Subject: Gain correctin details Dear Silvia, It looks like you've grasped all the major points. Here are some more details: 1) The weakness of the Ni peaks is not related to the dead anodes or any other gain problem as far as we know. It simply seems to be a case of the Ni foils being too thin to do the job, or perhaps having being dislodged during integration/launch. Whatever the reason we had to discard using the Ni peaks because though they can be seen in multiple summed calibration spectra, they're simply too small to be anything but noise in the individual spectra - hence the gain calibration software's horrible results in the beginning of the PVPhase. 2) Temporal changes in gain were always expected, and we did expect gain to vary in a regular way during each revolution, though the exact form of the gain shape was unknown at the time of launch, so we had a smoothing model that just consisted of time-dependant harmonics. The gain decay after switch on was known from lab calibration, but we hadn't realised this would be the major source of gain variation within each revolution. However, the two-stage HV switch on was implemented precisely because of this problem, because the hardware guys were worrried that turning on the full HV supply would cause such a high gain initially that it'd fry the microstrip plate. Now that this hi-gain/decay pattern seems to have died out, due to aging of the plate, we're considering doing only a one-stage switch on in the future. 3) Smoothing was originally considered necessary because of the inevitable statistical noise from single calibration spectra. The alternative is linear interpolation between the points at 4-minute intervals, which the user can still choose if they so wish. What made smoothing so necessary after launch was the appearance of gain glitches, which we know to be very localised, so that correcting with a glitch-gain value would be incorrect for all events except those very close to the calibration spectra collection site with the glitch. The gain decay itself does not make smoothing necessary, though it does mean that we have to have an early decay component in the smoothing model to get a good fit to the data. If there was not statistical noise, and no glitches, linear interpolation would be a fine way to find the gain between the 4-minute values, also in the case with gain decay. 4) The spatial gain tables were not used before the launch, though the tables themselves and the means of using them were available then. Energy determination with the calibration spectra is like the zero-th order energy correction, and this had to be functioning well before taking on board the finer corrections. The spatial gain correction tables (1st slice of SPAG-MOD) derived from pre-flight calibration is the first-order correction, and is quite significant for the gain resolution. The per-anode gain corrections (2nd slice of SPAG-MOD) are second-order corrections, and have very little affect on the gain resolution, but we keep them for completeness. 5) We do not use the fourth anode segment of JEM-X2 for gain determination, because, as you say, the calibration site seems to sit right on top of a dead anode strip. During the very first revolutions this anode segment had very strangely varying gain which did not mirror the behaviour of the other anode segments. However since revolution 30, the gain of the anode four segment calibration spectra have closely followed the behaviour of the other three sets of spectra, and there could be an argument for including the anode in the gain correction again. 6) Dead anodes are not the last word in gain resolution. We had hoped that implementing the anode-specific gain correction would really improve the gain resolution, but this was not the case. Where dead anodes are a nuisance is that they decrease the active detector area and can seriously screw up the calibration spectra if they pass through a calibration zone, as with JEM-X2. The greatest cause of excessive line widths is probably the multitude of mini glitches that occur continuously all over the detector plate, and therefore can neither be monitored nor corrected. We believe glitches are due to cosmic rays hitting the plate, depositing their charge and kinetic energy and hence charging the plate, which repels the electron cascade somewhat and causes a drop in gain. Dead anodes do not cause glitches: they happen all over the plate. However, cosmic ray damage may cause dead anodes. 7) The Xe line at 30 KeV is only used for offline calibration checking - like the single-anode gain determination done by Jerome. The line has to be integrated over long periods of time from each detector pixel to be useful, so this isn't done as part of the automatic calibration and correction. It does however, let us determine whether individual areas on the detector, away from the usual calibration sites, are behaving as they should - but only over long time periods. This means it cannot be used as a way on improving on poor gain resolution due to short glitches all over the plate. I hope this answers any questions you may still have about energy calibration. I'll make you a gzipped tarball of relevant figures and send it along later. Best wishes, Carol Anne .DANISH.SPACE.RESEARCH.INSTITUTE.DANISH.SPACE.RESEARCH.INSTITUTE.DANISH. Dr. Carol Anne Oxborrow Email: oxborrow@dsri.dk Homepage: http://www.dsri.dk/~oxborrow Telephone (direct): +45 35 32 57 33 Telephone (secretary): +45 35 32 57 01 Fax: +45 35 36 24 75