During the on-ground calibration measurements, performed at Laben in Milan,
the DU-EU assembly was connected to the Instrument Test Equipment (ITE)
developed by Laben. The ITE is a computer controlled system with a probe,
the On Board Data Handling Bus Emulator and Probe, that emulates the
satellite OBDH bus. It retrieves data from the HPGSPC-EU in the form of
telemetry packets and send those data to a VAX station via TCP/IP protocol.
The on-line data handling, the off-line processing and archival of data
from ITE were performed by the IFCAI calibration support system. A two axis
PC controlled translation stage was used to position collimated radioactive
sources in the relevant point of the entrance window of the detector. Test
and calibration of the HPGSPC were performed using four different
radioactive sources. Sources and relevant lines are given in
Table 3 (click here). During the collimated measurements graded shield
collimators (made of Pb-Sn-Cu-Al) were used. A 
(partially
collimated) gamma ray source was also used to simulate the gamma ray
induced background events.
Table 3: Calibration sources and lines
Burst length discrimination is a well established method to reduce the background counting rate in GSPCs. The method, that in the case of the HPGSPC is based on the ARC/CFT electronic chain (Leimann & Van Dordrecht 1979), essentially discriminates the pulse time duration produced by the localised electron cloud generated by the absorption of a genuine X-ray, against the extended ionisation track induced by background particles interacting in the gas. Also background events generated close to the detector walls by the interaction of gamma rays or particles in the detector shielding are discriminated because of their very short burst lengths.
For each event, the 20 - 80% rise-time of the integrated sum
signal is measured. Only events within the experimentally determined
upper and lower thresholds are recognised as potential genuine X-ray
events. The full area burst length spectra relative to 
and 
X-ray sources are shown in Figs. 5 (click here) and
6 (click here). The 
spectrum follows a Gaussian
distribution with a peak value at 
and a resolution
of 8.3% FWHM. The burst length spectrum of 59.5 keV photons of
is asymmetric with an extended tail towards longer burst
lengths. The peak position is at 
and resolution is
8.6% FWHM. The tail is due to events in which the residual and
fluorescence clouds overlap, scintillating for a longer duration.
Figure 5: Burst length spectrum of the 22 and 25 keV lines of
radioactive source. Continous line is the best gaussian
fit to the experimental measurement
Figure 6: Burst length spectrum of the 59.5 keV line of 
radioactive source. The long tail towards longer burst lengths is
due to events which give rise to overlapping residual and
fluorescence electron clouds
Figure 7: Burst length versus energy diagram for 22 and 25 keV
photons from 
radioactive source. See the text for a
description of the labels. Scale is logarithmic
Figure 8: Burst length versus energy diagram for 59.5 keV photons
from 
radioactive source
In Fig. 7 (click here) we show the burst length versus energy
diagram for the 22 and 25 keV line of 
. Events labeled
with (1) are associated with Cadmium events absorbed in the Drift
Region while the extended tail (2) is correlated with events that
penetrated deep into the Scintillation Region. These events,
scintillating for a shorter time, give rise to a detected events
with lower energy and burst length. For incident photons with
energies above the Xenon K-shell the situation is more complex.
Figure 8 (click here) is an exposure at 59.5 keV photons from
radioactive source. The events labeled with (1) are
associated with 59.5 keV events that have been detected as ``Single"
i.e. events originating a single electron cloud. The burst length
distribution relative to these events shows a very long tail (2)
towards longer burst lengths. As mentioned above, these are events
that generated a fluorescence photon whose cloud overlaps the
residual cloud. Events labeled with (3) are associated with the
and 
fluorescence photons and the
and 
escape peaks. Tails are not present for
these events associated with a single electron cloud. Events
absorbed in the Scintillation Region give rise to the ``strips" (4)
that connect the source region with the dense region at lower energy
and lower burst length (5). The latter is associated with the low-
energy features that are due to 59.5 keV 
events
absorbed in the Back Region which, scintillating close to the wires
of scintillation grid along paths of anomalous electric field,
generate small short pulses.
Figure 9: Burst length versus energy diagram relative to
illumination. Labels are explained in the text
All the features described above are well summarised by the
burst length versus energy diagram shown in
Fig. 9 (click here). All the ``genuine" X-rays peaks are distributed
in a very narrow region that run almost parallel to the corrected
energy axis, confirming that the burst length distribution is
independent of the energy of incident X-rays. Long tails are
associated with the primary release plus overlapping fluorescence
for the primary lines at 136, 122, 59 and 67 keV of Cobalt (1), (2),
(3) and (4) while tails disappear in the peaks at 92.3 keV and 88.3
keV (associated with residuals (5)) and at 29.7 and 33.8 keV
(associated with fluorescence photons (6)). The 58.9 and 67.6 keV
lines are respectively the 
and 
fluorescence
photons of the Tungsten (that shielded the Cobalt radioactive
source) induced by the 136 and 122 keV photons of the 
.
