Search

We demonstrate the digital predistortion (DPD) at a reduced sampling rate in high-loss-budget 10G-class LR-PONs. A >50-Gbps transmission was achieved over 30-70 km using fixed DPD with >87% complexity reduction, compared to fully optimized DPD.

We demonstrate an improved speckle variance optical coherence tomography (SV-OCT) pipeline for monitoring microvasculature dynamics. By accounting for time-varying and system-varying noise, we show robust quantitative SV readouts of human blood flow in 2D and 3D, with promise for long-term physiological monitoring.

High launch optical power can compensate for severe fading and power loss in long-reach passive optical networks (LR-PONs); however, it also aggravates nonlinear degradation, which necessitates the use of complex DSP-based nonlinear compensation techniques at optical network users (ONUs). DSP-related techniques also necessitate the use of additional hardware/software components by the receiver, which can greatly increase implementation costs and energy consumption, particularly when dealing with large-scale ONU deployment. This is the first study to propose artificial neural network (ANN)-based pre-distortion to eliminate the need for complex DSP at ONUs in a high-launch-power LR-PON, thereby permitting the use of a simplified architecture at the user end. In the first phase of the study, the proposed ANN-based pre-distortion scheme was implemented in a single-channel IMDD OFDM LR-PON, which achieved a data rate of >55 Gbps over 60-km transmission with a loss budget of 30 dB without the need for optical inline- or pre-amplification. In the second phase of experiments, the same scheme was applied to a 4-channel wavelength division multiplexing (WDM) OFDM LR-PON. Here, the proposed scheme achieved data rates of >200 Gbps using launch power of 18 dBm per lane, resulting in a loss budget of roughly 29 dB over 60-km single mode fiber transmission.

We propose a structured sparsity learning-based pruned retraining Volterra equalization for inter-dadta-center interconnects. Compared with conventional VE, we achieve 95% and 90.5% complexity reduction without signal degradation for B2B and 40-km at 80-Gb/s PAM4, respectively.

We proposed the ANN-based predistortion to mitigate nonlinear distortion in a 25- to 70-km IM/DD OFDM PON. Using 18-dBm launch power, the >55-Gbps downstream with 30-dB loss budget was demonstrated without nonlinear compensation at ONUs.

We propose an L1-regularized multilayer neural network nonlinear equalizer (L1PML-NLE) for inter-data-center interconnects. Compared with conventional and sparse VNLE, the L1PML-NLE reduces 81% and 57.8% complexity with improved BER performance for 40-km 112-Gb/s PAM4 transmission.

Volterra equalization (VE) presents substantial performance enhancement for high-speed optical signals but suffers from high computation complexity which limits its physical implementations. To address these limitations, we propose and experimentally demonstrate an elastic net regularization-based pruned Volterra equalization (ENPVE) to reduce the computation complexity while still maintain system performance. Our proposed scheme prunes redundant weight coefficients with a three-phase configuration. Firstly, we pre-train the VE with an adaptive EN-regularizer to identify significant weights. Next, we prune the insignificant weights away. Finally, we retrain the equalizer by fine-tuning the remaining weight coefficients. Our proposed ENPVE achieves superior performance with reduced computation complexity. Compared with conventional VE and L1 regularization-based Volterra equalizer (L1VE), our approach show a complexity reduction of 97.4% and 20.2%, respectively, for an O-band 80-Gbps PAM4 signal at a received optical power of −4 dBm after 40 km SMF transmission.

This paper demonstrates a novel approach to reducing in-band nonlinear distortion in DML-based OFDM transmission by introducing a frequency gap between the optical carrier and signal band. This scheme greatly increases tolerance to nonlinear distortion caused by the interaction between laser chirp and chromatic dispersion. Tuning the frequency gap between 2 to 5 GHz (in accordance with the transmission distance) makes it possible to mitigate a considerable proportion of dispersion-induced nonlinear distortion. Experiment results revealed that the proposed scheme can increase the achievable data rate by up to 136%, compared to conventional OFDM over 300 km of dispersion-uncompensated standard single mode fiber. Following transmission over more than 200 km, the proposed modulation scheme also outperforms conventional OFDM incorporating with a Volterra equalizer by up to 24%. We also conducted optimization of DML bias current and driving voltage. Our results show that operating at high bias and driving voltage can greatly improve data rates following the mitigation of distortion via gapped OFDM.

In this letter, we combine the use of dense-wavelength-division-multiplexing and mode-division-multiplexing (MDM) with advanced-modulation formats for on-chip optical interconnects. A three-mode MDM multiplexer (MUX) and de-multiplexer (DEMUX) was designed and used. Each wavelength carrier had an aggregate data rate of 192-Gbit/s using 4-level pulse-amplitude-modulation. The mode crosstalk of the MDM MUX/DEMUX was measured to be below −20 dB in the 40-nm wavelength band from 1520 to 1560 nm. Fourteen wavelength channels are multiplexed and successfully transmitted with an aggregate transmission capacity of 2.6 Tbit/s (14 wavelengths $\times$ 3 modes $\times$ 64 Gbit/s) with all channels satisfying the hard-decision forward-error-correction threshold. Numerical analysis of the mode crosstalk is also performed.

