ergs-cm2-sec).
From the initial selection of targets through the first GTO and GO proposal
rounds, it is clear that many of the favorite AXAF targets were discovered
by previous X-ray missions, and are much brighter than the AXAF limiting
source sensitivity.
In order to study these very bright sources we must accomodate the effects of pile-up, as described in the preceeding sections. If the objective is to obtain a spectrum of such a bright target, by using one of the transmission gratings, then pileup is not a problem, as pile-up effects will be greatest on the highest counting rate in the image (which is typically the zeroth order image). The critical quantity to be determined for the analysis of the dispersed spectrum is the centroid of this zeroth order image. Pileup effects will not affect the centroid determination of the zero order position of the spectrum. The grating has the property of dispersing the X-rays over a large portion of the ACIS-S array, so the projected flux in the dispersed spectrum is much lower than the zeroth order (as only a small portion of the spectrum falls on a single pixel).
If instead the intent is to obtain a spectrum of a very bright target, either integrated over time, or for time variable targets, to carry out phase dependent or time dependent spectroscopy, then by using the continuous clocking mode it is possible to reduce the pile-up substantially, since the effective exposure time is only 3.2 msec. (Note that the source must be bright, because the lack of timed exposures makes the incident photon's position ambiguous in one dimension. If we assume all photons detected originate from the target, then this loss of information about position doesn't matter.)
In the future, should high time resolution become important for certain
classes of observations, it is possible to reduce the exposure time to
about 0.1 msec. The spectral fitting of data obtained from continuous clocking
may not be as reliable as in normal faint mode operation, since the on-board
event reconstruction currently does not use all eight adjacent pixels to
the event pixel. Instead only a 3x1 region is examined for split threshold
crossings. Because some split events result in charge in vertically adjacent
pixels, split events are confused with single events in 3x1 event recognition.
A spectrum collected using 3x1 event data exhibits low energy tails as
compared to data processed by 3x3 event recognition. Figure 6.39
shows the observed spectrum of an Al K
line
spectrum collected in continuous clocking mode at XRCF.
Figure 6.39: Al K-alpha spectrum
taken in Continous Clocking mode
using 3x1 event processing (incident flux produced a rate of 96.3 events/sec).
Another effect of using 3x1 event recognition is multiple events produced by a single photon. In the data shown in Figure 6.39, for example, adding up only the peaks and multiplying by the number of piled-up events in each peak results in 26,562 events. The total number of events inferred from the flux measured by the Beam Normalization Detectors (BND) times the area times the time gives 27,742 events. On the other hand, the result of fitting the peaks and the continuum and multiplying by the number of piled-up events is to infer 36,368 events. The reason that total event rate exceeds the correct rate is that one event can be counted as many as three times, as it can split along both directions of the readout path.
Future software patches to the flight software will include the 3x3 event processing of each isolated event in contiuous clocking mode, so that the normal response matrices for spectral analysis should work and the inferred incident rate should accurately reflect the incident photon rate.