A simple rocking mechanism is used to move each cluster's detectors between alternate on- and off-
source fields-of-view. The rocking axes are fixed relative to the spacecraft (Figure 5.1), and the off-
source positions can be selected to be either 1.5° or 3.0° from the source, with the option of one- or
two-sided rocking. A beamswitch cycle consists of dwelling (i.e. accumulating data) at the on-source
position, moving to an off-source position, dwelling there, rapidly moving back to the on-source posi
tion, dwelling there, moving to the other off-source position, dwelling there, and finally rapidly mov
ing back to the on-source position. This cycle is then repeated continuously throughout any given
observation. The on-source dwell time can be selected to be 16, 32, 64 s or 128 s. These motions are
phased with respect to the data taking intervals such that a cluster spends the full dwell time when on-
source, but uses 2s of the dwell time allotted to off-source observing for the motion from and to the
on-source position. Data acquisition is inhibited for 2 s during these transitions off- and onto source;
consequently, 4 s less is available for the off-source dwell time each cycle. Movement of the two clus
ters is synchronized such that at least one cluster is taking data on-source at any time; the nominal
timing pattern is shown in Figure 5.4a. Since the rocking axes of the two clusters are orthogonal to
each other, four background regions are usually sampled around a given source position by HEXTE,
as shown in Figure 5.4b. Thus, the presence of a contaminating in a single background region can be
identified, and one other background field will still be available for the cluster in question. The
HEXTE clusters can be configured individually to avoid contaminating sources in advance through
use of one-sided and/or 3.0° rocking.
The rocking mechanism may also be commanded to dwell indefinitely (i.e.
stare) at any of the on- or off-source positions in order to
satisfy scientific objectives such as fast timing on bright sources where
background subtraction is not important. The cluster mechanism's position
during any data taking interval is indicated in telemetry by one of six
indications: -3°, -1.5°, 0°(two settings), +1.5°, or
+3°. A secondary position readout comes from a 10-bit shaft encoder
that provides a continuous mea sure of the modulator cam angle, and is used
as a diagnostic in case of a failure in the system.
The HEXTE detectors' field-of-view is defined passively by the honeycomb of hexagonal tubes that
make up the collimators. As for the PCA instrument, each HEXTE detector/collimator assembly in
has been aligned such that their peak response direction is close to the nominal look-direction of
RXTE. For sources off-axis, the detector open area visible through the collimators is decreased geo
metrically; the collimator response as a function of off-axis angle q and azimuthal angle is therefore
order the convolution of a hexagonal aperture with itself. For the dimensions given earlier, 50% of
on-axis response is reached at about q = 0.5, while zero response defines a roughly circular boundary
at q = 1.1. A model collimator response is shown in Figure 5.3, normalized to unity on-axis (q = 0).
The cross-sections are consistent with laboratory measurements using a real collimator. In practice,
the sharp peak of the collimator response is smoothed out somewhat by the deviations in the align
ment of the detectors within each HEXTE cluster (~1 arcmin, as measured during in-orbit checkout),
and by the slow residual motion (<0.1 arcmin) of the spacecraft's pointing axis. Ideally the collimator
response will be independent of energy, but in practice a small fraction (~1%) of the incident source
flux may be Compton-scattered off the interior of the collimator walls. The resulting detected spec
trum would then be a function of the source spectral shape, but for most sources this would prove to
be a negligible addition to the background.
