Single Point Source

This is the simplest case, a circular box should be used in xrtsort to extract the source counts into a spectrum (BIN_AXIS=7). There is one subtlety which must be considered. The PSF of the PSPC is very energy dependent (Hasinger et al 1992), e,g, on-axis the 95% radius of a point source is 0.65 arcminutes at 0.5 keV, but 8.8 arcminutes at 1.7 keV. Below $\sim 0.3 keV$ (channel 30), PSPC images suffer from an effect known as 'electronic ghost imaging'. This causes events to be recorded at the intersection of anode wires rather than at their true position and results in the image appearing like a rectangular grid rather than a point source. This effect varies from source to source and so is impossible to correct for in the software. It has been found that a source radius of 3 arcminutes usually includes most of the counts at all energies and produces good results as a rule. If low energies are not important and/or the source is very faint, this can be reduced to $\sim 1$ arcminute to improve the signal to noise. If very low energy counts are important for a particular source, e.g. if it is a very soft source, then the radius which includes most of the soft counts needs to be found. This may be done by displaying the low energy image with idisplay, centroiding on the source and producing a radial profile with iradial. The extraction radius should be set to the point where the source profile merges into the background.

Having produced the source spectrum, it should be background subtracted and exposure corrected as discussed earlier. After this a detector matrix needs to be created using the program xrtresp. The detector matrix is unique for each source, being a function of off-axis angle and source box radius. It is written directly into the spectral datafile. This file may be used directly in the ASTERIX spectral fitting package, sfit, or within xspec after format conversion by ast2xsp.

For weak or very soft sources, modelling the particle component of the background correctly is important. In this case the data should be "cleaned" before sorting by only using times when the Master Veto Rate parameter is between 0 and 170.