New observations for the complete Bonn link
were mainly taken at the double
refractor (
, 
) from 1978 to 1994.
103 a-O, IIa-O and IIIa-J emulsions have been used with a BG3 or
BG25 filter. Part of these plates has been
used by others (see Tucholke et al. 1997 and references
therein) for additional link fields. In addition
to the observations of our own refractor, we have used also plates of
the 2 m RC telescope at Rozhen observatory and the
1.23 m telescope at Calar Alto, mainly for
those fields, for which the objects were too faint
for our double refractor. Data of the plates and telescopes for
the complete observations may be found
in Geffert et al. (1989),
Brosche et al. (1991a) and
Dick et al. (1993).
Measurements were performed either by scanning the complete plate at the PDS2020GM1.15explus in Münster or by measuring about 100 stars equally distributed over the plate on the ASCORECORD of the Bonner Sternwarte. In the latter case each plate was measured in two opposite directions in order to avoid systematic errors of the measurements due to drift of the machine or personal magnitude equation.
For the reduction, to achieve a well defined local coordinate grid, a complete set of FORTRAN routines was established. The method of reduction includes variable plate constant models with polynomials of the rectangular coordinates x and y, colours and magnitudes. However the colour of our objects was generally unknown and not included in the reduction. An iterative central plate overlap algorithm was used. A more detailed description of the reduction procedure is given in LeCampion et al. (1992). The complete results of the Bonn part of the Hipparcos link (Tucholke et al. 1997) were based on these routines or modifications of them. In the beginning, the reduction was performed with the PPM catalogue. Later a study of the proper motions of the Pleiades cluster (Geffert et al. 1995) had shown that even using the central plate overlap algorithm the results depend in a strong way on the reference stars used at the beginning of the iteration. Therefore we used the Hipparcos catalogue for the 37 month solution of Hipparcos (H37) in fields with enough reference stars for each plate.
The use of this catalogue was the most promising method to avoid systematic errors in our proper motions by imperfect reference stars. Nevertheless we wish to point out, that the final proper motions determined by an iterative plate reduction procedure are mainly based on our plate material. Moreover, for our final proper motions in Table 2, only the link procedure described in Sect. 4 determines for each field separately the system of the proper motions, which is completely independent of the H37.
However, the number of reference stars in our reduction varies from only three to nine. For the field with nine Hipparcos stars (NGC 4147), we performed different independent reductions using only part of the available reference stars and compared the different results of our reduction. The comparison demonstrated the stability of the reduction using the H37 reference stars. The mean differences of the proper motions of the final solutions are below 1 mas/a. This result is of great importance for all future work on the Bonn plates, because these fields normally contain only 4 to 6 Hipparcos stars.
Table 2: 48 stars with absolute proper motion for the
calibration of the Hipparcos
proper motion system. The epoch
of the observations is 1950. The internal
accuracy of the proper motions is of the order
of 2 mas/a
