1 Introduction

Many large telescopes provide a Coudé focus to observe high-resolution stellar absorption line spectra in a format that was originally fitted to the size of photographic plates. While the photographic emulsion since then has been replaced by the CCD, the available spectrographs often do not allow the exposure of more than a very small part of a single spectral order. This requires multiple exposures as soon as a significant fraction of the visible spectrum is needed, either introducing a strong decrease of available telescope time or leading to an unacceptable shortening of the scientific programs.

The obvious solution is an échelle spectrograph installed at either Coudé or Cassegrain focus. Due to the surprising development of modern CCDs cross-dispersed échelle spectrographs can image the complete visible spectrum with moderate to high resolution on a single chip of 1024² or even 2048² pixels. A few échelle spectrographs have been built during the last decade and are now used with great success for a number of astrophysical programs. There are, however, large telescopes not equipped with an échelle spectrograph though there exists an increasing demand of spectroscopic telescope observing time. One of these sites was the German-Spanish Astronomical Center on the Calar Alto with its large telescopes of 2.2 m and 3.5 m diameter. The 2.2 m telescope which is a twin to the one on La Silla, has a large Coudé spectrograph with f/3 and f/12 cameras. Both can be equipped with a CCD camera providing resolution elements (2 pixel) of $\lambda / \Delta\lambda\ = 20000$ to 45000. The shorter camera is therefore not really appropriate to observe absorption line spectra of cool stars. Moreover, the CCD installed at the f/3 focus cuts out a small spectral range of $\sim\! 10$ nm only. For many purposes this is unsatisfactory. Abundance analyses or simultaneous investigations of spectral lines always require significant spectral coverage which could be obtained only with multiple exposures. An illustrative example is the Coudé spectroscopy of a 10th magnitude star. Near 5000 Å a single exposure of 60 min is required to reach a S/N ratio of 100. If a total of 1000 Å coverage is necessary one had to spend a whole night observing only this one star. Vice versa, a large amount of observing time can be saved by a spectrograph with resolution comparable to Coudé spectrographs but spectral coverage including the full visible wavelength range. We therefore decided to build an échelle spectrograph for use at the Calar Alto 2.2 m and 3.5 m telescopes for which we benefitted strongly from experience with existing échelle spectrographs such as the ESO CASPEC which was one of the few instruments being in regular use at large telescopes.

The final design of this instrument was driven by a number of constraints that are specified by the needs of the spectroscopy of stellar absorption lines,

• The definition and stability of the échelle image on currently available CCDs was a fundamental requirement. We used fibre optics to feed the spectrograph, which can thus be mounted in a thermally and mechanically isolated environment to provide an exceptionally stable echelle order image on the CCD. This allows a more accurate definition of spectrum positions than is possible for a spectrograph mounted at the telescope.
• A major problem encountered with existing échelle spectrographs such as the CASPEC is the different illumination of the spectrograph entrance slit for a stellar image and for the flatfield lamp. Most of that difference is removed by scrambling the light in the fibre.
• Scattered light background under uncontrolled conditions often constitutes a substantial fraction of the astronomical light. Such losses also tend to degrade the information contained in spectral orders. This has been discussed in a number of papers (Gehren & Ponz 1986; Hall et al. 1994). We found it much better to remove most of the straylight from the camera beam with the help of an intermediate slit. Replacing the usual cross-dispersion grating by a prism also keeps the straylight at a minimum.
• As discussed above the flatfield correction of échelle order spectra is problematic. Small drifts of order positions as observed in some échelle spectrographs mounted at the Cassegrain focus lead to a situation in which the order centers of astronomical object and flatfield exposure are systematically displaced, and a pixel-to-pixel flatfield calibration becomes impossible. Our solution to that problem is a combination of a white light flatfield resulting from an illuminated screen in front of the échelle and the usual one-dimensional order flatfield.
• Spectral coverage and adjacent order overlap are constraints to the échelle format. Spectral coverage has been one of the driving motivations to build FOCES. High-resolution stellar spectroscopy is dedicated to cool stars. The flux of such objects is heavily weighted towards the red which accounts for long exposure times to obtain a sufficient signal in the blue. During such exposures the red spectrum becomes overexposed and therefore useless. Depending on the type of cross-disperser the échelle orders become very crowded in the red or blue region of the spectrum, and this also constrains spectral coverage. Finally, large spectral coverage requires a highly efficient atmospheric dispersion corrector. We therefore decided to limit FOCES to a standard 380 to 750 nm coverage, with an option to shift the spectral range towards 430 to 900 nm. Implications for the selection of optical components, fibres and the CCD are discussed below.
• The use of optical fibres to transmit the astronomical light from the telescope focus to a remote spectrograph entrance slit has led to some experience during recent installations (Ramsey & Huenemoerder 1987). It is a source of potential light losses. Compared with a system of lenses and mirrors the throughput is inferior. However, The light-scrambling properties of optical fibres lead to an improved illumination of the spectrograph entrance slit, and this allows more accurate Doppler velocity determinations (Heacox 1986; Ramsey 1988). The disadvantages of fibre spectroscopy due to degradation and transmission properties may lead to a significant reduction of the performance in the near infrared, although more recent developments have led to considerably improved properties (Barden 1995).

We give a full description of the spectrograph in Sect. 2 including optical layout, electronic and mechanical systems. In Sect. 3 we discuss the telescope module and the fibre optics. In Sect. 4 we present some of the first observations obtained at the 2.2 m telescope of the German-Spanish Astronomical Center on Calar Alto. We present a data extraction software used to obtain échelle order spectra and compare the results of these test observations with the predicted performance of FOCES. The final section carries a discussion of the spectrograph in its present state.