XMM-Newton Science Analysis System: User Guide

6.7 Analysing OM data

As has been pointed out already, OM data in image mode and fast mode with the normal filters, are fully processed by the SAS pipeline. All data, including the grisms can be run through the corresponding chain: for each exposure of a given observation, all necessary corrections are applied to the data files. Then a source detection algorithm is used to identify the sources present in the image. Standard aperture photometry is applied to obtain the count rates for all detected sources. These rates are corrected for coincidence losses, dead time and time dependent sensitivity degradation of the detector, and finally OM instrumental magnitudes and standard colour corrections are computed as well as absolute fluxes. A final source list is obtained from all exposures and filters. The detector geometric distortion correction and astrometric corrections are applied to each source's position, and also the whole image is converted to sky coordinates space and rotated so as to have the North on the top. Default image windows are also combined to obtain a mosaic (per filter) of the FOV. In the case of grism data, the spectra are searched for, extracted and calibrated in wavelength and absolute flux. Astrometry is also performed on grism data to compute the astronomical coordinates of the sources whose spectra have been extracted.

However, one has to be aware of some peculiarities of OM data before starting the analysis.

• Artifacts

  OM images show several artifacts. These artifacts are generally only visible when viewing the OM images in a high contrast and in logarithmic display (see the OM dedicated section of the UHB for some example images). Artifacts can however affect the accuracy of a measurement, e.g. by increasing the background level. The following artifacts may be present in the raw OM images.

  • Out of time events: very bright sources with count rates of several tens of counts per second show a strip of events along the readout direction. As the OM is a shutterless detector, these events correspond to out of time photons arriving while the detector is read out.

  • "Smoke rings": bright sources generate smoke ring like structures, which are located radially away from the centre of the field of view. These rings are caused by photons which are reflected off the detector entrance window back onto the detector photocathode.

  • Fixed pattern noise: raw OM images show a modulo 8 regular pattern originating from imperfections of the event centroiding algorithm in the instrument electronics. This modulo 8 pattern is removed during image analysis by the task ommodmap, if sufficient statistics are available in the image but this is not always possible for low count images (e.g. some UV filter images).

  • Straylight features: straylight is caused by a chamfer in the OM detector housing which reflects light from outside the OM FOV onto the active detector area. The reflected celestial background light sums up onto a circular area with an increased background rate in the centre of the OM field of view. The light coming from bright sources at offaxis angles $\sim$12-13 arcmins and reflected by the chamfer can also produce loop like structures in an OM image. These loops can degenerate into long streaks depending on the source positions.

• Grism Data

  The complexity of grism data already mentioned earlier is increased by the presence of the artifacts described before (see § 6.4.4). Great care has to be taken when analysing the output products of grism data processing with SAS.

The following points should also be kept in mind:

1. The OM operates in photon counting mode but images are accumulated on board. Good time intervals (GTI) have no meaning in OM data. Either the full exposure is selected or discarded.

2. Contrary to the X-ray instruments, an OM exposure does not provide direct energy information except when grisms are selected.

3. As has been explained, a flatfield response of unity is assumed.

4. Coincidence losses can occur which depend on the source brightness and on the CCD frame rate. The frame rate itself depends on the selected configuration of the detector science and tracking windows. If two or more events are located close to each other within the same CCD readout frame, they are detected and counted as one event. Also when photon splashes overlap, a mis-location is given by the centroiding algorithm. Another consequence of coincidence losses is event depletion, occurring at high count rates around the central source position.

5. Corrections for coincidence losses and for detector deadtime are applied when aperture photometry is performed on the detected sources. These corrections are not applied to any of the produced images.

6. The OM filters do not form a proper photometric system. However the photometric calibration of the instrument, based on observations with OM and from the ground, allows the SAS to obtain standard U, B, V magnitudes and colours in the Johnson system. (This applies to stars. Extended objects should be treated with care.) For the UV filters (UVW1, UVM2 and UVW2) AB Magnitudes have been defined in addition to the instrumental system. An absolute flux conversion is provided for all filters.

7. The OM point spread function (PSF) has wide wings. This is taken into account in the application of the calibration with SAS through the proper setting of photometric apertures. A similar approach has to be taken if one wants to make an independent processing and analysis of OM data using any other data reduction package.

8. Straylight features, already mentioned, complicate the background subtraction in some cases, specially when the target star is located in or close to a straylight feature.

9. Imaging and fast mode data are corrected for time sensitivity degradation, so that data of a given source (count rates, photometry) obtained at different epochs can be compared. As of SAS 18, grism data processing also corrects for time sensitivity degradation.

10. In dense fields or fields with complex background variations, such as where nebulosity is present, contamination of source extraction apertures and/or background regions may not be adequately accounted for.

11. Although the OM SAS software attempts to identify problems, such as where sources are affected by bad pixels, read-out streaks or nearby objects, and flag sources accordingly, the flagging process is not foolproof and there are situations where sources may have issues but remain unflagged, or vice versa. For example, in some cases, if a bright source causing a read-out streak is not within the imaged area, sources that lie on its read-out streak may not be flagged.

In principle, after SAS has been run there is no need for further data reduction. The observation source list should contain the calibrated data with their errors. Therefore, the user can proceed with the analysis and interpretation of the processed data. However, some checking is recommended to verify the consistency of the data output.

