This Legacy journal article was published in Volume 3, May 1993, and has not been
updated since publication. Please use the search facility above to find regularly-updated information about
this topic elsewhere on the HEASARC site.
The On-Axis Point Spread Function:
In-flight comparison with
the PANTER results
Addendum: High Signal-to-Noise
Gunther Hasinger (MPE), T. Jane Turner (GSFC ROSAT GOF),
Ian M. George (HEASARC) & Gunter Boese (MPE)
Version: 1992 Oct 30 OGIP Calibration Memo CAL/ROS/92-001a
This memo is an addendum to OGIP Calibration Memo CAL/ROS/92-001 (Hasinger et
al 1992) in which we compare the ROSAT PSPC on-axis point spread
function psf with the best signal-to-noise data obtained to date for a
point source. We confirm that the parameterization detailed in CAL/ROS/92-001
does satisfactorily describe the observed psf at energies o(>,~)0.19
1 The Data & Model
As detailed in CAL/ROS/92-001, the psf of the ROSAT X-ray mirror
assembly (XMA) + PSPC is a convolution of 5 components, 3 of which (namely the
XMA scattering profile, the intrinsic spatial resolution of the PSPC, and Focus
and detector penetration effects) have been parameterized for on-axis
observations from PANTER ground calibration measurements. In CAL/ROS/92-001 it
was shown that the derived analytical functions satisfactorily described 5
moderate signal-to-noise datasets out to a radius ~1-2 arcmin in all but the
softest band (B-band; where `ghost imaging' was already known to be a problem).
Here we use a very high signal-to-noise (o(>,~)6 x 106 photons)
dataset of a (nominal) point source for a detailed comparison of the model
psf and data out to a radius ~10 arcmin.
1.1 The Dataset
The data used here is proprietary data (PI Thomas) thus the source is not named
here, nor is the observation described in detail. Radial profiles were
extracted, centered on the source, using the same method described in
CAL/ROS/92-001, with 5 arcsecond incremental radii, in the bandpasses as listed
in Table 1 of CAL/ROS/92-001. The lowest 11 PI channels were rejected to
exclude problems due to the variable lower limit discriminator for valid
events, due to the variable instrument gain which is folded into these data.
Any additional sources falling within the specified annuli were masked out of
the analysis. No background subtraction was carried out. Background rates were
measured from the images and were later folded into the predicted profile
template for each band.
1.2 The Model
As the in-flight data are affected by more uncertainties than the ground data
(aspect corrections, background subtraction, gain correction, etc.) it was not
possible to allow profile fitting with the parameters of the gaussian +
exponential + lorentzian components to be free. Instead we calculated the
psf for the source in each bandpass.
First, a spectrum was extracted for the source, in a circle of size 200
arcseconds radius (masking out a nearby source). Next, a psf was
calculated for each energy channel using the algorithm and parameters given in
CAL/ROS/92-001. A predicted psf
template was calculated in each band using the source spectrum to determine the
photon weighting to be applied to the psf component in each energy
channel. Thus, for each band and spectrum a combined psf was produced,
including a constant term for the background.
These predicted psf templates are overlaid on the appropriate datasets
in Figure 1. We stress that the normalization of the model psf in each
case was also calculated using the equations in CAL/ROS/92-001, thus no
fitting was performed. The normalization of the predicted psf was
calculated such that the integral under the predicted template is equal to the
integral under the observed psf. Discrepancies between the shape of the
two curves thus result in the slight discrepancy at the peak seen in some
panels of Figure 1.
It can be seen that even at this high signal-to-noise ratio, the psf
model provides a good parameterization of the source profile for all but the
B-band. The individual panels give an estimate of the error in the psf,
which it was not possible to illustrate with lower signal-to-noise data.
B-band (0.12-0.188 keV; panel a) illustrates the ghost imaging
problem present in the softest channels of the PSPC.
C-band (0.188-0.284 keV; panel b) illustrates a slight excess in
the data at radii ~1-2 arcmin, and a very small excess in the data at radii
~3-4 arcmin. Previously it was thought that the ghost imaging effect was only
important below PI channel 15 (0.15 keV), but recent analysis of other high
signal-to-noise data has suggested ghost imaging may be significant up to
channel ~30 and thus may significantly affect the C-band (Nousek & Harnden,
R1-band (0.284-0.5 keV; panel c) illustrates a notable excess in
the data at radii ~1.5-3 arcmin.
R2 & R3-bands (0.5-2.48 keV; panels d & e) demonstrate
good agreement between the data and model psf.
These small scale discrepancies may reflect some of the differences between the
ground calibration and in-flight data:
Although there has been no evidence to date to suggest an extended X-ray
component in this source, we note that the signal-to-noise and spatial
resolution obtained here are far superior to any previous observations of this
source (or different source of the same class). Thus we cannot rule out the
possibility that some of the above discrepancies are due to a contribution of a
genuine extended component associated with the source.
- the Gaussian component (due to spatial resolution of the detector): the
width in the model could be narrower than the data due to residual errors in
the attitude corrections;
- the exponential component (due to combined focus and penetration effects):
the geometry of the beam used in the PANTER calibration facility was different
to the parallel beam of the in-flight data. Therefore the detailed shape of the
penetration term could be slightly different in-flight;
- the Lorentzian component (due to the scattering profile of the XRT mirrors):
the average grazing angle is slightly larger in the finite beam than in orbit.
Approximately 40% less mirror scattering is expected in-orbit than on the
ground. In addition, the mirrors may have suffered dust contamination after
the PANTER tests, and may therefore have a greater microroughness in-flight.
The conclusions given in CAL/ROS/92-001 still hold, and these data give the
best estimate to date of the residual error in the psf model.
Work continues to refine and improve this model, and to extend this work to
include off-axis angles.
We thank the PI Thomas (MPE), for allowing his proprietary data to be used in
this way, and the many people at MPE involved in the determination and
interpretation of the PANTER data. We also thank Dave Davis (GSFC) for his help
extracting the data.
Hasinger, G., Turner, T. J., George, I. M. & Boese, G., 1992.
Legacy, 2, 77 (CAL/ROS/92-001).
Proceed to the next article
Return to the previous article
Select another article
HEASARC Home |
Last modified: Monday, 19-Jun-2006 11:40:52 EDT