Up: Introduction Previous: 2 Detailed comparisons
3 Impact on the daily calibrations
It is part of the regular ISO orbit to perform CAM calibrations. These are mainly used to monitor the trends in the detectors status. At the activation of the detectors, i.e. switch-on, we perform dark measurements (as well as a curing which I will not use here as it is close to saturation anyway). At the deactivation, i.e. switch-off, we perform dark as well as flat-field measurements.
Both the activation and deactivation are done in a part of the orbit which is close to the Earth's radiation belt and where we know that the glitch rate is higher than along the rest of the orbit, even under normal conditions. We also know already that dark currents at deactivation are correlated with the electron fluence as monitored by GOES. This will have some importance later on.
We also make LW dark measurements at the change of ground station, in the middle of the orbit. This is called the Handover. These handover darks have been shown to be more representative of the ``real'' dark than the activation or deactivation darks. They are actually used for scientific calibration purposes.
In the following, I have extracted from our archive the daily calibrations obtained for revolution 700 to 729. In all the plots that I will show, I have used some sort of normalization in order to amplify the changes from one revolution to the other.
- Darks: For each calibration sequence I construct a reference dark as the mean of the darks taken from revolution 700 to 709. The reference dark is then subtracted to all the dark images. The median and rms plotted in the figures 4 to 6 are computed on the whole image.
- Flat-Fields: Here again I construct a reference flat-field
for each calibration as the mean of the flat-fields taken from
revolution 700 to 709. Then all the flat-field images are
divided by this reference flat-field. The median and rms plotted
in figures 7 and 8 are computed on the
central part of the image, i.e. pixel area (10:21,10:21).
For various reasons, not all the daily calibrations are performed. Therefore in the figures I have used symbols to mark the location of the revolutions on which we do have data.
Let us now examine these calibrations in detail, starting with the dark as they are the most straight-forward data (flat-fields contain a further processing step, they have been dark-subtracted, and this will have its importance).
3.1 Dark current measurements
Figure 4: The activation dark measurements on the SW detector, left,
and the LW detector, right. The full line shows the evolution of the
median in the image (asterisks mark the revolution where we have
data) while the dashed line shows the evolution of the spatial rms in
the image (diamonds mark the revolutions where we have data). The
bimodal character of the median in SW is due to the automatic data
processing. The integration time for these sequences is 2.1s.
Figure 4 shows the activation dark for the SW and LW detectors. In SW, the solar flare produced no effects. This is most likely because the SW detector is much thinner than the LW one, and is thus less sensitive to glitches. The spike in the rms at revolution 717 is due to a glitch remaining on the reduced data. In LW revolution 722 is clearly detected, especially in the rms. This is due to an increase in the number of glitches remaining on the data after the reduction processes have been applied. That glitches can remain after the treatment even under normal conditions is exemplified by revolution 716 where here also the spike is due to a glitch that affected a large nunber of pixels.
Figure 5: The deactivation dark measurements on the SW detector at
integration times of 2.1s, left, and 6.02s, middle, and on the LW
detector at an integration time of 2.1s, right. The full line shows
the evolution of the median in the image (asterisks mark the
revolution where we have data) while the dashed line shows the
evolution of the spatial rms in the image (diamonds mark the
revolutions where we have data).
At deactivation, figure 5 paints a slightly different picture of the solar flare event: The SW data still show no specific change of the median value at revolution 722 though there is a peak in the spatial rms, probably indicating either that low-level glitches resisted the reduction processes or that the detector has suffered a bit from a revolution spent receiving particles. It is clear however from the data on following revolutions that the detector and associated electronics have recovered. One should also note that revolution 722 is rather negligible compared to the bump that appears betwen revolution 710 and 718 (which by the way can also be seen on the LW detector). This is what was mentioned earlier: the sensitivity of the deactivation dark measurements to the local space weather. Such events, most of them stronger, have already been recorded in the past and leave no trace on the detector.
