Yet another sampling issue? (White dots in eyes)

   2832   6   1
User Avatar
Member
71 posts
Joined: Oct. 2015
Offline
Hey guys, i just wanted to know what these spots are. Im sure its been asked before but i cant seem to find the right the answer or parameters to adjust it.
Edited by abvfx - April 16, 2020 01:34:02

Attachments:
emilyGroom_karma.jpg (357.4 KB)

Senior Groom TD @ Industrial Light & Magic - abvfx [abvfx.me]
User Avatar
Member
2045 posts
Joined: Sept. 2015
Offline
Yeah…I was going to post at some time…I think what your getting is same as my issue.

Karma Render Node:
Pixel Samples - 512
Light Sampling Quality - 5
Minimum Samples - 16

Render Geometry Settings Node:
Diffuse Samples - 14
Diffuse Limit - 14

And the settings I used are simply just throwing darts, not really knowing what I am doing, and these were the final ones I left it at after many trial and error attempts.

Scene is lit with single point light.

Attachments:
Cam2_DS14_PS512_2516x1440.png (868.7 KB)

User Avatar
Member
7837 posts
Joined: Sept. 2011
Offline
They look like indirect reflections. Try setting diffuse and reflect limits to 0.
User Avatar
Member
711 posts
Joined: July 2005
Online
Are you using Variance or Uniform for Pixel Oracle? Try Uniform if you aren't already using that.
User Avatar
Member
2045 posts
Joined: Sept. 2015
Offline
Thanks jsmack/DaJuice,

I turned the diffuse and reflect limits to 0 with not much change,

and the Pixel Oracle set to ‘Uniform’ instead of ‘Variance’ seemed to improve the matter although it still exists.

I'm going to try as a next step and change the tile size.

I made a new scene that I will post with some test results, but will refrain until the op has a chance to look back at this thread and get inputs for his situation…don't wan't to hijack his thread with lots of post from me.
Edited by BabaJ - April 17, 2020 15:17:39
User Avatar
Member
71 posts
Joined: Oct. 2015
Offline
BabaJ
I made a new scene that I will post with some test results, but will refrain until the op has a chance to look back at this thread and get inputs for his situation…don't wan't to hijack his thread with lots of post from me.

Please by all means share you findings. This thread is to share knowledge, you will not be hijacking at all.

Thanks for all your input. Ive found that just tuning up Pixel samples is enough and keeping everything set ti uniform kinda works.

But my skin examples are wayy too noisy now. When im doing SSS.
Senior Groom TD @ Industrial Light & Magic - abvfx [abvfx.me]
User Avatar
Member
2045 posts
Joined: Sept. 2015
Offline
So I finally got time to get back to this and a few more renders with the hip file included here. I couldn't use the original model because this was for a 3d printing project and promised a friend I made it for to keep the model itself under wraps.

In trying to recreate the same result I initially tried a number of spheres and elongated some of them but found the effect was much less pronounced. Deforming the geometry seemed to help bring back the desired replication of the ‘issue’.

As said before reducing both diffuse and reflect limits to 0 didn't appear to change anything and so for all the renders they are at zero.

I noticed the biggest gain in reducing the effect was increasing the pixel samples, going from 128 to 512. However, going from 512 to 1024 didn't make a difference.

Also changes in Light Sample quality did not appear to make any changes.

When the Pixel Oracle is set to variance there is a parameter available Minimum Samples that sits at a default value of 2. Increasing that to 16 seems to head in the right direction, however going to 32 doesn't do much.

Finally with Pixel Oracle set to uniform as DaJuice suggested there are parts which finally come a ‘resolve’ of the issue, yet some parts still remain like as if it is a texture rather than a smooth ‘reflection’. The initial image with the arrow points to the part of the image that never finally gets resolved.

One thing too, is that when using Pixel Oracle set to uniform, the render times shoot up really high relative to when set as variance.

So I am still wondering what could be done additionaly, preferably at first with Pixel Oracle set to variance before finally trying those new settings with uniform(if necessary).

Attachments:
Render_Testing.hiplc (1.1 MB)
Group_1.png (1.5 MB)
Group_2.png (1.5 MB)
Group_3.png (1.5 MB)
Group_4.png (1.5 MB)
Group_5.png (1.5 MB)
Group_6.png (1.5 MB)
Group_7.png (1.5 MB)
Group_8.png (1.4 MB)

  • Quick Links