Page 1 of 2

Cauchy B Coeff might produce wrong results?

Posted: Sun Nov 15, 2015 1:45 am
by Silverwing
Hi there,

I just rendered my glass scene again over night (17h) with a little dispersion.
However the Result is not giving me what I have expected.

Render algorithm was MLT-BiDir but it looks the same in Pathtracing-BiDir.

The Renderer was Indigo fo Cinema 4D 3.8.26.2
Cauchy B Coeff was 0.005
Indigo_Glass_Cauchy_Test.jpg

Here is the original render. Same Material but without the Cauchy B Coeff
It looks more natural (is not brighter then the background).
Also notice the silhouette of the glass thickens inside the glass is much more pronounced with the Cauchy B material.
Indigo_Glass_Test.jpg
Cheers,
Raphael

Re: Cauchy B Coeff might broduce wrong results?

Posted: Sun Nov 15, 2015 4:34 am
by zeitmeister
This is strange and definately shouldn't look that way. Did you check your scene scale?


Cheers, zeiti

Re: Cauchy B Coeff might broduce wrong results?

Posted: Sun Nov 15, 2015 5:48 am
by contegufo
Hi

The scene was lit by an environment map?

Re: Cauchy B Coeff might broduce wrong results?

Posted: Sun Nov 15, 2015 7:49 am
by Silverwing
zeitmeister wrote:This is strange and definately shouldn't look that way. Did you check your scene scale?
Cheers, zeiti
The Scene scale should match real world. I´m not sure if it fits exactly because I do not know exactly what scale is exported by cindigo. But at least in C4D everything is in scale.
zeitmeister wrote:This is strange and definately shouldn't look that way. Did you check your scene scale?
Cheers, zeiti
Yes the scene is lit entirely by a HDRI 360° map.
I should mention that this is the result of a linear output. (Untonemapped EXR)

It looks perfect with Cauchy B set to 0.

Cheers,
Raphael

Re: Cauchy B Coeff might broduce wrong results?

Posted: Sun Nov 15, 2015 11:55 pm
by thesquirell
Dear Raphael,

Cindigo will export in scale which is chosen in Render Setting, Enviroment, Scene scale. I think it just takes the numbers from the scene and it just takes them as they are, without conversion. But I am not sure about that. To avoid all the confusion about the scale etc. I make sure my project settings say the same scene scale as Scene scale in cindigo's render setting, which is the same as unit display in C4D's preferences. Play safe! :D

Would love to hear how this develops. Never had any problem with CBC before.

Re: Cauchy B Coeff might broduce wrong results?

Posted: Mon Nov 16, 2015 12:48 am
by zeitmeister
What Happens if you use the cindigo version before? And does it come correctly of you use a coefficient of 0.001?


Cheers, zeiti

Re: Cauchy B Coeff might broduce wrong results?

Posted: Mon Nov 16, 2015 2:10 am
by Oscar J
Oh my. Though Indigo at least had its basics right. :(

Re: Cauchy B Coeff might broduce wrong results?

Posted: Mon Nov 16, 2015 2:21 am
by OnoSendai
Hi, send over the scene, I'll take a look (eventually).

Re: Cauchy B Coeff might produce wrong results?

Posted: Mon Nov 16, 2015 4:56 am
by Silverwing
thesquirell wrote:Dear Raphael,

Cindigo will export in scale which is chosen in Render Setting, Enviroment, Scene scale. I think it just takes the numbers from the scene and it just takes them as they are, without conversion. But I am not sure about that. To avoid all the confusion about the scale etc. I make sure my project settings say the same scene scale as Scene scale in cindigo's render setting, which is the same as unit display in C4D's preferences. Play safe! :D
Would love to hear how this develops. Never had any problem with CBC before.
Yes that´s what I have been doing too. I checked the scene scale with the apperture. Whrong translations end up with the wrong DOF. So if the DOF is way to strong or not there with a Apperture 3.5 you know there is something off the scale.

I have set the Scale to mm within C4D and within the Exporter.
I changed it to cm and to m just to see if it makes a difference. But unfortunately it does not. Cauchy B always has that magenta tint.
zeitmeister wrote:What Happens if you use the cindigo version before? And does it come correctly of you use a coefficient of 0.001?
Cheers, zeiti
No I have not checked yet. Maybe I should.
I have instead exported it to the latest Indigo standalone v3.8.29

I added some pictures:
Indigo_No_Cauchy.jpg
Picture with Cauchy B at 0
Indigo_Chauchy.jpg
Picture with Cauchy B at 0.00001
I will upload the scene later.

Cheers,
Raphael

Re: Cauchy B Coeff might produce wrong results?

Posted: Wed Nov 18, 2015 12:40 am
by Silverwing
As Promised here´s the scene.
I hope this *.pigs file works!

If not just write me a line!

GLASS-SCENE

Cheers,
Raphael

Re: Cauchy B Coeff might produce wrong results?

Posted: Wed Nov 18, 2015 1:39 am
by OnoSendai
Thank Sliverwing.
Looks like it's a problem with the combination of dispersion (single wavelength rendering), and non-post-process aperture diffraction.
Easy workaround - use post-process aperture diffraction only (or turn it off).

Re: Cauchy B Coeff might produce wrong results?

Posted: Thu Nov 19, 2015 12:28 am
by Silverwing
OnoSendai wrote:Thank Sliverwing.
Looks like it's a problem with the combination of dispersion (single wavelength rendering), and non-post-process aperture diffraction.
Easy workaround - use post-process aperture diffraction only (or turn it off).
Oh such an easy fix. Thanks a lot for having a look at the scene!
Much appreciated!

Cheers,
Raphael

Re: Cauchy B Coeff might produce wrong results?

Posted: Fri Nov 20, 2015 1:07 am
by Silverwing
Sorry for being so picky.
Yes the Aperture Fix got rid of the coloring. But I still get this stange "Artifact" in the rendering with Chauchy B Coeff. The Silhouette of the glass gets much more pronounced (Marked Green). I think there is something off!

P.s. I used the old colored rendering for comparison. The test with the Aperture fix only gets rid of the color!
Glass_Comparison_01.jpg
Cheers,
Raphael

Re: Cauchy B Coeff might produce wrong results?

Posted: Fri Nov 20, 2015 1:37 am
by OnoSendai
That's no problem for being picky, good spotting :)
There's a known issue/limitation with the dispersion code that I should probably fix. It's to do with the computation of the fresnel reflection term and the actual (waveelength-dependent) IOR for the wavelength as opposed to the base wavelength.

Re: Cauchy B Coeff might produce wrong results?

Posted: Fri Nov 20, 2015 6:29 am
by galinette
What if you increase cauchy B to a value such as 0.005?