Abstract
We recall the principle of the correlation chopper and we develop a matrix formalism changing the deconvolution into the resolution of a linear system of N equations with N unknown roots. We consider the case of an ideal rectangular signal function and we solve the system when the number of analysing channels is equal to, or larger than the number of elements in the slot sequence { ai}. We show that, in practice, the correlation chopper improves the statistical accuracy with respect to a one slot chopper if the mean value of the diffraction function is smaller than one half the time independent background. We show also that an increase of the number of analysing channels, the experiment lasting the same time, improves the definition of the shape of the peaks, without modification of the relative accuracy of the measurement