Furthermore, the background extraction region is an annulus, possibly with circular areas excluded (see the OMREGION task description), and these excluded circles are not necessarily totally enclosed by the annulus. In addition, the FAST-mode OM window is small, and a part of the source and background areas falls out of this window. The source and background counts should be scaled to the CAL circular area in order to correct them for the coincidence loss. This scaling is sensitive to the currently used OM PSF. If the source and background areas are affected by the source proximity to the edge of the OSW or if the real OM PSF differs from that taken from the current calibration file (CCF), than the OMLCBUILD output count rates could vary from exposure to exposure, even if the source brightness remains constant.
This effect could be increased for bright sources due to the 8-pixel pattern noise. In order to avoid these variations and to obtain a uniform combined light curve, it is better to use a fixed radius for the source extraction area for all the exposures in the ODF. This radius should not be less than 6 pixels, although for faint source one it is possible to use smaller values (this would increase the signal-to-noise ratio). The value of 6 pixels is currently used as the default value for the chain in order to get an optimal signal-to-noise ratio for most of the sources. One may change this parameter by using the command `omfchain srcradius=...`. If srcradius is set to a negative value, it will be regarded as measured in unbinned pixels and will be fixed for all the exposures (its default value is negative).
When changing the srcradius parameter one should also properly establish the background extraction area setting bkginner and bkgouter parameters (see the description of the task parameters).
implot set= sky_image_file.FIT withsrclistset=y srclistset=osw_source_list_file.FIT device=your_device (e.g. /XW) itv=1 radius=3 maxsrc=10 colour=7.
While often there will be only one object in the image, one should check for spurious detections, which could affect the analysis of the target source and also for missed sources or prominent objects in the window, which may contaminate the target. Currently, bright contaminating sources are not treated properly, and, if present, affect the light curve of the target (and vice versa) - this will be rectified in a future version of the relevant SAS tasks. It is possible to change the detection efficiency via the boxscale, smoothsize, nsigma, and contrast parameters.
The visual inspection may also reveal that a bright source of interest is affected by modulo-8 noise (a rectangular pattern around the source). Obviously, the standard (calibrated) point-spread function is not applicable to these kind of images (the source counts could be found at any pixel within the Fast-window; and a large fraction of them fall outside of this window). For this kind of images the determining of background from the Fast-window becomes impossible, and it would be appropriate to consider the entire Fast-window as the source extraction region. Since the sources generating modulo-8 noise patterns are quite bright, the background becomes negligeable and, thus, could be set to zero. For this purpose one can set the source extraction region to the entire Fast window by setting the parameter srcradius to, say, 12 pixels (half of the Fast-window x- or y-size). The light-curve obtained in this way would be more reliable.