8 Conclusion

 The treatment described here has produced 2.1 10⁶ super-pixel light curves corrected for observational variations, with an error bar for each point. They are characterised by an average stability close to twice the photon noise: dispersions of 1.8% of the flux in blue and 1.3% in red are measured over a 120 days time period. To reduce the effects of the dispersion due to the observational conditions, averaging the images of each night turns out to be a crucial step. The fluctuations due to seeing variations have been corrected for. We associate an error bar with each measurement, and these careful estimates together with the study of possible systematics are used in the companion papers for the detection of intrinsic luminosity variations.

This study is the starting point for the comprehensive microlensing search described in Paper II. The error estimates enter the definition of the selection criteria and constitute an important ingredient for microlensing Monte-Carlo simulations required to quantify the efficiency of the pixel microlensing method. The study of the background of variable stars will be addressed in Paper III.

Acknowledgements

We thank A. Gould for extremely useful discussions and suggestions, and D. Valls-Gabaud for a careful reading of the manuscript. We are also particularly grateful to Claude Lamy for her useful help on data handling during this work. P. Gondolo was partially supported by the European Community (EC contract No. CHRX-CT93-0120). A.L. Melchior has been supported by grants from the Singer-Polignac Foundation, the British Council, the DOE and by NASA grant NAG5-2788 at Fermilab.