2. Observations and reductions
The observations were carried out partly with the Danish 1.54 m. telescope and DFOSC (Danish Faint Object Spectrograph and Camera) at the European Southern Observatory (ESO) at La Silla, Chile, and partly with the 2.56 m. Nordic Optical Telescope (NOT) and ALFOSC (a DFOSC twin instrument), situated at La Palma, Canary Islands. The data consists of CCD images in the filters U,B,V,R,I and H. In the filters BVRI and H we typically made 3 exposures of 5 minutes each, and 3 exposures of 20 minutes each in the U band. Both the ALFOSC and DFOSC were equipped with thinned, backside-illuminated 2 K2 Loral-Lesser CCDs. The pixel scale in the ALFOSC is 0.189" /pixel and the scale in the DFOSC is 0.40" /pixel, and the fields covered by these two instruments are and , respectively. All observations used in this paper were conducted under photometric conditions, with typical seeing values (measured on the CCD images) being 1.5" and 0.8" for the La Silla and La Palma data, respectively.
During each observing run, photometric standard stars in the Landolt (1992) fields were observed for calibration of the photometry. Some of the Landolt fields were observed several times during the night at different airmass in order to measure the atmospheric extinction coefficients. For the flatfielding we used skyflats exposed to about half the dynamic range of the CCD, and in general each flatfield used in the reductions was constructed as an average of about 5 individual integrations.
After bias subtraction and flatfielding, the three exposures in each filter were combined to a single image, and star clusters were identified using the daofind task in DAOPHOT (Stetson 1987) on a background-subtracted V-band frame. Aperture photometry was then carried out with the DAOPHOT phot task, using a small aperture radius (4 pixels for colours and 8 pixels for the V-band magnitudes) in order to minimise errors arising from the greatly varying background. Aperture corrections from the standard star photometry (aperture radius = 20 pixels) to the science data were derived from a few isolated, bright stars in each frame. Because the star clusters are not true point sources, no PSF photometry was attempted. A more detailed description of the data reduction procedure will be given in Larsen (1999).
The photometry was corrected for Galactic foreground extinction using the values given in the Third Reference Catalogue of Bright Galaxies (de Vaucouleurs et al. (1991), hereafter RC3).
2.1. Photometric errors
The largest formal photometric errors as estimated by phot are those in the U band, amounting to around 0.05 mag. for the faintest clusters. However, these error estimates are based on pure photon statistics and are not very realistic in a case like ours. Other contributions to the errors come from the standard transformation procedure, from a varying background, and from the fact that the clusters are not perfect point sources so that the aperture corrections become uncertain.
The r.m.s. residuals of the standard transformations were between 0.01-0.03 mags. in V, B-V and V-I, and between 0.04 and 0.06 mags. in U-B.
The errors in aperture corrections arising from the finite cluster sizes were estimated by carrying out photometry on artificially generated clusters with effective radii in the range pixels (0" - 1.6" on the DFOSC frames). 1" corresponds to a linear distance of about 20 pc at the distance of typical galaxies in our sample, such as NGC 1313 and NGC 5236. The artificial clusters were modeled by convolving the point-spread function (PSF) with MOFFAT15 profiles.
The upper panel in Fig. 1 shows the errors in the aperture corrections for V-band photometry through aperture radii pixels and pixels as a function of , while the lower panel shows the errors in the colour indices for pixels. At , the error in V-band magnitudes using pixels amounts to about 0.15 magnitudes. For a given , the errors in the colours are much smaller than the errors in the individual bandpasses, so that accurate colours can be derived through the small pixels aperture without problems. This convenient fact has also been demonstrated by e.g. Holtzman et al. (1996).
The random errors, primarily arising due to background fluctuations, should in principle be evaluated individually for each cluster, since they depend on the local environment of the cluster. Fig. 2 shows the random errors for clusters in NGC 5236, estimated by adding artificial objects of similar brightness and colour near each cluster and remeasuring them using the same photometric procedure as for the cluster photometry. Again it is found that the errors in two different filters tend to cancel out when colour indices are formed. The V-band errors are quite substantial, but we have chosen to accept the large random errors associated with the use of an pixels aperture in order to keep the effect of systematic errors at a low level.
© European Southern Observatory (ESO) 1999
Online publication: April 12, 1999