POV-Ray : Newsgroups : povray.text.tutorials : Radiosity Voodoo - Volume 3: Odds And Ends Server Time: 9 Dec 2018 21:35:28 GMT
  Radiosity Voodoo - Volume 3: Odds And Ends (Message 1 to 1 of 1)  
From: clipka
Subject: Radiosity Voodoo - Volume 3: Odds And Ends
Date: 31 Dec 2008 16:30:01
Message: <web.495b9db71d584ee0483cfa400@news.povray.org>
3.A Troubleshooting
    -------------------


smeared shadows:

horizontally smeared shadows are the results of samples being taken during the
final pass. This typically indicates that your pretrace is too coarse. Make sure
you have set "always_sample" to "off", and check your pretrace_end. You will
probably want to choose a smaller value (i.e. higher resolution), or increase
minimum_reuse.

Alternatively, you can render the scene twice, using the save_data keyword on
the first and load_data on the second run. Make sure to use exactly the same
settings, or you may end up with yet more samples to be taken during the second
run.


smooth meshes & triangles:

Radiosity is problematic with meshes and smooth triangles, as these objects'
effective normal differs from their true geometric normal.

Radiosity is designed to cope with effective normals differing from "raw"
geometric normals (e.g. due to bump mapping); however, the true geometric normal
is never reported by the mesh or triangle object in the first place, so
Radiosity never finds out about the discrepancy.

The problem results in radiosity samples occasionally picking up light from the
inside of a mesh, which may in turn be illuminated through radiosity samples on
the inside surface which happen to be picking up light from the outside.

As a workaround, use a recursion_limit of no more than 2. This should prevent
any light from leaking into closed meshes and back out again. Alternatively,
insert a slightly "slimmed down" copy of the mesh with a pitch black surface
texture.

To truly solve this issue, geometric shapes need to be extended by an interface
to report both the effective and true geometric normal of intersection points.


black splotches:

There is a bug in 3.6 and all prior versions (and a related flaw in 3.7.0 betas
up to .29) regarding the handling of max_trace_level during radiosity sampling,
almost inevitably causing stray black splotches, typically flocking near
reflective or refractive objects. The only reliable way to get rid of these is
to increase your scene's max_trace_level.


splotches in general:

Some people report that reducing both "count" and "error_bound" will help with
some very stubborn scenes.



    3.B Undocumented Details And Issues
    -----------------------------------


adaptive quality parameters:

Various quality settings are automatically reduced when taking "deeper"
radiosity samples. This includes the number of rays shot.

low_error_factor:

Contrary to documentation, this factor is effective during all pretrace steps,
not just the last one, and it affects error_bound, minimum_reuse, and a hard-
coded kind of "maximum_reuse" value.

Black Splotches:

There is a bug in 3.6 and all prior versions (and a related flaw in 3.7.0 betas
up to .29) regarding the handling of max_trace_level during radiosity sampling,
almost inevitably causing stray black splotches, typically flocking near
reflective or refractive objects. The only reliable way to get rid of these is
to increase your scene's max_trace_level.

Save/Load Limitation:

When saving previously gathered radiosity data, POV-Ray 3.6 only saves the "top-
level" samples. If the samples are not sufficient for the scene they are loaded
into, this will require a lot of "deeper" samples to be taken all over again.


---(End of Tutorial)---


Post a reply to this message

Copyright 2003-2008 Persistence of Vision Raytracer Pty. Ltd.