Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/docs/documents/CAL-TN-0019-1-0.ps.gz Дата изменения: Tue Jul 2 15:23:57 2002 Дата индексирования: Sun Apr 10 02:33:40 2016 Кодировка: Поисковые слова: zodiacal light

OM Calibration Status
Bing Chen with inputs from OM calibration team
please send any comments to bchen@xmm.vilspa.esa.es
ABSTRACT
This document re ects the status of the calibration of the OM instrument which is im-
plemented in SAS 5.3.3. Using data in SA95 eld, we demonstrate that a photometric
accuracy of 3% can be achieved with the OM. The astrometric accuracy is about 1 arcsec.
OM grism calibration is on-going.
Contents:
1) Introduction ................ 2
2) Throughput and Zeropoints ................ 2
3) Coincidence Loss ................ 3
4) Color Transformation ................ 3
4.1) UBV color transformation ................ 4
4.2)UV color transformation and Flux calibration ................ 5
4.3) Testing OM photometry with data in SA95 eld ................ 5
5) Point Spread Function ................ 9
6) Distortion ................ 9
7)Large Scale Sensitivity Variation ................ 9
8)Background ................ 10
9)Flat elds and dark frames ................ 14
9.1) Dark count rate ................ 14
9.2) Flat elds ................ 14
10) Straylight ................ 14
11)Fixed patterning ................ 14
12) Fast mode ................ 15
13) Grisms ................ 15
1

1 Introduction
The Optical Monitor extends the spectral coverage of the XMM-Newton observatory into
the UV and optical, thus opens the possibility to test models against data over a broad
energy band and makes an important contribution to XMM's science. It can provide
arcsecond resolution imaging of the whole eld of view and can simultaneously zoom
into a small area to provide timing information. Seven broad band lters allow colour
discrimination, and there are two grisms, one in the UV and one in the optical, to provide
low resolution spectroscopy. More detailed information about the instrument may be
found in Mason et al. (2001), and about OM science in Mason (2002). In this article, we
summarize the calibration activities and report results obtained by the OM calibration
team so far. This document re ects the status of the calibration of the OM instrument
that is implemented in SAS 5.3.3.
2 Throughput and zeropoints
The OM U, B, V throughputs and zero points are determined by a two-step approxima-
tion.
I): We observed spectrophotometric standard white dwarf stars and compared the
observed counts with pre- ight values to calculate the correction factors known as 'fudge
factors' for each lter. These discrete points (see Fig. 1) were tted with a polynomial
constrained to follow the form of the contamination curve, to give a response correction
curve. This correction curve, when multiplied with the OM pre- ight response, gives the
in- ight response.
The Zeropoint is the connection between the units found in your OM image (counts)
and the astrophysically important quantities of ux or magnitudes. The Zeropoint of the
OM is based on simulations of the Vega count rates. We have chosen the zero points such
that the OM ubv synthetic magnitudes for Vega match the observed UBV magnitudes,
respectively. For observed Vega magnitudes we have adopted the values 0.025, 0.03 and
0.03 for U, B and V lters. Since the UV lters do not match any existing photometric
system, the zero points can in principle be set to any values. Current zero points have
been dened in this way for Vega, UVW1 vega =UVM2 vega =UVW2 vega =0.025. With this
denition, we can derive the zero points of the OM by using equation
Zeropoint = 2:5log10(cts) +m vega (1)
II): The derived throughputs and the zeropoints can have few percent errors. First,
although these white dwarfs and Vega stars are spectrophotometric standards, there are
still uncertainties at a few percent level in the spectrophotometry. Another problem is
that response curves are derived by the least square polynomial t to the six points where
we have got the fudge factors, this t procedure can also produce few percent errors. Since
we have got both the ground-based and OM observations at two elds, we can revise our
zero-points based on several hundreds of common stars. Initially we used the zero points
derived above to calculate the transformation. This provides non-zero zero order terms in
the transformations. These terms represent corrections to the zero points which should be
made to switch from the simulated Vega-based values to the values based on observations.
We derive the zeropoint corrections 0.0873, 0.0615 and 0.0257 mag for V, B,and U lters,
respectively. Finally, we have the OM zeropoints given in Table 1. Once we have got
the corrected U, B and V zeropoints, we can update the fudge factor curve and response
curve of the OM. In Figure 2, we show the post-launch throughput curve.
2

