An improved layer peeling algorithm removes the constraint of equal segments length in a low-loss interferometric
sensor array. Consequently, the construction of the sensor array is greatly simplified while its performance remains
essentially unaffected. With unequal segments, the frequency response of the sensor array, which is based on a serial
concatenation of High Birefringence (HiBi) fibers, becomes effectively non-periodic. The original version of the layer
peeling algorithm was based on scanning one period of the frequency response and expanding it in the form of a Fourier
series. This led to decreased sensing performance since in practice it is virtually impossible to fabricate an array of
identical segments. In the non-periodic case, we show that broadening the scanning range, as well as windowing the
measured frequency dependent Stokes vector, prior to transforming it to the time-domain, greatly reduce sensing errors.
The new approach is demonstrated in a 14-segments array of HiBi fibers with a total length of 235m. Sinusoidal strain
modulation at up to 65Hz is applied to one of the segments. The improved layer peeling algorithm reveals the induced
signal in the perturbed segment with only little crosstalk in the signals extracted from the other segments.
A low-loss polarimetric sensor array, comprising a cascade of five polarization
maintaining fibers and a novel peeling algorithm, is experimentally demonstrated. With one
segment buried in a sand box, the array successfully detected pressure variations.