XMM-Newton Science Analysis System: User Guide

4.5.1 How to analyse a piled-up Timing mode observation

If a source count rate is greater than $\sim800~$counts/s for a pn Timing mode observation or $\sim100~$counts/s for MOS Timing mode then the source events are likely to be affected by photon pile-up (Note that in terms of flux, count rates for a given incident flux are higher for soft than for hard spectra; check Section 3.3.9 of the XMM-Newton User Handbook[3] for details).

Users can check if a given EPIC source is affected by pile-up by running the SAS task epatplot. Please, refer to the pile-up thread for advise on how to quantify the level of pile-up. A specific caveat applies to EPIC-pn Timing Mode exposures: the source which centroid of it is located at RAWY=190 or 191 could suffer from a possible systematic offset of a few percent which could be seen when comparing the distributions of observed and expected event shapes (the source position is contained in the keyword SRCPOS in the header of the EVENTS extension of the event list). The shift is in the positive (negative) direction for single (double) events. This shift is not due to pile-up. It is rather due to residual calibration uncertainties in the pattern fraction distribution as a function of energy in sources close to the first micro-pixel border. The effect is discussed in Guainazzi et al. 2009, CAL-SRN-0265. Users do not need to take any action to correct for this effect, because the calibration of these exposures is nominal.

If a source observed in EPIC fast mode (EPIC-pn Burst/Timing; EPIC-MOS Timing) is affected by pile-up, the boresight column must at least be excised from the region used to accumulate scientific products (this is analogous to removing a circle surrounding the PSF centroid in sources affected by pile-up in imaging modes). If the remaining event list is still affected by pile-up, one must increase the number of excised columns, until the resulting event list is pile-up free.

Users must use the specific procedure outlined below to calculate the effective area associated to a spectrum extracted from a box, from which one or more columns have been excised. The procedure is vald for EPIC-pn and EPIC-MOS. Let us assume that a pile-up free region in a EPIC-MOS Timing Mode exposure is defined by the following selectlib expression:

( RAWX in [280:298] || RAWX in [303:321] )

The BACKSCAL value can be calculated for the created spectrum in the normal manner using xmmselect or the backscale task. Similarly the tasks rmfgen and arfgen may be used to create the redistribution matrix (RMF) and ancillary response function (ARF) in the normal manner (see § 4.8.2).

For example, to produce the ARF where the 5 central columns (the boresight plus two columns at each side) are removed:

1. Produce a spectrum (spec_outer.ds)

   evselect table=EPIC_TimingEvts.ds withspectrumset=yes spectrumset=spec_outer.ds energycolumn=PI spectralbinsize=5 withspecranges=yes specchannelmin=0 specchannelm ax=11999 expression='(FLAG==0) && (PATTERN<=0) && ( RAWX in [280:298] || RAWX in [304:321] )'

2. Produce an ARF

   arfgen spectrumset=spec_outer.ds arfset=arf_outer.ds detmaptype=psf

The arf_outer.ds file can then be used as the ARF for fitting the spectrum created from the region:

( RAWX in [280:298] || RAWX in [304:321] )

together with the RMF produced earlier.

Note: Users working with version earlier than SAS14.0, must use the procedure outlined below to calculate the effective area associated to a spectrum extracted from a box, from which one or more columns have been excised. The calculation of the ARF is a little more complicated because the encircled energy (PSF) correction is not performed correctly by arfgen in earlier SAS versions. To overcome this problem the following steps need to be taken.

1. Produce a spectrum (spec_full.ds) from the full area without excluding the central columns, e.g. ( RAWX in [280:321] )

2. Produce a spectrum (spec_inner.ds) from the excluded region, e.g.

   ( RAWX in [299:303] )

3. Produce an ARF for both of these files

   arfgen spectrumset=spec_full.ds arfset=arf_full.ds detmaptype=psf
   arfgen spectrumset=spec_inner.ds arfset=arf_inner.ds detmaptype=psf

4. Subtract the inner ARF from the total ARF by using the ftool:

   addarf "arf_full.ds arf_inner.ds" "1.0 -1.0" arf_outer.ds

The arf_outer.ds file can then be used as the ARF for fitting the spectrum created from the region:

( RAWX in [280:298] || RAWX in [304:321] )

together with the RMF produced earlier.