Weak Caustics with BiDir, Path Caustics ok

[Basti]

I'm fixing this problem.

I fixed the problem and now this is the BIDIRCPU output:


bidir.jpg

Wow Dade, thanks for the fast reply and the fast fix, too.

[Basti]

Hello community,

sorry to bump this thread again, but i’ve downloaded the new BlendLux Version ( BlendLuxCore v2.0 )

And tried my old scene again (first post in this thread ), and noticed that the BiDir render-mode is still flawed.
Somehow the generated caustics in front and behind the sphere are not in line with the light source.
The path-tracer one works like it should.

I’ve attached some images to explain further.

[Dade]

And tried my old scene again (first post in this thread ), and noticed that the BiDir render-mode is still flawed.
Somehow the generated caustics in front and behind the sphere are not in line with the light source.
The path-tracer one works like it should.

I’ve attached some images to explain further.

Can you attach the test scene with that point of view ?

[Basti]

Can you attach the test scene with that point of view ?

Here are the files, and thank you for the fast response

LuxCore_Scene_BiDir_Caustics_Lightsource_direction_001.zip

.bcf and .blend

(3.68 MiB) Downloaded 309 times

The HDRI i've used is the same from here : HDRI Haven

https://hdrihaven.com/hdri/?h=furry_clouds

[kintuX]

IIRC, there's something wrong with HDRI rotation. At least last time i checked it was quite different than in Cycles. Gonna test now.

EDIT... looks fine, just intensity differs a bit, but was expected... sorry for the false alarm

EDIT2
I think Path caustics with small resolution (1k) HDRIs are a bit deceiving. Use larger one (at least 4K for nice caustics).
Also, if you create a Lamp (small light emitter) to coincide with the sun rays (position it on the sky where the brightest spot on HDRI is), you'll get to see 'correct' results much better. And you'll see HDRI caustics coincide with those of additional Lamp.

example (4K hdr used)

[Basti]

IIRC, there's something wrong with HDRI rotation. At least last time i checked it was quite different than in Cycles. Gonna test now.

EDIT... looks fine, just intensity differs a bit, but was expected... sorry for the false alarm

EDIT2
I think Path caustics with small resolution (1k) HDRIs are a bit deceiving. Use larger one (at least 4K for nice caustics).
Also, if you create a Lamp (small light emitter) to coincide with the sun rays (position it on the sky where the brightest spot on HDRI is), you'll get to see 'correct' results much better. And you'll see HDRI caustics coincide with those of additional Lamp.

example (4K hdr used)
HDRI+AREA comparison.gif

Hello kintuX,

thank you for your post, in your example gif : have you used Path or BiDir rendermode ?

Because i've changed the HDRI resolution ( tried 2k, 4k, an 8k ) and the results are always the same.
The focal point of the sun is always shifted to some degree and not in the vertical center of the shadow.

With BiDir, not Path of course.

[kintuX]

Hello kintuX,

thank you for your post, in your example gif : have you used Path or BiDir rendermode ?

Because i've changed the HDRI resolution ( tried 2k, 4k, an 8k ) and the results are always the same.
The focal point of the sun is always shifted to some degree and not in the vertical center of the shadow.

With BiDir, not Path of course.

Yes, only BiDir was used.

Wooow, after I had checked everything again just to be sure... now i FOUND the real cause WTF, this bug is caused by the receiver, the ground geometry. Also the glossy material used only exagerrated the effect. Subdiving, cutting, adding edge loops, setting epsilon... nothing really helps. Only scaling ithe ground plane above 5m^2 minimizes the flaw after as it seems as if it's correct for larger areas.

Could this stretching be somehow associated to triangulation, the Terminator thingy or...? That's just crazy. I only hope other lighting is not influenced, affected by this. Yes, only with BiDir.

We might be stupid, but at least we're not crazy

[B.Y.O.B.]

Does it also happen when the plane stays small, but a cube is placed farther and farther away from the origin?
The environment light sources (sky, HDRI) enclose the scene in a sphere that is a bit larger than the bounding box around all objects in the scene.
Maybe it's related to this.

[Basti]

Yes, only BiDir was used.

Wooow, after I had checked everything again just to be sure... now i FOUND the real cause WTF, this bug is caused by the receiver, the ground geometry. Also the glossy material used only exagerrated the effect. Subdiving, cutting, adding edge loops, setting epsilon... nothing really helps. Only scaling ithe ground plane above 5m^2 minimizes the flaw after as it seems as if it's correct for larger areas.

Wow nice find, i haven't even noticed that.
And you're right the bigger the plane the smaller the shifting.

Does it also happen when the plane stays small, but a cube is placed farther and farther away from the origin?
The environment light sources (sky, HDRI) enclose the scene in a sphere that is a bit larger than the bounding box around all objects in the scene.
Maybe it's related to this.

I made some test with a cube like you suggested.
And it's like you said, it looks like it is somehow the bounding box of the scene, which affects the focal point of the sun.

I rebuild the scene the groundplane, sphere and camera are in the center of the scene (x0, y0).

I made some images, which show the effect with the cube in different positions.

[Dade]

I'm looking into this problem.