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

21.5. A WORKED EXAMPLE 281
small. They are most noticeable for the DivSky method, or at the bad pixels removal stage
(see sec. 21.6).
Let us begin with the Perturbed Single Flat-Field method. One should note that this is
really to allow determination of the long-term transient. Since a subtraction is done, this is not
strictly speaking a flat-field correction and should not be used as such.
This method requires 4 parameters to be set with set red param: tdt, flat smooth window,
flat thresh, andnplanes.
• tdt: this is ISO-specific number which is located in the tdtosn field of your raster data
structure. How to extract this information is explained in section 21.4.1. In our example,
its value is 65801627, meaning the observation number 27 of that proposal, executed on
revolution 658 as the 16 th observation of the revolution.
• flat smooth window: this is a number of pixels used to define a smoothing window for the
filter that is going to be applied to the individual 32×32 images to remove the large scale
structure before the flat-field construction. Since it is working on 32×32 images, it should
be a rather small number, typically of the order or slightly larger than the radius of the
PSF (that is because the size of the filtering box will be 2×flat smooth window+1). Note
that smaller values of this parameter usually translate in small values of the residual noise,
but that is some sort of artifact since for small windows, most of the signal is removed. . .
• flat thresh: this is a percentage that is used to discard pixels from the flat field computation.
The histogram of the pixel readouts is built and we discard the bottom and top
flat thresh percent. This is useful when you have strong sources and structures in the
field to avoid having them propagate in the flat-field.
• nplanes: once the filtered and histogram-selected cube of readout is built, the flat-field
cube is built with a running average of nplanes. In that case, nplanes should be larger
than N exp but not as large as if no filtering had occurred.
Now that the definition of the parameters is clear, let’s proceed. Table 21.3 lists the parameters
setup for both filters. Note that both filters share the same tdt number since they
were obtained in a single AOT. For this data set, the sources are not very important so the
flat thresh value is not critical. As always in SLICE, choosing the parameters values requires
some tuning, and you’ll need more than one try to arrive at a good choice of parameters. The
commands line to issue to SLICE arethus,intheLW3case:
CIA> red param=set red param(tdt=’65801627’,flat thresh=10,nplanes=60,$
CIA> flat smooth window=4)
CIA> act=set act(/make map)
CIA> slice pipe
The results of this first flat-field determination are displayed on figure 21.2. At that stage,
the improvement from CIA is not yet striking.
Now let’s try the DivSky method. It is set with 5 parameters: tdt, flat thresh, nplanes,
size filter, anddivsky. Of these, only two are new. Let us explicit them:
• divsky: this is just a boolean keyword, i.e. true/false, set to indicate which method is to
be used (if it is not present on the command line as /divsky, then the Sliding Mean
Flat-Field method is used, which is not what we want.