We demonstrate a data-driven framework for emulating high-speed VCSEL-based four-level pulse amplitude modulation (PAM-4) optical interconnects using bidirectional long short-term memory (Bi-LSTM) networks. Unlike conventional rate-equation models, which are computationally intensive and often require difficult parameter tuning, our approach utilizes experimental waveforms to learn the end-to-end system dynamics. By employing transfer learning and weight interpolation, we extend the model to new operating regimes with a 20-fold reduction in computation time compared to independent training, while maintaining a normalized mean squared error below 0.04. This emulator provides a rapid, accurate tool for the design and optimization of short-reach optical links.

We analyze the symbol rate-code rate trade-off in bandwidth-limited IM/DD systems targeting LPO transceivers, using PAM-4/6/8 as candidate modulation formats. We use the capacity of the binary symmetric channel as an achievable information rate under hard-decision decoding, serving as a performance metric for 200G and 400G per-lane throughput targets. Our results show that reducing the FEC code rate below the KP4 baseline allows higher-order PAM formats to operate at substantially lower symbol rates than PAM-4 while meeting the throughput requirement.

The first outdoor field-trials of 3x1-distributed fiber-wireless mmWave antenna-system are experimentally presented, delivering uninterrupted real-time MEC-services at 2Gb/s user-rate and 0.195ms latency over an SDN-controlled coordinated multipoint X-haul with 7km fiber and 30m radio-distance. (c) 2025 The Author(s).

An SDN-orchestrated Fiber-Wireless FSO/mmWave X-haul with 100 Gb/s real-time, low-latency service-provisioning and 24hour-long weather-resilient daylight/night-time operation relying on automated zero-touch handovers is experimentally presented, achieving an order of magnitude speed enhancement in real-time hybrid links. (c) 2025 The Author(s)

Considering the challenges faced by the current Central Offices, this paper presents a novel architecture and related technological aspects of the Next Generation Central Offices (NGCOs), as envisioned within the OCTAPUS research project, towards the flexible support of the emerging capacity and latency demanding services. First, the main functional blocks of the OCTAPUS NGCO Data and Control planes are presented, with emphasis on the dynamicity of the optical paths within the NGCO and its reconfiguration capabilities. Then, focusing on the Control plane, we describe in detail its hierarchical structure, the role of the associated entities and the interactions among them. An example of network provisioning to support multiple concurrent next-generation services of realistic use cases in a real urban area is used to stress the need for NGCOs to adapt to different service requirements and temporal fluctuations in traffic demands, while catering for time-sensitive services with stringent delay requirements.

The tight channel filtering imposed by long cascades of reconfigurable optical add-drop multiplexers (ROADMs) represents, nowadays, one of the major performance-limiting aspects for optically-routed coherent optical fiber systems. This makes it vital to perform a ROADM-aware optimization of the network performance at the physical layer. In particular, the use of different modulation options is known to have a strong impact on the extent of the filtering-induced penalties. In that regard, a long debate between single- and multi-carrier modulation has been taking place during the last few years, sometimes leading to apparently contradictory results. Following the open scientific discussion on this topic, in this work, we investigate by simulation and experimentally the wavelength selective switch (WSS) filtering tolerance of single-carrier (SC) and digital subcarrier multiplexing (DSCM) signals. In order to promote a fair comparison, both modulation options are carefully designed to minimize the ROADM-filtering penalties, namely resorting to the use of entropy loading together with baud rate optimization. After some preliminary numerical assessment, a comprehensive set of experiments are carried out for the transmission of 21-WDM 95–105 Gbaud SC and DSCM signals over a 2040 km straight line of fiber with regularly spaced WSSs. In general, our results allow to conclude that the two modulation options yield similar performance if the overall baud rate is optimized for each filtering scenario, keeping the baud rate at $\pm$ 5% of the optimized value.

