Gbaud PAM-4 Research Articles

Design Considerations for 1.6 Tbit/s Data Center Interconnects: Evaluating IM/DD and Coherent Transmission over O-Band Transmission Window

As data center interconnects surge towards a 1.6 Tbit/s data rate, achieving cost-effective and technically viable solutions present challenges. Intensity-modulation and direct-detection (IM/DD) transmission over O-Band using standard single-mode fiber has emerged as a promising low-cost option. However, understanding the limitations imposed by factors like chromatic dispersion (CD) and fiber non-linearity (FWM) is crucial, particularly in different scenarios, such as operating at 8 × 100 GBaud PAM4 in an LWDM-8 configuration. In this paper, we adopt a statistical approach to assess outage probability and consider practical fluctuations in link parameters. Numerical modeling suggests IM/DD can span distances up to 5 km with transmission power under 0 dBm using this architecture. In addition, we evaluate recently proposed architecture to achieve 800 Gbit/s and 1.6 Tbit/s using an LWDM4 configuration and assess the impact of FWM to understand the role of zero-dispersion wavelength (ZDW) of the fiber. Coherent transmission leverages more powerful signal processing capabilities which extends the transmission range. Yet, reducing coherent transmission complexity is desirable for cost-effective and power-efficient data center applications. By exploring dual wavelength transmission and DP-16 QAM transceivers, akin to IM/DD counterparts, the feasibility of streamlining this architecture is also studied. The analysis indicates that the complexity of the coherent approach can be reduced without significant penalties for distances up to 10 km.

End-to-end learning strategy with channel-aided polar autoencoder in IM/DD optical interconnection.

An end-to-end (E2E) learning strategy with a channel-aided polar autoencoder (AE) is proposed and experimentally demonstrated in an intensity-modulation direct-detection optical interconnection system. With the global training of the E2E deep neural network, the learned fiber channel and transceiver characteristics are used to back-forward aid the channel indexes of AE, thus obtaining optimal polarization weight in polar encoding to mitigate signal impairments. The experimental results reveal that when the 50 GBaud PAM4 signals are transmitted over standard single mode fiber, the proposed polar AE by E2E learning strategy can effectively improve the received power sensitivity by 1.5 dB under conditions of overall adaptation of system parameters optimization.

Mixed domain coherent link with electrically reconfigurable IMDD and coherent modes for O-band data center applications.

We report the first O-band link with electrically reconfigurable intensity-modulation direct-detection (IMDD) and coherent operation using custom silicon photonic chips packaged with commercial electronic chips. Transmission below the KP4-FEC threshold is shown using commercial 53 Gbaud PAM4 digital signal processing (DSP) for 16QAM (200 Gbps/λ) and PAM4 (100 Gbps/λ). Efficient operation of the packaged full link at 12 and 11.2 pJ/bit is achieved for the PAM4 and 16QAM modes, respectively.

Large‐Scale High‐Density Multimode Optical Switch Matrix

AbstractOptical switching technology is emerging as a solution to the limitations of traditional electronic switching. Combining optical switching with mode‐division multiplexing effectively overcomes the limitations of expanding switching capacity solely by increasing the number of ports. This paper introduces a compact four‐mode 2 × 2 optical switch and a four‐mode reconfigurable non‐blocking 4 × 4 optical switch matrix based on the Beneš architecture. The multimode 2 × 2 switch achieves an extinction ratio larger than 15 dB for all modes in the wavelength range of 1530–1590 nm. The multimode 4 × 4 switch matrix has an extinction ratio exceeding 9 dB for all modes and states in the C‐band, with a compact size (1.46 × 0.23 mm2) and low power consumption (0–154 mW). It supports high‐integrity 50 Gbaud PAM4 signal transmission with a bit error rate below the forward error correction limit. These devices pave the way for enhanced integration density and capacity in optical switching systems.

