The algorithm described can be straightforwardly applied for screening ROSAT PSPC observations. Two major background anomalies that may cause a decrease in signal to noise ratio of sources can be identified in the background light curve of a ROSAT PSPC, namely: 1) high particle background as identified by high master veto count-rate, and 2) times of high solar background and other short-term enhancements. We will separately discuss the effect on the light curve of each of them, and we will show that, while the PB has to be screened by an "ad-hoc" cut, the SB can be effectively reduced by the procedure described in Sect. 2.
During a typical ROSAT PSPC observation, several kinds of charged
particles, including cosmic rays, hit the detector, and cause electron
cascades which are afterward marked as X-ray events by the read-out
electronics. The PSPC has been equipped with an anti-coincidence
counter, which detects, logs, and rejects these particle events. The
rate of the rejected events is called the Master Veto (MV) rate.
Nominally, 
% of particle events are rejected. S92 have shown
that, for typical MV values (MV 
cnt s-1), the residual
particle background (that is the particle events which are not rejected
by the anti-coincidence method) is negligible with respect to other
background components (e.g. cosmic).
Under some circumstances, the satellite enters regions of high particle
background (HPB) during an observing interval, signalled by a sudden
increase of the MV rate. In these cases, the residual particle
background may become comparable to other background rates (e.g.
cosmic). To reduce the contamination of HPB, S92 suggested the screening
out of
time intervals where the MV rate is >170 cnt s-1. This choice
rejects 
% of the exposure time. We have
verified that rejecting such an high fraction of the time may yield a
net loss of the number of point-like sources detected in the screened
observation and a systematic decrease of their SNR's.
On the other hand, Figs. 6 and 8 of S92 show that, for a MV rate up to
400 cnt s-1, the residual particle background is still lower than
2 cnt s-1 (40% and 15% of the minimum and typical X-ray cosmic
background, respectively), and that choosing the MV threshold 
cnt s-1 yields a significant reduction of the discarded time
intervals. However, as the MV rate increases, the detector dead time
also increases. This is clearly shown in Fig. 1 (click here), which
reports the MV light curve (left panel) of a PSPC observation pointed
to the NE rim of the Vela supernova remnant (Bocchino et al.
1994) along with the accepted event rate of the same time
interval (right panel). We note that, towards the end of the exposure,
the MV rate increases significantly; when it becomes larger than
about 250 cnt s-1, the accepted event rate shows an apparent drop
due to high detector dead time. This evidence suggested, as a
first step of our PSPC screening procedure, to reject the time
intervals with MV rate >250 cnt s-1. This choice is not
restrictive in terms of exposure times (according to S92, only less
than 1% of the exposure is rejected when only intervals with MV rate
<250 cnt s-1 are retained), and allows us to reject observation
intervals with critical dead time. Adopting this MV threshold, we can
estimate, on the basis of the typical and minimum rate of the cosmic
X-ray background and Fig. 8 of S92, that the PB contribution to the
total background is always < 25%, with a typical value of
<5%.
Figure 1: Left: Master Veto (MV) light curve of the first Observation
Interval (OBI) of the ROSAT PSPC sequence RP 200133 pointed toward
the NE rim of the Vela supernova remnant (3944 sec total exposure). On
average the MV rate is <250 cnt s-1, but at the end of the OBI the
satellite has entered an high particle background region which increases
the rejection rate above 1000 cnt s-1. Right: Accepted X-ray events
(AE) rate for the same OBI as before. There is an apparent decrease of
the rate corresponding to the high particle background interval of the
left panel, because the high rejection rate increases the detector dead
time. The dip is not real also because no source intrinsic variability
is expected in the case of the rim of the Vela supernova remnant on the
time scale of tens of seconds
A typical background light-curve of a ROSAT PSPC or HRI observation is characterized by occasional steep enhancements in the total accepted event count-rate. Some of these excess background intervals are due to solar radiation scattered by Earth's atmosphere and others have been identified as auroral X-rays. In Fig. 2 (click here) (left panel), for instance, we show a background light curve of a ROSAT/PSPC observation which suffers of several contaminating spikes. According to SF93, the spikes are due to scattered solar X-ray radiation which enters the mirror assembly when the pointing direction crosses the sunlit limb of Earth.
Figure 2: Left: Background light curve of a PSPC field pointed towards the
Pleiades open cluster. A large number of steep background enhancements
are present along the entire 
s observation interval.
Dashed line marks the median background level, dot dashed line is the
threshold level chosen by the screening algorithm described in the text
to operate the rejection of the contaminating spikes. Right: The
Efficiency Function defined in the text versus the fraction of the total
exposure time screened out. The maximum of the curve is automatically
selected by the algorithm and corresponds to the cut marked by the dot
dashed line in the left panel
Our screening algorithm is capable of removing these high background
intervals. To use Eq. (11 (click here)) in the PSPC case, we have
chosen a circular source region which includes 90% of the source
photons (
= 0.9) and an annular background region whose inner
radius is equal to the source region radius and whose area is three
times the source region area (r=3). From the integration of the PSPC
point spread function (PSF) at 0.3 keV given by Hasinger et al
(1993), we found for 
a typical value of 0.08 for off-axis
angles in the range 0'-50'. With these choices , , and r
are almost independent on off-axis position and the only relevant
parameters remain S, and B
.