Figure 1: The least square polynomial t (the dotted line ) to the measure fudge factors (the squares).
The fudge factor is the ratio of the in- ight count rate to the pre- ight predicted count rate.
Table 1: Zeropoints of OM
V B U UVW1 UVM2 UVW2 white
Zeropoints 17.9633 19.2661 18.2593 17.2965 15.8098 14.8615 20.2555
3 Coincidence Loss
The loss of counts through coincidence occurs whenever more than one photon arrives at
the CCD in the same place within the same frame. For the 20th magnitude stars, OM
coincidence losses are negligible. Losses become signicant for a point source at a count
rate of about 10 counts s 1 (about 10% coincidence). Coincidence losses for point sources
may be measured empirically if one observes the same standards with the OM as well as
from the ground. We have compared the in- ight and ground measurements of  400
stars to derive an empirical correction to this coincidence loss. This empirical curve has
been incorporated into the SAS.
4 Color Transformation
Originally the transformations from the OM instrumental system to Johnson's system
were based on simulations. Spectra from Bruzual-Persson-Gunn-Stryker were folded with
3

Figure 2: The current optical Monitor eective area curves for each of the lters
the response curves of the OM and with Johnson's UBV system to simulate the instru-
mental and standard colours.
To establish colour transformations to the standard UBV system based on real ob-
servations, several elds have been observed from the ground with the standard Johnson
UBV lters and with the XMM-Newtom OM camera. 363 cross-identied stars have been
used to make the color transformation.
4.1 UBV color transformation
We have derived empirical colour transformations from dedicated, ground-based photo-
metric observations of two calibration elds. The tting was limited to stars with less than
20% coincidence loss, and 5% statistical errors. To calculate the color-transformations, we
use the new zeropoints. Within the color intervals covered by the observational data, the
transformations may be described by a second order t. The transformation equations
are given as following. (The upper case letters denote the Johnson system; the lower case
letters the OM instrumental system).
B V = 0:0005 + 1:046  (b v) 0:023  (b v) 2 (2)
V v = 0:0008 0:006  (b v) 0:021  (b v) 2 (3)
B b = 0:0014 + 0:04  (b v) 0:044  (b v) 2 (4)
U B = 0:002 + 0:901  (u b) + 0:103  (u b) 2 (5)
B b = 0:0078 0:01  (u b) 0:019  (u b) 2 (6)
U u = 0:048 0:097  (u b) + 0:03  (u b) 2 (7)
U V = 0:053 + 0:91  (u v) + 0:033  (u v) 2 (8)
V v = 0:003 0:01  (u v) 0:004  (u v) 2 (9)
U u = 0:048 0:097  (u v) + 0:03  (u v) 2 (10)
We should point out that not all equations above have been implemented in the SAS.
For example, for deriving standard B magnitude, SAS uses (u-b) color transformation
4