The whole data processing can be repeated easily by a Guest Observer or XSA user, should any calibration file be updated and/or, importantly, in case there is doubt about the reliability of the results. The processing chains or the pipeline, apply default options in all SAS tasks, which eventually can be changed by the GO in order to improve the quality of the results. In particular, the source detection is very sensitive to the artifacts that are common in OM.

In what follows, some checks are outlined that a user should perform on OM processed data (by the standard Pipeline or by running SAS), and the use of one of the tasks, omdetect, where the user can modify parameters affecting the source detection and therefore the overall results of the data analysis.

1. Checking omichain output products:
   • The first thing to do is to overlay the image source list onto the sky image. This can be achieved with implot, which will overplot the detected sources on the corresponding image, and it will allow checks that all sources are real and that the one(s) of interested has/have been properly identified. A new task, srcdisplay does the same job and it is more friendly to use. If the REGION files produced by omdetect have been preserved (in manual run of the chain or the task), then ds9 can also be used to plot the detections on top of the image.

     If the background is strongly affected by straylight features, this check is very important.

   • Inspection of the combined source list (e.g. using fv) will allow us to check that the sources of interest have been picked up in all the filters where they are visible and that the combined list contains colours and standard magnitudes for them.

   • Special care should be taken in examining the quality flags assigned to all detected sources. Their meaning is described in detail in the SAS on-line documentation (task omdetect).

   • Check the tracking corrections: although the pointing stability of XMM-Newton is very good, one can verify it by examining the corresponding tracking history PDF file.

   • When there are several consecutive exposures in a given filter, different values of the frametime parameter (e.g. in the default image mode, where each of the 5 exposures has different frametime) can introduce a variation of a few percent in the count rate from one exposure to the next. This is due to the coincidence loss correction, which depends strongly on the frametime. The resulting jumps in average count rate may be more severe for bright sources.

   • When there are several exposures with the same filter, omichain combines the corresponding images and then it attempts a deeper detection in the co-added image. New sources may appear as a consequence of the increase in S/N. However, these sources do not have a proper coincidence loss correction. Since the new detected sources are very faint, the error is only a few percent. Some of these new sources may be fake detections due to errors in the alignment of the images or to enhanced noise patterns. All these new sources are flagged in the final merged list (OBSMER file).

   • If the observation was made after revolution 3224, check whether the source of interest is affected by the Jupiter patch (§ 6.7.7)

2. Checking omfchain output products:
   • Inspection of the light curves:

     One can look at the PDF files containing the light curves to check that there is some signal detected and measured (both in the source and in the background).

   • Checking the presence of source(s) in the fast window pseudo-image.

     This can be done easily by displaying this image with SAOImage (ds9) or fv. Another possibility is to overlay the detected sources onto the pseudo-image using the task implot or ds9.

   • Checking with ds9) or fv, the presence and centering of the source(s) in the fast window pseudo-image.

   • Using bkgfromimage=yes in omichain (or omlcbuild if running step by step) will give better results if the source is bright or if there are more than one sources in the fast mode window. This is the default since SAS 12.

   • Jumps due to coincidence loss correction. A fast mode observation consists of a series of exposures with one or several filters. As discussed for default image mode, the different values of the frametime in consecutive exposures can produce jumps in the average count rate level from one exposure to the next. The effect can be more significant in fast mode data where jumps in the mean count rate between exposures of up to 10% may be present.

   • Fast mode data can suffer from some issues related to poor centring of sources in the small fast mode window, such as can sometimes arise when field acquisition fails, when there are unusually large random pointing errors or where the provided target coordinates were inaccurate. These are described further in a technical note, Centring, drift and oscillation issues associated with OM fast mode observations.

3. Checking omgchain output products:
   • Checking the detection of zero and first orders

     The correct correlation of the zero and first order of the spectra is essential. It can be verified by displaying the rotated image and overlaying on it the detections from the SPCREG file with ds9. SAS 8.0 produces a picture of the grism image with superimposed extraction regions. The source list SPECLI file indicates also these correlations. If the users wants to analyse all detections, then REGION and SWSRLI files should be examined.

   • Position of zero order

     The position of the zero order (centroid) is the zero point of the wavelength scale. It should be verified, e.g. using ds9 as pointed out before.

   • Identifying the spectra.

     The final spectra present in the SPECTR.FIT file can be identified as said before by overlaying the SPCREG file on the rotated image or by looking at the produced .PS file. In addition, the astronomical coordinates of the sources showing spectra are computed when running omgchain.

4. Improving the source detection:
   • For image data, if the source of interest is close to straylight features or to other sources it may not be detected with the default settings of the omdetect task. The parameters have to be changed: nsigma and minsignificance (see SAS documentation, or run omdetect -h for details)

     omdetect set=your_image.fit outset=your_sourcelist.fit \
              regionfile=your_region.dat nsigma=p minsignificance=q
     

     where your_image.fit, should have been produced by ommodmap. The default values for nsigma and minsignificance are 2 and 1, respectively.

     Invoking omdetect with regionfile=your_region.dat will allow a fast checking by overlaying the newly detected source positions on the image with ds9 using the created region file.

   • In case of fast mode, if there is more than one source in the fast window, then the detection will be affected and therefore the whole light curve as well. Some parameters have to be changed in omdetect as in the case of image mode (see the online SAS documentation, or run omdetect -h for details).

     Invoking omdetect with regionfile=your_region.dat will allow a quick check by overlaying the newly detected sources positions on the image with ds9 using the created region file.

5. On grism data, as with normal image data, omdetect can be used to improve the detection. However, it is recommended to use omgsource to select the spectra interactively.