Experimental justification of a method for low-PMD measurements

When measuring very low differential group delay (DGD), standard techniques are affected by rather high uncertainties. This problem is usually solved by repeating many times the DGD measurements while randomly perturbing the fiber. Such an approach is founded on the basic assumption that polarization scrambling induces the DGD's decorrelation. Nevertheless, it has not been proved yet that the measured DGDs belong to an homogeneous statistical ensemble. We experimentally validate this approach by means of polarization-sensitive reflectometric measurements.