The LW data (right panel of figure 5) show a clear increase in the mean for revolution 722 (the fact that revolution 723 seems to be part of the peak cannot be considered significant, given the large variation of the dark current recorded on LW). Here again inspection of the image shows that this increase is related to glitches on the array, rather than to a global effect on the electronics. All other peaks in the rms lines are related to individual glitches that can be seen directly in the dark images.
Figure 6: The handover measurements on the LW detector at
integration times of 2.1s, left, and 5.04, right. The full line shows
the evolution of the median in the image (asterisks mark the
revolution where we have data) while the dashed line shows the
evolution of the spatial rms in the image (diamonds mark the
revolutions where we have data).
Finally figure 6 shows the trend observed during the handover. Here revolution 722 can be quite clearly seen, mostly because handover measurements are free from all other perturbing sources that affect activation and deactivation (nearby radiation belts, electrical effects related to the switch on and flooding of the detector...). Inspection of the individual images reveals once again that the increase in dark level and/or rms is due to individual glitches on the array. It is much stronger in the 5.04s data because these are almost impossible to deglitch correcly, whereas at two seconds the higher sampling rate helps.
These figures also show that the dark level was significantly lower
than ``normal'' on revolution 706, 713, 720
, 726, and
728. Inspection of the frames do not reveal any suspicious impact
that could explain these low values. The effect appears to be global,
i.e. affecting the whole array, as confirmed by the fact that the rms
has not significantly changed for these revolutions. Such drops were
already seen in previous revolutions, but their origin is not known.
Coming back to revolution 722, we can conclude that, from activation, deactivation or handover data, the electronic systems have not been affected by the array in another way than a simple, though large, increase in the number of particle impacts. Data taken in the following revolutions show that their properties have remained unchanged.
3.2 Responsivity measurements
What can be feared is that such a particle bombing leads to a permanent decrease in ISOCAM's responsivity, as some glitches have the ability to affect CAM's pixels gain (though for a limited amount of time). We can search for this effect using the flat-field measurements that are done during the deactivation sequence at the end of the revolution.
Figure 7: The deactivation flat-field of the SW detector,
obtained at 2.1s through the SW1, left, and SW3, right, filters.
The full line shows the evolution of the median in the image
(asterisks mark the revolution where we have data) while the dashed
line shows the evolution of the spatial rms in the image (diamonds
mark the revolutions where we have data). Note that we have added
1.02 ADU to the rms to plot it on the same graph.
Figure 7 shows the trend of the SW detector flat-fields. Nothing special can be seen apart from the apparent bimodality of the level, which again is related to the treatments of the data: The signal after the event shows the same properties as before the event, i.e. the SW array responsivity has not been affected by the solar flare.
Figure 8: The deactivation flat-field of the LW detector,
obtained at 2.1s through the LW2, left, and LW3, right, filters.
The full line shows the evolution of the median in the image
(asterisks mark the revolution where we have data) while the dashed
line shows the evolution of the spatial rms in the image (diamonds
mark the revolutions where we have data). Note that we have added
1.03, respectively, 1.01, ADU to the rms to plot it on the same graph.
Figure 8 shows the same trends for the LW detector. The most obvious feature is that the responsivity appears to drop on revolutions 707, 714, 719, 720, and 728 for the LW2 flat-field (left panel of figure 8) and 706, 714, 719, 720 and 728 for the LW3 flat-field. These revolutions are quite close to, or even right on, those where the handover dark showed unexplained low values (see figure 6, left panel). If indeed the dark level is significantly lower around these revolutions than the standard dark which is used to process the flat-fields, then the median of the flat-field image will also be significantly lower than on other revolutions. This is most likely what we see here.
Discarding these revolutions from our inspection of the figures does not make revolution 722 stand out much clearer above the trend. It in only in the LW3 data that we have a hint of an increase. But nothing is seen in the following revolutions that would indicate that the LW array has suffered from the event.
Therefore we can conclude that responsivity either of SW or LW was not significantly by the solar flare. ISOCAM has fully recovered from the event.
3.3 Memory dump
Just for completeness, I mention that the memory dump performed at the end of revolution 722 revealed no errors.
Up: Introduction Previous: 2 Detailed comparisons
Marc Sauvage