5 Astrometric reduction

The rectangular sky area covered by the EROS-1 chips is enclosed in a circle of radius $\sim$30 $\hbox{$^\prime$ }$ centred on the position $\alpha =$05$^{\rm h}$21$^{\rm m}$52.1$^{\rm s}$; $\delta=-69\hbox{$^\circ$ }34\hbox{$^\prime$ }16\hbox{$^{\prime\prime}$ }$(2000.0). In this region, the density of PMM astrometric standards is highly irregular, e.g. four $\sim$20 $\hbox{$^\prime$ }\ \times$ 15 $\hbox{$^\prime$ }$ areas are completely empty. This is mainly due to the crowding of the survey plates used at USNO. We therefore chose to use a U plate taken at the ESO 1 m Schmidt telescope to define the secondary astrometric standards. This plate was scanned with the MAMA microdensitometer, and reduced to the International Celestial Reference System (ICRS) with the ACT catalogue. Thanks to the modest sensitivity of the U emulsion/filter combination, the effects of crowding are reduced. A remarkably regular distribution of stars is detected with the SExtractor software (Bertin $\&$ Arnouts 1996), with on average 2700 stars per chip.

The secondary standards ($\sim 1500$/chip) are identified to EROS stars using a first linear transformation based on some 15 visually cross-identified stars per chip. The astrometric reduction software (Robichon et al. 1995) is then run using this set of references. After iteration, the final 2$^{\rm nd}$ order reduction keeps 1000 stars per chip with final rms deviation of the order of 0.2 $\hbox{$^{\prime\prime}$ }$.

According to the galactic model from Besançon, more than 95$\%$ of the secondary references belong to LMC. Furthermore, we check that the global proper motion of the LMC (Kroupa $\&$ Bastian 1997), together with upper limits on its internal velocity dispersion cannot affect the positions of our standards by more than 30 mas, over the 848-days separating the ESO plate from the EROS reference.