Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://xmm.vilspa.esa.es/docs/documents/CAL-SRN-0005-1-0.ps.gz Äàòà èçìåíåíèÿ: Wed Jan 3 15:10:22 2001 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 21:29:51 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: http astrokuban.info astrokuban

XMM­Newton CCF Release Note
XMM­CCF­REL­5
EPIC Energy Scale
D Lumb
October 3, 2000
1 CCF components
Name of CCF VALDATE List of Blocks
changed
CAL VERSION XSCS flag
EMOS1 ADUCONV 0008 2000­01­01T00:00:00 ADUCOEFF OFF­
SET GAIN EN­
ERGY COMBINE
YES
EMOS2 ADUCONV 0008 2000­01­01T00:00:00 ADUCOEFF OFF­
SET GAIN EN­
ERGY COMBINE
YES
EPN ADUCONV 0009 2000­01­01T00:00:00 ADUCOEFF
OFFSET GAIN
GAIN HIGH
REEMISSION
REEMISSION1
YES
EMOS1 CTI 0005 2000­01­01T00:00:00 CTI
CTI EXTENDED
NO
EMOS2 CTI 0005 2000­01­01T00:00:00 CTI
CTI EXTENDED
NO
EPN CTI 0008 2000­01­01T00:00:00 CTI
CTI EXTENDED
HOT PIXELS DIS­
CARDED PIXELS
CTI­HIGH
NO
2 Changes
First release
A number of factors are involved in determining the photon's energy. The intrinsic silicon
1

XMM­Newton CCF Release XMM­CCF­REL­5 Page: 2
photoelectric conversion process, the losses of charge during transfer to the output node, the on­
and off­chip amplifications and the event processing. As far as possible we aim to keep a physical
representation of all these items within the calibration files.
Although it is not the recommended practice for spectral fitting, we adopt a PI channel (i.e.
corrected apparent energy) domain analysis. This eases considerably the co­addition of data from
multiple CCDs, while the oversampling of energy resolution is probably sufficient to allow this
approach in practice.
The silicon conversion process is inherently linear, although some theoretical models predict
departure from linearity, especially at low energies and near the Si K absorption edge (1.84 keV).
Such features may be masked by other processes though.
Charge transfer losses are in principle measured, calibrated and monitored by using an internal
calibration source. This has Mn K ff and fi lines of 5.89 and 6.4 keV, with a fluorescent Alk target line
of 1.49 keV. Other much weaker fluorescent contaminants can also be detected. However all these
lines do not provide good leverage on energy­dependent CTI effects for low signal packet magnitudes.
In addition the calibration source flux is relatively large compared with most astronomical image
scenes, thus a systematic discrepancy due to a count­rate dependency might be encountered.
For the PN camera, the charge loss per transfer is significantly higher. Fortunately, an extremely
detailed model has been developed to account for energy and count rate dependence. This means
that single pixel events which are the first read out of a frame, can be corrected with the same fidelity
as the MOS camera. To the accuracy we can determine there has been (nor was there expected to
be) no degradation since launch. Furthermore, the response generation also has (at present) high
fidelity only for ''first singles'' events. These represent about 70% of all X­ray data, so the situation
is thought to be acceptable for the first calibration.
In the EPIC data­handling units, events are selected for transmission. A common occurrence is
the splitting of events between pixels. The energy determination is degraded by a combination of
noise quadrature summation and sub­threshold charge loss. The amount of charge splitting itself
is energy dependent. These effects are incorporated in the response matrix, however the on­board
and SAS processing makes the best attempt to reconstruct true energy in the first place. As part of
the in­orbit verification process, distenangling the effects of gain and response matrix is somewhat
complicated, and maintaining coherence between event processing algorithms, and their resulting
response distributions is not yet complete.
3 Scientific Impact of this Update
First release

XMM­Newton CCF Release XMM­CCF­REL­5 Page: 3
4 Estimated Scientific Quality
For the MOS camera, the pre­lauch CTI was good enough that a typical charge loss from CCD
edge to an on­axis target location was only 3--5 eV, largely correctable. Since launch, degradation
has been about this amount again, so that this portion of the energy determination budget is well
in hand. However the Small change in CTI with time is included in the CTI calibration files as
an avergae change over all CCDs, because the individual error measurement on the small change is
larger than the change itself.
The association of measured ADC value to an energy should be a linear translation due to
amplifier gain. However there are still episodic discrepancies at –5eV, which can be due to a
number of causes. A readout mode dependence has still to be ruled out, for we are only just starting
to gather enough data on modes other than full­frame with the calibration sources. The effect of
light loading can be present. For example if a blocking filter is chosen which is too thin, each optical
photon detected in a pixel produces a 3.6eV energy equivalent offset. In the PN camera this is in
principle calculated on a pixel­by­pixel basis, though associated noise must also be considered. In
the MOS a row/column average value is subtracted, so that local deviations can be considerable.
Such offsets can be checked for in the raw data files by histogramming E3 and E4 values around
the target of interest, and checking for gross deviations from zero. (of course the PSF for optical
photons is comparable with X­ray PSF's, so energy discrepancy will be significantly greater for the
pixels at the PSF core).
However, having laid bare all these caveats, the current experience is that energy determination
is generally good to about 5eV. Use of low energy SNR emission lines in principle are able to
supplement this knowledge, but plasma emission code uncertainties and scientific interpretation
start to play. In case of suspicious data, it is suggested that the user can double check gain values
by looking in the raw data for optical contamination, AND by checking parasitic emission lines in
the background (e.g. Al K (1.487keV) and Cu Kff (8.048 keV) are relatively uniform in the focal
plane and in long exposures should be centroided with reasonable accuracy).
For MOS Timing Mode, the gain was incorrect in hardware from mission start until ca. Septem­
ber 2000. Currently it is 10% incorrect.
5 Expected Updates
For EPIC MOS, the revision of the rmfgen code to match the low energy response may need some
modification of the low energy gain determination. This may be performed by reconciling the low
energy RGS emission line spectra with MOS.
When the MOS timing mode is properly calibrated the files will be updated.
For PN there will be changes due to the further analysis of mode and count rate dependence.
Only the FULL FRAME PN camera mode is properly calibrated, and future updates will address
the mode dependence. On ground this calibration for full frame mode required millions of photons
at a wide range of energies. In­orbit for the other modes we are restricted to the limited emission

XMM­Newton CCF Release XMM­CCF­REL­5 Page: 4
line set of the in­flight calibration source, which is necessarily faint.