1 Introduction

Since the great success of COBE-DMR results (Smoot et al.[1992]; Bennet et al.[1996a]; Górski et al.[1996]) and a large number of ground-based and balloon-borne detections of cosmic microwave background (CMB) anisotropy, many experiments with sensitivities and angular resolutions better than COBE are expected for the near future [e.g. Lasenby et al.([1998]) and De Bernardis & Masi ([1998]) for a review].

It is also clear that only a space mission free from unwanted contamination from ground and Earth atmosphere and with a nearly full-sky coverage (Danese et al.[1996]) can fully exploit the gold mine of cosmological and astrophysical (De Zotti et al.[1999]) information imprinted in the small scale pattern of microwave sky anisotropy. The two space missions MAP (Microwave Anisotropy Probe) (see Bennet et al.[1996b]) by NASA and PLANCK (Mandolesi et al.[1998]; Puget et al.[1998]) by ESA, planned to be launched respectively in the year 2001 and 2007, will play a definitive role to fully understand the properties and the evolution of the universe.

As for any CMB experiment, great attention has to be devoted to all the possible systematic effects. In the context of the PLANCK mission we have carried out a detailed study of one of these effects: the so-called 1/f, or low-frequency, noise which may lead to unwanted stripes in the final sky maps. Such stripes can increase the overall noise level and introduce correlations which affect adversely the statistical analysis of the CMB anisotropy. Recent analytical work by Seiffert et al.([1997]) has shown the dependency of the 1/f noise upon the radiometer characteristics such as the bandwidth, the noise temperature, payload environment temperature and other quantities properly related to PLANCK-LFI radiometers. They can be combined to define a representative parameter, the "knee-frequency'' $f_{\rm k}$, which has to be kept as low as possible compared with the spinning frequency $f_{\rm s}$ of the spacecraft. Janssen et al.([1996]) have indeed demonstrated that for $f_{\rm k} \,\lower2truept\hbox{${> \atop\hbox{\raise4truept\hbox{$\sim$ }}}$ }\,f_{\rm s}$ a degradation in final sensitivity will result.

In Sect. 2 we briefly outline the source of the 1/f noise and its spectral shape. In Sect. 3 we report the framework of our simulations, evaluate the possible impact of 1/f noise and present the technique for removing the 1/f noise in the context of PLANCK-LFI. The main results both for un-reduced and reduced 1/f noise are presented in Sect. 4. Our main conclusions and implications for the scanning strategy are discussed in Sect. 5.