> High!
>
> I wonder why the shadowed parts of the image are so strangely striated -
> although I use "save_file" and "load_file" within the radiosity statement!
>
> Here are my radiosity settings:
>
> radiosity
> {
> brightness 1.5
Should probably be left at 1, especialy with recursion_limit 5.
> count 300
Mostly exterior, but with important interior parts that requires a 
higher count.
Using version 3.7, you can use a larger value and use importance to take 
less samples where it don't mather, like the exterior ground and sky.
Proposition:
count 3000
interior parts and floor near the entrance: importance 1
EVERYTHING else: importance 1/30 (effective resulting count of 100)
> error_bound 0.05
Using something larger MAY diminish some artefacts. Try 0.07 to 0.09 or 0.1.
> gray_threshold 0.8
Personal taste, no problem.
> low_error_factor 0.2
May need a little smaler if you increase error_bound.
> minimum_reuse 0.015
May need to be smaller. Probably close to the value for pretrace_end.
> nearest_count 10
Maybe something like 20, 5
> recursion_limit 5
Very high, but may be needed for the interior part...
> adc_bailout 0.01
May need something smaller. Start with 0.005 (default value)
> max_sample 0.9 // -1
Usualy not used. Stay at 1.
> always_sample 1
Set to 0 or no. Default of no using version 3.7.
> pretrace_start 0.08
> pretrace_end 0.01
Normaly, I use something like 0.05 or 0.025
> load_file "whatmough_residential.rad"
> save_file "whatmough_residential.rad"
With version 3.7, it goes on the command line, not here.
> }
>
> What went wrong?
>
> See you in Khyberspace - www.khyberspace.de
>
> Yadgar

Using version 3.6.1?
Version 3.6 don't keep radiosity data when you interupt and resume a 
render, and even with always_sample no, you'll get new samples taken 
during the render phase unless you sue insane pretrace_end during the 
first pass.
The result is banding in any interupted/resumed radiosity render.

Using version 3.6, use pretrace_end 0.00125 or 000625 during the first 
pass. Using something like pretrace_end 2/image_width will give you 2x2 
tiles during the pretrace step.
It don't take any data from areas outside the current render. If you 
resume a render, there are missing radiosity data that where generated 
and used during the interupted render.

Cure: Use version 3.7. It will effectively save radiosity data in the 
*.povstate file. This will prevent many glitches.


Alain

Post a reply to this message