Due to the rapid development of techniques such as wavelength-division multiplexing (WDM), advanced modulation formats, and coherent detection with digital signal processing (DSP), per-fiber capacity growth of the single-mode fiber (SMF) communication systems has closely tracked the exponential growth of traffic demand over the past few decades [1]. However, this growth trend is becoming unsustainable as the SMF capacity approaches its limits imposed by the fiber non-linear effects [2], necessitating the development of novel technologies to keep up with traffic demand. In this context, space-division multiplexing (SDM) techniques [3–15], which explore degrees of freedom in the transverse spatial domain of optical fibers to enhance the per-fiber capacity and spectral efficiency (SE) of communication systems, have gained significant interest in recent years.

Several approaches have been implemented to increase the spatial channel counts and thus the capacity/SE per fiber in the SDM systems, including mode-division multiplexing (MDM) schemes utilizing few-mode fibers (FMFs) [3–6], core-multiplexing schemes employing multi-core fibers (MCFs) [7,8], and the combination of these two approaches based on few-mode multi-core fibers (FM-MCFs) [9–13,15]. Compared with the former two schemes, the FM-MCF SDM systems can potentially provide the highest number of multiplexed spatial channels per fiber for a given fiber cladding diameter. This diameter should be restricted within a certain range (e.g., no more than 230 μm) to ensure mechanical stability and longevity in practical deployments [16]. Furthermore, by carefully managing crosstalk (XT) among fiber cores and among non-degenerate mode groups (MGs) within each core, only small-scale, low-complexity multiple-input multiple-output (MIMO) processing is required to handle intra-core modal coupling, or even intra-MG coupling [6,9]. This demonstrates the excellent per-fiber channel scalability of FM-MCF-based SDM systems, achieved with relatively low MIMO complexity.

Recent laboratory SDM demonstrations using FM-MCFs have made significant progress, substantially increasing both per-fiber capacity and SE compared to SMF transmission systems, while managing MIMO complexity [9,11,12,17–20]. However, to the best of our knowledge, no SDM transmission over field-deployed FM-MCFs in realistic outdoor environments has been reported. This contrasts with SDM demonstrations using single-mode multi-core fibers (SM-MCFs) installed in practical fiber cable ducts, which show low inter-core XT and thus low MIMO complexity, but relatively lower SE [21,22], as shown in Fig. 1. Unlike the relatively stable laboratory environment, field-deployed optical cables are susceptible to external environmental interference, and the cabling process can introduce additional stress to the fibers. It remains unproven whether these factors would significantly worsen inter-core XT and particularly intra-core modal XT in the FM-MCFs, thereby affecting the capacity/SE and MIMO complexity of SDM transmissions.

In this paper, we present the successful demonstration of SDM transmission over a 5-km field-deployed seven ring-core fiber (7-RCF) with a cladding diameter of 178 μm. The measurements of attenuations, mode coupling, and transmitted power stability of the field-deployed 7-RCF are carried out, which show no significant differences compared to the 7-RCF before cabling, thereby confirming the fiber’s adaptability to cable deployment and enhanced resilience to environmental disturbances. Using the field-deployed 7-RCF, 84 orbital angular momentum (OAM) mode channels (7cores×6modes×2polarizations) are multiplexed in the SDM demonstration, each carrying 40 wavelengths. By implementing bidirectional transmission in the fiber, the number of channels per-fiber can be effectively doubled. Utilizing 12-Gbaud 8-quadrature amplitude modulation (8QAM) per data channel, a raw (net) SE of 2×241.92(2×201.6) bit/(s Hz) and a capacity of 241.92 (201.6) Tbit/s have been realized. To the best of our knowledge, this sets a record for the highest SE achieved in SDM demonstrations utilizing field-deployed fiber cables and is approximately 10× that of any reported field-deployed optical cable systems. Furthermore, these results are achieved using modular 4×4 MIMO processing with a time-domain equalization (TDE) tap number not exceeding 15, enabled by the simultaneous weak coupling among the fiber cores and non-degenerate MGs within each core, along with the high-degeneracy of the intra-MG modes in the field-deployed 7-RCF. The complexity per unit capacity of this MIMO equalization is comparable to that used in SDM demonstrations with weakly coupled SM-MCFs, as illustrated in Fig. 1.

The fabricated fiber cable was deployed in an outdoor environment, and the route is illustrated in Fig. 2(c). The installation commenced at our laboratory, traversing ducts within the State Key Laboratory of Optoelectronic Materials and Technologies (OEMT) building. Subsequently, the cable enters into the underground cable ducts on the Sun Yat-sen University (SYSU) campus. Following the path of the underground ducts, depicted by the blue line in Fig. 2(c), the cable formed a loop back to the OEMT building, ultimately reaching back to the laboratory. The total length of the installed fiber cable extended approximately 5 km, with ∼4.5km within the campus duct and ∼0.5km within the duct of the OEMT building. This configuration allows experimental setups at both the transmitting and receiving ends to be established within the same laboratory.

