PAM-4 Transmission Research Articles

50 Gbps PAM-4 Over Up to 80-km Transmission With C-Band DML Enabled by Post-Equalizer

In this letter, transmission of directly-modulated PAM-4 in C-band is experimentally demonstrated. Instead of introducing any optical devices for dispersion management, we employ receiver-side equalization consisting of Volterra-based FFE and Volterra-based DFE, to mitigate the signal distortions induced by the frequency chirp effect and chromatic dispersion. Results indicate that, PAM-4 transmissions of 56 Gbps over 60-km and 60 Gbps over 80-km are achieved respectively under the HD-FEC limit ( $3.8\times 10^{-3}$ ) and SD-FEC limit ( $2.2\times 10^{-2}$ ).

Application analysis of clipping and digital resolution enhancer in high-speed direct-detection PAM4 transmission.

In this paper, the application of a clipping and digital resolution enhancer (DRE) is numerically investigated in 56-GBaud direct-detection four-ary pulse amplitude modulation (PAM4) transmission systems. The influence of the tap length and the digital-to-analog converter (DAC) samples per symbol on DRE gain is first studied. After optimizing clipping probability, sampling jitter is introduced. The combination of clipping and DRE can increase jitter tolerance by ∼1.7% sampling time for 4-bit DAC. Then, the required optical signal-to-noise ratio (OSNR) at the hard-decision forward error correction (HD-FEC) threshold of 3.8×10-3 is investigated in back-to-back case and 80-km fiber transmission. It is shown that the combination of clipping and DRE can reduce the required OSNR significantly for 3-bit and 4-bit DAC. Finally, the performance gain introduced by clipping and DRE is analyzed from a perspective of the transmitted average signal power and the received mean-squared quantization error.

A Circuit-Level Implementation of Voltage-Tuning Scheme for Realizing Optical PAM-4 Using Three-Segment Microring Modulator

Silicon Photonics, as one of the solutions to satisfy ever-increasing data bandwidth growth, becomes more challenging due to the latest technologies such as Internet of Things (IoT). Higher order pulse amplitude modulation (PAM) schemes is one of the answers to push towards higher data transmission in the presence of bandwidth limited optical devices. In this paper, we have implemented a circuit-level PAM-4 transmitter design based on the voltage-tuning scheme for realizing optical PAM-4 using a three-segment microring modulator. Simulation results based on the extracted layout using TSMC 65nm LP technology and IMEC-ePIXfab SiPhotonics ISIPP50G technology show that our proposed circuit-level transmitter structure is able to achieve PAM-4 data rate of 25-Gb/s with extinction ratio of 9dB and PAM-4 energy efficiency of 0.5pJ/bit. The results also verify that the scheme is able to achieve high tuning flexibility, but the proposed transmitter will consume more power as a result.

35 km amplifier-less four-level pulse amplitude modulation signals enabled by a 23 GHz ultrabroadband directly modulated laser

The 4-level pulse amplitude modulation (PAM4) based on an 23 GHz ultrabroadband directly modulated laser (DML) was proposed. We have experimentally demonstrated that based on intensity modulation and direct detection (IMDD) 56 Gbps per wavelength PAM4 signals transferred over 35 km standard single mode fiber (SSMF) without any optical amplification and we have achieved the bit error rate (BER) of the PAM4 transmission was under 2.9 × 10−4 by using feed forward equalization (FFE).

Compressed Neural Network Equalization Based on Iterative Pruning Algorithm for 112-Gbps VCSEL-Enabled Optical Interconnects

Advanced nonlinear digital signal processing technologies, which bring significant performance gain for high-speed optical interconnects, are highly constrained by huge complexity in the actual deployment. Fully connected neural network-based equalizer has shown powerful efficacy to deal with the complex linear and nonlinear impairments for VCSEL-enabled multi-mode optical interconnects, but it also contains a number of redundancies with little impact on performance improvement. In this article, we experimentally demonstrate a compressed neural network equalization using the iterative pruning algorithm for 112-Gbps VCSEL-enabled PAM-4 and PAM-8 transmissions over 100-m MMF. We also study the impact of threshold and pruning span on the performance of proposed algorithms. The results show up to 71% connection compression by use of the iterative pruning algorithms and maximum 28.4% improvement compared with the one-shot pruning algorithm.

Passively mode-locked quantum dash laser with an aggregate 5.376 Tbit/s PAM-4 transmission capacity.

