[jep]

Hold on!

Did you guys take the time to read the papers after you watched the movie?
personally, I had to stop salivating first!

In any case - Luke, the "workaround" you are doing is very close to what the paper speaks of... The only difference is we have to implement an ad-hoc directional light (parallel rays) via "the spike" rather than having a true drectional light (no falloff - not location dependent)... that's it - brilliant really.

I wonder how we could make a true directional light in Maxwell... ideas???

They aren't doing a lookup of the true sun either! Look at this quote:
We utilize a simple solution, approximating the sun as a directional light, while using traditional sampling techniques for the remainder of the sky.

[jep]

This is just to noisey!!! Even if I just use ONE single poly plane as an emitter, the noise is through the roof!!! So, this workaround might work in theory, however in practice, it's not passable - anyone got any ideas? I wish pure white HDR maps cast shadows.....

[Micha]

... I think, HDRIs with high contrast areas will produce much noise.
It's standard rule of biased renderers: high contrasts produce much artfacts. Only use very blur HDRimages for lighting and detailed images for reflection/camera view only. In Maxwell is it approx. the same.

[3dtrialpractice]

yes still much noise,
heres my latest setup for this, i went away form the deform the dome idea, to having a very low poly dome, with a Open eneded spike (pyramid) mapped with a crop section from the original mxi of the sun, and the intensity turned up a bit compared to the dome sky mxi im using.

ok heres the setup.


pretty simple, but fairly noise free
and the render, to 12 samples about an hour. I found a very big varaiation of speed in the renderer i guess due to geometry, at one point this scene took 1 hour to get to 5 samples!!.. same scene, then i deleted the dome and the spike, and inserted clean new ones, and wahm, in an hour it went to 12 samples, so the geometry I was fooling with somehow slowed the render way down, it seems like others said, do all your uv mapping, texturing , modeling, before applying the maxwell plugin and shaders, once i applied the shaders to it, and deleted faces, and reprojected my textures it killed the speed.



ill post more soon.
-luke

[Micha]

You can not use a so low polygon count, because the image in the background show a very strong distortion now.
But I have a solution for you - use a cube cross environment on a six poly cube like I explain in this post:
http://www.maxwellrender.com/forum/view ... highlight=

The best could be, we got a special projection type, so that we can use a cube with a "spherical projection" of the standard environment image. So it was working in my good old HDRI shader for renderman, working on all geometries.

[3dtrialpractice]

You can not use a so low polygon count, because the image in the background show a very strong distortion now.
But I have a solution for you - use a cube cross environment on a six poly cube like I explain in this post:
http://www.maxwellrender.com/forum/view ... highlight=

The best could be, we got a special projection type, so that we can use a cube with a "spherical projection" of the standard environment image. So it was working in my good old HDRI shader for renderman, working on all geometries.

yes, you make a great point about the distrotion, I was going to render a seprete alpha and composite a clean distrotionless background.

Yet, you mention a box with spherical projection.
so I did just this, maily to see what a difference faces make on time.
BOY, I was shocked, your method is WAY faster...

here is jsut a render and time box Mxi emmiter VS. Dome mxi emmiter,,
box 25min to 10 samples
dome 1hour to 10 samples

box kicks dom's ass..
Scale is also Important, I test it so that the Pole in focus is 152cm (5feet) tall, so the ground extends to 3048cm's (100 feet) so this scene has Big scale, but thats what I need it to work with, a Big area.




-edit: added a 8 sided cylinder with a cap, same mxi emmiter mapped on the cyluinder with spherical projection. Just adding 4 faces (box was 5 faces total cylinder is 9faces total) added about 5 min to the final render time. But i like the cylinder over the box as the area has equal cover by the faces, like a sphere without allthe exta faces.



-Luke

[andronikos916]

intresting testing!

big thanx here!

cy,
Andronikos

[Micha]

Luke, interesting tests. You write, that the scale is important in one of your tests. Can it be, that your 3D application build automatic meshes of your objects (maybe you enable a polygon smooth option)?
In Rhino I can manual build a mesh from a sphere.

Big Scale: I use very big emitter objects around the scene, because, is the object to small, than you get no homogen lighting like in your image. The ground should show no light gradient like in the real world. I think, if the emitter sphere/box is 100 times bigger as the ground it works better.
Best is, if the outdoor MXI is extrem big and indoor MXIs are so big like the real room.

Also I see, you can use different projection types. In Rhino we have no projection types.

[jep]

The noise seems to be coming the emitter (sun light), itself. I can map a lat/long map on a fairly smooth sphere and render a fairly noise free image... Until I add the emitter to act as the directional sun - it gives me the beautiful harsh shadows but at the cost of major noise - when I take the "sun emitter" out, it goes back to fairly clean in a reasonable time....

[Swierk]

could you share that hdri or mail me, plz?

[ludenhud]

could you share that hdri or mail me, plz?

[img]http://lukeroman3d.cbeckleydesign.net/M ... R_hdri.jpg[/mg]

Or make it public and share it with me aswell

[vansan]

So simple, even I understood!
Nice way to achieve NORMAL shadows, not diffuse ones.

[Swierk]

once again...could somebody share that hdri,plz???