[Radiance-general] What's causing these artifacts???

[Gregory J. Ward]

Hi Marcus,

Are you using rad at all for your renderings, or are you setting  
parameters by hand?  You wouldn't be performing an overture  
calculation with different -a* settings (in particular -ab and/or - 
av), would you?  The dark splotches look suspiciously like bad values  
in the ambient file at the start of the calculation.  Is that  
possible?  What are your rendering parameters?  It helps a lot in  
debugging a problem.  Your floor looks overly bright and even  
compared to the rest of the surfaces.  How is it defined?  Why aren't  
there any shadows under the chair legs?  What is your overall scene  
size (run oconv -d on the octree) and your -ar setting?

The bright splotches in your last example could be caused by the  
bright, narrow opening surrounding the skylight.  I would probably  
insert a rectangle flush with the ceiling and use mkillum to compute  
its effective output distribution to eliminate this particular artifact.

-Greg

> From: "Marcus Jacobs" <marcdevon at hotmail.com>
> Date: December 5, 2005 10:07:11 AM PST
>
> Dear Group
>
> In some of my recent renders, I have encountered some very specious  
> artifacts. I generally specify a high number of ambient division  
> (usually around 3072) and ambient super samples (500-1000) for my  
> renders.
>
> The first ones are these mysterious dark splotches:
>
> http://pg.photos.yahoo.com/ph/marcdevon/ 
> detail?.dir=b3e9&.dnm=4bcbre2.jpg&.src=ph
>
> http://pg.photos.yahoo.com/ph/marcdevon/ 
> detail?.dir=b3e9&.dnm=c76fre2.jpg&.src=ph
>
> I really can't figure out what's causing these. I am unsure if this  
> is caused by poor interpolation. Also, occasionally I have come  
> across the following bright spots:
>
> http://pg.photos.yahoo.com/ph/marcdevon/ 
> detail?.dir=b3e9&.dnm=e02ere2.jpg&.src=ph
>
> No matter how many ambient divisions that I specify (I went up to  
> 16,000 ambient divisions with 2000 ambient super samples),  I can't  
> get rid of these bright spots. The problem with these particular  
> artifacts is that I can not predict when it will occur. What I need  
> the most when performing renders is predictability. Does anyone  
> have any ideas?
>
> Thanks
>
> Marcus