This paper presents an InAs/InP quantum dash (QD) C-band passively mode-locked laser (MLL) with a channel spacing of 34.224 GHz. By using this QD-MLL we demonstrate an aggregate 5.376 Tbit/s PAM-4 data transmission capacity both for back-to-back (B2B) and over 25-km of standard single mode fiber (SSMF). This represents the first demonstration of QD-MLL acting as error-free operation at an aggregate data transmission capacity of 5.376 Tbit/s for some filtered individual channels. This finding highlights the viability for InAs/InP QD lasers to be used as a low-cost optical source for data center networks.

100 Gb/s/$\lambda$ Duo-Binary PAM-4 Transmission Using 25G Components Achieving 50 km Reach

We demonstrate 100 Gb/s/ $\lambda $ duo-binary PAM-4 (DB-PAM-4) transmission over 50 km standard single-mode fiber (SMF) in the O-band using a 20G silicon-based transmitter and a 25G optically-amplified PIN+TIA-based receiver for 100G PON downstream applications. Receiver sensitivities of −15, −14.5 and −12 dBm are achieved at a bit error rate (BER) below the low-density parity-check forward error correction (LDPC-FEC) threshold (i.e., $1\times 10^{-2}$ ) in back-to-back (B2B), and after 20 and 50 km, respectively using a 131-tap linear feed-forward equalizer (FFE) at the receiver. According to the results, a power budget of 20 dB can be achieved for 20 km transmission. Furthermore, we report the performance of 99 Gb/s (33 Gbaud) PAM-8 transmission and show that DB-PAM-4 provides 5.3 dB better sensitivity compared to PAM-8 at the LDPC-FEC threshold.

On-Chip 4×10 GBaud/s Mode-Division Multiplexed PAM-4 Signal Transmission

Emerging 5G mobile networks and cloud computing applications are driving the demand for an ever-increasing capacity of short-reach optical communications. To meet this demand, mode-division multiplexing (MDM) has been proposed to scale up the bandwidth density by leveraging the spatial modes of an optical waveguide for transmitting multiple optical signals. On the other hand, the use of multi-level pulse amplitude modulation (PAM) can also increase the transmission bandwidth. Therefore, on-chip MDM in conjunction with PAM is an approach to enhance the transmission capacity in a photonic integrated circuit. In this paper, we report a silicon photonic integrated four-channel MDM circuit for high data rate on-chip communications with a low channel crosstalk and small insertion loss. To make the circuit have a small size that supports broadband operation, a mode multiplexer and demultiplexer are realized with the use of cascaded asymmetrical directional couplers on rib waveguides. By incorporating the MDM circuit in an optical communications system, the transmission of a 4 x 10 GBaud/s OOK and PAM-4 signal is experimentally demonstrated. The performance in terms of eye diagrams and power penalties is evaluated. The power penalties for the four-channel OOK transmission are 4.31, 2.38, 1.44 and 3.5 dB at a BER of 10−9. For the four-channel PAM-4 transmission, the power penalties are 7.98, 1.10, 0.66 and 5.14 dB. The required received optical power at a BER (<3.8 × 10−3) of the 7% overhead-hard decision FEC is −2.6 dBm. The key advantage of the approach is that high-capacity on-chip communications is enabled by the photonic integrated MDM circuit with a small footprint. Since the MDM circuit is implemented on a silicon photonic platform, the system holds high potential for full integration on a single chip.

Experimental Demonstration of 600 Gb/s Net Rate PAM4 Transmissions over 2 km and 10 km with a 4-λ CWDM TOSA

We propose to introduce a controlled amount of inter-symbol interference (ISI) during pulse shaping in a high symbol rate intensity modulation direct detection (IM/DD) system to increase the alternating current (AC) power of the received electrical signal after photo-detection and transimpedance amplification. We experimentally demonstrate high-speed 4-level pulse amplitude modulation (PAM4) transmissions with a 4-λ coarse wavelength division multiplexing (CWDM) transmitter optical sub-assembly (TOSA) using the proposed scheme. The intentionally introduced ISI is mitigated using either transmitter-side Tomlinson Harashima precoding (THP) or receiver-side feed forward equalization (FFE) aided with a 2-tap post filter and maximum likelihood sequence estimation (MLSE). Both methods enable 4 × 81 Gbaud PAM4 transmission over 2 km of single mode fiber (SMF), with a bit error rate (BER) below the 7% overhead hard-decision forward error correction (HD-FEC) threshold of 3.8 × 10−3, which corresponds to a net rate of >600 Gb/s. We also explore the limit of our system at 10 km. The result shows that an aggregated net rate of 600 Gb/s excluding the HD-FEC overhead can still be achieved by loading different symbol rates onto different wavelength channels.