(eq. 7), or making use of the UV lters (see next section). However, user can also
use (b-v) color transformation (e.q. 5) to get standard B magnitude. In Figure 3, we
plot the observed UBV color-transformation (the red dashed lines), together with the
simulated one (Crosses and solid lines). By comparing the observed and simulated color-
transformation, we can see that the agreements in (B-V, b-v) and (U-V, u-b) are excellent,
and the agreements in V-v and B-b are usually better than 2%. The main problem is for
the hot stars in the U lter, where the discrepancy can reach 10%. The spectral library
(BPGS) used in the simulations is combined from several sources. For the optical data,
the spectral atlas is based on Gunn and Stryker (1983) observations, which begins from
313 nm. Colina and Bohlin (1994) have demonstrated that the errors in normalization of
the spectra can produce several percent errors in the photometry. Other spectral libraries
will be used to investigate the discrepancy for the hot stars in the u lter.
4.2 UV color transformation and Flux calibration
Currently, the color-transformations for UV lters are based on the simulations because
we have not got enough calibration observations for UV lters. However a comparison be-
tween the simulated color-transformations and observational data shows a good agreement
(see Figure 6).
In Figure 4, we show the simulated color-color transformations and some of the simu-
lated UV color transformation equations are given as following,
B b = 0:0065 0:0085  (uvw1 b) 0:003  (uvw1 b) 2 (11)
V v = 0:0165 0:0059  (uvw1 v) 0:0008  (uvw1 v) 2 (12)
U u = 0:0094 0:0332  (uvw1 u) + 0:0233  (uvw1 u) 2 (13)
U B = 0:2386 + 0:5383  (uvw1 b) + 0:0157  (uvw1 b) 2 (14)
B V = 0:1446 + 0:2761  (uvw1 v) + 0:025  (uvw1 v) 2 (15)
Due to observing time constraints or due to bright source restrictions the optical colors
may not be available, then you can use UV color-transformation to derive the standard
UBV magnitudes and colors.
Users can convert OM countrate in the UV lters into uxes (expressed, e.g., in
erg/cm/cm/s/A) following a step-by-step recipe at XMM-Newton SAS watchout page:
http://xmm.vilspa.esa.es/external/xmm sw cal/sas frame.shtml.
4.3 Testing OM photometry with data in SA95 eld
A Landolt standard star eld (SA95 ) has been observed to test the OM photometric
accuracy. Since these standards have very high photometric accuracy ( 0.005 mag), the
standard deviation on the residuals between the Landolt magnitude and the magnitude
derived by OM provides a direct measurement of the OM photometric accuracy. The eld
has been observed with OM in rev. 407 in ve OM lters (u, b, v, uvw1, uvw2), and also
by Landolt (1992) and Stetson (2000) in Jonnson standard system. Since the magnitudes
reported by Stetson are more accurate, in the following comparison, we use the Stetson's
measurements ( U magnitude is not available).
In Figure 5, we plotted the dierence in standard magnitude between Stetson's mea-
surement and OM measurement as a function of Stetson standard magnitude. The error
5

Figure 3: The Observed (the red, dashed lines) and simulated (solid lines and symbols) color-
transformation for U, B and V lters.
6

Figure 4: The color-transformation based on simulation for UV lters.
7

Figure 5: The dierence in standard magnitude between Stetson's measurement and OM measurement
as a function of Stetson standard magnitude.
Figure 6: Comparing the simulated UV color-transformation with the data in the SA95 eld.
8

bars are the errors on OM instrumental magnitudes. From this gure, we found that the
OM photometric accuracies are 0.013 and 0.023 mag for B and V lters. There is a small
oset (  0.06 mag) in B lter. This needs to be conrmed from other observations.
In Figure 6, the simulated UV color transformations (the solid lines) are compared with
the data in the SA95 eld. The star with B-V  1.5 is very red. It is very faint (19.391)
in UVW1 and has very large error (1.199). This star should be neglected from gure
6. For other stars, we can see that the results are in agreement with the simulations. A
small oset between the observations and the simulations is probably due to the zeropoint
determination of UVW1 lter.
5 Point Spread Function
The XMM-OM point spread function (PSF) is radially symmetric in shape, and the
width increases with photon energy from 1.3" in V to 1.9" in UVW2 (see Figure 7). A
direct measurement of the curve of growth of the PSF is limited by the small number of
appropriate, isolated stars and the large scatter caused by coincidence loss and straylight.
Therefore we have used Daophot for our PSF analysis. This allows us to t the same
function to all the stars in a eld of view, thus giving a set of good average PSFs for a
range of count to framerate ratios (CFRR). This has beed achieved for the U, B, V lters
and incooperated into SAS. However, the UV PSF is not so well dened and has no count
rate dependence because we lack good calibration observations.
We are working on the UV PSF using the LH9-LH10 clusters which provide some
bright UV stars. The main problem so far is the crowded nature of the eld. Whilst
this provides a large number of stars with a large range of count rates, it is hard to nd
enough stars suфciently isolated to measure the PSF wings.
6 Distortion
The OM telescope optics, the lters and the detector itself, all contribute to a certain
amount of distortion in the nal image. At the edges of the eld of view, shifts of up
to 20" are seen (see Figure 8, the upper panel). A new distortion map for the U-lter
has been made by comparing the actual and predicted positions of 813 stars in an OM
calibration eld.
The map can be used to correct position to a level of about 1" rms error (see Figure
8, the lower panel). Analysis has shown that the distortion in other lenticular lters is
similar to that of the U lter. Therefore, this distortion map is used in the SAS for all
lters. We should point out, that the source positions determined by SAS are based on
the satellite attitude information, and are not cross-correlated with any catalogue. A few
arcsecs oset from the real positions can exist due to the pointing error.
7 Large scale Sensitivity variation
Two sets of oset observations of the eld EXO 0748 in V lter have been obtained in
rev. 212 and rev. 338. Apart from the central pointing, there are 4 additional pointings
shifted to the North, South, East and West by 2 arcmin. In the rst observation (rev.
212) the exposure time for each pointing is 1000 seconds, and in the second observation
(rev. 338), it is 4000 seconds. These data allow one to get local sensitivity gradients and
combine them together to obtain a CCD sensitivity map. Five measurements for every
star are used to determine the local gradients, whose values and directions are shown in
Figure 9 as arrows. Circles show the value of the statistical error for every star. If the
9

