This work presents a systematic approach for the characterization and numerical estimation of phase noise (PN) parameters at the output of a low-noise semiconductor external cavity diode laser using delayed self-heterodyne interferometry (DSHI) measurements. A parametric PN model comprising multiple colored noise contributions is introduced, and a numerical extraction algorithm is implemented to fit model parameters to measured power spectral densities (PSDs). Experimental results obtained for various optical fiber lengths demonstrate that the simulated PSDs, generated from the estimated parameters, closely match the measured spectra. The proposed method enables accurate PN characterization, providing a realistic PN model for electro-optical THz communication systems.
We report the application of inverse design for a 2×2 multimode interferometer (MMI) coupler on a 220 nm silicon-on-insulator (SOI) platform, with the aim of reducing the device footprint through an over 26% shortening of the multimode region compared to previously reported work and conventional design method, while maintaining a device which has high tolerance to process variation. Through shape optimization, we have designed and fabricated an MMI coupler with a length of 40 μm and power transmission variation across the C-band of less than 0.5 dB. We apply a further optimization process to the 2×2 MMI coupler and propose a design with a length of only 6 μm, while maintaining tolerance to fabrication variance.
The rapid growth in bandwidth demand in modern optical networks has driven the need for high-data-rate modulation techniques such as 16-QAM. However, these methods typically depend on complex and power-intensive coherent receivers. To address this, the Kramers–Kronig (KK) algorithm enables simplified demodulation using a single photodiode. This work proposes a hybrid receiver combining an all-optical Photonic Reservoir Computing (PRC) system with a KK direct-detection receiver to improve performance. The KK algorithm reconstructs phase information digitally, while PRC acts as an optical equalizer to mitigate dispersion and nonlinear effects over a 70 km fiber link, ensuring accurate phase recovery. Performance comparisons using various Support Vector Regression (SVR) kernels like Linear Kernel, Radial Basis Function (RBF) Kernel, Anova RBF Kernel, Arc Cosine Kernel, Regression models like Linear, Ridge, and Lasso regression, Ensemble Based Learning (Boosting) model and Classification model (Naive Bayes) at the reservoir’s readout layer was performed. It shows that the Naive Bayes classifier achieves the better performance in terms of BER and Q factor. Overall, results confirm that integrating PRC with the KK receiver significantly reduces BER compared to conventional KK detection, evaluated using BER, Q-factor, and constellation diagrams.
Fiber Bragg grating (FBG) sensors provide an effective approach for monitoring flow-induced dynamic strain (DS) in pipelines, where the measured DS signatures reflect the underlying multiphase flow behavior. In this work, DS responses were recorded from four representative three-phase flow regimes in a controlled pipeline flow loop using two externally mounted FBG sensors operating at a 5 kHz sampling rate and exposed to mixtures of CO2, water, and diesel oil. The acquired signals were pre-processed and transformed into the frequency domain using the fast Fourier transform (FFT), and principal component analysis (PCA) was used to guide binary slug/non-slug grouping. Cross-FBG slug/non-slug classification was then evaluated by training the models on one FBG channel and testing them on the other. The CNN achieved accuracies of 90.78% for FBG1-to-FBG2 and 85.83% for FBG2-to-FBG1 testing, while the RF baseline achieved 83.37% and 84.70%, respectively, using 17 handcrafted time- and frequency-domain features. These results indicate that FBG-based DS sensing combined with frequency-domain machine learning can support nonintrusive slug/non-slug discrimination in controlled three-phase pipeline flow.
This letter presents a 40-Gb/s fast-locking optical receiver with burst-mode clock and data recovery (BM-CDR). The proposed dynamic gain tuning algorithm (DGTA) mitigates the inherent trade-off between acquisition speed and loop stability. By dynamically stepping down the proportional gain of the CDR from a high state during the preamble to an appropriate value for steady-state tracking, the proposed BM-CDR achieves nanosecond-scale phase locking without degrading steady-state jitter performance. Fabricated in 28-nm CMOS, the prototype receiver occupies an active area of 0.073 mm² and consumes 159 mW. Operating at 40 Gb/s, it achieves a CDR lock time of 6.4 ns, and maintains a rapid lock within 7.2 ns even under a 50-ppm frequency offset.