A 40-Gb/s PAM-4 Transmitter Using a 0.16-pJ/bit SST-CML-Hybrid (SCH) Output Driver and a Hybrid-Path 3-Tap FFE Scheme in 28-nm CMOS

This paper proposes an SST-CML-Hybrid (SCH) output driver, and its corresponding hybrid-path feed-forward equalization (FFE) scheme, to enhance the energy efficiency of a PAM-4 transmitter (TX). Specifically, the SCH driver features one SST branch + one CML branch to co-synthesize the PAM-4 data, reducing substantially the signaling power, switching power and equalization power. The PAM-4 TX further integrates a half-rate serializer with 4-bit 3-tap FFE, duty-cycle correction circuits and a T-coil output matching network. Prototyped in 28-nm CMOS, the PAM-4 TX achieves a broadband return loss <−10dB up to 50 GHz, and occupies a compact die area of 0.0345 mm2. Operating at 40 Gb/s and at a 0.9-V supply, the TX dissipates 19.5 mW, of which 6.4 mW is due to the SCH driver. The corresponding energy efficiencies are 0.16 and 0.5 pJ/bit for the SCH driver and TX, respectively; both compare favorably with the prior art.

137 Gb/s PAM-4 Transmissions at 850 nm over 40 cm Optical Backplane with 25 G Devices with Improved Neural Network-Based Equalization

An improved neural network-based equalization method is proposed and experimentally demonstrated. The up-to-137 Gb/s transmission of four level pulse amplitude modulation (PAM-4) signals with 25 G class 850 nm optical devices is achieved over an in-house fabricated 40 cm optical backplane. An in-depth investigation is conducted regarding the impact of delayed taps and spans on equalization performance. A performance comparison of the proposed method with the traditional maximum likelihood sequence estimation (MLSE) and decision feedback equalization (DFE) is also undertaken. For the bit rate from 80 to 100 Gb/s, the proposed method achieves an adopted hard-decision forward error correction (HD-FEC) requirement at a received optical power (RoP) of −9 and −8 dBm, while DFE and MLSE cannot meet the HD-FEC requirement. When the bit rate increases from 120 to 137 Gb/s, the proposed equalization method still successfully maintains the acceptable system performance at an RoP of −4 and −2.5 dBm. Furthermore, the specific bit error rate (BER) performances for varied maximum achievable bit rate under different RoPs by applying MLSE and the proposed method are also analyzed. This provides an important potential solution to realize the future data centers.

A SiGe HBT BiCMOS 1-to-4 ADC Frontend Enabling Low Bandwidth Digitization of 100 GBaud PAM4 Data

Market forecasts for intra and inter data center interconnects predict high growth rates in the upcoming years. This is associated with increasing requirements in terms of costs, power consumption, and bit rate per wavelength. Today, PAM4 systems are frequently used at up to 100 Gb/s per wavelength. The next logical step is to double the symbol rate to allow bit rates of 200 Gb/s at the same modulation format and at the same direct detection scheme. Such systems require three-digit gigasample analog-to-digital converters with bandwidths larger than 50 GHz, which are not available today as CMOS components. To extend the bandwidth of CMOS converters, we suggest the use of ADC frontends, which sample and demultiplex the analog input signals simultaneously. At the outputs of such ADC frontends, the signals can then be digitized by parallel converters with lower bandwidth and at a lower conversion rate. In this article, we present the circuit implementation of a 1-to-4 ADC frontend in IHP 130 nm SiGe HBT BiCMOS and demonstrate its integration into a 100 GBaud PAM4 transmission system with only 14 GHz of digitizer bandwidth. For optimum performance, the mitigation of intersymbol and interchannel interferences is mandatory in the transmission system. We describe and compare two different receiver DSP architectures for reduction of these interferences and evaluate their performances in electrical and optical back-to-back experiments. Furthermore, we report on transmission experiments over 500 m, 10 km, and 40 km SMF at 1550 nm wavelength. With linear equalization, we obtained a BER below the HD-FEC threshold of 3.8 × 10−3 and with 3 rd order Volterra equalization, we even achieved a BER below the KP4 threshold of 2.3 × 10 −4 .

