I’m trying to model woven shades using a BSDF from the CGDB. To get better looking and more accurate results I am using a BRDF as the standing geometry that is based off of a .cal file that greg wrote for metal grating: https://www.dropbox.com/s/bzo9apzcl4ioika/rollershade.cal?dl=0
when I apply this material to geometry it behaves as expected and cuts off the direct relative to the angle of incidence (which I hope explains this funny test scene I’m using)
There is a major change in the ambient values once the rendering gets to the window (which is why the band is horizontal in the single process run and vertical in the multiple process run. I ran these using rad so there is an overture calculation. Anyone have an idea as to what is going on?
What is outside your space? Did you model a ground plane? What rendering parameters is rad handing to rpict and rpiece?
My best guess is that when your rays get to the upper sill of the window, they bounce outside and create ambient values that then leak some energy into your scene interior. It shouldn't really happen with the latest release, so if that's what is going on, then there may be a bug in the new interreflection code.
If you want to send me your complete model, etc. offline, I can try to have a look at it tomorrow.
Cheers,
-Greg
···
From: Stephen Wasilewski <stephen@coolshadow.com>
Subject: [Radiance-general] Ambient File and BSDF
Date: September 11, 2014 6:57:27 PM PDT
I’m trying to model woven shades using a BSDF from the CGDB. To get better looking and more accurate results I am using a BRDF as the standing geometry that is based off of a .cal file that greg wrote for metal grating: https://www.dropbox.com/s/bzo9apzcl4ioika/rollershade.cal?dl=0
when I apply this material to geometry it behaves as expected and cuts off the direct relative to the angle of incidence (which I hope explains this funny test scene I’m using)
There is a major change in the ambient values once the rendering gets to the window (which is why the band is horizontal in the single process run and vertical in the multiple process run. I ran these using rad so there is an overture calculation. Anyone have an idea as to what is going on?
This turned out to be a subtle problem that I've seen on rare occasions in the past related to the direct approximation in Radiance. If you set -dt 0, the issue goes away (and accuracy improves). In 1990 or so, I implemented an optimization in Radiance for many light sources that comes up with a statistical estimate of direct illumination. It usually doesn't cause any issues, but in scenes with one or two sources and a lot of indirect illumination, it can result in systematic errors when you compute a scanline at a time.
In your scene, the indirect rays are all inside, and never see the sun directly. (Even the rays that hit the window count as indirect.) Thus, the statistical estimate of hitting the sun is very small or zero. Once the rendering reaches the window, however, it treats it like a trans and computes direct rays from the opposite side, with nearly 100% of hitting the sun. This causes a sudden shift in the statistical estimate, so that subsequent rays that forgo testing get this higher estimate.
As a fix, I've changed the code so with 4 or fewer sources, shadow rays will be sent every time, even when indirect dominates. This should eliminate this problem and others like it. I wish I had thought of this fix a long time ago....
Cheers,
-Greg
···
From: Greg Ward <gward@lmi.net>
Subject: Re: [Radiance-general] Ambient File and BSDF
Date: September 11, 2014 8:01:19 PM PDT
Hi Stephen,
What is outside your space? Did you model a ground plane? What rendering parameters is rad handing to rpict and rpiece?
My best guess is that when your rays get to the upper sill of the window, they bounce outside and create ambient values that then leak some energy into your scene interior. It shouldn't really happen with the latest release, so if that's what is going on, then there may be a bug in the new interreflection code.
If you want to send me your complete model, etc. offline, I can try to have a look at it tomorrow.
Cheers,
-Greg
From: Stephen Wasilewski <stephen@coolshadow.com>
Subject: [Radiance-general] Ambient File and BSDF
Date: September 11, 2014 6:57:27 PM PDT
I’m trying to model woven shades using a BSDF from the CGDB. To get better looking and more accurate results I am using a BRDF as the standing geometry that is based off of a .cal file that greg wrote for metal grating: https://www.dropbox.com/s/bzo9apzcl4ioika/rollershade.cal?dl=0
when I apply this material to geometry it behaves as expected and cuts off the direct relative to the angle of incidence (which I hope explains this funny test scene I’m using)
There is a major change in the ambient values once the rendering gets to the window (which is why the band is horizontal in the single process run and vertical in the multiple process run. I ran these using rad so there is an overture calculation. Anyone have an idea as to what is going on?
