Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/docs/documents/CAL-TN-0081.pdf Дата изменения: Thu Aug 7 12:38:34 2008 Дата индексирования: Mon Oct 1 20:45:00 2012 Кодировка:

MPE Report Vol. 288, 159
c EPIC Consortium Meeting Ringberg 2005

Comparison of EPIC-pn ground-based and in-orbit calibration measurements
M. J. Freyberg1 , W. Burkert1 , G. Hartner1 , M. G. F. Kirsch2 , and E. Kendziorra3
1 2 3

Max-Planck-Institut fur extraterrestrische Physik, Garching, Germany Ё ESAC, ESA, Villafranca, Spain Institut fur Astronomie und Astrophysik, UniversitЁ Tubingen, Tubingen, Germany Ё at Ё Ё

1. Intro duction
The EPIC-pn CCD camera aboard XMMNewton has been extensively calibrated on ground as well as in orbit. While the formal flight spare model (FS) is in operation on XMM-Newton, the flight model (FM1) is situated at the Panter X-ray Test Facility. This allows additional tests of the detector performance and comparison of ground calibrations prior to launch with present measurements. Additionally, tests can be performed without loss of scientific XMM-Newton observing time. Here we present recent results on a comparison of the internal camera background at low energies in various instrument read-out modes, on the precise determination of frame times and EPIC-pn oscillator frequency, and on the determination of charge losses during fast shifts (read-out of EPIC-pn window modes).

2. EPIC-pn low-energy background
For various instrument modes the low-energy background spectra and intensity levels have been determined for EPIC-pn FM1 (PANTER) and FS cameras (XMM-Newton, closed filter observations). Figure 1 summarizes these results for FM1 (left) and FS (right), with spectra (top) and intensities (bottom) for each read-out mode. Spectra for each camera are comparable in shape. The in-orbit spectra (FS, after MIP rejection) are steeper than for FM1 below 30 adu (1 adu 5 eV). Low-energy background intensiSend offprint requests to : M. J. Freyberg, e-mail: mjf@mpe.mpg.de

ties are closely related to the number of readouts per second and thus the background is dominated by electronic read-out noise. A simple 2-parameter model has been fit to the standard imaging modes (labelled in blue). On ground longer integration times (extended fullframe mode with longer wait time between quadrant read-outs) are available than in orbit, the lower threshold can also be set to lower values (16 rather than 20 adu). For soft and faint diffuse sources the extended full-frame modes are therefore best-suited. For sources that have their emission maximum below the lower event threshold pile-up of X-rays with background could lead to event amplitudes above the lower threshold and thus mimic a low-energy excess. To investigate this further, a special short calibration measurement has recently been performed with a lower threshold of 16 adu (NRCO-49, 0979 9097900002 PNU002) to determine the low-energy background below the default event threshold of 20 adu. A comparison (see upper right panel in Fig. 1) of FS (filled red hexagons) and FM1 cameras (empty black hexagons) shows differences in the detailed spectral shape, however, at 16 adu the intensity levels agree.

3. EPIC-pn frame times and oscillator
EPIC-pn frame times for each instrument submode can be precisely determined independent of any oscillator properties using the integer-valued time stamps provided in the PNAUX1 FITS extension, FTCOARSE (full seconds) and FTFINE (sub-seconds). The internal EPIC-pn oscillator has a nominal frequency


