POV-Ray : Newsgroups : povray.binaries.images : D200 Photons : Re: D200 Photons Server Time
31 Jul 2024 12:26:07 EDT (-0400)
  Re: D200 Photons  
From: Alain
Date: 1 Nov 2009 17:26:05
Message: <4aee0afd$1@news.povray.org>

> Alain <aze### [at] qwertyorg> wrote:
>> You obviously have poor bounding. The bounding boxes for the sphere, all
>> 200 facets and all 200 numbers are all pilled up and inflated to that of
>> the sphere. Each ray must be tested against each and every litle component.
>> Try without the numbers, or use bump_maps for them. It should be much
>> faster.
>> Considering the size of the numbers, faking it should be good enough.
>>
>> The text object is relatively slow. Diferencing it from some object is
>> slower. Multiply that slowdown by 200, add photons, transparency,
>> reflection and dispersion, it's a perfect recipie for insane rendering
>> times.
> 
> Okay, I asked for feedback - I appreciate your taking the time to respond
> Alain. I read the reference material. I was aware that text objects are
> slow to render. The numbers are integral to my concept of the object,
> even though they don't add much to this picture, which I saw as an
> opportunity to assess how it might appear in that setting. In a simple
> and more 'realistic' setting it might appear like this:
> http://i34.tinypic.com/2ldvwqv.jpg
> Of course, this is just one frame from an animation, but you see...
I'd use an image pattern made after your numbers, and a bump map derived 
from the image pattern.
> 
> I'm happy to have insane rendering times if it gives me (or gets me
> closer to) what I want.
> 
> As for the poor bounding, the object (a sphere with bits removed) is
> bounded by a sphere of exactly the same size and location. Please
> enlighten me as to how it could be better bounded?
> 

To get beter bounding, you need to chop the sphere into many smaller 
parts and reassemble the sphere from those parts. A single part may 
consist of only a single facet, it's number and a bit of the surrounding 
spherical shape.
This will work as long as the object remains opaque. If the object is 
transparent, you'll see all the extra faces unless you use a merge, and 
then, the bounding problem returns.

Another way, would be to create a mesh equivalent of the faces and the 
numbers relief, make sure that it have an inside vector defined, and 
difference that mesh from the sphere.
If you go the mesh way, you may as well model the whole thing as a 
single mesh.


Alain


Post a reply to this message

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.