The design of photonic/opto-electronic structures is a crucial field of research due to the wide range of applications in which these devices are utilized, such as communications, sensing and imaging. The traditional approach to design, which involves numerical simulation and iterative adjustments, is both time-consuming and computationally intensive. The ability to design-emulate photonic structures quickly and efficiently could bring new technologies to market at a faster pace and can also facilitate the process of extracting critical performance metrics from complex fabricated devices. The Vertical Cavity Surface Emitting Laser (VCSEL) and its use as a transmitter broadly used in data center interconnects is an example of a rather complex nonlinear system also exhibiting bandwidth limitation mechanisms attributed to its frequency response and parasitic effects, and thus its numerical simulation based on conventional rate equations can only partially approximate its behavior. When it comes to modelling fabricated devices, such physical models are notoriously difficult to parameterize in order to fit to observed laboratory data; in contrast, the emulation of the experimentally recorded behavior with the use of nonlinear models based on neural networks provides a much more efficient path to modelling, and with potentially higher prediction accuracy [1]. In this work, two types of bidirectional recurrent neural networks (RNN), namely Long Short-Term Memory (bi-LSTM) and Vanilla (bi-VRNN) were used to simulate the dynamic behavior of an experimentally characterized VCSEL-based, PAM-4 transmitter, exhibiting a high prediction accuracy approaching 100%.

We experimentally demonstrate that symbol-rate optimization (SRO) provides nonlinear gains in multicarrier systems, even with PCS modulation and realistic DSP. Optimized carrier phase recovery is crucial to achieving 0.2 dB gain for 1400 km 800G transmission, out of the ∼0.7 dB theoretical maximum gain we measured.

We experimentally assess the use of super-symbol (SUP) transmission with different distribution matching methods in a 100 GBd PCS-256QAM digital subcarrier multiplexing system. We achieve 0.1 dB SNR improvement after 900 km, a gain which comes almost for free due to the low complexity of SUP.

We demonstrate that, in an actual network, maximizing the transmission spectral efficiency requires not only to take into account transceiver impairments, but also to consider WSS filtering and flex-grid specifications. In a WDM transmission experiment over a 10-span link covering $\sim 38$ nm, we study achievable rate, bitrate and spectral efficiency for different symbol rates of 62.5, 75, 87.5, 100 and 112.5Gbaud with corresponding channel spacings of 75, 87.5, 100, 112.5, 125GHz, therefore accounting for few GHz bandguard required for potential WSS filtering effect, as well as being compatible with flex-grid configuration. While maximum achievable information rate per transceiver (wavelength) is achieved for the lowest symbol rate, and operating at larger symbol rate maximizes the bitrate per wavelength, to maximize overall fiber spectral efficiency (and therefore, capacity), operation at intermediate symbol rates, from 75 to 100Gbaud in our study, is required as a trade-off between the penalties from transceiver impairments and WSS filtering.

We experimentally compare the WSS filtering tolerance of single-carrier (SC) and digital sub-carrier multiplexing (DSCM) at 95–105 Gbaud. Whereas DSCM tends to be advantageous when using excessive baudrates, the two modulation options yield similar performance if the baudrate is optimized.

Digital resolution enhancement is experimentally demonstrated for 800G 107 Gbaud 8-carrier PCS-256QAM employing waterfilling. 1.4 dB SNR and 1 dB Q2-factor gains are obtained with a 4-bit DAC at 1.125 samples/symbol. © 2022 The Author(s)

In a subcarrier multiplexing 125 Gbaud DP-PCS-64QAM system with a net bitrate of 800 Gbit/s, we numerically and experimentally study the performance of transmission-aware carrier phase estimation (CPE) for long-haul systems. The recently proposed transmission-aware dual-reference subcarrier CPE (DRS-CPE) is assessed against a more conventional per-subcarrier CPE approach and a transmission-agnostic joint CPE in 8- and 16-subcarrier transmission. At laser linewidths of $\sim$ 150 kHz, we achieve an experimental signal-to-noise ratio gain over a single-carrier of up to 0.5 dB after 1800 km transmission, enabled by the DRS-CPE.

