View Full Version : mental ray artifacts

09 September 2004, 02:52 PM
Hi there.
I'm setting up a scene and taking into account that Im not a mental ray master, and I have a slow machine, it has become a little difficult...

Well, I ve got an indoor scene lit by a mr spot light (the light comes from the outside through a window) and I added a Skylight with a value of 1.5

The scene is made in real life scale. In mental ray, I use a slow decay (1.5) to light the scene interior and I use Max exposure control to control the brightness and contrast.
I use GI. Sampling 500. Photon size 50 cm. (so they overlap each other).
OK, its cool so far...

The problem is with Final Gathering. Im using a sampling value of 500
and MAX radius 50 / MIN radius 5
With these settings some artifacts appear. When I increase the MIN radius to 10, a lot of them are gone but the resolution decreases... as you can see (too much smooth there)

Take a look at the images below and gimme a hand please...

How can I eliminate those artifacts? increasing sampling? or is there another trick?

Thanks a lot buddies

09 September 2004, 12:06 PM
come on guys! I need to finish my render I want it to look good!

i don't know what's going on. How can I eliminate that patchy look?
Im lighting the interior with only one light coming through a window. Maybe this scene configuration becomes the lighting patchy???

I tried a lot of tricks but no one works! if I just set the FG MAX and MIN value to higher values, the patches are gone but resolution go hel*!!!


09 September 2004, 02:29 PM
you're missing something here.. you should use basic fg setup for final renderings.. or generally understand how fg works.. there're tons of tuts and advices on the net and for sure also a lot on this same page.. settings like 500, 50, 5 are crazy for a final rendering.. you could use them just to get a diffuse light solution.. but not to get details.. with that setup you'll generally use an occlusion shader.. to get the whole only with fg you should use something near the brute force approach.. 1000, 0.01, 0.001.. and it can take up a lot also on a fast machine... finally you can try to leave the min, max at zero and let mr decide on a per-scene basis what radius your samples will have.


09 September 2004, 02:43 PM
oh thanks a lot francesca!!!
I tried with 500 FG samples and Auto MIN and MAX but there are some bright artifacts... (small ones, similar to photons) in the areas where there is light bounce.

I don't know I made render test with MR and took 9 minutes with that config (just a very basic scene), and the same scene with Vray took just 9 minutes (the same!) using irradiance maps but the quality absolutely better!

I just can't figure out this... oh man...
Why is so difficult in indoor scenes?

09 September 2004, 02:52 PM
oh boy, Im still with this issue...
Now i have two renders to compare.

FG: 1000 rays for sampling and AUTO Min Max

Second test:
FG: 2000 ! ! ! final gathering sample rays, and MIN MAX auto

What's going on buddies?
What kinda POWER i need to render this?

Tell me your opinions...

09 September 2004, 03:10 AM
It's the FG radius settings, don't use Auto!! Type in your own values like Max 0.5 Min 0.05 and try it with 1000 rays again.

09 September 2004, 01:29 PM
thanks JOzVex, I made my try here:

FG rays: 2.000 / MAX radius: 1, MIN radius: 0.1
Render time: 4 hours 40 minutes! for God sake! (640 x 480)

Quality is not perfect yet, but better. I have to less the energy to take out those strange bleeds... but mainly I have to reduce Render Time! ! ! ! !
Any ideas?
Could I reduce the ray amount to 1000 or less, and reduce the MAX and MIN samples? you think that this is going to speed up the rendering?
are there other ways to speed up this thing?

herbert west
09 September 2004, 05:19 PM
If your scene has been modeled "in real life scale" then you should set your system/scene units to to real life units. Min/Max radius is easier to visualize is you're working in a metric or US feet system. You can actually specify how many inches or cetimeters the "artifacts" or photon/FG hits are. Using Max's default units, you're kind of just taking stabs in the dark unless you use the mesurement tool to measure your objects.

09 September 2004, 09:10 PM
thanks herbert!
Im working in real life scale (cms.) BTW

what value do you think Ive got to use man for MIN and MAX values?
Is it better to increase the FG rays or decrease the MIN and MAX value size?

Here's another test render:
FG Rays 1.000
sample size: MAX 0.5 cm / MIN 0.05 cm.
Render time: 3 h 53 min. (640 * 480)

The artifacts are still there... but less noticeable (but still there :sad:) and the render time keeps close to 4 hours...

herbert west
09 September 2004, 05:08 AM
Hmmm. Since you're working in metric, try 1 or 2 meters for a maximum radius (for both GI and FG), and a minimum radius of 5 centimeters. Now that I know what units you're using, it looks like you may have your radiuses (radii???) set too small. That might be giving the splotchy look. As for FG samples, start with 250, then go up on increments of 100 or so until you hit a good render. Also, try setting the number of GI photons in the Global Light Properties to 1000, and then the Photons and Samples in the Global Illumination and Photon sections (respectively) to, like, 2500 to 5000 and see what happens, then tweak from there, setting GI Photons higher (increments of 1000) and maybe lowering the photons and samples. If I understand right, it seems that setting photons and samples high gives sharper GI results, while setting them lower smoothes them out. In the end, it's all about tweaking and fine tuning. I'm all about speedy renderings, because I don't have the patience or attention span. :) If you missed it, I posted a modification of a scene LS3D did (Mental Ray Shader Discussion, pg. 94). It uses caustics instead of GI, but the settings should work well for GI also. I took it from a 10 minute render down to a little over 1 minute. Hopefully with some more tweaking of your scene you can shave a huge chunk of time off your current render. Good luck.

(edit) Also, adjust energy and decay. That might help with the glowing areas.

09 September 2004, 07:29 AM
whenever i am inputting manual FG + min/max settings, i start with large numbers in the min max, and small in the FG. try sowmething like "FG: 200 min35 max25" to start, and see what that does for you. slowly work down to smaller digits in the min/max, and larger in the FG. for min/max, the smaller the numbers, the longer the render, and for FG, the larger the number the longer the render.

as i understand it, the min/max is the radius of a 'ray' being sampled. if you are specifying that the ray is a maximum of .05cm wide.. well, then it will override the FG rays in teh scene, and create more rays to accomodate itself.

i know that is more than likely totally far from the truth, but as a mental picture (image, heh) it seems to give me a decent frame of mind for setting up an image. also, dont be afraid of unrealistic setting in your realistic scene. if it works, who cares? :)

edit: i just totally missed herbert west's post above. along teh same lines i believe :shrug:

09 September 2004, 11:22 AM
thanks a lot herbert and opus.
Im trying with different settings, yet. And I realized of a couple of things (maybe it could be useful for all the mental guys here):

1. When you texture the thing, the splotchy look of final gathering decreases. I mean that the effect is less noticeable, so you can make a little twist in the settings...
2. Exposure Control (in MAX) generally increases the photon contrast, so I prefer to work with decay (GI settings) and no with exposure control for indoor scenes (the blooby look is less noticeable too).
3. and as you said opus, the larger the number for MIN and MAX settings the faster the render... but I can't help saying that with large MIN and MAX values, the quality for the GI decreases so much...!!! the smooth result is totally linked to the quality loss...

CGTalk Moderation
01 January 2006, 05:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.