In order to reject events absorbed in the Scintillation and Back Region,
the burst length lower threshold was set at 
, independent
of the energy of the primary events. The upper threshold will probably
depend on the energy of the incident photon. For events with energy below
the Xenon K-shell the upper threshold will be set around 
, narrowly around the peak of bl versus energy distribution. For
events above the Xenon K-shell the upper thresholds has to be set to
maximise the signal to background ratio which will depend on the in orbit
HPGSPC background.
Due to the hexagonal symmetry of the HPGSPC the look-up table need
in principle be determined only in a 
degrees sector
that is assumed as reference sector and then extended to the entire
area of the instrument with an appropriate set of rotations and
reflections. The physical symmetry of the HPGSPC was verified by
sampling all the sectors using the response to the 22 keV peak line
of 
in 20 different points per sector. Only the
reference sector was carefully mapped in 120 points.
Figure 10: Isocontours of the 
coefficient relative to the whole
detector open area. Dashed lines are from experimental mapping in all
sectors, while continuous lines are generated with a symmetrical
mapping of the reference sector
Figure 11: Isocontours of the 
coefficient. The two mapping
are in good agreement showing that lateral PMTs were correctly
equalised
Figure 12: The 
correction factor isocontours: asymmetry is
clearly visible
Figures 10 (click here),11 (click here),12 (click here) show the isocontours
relative to 
, 
coefficients and to the 
correction factor. The dashed lines plot is obtained from the
experimental measurements, while a continuous line map is generated
with a symmetrical mapping of the reference sector on the whole area
of the detector. While for the 
coefficient the two
mapping are in good agreement (differences are less than 2%),
showing that the gains of the lateral PMTs were correctly equalised,
the asymmetry in the 
contours is a consequence of the
shift present in the 
contours. This is due to a misalignment
between the geometrical centre of the detector and the optical
centre of the central PMT. Simulation show that even a quite small
misalignment (less than 1 mm and/or 1 degree) can have an impact (up
to 10% for r greater than 10 cm) in the Kc factor in some of the
sectors. The imperfect reconstruction in these sectors affects the
energy resolution and gives rise to a deviation from the gaussian
statistical distribution of the reconstructed lines.
To minimise the effect of the asymmetry, we derived directly a
(
, 
) distribution for a long measurement of a
59.5 keV source illuminating the full area of the
entrance window (see Fig. 13 (click here)). Of course,the (
,
) distribution does not depend on the energy of the
calibration line used.
The 
constant map can then be obtained just by subdividing the
(
, 
) coefficient distribution in rectangular areolas
whose vertices belongs to 
isolines, (see Fig. 13 (click here)).
Figure 13: The (
, 
) distribution for the 59.5 keV
line of 
For each areola i the 
factor can be defined as:
where 
is the correction factor of areola i, Peak(0)
is the peak of the 59.5 keV line of 
in the energy spectrum
collected in the central areola and Peak(i) is the peak of the 59.5 keV
line of 
in the energy spectrum collected in the i
areola. It must be stressed that the effects of asymmetry are averaged
because photons collected in each (
, 
) areola are indeed
physically detected in (
) positions, which differ from sector to
sector.
The areolas are iteratively determined in such a way that a finer
discretisation does not improve the energy resolution in the local
(
, 
) region. As a consequence, the variation of the
correction factor 
has been determined to be less than
1% in the outer areas of the (
, 
) distribution while
is less than 0.5% in the inner region (where the local
energy resolution is better due to the higher yield).
The correct working of the energy reconstruction algorithm is a key
point for the optimum scientific operation of the HPGSPC. This is
illustrated in Fig. 14 (click here) in which both the unreconstructed
and reconstructed full area raw energy spectra of a 
radioactive source, are shown. The source was positioned 3 meters
from the entrance beryllium window in order to illuminate the whole
entrance area of the detector. Both single and correlated events are
accumulated.
Figure 14: Reconstructed and unreconstructed energy spectra of
radioactive source
The unreconstructed spectrum is strongly distorted, shifted towards
the lower channels and the lines are not well resolved. In the
reconstructed raw spectrum the 
59.5 keV line is
clearly detected and peak position is correctly reconstructed. The
three, rather well resolved, lines at 25.7 keV, 29.7 keV and 33.8
keV correspond respectively to the 
residual peak
(
), the 
fluorescence of the
Xenon that overlap the 
residual peak
(
) and the 
Xenon fluorescence.
The two low energy features clearly visible in the spectrum below 10
keV are due to 
59.5 keV photons absorbed directly in
the Back Region which give rise to an electron cloud scintillating
very close to the scintillation grid wires along anomalous paths.
