Since Hubble ( 1926 ), galaxy number counts have been widely used as a statistical tool for probing the distant universe, with the hope of constraining both its geometry and the evolution of its content. Modern, high efficiency instruments and detectors reach impressive magnitude limits in the optical: (Tyson 1988 , Metcalfe et al. 1995a , Smail et al. 1995 ), although on a limited area. Galaxy counts done at brighter magnitudes ( ) are equally useful; their interpretation is much less model-dependant because of the smaller lookback-times and weaker cosmological effects. They thus constitute the link between models and deeper counts by providing a normalisation of both space densities and colours at low redshift.
The problem with bright galaxy counts is that they obviously require substantial solid angles to be surveyed in order to provide statistically significant samples. Until very recently, such areas could only be surveyed in a reasonable time using photographic Schmidt plates. Since the mid-eighties, fast microdensitometers coupled with image analysis computer programs have been employed to produce automatically highly complete catalogs from large areas of the high galactic latitude sky like the COSMOS (Heydon-Dumbleton et al. 1989 ), the MRSP (Seitter 1988 ), or the APM (Maddox et al. 1990b ) galaxy surveys.
The main difficulty when dealing with photographic material concerns flux measurements. Many scientific issues typically require the systematic errors to be kept mag.. This is quite a difficult task to achieve on large scales, and can be reached only if a large number of galaxy standards per Schmidt plate, over the whole magnitude range of the counts, are observed to calibrate the data (Metcalfe et al. 1995b ). Having such a high density of photometric standards is practically not possible with very large Schmidt plate surveys, because of the huge observing time it would require. The catalogs extracted from these surveys are therefore likely to be hampered by systematic errors in their photometry, rather than by limited statistics.
In this paper we re-examine galaxy number counts in the blue and red photographic passbands over the magnitude range and . Our surveyed area is modest (140 sq. deg.), but has been carefully calibrated using a fair density of CCD standards ( , half of them being galaxies), as an attempt to keep photometric systematic errors mag. over the whole magnitude domain of the survey. The procedure is described in detail in § 3 , as well as the general data processing. The galaxy number counts are derived and compared to previous studies in § 4 . Model predictions about number counts and galaxy colour distributions are tested in § 5 . We finally discuss the implications of our results in § 6 .
© European Southern Observatory (ESO) 1997