ROSAT Guest Observer Facility

ROSAT Status Report #54:

March 30, 1993



REPORT FROM TRIP TO MPE TO DISCUSS ROSAT PSPC CALIBRATION

1993 Mar 08 - 12


Contents

  1. Overview
  2. Spectral Response
  3. Off-axis Point Spread Function
  4. Boresighting
  5. SASS Summary
  6. Absolute flux calibration
  7. New Documentation

1.0 Overview

The trip was an informal visit to MPE by Jane Turner (GSFC ROSAT GOF), Ian George (HEASARC) & Julian Osborne (UK RDAC) to discuss the status of a number of ROSAT PSPC Calibration issues. Steve Snowden & Gunther Hasinger acted as primary hosts, but discussions were held with many other members of MPE, all of whom were extremely accommodating. This note is intended to summarize the discussions. Copies of the new technical memos in 7.0 can be obtained by request from the ROSAT GOF. It is intended that most technical documentation will be made available online in a very short time (< 1 month we hope).

2.0 The PSPC Spectral Response

As is fairly well known, a lot of work as been carried out by Hasinger & Snowden over the last 6 months in connection with the spectral response of the PSPC.

2.1 The Redistribution Matrices

2.1.1 Versions & Configuration Control

Several matrices of varying quality are being used by the community. The "official" PSPC detector redistribution matrices (ie excluding the effective area vector) are those supplied by SASS. For clarity it would be very useful to cross-identify all matrices available within SASS, EXSAS, PROS, ASTERIX & the OGIP Caldb. A direct comparison between the commonly used matrices and effective area curves will also be written up in the form of a memo including ratio plots etc

"DRM 36" (circa 92 Dec) released in the USA as pspcb_93jan12.rsp and pspcc_93jan12.rsp

This is the latest/current matrix and is recommended for use in all cases. It should be noted that this matrix has had a one channel (of the 256) downward shift applied with respect to all previous matrices. This is equivalent to a change in the zero-offest of the ADC (within the uncertainties given the lack of ADC zero-offset measurement pre-launch) and represents a 5% lowering of the gain. The necessity of such a change was determined from the fits to several datasets, with the formal actual best-fit actually being slightly less than a whole channel, but close enough (and uncertain enough) for one channel to be used for simplicity.

Gunther Hasinger is updating the "Archive User's Guide" to PSPC data to include some discussion of the construction of the latest matrix. This will be made public by the OGIP & RDAC as soon the updated version is available.

2.1.2 Current Uncertainties with "DRM 36" (pspcb_93jan12.rsp / pspcc_93jan12.rsp)

AT HIGH ENERGIES, there are two (competing) uncertainties: a) The gain saturation correction (used to correct individual events) could only be measured up to an energy of 1.7keV. However the parameterization of the correction factor has then to be extrapolated to higher energies. Furthermore, the correction factor was only measured at one (the nominal) gain value, but is expected to be a function of gain. b) The optical constants of gold are not well known, thus neither are the mirror effective areas at high energies. (Even if better data are obtained for the optical constants of gold (from other sources) there will still be the possibility of contamination of the mirror surface leading to a non-predictable deviation of the mirror properties).

AT LOW ENERGIES, there are also two competing uncertainties: c) the window transmission function (ie thickness) d) the ADC zero-offset mentioned above.

Gain changes of approx +/- 1 channel have been observed during an orbit. This is most likely due to a differential temperature effect between the counter gas and that in the tanks. Gas is supplied to the counter every 15s, the flow being density-regulated (ie Pressure & Temp measured). This is thought to be a small effect relative to other uncertainties. Also, detailed information of gain versus detector parameters are only available for 3 orbits.

A handful of PSPC observations have been found to be significantly better fit using the old matrix (pspcb_92mar11.rsp). However very little is currently known about whether the models assumed in these cases are astrophysically likely.

The bottom line is that further matrices are not being developed at present and hence that DRM 36 is likely to be the default until the results of the PANTER recalibration measurements have been digested (see 2.4).

Community comments on the current matrix are welcome and should be directed to LHEAVX::TURNER.

2.2 Total Spectral Response (Effective Area)

There are now 3 sets of official TOTAL effective areas which have been distributed by SASS. Just to be clear, the total effective area files contain: a) the on-axis mirror effective area b) the mirror vignetting function (see also 2.3), c) the window transmission (including the wire mesh), and d) the gas absorption function, with only c) changing between the 3 datasets. However d) is to be updated in the near future to include effects due to the bulge in the PSPC window due to gas pressure (which is likely to amount to a few % changes in the total effective area above 2-2.5 keV). It is likely that this change will be implimented into the next (major ?) SASS release.

