CCL: different MNDO parameter sets and partial charges

["Frank Hoffmann" <Frank_Hoffmann $#at#$ public.uni-hamburg.de>]

Dear CCLers,
 I'm new in the field of modelling and I want to know something more about the
 MNDO semiempirical method. Performing calculations of geometry optimization,
 partial charges and vibrational frequencies of a particular molecule with the
 the same method (MNDO) but with different programs results in different values
 for bond lenghts/angles, charges, and modes. Obviously, the different software
 packages use different MNDO parameter sets. I want to know, if they have names,
 what sets are usually used (is there a standard or reference set?), and what the
 "best" among them is; best means, of course, not a general  variable,
 but in my special case I would like to calculate accurate partial charges.
 Apropos: which other semiempirical or ab initio method do you prefer for the
 calculation of partial charges normally and why? So, this include implicit the
 question how accurate a non-observable value can be...
 Hope, these are not to many questions for one mail...
 Thank you very much in advance!
 Frank