About ROSAT	ROSAT Home Page	ROSAT Images

F.9 Comments of the PSPC Corrections

 

 

quantity min max mean st.dev. relevant corrections

-1.0156 +0.0312 -0.4942 0.2890 NLC

0.0000 +39.594 +0.3654 1.3463 GSC

-1.0156 +39.000 -0.1289 1.3768 NLC + GSC

+2.4844 +113.57 +17.250 13.409 TGC

+1.4844 +152.57 +17.121 14.364 NLC + GSC + TGC

-17.070 +18.945 -0.4353 1.1322 SGC

+1.3359 +162.56 +16.686 14.032 NLC + GSC + TGC + SGC

-9.0000 +20.750 +0.1771 2.6969 EPC

-21.875 +24.125 +3.3085 6.7258 WC

-22.750 +30.500 +3.4856 7.5448 EPC + WC

-2.0000 +1.7500 -0.2686 0.6963 FC

-21.500 +29.250 +3.2170 6.9006 EPC + WC + FC

-18.500 +19.625 +2.0454 8.7856 EPC

-21.625 +16.750 -4.5542 5.1255 WC

-36.625 +32.625 -2.5088 10.726 EPC + WC

-1.8750 +1.8750 +0.0447 0.7187 FC

-34.750 +30.750 -2.4640 10.296 EPC + WC + FC

Table F.3 shows the effect of each correction step. It does not give the dependence of a ``late'' correction on an ``early'' correction (e.g., how do changes in the effective ADC widths affect ). Nevertheless, at least some features become already clear: the mean shift , the large effect of the gain corrections in low-gain mode (maximum ), the large spread in the electronic position correction along the y-axis. Here we want to summarize the previously introduced corrections and possible improvements.

BAL:

Some improvements expected.

NLC:

It appears straight-forward to enhance the accuracy of the correction table. During the ROSAT mission far more than PSPC events have been recorded. Adding all RAW_AMPL information into a single spectrum and subtracting a smooth spline function as done with the ground calibration data would yield an improved correction. However, as the ADC non-linearities are small, this effort may be questioned.

GSC:

The rate-gain-effect is not yet fully investigated. Soft photons of strong point sources suffer from a saturation effect similar to harder X-rays (E > 1keV) at low count rates. First ground calibrations at the PANTER test facility (monochromatic X-ray source, at two energies) indicate an energy-dependent effect of the order of at 1.5-3keV for a rate of 500s . Further measurements had been suggested (variable intensities and more energies were necessary, also a variation of the high voltage).

TGC:

No detailed description of the temporal gain correction function has been found. The coefficients as well as the function itself occur somewhat arbitrary (e.g., the coefficients do not decrease rapidly for increasing order).

EPC:

The number of 4000 table elements seems to be sufficient, no improvements to be expected.

SGC:

Ground calibrations with slit masks were used to construct correction tables. 800 spectra along each detector axis at several energies were obtained (samples of 10 detector pixels). As the correction varies significantly from one entry to the next the spatial binning of 10 (all positions within one spatial bin are treated the same) introduces artificial steps in the spectral distribution. It is straight-forward to create improved correction tables using the already existing slit-mask data if a sliding-window technique is applied. Sampling all events with (with the window size ) one could obtain 8192 spectra and hence 8192 correction values (both in x and in y direction if desired). Also introducing a weight function (more distant positions are rated less than positions closer to y) could keep the statistics high and still lead to an improved correction. This does not require any further measurement. The computational effort seems to be justified as the SGC has a major effect on the data quality as well as on calibration tasks itself (like the construction of a detector response matrix).

Additionally, the energy-dependent correction table has been determined using a physical model, it should not be a principal problem to extend the table beyond channel to make full use of data obtained in low-gain mode. At the moment high-energy events are under-corrected in amplitude.

WC:

The correction procedure implemented in the SASS routine does not fully work correctly [Englhauser1995]: this is due to the (apparently) unnecessary use of random numbers and due to insufficient interpolation of the correction data. Englhauser has developed an alternative algorithm without these weak points.

Similar to the SGC the energy-dependent correction table has been determined using a physical model; it should not be a principal problem to extend the table beyond channel to make full use of data obtained in low-gain mode. At the moment high-energy events are over-corrected in position.

FC:

No effect on spectra, spatial shift detector pixels. The case is not treated correctly due to an improper IF-clause in the routine: it then uses the computed correction of the previous photon.

Further items:

Apparently there is a shift in amplitude during the gain correction process. The NLC shifts the amplitude (on average) by -0.5, the final introduces another (statistical) shift to lower amplitudes. There is doubt whether this is intentional and/or whether this is accounted for at some other place. This may be important as other spectral corrections are minor at lower channels. This effect predicts a significant excess at lower energies because there is a steep fall-off of the spectra in this range; such an effect has been found. At higher energies the small offset is smeared out due to the large spectral corrections. The observed excess at higher energies cannot be explained by this effect as it would predict a small deficiency.