Optical Amplifier-Free 100 Gbaud PAM-4 Transmission Over 10 Km SSMF Based on O-Band SiP Transceiver

Optical Amplifier-Free 100 Gbaud PAM-4 Transmission Over 10 Km SSMF Based on O-Band SiP Transceiver

Bias-drift insensitive full-field frequency response characterization of a thin-film lithium niobate-based intensity modulator.

We propose a rapid and precise scheme for characterizing the full-field frequency response of a thin-film lithium niobate-based intensity modulator (TFLN-IM) via a specially designed multi-tone microwave signal. Our proposed scheme remains insensitive to the bias-drift of IM. Experimental verification is implemented with a self-packaged TFLN-IM with a 3 dB bandwidth of 30 GHz. In comparison with the vector network analyzer (VNA) characterization results, the deviation values of the amplitude-frequency response (AFR) and phase-frequency response (PFR) within the 50 GHz bandwidth are below 0.3 dB and 0.15 rad, respectively. When the bias is drifted within 90% of the Vπ range, the deviation fluctuation values of AFR and PFR are less than 0.3 dB and 0.05 rad, respectively. With the help of the full-field response results, we can pre-compensate the TFLN-IM for the 64 Gbaud PAM-4 signals under the back-to-back (B2B) transmission, achieving a received optical power (ROP) gain of 2.3 dB. The versatility of our proposed full-field response characterization scheme can extend to various optical transceivers, offering the advantage of low cost, robust operation, and flexible implementation.

Local oscillators-free chip-enabled W-band wireless IM-DD PAM-4 data transmission with simultaneous dual-laser modulation and envelope detection.

Wireless data traffic is expected to exponentially increase in the future, and meeting this demand will require high data rate photonic-wireless links operating in the W-band (75-110 GHz). For this purpose, pulse-amplitude-modulation with four levels (PAM-4)-based intensity modulation and direct detection (IM-DD) photonic-wireless systems are preferred due to their simplified configuration. In this Letter, we present an experimental demonstration of an IM-DD PAM-4 photonic-wireless link in the W-band, leveraging a monolithic dual-laser photonic chip to enhance integration. Through injection-locking by an optical comb, the chip generates a W-band wireless signal via photo-mixing with a photodiode. This comb injection approach facilitates the phase correlation of the chip's two modes, resulting in a stabilized beat note. Additionally, the on-chip integration of the dual lasers enables the modulation of the two modes with a single modulator, improving the signal-to-noise ratio (SNR) while eliminating the need for extra splitters or combiners. Meanwhile, the envelope detector (ED) plays a crucial role in the simplified configuration, contributing to the overall decrease in size, weight, power, and complexity of the system. The integration of the chip-based phase-locked light source and the utilization of the ED thus signify noteworthy features of our experimental setup, which functions without the necessity of both optical and electrical local oscillators. PAM-4 signal modulation is simultaneously applied to the two coherent optical carriers. Our experiments have effectively transmitted 5 and 10 Gbaud PAM-4 W-band wireless signals in a cost-effective, lightweight, and straightforward configuration, achieving a line data rate of up to 20 Gbit/s economically. These experimental results demonstrate the practical potential of implementing fully integrated photonic-wireless transmitters.

Multiplication-free and hardware-efficient baud-rate timing error detector for Nyquist and faster than Nyquist optical IM/DD systems.

Clock recovery (CR) algorithms that support higher baud rates and advanced modulation formats are crucial for short-distance optical interconnections, and it is desirable to push CR to operate at baud rate with minimal computing resources and power. In this Letter, we proposed a hardware-efficient and multiplication operation-free baud-rate timing error detector (TED) as a solution to meet these demands. Our approach involves employing both the absolute value of samples and the nonlinear sign operation to emphasize the clock tone, which is deteriorated by severe bandwidth limitation in Nyquist and faster than Nyquist (FTN) systems. Through experimental investigations based on a transceiver system with a 3 dB bandwidth of 30 GHz, the proposed baud-rate TED exhibits excellent performance. The proposed scheme successfully achieves clock synchronization of the received signals with the transmitted signals, including 50 GBaud PAM4/8, 80 GBaud PAM4, and up to 120 GBaud PAM4 FTN signals. To the best of our knowledge, the CR based on the proposed baud-rate TED is the most optimal solution for ultrahigh-speed short-reach IM/DD transmission, comprehensively considering the timing jitter, bit error rate (BER), and implementation complexity.