However as described in the previous paragraph this undesirable
features can be almost totally rejected by burst length selection.
The HPGSPC resolving power in the entire operative range is shown
in Fig. 15 (click here) where both unreconstructed and reconstructed
energy spectra of a 
radioactive source are given.
Figure 15: Reconstructed and unreconstructed energy spectra of
radioactive source. Low energy features are due to events
absorbed in the Back Region. 
and 
lines,
emission from Tungsten is also visible
Low energy features are quite enhanced as expected because of the longer
penetration depth of higher energy photons. In the reconstructed spectrum
two lines at 13.5 keV and 16.8 keV are visible. They are due to Uranium
contamination (
and 
lines) in the beryllium window.
In Fig. 16 (click here) we show the full area corrected energy spectrum of
the 22 keV and 25 keV lines of a 
radioactive source in which
background has been subtracted and
burst length thresholds have been applied. A double gaussian fit is also
shown. The agreement is not perfect
because of residual distortion of the reconstructed spectrum in some of the
sectors and due to discretisation in the 
look-up table that
can introduce a deviation of the counting statistics
from the poisson
distribution. The deviation from
a gaussian nature is modeled in the response matrix of the instrument.
Figure 16: Full area energy spectrum of 22 and 25 keV lines of
radioactive source
In Fig. 17 (click here) we report the full area energy corrected
spectrum of an 
radioactive source after background
subtraction and burst length selection. The spectrum has been
accumulated in Single Mode. Line at 59.5 keV contains, then, only
events that have been detected as ``Single". The tail towards lower
energies is due to double events that scintillate in spatially
separated centres but overlap in burst time and then give rise to a
non correctly detected light production centre. The line resolution
is 3.3% +/- 0.1% FWHM. The other two peaks at 25.7 keV and 29.8
keV correspond to the Escape 
and Escape 
i.e. to events for which the fluorescence photon exit from the
detector. The low energy features relative to the events absorbed in
the Back Region are much reduced in intensity. The Fluorescence
Gated spectrum is shown in Fig. 18 (click here) (the background
has been subtracted and burst length thresholds have been applied).
Only events detected as ``Double" are accepted if at least one of the
events has an energy determined to lie within 
of
either the 
or 
fluorescence energy, and if
this is the case, the exactly known energy of the 
or
line is summed to the measured energy of the residual.
As can be seen in the figure, the energy resolution in FG mode is
improved with respect to that obtained in SE mode by up to 2.4%.
Figure 17: Corrected energy spectrum of 59.5 keV line of
as seen by HPGSPC in Single Event Mode
Figure 18: Corrected energy spectrum of 
in Fluorescence
Gated mode
The spectroscopic performances of the HPGSPC summarised in Fig. 19 (click here) where the energy resolution as a function of the energy is shown over the entire energy range of the HPGSPC. Both SE and FG mode are shown. The figure also shows, for comparison, the energy resolution as a function of energy for narrowly collimated ``on-axis" X-ray sources. The full area curve can be fitted with the following relation:
The deviation from the 
law,
observed with narrowly collimated illumination, results from the
discretization (Giarrusso et al. 1989).
Figure 19: Energy resolution versus energy: both full area and collimated
case are given. Small squares show the energy resolution in FG mode
The detector's linearity has been verified for the response to both collimated and full area X-ray illumination. Results are shown in Fig. 20 (click here) where all lines and residuals are reported. The channel to energy conversion is given:
for the SE mode and
for the FG mode.
The slight difference between the two mode is due to a small offset
difference between the two electronic chains. Because of the limited number
of energy lines we could not find any evidence for the gain discontinuity
around the Xenon K-edge where a jump of 177 eV has been recently measured
(dos Santos G.M.F. et al. 1994). A L-jump discontinuity at
4.78 keV of about 110 eV has been also measured by many authors (dos
Santos et al. 1993; Lamb et al. 1987; Boella et
al. 1996). This effect is small for the HPGSPC, because, due to
the window thickness, the detection efficiency at that energy is low. Both
the K and L jump have been, however, modelled and introduced in the
response characteristic of the detector.
The effective area of the HPGSPC is mainly limited at low energies by the transmission of the beryllium entrance window, while at higher energies it is limited by the efficiency of absorption of X-rays in Xenon. in Fig. 21 (click here) we show the HPGSPC effective area as a function of energy. In the calculation have been taken into account the predicted transparency of collimator (around 75%), the beryllium window transmission and the reduction factor due to the efficiency of electronics to recognise a true X-ray.
