The science fibres and fiducial bundles are terminated at their input ends with SF5 microprisms whose input faces are wide-band coated. Their 92-degree angle provides median compensation for non-telecentricity, which amounts to four degrees at the edge of the field. The residual maximum tilt of two degrees is still enough to produce a field-dependent degradation of focal ratio from f/2.8 to f/2.3 in addition to the intrinsic focal-ratio degradation (FRD) of the fibre. This is one reason why an FRD-insensitive design was chosen for the spectrograph.
Each fibre is equipped with a magnetic button which holds it onto the (flat) field plate. The circular plate and its entourage of radially-disposed fibres are supported in a box from which trail the fibre cables to the spectrograph; the complete assembly constitutes a `fibre module', which is easily removed from the robot to be interchanged with another. Coded proximity switches tell the Autofib control system which fibre module is in place.
The first fibre module contains 126 high OH fibres in 14 individually-cabled bundles of nine fibres, but it will be possible to image only about 120 of these onto the spectrograph detector at any one time. The fibres subtend 2.7 arcsec on the sky, and will be used principally for galaxy observations. Each of the 14 bundles has a nine-way fibre connector some 8.5 metres from the Autofib end to allow the lower part of the feed to remain permanently attached to the telescope.
At their output ends, the fibre bundles are brought together in a slit unit, where each group of nine is disposed in a skewed 33 array. The 14 arrays are then lined up in the collimator focus. The spectrograph thus sees three staggered fibre `slits', an arrangement needed to accommodate the 2-mm diameter sapphire balls that terminate each fibre. These are used in a reverse-Fabry configuration, so that the fibre is imaged accurately onto the collimator while a pupil formed 0.9 mm from the surface of the microlens acts as the entrance slit. The pupil has a nominal diameter of 0.5 mm, but varies with the focal ratio of the beam emerging from the fibre; thus a change in focal ratio alters resolution rather than throughput.
Because the large fibres admit a high level of sky background, and because point images are not well scrambled at fast focal ratios (Watson & Terry, Gemini, No. 42 (1993), 32), a smaller second set will be provided (probably late in 1995) for observations of point sources. They will capitalise on the throughput-stability of the spectrograph for accurate sky-subtraction, and will make the best use of the site's sky conditions. For best possible performance, the connectors will be dispensed with and a wide-band hydrogenated low OH fibre-type will be considered as an alternative to the current one. It may also be possible to replace the sapphire balls with graded-index lenses.