26.5625 Gbaud PAM-4 short-reach transmission at 75°C using 850 nm VCSELs with strategic oxide guiding layer positioning.

This article presents an all-epitaxy approach to reduce the root mean square spectral width (ΔλR M S) of 850 nm oxide-confined vertical cavity surface-emitting lasers (VCSELs) with a large aperture of 7 µm through strategic optimization of the oxide guiding layer within the epitaxy structure. At 75°C, the VCSEL demonstrates a ΔλR M S of ∼0.3 nm at a bias current of 7.5 mA. Furthermore, the VCSEL achieves successful transmission of 26.5625 Gbaud PAM-4 modulation over a short-reach (SR) OM4 fiber link while maintaining a TDECQ budget below the 4.5 dB specified by 50G IEEE Ethernet standards.

Low-complexity and low-latency equalization technique - probabilistic noise cancellation.

Both inside data centers (DCs) and in short optical links between data centers (DC campuses), intensity-modulation and direct-detection (IMDD) systems using four-level pulse amplitude modulation (PAM4) will dominate this decade due to low transceiver price and power consumption. The next DC transceiver generation based on 100 Gbaud PAM4 will require advanced digital signal processing (DSP) algorithms and more powerful forward error correction (FEC) codes. Because of bandwidth limitations, the conventional DC DSP based on a few-tap linear feed-forward equalizer (FFE) is likely to be upgraded to more complex but still low-complexity Volterra equalizers followed by a noise whitening filter and either a maximum likelihood sequence estimation (MLSE) or a maximum a posteriori probability (MAP) algorithm. However, stringent power consumption and latency requirements may limit the use of complex algorithms such as decision feedback equalizer (DFE) or MLSE/MAP in DC networks (DCN). In this paper, we introduce a low-complexity, low-latency algorithm based on a feedforward structure, yielding a performance between DFE and MLSE. We call the novel equalization algorithm probabilistic noise cancellation (PNC), since it weights noise patterns based on their probabilities in the presence of bandwidth limitations. The probabilistic weighting is efficiently exploited in correcting correlated errors caused by noise coloring in the FFE.

An optimized LSTM-based equalizer for 100 Gigabit/s-class short-range fiber-optic communications

<p>Intensity modulation/direct detection (IM/DD) remains to be the preferred optical transmission scheme for short-range applications for its simplicity of application, inexpensiveness, and small footprint. However, the impairments of low-cost device and fiber chromatic dispersion lead to the limitation of system performance when the data rate rises to 100 Gbps or higher. In this paper, we demonstrated that an equalizer using neural networks can effectively improve the transmission performance of high-speed IM/DD systems. An optimization of a long short-term memory (LSTM) structure in terms of network depth and distribution of neurons in hidden layers leads to an enhancement of the overall performance of the 50 Gbaud PAM4 communications. Furthermore, the results for a system using a LSTM-based equalizer give the better outcome than the traditional feed-forward equalizer (FFE) or artificial neural network (ANN)-based equalizer.</p>

Deep Learning Equalizer Connected with Viterbi-Viterbi Algorithm for PAM D-Band Radio over Fiber Link.

