OGIP Calibration Note CAL/ROS/93-007
Stability of the PSPC Spectral Calibration
Calibration Observations of N132D
Version: 1995 Feb 01
Code 668
NASA/GSFC,
Greenbelt, MD20771
The ROSAT PSPC data processing software SASS, gain corrects the pulse height analyser (PHA) spectral data to pulse invariant (PI) data. This study note tests the accuracy of those corrections by making a simple ratio comparison between spectra of a constant X-ray source, N132D taken at different epochs.
Release | Sections Changed | Brief Notes |
Date | ||
1993 Jul 26 | First public version | |
1995 Feb 01 | All | Made compatible with LaTeX2HTML software |
The calibration corrections performed on PSPC events in the SASS processing
are described in detail in "Calibration Corrections to individual PSPC
events" by Hasinger and Snowden (1990). The correction from PHA
to PI for the spectral data, essentially folds in calibrated
gain variations. Quantities such as the detector
gain (pulse height as a function of energy) change with time due to
variations in the detector high voltage, temperature, gas pressure, density
and composition (contaminants arise due to outgassing of the inner wall of
the tank as the internal pressure drops). An additional variable will be the
reduced gas flow rate during AO4. The temporal gain
variations are currently parameterized by fitting Prescott functions to the
Al Ka calibration data. Al Ka calibration observations before
and after the target observation are linearly
interpolated to estimate the appropriate temporal
correction required.
A spatial gain adjustment corrects for the
variations due to the variable distance of the anode wires and event
position relative to these (see Hasinger and Snowden 1990). The PSPC also
shows some saturation of gain, which is a minor effect also corrected for
in the SASS processing.
These corrections are performed in SASS and, if accurate, one should see no
significant variations between spectra of a constant soft X-ray source
extracted in a consistent manner. Some post-SASS corrections may need to be
applied, ie a vignetting correction for off-axis data, and an exposure
correction. This note examines the accuracy of the PHA to PI correction.
The constant X-ray calibration source, the supernova remnant
N132D was used for this analysis.
Only a limited number of the calibration and maintenance day data are
currently available to the U.S. GOF.
Source spectra were extracted in a (conservative) 12 arcminute cell (the
source is extended and 40 arcminutes off-axis)
and were background subtracted. These were high gain
state data, and taken with PSPCC.
All of the following observations were made with PSPCB.
Four datasets are currently available:
Sequence Date Gain state
rp160084 May 1991 High
rp141800 Dec 1991 Low
rp500004 Apr 1992 Low
rp141937 Apr 1993 Low
RP500004 is available for calibration studies thanks to PI J. Hughes.
Spectra were extracted in a 5 arcminute cell centered on N132D, while a
nearby source was excluded (at 4 arcminutes away, and with a count rate
only 0.4% of N132D this nearby source is insignificant in this analysis).
The PSPC gain was lowered from 3060
volts to 3000 volts on October 14 1991, resulting in a gain drop of ~
30%. These data cover both gain states. Although the lower discriminator
for valid events was raised from channel 9 to channel 11 (of 256) following
this gain drop, thoughout this analysis we consistently used only channels
12-211 binned by a factor of 10.
Each spectrum in turn was compared to the most recent spectrum from
rp141937, by making a
simple ratio of pairs of spectra with the April 1993 spectrum as the
denominator. Figures 1a,
1b,
1c,
1d
show these ratios in chronological order with
respect to the numerator.
The early PVC spectrum is the numerator in Figure 1a,
and more recent
spectra appearing in subsequent panels. It can be seen that the
further apart in time the spectra, the greater the discrepancy, although
the most recent two (Figures 1c and
1d) agree acceptably well. Overall the data
imply a time
dependant gain variation has occured, which has not been taken into account
in the gain correction procedure. The largest differences are between the
high and low gain state spectra (Figures 1a and
1b), the SASS
corrections, were they adequate, should make these spectra equivalent.
Inaccuracies in the vignetting correction function used may be partly
responsible for the large discrepancy in Figure 1a, as
may some (unknown)
difference in the response of PSPCC and PSPCB. We cannot easily
deconvolve the different contributions to the discrepancy in
Figure 1a using these
data.
Silvano Molendi has taken Steve Snowdens Exposure Maps, constructed from the
survey data and (masking the ring and ribs) extracted counts in annuli out
to the edge of the detector. The count rate ratios can then give the
vignetting function. The survey data are of very high S/N, and show
excellent agreement with the preflight vignetting function (ie double
power law). Theres a small discrepancy above 40 arcmin, where they observe a dip
in the survey data versus the model, this rises up again and goes above the
model, crossing it at about 54 arcmin. The size of this discrepany and the
radius where is starts seems to be a function of energy. The origin of this
effect is not yet understood but the discrepancy between the preflight and
in-flight vignetting functions seems to be less than 4%, hence this cannot
be the predominant effect in Figure 1a.
Guenther Hasingers independant
examination of N132D data yield the conclusion that
there is some variation of gain, which is not currently corrected
for in the conversion to PI data. GRH
observes that the peak energy of the counts distribution in the N132D
corrected data decreases monotonically as a function of
time.
There appears to be a
significant time dependance of the PSPC gain, which is not
accounted for in the SASS correction from PHA to PI data.
Although the data are temporally corrected using the position of the
Al Ka calculated from a linear interpolation between calibration
line observations either side of the target observation, this does not seem
to provide an adequate correction across the PSPC bandpass. Work at MPE
suggests that the peak channel for the N132D spectrum is
decreasing monotonically with time even after
the nominal corrections have been applied.
A more sophisticated gain correction
should be possible after examination of much more
calibration data.
This effect may explain some of the apparent anomalies noted in the
PSPC spectral data.
Recent correspondance with MPE also suggests that a spectral recalibration
is required to fold in a rate-gain correction to Mkn421 (estimated
at 1-2%).
1 INTRODUCTION
2 DATASETS USED
2.1 PVC data
A large number of brief off-axis spectra were taken during the pre-survey
calibration phase of June 1990, although no useful unfiltered
on-axis data from N132D are available for the same epoch. Comparison
between the ~ 200 second off-axis spectra showed no significant variations
between these, ie on a timescale of days (albeit at low signal to noise).
Thus, we made a mean spectrum for the 40 arcminute offset spectra taken in
June 1990, and applied a weighted
vignetting correction to make it equivalent to an on-axis spectrum.
Sequences used were: rp110063; rp110071; rp110076; rp110083; rp110088;
rp110095 and rp110107.
2.2 Maintenance Day Data
3 RESULTS
4 WORK AT MPE: based on email correspondance of 1993 Jul 23 with GRH
5 CONCLUSION
FIGURES
The following useful links are available (in the HTML version of this document only):