Scintillation pulses are analyzed by the event selection logic and assigned a Pulse Height Analyzer
(PHA) channel number from 0 to 255 which is a measure of incident photon energy (in fact, the PHA
channel number ~= peak energy in keV). At normal gain, these channels are roughly 1 keV wide and
arranged such that the PHA channel number corresponds to the peak photon energy in keV. The min
imum acceptable energy is defined by the Lower Level Discriminator (LLD) which may be set to any
value from 5-50 keV, though values in the range 10-30 keV prove the most useful. The Upper Level
Discriminator (ULD) level defines the upper bound of the instrument, which corresponds to about
250 keV at nominal gain
The on-axis effective area of a HEXTE detector is defined as its efficiency in producing an a detected
scintillation pulse event for each incident x-ray photon, multiplied by its open area to the sky in the
on-axis look direction. Non-detected photon interactions include absorption in the detector window,
those partial energy loss events in the NaI (Compton scattering) which are captured and rejected by
the CsI layer, and photons which pass through the NaI undetected. Extensive Monte-Carlo simula
tions of the phoswich detectors have been performed. Figure 5.5 shows the effective area as a func
tion of energy for "photopeak" interactions, i.e. complete x-ray energy loss in the NaI, which is the
significant contributor to the HEXTE's sensitivity (these events appear as diagonal elements in the
response matrix); this curve has been scaled to Cluster A's net open area of 890 cm2. The sharp edge
at 33.17 keV is the K-escape energy of Iodine. Note that in order to estimate the measured count rate
for a given source, this effective area must be multiplied by the detector live-time fraction
(Section 5.3.8).
X-ray photons of a given energy will produce a spread in detected pulse height due to the Poisson
fluctuations in the number of photo-electrons produced per event. The intrinsic energy resolution of
the phoswich detectors can be described roughly by a gaussian function with FWHM increasing as
, to which must be added a small term proportional to energy due to the gain variations
across the face of each detector. For an average phoswich detector, the resolution FWHM (in channels
or keV) at PHA channel e (~=E keV) is given approximately by
Summing data from the detectors in a HEXTE cluster degrades this resolution somewhat due to this
lightly different PHA channel centroids of the line-spread function in each detector at a given energy.
Nevertheless, the 256 PHA channels (numbered 0-255) still over-sample the HEXTE resolution by at least a factor of 2. The user may select contiguous sub-ranges of channels and/or group them into
larger spectral bins, according to the Science Mode telemetry format in use.
Even with the magnetic shielding around each detector, the gain (or pulse-height/energy relation) is
affected by the magnetic fields encountered throughout an orbit. Secular changes in the gain also
occur due to aging of the phototubes. To counteract these effects, the AGC system is designed to stabilize each phoswich detector's gain, such that photon events of a given energy will always produce
counts in the same PHA channels. X-rays of 59.6 keV energy from the 241Am source interact in the
NaI in coincidence with the associated alpha particle interaction in the gain control detectors
(described above). These NaI events are used to provide the gain control feedback signal that adjusts
the PHA conversion gain to preserve the energy/PHA channel relation. Coarse gain settings are
accomplished by selecting one of 256 high voltage steps (5% gain change per step) and fine steps can
be commanded by the AGC system within the range 0.2% to 0.0125%. The AGC updates the gain
every 0.5 s. In this manner short-term gain variations are kept to <1%.
All events detected in the phoswiches in coincidence with the a-particle events in the gain control
detectors are separately accumulated into a calibration spectrum which is acquired over 32 Instrument
Data Frames (8.5 minutes). The result is a very clean 241Am spectrum with a line at 59.6 keV, a blend
around 25 to 30 keV, and the L-shell blend around 17 keV (Figure 5.7). These spectral features will
provide constant monitoring of the gain and resolution for each phoswich detector throughout the
mission. Therefore, with this system meaningful comparisons, and co-additions of datasets, may be
made between spectral observations taken many months apart.
The anti-coincidence shielding around each HEXTE cluster vetoes almost all
particle scintillations in the detectors. The remaining background spectrum
is intrinsic to the HEXTE and is dominated by x- rays emitted in the decay
of radioactive products, which are produced by high energy particles inter
acting with the detector materials (principally lead in the collimators and
iodine in the phoswiches themselves). During an orbit of the RXTE this
background will vary by a factor of 1.5 to 2. To a first approximation the
HEXTE background rate outside the SAA region is proportional to the cosmic
ray particle flux, which varies from point-to-point due to geomagnetic
cutoff. There are delayed compo nents after passage through the SAA,
however, from the activation of radioactive daughters in the detector
material itself, as well as in surrounding matter, by SAA protons.