D-band (110-170 GHz) has been regarded as a potential candidate for the future 6G wireless network because of its large available bandwidth. At present, the lack of electrical amplifiers operating in the high frequency band and the strong nonlinear effect, i.e., the D-band, are still important problems. Therefore, effective methods to mitigate the nonlinear issue resulting from the ROF link are indispensable, among of which machine learning is considered the most effective paradigm to model the nonlinear behavior due to its nonlinear active function and structure. In order to reduce the computation amount and burden, a novel deep learning neural network equalizer connected with typical mathematical frequency offset estimation (FOE) and carrier phase recovery (CPR) algorithms is proposed. We implement D-band 45 Gbaud PAM-4 and 20 Gbaud PAM-8 ROF transmission simulations, and the simulation results show that the real value neural network (RVNN) equalizer connected with the Viterbi-Viterbi algorithm exhibits better compensation ability for nonlinear impairment, especially when dealing with serious inter-symbol interference and nonlinear effects. In our experiment, we employ coherent detection to further improve the receiver sensitivity, so a complex baseband signal after down conversion at the receiver is inherently produced. In this scenario, the complex value neural network (CVNN) and RVNN equalizer connected with the Viterbi-Viterbi algorithm have better BER performance with an error rate lower than the HD-FEC threshold of 3.8 × 10-3.

Transmitter dispersion eye closure quaternary assessment based on linearCNN with 1 × 1 convolutional kernel.

Transmitter dispersion eye closure quaternary (TDECQ) is a vital metric to characterize the quality of four-level pulse amplitude modulation (PAM-4) optical signals. However, the traditional TDECQ assessment scheme is complex and time consuming, with heavy iterative operations. Therefore, accelerating the TDECQ assessment has great significance for photonic data-center interconnection (DCI) applications. Here, we propose and experimentally demonstrate a TDECQ assessment based on linear-convolutional neural network (L-CNN) with the 1 × 1 convolutional kernel to reduce the implementation complexity. Our experimental results verify that the lightweight L-CNN can realize the accurate TDECQ assessment, without the involvement of nonlinear activation functions (NAFs). The mean absolute error (MAE) of 26.5625 and 53.125 GBaud PAM-4 signals are 0.16 dB and 0.18 dB, respectively, over a TDECQ range from 1.5 to 4.0 dB. Meanwhile, in comparison with existing CNN-based schemes, the L-CNN based TDECQ assessment scheme only needs 2048 multiplications, which have been reduced by five orders of magnitude.

High Spectral Efficiency Long-Wave Infrared Free-Space Optical Transmission With Multilevel Signals

This study explores the potential of long-wave infrared free-space optical (FSO) transmission that leverages multilevel signals to attain high spectral efficiency. The FSO transmission system consists of a directly modulated-quantum cascade laser (DM-QCL) operating at 9.15 μm and a mercury cadmium telluride (MCT) detector. To fully understand the system, we conduct measurements on the DM-QCL chip and MCT detector and assess the overall amplitude response of the DM-QCL, MCT detector, and all electrical components. We apply various signals, including on-off keying (OOK), 4-level pulse amplitude modulation (PAM4), 6-level PAM (PAM6), and 8-level PAM (PAM8) to maximize the bit rate and spectral efficiency of the FSO transmission. Through a two-dimensional sweeping of the laser bias current and MCT detector photovoltage, we optimize the transmission performance. At the optimal operation point, the FSO system achieved impressive results which are up to 6 Gbaud OOK, 3.5 Gbaud PAM4, 3 Gbaud PAM6, and 2.7 Gbaud PAM8 signal transmissions, with a bit error rate performance below 6.25% overhead hard decision-forward error correction limit when the DM-QCL operates at 10°C. We also evaluate the eye diagrams and stability of the system to showcase its remarkable transmission performance. Our findings suggest that the DM-QCL and MCT detector-based FSO transceivers offer a highly competitive solution for the next generation of optical wireless communication systems.

Si3N4-plasmonic ferroelectric MZIR modulator for 112-Gbaud PAM-4 transmission in the O-band.

