Astrometric demands of fibre-input spectrographs for wide-field quasar redshift surveys

A potential problem with using wide-field multi-object spectrographs for quasar redshift surveys is the introduction of positiondependent selection effects into the survey catalogue because of variations in the observed signal-to-noise ratio (SNR) across the survey area. A function relating signal losses in instruments using optical fibre input to the fibre-to-image position error, the fibre diameter and the seeing conditions is given and plotted for a range of typical values of fibre sizes and seeing found in current instruments. 1. Why this is interesting The new era of wide-field redshift surveys, exemplified by 2df and SDSS (see Boyle et al. and Lupton et al., these proceedings), relies on multi-object spectrographs to observe thousands of candidates per night. Constructing homogeneous catalogues from such data requires automated analysis to achieve consistent and reliable spectral classifications and redshift measurements. The probability of correct results depends on the SNR achieved during observations, which is strongly affected by how much of each target image actually falls on a spectrograph aperture. Compared to single-object spectrographs, multi-object instruments thus introduce problems in reaching a consistent minimum SNR for faint objects across the survey. For fibre-input instruments observing point sources, the seeing, fibre diameter and fibre-to-image position error are the important parameters. Typical galaxy images are ∼> 4 h arcsec across, giving some tolerance to fibre positioning errors, but quasar images are usually seeing-limited at ∼ 0.5–2.0 arcsec, demanding greater precision. Fibre-to-image offsets will clearly reduce the recorded signal and may also vary systematically across the field, leading to position-dependent selection. 2. Signal reduction due to fibre position errors Most fibre spectrographs use blind positioning of individual fibres in (or close to) the telescope focal surface before observation, and employ a small number of coherent fibre bundles positioned on guide star images to centre the telescope on the desired field, adjust fibre mount plate rotation and track the field during exposure. Exposures are typically 45–90 min. Sources of fibre-to-image position errors in such systems include the fibre positioning robot or plate drilling machine, variations in differential atmospheric refraction and dispersion during 1