 |
 |  |
|
| |
 |
Summary of Slow Slew Survey Test of Rev 1242
Select Sky view for 
Table of contents
IntroductionThe aim of the test was to address the capability of XMM-Newton to map a predefined region of the sky in slewing mode. Practical questions addressed in the tests have been:
Before going into the details of the test, the following table provides a summary of the main results:
Design of the test
Sky RegionFor the slow slew survey mode to be efficient, there must be a compromise in the size and shape of the mapping region: the shape should be close to rectangular so that long slews are performed along the long side of the rectangle. Efficiency (overheads and accuracy) arguments require that the duration of these long slews must be kept, roughly, within 30 to 90 minutes. At a slew rate of 30arcsec/s, the slew paths would be in the range 15-45 degrees (or 30-90 times the XMM EPIC field of view). The total duration of slow slew survey is finally determined by the other side of the rectangle and the overlapping required between adjacent slews. Depending on a number of constraints, the number of slow slew exposures that can be done within a single XMM-Newton revolution is 15-20. To reach a roughly uniform exposure in the sky region to be mapped, adjacent slews should overlap by about half a field of view (~30arcmin). Thus, slow slew surveys would be able to map rectangular regions of, typically, ~30x4 degrees.The analysis of slews during normal operations and the results of some specific tests show that the slew errors, i.e., the departures from the expected path, depend on the slew length and the direction of the path. To ensure that the requested area is actually mapped, the slew errors must be significantly smaller than the size of the EPIC field of view, otherwise the probability of having missing regions or "holes" in the exposure map will not be negligible at all. "Best" slew paths in terms of accuracy are those performed around "sun-lines", i.e., holding the angular distance to the Sun constant, and among these, those with the Sun at 90 degrees record the smallest errors. The paths of crossing sources in a sunline slew have a fixed orientation in the scientific instruments layout: parallel to the long sides of individual EPIC-pn CCDs and perpendicular to the dispersion direction in RGS's. Two main criteria have been followed to select a region on the sky to execute the slow slew survey test. First, it had to be a region with a large number of "well" known X-rays sources, i.e., sources whose positions are known to better than a few arcsec and bright enough to be detected in rather short XMM-Newton exposures. These sources can be used as a control sample to analyze the results of the test. The second criterion was to visit a region expected to have a significant fraction of hard sources. There exist several sky surveys in the X-ray range, but all of them restricted to soft bands (<4keV); XMM-Newton is the first X-rays mission suitable to perform some sort of survey in the hard X-ray band. The results of the test may then be used to compare with surveys in the soft and hard bands. The Galactic Bulge naturally fits all conditions outlined above: It has the right size, the (almost) correct orientation with respect to the Sun in September, contains a large number of well know sources and the high extinction results in hard sources. Therefore, it was decided to use as target region the Galactic Bulge. In total, 15 slew exposures were scheduled in two sets of 8 and 7 slews. Since the overlap between adjacent slews was established at 15arcmin and the length of each slew to ~43 degrees, the total area of the sky mapped in the test was, approximately, 2x(43x2degrees). The reason for having two separate regions is the approach of the Earth constraint by the end of the revolution.
Figure 1: Sketch of the slew paths for the Slow Slew Survey Test in rev. 1242. Consecutive slews are shown in red and blue, alternately. Solid circles and open squares indicate the initial and final attitudes, respectively.
Instruments ConfigurationA summary of the instruments setup is given in Table 1. Details for each instrument will be briefly commented, but the overall idea was to go for short frametime, to increase the number of different samples for each single crossing source. It should also be noticed that, during this test, the instruments were exposed, i.e., collecting science data, ONLY during long slews but not during fine tuning slews or spacecraft stabilization periods.
EPIC-pnEPIC-pn was commanded in Full Frame. All EPIC-pn exposures used the same offset map, that computed at the initial attitude of the first slew. This initial attitude was chosen at planning time to have no bright optical source within the EPIC-pn field of view. In addition, the Medium filter was selected, as in the case of MOS, to reduce the potential impact of optical loading during the slew exposures.
MOSThe frametime for Full Frame mode of the MOS cameras is 2.6s. During this time a source would cross 78 arcsec if the slew is executed at 30 arcsec/s. In order to reduce such a frametime and improve the final spatial resolution and astrometric accuracy, MOS cameras were operated in a non-commissioned Full Frame mode where a binning of 3x3 pixels was performed on-board. Such binning results in an effective pixel size of 3.3 arcsec but the time resolution would have reached 0.4s. In practice, the 0.4s readout was just beyond the capability of the electronics, and all MOS exposures finally worked with a frame time of 0.8s.
RGSRGS spectrographs were operated in the nominal Spectroscopy mode, except for the readout sequence of the CCDs: chip number 5 (∼20-24Å) was read out consecutively seven times and, then, chip number 9 (∼5-7Å) was read. With this configuration, CCD#9 was operated exactly as in default Spectroscopy mode (frametime=4.6s) and it was taken as reference to monitor the radiation level. On the other hand, the frametime of CCD#5 was reduced to 0.57s.