This paper presents a simulation-based analysis on the performance of plasmonic ferroelectric Mach-Zehnder in a ring (MZIR) versus symmetric Mach-Zehnder modulators (MZMs) on Si3N4 targeting O-band operation. The detailed investigation reveals the tradeoff between Au and Ag legacy noble metals providing lower modulator losses and CMOS compatible Cu featuring low cost. The numerical models also show that by opting for the MZIR layout there is a reduction in the Vπ x L product of 46% for Ag, 39% for Au and 30% for Cu versus MZMs. Time-domain simulations verify the successful generation of 112 Gbaud PAM-4 Signals from both MZIRs and MZMs for as low as 2 × 1.3 Vpp and 5µm long plasmonic phase shifters (PSs) with MZIRs providing a ΔQ signal improvement over MZMs of 2.9, 2.4, and 1.3 for Ag, Au, and Cu metals respectively. To the best of our knowledge, this is the first theoretical demonstration of such a low-loss, low-voltage, high-speed, and CMOS compatible plasmonic modulator on Si3N4, in the O-band.

Accelerating Transmitter Dispersion Eye Closure Quaternary Assessment by Deep Transfer Learning Technique

Transmitter dispersion eye closure quaternary (TDECQ) is one of the most important parameters that can systematically evaluate the signal quality of four-level pulse amplitude modulation format (PAM-4) transmitters. We experimentally demonstrate the acceleration of TDECQ assessment by deep transfer learning (DTL) technique. With the help of amplitude histograms (AHs) generated by the received PAM-4 signal, we are able to characterize the TDECQ value directly by the deep neural network (DNN). To further reduce both the computation time and the experimental data used for the training, simulation data can be used to provide the suitable source domain for the DTL implementation. The experimental results of 25 Gbaud PAM-4 signals indicate that, the mean absolute error of TDECQ assessment is below 0.18 dB over a TDECQ range from 1.9 to 3.6 dB. Meanwhile, 33.3% amount of the used experimental data and 40% epochs can be simultaneously reduced during the training process, due to the use of DTL.

800 Gbit/s QSFP-DD Transceiver Based on Thin-Film Lithium Niobate Photonic Integrated Circuit

800G Ethernet is expected to be the dominant solution for the next generation inter- and intra-data-center connections. To boost the transmission capacity to 800Gb/s, utilizing multi-level pulse amplitude modulation (PAM) transmitting format and adopting multiple lanes are competitive solutions. In this paper, we demonstrate a PAM4-modulated 800G QSFP-DD transceiver integrated with two 4×100G thin-film lithium niobate (TFLN) DR4 modulator chips. The DR4 modulator chips have a 13.5 dB insertion loss (including the inherent 3 dB modulation loss and 6 dB splitting loss) and one DR4 chip requires one 17 dBm light source. The EO response has a 40 GHz bandwidth at 2.3 dB and the electric reflection keeps below -12 dB. The eight channels of 800G transceiver have a clear open eye patterns at 1.5 Vpp driving voltage under 53.125 Gbaud PAM-4 modulation with all the ERs above 4.03 dB and TDECQ's below 2.13 dB. The sensitivity of the receiver is around -8.6 dBm after 2 km and back-to-back transmission with the BER of all the channels below the DR4 forward error correcting threshold. It is demonstrated that our 800G QSFP-DD transceiver meets 400GBASE-DR4 standards and packaging requirements simultaneously. This is the first time that 4×100G DR4 modulator array based on TFLN and 800G QSFP-DD transceiver based on TFLN are reported to the best of our knowledge.

200 Gb/s Optical-Amplifier-Free IM/DD Transmissions Using a Directly Modulated O-Band DFB+R Laser Targeting LR Applications

