Programmable Reconfigurable Packet-Optical 6G Front-/Mid-Haul Infrastructure

Programmable Reconfigurable Packet-Optical 6G Front-/Mid-Haul Infrastructure C. Christofidis Electrical and Computer Engineering University of Patras Patras, Greece [email protected] K. Moschopoulos Electrical and Computer Engineering University of Patras Patras, Greece [email protected] V. Tsourtis Electrical and Computer Engineering University of Patras Patras, Greece [email protected] D. Uzunidis Electrical and Computer Engineering University of Patras Patras, Greece [email protected] D. Marom Applied Physics Department Hebrew University of Jerusalem Jerusalem, Israel [email protected] M. Nazarathy Electrical and Computer Engineering Technion, Israel Institute of Technology Haifa, Israel [email protected] R. Munoz CTTC/CERCA Centre Tecnologic de Telecomunicacions de Catalunya Castelldefels, Spain [email protected] I. Tomkos Electrical and Computer Engineering University of Patras Patras, Greece [email protected] Abstract—The 6G networks need to rely on a dynamic, flexible, scalable, high-bandwidth, and low-latency packet-optical front-/mid-haul infrastructure that can dynamically manage the Radio Access Network (RAN) functional split options. Its main aim is to offer the desired services with the best possible performance at the lowest power consumption and cost. To implement the flexible functional splitting (FFS), that maximizes the efficiency of the RAN by dynamically selecting the optimal split between Central Units (CUs) and Distributed Units (DUs) for each cell and each user, the underlying 6G front-/mid-haul network segments should posses the ability of reconfigurability under dynamically customizable operational conditions. Such an optical front-/mid-haul infrastructure needs to be based on novel optical processing/switching schemes that are ultra-low energy (at the order of pJ per bit), ultra-high capacity (>1 Tb/s), and fast-reconfigurable (sub-ms) software-programmable photonic subsystems (i.e., transceivers, multiplexers, and fast switches integrating tunable lasers/filters). At the same time, the existence of a novel intelligent control plane to optimize the utilization of network resources is mandated. In this invited contribution, we discuss the proposed innovations and their associated control as well as data plane solutions currently under development within the EC-funded project PROTEUS-6G, with a particular focus on novel transceiver and switching innovations. Index Terms—component, formatting, style, styling, insert. I. INTRODUCTION The 6G use cases categories bring forth new prerequisites, urging optical transport systems to evolve in order to meet This work was supported by the European Commission under the Horizon Europe project PROTEUS-6G (Grant Agreement No. 101139687). new demands, as outlined in IMT-2030) [1], . To realize the 6G vision, optical networks must address multiple challenges: i) achieving high-capacity transport with hundreds of Tb/s per link and beyond, ii) maintaining low latency at the sub-millisecond level, iii) enabling swift reconfigurability within microseconds, iv) boosting programmability and operational optimization of components such as transceivers and switches, v) decreasing energy consumption per bit, especially transceiver related, and vi) improving the end-to-end allocation of communication and computing resources using Artificial Intelligence (AI) [2]. This work shares our perspective on 6G network design by proposing novel components necessary to meet these challenges, within the PROTEUS-6G EU funded research initiative: i) The SDPtMP network: leverages a degree-four Space Division Multiplexing (SDM) feed and Wavelength Division Multiplexing (WDM) for scalable capacity. By allocating digital subcarriers per Remote Unit (RU) and Central Unit (CU), multiple RUs can share a wavelength without interference. This is enabled by a circular subcarrier mux/demux built from sharp optical interleavers. When SDM capacity is reached, Wavelength Selective Switches (WSS) assign new wavelengths to RUs. The architecture supports spatial/spectral RU-CU routing while eliminating lossy splitters or combiners. ii) Novel