ISOCAM Interactive Analysis User's Manual Version 5.0 - ISO - ESA

21.5. A WORKED EXAMPLE 283
Table 21.4: Our choice of parameters for the DivSky method.
Filter tdt size filter flat thresh nplanes
LW3 65801627 15 10 60
LW2 65801627 15 10 30
• size filter: this is an integer whose meaning is rather similar to the flat smooth window
of the previous method. Here also the data are filtered and removed from the readouts
before computing the flat-field. However, it is the sky image (in map) that is smoothed and
then divided out. Note that contrary to flat smooth window, the size of the smoothing
box is really size filter. It is rather hard to determine aprioriwhat should be the value
of this parameter. One way to proceed is to try different values and judge from the results,
from a size slightly larger than the PSF to one larger than the raster step. Remember to
inspect not only the resulting map, but also the flat-field: it is generally in the flat-field
that you can judge the success of your parameter choice. If features reminiscent of your
source appear in the flat-field cube, then size filter is wrong (see section 21.4.1 to see
how to recover the flat-field cube).
Table 21.4 summarizes our choice of parameters for the DivSky method while figure 21.3
shows the results for the two filter. For this particular method, and in the LW3 case, the
command lines are:
CIA> red param=set red param(tdt=’65801627’,flat thresh=10,nplanes=60,$
CIA> size filter=15,/divsky)
CIA> act=set act(/make map)
CIA> slice pipe
An obvious improvement is seen here: the gradient of “emission”, which is in fact the longterm
transient, is much smoother now, and pointing imprints have disappeared. In your data
reduction session, we suggest that you play around with both flat-field methods, before you
select the one to use in the first pass. It is also important that you check that your choice of
flat-field method is compatible with the long-term transient determination (see sec. 21.5.3).
21.5.3 Long-term transient determination
Now that we have seen how to derive the flat-field in our data, we are going to try and determine
the long-term transient which is assumed to be a global (i.e. pixel independent) additive time
drift of the signal.
It is important to understand that the long-term transient determination is made by comparing
the data cube to an estimation of the sky. Thus a flat-fielding method is necessary, and,
among the parameters to set for this step, we will find again those which have been encountered
in the previous section. It is extremely important that a good flat-fielding method is chosen,
otherwise you will see very strong artifacts appearing in the long-term transient curves. These
artifacts are quite characteristic: they form an oscillating signal with a number of peaks equal
to half the number of raster legs. An example of such artifacts is shown in figure 21.4. If that is
your case, try and tune the flat-field parameters better or select another flat-field method (this
is why we do that exercise in sec. 21.5.2).