It should be noted that up until the latest datasets (SASS_AREA_B_NEW2 & *.._C_NEW2) rigorous checks were not made that all components were included correctly. The 1st datasets (SASS_AREA_B etc) are likely to have a problem at low energies. Based on high gain HZ43 observations, the Ppl thickness (within the window) was reduced by 1sigma to produce SASS_AREA_C_NEW1. SASS_AREA_B_NEW1 is thought to have been unchanged. Unfortunately there certainly is a low energy problem in these data. The SASS_AREA_B_NEW2 & *.._C_NEW2 have been adjusted for PI < 10 to reduce this low energy problem. There is also a few % change at the C edge cf SASS_AREA_B_NEW1 etc (which had gone unnoticed).

It should be noted that the detector maps do show variations at the few % level on scales of tens of arcminutes. This is consistent with window thickness variation at fabrication. The 0.25keV map may also have a top-bottom (in detector coordinates) gradient at a similar level. Furthermore, some ablation is possible within the central, most-used region of the detector. Currently however, it is not intended to fold these effects into the effective area datasets, although the effects are under confirmation/investigation.

See also section 2.7

2.3 Vignetting

Silvano Molendi has taken Steve's Exposure Maps, constructed from the survey data and (masking the ring and ribs) extracted counts in annuli out to the edge of the detector. The count rate ratios can then give the vignetting function. The plotted data are of very high S/N, and show excellent agreement with the preflight vignetting function (ie double plaw). Theres a small discrepancy above 40 arcmin, where we observe a dip in the survey data versus the model, this rises up again and goes above the model, crossing it at about 54 arcmin. The size of this discrepancy & the radius where is starts seems to be a function of energy. The origin of this effect is not yet understood.

2.4 Recalibration Plans

MPE may be able to obtain a week or two of PANTER time in April/May to perform a number of recalibration experiments of the PSPC using the engineering model. It should be noted however that this time is not guaranteed. Specifically measurements will be made:

  • at 3 or 4 gain states in order to investigate the amplitude of all gain- dependent effects,
  • using more energies than the original pre-launch cals to better determine the energy dependent effects,
  • and to explore the effects of very high localized count rates (employing an X-ray mirror) to explore the turn-on of the "rate-gain effect".

MPE does not yet have any formal plans to measure the transmission of the PSPC window. However, Leicester have a forthcoming run on the Daresbury synchrotron facility, and Steve Snowden has delivered a piece of window to Leicester.

2.5 Changes to SASS Calibration Inputs:

Since 1992 Nov, the following changes have been made to SASS:

1) The "Golden Disk" is the term used to describe the spatial non-linearities due to the pillow-shaped bulging of the PSPC window due to gas pressure. This results in the electric field lines along which the electron cloud travels no longer being perpendicular to the plane containing the anode wires. Thus a spatial distortion is introduced, which is also a function of penetration effect & hence energy. The change is that the previously derived correction factor (as plotted in Fig4b of TN-ROS-ME-ZA00/027 ("Calibration Corrections to individual PSPC Events") has been smoothed using the resolution of the detector. The reason was that the sharp disconts in the original curves at absorption edges lead to high-low channel effect in the PI spectra.

2) A similar smoothing was also applied to the Spatial Gain correction (fig 3 of the above memo) for the same reason. Previously channels 69 and 70 gave a marked glitch (spike) in high S/N data.

3) The "E_k values" (which parameterize the depths of the discontinuity in the gain curves at absorption edges) were changed for DRM 36. These were minor adjustments and should not have noticeable effects on spectral results (particularly if the raw 256 channels are NOT used -- see 2.4)

2.6 Spectral Analysis Techniques

It should be noted that 256 PI channels were not originally intended for use during scientific analysis, rather it was always intended that these raw channels be binned up so that known remaining digitization effects could be binned away. There appear to be two subtly different philosophies regarding the correct procedure for this rebinning: After excluding "bad" raw channels, the 256 channel data should be

  1. binned up such that each new bin contains the same oversampling of the spectral resolution. This is to prevent dramatically different oversampling factors at various energies skewing the spectral fit. The SASS channels were selected to provide this, but a number of people suggest 19 channels are (statistically) the most appropriate number. Thus rebinning the SASS chans by a factor of two is close.
  2. binned up such that there are sufficient counts in each new bin such that Chi-squared analysis can be used (if desired).

However, it should be noted that the skewing of the spectral analysis is only likely to be significant for high S/N data from very steep (or absorbed, flat) sources. The best recipe may therefore be to use a combination of both methods.

