A User’s Guide to CCD Reductions with IRAF Philip Massey (Feb 1997) Contents Introduction Why Your Data Needs Work And What to Do About It Doing It Outline of Reduction Steps Examining Your Frames to Determine the Trim and Bias Sections Setting things up: setinstrument, parameters of ccdproc, and ccdlist Combining Bias Frames with zerocombine The First Pass Through ccdproc Constructing a bad pixel mask Dealing with The Darks Combining the Flatfield Exposures Normalizing Spectroscopic Flats using response Flatfield Divisionm ccdproc Pass Getting the FlatFielding Really Right Combing the twilight and blank-sky flats Creating the Illumination Correction Finishing the Flatfielding Fixing Bad Pixels A How Many and What Calibration Frames Do You Need B The Ins and Outs of Combining Frames C Summary of Reduction Steps C.1 Spectroscopic Example C.2 Direct Imaging Example 1. Introduction This document is intended to guide you through the basic stages of reducing your CCD data with IRAF: be it spectroscopic or direct imaging. It will take you through the removal of the “instrumental signature,” after which you should be ready to extract your spectra or to do your photometry. Additional resources you may wish to review are: A Beginner’s Guide to Using IRAF by Jeannette Barnes Once you are done with this manual you may wish to go on to do stellar photometry on direct frames or to reduce slit spectrograph data. You can find details and sage advise in the following sources: A User’s Guide to Stellar Photometry with IRAF by Phil Massey and Lindsey Davis A User’s Guide to Reducing Slit Spectra with IRAF by Phil Massey, Frank Valdes, and Jeannette Barnes Copies of all these documents are available over the net; contact the IRAF group for details. In Sec.2 we will discuss whys and wherefores of CCD reductions in general terms. In Sec. 3 we will go through the actual IRAF steps involved in implementing these reductions, using spectroscopic data as the primary example. In the Appendices we provide a guideline to what calibration data you may want to collect while you’re observing (Sec. A), and discuss the nitty-gritty of the algorithms available to you in IRAF for combining images (Sec. B). Finally we close with a summary of the reduction steps for spectroscopic data (Sec. C.1), and the reduction steps for direct imaging (Sec. C.2). 2. Why Your Data Needs Work And What to Do About It This section will briefly outline how and why your CCD images need work. For a less heuristic treatment, see Sec. A, which discusses how many of what kind of calibration frames you need.Most of the calibration data is intended to remove “additive” effects: the electronic pedestal level (measured from the overscan region on each of your frames), the pre-flash level and/or underlying bias structure, and, if necessary, the dark current. The flatfield data (dome or projector flats and twilight sky exposures) will remove the multiplicative gain and illumination variations across the chip. Fringes are an additive effect that must be removed last. When you obtained your frames at the telescope, the output signal was “biased” by adding a pedestal level of several hundred ADU’s. We need to determine this bias level for each frame individually, as it is not stabilized, and will vary slightly (a few ADU’s) with telescope position, temperature, and who knows what else. Furthermore, the bias level is usually a slight function of position on the chip, varying primarily along columns. We can remove this bias level to first-order by using the data in the overscan region, the (typically) 32 columns at the right edge of your frames. We will average the data over all the columns in the overscan region, and fit these values as a function of line-number (i.e., average in the “x” direction within the overscan region, and fit these as a function of “y”). This fit will be subtracted from each column in your frame; this “fit” may be a simple constant. At this point we will chop off the overscan region, and keep only the part of the image containing useful data. This latter step usually trims off not only the overscan region but the first and last few rows and columns of your data. If you pre-flashed the chip with light before each exposure, there will still be a non-zero number of counts that have been superimposed on each image. This extra signal is also an additive amount, and needs to be subtracted from your data. In addition. there may be column-to-column variation in the structure of the bias level, and this would not have been removed by the above procedure. To remove both the pre-flash (if any) and the residual variation in the bias level (if any), we will make use of frames that you have obtained with a zero integration time. These are referred to in IRAF as “zero frames” but are called “bias frames” in KPNO and CTIO lingo. We need to average many of these (taken with pre-flash if you were using pre-flash on your object frames), process the average as described above, and subtract this frame from all the other frames. “Dark current” is also additive. On some CCD’s there is a non-negligible amount of background added during long exposures. If necessary, you can remove the dark current to first-order by taking “dark” exposures, long integrations with the shutter closed, processing these frames as above, and then scaling to the exposure time of your program frames. However, it’s been my experience that the dark current seldom scales linearly, so you need to be careful. Furthermore, you will need at least 3 dark frames in order to remove radiation events (cosmic rays) and unless you have a vast number of dark exposures to average, you may decrease your signal-to-noise; see the discussion in Sec. A. The bottom line of all this is that unless you really need to remove the dark current, don’t bother. The next step in removing the instrumental signature is to flatfield your data. The variations in sensitivity are multiplicative, and we need to divide the data by the flatfield to remove the pixel-to-pixel gain variations, and, in the case of long-slit spectroscopy and direct imaging, the larger-scale spatial variations. If you are doing direct imaging, or plan to flux calibrate your spectroscopic data, then you are probably happy to just normalize the flatfield exposures to some average value; but if you are interested in preserving counts