Transmission of C-band 40×800 Gbit/s signals using 125 Gbaud PCS-64QAM is demonstrated by performing power allocation optimization at the input of 3 OMSs along a 15-span link (>20 dB loss each) with EDFA-only amplification.

We investigate the benefit of entropy-loading for digital multicarrier systems operating under colored noise conditions induced by transceiver impairments. We demonstrate up to 0.35 dB Q 2 -factor gain for 1 Tb/s capacity with 8-subcarrier 125 GBd PCS-256-QAM.

Theoretical gains of digital multi-carrier systems are hindered by the use of sub-optimal conventional phase recovery, especially after fiber transmission. We experimentally validate an advanced, dispersion-aware algorithm that addresses this issue, achieving SNR gains up to ∼0.5 dB with 800G 125 Gbaud 16-carrier PCS-64QAM, transmitted over 1800 km.

We experimentally compare PCS-16QAM and QPSK for 400G transmission at 128 Gbaud. A realistic, full system implementation that accounts for penalties from the FEC, distribution matcher, transceiver impairments, fiber nonlinearity, and DSP, reveals that the theoretical 0.8 dB gain of PCS- 16QAM is reduced to only 0.1 dB.

We present and validate a statistical method able to separate nonlinear interference noise (NLIN) into a residual Gaussian (ResN) and a phase noise (NLPN) component. We take into account the interaction of the NLIN with the receiver’s DSP, mainly through carrier phase recovery (CPR), by considering the amount of correlation of the NLPN component. This allows obtaining in a straightforward way an accurate prediction of the achievable post-DSP transmission performance. We apply our method on simulated data in different scenarios. For this purpose: (i) several different quadrature amplitude modulation (QAM) and probabilistically shaped (PS) formats are investigated and (ii) simulations with standard single mode fiber (SSMF) and dispersion shifted fiber (DSF) are performed. In all these cases we validate the results provided by our method through comparison with ideal data-aided CPR and a more practical blind phase search (BPS) algorithm. The results obtained are finally compared with the predictions of existing theoretical models and the differences with our approach are pointed out.

Capacity demand for network connections within and between datacenters is increasing relentlessly, fueling the need for deployment of new and improved optical communications equipment. Confronted with the task of developing innovative solutions to address this challenge, engineers must deal with and consolidate countless design choices that are influenced by a large variety of constraints. To name just a few, an optimum solution may depend on technological requirements such as minimum data rate, maximum latency, electrical and optical bandwidth, link distance, upgradability (to higher speeds and/or other/more wavelengths), as well as the need to comply with standards and how these evolve. In this sense, automated design tools for simulating and comparing alternative solutions are indispensable. We present design examples at the system- and component-levels, illustrating the challenges in modeling, analyzing and optimizing technology choices and equipment parameters of optical interconnects for intra- and inter-datacenter applications. Of critical importance for tuning the performance of transceiver components is the integrated co-design of the corresponding electronic and optical parts. We demonstrate a seamless design flow linking simulations of the electronic circuits at the transmitter/receiver (such as serialization/deserialization, DAC/ADC, driver amplifier/TIA, etc.) with simulations of the optical fiber link, enabling investigation and optimization of the overall system performance. Further, we compare advantages and challenges of multimode infrastructure solutions utilizing, for instance, PAM4 modulation of multimode VCSELs with transmission over wide-bandwidth multimode fibers, and single-mode solutions employing Mach-Zehnder modulators with tunable DFB lasers in WDM operation over SMF-links.

We experimentally assess a series of four multi-dimensional modulation formats for 100 Gbaud single -span unrepeatered multi-rate applications. We show the variation of maximum acceptable span loss for entropies between 1 and 2 b/S/pol, with a granularity of 0.25 b/S/pol.