Before we describe operatively the screening procedure, we discuss, on the basis of results of the simulation tests, the dependence of our screening procedure on the choice of the parameters on which the screening procedure is tailored (i.e. off-axis angle and count-rate of the point source of which the efficiency function is to be maximized). For these purposes we have generated a simulated PSPC image with exposure time and background light curve similar to a PSPC field of the Pleiades young open cluster with nearly 40 000 s total exposure time (rp200068, PI R. Rosner) presenting many steep background enhancements (Fig. 2 (click here)). We have conducted extensive tests on the simulated Pleiades PSPC field under study, running the screening procedure tailored to sources of different off-axis angles (hereafter, the template sources). For each run, a template source was generated with a number of counts between the detection threshold and 10 times the threshold. C, T, S and B were computed accordingly, and the optimal screening was found by maximizing the efficiency function (Eq. 11 (click here)). This approach has allowed us to study the behavior of the algorithm by varying the characteristic of the template source.
The tests show that when the screening is tailored to fainter template sources, more time is screened out to obtain the maximum SNR. This means that, when the screening procedure is tailored to obtain maximum SNR for faint sources, more time intervals are screened out than necessary for obtaining maximum SNR for brighter sources. Nevertheless a screening procedure tailored to faint sources usually produces also an increase in SNR for bright sources, although not the maximum obtainable for these latter sources.
Once it is clear that it is advantageous to tailor the screening procedure to the faintest source that can be detected in the field under study (i.e. S is equal to the detection threshold for a given image), we have to investigate whether the procedure is sensitive to the choice of the off-axis angle of the template source. In fact, as we move outside from the center of the field of view, the width of the PSF of the detector increases, and the detection cell needs to be increased in order to keep the fraction of source counts falling within it constant. This implies an increase of the background counts B, whose collection cell scales with the area of the detection cell. Furthermore, the larger the off-axis angle, the higher the detection threshold (due to PSF widening and vignetting effect); therefore, the behavior of S/B in the efficiency function E to be maximized is not directly obvious.
| Off-axis | Cell Radius | S | X |  |
 |
 |
 |
| (arcmin) | (arcmin) | (cnt/s) | (%) | ||||
| 0 | 0.475 |  | 3.7 | 1.02 | 1.02 | 1.02 | 1.02 |
| 2 | 0.492 |  | 3.9 | 1.02 | 1.02 | 1.02 | 1.02 |
| 4 | 0.492 |  | 3.7 | 1.02 | 1.02 | 1.02 | 1.02 |
| 6 | 0.542 |  | 3.9 | 1.03 | 1.02 | 1.02 | 1.02 |
| 8 | 0.592 |  | 4.2 | 1.03 | 1.03 | 1.03 | 1.03 |
| 10 | 0.692 |  | 4.5 | 1.03 | 1.03 | 1.03 | 1.03 |
| 15 | 0.992 |  | 5.4 | 1.04 | 1.04 | 1.04 | |
| 20 | 1.408 |  | 5.4 | 1.04 | 1.04 | 1.04 | |
| 25 | 1.875 |  | 5.6 | 1.05 | 1.05 | ||
| 30 | 2.391 |  | 5.6 | 1.05 | 1.05 |
Table 1 (click here) summarizes the results of these tests. Column 1 is
the off-axis angle; Col. 2 gives the associated detection cell
radius; Col. 3 is the detection threshold for that off-axis angle
derived using the Wavelet transform detection algorithm of Damiani
et al. (1997a); Col. 4 gives the amount of screened time (x); Col.
5 gives the ratio 
i.e. the SNR gain for the
source to which the screening is tailored; Cols. 6, 7, and 8 give the SNR
gain which would be obtained for the source if the procedure had been
tailored to 10', 20' or 30' off-axis angles. This table shows that
the choice of tailoring the algorithm to large off-axis angles does not
affect the gain in SNR of sources at smaller off-axis angles. For instance,
an on-axis source (first row in in Table 1) would have a gain of 2% in SNR
with the choice of tailoring the algorithm to 30' off-axis angle (Col. 8),
which is the same gain which would be obtained tailoring the algorithm to
the on-axis source (Col. 5). The table also shows that the optimal screening
of sources at large off-axis angles is obtained tailoring the algorithm to
template sources also at large off-axis angles. In fact, the last row of the
table shows that the optimal screening for a 
off-axis source is
obtained with a cut of 5.6% of the exposure time (Col. 4), while
tailoring the algorithm at smaller off-axis angle would have yielded a
smaller fraction of rejected time, and thus a SNR gain for the
source which is lower than the maximum achievable (Fig.
2 (click here)).
However, it is evident that the screening algorithm is not very sensitive to the chosen off-axis angle. This indicates that the increase of the detection cell size, and thus B, and the increase of the detection threshold that occurs at larger off-axis angles, in some way compensates, keeping S/B nearly constant. On the basis of these results, we have chosen to tailor our screening procedure to an off-axis angle of 30', where we have the highest gain in SNR and we still have acceptable PSPC performances both in terms of sensitivity and spatial resolution.