160

M.J. Freyberg et al.: EPIC-pn calibration

Fig. 1. Top left: internal background spectra of various instrument modes at the focal position for the FM1 camera at the Panter X-ray test facility. Bottom left: background intensity in the 16 - 40 adu energy range for FM1 for various instrument modes. The blue dotted line is a two-parameter fit using the standard imaging modes only (SW, LW, FF, eFF3 , eFF5 ; blue labels). Right: similarly for FS camera aboard XMMNewton for the 20 - 40 adu energy range. In the upper right panel a SW mode observation with the FS camera with lower threshold 16 adu is shown (filled red hexagons, NRCO-49) along with the corresponding spectrum for the FM1 camera (empty black hexagons). For details see text. Table 1. EPIC-pn frame time analysis using the PNAUX1 file for all read-out modes: the given nominal clock numbers (column 2) have been obtained from the nominal frame times (CCF values) multiplied by fpn = 25 MHz; the quantity M1 is the average measured distance between two events in two consecutive frames within the same second in FTFINE units; multiplied by 512 it gives the number of clocks of the internal pn-oscillator for each read-out mode. The derived values differ from the values in the EPN TIMECORR CCFs. In the columns at the right-hand-side we give recommended values for the length of each read-out cycle as well as for the frame time assuming a fpn = 25 MHz internal pn-oscillator frequency.
Mode FF eFF5 eFF3 LW SW TI BU clocks 1834108 7076988 4979836 1191616 141794 149116 108614 Observation 0078 0124100101 0044 0119710201 0469 0108260201 0537 0136540701 0908 0158961001 0807 0158971201 0411 0153750501 PNS003 PNS008 PNS003 PNS008 PNS013 PNS003 PNS001 512 M1 1834123.988 7077004.295 4979852.043 1191595.994 141795.019 149115.990 108612.031 +16 +16 +16 -20 +1 0 -2 recommendation 1834124 clocks 73.36496 7077004 clocks 283.08016 4979852 clocks 199.19408 1191596 clocks 47.66384 141795 clocks 5.67180 149116 clocks 5.96464 108612 clocks 4.34448 ms ms ms ms ms ms ms

of fpn = 25 MHz, but the exact value is unknown and sub ject of this analysis. The CCD sequencer and the FTFINE counter are both clocked by this oscillator. An external oscillator triggers the increment of FTCOARSE and the reset of FTFINE. aux 512/fpn is the length of

a FTFINE unit, nominally aux = 20.48 µs . As the mode sequences are not an integer multiple of 512 clocks the time stamps in the PNAUX1 file show a jitter of 1 FTFINE unit. For times of two events we have: T (n) = Tcoarse (n) 1s + Tfine (n) aux


M.J. Freyberg et al.: EPIC-pn calibration

161

Fig. 2. Total difference of the effective frequency of internal EPIC-pn oscillator (combined with external
OBT oscillator effects) from the nominal value of fpn = 25 MHz as function of time and instrument mode. External events like eclipse seasons or RGS switch offs are indicated; dotted vertical lines show CDMU re-syncs. The long-term increase of the frequency is probably due to external temperature effects or to ageing the EPIC-pn oscillator. There is no apparent mode dependency.

T (n + 1) = Tcoarse (n + 1) 1s + Tfine (n + 1) aux . The time difference is then T (n + 1) - T (n) = [Tcoarse (n + 1) - Tcoarse (n)] 1s + [Tfine (n + 1) - Tfine (n)] aux . If these two events are within the same second and in two consecutive read-out cycles this reduces to: Tframe = 0s + [Tfine (n + 1) - Tfine (n)] aux Tfine (n + 1) - Tfine (n) = Tframe /aux . When we average this (integer) quantity over the total exposure we can discard the jitter effect and the real value M1 is M1 := Tfine (n + 1) - Tfine (n) = Tframe /aux . If we now consider the similar case of 2 events separated by 1 frame time but for events with full-second increment in between, we get: Tframe = 1s + [Tfine (n + 1) - Tfine (n)] aux Tfine (n) - Tfine (n + 1) = (1s - Tframe )/aux and to discard the jitter effect by averaging over all such occurrences within an

exposure M2 := Tfine (n) - Tfine (n + 1) = (1s - Tframe )/aux . From these quantities M1 and M2 (in FTFINE units or 512 oscillator units) one can derive the effective oscillator periods and frame times via aux = 1s/(M1 + M2 ) and Tframe = 1s M1 /(M1 + M2 ). Table 1 illustrates the results of the frame time analysis for representative exposures. The differences in units of the 25 MHz quartz clocks compared to the CCF values are negligible for exposure times (e.g., 1 clock of 40 ns for SW mode per frame). However, the corrected numbers help to identify time jumps in the data over longer time spans ­ as deviations from integer number of frame times in time differences of two consecutive events can be detected more reliably. While the values of M1 are constant with time, the M2 values contain an explicit relation to the EPIC-pn oscillator and implicitly to the onboard time (OBT) oscillator as trig-


