Subject: Re: IUG Meeting ------------- Begin Forwarded Message ------------- Subject: Re: IUG Meeting Hello Chris and Arvind: I hope you will excuse my running away from the Users Group meeting in the middle of the session. I had planned my trip as a one day excursion from our vacation place near Jonkoping, and this was as much time as I could get under these circumstances. I appreciate your offers of assistance (moral assistance initially, but hopefully also more concrete assistance later) for the future development of the JEM-X software - in particular concerning the spectral extraction software. After coming back to Denmark I have discussed the problems we face with Soren Brandt (everybody else is off on vacation). During my absense Soren has made an analysis which supports my suspicion that we need to take a much more serious look at the pixel to pixel variations of the JEM-X detection efficiency as function of time, detector position and energy. So far our spectral extraction software has assumed that we could use one common response matrix for all observations made within a limited time period (typically the 6 months between any two Crab calibration runs). This matrix was assumed to be valid across the entire detector surface, and that it could be applied to data already corrected to a 'pulse invariant' energy scale. All these assumptions are now called into question by the results obtained by Soren. The attached note give a brief summary of Sorens analysis. So I am sure that we must rebuild our spectral extraction software such that it uses the actual signals corresponding to each event rather than the pulse invariant signals, and that we must construct a specific response matrix for each individual source in each science window. We must then learn how to combine spectral data obtained in several science windows and characterized by different response matrices. Unfortunately we have lost from the JEM-X team the person who developed our present spectral extraction software. We must therefore start pretty much from scratch here at DNSC. But in one sense this may turn out to be an advantage because it will then be easier to assure compatibility between the new OSA5 ima-iros imaging software and the new spectral extraction software. (Some of the essential elements in the new spectral extraction package can be taken over from ima-iros). So the obvious area where we could benefit from support is with the problem of spectral extraction from an instrument with significant local variations of the efficiency. We are pretty confident that this is not the first time such problems are encountered in X-ray astronomy - but personally I do not know where to find this person. Before we embark on this project it may be very useful to arrange small workshop here at DNSC where we may invite 2 or 3 experienced persons, where we could present the problems as we see them and put forward our plan of action. Most likely we should include people from ISDC and from the IBIS team - and possibly SPI as well. To get things going reasonably quickly I would hope that we could arrange such workshop in the September timeframe. Regarding further calibrations we could benefit from Crab observations taken with different detector voltages (gain). We will elaborate our requirements for this before the next Crab calibration. But such data are not a prerequisite for starting on the definition and coding of a new spectral extraction package. With best regards Niels ********************************************************************* Technical background: Soren have looked at the count rates from our Fe-55 calibration sources (in JEM-X1) as function of the detector gain. The detector gain varies as function of temperature and time due to the ageing of the glass microstrip plates. By occasional reductions in the detector voltage we try to maintain these gain variations within a +/- 10% range. Now Soren has found that in the first part of the mission, up to orbit 102 where we changed our on-board selection criteria, the count rate of the Fe-55 sources (read: detection efficiency at 6.4 keV) varies in a linear fashion with the gain. The efficiency variations are typically about 15% for a 10% gain shift at 6 keV, they are larger for lower energies and smaller for higher energies, but will disappear only for energies in excess of 12 keV. After the modification of the selection criteria in orbit 102 only energies up to about 8 keV will be affected by typical gain changes. Recalling that local gain corrections of up to 40% are applied to the individual events based on their position in the detector it is clear that the assuption of a fixed response matrix valid everywhere on the detector and independent of the instantaneous global microstrip gain is very questionable. Figure Captions: JEMX1_Fe_gain_orbit.pdf ----------------------- The plot shows the PHA position for the two Fe calibration sources in JEM-X1 averaged per revolution. The periods of gradual increase are periods of constant high voltage settings. The gain jumps down when the HV is adjusted down. NOTE that JEM-X1 was operated at a very low gain up until the end of orbit 102. JEMX1_response.pdf ------------------ The plot shows an attempt to reconstruct the JEM-X1 response as function of detected pulse height (PHA) based on data from the Fe calibration sources. Ten calibration spectra are used from the beginning of each orbit. (Each spectrum is integrated over 256 s). Just after activation in each orbit JEM-X is running at a voltage level somewhat below the nominal setting. Additionally, various effects of the detector behavior allows the sampling of a relatively wide range of detector gain with these data. Two curves are shown: One curve showing the response for the "new" selection criteria in effect since orbit 102. This is based on data from orbits 170 to 330. The other curve shows the response for the "old" selection criteria in effect before orbit 102, is based on data from orbits 20 to 45. The fitted intensity of the calibration sources have been corrected for the 2.7 year half-life of Fe55. In the early part of the mission, up to orbit 115, the full set of selection criteria was applied at all energies ("raw-PHA"-values). This led to a gradual loss of events below raw-PHA channel 115. The change of selection criteria relaxed the onboard selection criteria below raw PHA 115, while maintaining the existing selection criteria above this value. With the new criteria the linear decrase in efficiency below raw-PHA 115 was replaced by a 10% discontinuity around PHA channel 115. The steep drop in efficiency is still present, but is now confined to energies below raw-PHA 85 (about 8 keV). The corrected PHA, which is transmitted to ground is within +-10% of the raw PHA and therefore the discontinuity around channel 115 is smoothed over the range from approximately 105 to 125 in corrected PHA. The curve has been normalized to 1 at maximum for the new selection criteria. The response above PHA 125 is about 90% on this scale. The data indicate that the response for the old selection criteria was quite poor and rapidly varying from effectively zero at PHA below 50 to about 90% at PHA 115. This effect, plus the fact that the instrument was operated a very low gain, the variation in gain due to micro-strip plate ageing over time, and the fact that the standard analysis software only uses the gain corrected "pulse invariant" and not PHA, all conspires to make spectral fitting unreliable for JEM-X1 up to maybe 12 keV for dat taken prior to orbit 102. For JEM-X2 the situation will be similar, except that JEM-X2 was operated at a higher gain value. The spectral problems are therfore expected to be confined to lower energies, this analysis remains to be done. The new selection criteria and the increased gain moves the boundary to the less reliable end of the spectrum lower. However the use of PI in the OSA standard software still prevents the proper use of the information about the response as function of gain as presented in this graph. JEMX_curtain.pdf ---------------- The plot shows the JEM-X "curtain" correction used onboard to derive a corrected PHA from the "raw PHA" calculated from the "slow anode" signal. The correction varies as a function of the event position between the 11 amplifiers in the cathode plane (DETX). The correction is between about 0 and 20% of the anode signal. This will effectively mean a position dependent variation in the detection efficience of events with low PHA. ------------- End Forwarded Message -------------