High Capacity and Ultra Energy Efficient transceivers: Innovative ultra-high-speed, low-latency, low-cost, and power-efficient Lite-Coherent (LITE-COH) transceivers (TXR) will be developed as key enablers for cell-free MIMO. These transceivers aim to support the fronthaul capacities required by the functional split option 8, enabling 6.4 Tb/s per fiber over 8 wavelengths (λ). The LITE-COH TXR will feature 0.8 Tb/s per λ and contribute to reduced latencies, cost savings, and power efficiency. This will be achieved through alloptical-signal-processing (AOSP) functionality, displacing bandwidth-limited and power-hungry digital signal processing (DSP) used in conventional coherent TXRs for IQ quadrature and orthogonal polarization processing. The power consumption will be reduced by approximately 50%. iii) Novel service management, orchestration, and control system for packet-optical x-haul networks: This system will support dynamic reconfiguration of functional splits across RUs, DUs, and CUs, and adapt the packet-optical x-haul network to meet transport demands accordingly. It will also define data models and protocols for full programmability and monitoring of key optical elements. To enhance energy efficiency, autonomous networking architectures and optimization algorithms will be developed for resource management, orchestration, and service provisioning. II. INNOVATIONS A. Spatially-Diverse Point-to-Multipoint (SDPtMP) Optical Fronthaul Distribution Network The SDPtMP node establishes bidirectional connectivity with cell sites via dedicated optical fibers, which is a standard setup in such deployments [3]. In contrast, fiber links connecting to the SDM/WDM ROADM gateway and the access CO operate in a unidirectional manner. The PROTEUS-6G fronthaul enables multi-level dynamic traffic provisioning. To each cell site subcarriers are assigned from four different access CO sources, as illustrated in Fig.1. Typically, the CO transceivers maintain spare subcarriers that can be electronically activated or deactivated on demand, ensuring adaptive capacity allocation [4]. The subcarrier allocation considers that if the subcarrier demand of a transceiver exceeds its available capacity, cell sites can be assigned to optical channels with higher availability of subcarriers. The access CO typically operates multiple transceivers within the WDM grid, dynamically allocating wavelengths to optimize resource utilization and reduce energy consumption [5]. To enable coherent detection and upstream transmission, cell site transceivers must adjust their designated wavelengths using a rapidly tunable laser. The WSS plays a key role in this process by either establishing a connection prior to the transmission, maintaining connectivity alongside an active transmission, or using a circular multiplexer/demultiplexer that supports multiple free spectral range (FSR) wavelengths between input and output ports. This architectural approach mirrors the rapid dynamic switching techniques employed in data centers based on wavelength selection, which PROTEUS-6G extends to fronthaul networks. Another dynamic provisioning scenario involves transitioning from DSCM multi-access to direct-link LITE-COH, enabling the efficient use of significant edge computing resources at the CO. (a) (b) (c) Fig. 1: (a) DSCM transceivers connected to the inputs of the optical filter network. The interlacer (RAMZI filter network) allocates digital subcarriers from all inputs to the outputs. (b) EVM degradation vs. baud rate relative to 3.125 GHz baseline [3]. (c) EVM under SDPtMP-induced crosstalk as low as –15 dB [4]. The circular subcarrier-group interleaver is structured using a network of precise wavelength interleavers, implemented with ring-assisted Mach-Zehnder Interferometers (RAMZIs) as shown in the inset (iii) of Fig. 1(a). Traditional PtMP networks utilize passive splitters. A significant drawback of this approach is the substantial inherent losses caused by the power splitters. A candidate alternative, which is the use of a WSS to allocate RU capacity at the wavelength or carrier level, fails to support subcarrier multi-access due to its resolution confined to the