Christopher James Huff wrote:
> In article <422db829$1@news.povray.org>,
>  Sascha Ledinsky <sas### [at] userssourceforgenet> wrote:
> 
> 
>>>  But the values in the matrix may not be accurate enough.
>>
>>That's double precision (64bit)... How accurate does POV need them?
> 
> 
> No, that's decimal ASCII text. Those are *not* the exact values your 
> program computed.

Yes, sure, I know that representing (binary) floats as decimal is always 
problematic, but I think that's about as accurate as possible...
I don't think that POV likes hexadecimal IEEE754 representations?

> Rounding error can easily accumulate and result in a 
> non-orthogonal final matrix.

I've used the vecmath package from Java3D - All I do is creating a 4x4 
matrix from a unit-quaternion... From looking at the images it renders 
(with -UV) everything looks right.
Could it be that the test for perpendicular vectors in POV-Ray is a bit 
too pedantic?

>>Do you know how to convert orientations represented as quaternions or 
>>axis-angle pairs into fixed-angle representations (for use in POV's 
>>rotate <x,y,z> command)?
> 
> 
> <http://www.mathworks.com/access/helpdesk/help/toolbox/aeroblks/quaternio
> nstoeulerangles.html>

Thanks a lot for that link, I'll try that...

-sascha

Post a reply to this message