2 Observations and data reduction

Our observations were carried out using the 90 cm Dutch telescope and its CCD camera at ESO La Silla, Chile, between the $28^{\rm th}$ of November and $3^{\rm rd}$ of December, 1991. The CCD chip installed in the camera had a very low quantum efficiency in U, which caused minutes long exposures in this filter. Its dimensions are $420\times 580$ pixels. Its linearity went up to 12000 ADUs remaining always better than 1%. These factors made it an acceptable tool for observing our relatively bright targets in the U filter as well.

We have selected a fixed window of $180 \times 180$ pixels on the CCD chip for all nights and we centered all double stars observed and the photometric standard stars in this window before starting the exposures. On the other hand, we used the whole surface of the CCD chip for the exposures aimed at its astrometric calibration (stellar traces).

All objects were observed in all three filters during each night. It may be remarked that observing such targets in different filters in the same night with CCDs was (and remains) a time consuming method, as the focal length of the telescope depends on the filter properties so that the instrument must be refocused whenever a different filter is used.

Two to four exposures per double star were taken in V. The duration of each exposure was as long as possible, at least two seconds and typically five to ten seconds. The same number of exposures per double star was taken in B. Their duration was at least ten seconds and typically around thirty seconds. Finally, six to ten exposures per double star were taken in U. The duration of each exposure was not longer than thirty seconds in order to avoid bad quality images due to telescope guiding errors.

The astrometry was performed in the V filter only. We chose to use some of the stars from the catalogues of Brosche & Sinachopoulos (1988, 1989) for the estimation of the scale of the CCD. For the calculation of the instrumental position angle of the CCD camera we used traces of equatorial bright stars. During the first night (November 28 to 29) photometric standard stars from the E regions (Graham 1982) were observed, in all filters, for the calculation of the transformation coefficients into the UBV system. Five exposures per filter were taken.

The atmospheric conditions under which the observations were carried out were very good. During all nights seeing at the 90 cm Dutch telescope was varying from 1.3 to 1.6 arcseconds, while the temperature was fluctuating from $\rm 10.2\; ^{\circ}C{-}16.3\;^{\circ}C$.

The data reduction was performed by using the ESO-MIDAS image processing software. Bias offset subtraction and a flat-field correction were included.

  

Figure 1: Accuracy of components magnitude difference determination

Two two-dimensional Moffat profiles were fitted simultaneously to the double star components on each CCD frame according to the classical least square technique, with our FORTRAN program.

The distribution of the accuracy of the components' magnitude difference that resulted from the photometry with this chip can be found in Fig. 1 for the three filters used. We obtain a standard deviation of $\sigma_{\Delta V} =$ 0.026 mags, $\sigma_{\Delta B} =$ 0.009 mags, and $\sigma_{\Delta U} =$ 0.014 mags. Accuracies on all filters are higher than 0.025 mags when the angular separation between the binary components is larger than 8 pixels (3 arcseconds), dropping often down to 0.07 mags for smaller ones.
On the other hand accuracy decreases significantly for component magnitude differences higher than two. Not surprisingly a combination of components' small angular separation plus high magnitude difference results in a low accuracy measurement.