WDM channel level. Likewise, a circular arrayed waveguide grating router (AWGR), proposed for data center connectivity, cannot resolve below the WDM channel level (making it unsuitable for subcarrier access) and has a channel passband restricted by a Gaussian profile. An asymmetric MZI produces outputs that vary according to sinusoidal and cosine functions based on differential phase delay, but when adding one or two ring resonators within the interferometer arms greatly enhances interleaver performance. The single and dual RAMZI setups introduce poles in the transfer function, optimizing for a sharper interleaver response and achieving spectral utilization of 76% and 94%, respectively, with respect to the crossing point of 3dB loss [6] as shown the transfer function in the inset (ii) of Fig. 1(a) . In prior art, interleavers arranged in a binary tree structure with varying spectral periodicities have been employed for demultiplexing, PON seeding, and OFDM demultiplexing [7]. A four-way subcarrier-group demultiplexer utilizing identical interferometers helps mitigate double-filtering artifacts. Dual-ring RAMZI interleavers with a 64 GHz free spectral range define 30 GHz-wide channels accommodating seven 3.75 Gbaud subcarriers. In [3] we have estimated the EVM penalty across various baudrates as illutrated in Fig. 2(b). In a dual-stage mesh configuration with a quarter-cycle offset, four distinct groups of four subcarriers emerge, with one edge subcarrier sacrificed per group. This tree-like structure forms a circular subcarrier-group interlacer that uniformly distributes subcarriers from four inputs to four outputs, maintaining a complementary unitary behavior [8]. While the interleaver periodicity suggests a 64 GHz WDM grid, the circular routing pattern naturally aligns with a 16 GHz sub-band spacing. This enables efficient DSCM packing, with 4 GHz subcarriers and 16 GHz alignment, albeit at the cost of narrow guard bands. Such granularity exceeds the resolution of standard WSS hardware, which cannot reliably switch channels with 4 GHz guard bands. However, allocating a full 16 GHz subcarrier group as a guard band restores compatibility, allowing seamless operation with 80 GHz WDM spacing for 64 GHz DSCM channels [9]. The access CO’s SDPtMP element in the fronthaul is supplied by four parallel fibers that pass through an SDM/WDM ROADM node. This node offers comprehensive optical connectivity for all spatial and spectral channels and links with the mid-haul network segment (metro ring, capable of further supporting SDM, depicted as dashed lines). The study in [4] investigates the effect of crosstalk on performance by analyzing the EVM penalty as the crosstalk level increases in the transmission path, where the signal is filtered through the midhaul’s network nodes as shown by the arrows in Fig. 1(a) . B. Linear-Drive Plasmonic (LDO) based Lite-Coherent Optical Transceivers LDO-based TXRs eliminate the power-hungry DSP, reducing power consumption by half, as well as cutting down cost, footprint and latency. Beyond those major benefits, PROTEUS-6G aims to unlock additional advantages by developing two novel Photonic Integrated Circuit (PICs) featuring All-Optical-Signal Processing (AOSP), namely the Transmit/Receive Optical Sub-Assembly (TROSA) and Receive Optical Sub-Assembly (ROSA), augmented by an Analog Signal Processor (ASP) RFIC. (i): the proposed development (Dev) stage-I aims to double data rate per λ(relative to the extrapolated-SotA) by developing an advanced PIC-based AllOptical (AO) Dual-Polarization (DP) PAM4 Intensity Modulated Direct Detection (IMDD) LDO TXR (Fig. 2) carrying a pair of independent POL-channels per each λ(whereas the datacom industry is set to continue along its LDO TXR development path, aiming for a single IMDD-channel per each -λ). The beyond-SotA baud rate experimentally available at our disposal is 200GBd. At 200GBd per λ, the novel LDO TXR will multiplex a pair of optical PAM4 (oPAM4) signals each carrying 400Gb/s, conveying 800Gb/s per λ. Upon aggregating 2,4,8 of our λby means of Coarse WDM (CWDM), the total rates per fibre would be 1.6Tb/s, 3.2Tb/s, 6.4Tb/s, respectively, full-duplex, suitable