2. Data analysis
Three observations have been made of X 1755-338 with the Rosat PSPC. In two of these there is little or no sign of dipping. We present results for the other longer observation made on 1993 March 29, lasting 18 h. Source data were extracted from a circle of radius 2 arcmin, taking into account the dust scattered X-ray halo of the source discussed by Predehl and Schmitt (1995). Photons scattered in the halo are not expected to show dipping because of the variable time delays introduced which smear out time-variability. From the figure given by Predehl and Schmitt, it can be estimated that a radius of 2 arcmin includes 96% of the unscattered source counts, but excludes 97% of the halo, which should contribute about 3 c/s in the total PSPC band. For source extraction from a 2 arcmin region, background subtraction is not very important in this bright source, normally less than 0.1% of the total count rate, except for the subtraction of sharply rising background at the ends of each of the first 3 sections of data in Fig. 1 which in fact caused the switch-off of the detectors in the following data gaps. We obtained background data from an annulus between 0.22 and 0.28 , excluding point sources outside this region, and excluding most of the halo which Predehl and Schmitt show is becoming small at 0.22 . It is important for spectral fitting that background subtraction is made from a region with sufficiently high background count, and this region is the best for this requirement.
The light curve of the source in the band 0.5 - 2.0 keV is shown in Fig. 1a with 160 s timebins. Ideally, longer binning would be better, but the shortness of some of the data sections do not allow this. Consequently, it is still possible to see the effects in the light curves of the wobbling of the telescope to prevent occultation of sources behind the wires, and this effect dominates over Poisson noise in the light curves. A section of data can be seen with count rate reduced by 10% , which, allowing for data gaps, lasts between 25 and 68 m. The duration is consistent with the duration of dipping seen in the Exosat ME of 40 m. The depth of dipping in the ME was typically 20% (1 - 10 keV), although one dip at 12% was seen. Thus there is an indication that the depth of dipping is less in the 0.1 - 2.0 keV band. The lack of other dips is consistent with the orbital period of about 4.4 h which predicts that other dips would occur during the data gaps (although there is a possibility that the first section of data in the light curve is contaminated at the end by the onset of dipping). Finally, our spectral fitting results are fully consistent with this low intensity period being a dip.
We extracted spectra for each of the non-dip sections of data in the light curve lasting more than 1500 s, and also dip data (1700 s). These were corrected by subtracting the background and correcting for deadtime and vignetting. The spectra were rebinned into 24 channels (after excluding channels below 0.1 keV and above 2.0 keV), and a systematic error of 2% was added conservatively to each spectral channel. The results were completely consistent with those obtained fitting the spectra in primitive channels without rebinning. We present results below obtained using the January 1993 instrument response as appropriate to this observation in the later part of the Rosat mission (see Fiore et al. 1994). However to assess systematic errors, the data were also analysed using the alternative March 1992 response. It was found that parameters derived from spectral fitting changed by only 1.5% to 5%, giving confidence in the results.
© European Southern Observatory (ESO) 1997