[Radiance-general] Re: -ar parameters

Greg Ward radiance-general@radiance-online.org
Sun, 7 Sep 2003 21:43:37 -0700


Hi Rob,

Seems that no one answered this query, either...

The -ar parameter is a bit tricky, as it's proper setting requires 
knowing the ratio between the scene octree size (which you can get from 
the octree with getinfo -d) and the smallest important details, which 
only the scene author knows.  Divide the former by the latter, and this 
will give you a reasonable value for -ar.  For example, if "getinfo -d 
scene.oct" gives you:

	scene.oct: -123.8831 -.00315 -202.67102 431.0032

then the last number (431.0032) is your scene cube size.  Dividing this 
number by the thickness of one of your cushions, say 3.75, you get 
about 115, which is what you should set for -ar.

Setting -ar 0 sometimes helps but usually does not.  In your scene, I 
think increasing -ad would also help, and this is almost always a good 
idea when you are seeing artifacts in the indirect calculation.  
Increase -ad first, then -as, then try increasing -ar, finally decrease 
-aa.  This is the order I do things in, and you can usually find a good 
place for these settings with a little experimentation.  If you don't 
like waiting for each run to finish, try rendering a lower resolution 
image of just the problem area in your scene, like the seat cushions in 
yours.

-Greg

> From: "Fitzsimmons, Rob" <Rob.Fitzsimmons@Summit.Fiserv.com>
> Date: Wed Sep 3, 2003  10:55:43  AM US/Pacific
>
> I'm trying to understand the -ar parameters
> I'm thinking thats what behind the artifacts in between the couch 
> cushions
> in this image
> http://rastermon.com/Rad/image/july1_livdbig.jpg
> Unfortunately, I don't have the actual -ar value used with me now (I'm 
> at
> work) -- I used a rif with high quality, so I'm thinking it  was 
> around 93
>
> Acording to this page,
> http://radsite.lbl.gov/deskrad/param_table.htm
> -ar minimum value is 8, fast is 32 and accurate is 128, yet maximum is 
> 0. So
> if I want better quality than 128, can I go to 256, 512, etc or is 0 
> the
> value I should use?
>
> Or is that max 128 only in Deskrad?
>
> From the Radiance manpages, rpict -ar
> Error will start to increase on surfaces spaced closer than the
> scene size divided by the ambient resolution. The maximum ambient value
> density is the scene
> size times the ambient accuracy (see the -aa option below) divided by 
> the
> ambient resolution.
>
> This seems to imply you can set it higher than 128, and also decrease 
> the
> -aa
>
> thanks
>
> rob fitz