There is a comprehensive Starlink Guide for those preparing for an observing run—Preparing to Observe (Starlink document SG/10).
In this section some notes on what to be aware of prior to an observing run are given. In outline: to successfully extract and calibrate an échelle spectrum a complete set of reference frames should be obtained at the telescope.
You might want to refer to the more extensive discussion of CCD data calibration in: A User’s guide to CCD Reductions with IRAF, the section entitled “How Many and What Calibration Frames Do You Need?”
This is what you need to attempt ‘textbook’ échelle data preparation:
The arc and object frames will be processed by the échelle data reduction software. The bias and flat-field frames are used in the preparation of the CCD data.
A more complete introduction to the handling of CCD data can be found in the CCD Reduction Cookbook (SC/5), a brief outline is given here.
In order to remove detector-related effects a complete set of bias, dark, and flat-field frames should be obtained. It is important to bracket the science data exposures with complete sets of CCD characterisation frames. A post-observation review of these will reveal any image shifts. Having bracket frames allows the data to be accurately prepared even in the event that some time-dependent variation is found (as long as its a small, slow-varying effect).
See §4.1 for outline details of CCD data preparation.
In some cases it may be possible to not use any bias frames. Instead, a median value for the bias level is obtained by inspecting the overscan region in some or all of the object/arc frames. You can use fewer bias frames when you have a high signal-to-detector-noise ratio.
For exposure times limited by cosmic-ray event counts, the ?? in most CCD cameras is not a significant factor. The simplest way to decide whether to take dark frames is to take one of exposure time similar to that you are using for object frames, and check the signal level.
Beware of flat fielding in échelle spectroscopy. For a stable spectrograph (e.g. UCLES which is in the AAT coudé room) you can take flat-field frames at any suitable time. For less-stable instruments it may be difficult to get a useable flat field. The problem is getting the object and flat-field orders to fall on the detector in the same place. In this case, the best procedure is to take flat fields immediately before and after each science exposure in the same way as you would take arcs.
When preparing flat-field frames for your data ensure that a dekker size (length) larger than that used for the science exposures is used—avoid order overlap in the image, though. This ensures that a reasonable flat-field is available across the complete profile of each order. If a choice of gratings is available, use the one which will give you the widest order separation.
In some areas of the échellogram the brightness of a single flat-field may fall off or vary rapidly due to the characteristic spectrum of the ?? used or variations in the efficiency of the detector (some front-illuminated CCDs have 20–30% variations in efficiency over a small wavelength band). There are two ways to overcome these effects: use a different lamp to produce the flat field in those orders—but usually a different lamp is not available—or, obtain as many unsaturated flat fields as possible and sum or average them.
It may not always be necessary to flat field; a flat-field frame taken with a modern CCD can be flat in response to only a few percent. You may only need to flat field if the signal-to-noise ratio you require is particularly high. The precise figures will depend upon several factors, and should be estimated for each science object.
Accurate determination of the path of the échelle orders across the images is vital to achieving the best extractions. Processing software requires a bright, clean (i.e., cosmic-ray free) image from which to trace the orders.
For well-exposed continuous spectra the object frame can be used for tracing. However, in the case of faint objects, or objects with strong absorption features in their spectra, tracing of object frames will not be easy. A flat-field frame can be used for tracing as this is likely to give a good signal. Therefore, if necessary, obtain a few flat fields with a narrow dekker to improve the traces. It may be worth having several possible frames for tracing—in case some of them are badly contaminated by cosmic rays and so difficult to use.
As with the CCD characterisation frames, wavelength-scale reference (arc) frames should be taken bracketing the science exposures if precise wavelength scales are required. At the high dispersions used in échelle spectrographs a small change in the optical system can lead to a detectable shift between the bracketing images. Using both arc spectra, a time-weighted mean wavelength scale can be produced and applied to the science data.
Existing standard star data is often based upon a system in which the band size is much larger than the wavelength coverage of a single échelle order. In practice, this can make the application of proper flux calibration to high-resolution échelle spectra difficult or infeasible. Proper high-resolution spectral standards are now starting to become available (notably for HST). If you intend to flux calibrate your data you should ensure that suitable standards are available.
Refer to §4.9 for more information on the problems associated with flux calibration.