We propose and demonstrate a 64-Gbit/s four-level pulse-amplitude-modulation transmission by using a limited modulation bandwidth silicon microring modulator (SiMRM) and the Volterra equalization scheme. Severe intersymbol interference caused by both the chromatic dispersion of the optical fiber and the bandwidth limitation of components can be compensated. Besides, several equalization schemes are compared and analyzed in terms of performance and complexity. In addition, the design and characteristics of the SiMRM used in the experiment are also discussed.

We employ ℓ0-regularization to reduce Volterra filter complexity by up to 90% in 112-Gbps PAM-4 VCSEL transmission. Compared to ℓ1-regularization, ℓ0-regularization achieves lower complexity and more precise weights without retraining after sparse identification.

We propose and demonstrate a distortion mitigation of 50-Gbitls PAM-4 signal generated by a silicon-based Mach-Zehnder-modulator (SiMZM) after 20-km single-mode-fiber transmission via deep-neural-network (DNN). The implementation of a 2-layer DNN is discussed.

In this Letter, we propose and experimentally demonstrate a novel, to the best of our knowledge, sparse deep neural network-based nonlinear equalizer (SDNN-NLE). By identifying only the significant weight coefficients, our approach remarkably reduces the computational complexity, while still upholding the desired transmission accuracy. The insignificant weights are pruned in two phases: identifying the significance of each weight by pre-training the fully connected DNN-NLE with an adaptive L2-regularization and then pruning those insignificant ones away with a pre-defined sparsity. An experimental demonstration is conducted on a 112 Gbps PAM4 link over 40 km standard single-mode fiber with a 25 GHz externally modulated laser in O-band. Our experimental results illustrate that, for the 112 Gbps PAM4 signal at a received optical power of − 5 d B m over 40 km, the proposed SDNN-NLE exhibits promising solutions to effectively mitigate nonlinear distortions and outperforms a conventional fully connected Volterra equalizer (VE), conventional fully connected DNN-NLE, and sparse VE by providing 71%, 63%, and 41% complexity reduction, respectively, without degrading the system performance.

A pruned Volterra equalizer that reduces computational complexity by up to 85.2% in 112-Gbps PAM-4 optical interconnects is proposed. The BER performance is competitive with LASSO-based Volterra equalizers, and training time is twice as fast.

A 400-Gbit/s Ethernet (GbE) pulse-amplitude-modulation-4 (PAM-4) over 10-km single-mode-fiber (SMF) transmission direct-detected by an integrated silicon-photonics-receiver satisfying the IEEE P802.3bs requirement; and its feasibility to support 512-Gbit/s were illustrated.

192-Gbit/s pulse-amplitude-modulation-4 (PAM-4) signal transmission for on-chip optical interconnects using mode-division-multiplexing (MDM) is experimentally investigated. 3 channels are successfully mode multiplexed and de-multiplexed satisfying the forward-error correction (FEC) requirement.

We have designed a novel convolutional neural network based nonlinear classifier that outperforms traditional Volterra nonlinear equalizers. A BER of 3.50 × 10-6 is obtained for a 112-Gbps PAM4 EML-based optical link over 40-km SMF transmission.

We implement a deep neural network (DNN) to attain a 64-Gbps 4-PAM, 4-km MMF link by using 850-nm VCSEL, with BER of 4.41 ×10−5. The data-rate-distance product achieves recorded high of 256-Gbps-km for Inter-data center applications.

We applied partial transmit sequence and Volterra filtering to achieve 67.6% improvement in capacity in a high-launch-power 60-km OFDM LR-PON. 61-Gbps capacity and 30-dB loss budget were realized to support 64 ONUs with 0.95 Gbps/ONU.

This work demonstrates a 56-Gbps 4-PAM EML-based transmission system. Using an O-band SOA to increase power budget and FFE/DFE to compensate SOA-induced ISI, 37.8-km transmission is successfully implemented with BER below the FEC limit (2 × 10−4)

We demonstrate the feasibility of using silicon-micro-ring-modulators for 50-Gbit/s and 64-Gbit/s pulse-amplitude-modulation-4 (PAM-4) communications in a 20-km single-modefiber (SMF) link with Volterra-filtering. The integrated transmitter may be used in passive-optical-access-networks with 20-km-reach and 64-split-ratio.

We demonstrate a 4-level pulse-amplitude-modulation (PAM-4) transmission system direct-detected by germanium-silicon (Ge-Si) photodiode (PD). Volterra-equalization is used to compensate signal distortion and 40-Gibt/s and 50-Gbit/s PAM-4 signals after 20-km and 10-km fiber transmissions are achieved.

A 400-Gbps WDM-OFDM LR-PON over 60-km SMF is demonstrated with a 10G-class EAM and PIN. 25-dB loss budget is realized without in-line or pre-amplifier to economically support 128 ONUs with 3.1-Gbps/ONU capacity.

A noise-subspace-based adaptive beamformer is modified by introducing the triangular systolic array (TSA). A set of TSAs is used to derive the least-squares problem, thus eliminating the need to compute the least-squares weight vector. This modified structure is advantageous in that it requires only limited input data to accurately describe the external environment. Computer simulation results are presented. >