After the encouraging results of the DMR experiment on the COBE satellite (Smoot 1992), there has been a burst of renewed interest in the anisotropies of the Cosmic Microwave Background Radiation (CMBR), both on the experimental side and on the theoretical side. While theorists refined calculations to evaluate how individual parameters of the theories affect the expected properties of the tiny fluctuations of the CMBR, experimenters, in answer to announcements of opportunities by several space agencies, proposed sophisticated new generation satellites to map the anisotropies of the CMBR with a sensitivity and angular resolution an order of magnitude better than those of COBE. Of these, the Microwave Anisotropy Probe (MAP) experiment has been selected by NASA as one of the next medium-size explorer, or midex, and the PLANCK satellite (formerly COBRAS/SAMBA) has been selected by ESA as the next medium-size mission M3.
The accuracy with which the useful cosmological information can be deduced from the data of such a mission depends on the global characteristics of the instrument and on the observing strategy: sensitivity of the detectors, spectral coverage, resolution, susceptibility to systematics, scanning strategy... the optimal solution is often a trade-off between several marginally compatible constraints, and very different strategies can be adopted.
Because a large telescope is necessary in order to achieve the high angular resolution that is mandatory to distinguish between cosmological models, the option of differential measurements implies complicated optics. Fortunately, the availability of space-qualified cryogenic devices (Benoit et al. 1994) and the development of new readout electronics (Gaertner et al. 1997) now permits to use in space bolometers cooled to 0.1 K. These are much more sensitive than available radiometers at the frequencies most interesting for cosmology, and so much more stable that for the PLANCK bolometer instrument (HFI, for High Frequency Instrument), the conservative differential approach has been abandoned in favour of total-power measurements.
In the case of the nominal PLANCK mission, the scanning of the sky is performed very simply, by rotating the satellite at 1 rpm around a spin axis which position on the sky is roughly anti-solar (to first order the spin axis remains in the ecliptic plane, and its position is shifted by every 2 hours). There is some flexibility as to the direction of the spin axis, however, so that the scanning strategy can be adapted for optimal sky coverage or rejection of systematics within technical constraints (the thermal stability of the payload puts a limit on the solar aspect angle of , and the telemetry rate a limit on the earth aspect angle of 15 during the dumping of the data).
The beam axis makes an angle of 70 with the spin axis, and thus scans 140 diameter circles on the sky. The PLANCK orbit is a Lissajous orbit around the sun-earth Lagrange point L2. More details on PLANCK can be found in the COBRAS/SAMBA report on the phase A study (1996).
For this observing strategy, there are several important characteristic time-scales. One scan corresponding to one complete rotation of the satellite around itself is performed in a time minute. Data circles are obtained by averaging 120 such scans, and correspond to a period of two hours. Each of these data circles crosses in two points all other data circles obtained less than about 20 weeks before or after. Thus, they share a common area on the sky of at least two pixels (and more for circles that are tangent or nearly tangent). Finally, data circles corresponding to measurements separated by a 1 year period coincide on the sky. Figure 1 (click here) shows four PLANCK SURVEYOR circles on a sinusoidal projection of the sky in ecliptic coordinates.
Figure 1: Sinusoidal projection of four scans for the nominal PLANCK SURVEYOR scanning strategy. The second, third and fourth scans from the left (long-dashed, short-dashed, and dotted line) are obtained 2 weeks, 2 months, and 5 months respectively after the first (plain line). Because these circles have 140 diameters, intersections are distributed everywhere along circles, not only at ecliptic poles as for great circles. Not all pairs of circles intersect: here for instance the first and last circles, separated by a period of 5 months, have no pixel in common
All these redundancies at very different time-scales make it possible to minimise low-frequency noise effects by comparing the values the signal takes at times where the useful astrophysical signal is supposed to be the same (because the antenna is pointed at the same place on the sky) and thus estimating and correcting for low-frequency drifts.
It has recently been suggested that the scanning strategy and destriping method adopted for PLANCK might lead to striping on the maps due to excess low-frequency noise even if there were no intrinsic low frequency noise in the measurements (Wright 1996).
In that paper, the author argues correctly that relying on no more than a pixel or two per scan (namely, ecliptic poles) to readjust relative offsets might cause striping in the maps. However, when suggesting that for this reason the PLANCK maps will be striped, he seems to disregard completely two essential characteristics of the PLANCK SURVEYOR scanning strategy, which are the 15 freedom of motion of the spin axis and the 70 off-axis spin angle, any of which characteristics modifies completely the way circles on the sky intersect each other. Thus, his suggested conclusions should be regarded with extreme caution.
In the following, we investigate how well a simple destriping technique can remove the striping in the maps in the context of the PLANCK SURVEYOR mission.