Spectrograph costs have become the limiting factor in multiplexed fiber-based spectroscopic instruments, because tens of
millions of resolution elements (spectral x spatial) are now required. Catadioptric (Schmidt-like) designs allow faster
cameras and hence reduced detector costs, and recent advances in aspheric lens production make the overall optics costs
competitive with transmissive designs. Classic Schmidt designs suffer from obstruction losses caused by the detector
being within the beam. A new catadioptric design puts the detector close to the spectrograph pupil, and hence largely in
the shadow of the telescope top-end obstruction. The throughput is competitive with the best transmissive designs, and
much better in the Blue, where it is usually most valuable. The design also has milder aspheres and is more compact than
classic Schmidts, and avoids most of their operational difficulties.
The fast cameras mean that with 15micron pixels, the PSF sampling is close to the Nyquist limit; this minimises the
effects of read-noise, which for sky-limited observations, far outweighs any difference in throughput. It does introduce
pixellation penalties; these are investigated and found to be modest.
For 4MOST, low and high resolution designs are presented, with 300mm beams, 3 arms with f/1.3 cameras, and standard
61mm x 61mm detectors. Coverage is 380-930nm at R=5000-7000, or R~20000 in three smaller ranges. A switchable
design is also presented. For Hector, a design is presented with 2 arms, 380-930nm coverage, and R=3000-4500; a 4-
armed design with smaller beam-size and detectors is also presented. The designs are costed, and appear to represent