[Radiance-general] Rendering large space with small detail (small picture now attached!)

[Paul Chilton]

Hi Rob,

Thanks for your help.

I'll try to explain better what is happening in the image. As you look at
the image you are looking down the spine of a supermarket mall. Right in
front of the camera position is a void in the floor to the ground floor
below. On the immediate left (I assume this is the 'cove' you talk about) is
another space perpendicular to the mall which is a food court. To the right
and left are glazed shop fronts with the first shop front on the left
showing a reflection of the scene behind the camera position. The cubic
looking geometry are 1m high kiosks in the centre of the mall space. The
baffles hanging from the ceiling are internal shading devices which are
hiding a horizontal skylight above which stretches the length of the mall.
The left top strip of baffles are external shading devices covering the
vertical clerestory windows which again run the length of the mall.

The electric lighting you can see inside the shops fronts are very generic
and are simply there so that the shops don't look like dungeons for the
purpose of the image. The primary focus and motivation for the image is the
day light conditions in the mall space and so the mall space has no electric
lighting - just the shops.

The outside ambient conditions are a uniform sky.

You say;

'the easiest way to avoid splotchy artifacts is to bring the direct light
contributions closer and closer to the surfaces visible in the image'

I'm unsure exactly what you mean here as the direct light contributions are
the external sky and therefore how would you bring this closer to the
surfaces to create better resolution?

Thanks for your reccomendations on the variables, I'll try doubling the
variables you said were too low and seeing if that has a better effect.

Cheers,

Paul.





On Tue, May 19, 2009 at 3:03 PM, Rob Guglielmetti <rpg at rumblestrip.org>wrote:

>  Hi Paul,
>
> Bottom line, the easiest way to avoid splotchy artifacts is to bring the
> direct light contributions closer and closer to the surfaces visible in the
> image. I'm having a hard time understanding what the lighting design is in
> this image. There seems to be a cove or something on the left, and some
> baffled ceiling condition to the right of that, that changes in plan. maybe
> you could describe the lighting design and the section of that ceiling?
> Where is the skylight and what is the section of that in relation to this
> scene? What are the characteristics of the media between the external
> daylight and the interior scene? That information might inform a solution to
> these rendering issues (because there may be a solution to them that takes
> less than 20 hours to get results better than what you have posted). Either
> way, for a scene this apparent size, the -ar seems low right off the bat;
> also, the -ab is way high for an image (generally speaking); and the -ad and
> -as are probably too low.
>
>
>   On May 18, 2009, at 6:32 PM, Paul Chilton wrote:
>
>   Hello,
>
>
> I’m using Ecotect as the geometry interface and file generator to Radiance.
>
>
> For day-lighting purposes, I’m modelling a large space which includes areas
> of small detail. After many lengthy iterations (20+ hours for each
> simulation) the best image I could generate is the image attached.
>
>
> My problems are the splotchy nature of the light being reflected off the
> ceiling space and the baffled louvers under the skylight. The render
> settings that I used to generate the image are as follows;
>
>
>   *-dp=*
>  *-ar=*
>  *-ms=*
>  *-ds=*
>  *-dt=*
>  *-dc=*
>  *-dr=*
>  *-sj=*
>  *-st=*
>  *-ab=*
>  *-af=*
>  *-aa=*
>  *-ad=*
>  *-as=*
>  *-av=*
>  *-lr=*
>  *-lw=*
>  512
>  128
>  3.4
>  0.3
>  0.1
>  0.5
>  1
>  0.7
>  0.1
>  8
>  RCP.amb
>  0.1
>  1023
>  512
>  0.01 0.01 0.01
>  6
>  0
>
>
> I’ve increased the accuracy of the variables which I thought were critical
> to the accuracy of the render. These were; ambient resolution (ar), ambient
> bounces (ab), ambient accuracy (aa), ambient divisions (ad) and ambient
> super-samples (as).
>
>
> Is it a case of increasing the accuracy of these specific variables further
> or is there an easier or quicker way to achieve better rendering?
>
>
> Any help would be greatly appreciated.
>
>
> Thanks,
>
>
>  --
> Paul Chilton
>
> Renewable Energy Engineer
>
> [m] 0400 306 791 | [e] p.chilton at gmail.com
> <7.jpg>_______________________________________________
> Radiance-general mailing list
> Radiance-general at radiance-online.org
> http://www.radiance-online.org/mailman/listinfo/radiance-general
>
>
>
> _______________________________________________
> Radiance-general mailing list
> Radiance-general at radiance-online.org
> http://www.radiance-online.org/mailman/listinfo/radiance-general
>
>


-- 
Paul Chilton

Renewable Energy Engineer

[m] 0400 306 791 | [e] p.chilton at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://radiance-online.org/pipermail/radiance-general/attachments/20090519/886b504f/attachment-0001.html