PDA

View Full Version : Prism rendering



Polygons
10-28-2018, 12:53 PM
I tried rendering a realistic prism refracting some light rays without any luck (in 2018). Does anyone know if this is possible in the new LW?
I tried using several techniques to get the light color splitting effect. I even turned on caustics but nothing worked. I used the dielectric shader and the pbr shader. Even with the dispersion turned to 10000000 the refracted licht is still white. Also giving the dielectric shader some roughness did change a thing which in my opinion is strange because I would expect the licht to get blurred by the refraction/roughness but is did not change anything. I tried using GI on/ Gi off ... still no color separation.
Any ideas?

Kryslin
10-29-2018, 08:07 AM
I think while it can be faked in Lightwave - any version, use a gradient and a weight map - it can't be done following the laws of physics in the renderer.

I suspect that there are rendering methods that can do it (spectral rendering comes to mind), though.

prometheus
10-29-2018, 01:38 PM
You should be able to get color dispersion with dielectric, you do need some roughness in dielectric for any dispersion to get the effect working, if you have that set to zero..it will not work.
Not sure distant light would work, try spherical and area lights.

I donīt think the native lights is actuall doing anything, using sunsky environment or hdr imaging to get a full spectral lighting condition.
you can check use background color to hide the background environment.

MonroePoteet
10-30-2018, 10:37 AM
As far as I can tell, LW doesn't implement correct refraction for rays of Light, which would provide spectral dispersion of Light rays in LW2018. In both LW2015 and LW2018, enabling volumetrics to see the beam, the beam of the Light passes straight through an object with an Index of Refraction (IoR) above 1.0, implementing refraction as if the Light beam were a physical object.

For example, here's LW2015 with a Spotlight with Volumetric Lighting enabled and a luminescent red straight Rod parented to the Light. The rod passes through the Prism (IoR of 1.52), and you can see the refracted image of it. The refracted image of the Light beam follows the same path as the refracted image of the rod, which is physically incorrect.

143248

In the real world, when a beam of Light enters a refractive volume (e.g. something made of glass) with an IoR that's above 1.0 (IoR of vacuum is 1.0, for air it's about 1.0003), the beam changes direction towards the normal of the plane of the surface. Because different frequencies (colors) of Light refract at different angles, the result is a spectrum / rainbow.

Hypothetically, LW2018 implements a "physics based renderer", but the behavior of refraction for beams of Light appears to be the same. Here's the same scene converted into LW2018 with Volumetric Scattering enabled to see the beam of Light. Intensity Falloff has to be set to Inverse Distance squared (rather than OFF as imported) to see the Volumetrics!

143250

So, I think to get a spectrum exiting from the prism as well as the correct change in direction for the refracted beam of light, it'd have to be faked up.

mTp

P.S. I should mention that "faking it up" isn't very easy for an animation where the relative angle of the prism and the light changes during the animation, and it may be impossible (or more difficult) in LW2018 because LWDG removed the Light-specific nodal control over the Light's volumetrics (a really bad decision, IMO!), and the nodal setup for the Light itself doesn't allow real-time control of Spotlight cone angle or other parameters. Here's a nice video to show the changes in the spectrum intensity and shape as the prism rotates relative to the incident light:

https://www.youtube.com/watch?v=8eK-kFBgQU0

Note the shadow of the prism to identify the location of the light source.

Tobian
11-03-2018, 04:39 AM
Lightwave has a new physical renderer, but it lacks a number of things to make it do that trick. Dielectric still lacks true dispersion, it also needs a pathtracing engine, and to be truly accurate, spectral rendering. There's only s handful of renderers which support such advanced physical simulations, and mostly they are incredibly slow!

MonroePoteet
11-03-2018, 09:18 AM
Yes, I can believe it'd be very compute intensive to raytrace that degree of accuracy. It can be faked up with results that are adequate for a few simple (and *inaccurate*) animations, but trying to do (for example) the physically-accurate simulation of a multi-faceted prism, a cut diamond, etc. would be very slow.

Here's a simplistic animation in LW2018:

143279 143280

MOV file: 143281

To re-iterate, the behavior of the refracted light (the spectrum) is WRONG. To have it behave accurately, the node network in the Nodal Motion of the SpectrumLight would need to calculate the angle between the incident SourceLight and the refractive surface's normal both for the beam entering the prism and exiting the prism. As well, the intensity and shape of the spectrum / light exiting the prism would change as the incident angle changes. Since LW2018 removed the Light specific volumetrics and the Light's nodal network doesn't provide access to anything but its various Colors, I don't think a realistic, moving prismatic refraction would be feasible even in this simple case. I think! :)

mTp

Polygons
11-04-2018, 12:07 PM
The problem in LW is that there is no way of having slightly different refracting indexes for rgb colors I would expect dispersion to fake this effect but apparently it doesn't.

This is what I'm after. https://vimeo.com/95755327

Maybe we need a dielectric shader 2.0.

There seems to be something terribly wrong with refractions in LW2018. If you shine a light through a surface with some amount of refraction blur is doesn't blur the light. I would expect the light to be refracted and be blurred but it's not for some reason.

