Announcement

Collapse
No announcement yet.

Noisy spec pass/highlights

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

  • Noisy spec pass/highlights

    I hope you can help me out in diagnosing these firefly-looking highlights I'm getting on DoF-renders.

    I have prepared some screenshots so I hope they help you in understanding my problem.
    I'm also including the scene, as that seems to be the first thing you ask for when diagnosing.

    First: the scene.
    It is a still frame from an animation course, that must include DoF by animating the camera target in a VRayCam. (The big ball is (almost) in focus, the bg is blurred)
    (Only 5 images pr. post, so you have to go here to see the scene, sorry: https://drive.google.com/file/d/1GC3...ew?usp=sharing )

    Rendered pretty fast just to show the DoF working, with the following settings:
    Click image for larger version

Name:	02_quick_render.jpg
Views:	388
Size:	374.3 KB
ID:	1114542

    I am not a beginner so I know that DoF in V-Ray requires a lot of samples and I (think) know where to fix it.
    So with upped samples:
    Max subdivs @ 24 (I'm not allowed to post more than 5 images. Sigh.)

    Beautiful Dof!
    But the highlights on the glass- and chromeball does look firefly-ish still.
    The materials are straight forward - you can peek in the attached file to see if I made any mistakes.
    The only thing I guess that can be troublesome is the glossiness for both relfec and refrac is 1.0 (for now!)
    With the amount of samling required to fix DoF in V-Ray, I upped it further:
    Click image for larger version

Name:	04_even_higher_samples.jpg
Views:	311
Size:	318.5 KB
ID:	1114543

    This still isn't pretty.
    I backed of the sampling a bit as I really wouldn't want to render the scene with max subdivs @ 100.
    I then introduced some glossiness to all the specularities as shown here - partly to make the job easier for the render and partly to get more realistic results.
    The reflection glossiness is also changed to 0.98 in the chrome material.
    Click image for larger version

Name:	05_saner_sampling_with_glossiness.jpg
Views:	269
Size:	611.1 KB
ID:	1114544

    More or less the same. (But notice that refractions are now behaving nicely! Not reflections/specular though)
    There a some slight nuances in the highlights now though and that is nice.
    I then tried upping the min. shading rate:
    (Unfortunately it is not allowed to attach more than 5 images, so this one is out. But as I said: no noticeable difference.)


    This did almost nothing - maybe it should have been higher, but I really don't see much improvement.
    Backing off min. shading rate to way too low levels for a production render, I continued to find the culprit and lowering the glossiness further did do something, but I'm not interested in having the glossiness values this low!
    Click image for larger version

Name:	07_more_glossiness_low_shading_rate.jpg
Views:	264
Size:	611.8 KB
ID:	1114545

    And this is where I'm currently at.
    I have monitored the sample rate while doing this and is behaves nicely by maxing samples the correct places.
    Even with max subdivs @ 100 it maxes out, which is expected I guess, as it requires all the samples it can get.
    (Unfortunately it is not allowed to attach more than 5 images...)

    I poked around on the forums before creating this thread and I have picked some nifty little tricks and tried to apply then, but the results are not really fixing the problem.
    You can see the two "nifties" highlighted below:
    Click image for larger version

Name:	09_various_tricks_applied.jpg
Views:	272
Size:	337.7 KB
ID:	1114546

    If you made it here: thanks!
    So to conclude:
    How do I get better looking highlights in my out of focus DoF, without just setting max subdivs to like a thousand?
    And can I control it more individually - for example with a setting like subdivs multiplier in the V-Ray object properties, which also did nothing in my scenario (I guess it's because it modifies min. shading rate?)

    Maybe I missed something completely elementary, so please do explain in great details if you can.

    Link to the file:
    https://drive.google.com/file/d/15YA...ew?usp=sharing

    Thank you in advance!

  • #2
    Maybe you already saw this post https://forums.chaosgroup.com/forum/...of-bokeh-noise

    So I just did your region with 24min/256max subdivs for the attached result. I also set noise to .008.
    Progressive wins for time. I have no idea why it is so much quicker for exactly the same result (barring one pixel difference that I can see),
    though I suspect Lele or others will be along shortly to explain why that is

    This is with your setting of.98 glossy by the way.
    DOF is costly, so no getting away from high settings unfortunately.

    Hopefully that helps
    Attached Files
    https://www.behance.net/bartgelin

    Comment

    Working...
    X