The design of the 7-RCF aims to achieve low loss and low XT transmission with good mechanical stability [9], while also considering the stability of the fiber’s performance after cabling. As shown in Fig. 3, a proper refractive index profile (RIP) of the ring core allows the 7-RCF to maintain differential refractive index (Δneff) between high-order MGs greater than 1.5×10−3 and nearly identical neff of intra-MG modes, in order to achieve weak coupling between high-order MGs with topological charge |l|>0 and high degeneracy intra-MG modes [6]. In addition, the trench structure outside the ring core reduces the impact of macro-bending and micro-bending deformations caused by environmental disturbances and cable deployment on inter-core XT and fiber attenuation [23,24]. With a cladding diameter of 178 μm, the core pitch of the 7-RCF is set to be 50 μm, in order to suppress the inter-core XT of high-order modes. Meanwhile, the outer cladding thickness is appropriately set at around 40 μm to mitigate micro-bending and excess loss for the outer cores [6]. The 7-RCF within the fiber cable supports a total of 35 OAM MGs, encompassing seven cores with five OAM MGs each. Notably, each MG comprises degenerate OAM modes with topological charge values +l and −l (l=1, 2, 3, and 4) alongside the OAM MG with a topological charge l=0. Each of these modes carries two orthogonal polarizations.

After the 7-RCF was fabricated and cabled, its performance was measured to ensure it meets the required specifications. The impact of cabling on fiber performance is assessed by measuring characteristics such as mode attenuations, inter-MG XT, inter-core XT within each fiber core, differential group delay (DGD), and mode partition noise, which are induced by strong intra-group modal coupling in a 7-RCF within a field-deployed cable. These measured results are then compared with those obtained from the fiber spool prior to cabling.

The attenuation of all the MGs at the wavelength of 1550 nm is measured using optical time-domain reflectometry (OTDR) [18]. As depicted in Fig. 4, the attenuation for the OAM MGs of the 7-RCF within the installed cable averages 0.32 dB/km (see red triangles in Fig. 4). This is ∼0.02dB/km higher than that of the 7-RCF spool (as indicated by the blue circles in Fig. 4). This slight increase can be ascribed to the minor micro-bending and additional stress incurred during the cabling process. Notably, there is a pronounced attenuation difference for the highest-order OAM MG with a topological charge of |l|=4 in fiber core #2, observed in both the fiber spool before cabling and the installed cable. This discrepancy could stem from deformation in the structure of this particular core during the drawing process.

The inter-MG XT within an individual fiber core, as well as the inter-core XT of the 7-RCF, has been experimentally characterized for both the installed cable and fiber spool before cabling based on power measurements [18]. For this measurement, all modes of all cores are excited simultaneously at the transmitting end. In the receiving end the transmitted mode is bifurcated into two branches. Each branch subsequently passed through a commercial vortex phase-plate (VPP) with opposite topological charges, denoted as l and −l. Notably, only the OAM MG matching the topological charge |l| could be successfully demultiplexed, converted into Gaussian beams, and coupled into two SMFs. Therefore, the crosstalk power between each mode pair can be measured separately. In Fig. 5(a), the sum of XT represents the total crosstalk from OAM MGs that have different topological charges with the received VPP topological charge. In Fig. 5(b), the sum of XT represents the total crosstalk from cores with different core indices with the received core index. As illustrated in Fig. 5, the aggregated XT among OAM MGs with |l|=2, |l|=3, and |l|=4 within the same 7-RCF fiber core of the installed cable is below −12dB at a wavelength of 1550 nm, while this value is around −20dB among different fiber cores. In this case, MGs |l|=0 and |l|=1 are not used as spatial channels in this work due to their strong inter-MG XT. Therefore, the ring-core design can be further refined to reduce XT between these two MGs, ensuring that all MGs exhibit low inter-MG XT and can be utilized effectively [25]. In addition, as illustrated in Fig. 5, the measured data show no significant variation in either inter-MG or inter-core XT for the field-deployed fiber cable relative to the laboratory fiber spool before cabling. Even with optical fibers of varying lengths, the inter-MG and inter-core XT exhibit similar performance. This consistency arises because the system’s overall crosstalk, which includes both the cascaded MUX/DEMUX modules and the 7-RCF, is primarily influenced by the MUX/DEMUX modules, rather than the fiber itself. Previous measurements showed typical in-fiber XT of −30dB/km or less in the fiber spools [9,26]. Consequently, the accumulated crosstalk in the 7-RCF over moderate or short transmission distances is lower than that introduced by the MUX/DEMUX modules.

To better characterize the weak coupling between non-degenerate MGs, the DGDs of the OAM MGs are measured with a vector network analyzer (VNA) using a time-domain impulse response method [27]. As depicted in Fig. 6, large DGDs with values of more than 5 ns/km between adjacent pairs of OAM MGs of topological charge |l|>1 can be achieved in each fiber core. This observation indicates weak coupling among high-order OAM MGs within each fiber core [27]. This performance also closely aligns with that of the 7-RCF spool, implying that the effect of cabling and installation on the DGD is negligible.