Figure 7: OM PSF in dierent lters
local gradient is larger than the statistical error, the arrow must be longer than the radius
of the corresponding circle. Even if the local gradients are smaller than poisson errors but
systematic, we would see similar orientation of arrows in some areas of the plot. This is
not the case, the directions of the arrows seem to be chaotic. Also, if one compares the
directions of the gradients in the same areas of the CCD in two observations, one nds
no similarity. These results tell us that large scale sensitivity variations, if present, are
smaller than the current photometric accuracy of observations and do not exceed a few
per cent. Current SAS assumes no large scale sensitivity variation and this should no
degrade the accuracy of OM data.
8 Background
The background count rate in the OM is dominated by the zodiacal light in the optical.
In the far UV the intrinsic detector background becomes important. The zodiacal light
level depends on space coordinate. As an approximation, we adopt the sky background
as 100 times the brightness of one V=10 mag solar spectral- type star per square degree,
and the values (in unit cts/s/arcsec 2 ) for each lter are shown in Table 2. In Table 2,
we also include several measurements from real observations. We can see that the values
derived from the simulations and the observations are comparable in three optical lters.
However, for the UVW2 and UVM2 lters, the simulated values (  10 5 c/s/arcsec 2 ) are
much smaller than that from observations. This is because, for these lters, the zodiacal
light level is negligible. The OM dark count is 0.00025 cts/s per pixel, and each pixel
is about 0.5"  0.5" on the sky, therefore, the Dark count contributes 0.001 c/s/arcsec 2
to the background count rate and is the most important for UVM2 and UVW2. For
the UVW1 lter, it seems that both the zodiacal light level (0.0007 c/s/arcsec 2 ) and
the dark count (0.001 c/s/arcsec 2 ) contribute the total background count rate (  0.002
c/s/arcsec 2 ).
10

Figure 8: This is a 5th order polynomial t to the distortion measured from star positions (the upper
panel). The lower panel is the distortion map after the position correction.
11

Figure 9: Local gradients vs. Statistical noise. Left: rev. 212, right: rev 338. The crosses show the scale
of the oset pointings. The sizes of 1 per cent count rate dierence are also shown.
Figure 10: The histogram of dark counts
12

Figure 11: Out-of-focus ghost image ("smoke ring") caused by re ection of bright eld stars during SA95
observation (B lter). The backgound is also enhanced in the central region due to re ection of the diuse
sky light from outside the eld
Figure 12: The BPM16274 eld taken with the optical grism. The spectrum for the target has been
extracted and the extraction region is illustrated with a box
13

