„An astronomer is someone to whom a factor of 10 is important, but a factor of 2 is not.“ If we agree about this, we may say that stellar polarization is 1% and that we wish to measure that polarization to l% accuracy. Polarimeters therefore need an accuracy of 0,01 % (of the total intensity).

In general, stellar polarimetry has been rather a delicate measurement. This is because its nature is differential, i. e. the information required is the small difference between two large observables. There are two general classes of polarimeter, single-detector and two-detector. In the single-detector polarimeters the observables are measured one after the other by the same detector and time-varying attenuations cause errors. In two-detector polarimeters, differences between the two channels cause errors. In both cases extreme stability is required. The solution is to use the single-detector type, but to alternate between the two observables faster than any of the time-varying attenuations. This type of polarimeter uses some sort of polarization modulation.