DnCNN Assisted Phase Noise Cancellation for Fiber-Optic Interferometric Sensors

An effective method for reducing phase noise is to employ an auxiliary reference interferometer to independently measure the phase noise. The common-mode phase noise can then be directly subtracted from the sensor signal. However, due to difficulties in achieving consistent optical path length differences (OPD), two interferometers exhibit different modulation depth and hence leading to different phase noise and nonlinear distortions. Previous studies have demonstrated that, after the nonlinear distortion is corrected, effective phase noise reduction can be achieved by linearly scaling the noise signal such that the phase noise in both interferometers becomes identical. However, the nonlinear distortion is hardly to be completely suppressed. In this paper, we figure out that, the linearly scaling approach is valid when the remained nonlinear distortion is negligible compared to phase noise. When the magnitude of the remained nonlinear distortion is nonnegligible, linear scaling fails to match the non-identical distortions, leading to incomplete phase noise cancellation. To address this challenge, we propose a DnCNN assisted phase noise cancellation approach, in which DnCNN is employed to match the non-identical residual components arising from the two interferometers. Experimental results in this paper show that the proposed approach achieves an average of 17.8dB reduction in total harmonic distortion (THD) and 12.7dB improvement in signal-to-noise and distortion ratio (SINAD).