Complexity Reduction With a Simplified MIMO Volterra Filter for PDM-Twin-SSB PAM-4 Transmission

We propose and experimentally demonstrate a simplified multiple-input multiple-output Volterra filter (MIMO-VF) in a polarization division multiplexed (PDM) twin single side-band (SSB) scheme based on two offset carriers. The right side-band (RSB) and left side-band (LSB) signals of the twin-SSB in each polarization are separated by an optical filter, and the crosstalk between the RSB and LSB is mitigated in the receiver digital signal processing (DSP) by the proposed simplified MIMO-VF, which is combined with one two-stage equalizer, i.e., a linear 2 × 2 butterfly MIMO equalizer and two independent VFs. With the proposed simplified MIMO-VF, we experimentally demonstrate a 400-Gb/s four-level pulse amplitude modulation (PAM-4) signal transmission over an 80-km standard single mode fiber (SSMF). Compared with the traditional MIMO-VF, the proposed simplified MIMO-VF can retain the same transmission performance and reduce the computational complexity by approximately 57.9%.

Interband Short Reach Data Transmission in Ultrawide Bandwidth Hollow Core Fiber

We report a Nested Antiresonant Nodeless hollow-core Fiber (NANF) operating in the first antiresonant passband. The fiber has an ultrawide operational bandwidth of 700 nm, spanning the 1240–1940 nm wavelength range that includes the O-, S-, C- and L- telecoms bands. It has a minimum loss of 6.6 dB/km at 1550 nm, a loss ≤7 dB/km between 1465–1655 nm and ≤10 dB/km between 1297–1860 nm. By splicing together two structurally matched fibers and by adding single mode fiber (SMF) pigtails at both ends we have produced a ∼1 km long span. The concatenated and connectorized fiber has an insertion loss of approximately 10 dB all the way from 1300 nm to 1550 nm, and an effectively single mode behavior across the whole spectral range. To test its data transmission performance, we demonstrate 50-Gb/s OOK data transmission across the O-to L-bands without the need for optical amplification, with bit-error-rates (BERs) lower than the 7% forward error correction (FEC) limit. With the help of optical amplification, 100-Gb/s PAM4 transmission with BER lower than the KP4 FEC limit was also achieved in the O/E and C/L bands, with relatively uniform performance for all wavelengths. Our results confirm the excellent modal purity of the fabricated fiber across a broad spectral range, and highlight its potential for wideband, low nonlinearity, low latency data transmission.

Comparison on OM5-MMF and OM4-MMF Data Links With 32-GBaud PAM-4 Modulated Few-Mode VCSEL at 850 nm

Based on the pre-emphasis 4-level pulse amplitude modulation (PAM-4) format at a symbol rate of 32 GBaud, the comparisons among optical PAM-4 data transmissions through back-to-back (BtB), 100-m OM4 multi-mode fiber (MMF), and 100-m OM5-MMF links are discussed in detail by employing the directly modulated few-transverse-mode (FM) vertical cavity surface emitting laser (VCSEL) chip at 850 nm as the carrier. The bare FM-VCSEL chip exhibits an oxide-confined aperture of 5 μm to provide 3–4 transverse modes with low threshold current (Ith) of 0.25 mA and high optical power of 1.2 mW at a bias current of 38 Ith. The impendence of 65 Ω and the 3-dB bandwidth of 24 GHz are obtained by setting the DC point of the FM-VCSEL at 6.25 mA. With the bias current optimization and the pre-emphasis, the receiving bit error ratio (BER) of the 32-GBaud PAM-4 data meets the KP4-forward error correction criterion of 2.4 × 10−4, which enables the FM-VCSEL successfully achieving the 64-Gbit/s data stream transmission through the BtB and 100-m OM4-/OM5-MMF links. As compared to the BtB transmission with a BER of 2.5 × 10−8, replacing OM4 with OM5 for the 64-Gbit/s PAM-4 transmission reduces the receiving power penalty from 3.41 to 0.86 dB and improves the BER from 3.8 × 10−5 to 3.8 × 10−6. That is mainly attributed to the higher effective modal bandwidth and compensated modal and chromatic dispersions of the OM5-MMF as compared to the OM4-MMF when delivering the FM-VCSEL carried PAM-4 data stream.

A 112-Gb/s PAM-4 Transmitter With a 2-Tap Fractional-Spaced FFE in 65-nm CMOS

This letter presents a 112-Gb/s four-level pulse amplitude modulation (PAM-4) transmitter with a 2-tap fractional-spaced feed-forward equalizer (FFE) implemented in a 65-nm CMOS process. The tap delay is adjusted by a coarse-fine capacitor array-based delay cell located in the clock path. Dynamic latches, , and bandwidth-enhanced 4:1 MUXs are employed to guarantee an adequate timing margin and a sufficient bandwidth for the 112-Gb/s design. The measurement results show that the fractional-spaced FFE can significantly optimize the eye opening. The fabricated PAM-4 transmitter achieves a maximum data rate of 112 Gb/s with an energy efficiency of 2.17 pJ/bit.