2.7 After Pulses

Due to contamination of the gas supply, a number of observational datasets are effected by "After Pulses"(APs). These are low PH events due to either X-ray photons or particle bkgd events, and occur within approximately 0.35 ms of the primary photon/event (cf PSPC deadtime of 0.25 ms). The most likely cause of APs is negative ion formation in the vicinity of the anodes. They have a soft, steep spectrum effecting only PI channels 15 and below, in which they can contribute up to a few % of the total count rate. The spectrum changes slightly as a function of time, but is always approximately a single electron spectrum. The number of AP events also changes as a function of time and this along with it's correlation to the PSPC gas tank swaps etc is currently under investigation.

APs caused by photons can be rejected, but those caused by the particle bkgd often cannot since the precusor event is usefully vetoed by the on-board s/w and hence not within the telemetry. The number of APs appears to be:

  • worse for high gain datasets
  • worse for harder spectra

The effects counts are not always adequately taken into account using standard background subtraction techniques, and the effect can be quite significant for extended sources (Snowden et al "Analysis Procedures for ROSAT XRT/PSPC Observations of Extended Objects & the Diffuse Background")

The afterpulses can be largely removed by excluding all events occurring within 0.35 millisec of a nearby (parent X-ray or particle) event, this fails when the parent event is a rejected particle event or when the source of interest is soft and bright enough that source events occur within 0.35 millisec of each other.

Steve has produced a recipe for the analysis and minimization of the AP problem (Ref 4) involving the construction of a Polya function spectral component (which should NOT be folded through the redistribution matrix).

2.8 The Rate-Gain Effect

At very high count rates, such as obtained for the Crab (count rates ~800 ct/s) the detector gain saturates due to the reduction in the elective field due to the pseudo-continuous presence of a substantial electron cloud. It is currently thought (hoped) that such an effect does not become important under count rates greatly exceeding those obtained for other bright sources especially Mkn421 (which was used to "fudge" the matrix, ~ 150 ct/s). It is intended that this effect will be investigated further in the PANTER measurements (see 2.4).

3.0 Off-axis Point Spread Function

The PANTER data and in-flight data look very different for off-axis sources. In the PANTER data the peak of the counts distribution lies outside the "hole", in the in-flight data the opposite is true. In addition, the shape of the radial profiles are different in the two datasets. These differences are because the PANTER data were not a parallel beam. This means that one has to use in-flight observations to calibrate the off- axis dependence of the point spread function. Taking 50 fields, Guenther has made a first cut parameterization of the off-axis psf. Basically the most important thing is to fold in a Gaussian due to the "coma" of the mirror, whose width depends on the off-axis angle. As a mathematical approximation it is easier to add terms than convolve them. Guenther has found the easiest way to parameterize the off-axis dependence is to add an angular dependence to the Gaussian term of the psf and to the exponential term. This mathematical approximation is not very good around 15-20 arcminutes radius, where the psf starts to get quite large, but as this is where the detector rib is, analysis at that radius will be subject to large errors anyway.

We have agreed that Guenther will send GSFC his best parameterization of the off-axis dependence after the German proposal review (ie. in 1-2 weeks time) and that will add the angular term to the new version of the psf software, and test it against datasets we have at GSFC (weighted with the appropriate source spectra). We expect several weeks of iterative changes to the model at this stage, before the off-axis parameterization is publically available.

4.0 Boresighting

Previously, systematic discrepancies have been noted between the optical and X-ray positions of ROSAT targets. In an MPE study the 1-sigma scatter of the distribution of optical versus X -ray positions was 6.1 arcsec for the PSPC (B) and 6.4 arcsec for the HRI. The PSPC data showed a systematic offset of 6.9 arcsec while the HRI showed negligible systematic offset (the HRI always had a boresight correction of 10 arcsec in the Y-direction). For a long time, no correction was made to SASS to account for this systematic effect in the PSPC data. This was because there was a large scatter in the offsets noted (up to 20 arcsec), and the effect was not understood. Finally though, it was decided to correct the SASS for the mean offset of 6 arcsec in the PSPC data. The correction was implemented in December 1992 at MPE (SASS 6-2), and affects data processed since then (or since Jan 1993 in the USA), which are data taken since ~ September 1992. Also, all REPROCESSED (Rev1) data will have this (or an updated) boresight correction folded in.

There were some inaccuracies in the original study however, notably that the nearest SIMBAD source was always taken as the optical counterpart, when in fact that was not always the ID with the best position. This meant of course that the boresight error was originally underestimated. A new study then showed the deviations from expected (SIMBAD) positions still lay clustered around 3-4 arcsec on the positive detector y axis. Another 3-4 arcsec correction is required, in the same sense as the original correction.

