive to render clean at 10.000.
This float value has a sweetspot, that means you need to sample different values to get a sense about which amount of samples is needed and how pingy the highlight is before grading the highlight.
I usually enable the Clamp AA Samples as the first thing, then set both AA Clamp value and Indirect Clamp Value to 1.0.
Then work my way to a nice sweetspot that does not compromise everything in terms of look, while still being very snappy to clean up over time.
I can have cases where they both are below 1. I have cases where 0.01 indirect is spot on. It depends on the actual final product since that is an arbitrary thing, you cannot state any pair of values will be best for everything, it is very much a case by case thing.
You are essentially allowing to let rays parse at very high magnitudes, so the problem comes from the other end of the ray, not the hitpoint, it stems from very bright surfaces sending rays to the area that is difficult to make more uniform and less spikey.
All we do is tone down the spikes and doing that with fewer Samples.
In addition to improving your scene and/or settings, perhaps you can upgrade to the latest mtoa, which has a much improved adaptive sampler, and then just let the adaptive sampler handle those noisy bits?