for the elevated 6G data transport of option-8 functional splits, high-end massive network MIMO, and cell-free configurations. Current 120GBd PAM4 systems carrying 200Gb/s-per λwill be deployed, to be future-extended to 200GBd per-PAM4. Thus, PROTEUS6G TXR target of 800Gb/s-per-λleaps ahead by a factor-offour and promises to maintain a future-proof factor-of-two lead in bitrate per λ, even as industry baudrates would eventually catch up and evolve beyond 200GBd (as the candidate dualPOL-(de)mux is agnostic to baud rate and will always double the rate relative to using a single polarization, at any future baud rate). Fig. 2: Novel DP-PAM4-IM LDO transceiver (TXR) with phase-diverse LITE-coherent detection. This design doubles the data rate while enhancing the loss-budget margin by more than 10 dB through All-Optical Signal Processing (AOSP). Additionally, it provides frequency selectivity, enabling all-optical switching within the 6G front-haul photonic infrastructure. Abbreviations: TROSA—transceiver optical sub-assembly, AO-EQZ—analog optical equalizer, AOPOLtrack—analog optical polarization tracker. Considering the TXR PICs structure (Fig.2), the innovative AOSP ROSA PIC, implemented on the low-loss Silicon Nitride platform, all-optically mitigates Chromatic Dispersion (CD), Differential Group Delay (DGD) and to POL-demux (decouple) the two X and Y signals that got linearly crosscoupled upon propagating along the fibre link. This PIC comprises two modules in series, following spatial separation of the X,Y incoming POLs by a Polarization-Splitter Rotator (PCR): the analog optical Equalizer (EQZ) and the analog optical POL tracker. The latter module concatenates alternating 2:2 directional-couplers and push-pull (differential) slow Phase Modulators (PM). The POLtracker acts as a tunable optical linear 2×2 MIMO EQZ implementing an arbitrary transfer matrix acting on the received (X,Y) POL vector, capable of compensating endless runs of the State-of-Polarization [10], [11]. The chromatic dispersion equalizer (CD-EQZ) module features a lattice filter, which is essentially a sequence of single-tap cells. Each cell contains asymmetrical waveguides of varying lengths in combination with tunable directional couplers. Additionally, rings will be integrated with the MachZehnder Interferometer (MZI) structure or arranged in a cascaded manner. C. Service management, orchestration and control for adaptive management of fronthaul and midhaul transport (packet/optical) connections The conventional packet-optical Software Defined Networking (SDN) orchestration model, where packet and optical controllers operate in silos with limited coordination, is inadequate for managing complex and disaggregated networks like the PROTEUS-6G x-haul. In this model, the multi-layer SDN orchestrator merely delegates tasks to separate controllers—the optical SDN controller manages transponders, and the packet controller configures switching nodes—without unified oversight. To support dynamic, integrated x-haul services, a more cohesive architecture is required, wherein the orchestrator actively supervises both layers. This tighter integration is especially crucial for managing advanced optical nodes and transceivers, including pluggable Sub-Carrier Multiplexing (SCM) modules embedded within IP routers [12]. The proposed PROTEUS-6G SDN orchestration framework introduces a hierarchical model, comprising packet and optical layer controllers under a unified orchestrator. The orchestrator governs routing policies and resource coordination across the fronthaul and midhaul, ensuring end-to-end connectivity between RU, DU, and CU. The packet controller handles routing, QoS enforcement, and security policies across cell-site, metro, and regional routers, executing the orchestrator’s directives. Simultaneously, the optical controller manages optical path provisioning through components such as SDM/WDM ROADMs, subcarrier mux/demux devices, WSSs, and various coherent transceivers. SCM transceivers, integrated in IP routers, are monitored and configured via standard APIs (e.g., NETCONF, REST), while telemetry protocols like gRPC or OpenConfig provide real-time performance