162

M.J. Freyberg et al.: EPIC-pn calibration

Fig. 3. Evolution of EPIC-pn quadrant box temperature for quadrant Q1 (which includes the target
CCD) for FF mode, averaged over an exposure. The temperature sensor has only a coarse sampling (1 step = 0.88 C), for shorter exposures only one value during an exposure is realized. The ordinate has been intentionally inverted, higher temperatures are toward the bottom. When comparing with Fig. 2 a similarity in terms of general trend is suggestive: a decreasing quadrant box temperature increases the frequency of the oscillator (which is not temperature stabilised).

ger of the full-second increment. Figure 2 shows the evolution of the effective EPIC-pn oscillator frequency with time expressed in XMM-Newton revolutions since launch, relative to the nominal 25 MHz frequency; the effect is not cumulative as after 1 second a reset is triggered externally. This plot does not use any time correlation such as corrections for OBT oscillator drifts (ageing, temperature). Kirsch et al. (2004) show that the OBT oscillator frequency decreases with time by only about 1 Hz per 30 revolutions; extrapolated to launch the deviation from nominal frequency (223 = 8388608 Hz) was already -44.5 Hz, i.e. -5.3 10-6 ; after 1000 revolutions a total additional change of -25 Hz (i.e. -3 10-6 ) is expected. It is unknown yet how much temperature changes in the EPIC-pn environment influence the EPIC-pn oscillator. The corresponding total change of the effective EPIC-pn frequency is about +180 Hz (i.e. +7.2 10-6 ). It is therefore likely that the dominating effect of the effective

frequency shift is due to the EPIC-pn oscillator (e.g., ageing, temperature). In Fig. 2 there are external influences visible, like switch-off of RGS at revolutions 110 - 180 or CDMU resets. The external OBT effects are corrected for using time correlation analysis. It is known that the EPIC-pn quadrant box temperatures generally decrease with time, which is believed to be due to cooler satellite environment caused by reduced reflected emission from Earth at perigee. Figure 3 shows the temperature of quadrant box Q1 averaged over an exposure from launch until present, with inverted ordinate. The temperature is decreasing (with a slope of about 3.2 mK per revolution for revolutions > 150); there is an apparent similarity to the trend in Fig. 2. There is at the moment no tool in XMMSAS to correct for EPIC-pn internal oscillator drifts; note that this is not really needed for normal timing analysis as this small effect is limited to times within a second.


M.J. Freyberg et al.: EPIC-pn calibration

163

Fig. 4. Right: design of pinhole measurements at the Panter X-ray test facility: 3 pinholes with different
sizes (4, 2, 1 mm 1.83, 0.92, 0.46 arcmin diameter) disjunct in RAWX and RAWY coordinates are moved relative to the FS camera into positions characteristic for various imaging modes. Each colour represents a certain position. The read-out is toward the bottom. Top row: intensity images for these 6 positions for CCD 4 in FF mode. Bottom row: as above but for LW mode; the green circles show the positions of the pinholes in front of the CCD.

Fig. 5. Left: pinhole measurements for the FM1 camera at Al-K for various instrument modes. Right:
comparison of the fast shift CTI contribution for FS (lines: CCF models) and FM1 (dots: Panter measurements).

