[About EXOSAT] [Archive] [Software] [Gallery] [Publications]


Although the present calibration of the ME detector is adequate for the majority of sources, some problems have been noticed when carrying out spectral fits to bright sources. In particular:

1. An excess of counts above channel ~ 60 in, the argon chamber.

2. Systematic trends in the residuals around 3 keV.

3. Poorer fits to the spectra of bright sources obtained during the last year of the mission.

4. Differences in the relative effective areas of the argon and xenon chambers when simultaneously fitting argon and xenon data to sources with quite different spectral shapes.

These problems reflect three basic uncertainties in the calibration of the ME. 1) The value of the effective area of the argon and xenon areas corrected for collimator transmission effects, 2) the thickness of the windows to the argon chamber for each detector and 3) the non-linearity of the gain of the argon detectors. In order to overcome these problems the ME is being recalibrated using a new three stage approach in which each of the above uncertain parameters are determined by matching the predicted count rate and spectrum to those observed from the Crab. The measurements used for this recalibration are the 9 Crab Nebula observations made at various intervals through the mission, the original ground calibration and the in-flight radioactive calibration data. Final details of the implementation of this new calibration will be provided in the near future when a calibration update tape is issued. The following gives an overview of the methodology that is being used.

1. Determine the geometric area of the detectors

Each of the eight xenon chambers is located behind an argon chamber and so the xenon sensitivity is modified by the absorption properties of the argon chamber and the associated detector windows. The overall count rate in any particular xenon detector is determined by the following:

a. the transmission of the collimators.

b. the detector geometric area.

c. the thickness of the front window.

d. the depth of the argon chamber.

e. the pressure in the argon chamber.

f. the thickness of the window seperating the argon and xenon chambers.

g. the depth of the xenon chamber.

h. the pressure in the xenon chamber.

i. the loss of area associated with the xenon guard wires (1-2% the total area).

j. sampling losses from the onboard computer (0BC).

The overall counting rate from the Crab in each xenon chamber is not strongly dependent on the gain or resolution functions because at low energies the overlying material (i.e. the argon chamber and the intermediate window) limits the count rate, while at high energies the steeply falling Crab spectrum means that few counts are observed above ~40 keV. In the above list the largest uncertainty is associated with the total effective area and the efficiency of the collimator. The low energy response is principly determined by the depth of the argon chamber, and the thickness of the intervening window and are both well known. Uncertainties in the other properties that effect the low energy response such as the thickness of the windows, the kapton foil, the column density to the Crab and the pressure of the argon chamber do not significantly effect the total count rate. The loss of area associated with the guard wires is small and well known from ground calibrations. The OBC sampling losses are well calibrated. A power law photon spectrum with an index of -2.10 and a normalization of 9.7 (the generally accepted values for the Crab) was folded through the response of each detector and the geometric area/collimator transmission adjusted until the predicted count rate agreed with that observed.

2. Determine the Front Window Thicknesses

One of the critical parameters for the argon detectors is the thickness of the window. Small variations in thickness can lead to large changes in the spectral response at low energies. Again the overall counting rate is relatively insensitive to uncertainties in the gain. The counting rate in the argon chamber depends on:

a. The column density to the Crab.

b. The thickness of the kaptan foil.

c. The front window thickness.

d. The area/collimator transparency product.

e. The depth of the argon chamber.

f. The loss of area associated with the argon anticoincidence guard wires.

g. The pressure in the argon chambers.

h. The losses associated with the risetime vetoing.

i. OBC sampling losses.

The geometric area/collimator transparency product was assumed to be the value obtained from matching the observed and predicted xenon counting rates. The value used for the column density to the Crab was the value of 3 x 1021 H atoms cm-2 measured by EXOSAT by fitting various absorption models to the count rates obtained in the low energy telescope using different filters. This value is in good agreement with previous measurements. The risetime and guard losses were measured by observing the Crab with each turned on and off. Pre-launch measurements of the kaptan foil thickness and gas pressure were assumed. All the other parameters are relatively well known from ground or inflight calibration. The front window thickness could then be optimized by varying the window thickness of each detector until the predicted and observed counts from the Crab were equal. Non-uniform window thicknesses did not significantly improve the fitting procedure.

3. Determine the Gains and Energy Resolutions. Once the geometic areas and window thicknesses are known the gain and energy resolution functions can be determined by iteratively comparing an observed spectrum with that predicted using trial resolution and gain functions until agreement is reached. In the current calibration data the gains and resolutions are given analytically. In the ground calibrations it was noticed that there was a small dependence in the gain with position on the face of each detector. The dependence was such that the gain was highest over the anode wires and lowest between them. In the case of the argon chambers gain changes as large as 10% were noticed, while for xenon the effect was about a factor of two smaller. The effect is likely caused by space-charge effects. Since the gain measurements were not made at energies above 8 keV it is possible that the effects are larger at higher energies. The new calibration data includes the effects of this gain modulation by allowing for a more complex resolution function than hithertoo used. Such a gain modulation seems consistent with a decrease in resolution seen at high energies and the observation of genuine x-ray counts in high channel numbers in the argon chambers. Possible origins for the systematic "wiggle" in the residuals at around 3 keV have been considered. Using the present calibration data these tend to have a characteristic shape with an excess of observed counts at ~3 keV and a deficit at slightly higher energies. It is likely that this effect is caused by variations in the widths of individual energy channels. In the new calibration data this is accounted for by providing individual channel boundaries, rather than analytic functions for the gain curves. Channel boundaries are obtained by fits to each of the Crab observations and the time dependence is modelled by assuming a linear change between observations. The new calibration scheme will involve 1) slightly revised areas and window thicknesses, 2) Individual energy channel boundaries as a function of time, 3) Revised resolution functions and 4) new electronic acceptance values (in fact unmodified guard acceptance values). Currently we are comparing the results of fits to a range of bright sources using the current calibration data and a preliminary set of revised data. Another cross check being made is a comparison of GSPC and ME results on bright sources.

A. Smith
A. N. Parmar

[About EXOSAT] [Archive] [Software] [Gallery] [Publications]