IRAY rendering, Glossy anisotropy + Glossiness, incorrect outputs

JD_Mortal

I was not sure where to place this, as it is not honestly a "bug", but it is something that is causing incorrect output in rendering.

To explain, "Incorrect output", I must use a rendered example.

Example, one "Cube Primitive", with "Emissive" surface, "Iray". Plus, one "Plane Primitive", with "!Iray Uber Base" surface. Rendering, default IRAY setup, rendering only the "Scene", and no "Ground", and no "Headlamps".

"Plane Primitive" material value for "Base Mixing" set to "PBR Specular/Glossiness"

1: Turn-on "Nvidia IRAY" for the preview window. Position the plane centered and directly under the cube, with a gap large enough to fit three cubes between the plane and the single cube above. On the plane surface, change IRAY surface value for "Glossiness" from 1.00 to 0.99 (Tap the "-" symbol on the slider.)

- Not much change. "Glossy Anisotropy" pops-up with 0.0 as a value. You should see the plane reflecting the glowing cube above, with a slight haze in the reflection.

2: Change the value of "Glossy Anistrophy", increasing it a bit, taking note of the changes. They are small, but I will amplify those later. Get to the value of 0.99, see there is not much change from 0.00 to 0.99... Now go one more, to 1.0

- Now, depending on the camera-angle, the camera rotation, and the plane rotation... You may see one of a few things.

- A: An upside-down "V" shape, with ramped-up curves, going from the edge of the plane to the place where the reflection should be...

- B: A "Waterfall" of light, streaming across the plane, then dropping down, as if it is casting over a rounded cube edge, though it is in the middle of the plane.

- C: A vertical line with an apparent "falloff", that infers the light is dropping in a linear path, through the tilted plane. (Devoid of any perspective, no matter what the orientation of the objects is.

- Expected results... 100% Anisotropy, not an object/camera/polarity oriented directional shifting, inconsistent light. A light which only seems to reflect/cast off two sides, into two directions, which changes if you rotate the "View", or the "Surface" but not the "light". (100% Anisotropy being a slight nudge more than 0.99, as 0.00 to 0.99 was "slight". (I am about 99.9% certain that the "reflection", should be oriented "fixed" to the camera "direction", and altered with camera-inclination or surface inclination only. Thus, the two streaks, one forward, oriented down without perspective, (symbolic of mirror-blaze, or ocean blaze), with the other streaking into the distance, after having pulled from the sides, creating that perspective surface glow, continuing to the center of the screen horizon, based on the "tilt" up or down of the camera. (Unless this should be streaming horizontal across the surface, side to side... Which wouldn't make sense in any respect.)

3: Failure seen slightly at 0.50 "Glossiness", and more apparent as you go from 0.50 to 0.99... Until you hit 1.00 Glossy, then it kills the "Glossy Anisotropy" slider.

4: This happens to "Backscattering Glossiness" and "Backscattering Weight" also. It works up to about 90% 0.9, then fails, reflecting to one specific direction, instead of "Back", as the other should be reflecting "Forward and back", or "Side to side"???

- NOTE: "Backscattering Weight", steals "Anisotropy Glossy" value, and directs it backwards (away from the camera) Not sure why it would reduce Anisotropy and glossiness at all, it is just another form of micro-surface illumination, not an alternative to anisotropy and glossiness.

Thus, with the noted pictures, this is not what the expected output for lights should be. Thus, items which use these settings, near these limits, or just with conflicting values between backscattering and anisotropy, are being displayed incorrectly. Especialy where formulas push, or expect correct values for shaders, in those range limits, like metals, mirrors, water, glass, chrome, fog, and even "flat" surface, plastic and emissive lights.

Attached pictures

1: Upside-down "V" shape. (Glossy Anisotropy), with right-side "Backscattering" which is somehow coming off the cube itself, not the surface below. Both rotate with the camera or the surface, as if painted on the surface when rotating the surface, or painted on the surface when rotating the camera.

2: Image shows that the backscattering is absolutely coming off the object, but it is displayed on the surface. Also shows the waterfall linear drop-off mentioned above. (Again, this is a solid flat surface below the cube, a Plane primitive.)

3: To show you that backscattering is no-longer "behind" or "back", I turned it around more, so it points forward. Now the waterfall points in the other direction too.

4: This last picture shows "backscattering Anisotropy 0.50" within the slightly normal values, and "Glossy anisotropy 0.50" with a more realistic mid-value, where both show, as one consumes the other as the value rises. (You still see it pointing in a direction that it should not be. It is stuck always pointing to that side of the cube, or the other, when it isn't pointing directly down.)

I assume it was tested with a generic camera, in a generic 0,0,0 orientation, on an object which was aligned 0,0,0, where the "left and right" or "fore and aft" were perfectly aligned, so these two "effects", seemed to function as desired. However, no-one turned the camera, or the object, to see these two big biased directional errors. Now, they just can't figure out why nothing with these settings looks quite right, and just avoid them, or live with fixing the errors in post-edits.

P.S. Yes, I have rendered these with and without "Architectural" and "Caustic" samplers.

5: Last image is 0,0,0 orientation, which is what the "effects" should look like at all rotations/orientations. (Keeping in mind that the cube is floating above the surface by about 3 meters and is a 1x1x1 meter cube.

6: Added one more to show the plane and cube as they are displayed in all renderings, roughly... With the settings for the last render too.

JD_Mortal

Replying to say that you may have to enhance the brightness on the images. Daz (website), seems to have "altered" the images colors and brightness. Repeating the steps in daz3d is more clear what is going on. Images can only show so much. :P

I am uploading them "enhanced", and hope the website image compressor doesn't alter them again.

In the last image, to make it more clear... You should see this same image no matter what orientation the camera or surface is rotated... Like water reflecting the land above, it does not bend land if you look to the side. If this were a glowing box floating above a ground, you would see the same thing that a guy standing on the left or right or back would see. You would all see a tree, upside-down in a puddle of water that same way too.