OMOptical Monitor timing capabilities are provided by its so-called Fast Mode, in which photon events are time tagged with a maximum resolution of 64 ms. However this mode requires using a small window of only 22x23 pixels (11.5 x 11.5 arcsec2). We have taken the opportunity of the Slow Slew Test to check the detection capability of OM while slewing. It is clear that in a real slew operation, OM could be used only if the Fast mode could be used in a much larger window, but this requires a change in the on-board software whose feasibility is being studied. Contrary to the X-ray instruments, OM exposures could not be included in the schedule and they had to be executed via manual commanding in Real Time. Four exposures of 2,200s duration were commanded, but only three of them were succesfully downloaded to ground. Each exposure includes one IMAGE (7.5x7.5 arcmin) and two FAST (22x23pixels) windows. The time resolution of the FAST window data was set to 0.125s. Tha data in table 1 refer to Fast mode.
Results of the testThe results to be discussed here will be based on the first set of 8 adjacent slew exposures. As will be shown below, the background radiation became unacceptably high during the last three exposures, leaving a significanly smaller surveyed region in the second set of adjacent slews. Slews accuracyAs discussed above, a high accuracy in the slews is critical to ensure the correct coverage of the sky region to be surveyed. The results confirm that such accuracy is achieved in the test performed in revolution 1242. The errors in the direction perpendicular to the slew path (∼3arcmin) are significantly smaller than the instruments field of view (∼30arcmin) and the overlap between adjacent slews (∼15arcmin). The errors along the slew path were much larger, but always in the same sense as the slew, so that each slew was somewhat longer than planned. The fact that no slew was shorter than planned assures that there was no missing region in the area surveyed. It is worth noting that the errors along the slew path were significantly smaller (∼20arcmin) in the first set of slews, performed at Solar Aspect Angle (SAA) close to 90 deg, than in the second set of slews (∼40arcmin) at SAA ∼ 77deg. Other slew orientations with respect to the Sun have not been tested yet, but the results of this test and the knowledge of the way XMM-Newton executes long slews strongly suggest that the uniformity in the coverage of the surveyed region will be severely compromised if the scanning slews are not performed as sunline slews. SAS ProcessingEPIC-pn and RGS slew survey data can be fully processed with SAS7.0.0. Here, "fully" means that in the case of EPIC-pn SAS7.0.0 can provide correct astrometric coordinates and correct source spectra and detector responses. Similar spectral products can be obtained for RGS sources with the same SAS version. It has also been possible to get similar products for MOS data, but not with the public release of SAS. Finally, SAS7.0.0 can only provide intermediate products for OM exposures; astrometric and photometric results still require some software development. Even in the case of EPIC-pn and RGS, some precautions have to be taken to ensure a proper processing of the data from any XMM-Newton instrument: Attitude History File: SAS must be forced to deal with the finest time resolution possible. This is achieved in the Raw Attitude history file (RRRR_9RRRR000LL_SCX00000RAS.ASC) and not in the Attitude History file (RRRR_9RRRR000LL_SCX00000ATS.FIT). There are two ways to force SAS to use one history file or the other: (a) set SAS_ATTITUDE environment variable to one or the other file (b) delete the ATS file and re-run the odfingest task. Astronomical coordinates/Sky projections: Astronomical coordinates in imaging products are provided by SAS in the spherical gnomonic (TAN) projection. For regions far (≥ few degrees) from the reference zenithal attitude, the astrometry based on these projected coordinates is not reliable. Similarly, SAS tasks dealing with RGS source spectral products use some approximations that are not very accurate for large angular offsets. Therefore, in order to avoid the inaccuracies introduced by these approximations large angular fields should be avoided when dealing with single sources. In practical terms, the data should be split into "small fields"; such small fields may be selected either via sky region selections or via "good time intervals" (GTI) filtering. The second option might seem somewhat more cumbersome, but it is more consistent with the SAS processing scheme. RGS Data processing RGS source spectra are better produced if a few parameters of the rgsproc task are not set to their default values:
rgsproc withsrc=yes srcra=Src_RA srcdec=Src_Dec finalstage=fluxing rejflags='' \
xpsfincl=100 withpointingcolumn=yes rebin=100 rmfbins=500
MOS Data processing requires some data manipulation not supported yet by the current public SAS version. The first issue is to recast the 203x201 MOS pixels onto the original 610x602 pixel grid, including some randomization. In addition, some parameters of some MOS tasks must be set to non-default values: emenergy rejectbad3e4=NO detectbadoffset=NO Finally, some MOS CCFs are likely to require update to account for the new mode. OM processing tools are designed for fixed pointing observations and therefore cannot be applied to the slew data, even in fast mode. The data obtained with OM in this test correspond to three sky bands of only a few arcsec width. An analysis of the raw data allows us to "see" the passage of several stars through the fast windows. The estimated detection sensitivity (2 sigma) in the UVM2 (2310 nm) filter is 4.4 e-14 erg/cm2/s/A, which corresponds to a 15 mag B0 star.