The estimated typical background rate for the HEXTE as a function of
detected photon energy is shown in Figure 5.8, and is based upon
experience with HEAO-1 A4, taking into account the design differences such
the HEXTE's smaller field of view (significantly less diffuse x-ray flux)
and less shielding (the A4 detectors were surrounded by a massive CsI
anticoincidence shield, while the HEXTE has a plastic anticoincidence shield
and small amounts of lead shielding).
Background estimation is provided to a very good first order by the
source/background switching. The current DEFAULT dwell time for the HEXTE is
16 s, which is the least efficient; however the effectiveness of longer duty
cycles in removing systematic background variations is still under inves
tigation established. Estimates of secular variations in the HEXTE
background have been made based on HEAO-A4 data, and predict that systematic
errors in background subtraction can be kept below 0.2% over a 105
s observation of a faint source; longer exposure times than this will
not be able to elicit a fainter detection since they will be dominated by
systematic errors.
For a given 16 s on-source data accumulation, the background rate estimate will be the average of the
last 6 s of the previous off-source accumulation plus the first 6 s of the following off-source accumu
lation. (Remember that 2 s is lost at each end of the off-source observation while the rocking mecha
nism is in motion). For a typical background rate of 90 count/s per HEXTE cluster, the resulting
interpolated background rate is limited by Poisson statistics to about 3% accuracy per cycle.
The observer will typically co-add the two off-source datasets for each cluster, after testing for the
presence of a contaminating source. Users may also wish to make an empirical fit to the off-source
data versus time for estimating the on-source background contribution. For those observations where
off-source observations are not made (e.g. dwelling on-source for temporal investigations of bright
sources), the HEXTE background may be modelled using orbital parameters and particle veto rates,
but not to the precision available when beamswitching. If background estimates are important in this
case, users are encouraged to contact the HEXTE team for advice.
The HEXTE is capable of time-tagging events to 7.6 ms when used in Event List mode. For high
count rate sources, which would saturate the telemetry in this mode, users may still perform fast tim
ing in Temporal Bin Mode, which can produce light curves with time bins as small as 1ms
(Section 5.6). Calibration of the absolute timing accuracy is performed via the spacecraft clock, while
relative timing tests between the HEXTE and the PCA have verified the coincidence to within the 7.6
ms precision of the HEXTE measurements.
Dead time in the HEXTE originates in the detector electronics, which takes 16-30 ms to process a
typical scintillation pulse (but much longer for MeV particles, see below). This dead time is measured
for each HEXTE detector assembly by a gated counter which increments for each ½ spacecraft
clock cycle (1.91 ms) for which the detector electronics is busy. The dead time counter's value is nor
mally telemetered every 16 s (but as fast as every 1 s in Spectral Bin mode). For dead time estimates
on shorter timescales, modelling of the source behavior becomes necessary.
While an accepted event is being pulse-height analyzed after passing the event-selection criteria, the
HEXTE processor can also register the detection of up to 3 Lost Events, which are those accepted
events which were received while the analyzer electronics were still busy (and the dead time counter
was incrementing). Although no energy information is available for these events, they can be used to
model bright, rapid (<1s) changes in source flux, and as an aid to absolute flux determination.
The dead time counter can therefore be used to provide an accurate estimate of dead time on 16 s to 1
s intervals, depending on Science Mode, while the Lost Events bits provide a crude means of estimat
ing precisely when that dead time occurred on much shorter time-scales. In practice, though, it is
unlikely the Lost Events data will be required except for operations under extreme high background
conditions, since the dead time per event is normally short compared to the count rates from the
brightest sources.