Some test


#1

Hi to all.
I tried to render an old scene, doing some tests. I’m quite happy, but there are some points that do not convince me.

To be a 1920x1080 scene the calculation times are too high, maybe I exaggerated with the passes.
Moreover, there are the lower parts that do not receive any light, and not the low value of roughness are flat. What is it due to?



#2

Hi Magog,
Can you render the scene without denoiser, maybe the flat roughness patches come from the denoising. Another point, you have set unlimited bounces, maybe try limiting diffuse to say 3 and glossy to 8 and see if the scene improves.


#3

Tnx Mango!
I will try this evening!!


#4

I wonder if maybe the thin black wires / struts have their normals reversed? Could this be?


#5

I thought about it too, sometimes it’s nonsense, but everything was ok…


#6

What’s the material like on those struts?


#7

Now I’m doing a test with the new sets that Mango recommended.
Here the material.

This is the same scene, render at 2900x1700 px at 3000 sample with Cycles, 1h 20" .


#8

And this is the final result without denoise… :disappointed_relieved:

here you can see the time and material.


#9

appleseed is going to be slower than Cycles since it’s currently CPU only, there’s no doubt about that.

Maybe the rendering settings can be optimized somewhat, maybe @Jonathan_Dent can help here.

Note that the master version of appleseed has an important optimization for glossy materials, it would be interesting to check if/how that helps in your scene. I’ll try to give you a preview build. What operating system are you using?

I’m also hoping that we can figure why the struts are so dark. Can you try making them purely diffuse just to see if they receive enough light?


#10

Oh, yeah, I know… Moreover, I use it to calculate the image both cpu + gpu.

I’ve Win10 64bit.
If you want I can test the scene with Release 2.79b that 2.79.4 …


#11

I don’t undernstand…

I tryed with diffuse mtl and disney …


#12

The material you use is for scattering in transmission (BTDF Bidirectional Transmission Distribution Function) that’s why it looks like an absorber. Choose Lambertian BRDF instead for a diffuse surface.
I agree the naming should probably be changed to avoid technical terms which cause confusion.


#13

We’re in the process of replacing technical jargon by friendlier terms. Diffuse BTDF is a good candidate, it should be called something like Diffuse Transparency.


#14

I see that you tried the Disney BRDF for the strusts. Try making it completely diffuse instead of completely metallic.

I still suspect that normals are not correct on the struts (either they’re flipped, or they’re bent, or something like that). Any chance to display them?


#15

Maybe I solved…
I separated the affected mesh, renamed the material. Made quick test with Lambert mtl and then Disney.
Here are the screenshots

Lambert_

Disney_


#16

Are you willing to share the scene? :DD


#17

Hi Juan.
I don’t think I have more access to the scene for Appleseed.
Surely I have the original version for Cycles, if you are still interested I can gladly share it.


#18

Please! It would be cool to modify it a bit. :smiley:


#19

Try with this link:
https://drive.google.com/open?id=1wk9CfhWnhWmzP9fDbJa8CmNHHmPN7sVW


#20

Thanks man!