next up previous contents
Next: The on- and off-line Up: The Late SWTCs Previous: Automatic computation of the   Contents

Benefits and Drawbacks

The requirements for matching the SWTC 1 are similar to the case of the early SWTC 1, i.e., that in at least two units the same couple of thresholds, one for each energy band, is exceeded; the only difference resides in the numerical values: in this case, all these thresholds ($n_1^{(G)}=3.7$ and $n_1^{(A)}=2.5$) are lower than their corresponding parameters in the case of the early SWTCs ($n^{(G)}=6.0$ and $n^{(A)}=3.0$): this does not increase the number of false events due to both statistical fluctuations and passages of high-energy particles, thanks to the best background fit performed in this case: this limits the number of spurious detections, making it possible to adopt lower thresholds.

The SWTC 2 is a more sensitive version of the early SWTC 1: in fact, it requires that in at least two detector units, both the GRBM and AC thresholds are exceeded, but in this case, the threshold parameters for the second brightest unit are little lower than for the brightest one: $n_2^{(G)}=2.8$ instead of $n_1^{(G)}=3.7$ and $n_2^{(A)}=2.0$ instead of $n_1^{(A)}=2.5$. Whenever the SWTC 2 is matched, the SWTC 1 is, too; this does not make the SWTC 1 useless, which is therefore stronger, because the candidate GRBs caught with SWTC 1 are more likely to be true GRBs than those caught only with SWTC 2.

The SWTC 3 has been thought to catch all those GRBs, that produce a weak signal in each single unit, due either to an intrinsic burst faintness or to the arrival direction far from all detectors' axes.

The SWTC 4 is particularly sensitive for those detector units co-aligned with the WFCs, i.e. nn. 1 and 3: whenever the same thresholds as those of SWTC 1 are matched in at least one of these two units, and this holds for at least three contiguous bins, the SWTC 4 triggers a good burst candidate: the reason why the GRB quest pays more attention to these units is clear: since the SWTC 4 is suitable for detecting those bursts (except for the very short bursts, i.e. with $T_{90}\ll 1$s), whose directions are likely to be within the fields of view of the WFCs, on principle, this SWTC could contribute to increase the number of weak bursts with a prompt X-ray counterpart localized by one WFC.

Actually, this happened indeed in some cases: as already described above, the case of GRB010518 (fig. [*]), was localized thanks to the on-line detection from the SWTC 4; other similar cases are presented in section [*].

As regards the threshold on the Hardness Ratio $\overline{HR_w}$ (eq. [*]), there are several benefits: this condition exploits the spectral hardness of the GRBs, which marks them from the majority of all the other transient events in the 40-700 and $>$ 100 keV energy ranges. In particular, this threshold automatically rejects almost all the spikes caused by high-energy charged particles, since typically in those cases it is $\overline{HR_w} \leq 0.1$. Another class of transient events mostly rejected by this constraint is represented by the solar X-ray flares, many of which are often very bright (total counts of $\sim 10^{4\div 5}$) and lasting a few hundreds seconds). Therefore, the possibility, offered by this spectral constraint, of automatically distinguishing the true GRBs from other events has turned out to play a key role in the on-line GRB quest.

The quest performed with the late SWTCs has two main drawbacks with respect to that with the early SWTCs: a greater number of time bins that cannot be scanned, because the background cannot be fitted, and some biases in the automatic calculation of the total counts for the longest GRBs.

As for the first point, the capability of the early SWTCs of scanning the same bins that are taken into account to compute the average background counts, allows the discoveries of several bursts close to gaps, but, on the other side, yields a great number of fake detections: since the late SWTCs are applied to the on-line quest, it is preferable to keep the number of false events automatically detected low, at the cost of some lost bursts. Typically, the fraction of unscanned bins is $\sim$ 75-90% in the case of the on-line quest, which represents the worse case from this point of view, because the ratemeters are grouped orbit by orbit and have more breaks than the archive data.

The unscanned bins left by the late SWTC quest in the archive ($\sim$ 80-95%), have been covered by the early SWTCs, thus strongly decreasing the number of GRBs missed by the previous quest.

The second deficiency comes out when the GRB detected lasts longer than $n_a$ time bins; in this case, the post-burst interval used to fit the background includes the last part of the burst itself, biasing, in particular, the automatic computation of the total counts due to the GRB in two ways: first, by biasing the background level estimate; second, by ignoring the counts measured outside the time interval $i-n_b+N_b \leq k < i+n_a$. On the other hand, the greater $n_a$, $N_a$, $n_b$ and $N_b$, the greater the fraction of unscanned bins: thus, the best choice for the value of $n_a$ is achieved with a fine tuning between the two limits. In our case, $n_a=50$ (on-line) and $n_a=91$ (off-line), this bias takes place more frequently in the on-line case, whenever the burst lasts longer than 50 s: however, when the automatic estimate of the burst duration is longer than 50 s, it seldom happens that that last part ($>$ 50 s) contains a significant fraction of the total counts; thus, this severely limits the number of biased bursts caught on line.

On the other hand, every NTB detected during the off-line quest, is examined by visual inspection, to check the goodness of the background fit: whenever the fit and/or the total counts' estimate are judged incorrect or even little biased, the operator manually fits the background, by properly choosing the values for $n_a$, $N_a$, $n_b$ and $N_b$, and, in some rare cases, by fitting with higher degree polynomials.

A final remark concerns the time resolution used for the off-line quest: while for the on-line case only the quest is performed only on the original 1 s bins, in the off-line case the same quest applies with four different rebin times: 1, 2, 4 and 8 s, respectively. The motivations and the results of this choice are discussed in section [*].

next up previous contents
Next: The on- and off-line Up: The Late SWTCs Previous: Automatic computation of the   Contents
Cristiano Guidorzi 2003-07-31