Announcement

Collapse
No announcement yet.

Noise when rendering round Rect Light - questio about settings

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Noise when rendering round Rect Light - questio about settings

    Hello,

    1,
    I experience a lot more noise than usual, when using a round directional Rect Light (disc) on a background.
    I had to increase the sampling

    from min1 max4 (CPU, bucket), threshold 0.01, minimum shading rate 6

    and to get rid of the noise, I ended up with:
    max10, threshold 0.005 and min Shading rate 14.

    Is that typical? Is there any setting which I might have missed?

    the scene has several HDR-lights and other rect lights

    2.
    Question about the Minimum Shading Rate:
    The minimum shading rate has no disadvantages (with bucket CPU), when settings this higher? (only of course render time)

    Because I read, that it in progressive mode, when it goes higher it takes time away from sampling to enhance other shader details.

    Thanks!

  • #2
    Try resetting the noise threshold and min shading rate to the first setting and just crank the max subdiv to like 100 and see what happens. A lower noise threshold doesn't do anything if the max subdivs aren't high enough, and even 10 sounds low for a 0.01 threshold. I use 24 by default for a 0.02 threshold and often have to crank it higher.
    Also, you shouldn't have to mess with the minimum shading under normal circumstances in the latest versions of VRay, I'd guess that whatever advice you've read about it is likely outdated.
    __
    https://surfaceimperfections.com/

    Comment


    • #3
      The issue is that of any threshold based image samplers. The really low values of the gradient mean that only mean subdivs are being used. If you measure the noise value in the areas where you see noise you will most likely see values well below 0.005 or even values that equal that meaning the max subdivs will not be used or reached. One trick is to increase the gamma curve for the color adaptation and set it to higher value like 3-4 for example. That will result in the sampler to be much more aggressive in the darker areas as if its been brighter there.

      As a general rule though, the image sampler does not like darks. In fact if you were to invert the color where it went from black to white, the image sampler would do a much much better job.
      Attached Files
      Dmitry Vinnik
      Silhouette Images Inc.
      ShowReel:
      https://www.youtube.com/watch?v=qxSJlvSwAhA
      https://www.linkedin.com/in/dmitry-v...-identity-name

      Comment


      • #4
        Thanks.
        I was used of course from oolder VRay always to use higher values in Max Subdivs (standard 1min and 24 max). But wasn't aware, that the slider could be set higher than 10 with numerical input.
        Indeed the simple max value gave the best efficiency.

        About the Color mapping > Gamma threshold, that's interesting to know. I will check that next time.

        p.s.
        Today I was wondering, that even the IPR render in VFB is steered by the sampling values of the CPU bucket rendering. Despite IPR max noise is set to 0 and render time unlimited.

        Comment


        • #5
          But still I wonder for what "minimum shading rate" is good for then.
          E.g. I thought it's good for Fog with GI scatter, to set the value lower, that time is used better.
          But I'm not sure about this and tests take always really long time

          Comment


          • #6
            But still I wonder for what "minimum shading rate" is good for then.
            It's not good for much anymore. I've tried testing a lot of different scenarios, and sometimes tweaking it helps a little, but often it just makes things slower, and it's hard to predict how and when. Vlado has talked about removing it from the interface to avoid confusion.
            __
            https://surfaceimperfections.com/

            Comment

            Working...
            X