Table 2: The simulated (100 S10) and observed background count rate in the OM.
Filters 100 S10 BPM16274 Her X-1 PKS2155-304 Lockman Halo PSK0558 SA95
V 0.0124 0.0166 0.0362 0.0186 0.0152 0.0258
B 0.0167 0.0242 0.0517 0.0394
U 0.0040 0.0085 0.0190 0.00923 0.0154
UVW1 0.0007 0.00263 0.00254 0.00284 0.00258 0.00427
UVM2 1.7 10 5 0.00120
UVW2 7.2 10 6 0.00125 0.00100 0.00133 0.00103
9 Flat elds and dark frames
Dark frames are obtained by moving to the blocked lter and making an engineering
observation with the detector. For a at eld, the LED is turned on in addition so that
photons are re ected from the back of the blocked lter into the detector.
9.1 Dark count rate
The dark frames are regularly taken with the blocked lter and no LED illumination to
measure the detector dark counts. The dark count rate is very low, at 0.00025 count s 1
per pixel as shown in Figure 10. The variation in dark count rate across the detector is
at a level of 7% and there is no signicant trend with time.
9.2 Flat elds
The LED illumination provides a count rate of 0.0078 cts/s, but it is not a uniform
illumination. A super at has been produced which includes more than 70 hours of at
eld time, and reaches the 2% statistical level. From this super at, we are able to recognise
bad pixels and an update has been made to the SAS calibration le. After removing the
illumination signature, the super at can be used to measure pixel to pixel non-uniformities
caused by small-scale variations in the detector sensitivity over the eld of view. The work
on the photometry has shown us that the pixel-to-pixel variation is at a level of less than
5%. Current SAS uses a unity at eld, this does not aect the accuracy of the results
in more than a few percent.
10 Straylight
There are two causes for the straylight appearing in the OM images. In the rst, there is
an internal re ection of light within the detector window, which leads to an out-of-focus
smoke ring image of a bright star displaced in the radial direction away from the primary
image. The second eect is due to the re ection from part of the detector housing, of
o-axis starlight and sky background lying between 12.1 and 13 arcmins from the center
of the eld of view. The re ected stars cause low-level loops and streaks in the image, the
re ected background leads to a diuse annulus in the center of the image. Since SAS uses
aperture photometry to measure the source ux or magnitude, straylight only increases
the local background, thus reduce the sensitivity, but not aect source detection and
magnitude too much. This can be seen from Figure 11. This is the OM SA95 observation
in B frame. The standard star (SA95-42) at the center is just in the position of the
bright straylight ring. Stetson found its standard V and B magnitudes are 15.606 and
15.392, which is in agreement with our determination from OM observations (V=15.62
and B=15.50).
14

11 Fixed patterning
Fixed patterning (also known as mod8 noise) is a side eect of the centroiding of the
photon splash on the CCD resulting from the one photon incident on the front of the
detector. By centroiding, the 256 x 256 real CCD pixels are diveded into 8 x 8 subpixels,
giving 2048 x 2048 in total. As the centroiding has to be done in real time, a lookup table
is used, which cannot take into account variations in photon splash shape over the surface
of the detector and thus leads to "pixels of unequal sizes", and a pattern which repeats
on an 8  8 subpixel grid. However, the average mod8 noise can be measured from the
background sky, and then the counts may be re-sampled based on the true pixel sizes, as
it is done in the SAS mod8 correction routine.
12 Fast mode
In fast mode, the OM does not produce accumulated two-dimensional images, but in-
stead produces event lists like the X-ray instruments. This mode is useful for monitoring
rapidly variable sources, for example AGN or accreting binaries. One or two windows
approximately 10 10 square arcsecs can be dened. The fastest time sampling is 0.5 sec.
OM fast mode is fully supported with SAS 5.3.3.
13 Grisms
The OM grism calibration is on-going. Figure 12 shows a typical grism image. For each
source the grism gives a dispersed rst order image and a zeroth order image, displaced
from one another in the dispersion direction. Default windows are available for selecting
the region of the eld of view which will contain the target zeroth and rst order spectra.
The current version of SAS 5.3.3 does not include any utility to deal with spectra
obtained with the OM grisms. A tool based on IDL has been developed by the OM
calibration team as an aid to their on going work on the calibration of the grisms. It is
planned to incorporate this tool into the SAS soon for processing the grism data; initially
this will be for the spectrum of the object lying at the boresight.
14 Acknowledgements
I would like to thank all the people who have contributed to the operation and calibration
of XMM-OM. This includes many people at MSSL, Liege University, UCSB, ESTEC as
well as VILSPA.
15 References
Colina L., Bohlin R.C., 1994, AJ, 108, 1931
Gunn J.E., Stryker L.L., 1983, ApJ Supplement, 51, 121
Mason K.O., Breeveld A.A., Much R.M. et al. 2001, A&A 365, L36-L44
Mason K.O., 2002, ESA SP-488, ESTEC, The Netherlands
Landolt A. U., 1992, AJ, 104, 340
Stetson P. B., 2000, PASP, 112, 925
15