Figure 21: HPGSPC effective area
In order to obtain the complete response matrix of the instrument, the
redistribution matrix has been experimentally determined. The
redistribution matrix depends, of course, on the energy resolution and on
the distribution of the response to a monochromatic line, but it depend
also (above 34.6 keV) on the escape fraction probabilities of 
and 
fluorescence photons. Table 4 (click here) shows the
experimentally determined escape fraction probabilities for the
and 
fluorescence photons for an incident energy
equal to 59.5 keV (Americium source) and to the 122 and 136 keV of the
lines. Escape probabilities have been normalized with respect
to the total number of detected events in SE mode.
Table 4: Escape probabilities for the Xenon 
and
fluorescence photons
Several components contribute to the background level of the HPGSPC. A major background source (Mason et al. 1983) is the diffuse isotropic atmospheric and Cosmic X and gamma ray component that manifests itself as an aperture flux that depends upon the Field of View of the collimator and as an indirect background induced through shield penetration. A second component is due to primary cosmic rays (mostly protons) and geomagnetically trapped particles that interact directly in the detector or, interacting with the spacecraft, giving rise to gamma rays which in turn penetrates the detector through the shielding and interacting with the detector walls or filling gas producing energetic Compton electrons. A third component is due to electrons and neutrons mainly produced as a result of the primary cosmic ray interaction with the atmosphere.
As the HPGSPC is a background dominated instrument for most sources, background reduction is extremely important. It is be achieved in three ways:
Precise simulation of the in-orbit background is practically
impossible on the ground. However, to provide an insight into the
background rejection capabilities of the HPGSPC two ``on ground
available" background sources have been used: the environmental
background at Laben Clean Room (that is the sea level cosmic ray
flux) and a 
gamma ray source (1.17/1.33 MeV gamma
rays) that can mimic the in-orbit background component of energetic
Compton electrons. In Fig. 22 (click here) we show the energy
reconstructed spectrum of the background (both environmental and
induced background) as seen by the HPGSPC.
Figure 22: Environmental and 
induced background as seen
by the HPGSPC
In order to evaluate and optimise the background rejection efficiency of a X-ray detector, such as the HPGSPC the statistic S defined as:
must be maximised. In the previous equation 
) is
the X-ray acceptance efficiency i.e. the fraction of X-rays, with energy
, accepted by burst length selection with respect to the total
number of X-rays events recorded by the HPGSPC and 
)
is the fraction of background events rejected because of burst length
selection with respect to the total background. The variation of burst
length efficiency 
with 
is shown in
Fig. 23 (click here) along with the Statistic S, for the 22 keV line of
source. The two curves correspond to the environmental
background at Laben and to a 
induced background. As S should
be maximum, the optimum rejection efficiency 
occurs at 55%
with an acceptance 
of 90%.
Figure 23: Background rejected and Statistic S versus energy for
source: dashed lines correspond to environmental
background; continuous line is for 
induced background.
The vertical line corresponds to a choice of burst length window
narrowly selected around the burst length of genuine X-ray events
Above 34.6 keV the situation is different. In Fig. 24 (click here)
the variation of 
is shown for the 59.5 keV line of
. A Maximum is obtained when the upper burst length
thresholds is fully open. This is due to the long tail that is
present in the Americium burst length spectrum. The vertical line in
Figs. 23 (click here) and 24 (click here) corresponds to a choice
of burst length window narrowly selected around the burst length of
genuine X-ray events.
Figure 24: Background rejected and statistic S versus energy for an
source
Finally in Fig. 25 (click here) we show the burst length rejection efficiency as a function of X-ray photon energy for an X-ray acceptance efficiency of 80%. Only events with energies within of the X-ray lines are considered.
Figure 25: Burst length versus energy for an X-ray acceptance efficiency of
80%
The background rejection efficiency 
in Fluorescence Gated mode
has been measured using both a 
gamma ray source and the
environmental background (Sims et al. 1983). 
has
been measured in a narrow band around the reference energy and the results
are shown in Table 5 (click here).
Several components contribute to the residual FG background.
First, fluorescent photons can be produced by the interaction of
charged particles (mostly protons and Compton electrons) with
the Xenon K shell electrons (Manzo et al.
1980; Middleman et al. 1967).
Secondly, Compton scattered gamma rays have a 4% probability of
producing fluorescence photons by interacting with the Xenon
K-shell electrons. However, only a minor difference has been
found between the 
and cosmic ray induced
backgrounds. This suggests, in good agreement with Sims
et al. (1983) and Ramsey et al.
(1990), that, for the HPGSPC, the majority of the FG
background is produced by high energy photons which scatter many
times in the detector shielding and body and are ultimately
absorbed in the Xenon via photoelectric interaction, showing the
same indistinguishable ``fingerprint" as genuine X-rays. An
estimate of the influence of the above described components is
currently being determined. Table 5 (click here) summarises, the
background rejection efficiency at various incident photon
energies 
using both burst length discrimination and
the FG technique.