Radiation backgroundEven though a high background level is always a concern, it becomes critical for the short effective exposure times in slew exposures. The behaviour of the background radiation during the whole revolution can be appreciated from Figure 2. Only the very last exposure was so severely affected by radiation as to suffer telemetry problems. This was not unexpected since it was scheduled out of the low radiation window predicted by the current radiation environment model. There were two other slew exposures (9124200027 and 9124200029) for which the radiation was significantly higher than in the remaining twelve. This leaves two sets of low radiation slew exposures: the first set with eight overlapping slews and the second set with only three. The average background in EPIC-pn during the first set of exposures was 5x10-6counts/s/square arcsec for the energy range [0.3-10]keV. Figure 2: The top panel of the figure shows the slew rate of XMM-Newton along revolution 1242. Slew exposures are identified by their slew id in blue. Red dashed vertical lines encompass the low radiation period according to the predictions of the operational environmental radiation model. The light curves of the count rates in EPIC-pn and RGS1 are shown in the middle and bottom panels, respectively.
Effective exposure timesThe total effective exposure time for a given source depends on the instrument (field of view and vignetting), the slew rate and the overlap between adjacent slews. Typical values achieved in this test for the different instruments are shown in Table 2.
An example of a typical EPIC-pn exposure map is shown in Figure 3. The two exposure maps correspond to the same region on the sky: On the left side only one exposure is taken into account and on the right side the exposure map is built up from all (three) exposures overlapping this region. Probably the first remarkable feature is the presence of dark parallel stripes, more noticeably in the exposure map of a single slew/pass. These lines are chip columns flagged as bad and inter-CCD gaps, in particular the gaps between the long sides of adjacent chips, and appear as stripes in the exposure map because they are parallel to the direction of the slew. When the exposure maps of individual slews are added up to form the total exposure map (a) more stripes appear since all slews are made along the same direction but (b) the exposure in every single stripe is longer because it is very unlikely that stripes from different slews match exactly. For the particular case of the test performed in revolution 1242, the exposure map on the right side of Fig. 3 indicates that 15 arcmin overlap is probably too short: there is a strip between -1 and +1 arcmin along the X-axis that, in practice, has been sampled in only one slew. A closer overlap is needed to make sure every region is sampled at least twice in the survey.
Figure 3: Example of typical exposure maps achieved during the slow slew survey test. Both exposure maps correspond to the same region on the sky. On the left side only one exposure is taken into account and on the right side the exposure map is built up from all (three) exposures overlapping this region.
Figure 4: Same as fig. 3, but for MOS2 The presence of the stripes indicates some degree of inhomegeinity in the final exposure map. One way to reduce that inhomogenity would be make the slews through "random" directions. However, it must be recalled here that it is necessary to use a particular "good" slewing direction to ensure its accuracy and the proper coverage of the sky region to be surveyed.
Detection limitsThe background levels mentioned above, together with the short effective exposure times, may lead to some problems in the source detection process. The total number of background counts expected within the typical extraction regions for point-like sources (∼arcmin2) may well be below 1, but this does not mean that the background is null. Simulations based on EPIC-pn data and the analysis of the results from the emldetect task applied to MOS data point to a lower limit of 10 net counts for the source to be reliable at these background levels. For the effective exposure times shown above the combination of the two MOS and the EPIC-pn cameras should give a flux limit of 2×10-13 erg/s/cm2 for the soft band and 1.3×10-12 erg/s/cm2 for the hard band. This is three times deeper than the 'fast slew' catalogue (XMMSL1)
On the high flux side, the slow slew survey makes little difference to the pile-up limit relative to pointed observations. At a slew rate of 30 arcsec/s the time to cross one EPIC-pn pixel (0.14 s) is still signicantly longer (2x) than the chip(s) frametime (0.073 s); in other words, sources move by less than one pixel during a single readout frame time. The source detection process is commonly associated with images coming out of the EPIC cameras. Source products from RGS spectrographs are generally obtained from sources previously detected in the EPIC cameras. It might be worth remarking that during this test there was a case where a source was clearly detected in RGS data, but not seen in simultanous EPIC images. The case is readily seen in Figure 2
Astrometric accuracyThere are ∼20 EPIC-pn sources detected with more than 10 counts in consecutive slews. The consistency of the coordinates obtained for these ∼20 pairs is within a few arcsec (2.5±1.6′′) There are 44 sources in the final source list for which a reasonable Simbad counterpart exists. The median distance between the Simbad coordinates and the coordinates in the SSS is 7.8′′. This median distance is reduced to 3.6′′ when it is computed for those sources whose Simbad counterpart has a measure of its optical magnitude (indirect way of selecting optical coordinates). It is also worth remarking that the median distance increases to 15.6′′ when considering only Simbad counterparts having in their names the RXS identification of ROSAT X-ray sources. Such a larger value is most likely due to the poorer accuracy of the ROSAT coordinates. In summary, the accuracy of the asbolute astrometry in this slow slew test is better than 4′′. Appendix: EPIC-pn source list
Authors: P. Rodriguez Pascual, R. Gonzalez Riestra & R. Saxton |
| Copyright 2012© European Space Agency. All rights reserved. |