A 112 Gb/s PAM4 Silicon Photonics Transmitter With Microring Modulator and CMOS Driver

Microring modulators (MRMs) with CMOS electronics enable compact low power transmitter solutions for 400G Ethernet and future on-package optical transceivers. In this paper, we present a 112 Gb/s PAM4 transmitter using silicon photonic MRM, on-chip laser and co-packaged 28 nm CMOS driver. We describe the impact of static and dynamic MRM nonlinearity on PAM4 signaling and present a dual path nonlinear pre-distortion technique to compensate both effects. PAM4 measurement results of our transmitter at 112 Gb/s show that TDECQ <0.7 dB is achieved from 30 °C to 60 °C while dissipating 6 pJ/bit. We also present link level measurements at 112 Gb/s PAM4 obtained by coupling this transmitter with our previously published CMOS TIA-based receiver, to demonstrate the feasibility of low cost optical transceivers through CMOS integration of optical interface circuits.

Machine learning classifier based on FE-KNN enabled high-capacity PAM-4 and NRZ transmission with 10-G class optics

Modified by special feature engineering, a powerful and low-order equalizer based on K-nearest neighbors (KNN) classifier is applied to improve performance of high-speed system with bandwidth-limited optics. The feature construction and feature weighting are specially designed to conduct an appropriate a feature engineering-based KNN (FE-KNN) scheme, which contains more data characteristics to enhance the equalization performance. Experimental comparisons of KNN classifier with/without feature engineering, decision feedback equalizer (DFE) and feed-forward equalizer (FFE) are implemented to prove the feasibility of our scheme in both 25-Gb/s NRZ and 50-Gb/s PAM-4 transmission experiments with 10-G optics system. The corresponding results show that, without the feature engineering, the performance achieved by the common KNN is not improved even in the case of hard decision (HD). In contrast, compared to the common 11-taps DFE, the performance achieved by FE-KNN with only 5 taps is improved by 1-dB at KP4-FEC threshold (BER=2.2E-4) for 25-Gb/s NRZ transmission. While, for 50-Gb/s PAM-4 case, 0.5-dB sensitivity improvement is achieved by our scheme compared to the common 11-taps DFE under the HD-FEC limit (BER=3.8E-3).

C-band 56-Gb/s PAM4 transmission over 80-km SSMF with electrical equalization at receiver.

In this paper, a memory polynomial equalizer combined with decision feedback equalizer (MPE-DFE) is proposed to eliminate channel distortions for intensity modulation and direct detection (IM/DD) systems. Compared with traditional feedforward equalizer and decision feedback equalizer (FFE-DFE), the proposed MPE-DFE introduces extra square terms and cubic terms to jointly equalize chromatic dispersion and nonlinear distortions. We demonstrated a C-band 56-Gb/s four-level pulse-amplitude modulation (PAM4) system over 80-km standard single mode fiber (SSMF) transmission. Experimental results show that the proposed MPE-DFE achieved up to 6.2 dB higher SNR than traditional FFE-DFE. Moreover, the achieved bit error ratio (BER) with MPE-DFE reaches 3.1 × 10-3, which is below 7% feedforward error correction (FEC) threshold of 3.8 × 10-3. To the best of our knowledge, we achieved a record transmission distance for C-band 56-Gb/s PAM4 signal with only electrical equalization at the receiver.

Fast remodeling for nonlinear distortion mitigation based on transfer learning.

We have proposed and demonstrated a transfer learning (TL)-assisted deep learning network (DNN) for nonlinear distortion compensation in optical side-band PAM-4 modulation and direct-detection transmission. Since there exists partial correlation of nonlinear distortions, we can transfer the parameters of the trained DNN to the target model to speed up remodeling and reduce complexity. We conduct experiments to demonstrate the effectiveness of the proposed scheme in Nyquist PAM-4 transmissions. The required iterations or train size with TL can be less than half of that with retraining without any performance degradation in single-channel transmission. We also extend the proposed scheme into multi-channel transmissions. Since all channels co-propagate on the same fiber, the correlation of nonlinear distortions enables TL to share the parameters among different channels and realize fast remodeling with a better starting rather than a stochastic beginning. The experimental results show that the iterations of TL are one-fourth that of retraining without performance penalty in five-channel transmission.