LuxCoreRender Forums

Hi,

I'm still not sure how to interpret the absorption depth value inside the volume definition.

The attached blend file shows 2 cubes. Both have the same size (10mm edge length).
The volume description that is used is "Color R1 G0 B0" and depth at 1mm.
My expectations were a that all light colors that are different from red is absorbed after 1mm of volume.
The render shows that the cubes are only slightly red tinted. Whats wrong with my interpretation of this values?

BR

I think you found a bug in BlendLuxCore.
It is caused by the world scale.
If I set world scale to 1 and scale everything to have the values you used, it works correctly.

I have the feeling that this is the easiest way to find bugs: change worldscale and then test a few things, and boom you found a bug

https://github.com/LuxCoreRender/BlendL ... issues/179

Fixed it.
https://github.com/LuxCoreRender/BlendL ... df31f3d93b

Hi,

thanks for quick response and even more quick bug fixing.

Once we're in that topic, is the volume diffusion parametre set also dependent on the worldscale?
In case of yes, could the fix be applied to them as well?

I think the mapping nodes (in material) also deal somehow dimensions and scales - do they work correctly?

Thanks you.

BR.

lighting_freak wrote: ↑Wed Jun 20, 2018 10:01 am Once we're in that topic, is the volume diffusion parametre set also dependent on the worldscale?

You mean the scattering? I don't think it is dependent on worldscale, if I remember correctly it is just a probability (unitless).

lighting_freak wrote: ↑Wed Jun 20, 2018 10:01 am I think the mapping nodes (in material) also deal somehow dimensions and scales - do they work correctly?

Good catch. Fixed: https://github.com/LuxCoreRender/BlendL ... 19c6126d7d

Hi

You mean the scattering? I don't think it is dependent on worldscale, if I remember correctly it is just a probability (unitless).

No. I mean the path length in heterogenous volume.

The whole topic makes me thinking of thid thread:
viewtopic.php?f=4&t=87&p=1670&hilit=Spread#p1670

I hope we will find a way to solve this.

Currently, using the world scale is more or less pointless.
I explained why in another thread: usually you use it to pretend that 1.0 as a float number is not 1 meter, for the sole reason that floating point precision is limited.
So BlendLuxCore should do it like Cycles and ignore the world scale value on object transformations (however we should do better than Cycles and use it on other stuff, like camera f-stop and everything else that's affected).

It requires some big changes and I currently have nearly no time, so maybe I will get to it in 2 months or so.
In the meantime, I would recommend to do this: leave world scale at 1.0, and if you encounter floating point problems, adjust the min/max epsilon settings (there is a toggle button for "Advanced Settings" in the unit settings panel where you can adjust min/max epsilon).