How to average angles?

Rainer Gratias rg at joshua.org.chemie.tu-muenchen.de
Fri Feb 14 07:31:09 EST 1997


Hallo altogether.
I've got a little problem, which might sound a little bit stupid.
How do I correctly average angles, keeping the periodicity in mind?
The only (more or less) correct way I found is derived from averaging
complex numbers, using the transformation from polar to cartesian 
coordinates:

<ang>=arctan(<sin(ang_i)>/<cos(ang_i)>)
(in Fortran: atan2(<sin(ang_i)>,<cos(ang_i)>))

But, since it's derived from vector averaging, it has quite a strange
behavior for averaging:
0, 0, 180, 180      => -90
0, 0, 180, 180, 180 => 180
0, 0, 0, 180, 180   => 0

or in more detail (with n>=1):
n*0, n*180     => -90
n*0, (n+1)*180 => 180
(n+1)*0, n*180 => 0

Hence, if the total number of values is odd and the difference between 
the values is 180, nearly half of all values would be neglected. That's
my problem.

Has somebody a better soulution for this problem?
Any hints are welcome.

Bye,

Rainer Gratias.
 
********************************************************************************
				Technische Universitaet	Muenchen
				Institut fuer Organische Chemie und Biochemie
                                Lehrstuhl fuer Organische Chemie II
				Rainer Gratias
				Lichtenbergstr. 4
				D-85747 Garching
                                Tel :   ++49 +89 289-13324
                                FAX :   ++49 +89 289-13210
		  		e-mail: rg at artus.org.chemie.tu-muenchen.de
********************************************************************************



More information about the Molmodel mailing list