In order to properly process and calibrate your data you need to acquire at least the following images:
Note that for sequences of identical images like biases, darks, dome flats, we have provided a "combine" option in the EXPOSE form to compute a combined frame from the individual exposures, using the IRAF/COMBINE task (see " A Users's Manual for the CCD Data Acquisition User Interface@", accessible via the hyperlink). This process, combines the images by using the average sigma clipping algorithm, which has proved to provide slightly better results than a plain median algorithm (same mean value, but better r.m.s.). To select this option, click on the "median" button in the "EXPOSE" form, before starting the sequence. A combined image will be computed as soon as the image acquisition sequence is completed, and written to the disc as a FITS file with "B", "F", "D" affixes depending if they are combined biases, flats or darks, respectively.
Each exposure is assigned a unique sequence number. The name of the file containing an image is this exposure number followed by a letter and given the "fits" extension. The option by default is a "object" frame for obvious reasons. Make sure you select the appropriate option, if you are not taking an "object" frame, each time you re-open the "EXPOSE" form. Be careful in using the right exposure type (######d.fits, ######b.fits, ######c.fits, ######o.fits) since it is used in the CFHT archival scheme, iqe scheme and also for automatic image processing of archived images. Besides, it will make your life easier for your own processing especially if you develop a bulk automatic procedure.
A set of (at least) 10 bias frames at the beginning of the night, plus another set of 10 at the end should be taken.
As an indication of the CCD associated electronic noise, the bias level may vary by a few ADCUs in its mean value during the course of a night. To be able to follow these variations, we have provided an overscan region for each CCD (for the GEN III systems, the overscan consists in 40 columns of pixel added to the frame; if you chose a raster of 500x500 pixels, your final image will have a format of 540x500 pixels). The mean ADCU value in this overscan is identical to the mean bias value at the time the exposure was taken. To correct for the bias level in a given frame you need to follow the following sequence:
To take a sequence of bias frames proceed as follows:
Dark Current frames
A set of dark current frames, obtained by integrating for as long as your typical exposure times without opening the CCD shutter, may be taken for the RCA2, RCA4 CCDs, since the dark counts will amount to several ADCUs for typical FOCAM exposures. As for any step in the data processing, the master dark frame should add a negligible noise when subtracted from the object frames. For this reason you need to take as many dark current frames as possible. Some observers consider that the dark current level for the above CCDs is sufficiently low, and with almost no 2D dark current pattern, that there is no need to correct for the dark current with a full 2D dark frame. Instead, a correction to the bias level is made in the form of an additive constant representing the mean dark count in ADCU.
For CCDs like Loral3, Lick2, MOCAM the dark current is so low, with a very uniform 2D distribution, that you would probably introduce more noise in the data by using a dark current frame. However, Orbit1 has shown definite time varying dark current, especially after pinning (see the detector page@ for a better description of this problem and its symptoms.)
To take a sequence of dark frames proceed as follows:
Flat Field Frames
Flat fielding is one of the most important aspects of frame correction. THE best way to insure that flat field exposures match the program frame exposures is to use the ``expose and shift'' technique wherein multiple shifted exposures of a given field are taken. Later, a flat field exposure is formed from the median of the multiple exposures in a given band pass. The median filtering is a way to do a form of ``pattern recognition'' in the frame, that is to find the invariant pattern due to the chip response while ignoring the transient pattern (stars, galaxies, cosmic rays, etc.). However, this technique assumes that your objects are relatively small compared to the field size (i.e. it would not work for observations of a nearby galaxy filling the entire field of view). An additional complication can arise from variations in sky illumination from frame to frame, however. This idea is not particularly new but it is most applicable to data where the signal-to-noise in the individual frames is high enough so that the background is not read-noise limited.
In many instances a flat field exposure that exactly duplicates the program frame illumination cannot be obtained. Twilight sky exposures are useful because (1) they can be obtained in relatively bright sky conditions, (2) they provide flat illumination, and (3) the optical path is the same than the science frames. However twilight sky exposures do not exactly match dark sky illumination in that polarization and spectral content may be very different between the two. The twilight emission changes rapidly with time causing variations from one exposure to the next. This has the secondary effect that the level of the illumination is difficult to control.