We experimentally demonstrate an O-band single-lane 200 Gb/s intensity modulation direct detection (IM/DD) transmission system using a low-chirp, broadband, and high-power directly modulated laser (DML). The employed laser is an isolator-free packaged module with over 65-GHz modulation bandwidth enabled by a distributed feedback plus passive waveguide reflection (DFB+R) design. We transmit high baud rate signals over 20-km standard single-mode fiber (SSMF) without using any optical amplifiers and demodulate them with reasonably low-complexity digital equalizers. We generate and detect up to 170 Gbaud non-return-to-zero on-off-keying (NRZ-OOK), 112 Gbaud 4-level pulse amplitude modulation (PAM4), and 100 Gbaud PAM6 in the optical back-to-back configuration. After transmission over the 20-km optical-amplifier-free SSMF link, up to 150 Gbaud NRZ-OOK, 106 Gbaud PAM4, and 80 Gbaud PAM6 signals are successfully received and demodulated, achieving bit error rate (BER) performance below the 6.25%-overhead hard-decision (HD) forward-error-correction code (FEC) limit. The demonstrated results show the possibility of meeting the strict requirements towards the development of 200Gb/s/lane IM/DD technologies, targeting 800Gb/s and 1.6Tb/s LR applications.

420 Gbit/s optical signal reception enabled by an inductive gain peaking Ge-Si photodetector with 80 GHz bandwidth.

Based on the commercial silicon photonics (SiPh) process platform, a flat 3 dB bandwidth of 80 GHz germanium-silicon (Ge-Si) photodetector (PD) is experimentally demonstrated at a photocurrent of 0.8 mA. This outstanding bandwidth performance is achieved by using the gain peaking technique. It permits an 95% improvement in bandwidth without sacrificing responsivity and undesired effects. The peaked Ge-Si PD shows the external responsivity of 0.5 A/W and internal responsivity of 1.0 A/W at a wavelength of 1550 nm under -4 V bias voltage. The high-speed large signal reception capability of the peaked PD is comprehensively explored. Under the same transmitter state, the transmitter dispersion eye closure quaternary (TDECQ) penalties of the 60 and 90 Gbaud four-level pulse amplitude modulation (PAM-4) eye diagrams are about 2.33 and 2.76 dB, 1.68 and 2.45 dB for the un-peaked and peaked Ge-Si PD, respectively. When the reception speed increase to 100 and 120 Gbaud PAM-4, the TDECQ penalties are approximatively 2.53 and 3.99 dB. However, for the un-peaked PD, its TDECQ penalties cannot be calculated by oscilloscope. We also measure the bit error rate (BER) performances of the un-peaked and peaked Ge-Si PDs under different speed and optical power. For the peaked PD, the eye diagrams quality of 156 Gbit/s nonreturn-to-zero (NRZ), 145 Gbaud PAM-4, and 140 Gbaud eight-level pulse amplitude modulation (PAM-8) are as good as the 70 GHz Finisar PD. To the best of our knowledge, we report for the first-time a peaked Ge-Si PD operating at 420 Gbit/s per lane in an intensity modulation direct-detection (IM/DD) system. It might be also a potential solution to support the 800 G coherent optical receivers.

28 Gbaud PAM-4 Burst-Mode CDR With Reconfigurable Sampling Scheme

This paper presents a fast-locking 28 Gbaud PAM-4 CDR targeting for burst-mode operation. By implementing an oversampling scheme tailored for the preamble stage signal according to the pattern appearance regulation, the trade-off between the loop bandwidth and the convergence time in the conventional CDR is avoided. In addition, to reduce power consumption in the oversampling mode, a 3-phase sampling scheme equivalent to 3 times-oversampling with a minimal phase shift range of 30 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{\circ}$</tex-math> </inline-formula> is proposed. The PAM-4 burst-mode CDR also includes a conventional Bang-Bang phase locking loop with a bandwidth of 5 MHz. Realized in a 65nm bulk CMOS technology, the CDR chip achieves 28 Gbaud PAM-4 burst-mode clock recovery within 10 ns. With integrating slicers, phase interpolators and output buffers, the total power consumption of the CDR is 154 mW. Compared with the prior arts, to the best of authors’ knowledge, the proposed CDR first achieves on-chip phase-locking with PAM-4 Burst signals, based on the proposed PAM-4 burst-mode signal tailored 3-phase oversampling scheme.