Index of refraction < 1 renders wrong in cycles #38439

New Issue

Anna Celarek · 2014-02-01T23:31:51+01:00

Anna Celarek commented

2014-02-01 23:31:51 +01:00

System Information
Windows 7, Nvidia gtx550Ti

Blender Version
Broken: Cycles Render, 2.69 fd0b104
Worked: Internal Render, 2.69 fd0b104

Short description of error
In cycles, the glass shader with IOR < 1 does not render correctly. It behaves like IOR = 1. It should behave according to Snell's law http://en.wikipedia.org/wiki/Snell's_law and bend the at surfaces, no matter from which direction the light is coming.
In Blender Internal renderer, the behaviour is correct.
IOR < 1 is necessary for example to render underwater air bubbles. Then the IOR of the air bubble is 0.75 (the inverse of the water IOR).

Exact steps for others to reproduce the error
See attached file and screenshots.

Note: The drag and drop thing in the new bug tracker does not work for me, it asks me how I want to open it. Opera 12.10. I'm uploading it elsewhere. Would be nice if the old standard attach and browse button would come back (optionally).
http:///www.ania.xibo.at/tmp/ior.jpg
http://www.ania.xibo.at/tmp/cycles_ior.blend
http://www.ania.xibo.at/tmp/internal_ior.blend

**System Information** Windows 7, Nvidia gtx550Ti **Blender Version** Broken: Cycles Render, 2.69 fd0b104 Worked: Internal Render, 2.69 fd0b104 **Short description of error** In cycles, the glass shader with IOR < 1 does not render correctly. It behaves like IOR = 1. It should behave according to Snell's law http://en.wikipedia.org/wiki/Snell's_law and bend the at surfaces, no matter from which direction the light is coming. In Blender Internal renderer, the behaviour is correct. IOR < 1 is necessary for example to render underwater air bubbles. Then the IOR of the air bubble is 0.75 (the inverse of the water IOR). **Exact steps for others to reproduce the error** See attached file and screenshots. Note: The drag and drop thing in the new bug tracker does not work for me, it asks me how I want to open it. Opera 12.10. I'm uploading it elsewhere. Would be nice if the old standard attach and browse button would come back (optionally). http:///www.ania.xibo.at/tmp/ior.jpg http://www.ania.xibo.at/tmp/cycles_ior.blend http://www.ania.xibo.at/tmp/internal_ior.blend

Anna Celarek commented

2014-02-01 23:31:51 +01:00

Changed status to: 'Open'

Anna Celarek commented

2014-02-01 23:31:51 +01:00

Added subscriber: @AnnaCelarek

Anna Celarek commented

2014-02-01 23:33:28 +01:00

http://www.ania.xibo.at/tmp/ior.jpg

Thomas Dinges commented

2014-02-02 03:47:42 +01:00

Added subscribers: @brecht, @ThomasDinges

Thomas Dinges commented

2014-02-02 03:47:42 +01:00

Ania: You can submit the report without files, and then upload them in the first comment (Choose "Upload File" in the Action menu.

Developer Note:
@brecht: We clamp the IOR to "1.0f + 1e-5f" in svm_closure.h. We could change it to "0.0f + 1e-5f"? This way we would still avoid a division by zero later. The rendered result then looks similar to BI.

Ania: You can submit the report without files, and then upload them in the first comment (Choose "Upload File" in the Action menu. Developer Note: @brecht: We clamp the IOR to "1.0f + 1e-5f" in svm_closure.h. We could change it to "0.0f + 1e-5f"? This way we would still avoid a division by zero later. The rendered result then looks similar to BI.

Dries van Oosten commented

2014-02-02 18:09:28 +01:00

Added subscriber: @DriesvanOosten

Dries van Oosten commented

2014-02-02 18:09:28 +01:00

Would it be an idea to just avoid zero? A negative IOR is interesting to render metamaterials.

Brecht Van Lommel was assigned by Sergey Sharybin

2014-02-03 13:42:00 +01:00

Sergey Sharybin commented

2014-02-03 13:42:00 +01:00

Added subscriber: @Sergey

Brecht Van Lommel commented

2014-02-03 13:54:55 +01:00

You can just flip the normals on the object to get the same result.

I'm fine with changing the limit to 0.0 + 1e-5f though. Negative IOR I'm not sure about, I don't think our fresnel formulas work with that.

You can just flip the normals on the object to get the same result. I'm fine with changing the limit to 0.0 + 1e-5f though. Negative IOR I'm not sure about, I don't think our fresnel formulas work with that.

blender-admin commented

2014-02-03 17:08:34 +01:00

This issue was referenced by blender/blender-addons-contrib@eff3bd4e98

This issue was referenced by blender/blender-addons-contrib@eff3bd4e9880efc0c767e3098baa86ab23d77bc7

blender-admin commented

2014-02-03 17:08:34 +01:00

This issue was referenced by eff3bd4e98

This issue was referenced by eff3bd4e9880efc0c767e3098baa86ab23d77bc7

Brecht Van Lommel commented

2014-02-03 17:10:19 +01:00

Changed status from 'Open' to: 'Resolved'

Brecht Van Lommel closed this issue