5 Conclusion

In this paper we have shown that, if the noise of the polarimeters have nearly equal levels and are approximately uncorrelated, these detectors can be set up in "Optimised Configurations''. These configurations are optimised in two respects: first the volume of the error ellipsoid is minimised, and second the error matrix of the inferred Stokes parameters is approximately diagonal in all reference frames. This remains true even if one combines information from numerous measurements along different scanning circles. Such "Optimised Configurations'', with 3 and 4 polarimeters, have been proposed by the Planck High Frequency Instrument consortium (HFI Consortium, 1998) for the three channels where it is intended to measure the CMB polarisation.