As Cosmic Microwave Background (CMB) measurements are becoming more ambitio
us, the issue of foreground contamination is becoming more pressing. This i
s especially true at the level of sensitivity, angular resolution and for t
he sky coverage of the planned space experiments (MAP, 1998) and (Planck, 1
999). We present in this paper an indicator of the accuracy of the separati
on of the CMB anisotropies from those induced by foregrounds. Of course, th
e outcome will depend on the spectral and spatial characteristics of the so
urces of anisotropies. We thus start by summarising the present knowledge o
n the spectral and spatial properties of Galactic foregrounds, point source
s, and clusters of galaxies. This information comes in support of a modelli
ng of the microwave sky including the relevant components. The accuracy ind
icator we introduce is based on a generalisation of the Wiener filtering me
thod to multi- frequency, multi-resolution data. While the development and
use of this indicator was prompted by the preparation of the scientific cas
e for the (Planck, 1999) satellite, it has broader application since it all
ows assessing the effective capabilities of an instrumental set-up once for
egrounds are fully accounted for, with a view to enabling comparisons betwe
en different experimental arrangements. The real sky might well be differen
t from the one assumed here, and the analysis method might not be in the en
d Wiener filtering, but this work still allow meaningful comparative studie
s. As a matter of examples, we compare the CMB reconstruction errors for th
e (MAP, 1998) and (Planck, 1999) space missions, as well as the robustness
of the (Planck, 1999) outcome to possible failures of specific spectral cha
nnels or global variations of the detectors noise level across spectral cha
nnels. (C) 1999 Elsevier Science B.V. All rights reserved.