Tobian
11-04-2018, 01:40 PM
The dispersion effect in dielectric does split the 3 channels, but it's inefficient and causes splits, because it the 3 refraction indeces break up, instead of spreading into a spectrum. Spectral rendering us probably a while away, but stochastically sampling the dispersion is possible.

There's no way you're going to get it working in volumetrics, without bi-directional path tracing.

jwiede
11-04-2018, 07:21 PM
There seems to be something terribly wrong with refractions in LW2018. If you shine a light through a surface with some amount of refraction blur is doesn't blur the light. I would expect the light to be refracted and be blurred but it's not for some reason.

How are you setting up the surface/material where you're trying to cause refraction blurring?

Polygons
11-05-2018, 11:29 AM
How are you setting up the surface/material where you're trying to cause refraction blurring?

It's just a prism with a spotlight on it. The prism has a dielectric shader on it. The spot that shines through doesn't seem to be influenced by it at all except from coloring it if you add some color to the dielectric.

madno
11-08-2018, 12:46 PM
Just for fun a test with Iray in free Daz Studio 4.11.
imported geometry -> set a Daz Studio build in spot light -> assigned Iray Ubershader to materials -> adjusted some sliders -> added an HDR environment -> rendered.
Time was 18 min (one GTX 1070, one old Quadro K6000).

143311

madno
11-09-2018, 02:05 AM
Another quick test.
I am really impressed. It just works. No hacks, just a spot light and adjusting settings and materials.

143326

prometheus
11-09-2018, 04:05 AM
Another quick test.
I am really impressed. It just works. No hacks, just a spot light and adjusting settings and materials.

143326

Cool ..will have to take a look, if it works with daz studio 4 that is.
Have to check what it is that does the trick, thanks for the heads up.

Ztreem
11-09-2018, 05:39 AM
Cool ..will have to take a look, if it works with daz studio 4 that is.
Have to check what it is that does the trick, thanks for the heads up.

There's no trick. Iray is a bidirectional path tracer that's why it can do nice caustics.

When a software like Daz studio has a more advanced renderer than Lightwave you start to realize which state Lw is in, it's a sinking ship(or more like a sinking small boat).

lardbros
11-09-2018, 06:05 AM
There's no trick. Iray is a bidirectional path tracer that's why it can do nice caustics.

When a software like Daz studio has a more advanced renderer than Lightwave you start to realize which state Lw is in, it's a sinking ship(or more like a sinking small boat).

Of course... very logical. Arnold doesn't do this either, so what does that mean?

prometheus
11-09-2018, 06:06 AM
There's no trick. Iray is a bidirectional path tracer that's why it can do nice caustics.

When a software like Daz studio has a more advanced renderer than Lightwave you start to realize which state Lw is in, it's a sinking ship(or more like a sinking small boat).

Thatīs the trick :D
Thanks for explaining it well.

Ztreem
11-09-2018, 06:17 AM
Of course... very logical. Arnold doesn't do this either, so what does that mean?

The logic is that Autodesk didn't spend their time on a new renderer (they bought one). Newtek spent their developing time on a new renderer instead of just license one or use a open source renderer. The time could have been spent on all other missing aspects of Lightwave. Lightwave is just light years behind and with this developing speed they will never catch up. If you love convoluted workflows with no undo and no unificated work environment then LW is perfect.

- - - Updated - - -


Thatīs the trick :D
Thanks for explaining it well.

Luxcore for blender can also do this if you want to test.

prometheus
11-09-2018, 10:25 AM
The logic is that Autodesk didn't spend their time on a new renderer (they bought one). Newtek spent their developing time on a new renderer instead of just license one or use a open source renderer. The time could have been spent on all other missing aspects of Lightwave. Lightwave is just light years behind and with this developing speed they will never catch up. If you love convoluted workflows with no undo and no unificated work environment then LW is perfect.

- - - Updated - - -



Luxcore for blender can also do this if you want to test.

Thanks..cool, though a bit messy with the installation..I happened to choose the standalone and download, did not carefully read the the whole section in the page, just went for win 64 opencl.
so it didnīt show up in the addons, once that was tracked, I managed to find it in addons..though I got an error...could be that I do not have the right redistributin packs.

So a bit messy ..may try tomorrow again.
The page to download the right version is a bit messy, I honestly do not know how to seperate the addons properly from standalone builds, etc...whatīs so hard about dividing it properly.

lardbros
11-09-2018, 01:14 PM
The logic is that Autodesk didn't spend their time on a new renderer (they bought one). Newtek spent their developing time on a new renderer instead of just license one or use a open source renderer. The time could have been spent on all other missing aspects of Lightwave. Lightwave is just light years behind and with this developing speed they will never catch up. If you love convoluted workflows with no undo and no unificated work environment then LW is perfect.

- - - Updated - - -



Luxcore for blender can also do this if you want to test.

But not sure you understand my point.

You said that LW doesn't have as advanced a renderer as DAZ Studio, purely because it can't do spectral rendering. You seem to have brought up a load of other unrelated issues that aren't rendering related to smear the discussion.

All I said was, Arnold (which is one of the best renderers out there) also doesn't do spectral rendering, and it's been in development for years and years.

LightWave's new renderer is more than capable and I'm really enjoying using it.