metrics—e.g., power levels, BER, and SNR—enabling fine-grained performance tuning. III. CONCLUSION By combining novel transceiver architectures, dynamic subcarrier multiplexing, and intelligent network control, PROTEUS-6G paves the ground for a scalable, energyefficient, and ultra-high-capacity transport system. This architecture not only meets the demands of cell-free massive MIMO and dynamic functional splits but also ensures that network resources can be utilized optimally under dynamically customizable operational conditions. Future research will focus on refining photonic integration strategies, enhancing DSPless architectures, and further optimizing AI-driven network automation. ACKNOWLEDGMENT We acknowledge the support of the European Community’s Horizon Europe Program (2021-27), Smart Networks and Services Joint Undertaking (PROTEUS-6G), (Grant agreement ID:101139134) REFERENCES [1] International Telecommunication Union (ITU), “Framework and overall objectives of the future development,” ITU-T Recommendation, Geneva, Switzerland, 2024. Available: https://www.itu.int/ [2] I. Tomkos, D. Uzunidis, C. Christofidis, K. Moschopoulos, Ch. Papapavlou, K. Trantzas, D. M. Marom, and R. Munoz, “Challenges and Innovations of Transport Networks to Support 6G Use-Cases,” in Proc. 2024 24th Int. Conf. on Transparent Optical Networks (ICTON), Jul. 2024, pp. 1–4, doi: https://doi.org/10.1109/ICTON62926.2024.10647904. [3] C. Christofidis, G. Gorgias, H. Georgopoulos, K. Moschopoulos, D. M. Marom, and I. Tomkos, “Spatially-Diverse Point-to-MultiPoint Optical Distribution Network for Enhanced 6G Fronthaul,” in Proc. 2023 Int. Conf. on Photonics in Switching and Computing (PSC), pp. 1–3, IEEE, 2023. [4] C. Christofidis, G. Gorgias, H. Georgopoulos, A. Napoli, D. M. Marom, and I. Tomkos, “Feasibility Study of Nyquist-Switching Node to Enable Mesh RU Fronthaul Interconnections and Flexible 6G Mobile Network Operation,” in Proc. 2024 International Conference on Optical Network Design and Modeling (ONDM), pp. 1–3, IEEE, 2024, doi: https://doi.org/10.23919/ONDM61578.2024.10582596 [5] D. Uzunidis, et al., ”Quantifying the Operational Benefits of Deep Learning based Dynamic Traffic Prediction using Real-World Dataset,” in Optical Fiber Communications Conference and Exhibition (OFC), 2025. [6] Y. Xie, Z. Geng, L. Zhuang, et al., “Programmable optical processor chips: toward photonic RF filters with DSP-level flexibility and MHzband selectivity,” Nanophotonics, vol. 7, no. 2, pp. 421–454, 2017. doi: https://doi.org/10.1515/nanoph-2017-006410.1515/nanoph-2017-0064. [7] M. Bakaul, M. Masuduzzaman, et al., “Simultaneous multiplexing and demultiplexing of wavelength-interleaved channels in DWDM millimeter-wave fiber-radio networks,” Journal of Lightwave Technology, vol. 24, no. 9, p. 3341, 2006. doi: https://doi.org/10.1109/JLT.2006.87594910.1109/JLT.2006.875949. [8] D. M. Marom, C. G. H. Roeloffzen, R. Botter, C. Christofidis, and I. Tomkos, “Fine-Resolution, Four-Port Optical Interlacer for SubcarrierLevel Optical Front-Haul Networking,” in Proc. 2024 IEEE Photonics Society Summer Topicals Meeting Series (SUM), pp. 1–2, IEEE, 2024. [9] A. Otero-Casado, P. Torres-Ferrera, C. Christofidis, G. Parisi, D. M. Marom, I. Tomkos, P. Monti, and A. Napoli, “Impact of SubcarrierRouting Devices on the Performance of DSCM Coherent Systems Evaluated with Analytical Models,” in Proc. of the 29th Int. Conf. on Optical Network Design and Modeling (ONDM), Toledo, Spain, 2025, (accepted for publication). [10] Z. Lin, Y. Lin, H. Li, et al., “High-performance polarization management devices based on thin-film lithium niobate,” Light: Science & Applications, vol. 11, p. 93, 2022. doi: https://doi.org/10.1038/s41377-022-00749x10.1038/s41377-022-00749-x. [11] A. Nespola, et al., “Proof of Concept of PolarizationMultiplexed PAM Using a Compact Si-Ph Device,” Photonics Technology Letters, vol. 31, pp. 62–65, 2019. doi: https://doi.org/10.1109/LPT.2018.288169910.1109/LPT.2018.2881699. [12] R. Vilalta, R. Mu˜ noz, et al., ”Teraflow: Secured autonomic traffic management for a tera of SDN flows,” in *2021 Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G Summit)*, June 2021.