- 3.1 The used method of PSF fitting the data
- 3.2 CCD scale determination
- 3.3 Position angle determination

More than 350 CCD frames were taken during the four nights. The data reduction was performed by using the ESO-MIDAS image processing software. It evidently included a BIAS offset subtraction and a flat-field correction has also been performed for "cleaning'' the observations.

Since the angular separation of the components is of a few arcseconds only, their images have identical shapes. So, the two fitted MOFFAT functions should have the following profile determining parameters in common: b -that is related to the FWHM-, e and f: parameters of the ellipticity, and q: the exponent of the MOFFAT function (see Cuypers 1997).

Figure 3 shows the strong correlation between seeing and the exponent of the MOFFAT function "q''. The symbols used in Fig. 3 have the same meaning as the ones of Fig. 1, corresponding to the four different nights. The corresponding slope of the linear regression was for the second night, very significantly different from that of the third and the fourth .

One may remark that the uncertainies of angular separations due to the instrumental on CCD chip are around 0.02 pixels, equivalent to 0.006 arcseconds. Since our astrometric standard double stars have an angular separation of around 50 pixels, we expected an internal scale accuracy (due to the used CCD pixel size and the telescope focal length) around 0.00006 arcseconds/pixel. It is then clear that the resulted uncertainty of the scale (0.00056 arcseconds/pixel) is dominated by the HIPPARCOS catalogue uncertainties. This is mainly due to the rapid deterioration of the proper motion accuracies in this catalogue with time, which is the reason, in our case, for the relatively high uncertainty of the scale.

We calculated the same using data from the Brosche & Sinachopoulos (1988, 1989) catalogues and we found a scale arcseconds/pixel. This value is four times less accurate than the value of CCD scale based on the HIPPARCOS results and thus not used.

The precision achieved in both the angular separation and the position angle has been often discussed in our previous papers of this series, especially by Sinachopoulos & Seggewiss (1990, Paper II), and by Nakos et al. (1995, Paper IV, and 1997).

It is nevertheless necessary to mention that we took
as an approximation of the mean error due to
refraction for the used *V* filter. According to Herzsprung (1920),
this value of is the
expected difference in position
due to refraction between an A-type star and a K-type star
at zenith distance
for observations with yellow sensitive plates,
which correspond to the *V* filter we used
(see also in Sinachopoulos &
Seggewiss 1990, for more details). Since we always observed at
smaller zenith distances and the colour difference between the two
components of our targets is probably not often that large, we
consider the value of we adopted to be always larger than
the correct value.

Copyright The European Southern Observatory (ESO)