On 5/29/19 9:22 AM, William F Pokorny wrote:
> On 5/29/19 2:31 AM, Paolo Gibellini wrote:
>> Thomas de Groot wrote on 28/05/2019 08:35:
> ...
> 
> My maze play turned up two issues. One marked with the '?s'. It looks 
> like the isosurface is getting pulled apart (shadows? normals?). No luck 
> thus far running the cause down. Only that one place and with smallish 
> changes it goes away.
> 
...

OK. In the hopes of helping by providing a later reference post.

The '?s' artifacts in the posted image can be significantly mitigated by 
increasing isosurface accuracy.

I usually set my isosurface accuracy a step beyond 
1/max(image_width,image_height) the resolutions I typically expect to 
render. I used 'accuracy 0.0005'. Increasing it to 0.00001 mostly solves 
the problem, but things are, for me, still a little grainy.

Grainy for me because I'm running with a shadow tolerance of 1e-5 
instead of the 1e-3 long used in POV-Ray to avoid light bleed through 
during shadow testing. Why not just increase accuracy to say 1e-6 you 
ask! This indeed solves my shadow graininess, but that accuracy also 
reveals a subtle ripple in the underlying isosurface function. Yep, in a 
box.

Starting to believe the shadow tolerance in trace.cpp should be a user 
controllable setting!

The last bit is the df3 being used is not deep in z relative to x,y to 
save space. The black hole warp is rolling that lessor z resolution into 
view at the problem location.

Bill P.

Post a reply to this message