I discovered that the reason is that the anti-aliasing parameters are not
being read the same way in 3.1 and 3.5.11. Maybe it's a bug and maybe it's
intentional. I'll describe it.

For some special purposes, I had added entries in the quickres.ini file. The
first entry below is a standard and the second is what I set up.

[640x480, AA 0.3]
Width=640
Height=480
Antialias=On
Antialias_Threshold=0.3

[768x494, AA 0.3]
Width=768
Height=494
Antialias=On
Sampling_Method=2
Antialias_Depth=4
Antialias_Threshold=0.3

This "bug" was detected at 640x480, but was solved at 768x494. So I looked
deeper.

I have a povray.ini file in the same directory as the fiducial.pov source
file. It also specifies the additional anti-aliasing parameters. 3.1 reads
that file, but 3.5.11 does not. I renamed the file to fiducial.ini; neither
version read it. (This would be what I would prefer to work in both cases,
since I can have more than one .pov in a directory and not force them the
share the same povray.ini file. I don't like having to render the .ini
file.) I added the fiducial.pov name into the .ini file and rendered the
.ini file. 3.5.11 worked and 3.1 complained. I added Input_File_Name =
fiducial.pov into the .ini file, and rendered the .ini file and both
versions work. Of course the bad thing is I have to keep the .ini file open
in the editor and remember to switch from the .pov file to the .ini file
before pressing RUN. That's why I'd prefer that when rendering name.pov, the
renderer looked for name.ini to find any additional parameters.

At any rate, that's what happens. Is this a bug or it is an intentional "new
feature"?

Bill



"Anders K." <and### [at] prostard2gcom> wrote in message
news:3c759cc2$1@news.povray.org...
> > That's odd, because the object rendered in 3.1 with AA is placed exactly
> at
> > 0, 0 and the camera is looking at 0, 0 and the object is rendered quite
> > symetrically. Notice the cross at the center of the circle; the pixel
> colors
> > representing where the edge lies seem equally bright both left and
right,
> > and top and bottom.
>
> This is because the object is exactly centered at (320.5, 239.5), which is
> the center of pixel (320, 239). So it will look symmetrical, but it's
> centered around the wrong place.
>
> In fact, the 3.5 beta 11 image is also centered at (320.5, 239.5). Did you
> set #version to 3.1? That would disable the half-pixel bug fix.
>
> Anders
>
>
>

Post a reply to this message