4 The observations

ELODIE (Baranne et al. 1996) is an échelle spectrometer physically located in a coudé room at the 1.93 m telescope at Observatoire de Haute-Provence (OHP). For this programme, the spectrometer was fed via one optical fibre from the Cassegrain focus. The instrument FWHM is $\simeq$7.2 km s^-1, corresponding to a resolving power of $R \simeq 42~000$. The present observations were made 1997 in two separate runs on February 18-23 and October 15-23. The campaign targeted mainly stars in the Hyades and Ursa Major open clusters, but also included a set of IAU radial-velocity standards, some low-metallicity stars, Procyon and 51 Peg. Lunar spectra were also obtained for the purpose of calibrating the absolute wavelength scale.

The strategy during the observations was to have as good signal-to-noise ratio (S/N) as possible, in order to allow also weak lines to be used for the extraction of differential velocity information (Gullberg 1999). With the gain factor used, 2.65 e^- ADU^-1, the maximum S/N is about 300 before non-linearity and saturation effects occur in the most flux-rich orders.

The normal operation of ELODIE, when used as a radial-velocity machine, is to obtain spectra of modest $S/N \simeq 50$ in short exposures, with Th-Ar calibration spectra obtained simultaneously occupying the inter-order spaces of the CCD image. For the present programme it was considered important to avoid any possible light or charge leakage from the Th-Ar exposure; therefore separate Th-Ar exposures were made, leaving the inter-order space of the stellar spectra empty. Several such calibration exposures were obtained during each night, ideally between each stellar exposure.

Observations of the Moon were needed to derive an absolute wavelength scale and to correct for any long-term instability of the instrument, in particular between the February and October sessions. During the lunar observations the intended target was a well-defined crater or bright surface near the selenographic centre, although this turned out to be difficult to achieve in practice. Fortunately, even an offset by several arcmin would not cause any significant zero-point error (Sect. 5.4).

Figure 2: A series of diagrams illustrating the conditioning of an observed spectrum before it is correlated with the template. a) The raw one-dimensional spectrum versus pixel number, as extracted from the CCD image. Note the characteristic parabolic envelope of each spectral order. b) The wavelength scale has been set and the gross variation within each order removed using calibration observations of a tungsten lamp. c) The spectrum has been normalised through division by an estimate of the continuum intensity. d) The spectrum has been inverted, its average subtracted, and the windowing function applied to taper off the ends. This is what goes into the digital correlation. The Solar Flux Atlas, used as a template for the observations of the Moon, is treated similarly, going from c) to d).