4. EPIC-pn window mo de CTI correction
The SW and LW modes are a integration period, a fast shift tion window to the read-out no read-out like in FF mode, and sequence of an of the integrade, then a slow finally another

(erase) fast shift of the window. Times are for LW mode for integration of 100-row window (45.14 ms), fast shift of window area toward CAMEX (0.072 ms, twice), and read-out as in full frame mode (2.45 ms). Charge transfer losses are consequently attributed to fast shift CTI and


164

M.J. Freyberg et al.: EPIC-pn calibration

to normal (slow) shift CTI. Using pinhole masks on ground dedicated areas of the CCD can be separately illuminated with monochromatic Xrays (for details see Freyberg et al. (2005)). The resulting peak positions (corrected for gain column variations only) can then be used to determine the CTI parameters. A pinhole mask (2 mm Al) was put into the X-ray beam at the Panter facility in front of the EPIC-pn FM1 camera. Three holes with 1 mm, 2 mm, 4 mm diameters (for a focal length of 7.5 m this relates to 27.5 , 55 , 110 ) were drilled disjointly in RAWX and RAWY pro jections so that they can be considered as independent. The EPIC-pn camera was then moved up and down behind the fixed mask to avoid any possible beam effects (like energy dispersion of monochromator). In Fig. 4 (left) we show images of FF and LW modes. The green circles in LW mode mark the positions of the pinholes above CCD 4, similarly to the FF mode. On the right-hand side the various pinhole positions are indicated with different colours. With various line positions (corrected for column gain variations, in adu) of monochromatic input as function of position for various instrument modes we can derive the CTI losses for fast shifts and slow read-out. For LW mode, e.g., we have for CTE = 1 - CTI: Y > 100 : Y 100 PHA(Y ) = PHA0 в CTEfast в CTEslo-100 w Y < 100 : Y PHA(Y ) = PHA0 в CTEslow This leads to the relation 100 PHA(Y )/PHA(Y - 100) = CTEfast and therefore 1/100 CTEfast = (PHA(Y )/PHA(Y - 100)) In window modes the peak positions show characteristic steps which are related to the difference between fast and slow shift losses (see Fig. 5 (left)). The gain (as determined from the amplitude value at RAWY = 0) seems to be slightly mode dependent for the FM1 camera. Figure 5 (right)) compares CCF values for the fast shift CTI contribution for the FS camera with ground measurements for the FM1 camera. The pinhole procedure has been successfully applied to the FM1 camera model. As the operating conditions as well as detailed CCD properties are different, an in-orbit verification measurement is necessary. An extended multi-line target with intensities below pile-up is wellsuited, such an example is N132D. Such a verification measurement was performed but background conditions were very high and thus spec-

tral parameters (i.e. line shifts) could not be derived (the verification observation was repeated after this meeting, NRCO-47).

5. Conclusions
The EPIC-pn low-energy background is dominated by electronic read-out noise. Effects of possible sub-threshold pile-up will further be investigated (NRCO-49). Nominal EPIC-pn frame times in the CCFs (EPN TIMECORR) will be revised; this will hopefully lead to significant improvements in the detection of time anomalies by the XMMSAS OAL library. The EPIC-pn fast shift CTI correction will be further analysed using an in-orbit calibration measurement (NRCO-47).
Acknow ledgements. The XMM-Newton pro ject is supported by the Bundesministerium fur Bildung Ё und Forschung / Deutsches Zentrum fur LuftЁ und Raumfahrt (BMBF/DLR), the Max-PlanckGesellschaft (MPG) and the Heidenhain-Stiftung.

References
Freyberg, M.J., Burkert, W., Hartner, G., Kirsch, M. 2005, Calibration Report, ftp://epic3.xra.le.ac.uk/pub/cal-pv/ meetings/mallorca-2005-02/mf fastshift cti.pdf Kirsch, M.G.F., Becker, W., Benlloch-Garcia, S., Jansen, F.A., Kendziorra, E., Kuster, M., Lammers, U., Pollock, A.M.T., Possanzini, F., Serpell, E., Talavera, A. 2004, Proc. SPIE 5165, 85, http://xmm.vilspa.esa.es/docs/ documents/CAL-TN-0045-1-0.pdf Note added: In Tab.1 a typesetting error of printed version has been corrected, with proper recommended frame time value for SW mode of 5.67180 ms, all other entries been correct (especially the number of clo i.e. 141795 clocks of 40 ns each). the the the had cks,