For the majority of the link fields only two plate pairs, i.e. two first epoch plates and two second epoch plates were measured. Plate pairs were also used in the fields of M31 (4 pairs), the dSph fields UMi and Draco (3 pairs, for more details, see Scholz & Irwin 1994) and M92 (5 pairs, see Scholz et al. 1994). For details on the reduction within the MEGA programme, see Schilbach & Scholz (1991) and Kharchenko et al. (1994).
The proper motion determination in all these fields was based on independent plate-to-plate solutions using polynomials of up to fourth order and the method of stepwise regression of Hirte et al. (1990). The order of the polynomial was chosen according to the number of reference objects (field stars or galaxies) used. Generally, the reduction was done with large numbers of reference galaxies selected from the full plate scans by APM, MAMA and PDS and with third order polynomials. In cases of relatively small numbers of reference galaxies we used field stars in the differential plate-to-plate solutions. The absolute proper motions were then obtained by subtracting the mean "proper motion'' of the galaxies per field. Plate pairs measured with ASCORECORD and PARSEK were reduced using the galaxies directly as reference points.
The reduction of the globular cluster fields M5, M12 and M15 is described in Scholz et al. (1996). Here we preferred to use an iterative plate overlap method (Eichhorn 1960) with a catalogue of positions and zero proper motions of the selected reference galaxies as starting point. The plate overlap method was also applied by Tucholke (1995) for the field 3C345. His results based on 3 first and 8 second epoch Tautenburg plates are also included in the present paper.
A new reduction of the field around the globular cluster M3 (Scholz
et al. 1996) was carried out with the aim of a QSO survey
based on variability and zero proper motions. Here we have used the proper
motions of the Hipparcos stars obtained from 53 blue Tautenburg plates. For
this special case of a large number of plates taken over a time baseline of 30
years, we built up a time series of coordinates for each object. The measuring
coordinate frame of each plate was transformed to that of the master
plate by using a quadratic plate constant polynomial relationship.
On each comparison plate the transformed coordinates were shifted
by the negative mean coordinate differences of all available galaxies.
The absolute proper motions of all objects were then determined
from the linear regression of their coordinates over the time.
The absolute stellar proper motions in the 26 Hipparcos link fields were determined with respect to 30 up to 2000 reference galaxies per field (only 7 fields with less than 100 galaxies, see Table 1 (click here)). With epoch differences of about 20 to 40 years an internal accuracy of 3 to 5 mas/yr was achieved for undisturbed (by overlapping images) and relatively faint (B > 9) Hipparcos stars. Serious problems in the internal proper motion accuracy of bright stars appeared in some of those cases, where we combined plates from different Schmidt telescopes. Especially in the fields of the dSph's Draco and UMi, each with two first epoch POSS1 plates, two second epoch Palomar plates with different plate centre and three third epoch Tautenburg plates, we failed to obtain accurate proper motions of all bright stars (B < 12). Therefore, these two fields were excluded from the extragalactic link of the Hipparcos proper motions. In the field of M15 similar problems arised so that we re-reduced the proper motions of the Hipparcos stars in this field without the use of the POSS1 plates.
From 706 Hipparcos stars in all 26 fields H37cr data of 651 (92%) stars were provided for the link. From all measured H37cr stars we excluded those stars which showed large errors in their proper motions because of bad measurements or nonstellar image classification. 360 stars remained for the determination of the link.
Considering the final 24 Hipparcos link fields, the following "success rates'' were obtained for the measuring machines used: 34% of all available Hipparcos stars in 8 fields reliably measured with the APM, 57% in 7 fields measured with MAMA, 65% in 4 fields measured with the PDS, 67% in 3 fields measured with PARSEK and 77% in 2 fields measured with the ASCORECORD. The smaller "success rate'' of the APM in measuring accurate proper motions of Hipparcos stars confirms our results from the comparison of APM and MAMA (Schilbach & Scholz 1992).
Besides of the random errors arising from the measuring accuracy, all proper motions in a given field are affected by a systematic error which depends on how accurate the extragalactic reference frame was defined by the measurements of the galaxies in that particular field. Due to the large number of galaxies, the formal zero point error in a given field is usually less than 1 mas/yr. However, in a few fields with globular and open clusters systematic magnitude dependent errors in the proper motions were detected and removed (Scholz & Kharchenko 1994; Kharchenko & Schilbach 1995). In some of the link fields this kind of error was also found and corrected for. But it can not always be detected so that on the average, residual systematic errors of about 2 mas/yr per field may be expected.