> Alain <kua### [at] videotronca> wrote:

>>> Thomas de Groot <tho### [at] degrootorg> wrote:
>>>> On 12-4-2013 15:03, s.day wrote:
>>>>> Would be nice to see a radiosity version of this image as well (I assume the
>>>>> lighting is ambient currently?)
>>>>
>>>> Strange you say this. This *is* a radiosity version ;-)
>>>>
>>>> Currently:
>>>>
>>>>      radiosity {
>>>>        pretrace_start 0.08
>>>>        pretrace_end   0.004
>>>>        count 50, 500
>>>>        nearest_count 10, 5
>>>>        error_bound 0.6
>>>>        recursion_limit 2
>>>>        low_error_factor .3
>>>>        gray_threshold 0.0
>>>>        minimum_reuse 0.010
>>>>        maximum_reuse 0.1
>>>>        brightness 1
>>>>        adc_bailout 0.01/2
>>>>        normal off
>>>>        media off
>>>>        always_sample off
>>>>      }
>>>>
>>>> Thomas
>>>
>>> Oops.. sorry, I guess looking at the inside of the windows it shows radiosity
>>> effects, I think the shadows of the buildings on the road and to a certain
>>> extent the shadows on the walls in many areas look too consistent, I would
>>> expect them to darken a bit as they get into corners etc..
>>>
>>> I assume you have ambient_light set to zero in the scene? (looking at the
>>> darkness inside the windows would point to this being set I guess).
>>>
>>> Will look more closely before I comment again ;-)
>>>
>>> Sean
>>>
>>>
>>
>> Rendered with version 3.7, so, ambient get automativaly turned OFF as
>> soon as radiosity is activated.
>> You can tell it use version 3.7 when you see double values after count
>> and nearest_count, as well as the presance of maximum_reuse.
>>
>>
>> Alain
>
> If you have heard the saying ignorance is bliss, then I must be very happy ;-)
>
> I have not tried the new double values on 3.7 do they make much difference to
> performance? (always looking for ways to speed up a render).
>
> Sean
>
>
>

The second value for count set a larger direction pool for the radiosity 
sampling. It should not have any important effect on the speed, but can 
affect the quality. In some cases, it allow you to use a smaller count 
value, making the render faster. If used, it MUST be larger than the 
first value. Those who use it seems to tend to use a value at least 10 
times larger than the first one.

The double nearest_count can affect the quality and the speed. The first 
value set the maximum bound to be used in areas where the samples have 
more variation. The second value is the minimal value to be used in 
areas where everything is prety uniform, like a large, plain wall.

maximim_reuse set the largest distance any sample can be reused. It's 
grosly a ratio of the width ov the scene.

If you use a smaller pretrace_end value, you often can do with more 
relaxed settings elsewhere. The default of  0.04 is often much to large, 
I usualy use 0.01 or 0.005. low_error_factor also play a large role in 
reducing the sampling done during the final trace and a small value can 
greatly reduce artefacts on multi-core systems. Keep it at or under the 
default value of 0.5.


Alain

Post a reply to this message