These boresight corrections do not get rid of the scatter in the difference between optical and X-ray positions. It is thought that this may be due to star tracker problems. It seems that the startracker has some variations in quantum efficiency between the pixels. This means that the centroiding to get the position of the guide stars can be skewed. It seems unlikely that this can be corrected for as not all of the star tracker pixels are calibrated. This seems like the most likely origin of the large scatter in the discrepancies in X-ray positions mentioned above.

5.0 SASS Summary

Newest SASS output cover pages will include the version numbers of the software components of the SASS processing. These components are:

ROMAN
the ROSAT managing software. This handles the input (and output) parameter files and lists. Thus this refers to changes in parameter settings (eg for data screening) and things like which calibration files are used by the SASS (matrix, effective area etc.) and a revision indicates a change in one of the parameter lists. Thus important info on things like attitude screening are related to the ROMAN version numbers.
STEERING
directs data to the database etc
ROSAT
splits HRI/PSPC data etc.
HRI & PSPC
instrument specific software
MPE
internal MPE routine related to utilities

Each SASS number relates to a specific combination of versions of the aforementioned routines.

Wolfgang Voges has agreed to provide us a list of SASS releases with the corresponding version numbers for the routines listed above. He will include installation and distribution dates. We will also receive a list of which calibration files were used as input files for each SASS release, and what the screening criteria were for each SASS release. In this way we hope to compile a " SASS History for Data Analysts", so PIs know what has been done to each of their datasets.

6.0 Absolute flux calibration

The simultaneous PSPC (survey) + Ginga data indicate reasonably good agreement at 1 keV, indicating a good absolute flux calibration of the PSPC. The most useful datasets seem to be the AGN, many of which have been analyzed and written-up (proceedings of the the Garching AGN meeting Nov 91). These agree to within 10% in the overlap regime. While many of those sources have complex spectra and cannot be used for this study, there are some good calibration sources such as 3C272 ( where the break to a soft excess is well below the overlap region), Mkn421 which has a curved spectrum but has very high S/N and the curvature can be modeled. The background spectrum and flux were compared to the Wisconsin background data and the agreement is again good to within about 10 %.

The Galactic sources are disappointing. Many are weak hard X-ray sources and in many cases the analysis of the ROSAT/Ginga data is incomplete.

A ratio plot of two N132D spectral PI files indicates no significant spectral variations on long (6 month) timescales. This will be followed up in more detail.

7.0 New Documents

The following assorted set of documents, memos, preprints etc were also brought back:

  1. "Mission Planning with ROSAT" Reprint of Snowden & Schmitt 1992 Data Analysis in Astron IV, 121.
  2. "Calibration Program for the Spectral Response" Notes by Steve Snowden on the forthcoming PANTER calibration measurements using the engineering model PSPC.
  3. "In orbit PSPC entrance window ablation by atomic oxygen bombardment" TN-ROS-ME-ED52/24 by Schmitt 1987 Aug.
  4. "Analysis Procedure for ROSAT XRT/PSPC Observations of Extended Objects and the Diffuse Background" Snowden etal, submitted to ApJ.
  5. " The Effects of ROSAT Orbit Inclination on the all-sky survey" Snowden and Schmitt 16 Feb 1988 TN-ROS-ME-ED52/29
  6. "Study of the ROSAT XRT startracker problem" Kuerster and Hasinger 23rd October 1992 TN-ROS-ME-ZA00/029

These will be added to our master list of technical memos, and Steve Snowden is collecting the online text for all of these in a directory at MPE, as soon as we have a final approval to put these on the ftp account, we will copy them all over. Steve wishes to regenerate some of the figures for these memos (at least for his memos) and can make us figure post script files which can also go online.

Menu Select another report ROSAT GOF Return to the ROSAT GOF Curator: Michael Arida (ADNET); arida@milkyway.gsfc.nasa.gov
HEASARC Guest Observer Facility


Please use the Feedback link if you have questions on ROSAT.

This file was last modified on Tuesday, 14-Sep-1999 11:47:03 EDT

NASA Astrophysics

  • FAQ/Comments/Feedback
  • Education Resources
  • Download Adobe Acrobat
  • A service of the Astrophysics Science Division (ASD) at NASA/ GSFC

    ROSAT Project Scientist: Dr. Robert Petre

    Responsible NASA Official: Dr. Andy Ptak

    Privacy Policy and Important Notices.