It is clear that in order to obtain maps of the CMBR fluctuations alone it is necessary to separate the emission due to these various components. The removal of point sources from the satellite observations is perhaps the most troublesome aspect of this separation, since our knowledge of the various populations of sources is incomplete. Nevertheless, at observing frequencies in the range 10--100 GHz, we expect the point sources to be mainly radio-loud AGN, including flat-spectrum radiogalaxies and QSOs, blazars and possibly some inverted-spectrum radiosources. At higher observing frequencies in the range 300--900 GHz, the dominant point sources should be infrared luminous galaxies, radio-quiet AGN and smaller numbers of high-redshift galaxies and QSOs. However, since the frequency spectra of many of these extragalactic objects are, in general, rather complicated, any extrapolation of their emission over a wide frequency interval must be performed with caution.
For small fields, a straightforward and effective technique for removing point sources is to make high-resolution, high-flux-sensitivity observations of each field, at frequencies close to those of the CMBR experiment. The point sources can then be identified and accurately subtracted from the maps (O'Sullivan et al. (1995)). For multifrequency all-sky satellite observations, however, such a procedure is unfeasible. For the Planck Surveyor, it is expected that a significant fraction of point sources may be identified and removed using the satellite observations themselves, together perhaps with pre-existing surveys. Based on the estimated sensitivity of the Planck Surveyor to point sources, De Zotti et al. (1997) find that it is straightforward, at each observing frequency independently, to subtract all sources brighter than 1 Jy and that it may be possible to subtract all sources brighter than 100 my at intermediate frequencies where the CMBR emission peaks. Careful modelling of the likely point source contamination also suggests that the number of pixels affected at each frequency should only be a small percentage of the total number. Moreover, De Zotti et al. (1997) find the level of fluctuations due to unsubtracted sources to be very low. Using simulated maps of point sources (Toffolatti et al. (1998)), a full investigation of their effects on Planck Surveyor observations will be presented in a forthcoming paper (Hobson et al, in preparation).