In this paper, we proposed a pressure sensor using a Sagnac interferometer based on a side-hole fiber (SHF) with the assistance of deep learning. A convolutional neural network (CNN) was built to identify the spectra of different pressures since the traditional tracing method will face spectral overlap problems when the shift of the spectrum exceeds the free spectral range (FSR). The spectra of pressures ranging from 0 Mpa to 5 MPa with a step of 0.1 MPa will be normalized firstly and then sent to the CNN model for training. The precited result shows that the coefficient of determination$R^{2}$is 99.99987% with the root mean square error (RMSE) equal to$1.6537\times 10^{-3}$MPa. Additionally, a similar structure was constructed to demonstrate the universality of the proposed CNN model. The model can also get a good performance, although the receiving device has a low resolution, showing its great potential for developing a low-cost sensing system.

We perform theoretical analysis and systematic measurement of the degree-of-polarization and eye-closure penalty for optical signals with orthogonal polarizations. Both the theory and experiment show that the symmetry of the DOP is maintained for the orthogonal polarizations under both first and higher-order PMD, whereas the symmetry of eye-closure penalty is broken under second-order PMD. As a result, an orthogonal polarization pair can have large disparity of eye-closure penalty despite an identical degree-of-polarization. We also demonstrate a novel approach to estimate the maximum eye-closure penalty asymmetry with three orthogonal polarizations on the Poincare sphere.