Dome flats can be obtained more easily than any other type of flat field illumination. Each telescope dome is different in its characteristics. As well, the lamps used for dome illumination differ in color and spatial illumination. Possible `red leaks' in some filters can wreak havoc when trying to properly flat field exposures using `dome flats'. Proper, or at least adequate, illumination of the primary is difficult to achieve and what works at one observatory does not necessarily work at another. The best results at CFHT were consistently obtained with several flood lamps pointed down near the inside dome catwalk. In order to color balance the necessarily large lamps, blue plexiglas that cuts off red wavelengths was purchased and installed. Four lamps are routinely used. The CCD's flat field best when the telescope is pointed at a diffusely lit portion of the dome. Concentrated illumination has never worked in the CFHT dome for CCD's attached to FOCAM. Notice that the Cassegrain upper end has some flood lamps installed on it. This means that with FOCAM mounted at the Cassegrain focus, you can obtain flat field with the telescope at zenith. The lamps can also be switched on from the control room (module to the left of the data acquisition computer), and offer varying degrees of illumination. This does not require the presence of a TO; just remember to ask the TO to keep the compressors on at the end of a night so you can open the mirror covers yourself the next day. A similar setup is being implemented on the prime focus upper end.
Flat field exposures taken at the anticipated level of the sky in program frames work well (levels 5000-30,000 e-). Another school will tell you that it is better to expose the flat fields to the maximum number of electrons allowed to stay in the linear regime of the CCD. A minimum of 10 exposures at this level can be combined to produce a given flat field, but, "more is better". Obviously, if more flat fields are obtained the statistics will be improved. A sufficient number B, V, R and I exposures can be obtained for most programs in 1 hour including 1 test exposure to verify the exposure time; the typical exposure time is 2 to 10 sec with the 4 lights on. Narrow band filters and Ultraviolet filters are much more time consuming. For sky flat fields, exposure times greatly vary with sky brightness: use the "trial and error" method with a small CCD raster (faster readout time) to adjust to the proper exposure time and then take full frames as needed, while increasing or decreasing the exposure time between each as the sky gets brighter or darker.
Low frequency variation in the form of an illumination gradient of 1-2% has been observed at prime focus in sky frame corrected with the dome flat fields alone and no illumination correction. These low frequency variations can be traced back to the dome flat fields (comparison of dome flats and sky flats), and can be corrected easily by heavily smoothing the dome and sky flats to form a secondary low spatial frequency illumination correction frame.
Residual night sky fringes (noticable in RCA2 and even more with RCA4 in the near IR, non-existent for the thick, non coated Loral3) should be obtained by median filtering program frames. Fringing results from interference patterns generated by incident light impinging on the silicon CCD substrate and reflections either between the CCD and its support glass or within the CCD itself. Thin chips exhibit fringing more readily than thick ones, and the RCA chips with glass substrate have sometimes suffered from severe fringing. The fringing pattern in a given bandpass (for thin chips 8500 A) is a superposition of the fringing from all wavelengths in the bandpass. If a narrow feature such as a night sky line occurs in the incident wavelength range, it can produce a fringing pattern along with the normal `continuum' pattern. Variable night sky lines such as [OI] at 6300 A can cause the `flat field' pattern on the CCD to vary from frame to frame. Some correction for such variable patterns can be made by first dividing by the dome flat and then subtracting a `fringe' frame. The fringe frame is scaled to the variance of the residual fringing in the program frame. However this procedure is not strictly correct at all wavelengths. For bluer wavelengths the original flat field should have a fringe frame added to it and then this new flat field should be divided into the program frame.(!)
Flat field sequence:
You have an alternate option for the first step if you are observing with FOCAM at the Cassegrain focus. This option is to keep the telescope at zenith, and switch the upper end flat field lamps using the module by the data acquisition computer.
Photometric Calibration Frames
At least two standard fields should be observed per night at different air masses to be able to correct for atmospheric absorption. A selection of photometric standard fields is given in Appendix E. We recommend two exposures per field per filter, in order to obtain optimum S/N on faint as well as bright stars of a sequence. Some people defocus the telescope in order to get better photometric measurement accuracy. For the fields in Appendix E, exposures times are typically 30 and 90 sec for B, V, R and I filters. Appendix F presents the average photometric coefficients derived from FOCAM observations.