Ard nous apporta ses lumieres en ce 2005-12-06 17:59:
> Hi folks.  The floor of the gap between these two boxes is in total
> darkness, so I'm using radiosity to bring it out.  The default
> radiosity settings produce the inset image, with obvious horizontal
> artifacts.
> 
> After a few renders I ended up with
> 
>  count 80 nearest_count 8 error_bound 0.1
>  pretrace_end 0.02 adc_bailout 0.01
> 
> ....which rendered the attached image in 6h45m.  It still isn't good:
> it looks like my graphic equaliser is leaking coal dust.  And it takes
> five times as long to render as radiosity{}.
> 
> I've had about enough of the tweak/render loop, so I'm turning to you
> for advice.
> 
> Also, the finer settings introduced the black splotch arrowed.  Ummm, ?
> 
> Actually, I'm not too cut up about the final result - these models are
> part of a much larger project and will not be viewed too closely, and
> certainly not on a green and yellow chequered plane.  That's why I'm
> not fixing the image maps or the cable texture.  However I would like
> to understand radiosity a little better, so I'm looking forward to
> your suggestions.
> 
> 
> Many thanks in advance.
> 
> 
> 
> ------------------------------------------------------------------------
> 
Increasing error_bound usualy lead to smoother results. You can try something around
0.4 to 0.8. Low 
values also take more time and memory that larger ones.
When using low error_bound value, you frequently need to proportionaly increase the
count, you may 
need a value higher than 800.
Lowering low_error_factor result in more samples at the end of the pretrace, it can
improve your 
results.
Lowering pretrace_end can, in some instances, provide higher quality.

-- 
Alain
-------------------------------------------------
Islam #3: If shit happens, blame Israel.

Post a reply to this message