Map Destriping Methods

One important issue in the context of the Planck experiment is the 1/f
noise induced gain and noise temperature fluctuations in the LFI
radiometers. When these low frequency fluctuations are coupled with the
Planck scanning strategy they produce artefacts in the final CMB maps.
These artefacts express themselves as stripes which are clearly visible in
the maps and they follow the circular trajectories that the telescope beam
scans in the sky. A typical CMB map containing the instrument noise is
depicted in the figure on the right (for a larger image click here). The noise
stripes clearly follow the pole-to-pole scanning circles. The stripes will
increase the overall noise level and introduce correlations that affect the
statistical analysis of the CMB pattern in the sky. The CMB pattern is
an important input data when the angular spectrum (C(l)) of the CMB is
determined. The stripes will degrade the accuracy of these angular spectra.

The levels of stripes will be reduced when maximum likelihood based map
making methods are applied. In CMB recordings like Planck where the number
of measurement points in the sky is high (> 10 000 000) the maximum
likelihood methods become unfeasible to the existing computers. We have
studied an alternative approach. In the nominal Planck scanning strategy
each point in the sky is scanned several times during the mission. The
scanning is organized in successive closed nearly polar circles. A large
number of points in the sky will be measured during different scan
circles. The low frequency drift is modelled with a set of baseline
values. There is one fixed baseline value for each scanning circle. During
a nominal 14 months mission time there will be around 10 000 scanning
circles. Each scanning circle has its unique baseline value. When the
baseline values are appropriately determined they can be subtracted from
the original map and a new map is produced where the stripes h! ave been
removed. The key point here is how to determine the baseline values. There
are different possibilities but common to all of them is that they utilize
those recordings where the same point in the sky has been measured during
two or more scanning circles. By suitable subtraction we can remove the
CMB and use only the noise information to determine the baseline values.
The methods we use to determine the baseline values are typically based on
maximum likelihood techniques.

We have a full simulation pipeline running in a supercomputer environment.
The pipeline contains software that produce the CMB and noise streams.
These are input to the destriping code. Finally spherical harmonics
expansion is carried out to the maps in order to recover the angular
spectrum. Currently we are studying different methods to determine the
baselines. Another important topic that we are working on is the accuracy
of the angular spectrum when determined from the destriped maps. When the
map shown in Fig. 1 is processed through the destriping pipeline the
resulting stripe free map looks like the map shown in the figure on the right
(for a larger image click here).