The intra-MG degenerate modes, with closely spaced neff, are theoretically sensitive to environmental disturbances due to strong coupling. This sensitivity results in noticeable mode power fluctuations known as mode partition noise [28]. In this section, the mode partition noise of one mode in a selected MG of one fiber core in the field-deployed 7-RCF is experimentally evaluated. The results are then compared with those from fiber spools before cabling to assess their differential responses to environmental disturbances. In this experiment, a selected OAM mode with a topological charge of l=3 in the central core of the field-deployed 7-RCF was excited at the transmitter end. This mode, after transmission through the fiber cable, was transformed into a Gaussian beam in free space, and then coupled into an SMF connected to a power meter. Power measurements were taken every 5 ms over a period of 20 s, and the process was repeated multiple times to gather a larger statistical sample. The field-deployed 7-RCF was then replaced with a laboratory 7-RCF spool to replicate the same measurement procedure.

Figure 7(a) compares the random time-domain power fluctuations recorded over a 20-s period for both the field-deployed 7-RCF and the laboratory 7-RCF spool. Figures 7(b) and 7(c) show the averaged normalized power fluctuation spectra, derived from the time-domain data collected across multiple measurements. As the results illustrated in Fig. 7, the selected mode in the field-deployed cable exhibits more frequent power fluctuations (or mode partition noise) compared to the laboratory fiber spool. This is primarily because the field-deployed cable is installed in standard fiber conduits without special fixation, and some sections are suspended [see Fig. 2(e)]. These conditions make the cable more susceptible to environmental perturbations. In contrast, the environment surrounding the laboratory fiber spool is relatively stable. However, the fluctuations in the field-deployed 7-RCF are of smaller magnitude compared to those in the laboratory 7-RCF spool, likely due to the protective outer sheath of the fiber cable. Mitigating such mode partition noise can be achieved by simultaneously detecting all intra-MG degenerate modes and equalizing them through MIMO processing, which will be discussed in the following section. The more frequent variations in mode partition noise necessitate a faster adaptive speed for the MIMO algorithm.

In practical systems, the adaptive time for MIMO compensation depends on the specific deployment of processing hardware. It is generally assumed that the parameter update speed of the MIMO adaptive filter aligns with the clock frequency of the hardware [29]. For example, traditional FPGAs have a clock frequency around 100 MHz, while ASIC chips typically operate at approximately 1 GHz [29]. This ensures that the time for a single MIMO parameter update does not exceed the 10 ns range. Our MIMO algorithm converges after 30,000 parameter iterations (see Appendix C for details), maintaining the adaptive time at around 0.1 ms or even lower. This is shorter than the millisecond-level mode power fluctuations observed in the field-deployed fiber cable, as illustrated in Fig. 7.

Additionally, in our data transmission experiments, MIMO digital signal processing is performed offline (see the next section for details). Due to the limited storage depth of the oscilloscope, each processing session collects 100,000 data symbols, with each symbol having a period of 83 ps. This results in a total time of the order of microseconds, which is significantly shorter than the mode power fluctuation time. Therefore, it can be assumed that the channel parameters remain constant during a single data acquisition session. There is no issue with the MIMO update speed being unable to keep up with channel changes.

The experimental setup of the bidirectional OAM-WDM-SDM data transmission is depicted in Fig. 8. The setup comprises five main parts: a WDM signal transmitter with 8QAM modulation, an OAM space and mode multiplexing (MUX) module, the 5-km field-deployed 7-RCF, an OAM mode demultiplexing (DEMUX) module, and coherent optical receivers followed by DSP including the offline 4×4 MIMO equalization.

At the transmitter, to circumvent laboratory resource limitations, sliding test channels with high optical signal-to-noise ratio (OSNR) and dummy channels with low OSNR are multiplexed together to generate 40 WDM carriers. Five optical carriers with wavelengths in a 0.1 nm/12.5 GHz grid, one of the standard WDM channel spacings specified in ITU-T G.694.1, from external cavity lasers (ECLs) are employed as sliding test bands [30]. The light source of the dummy wavelength band is generated by a WDM carrier generator based on multiple seed light sources that undergo modulation through a cascaded phase modulator and a Mach–Zehnder modulator (MZM) [28]. Following this, the test and dummy bands are combined by a wavelength-division multiplexer. Subsequently, the WDM carriers are modulated by a 12-Gbaud 8QAM electrical signal sourced from an arbitrary waveform generator (AWG) via an in-phase/quadrature (I/Q) modulator. This modulation process generates a set of 40-channel WDM signals spanning from 1548.16 nm to 1552.16 nm, maintaining a grid of 0.1 nm/12.5 GHz. The sampling rate of the AWG is set to 120 GSa/s and the data sequence is the pseudo-random binary sequence with a pattern length of 218−1. The electrical signals are digitally pre-shaped by a Nyquist filter, specifically a raised cosine filter, with a roll-off factor of 0.01 to align with the 12.5 GHz WDM grid. It is noted that due to device resource constraints, all WDM channels are modulated by a single electrical signal. This configuration may result in limited decorrelation among the WDM channels, potentially leading to an overestimation of system performance due to inter-WDM-channel nonlinear effects [31]. However, the high nonlinear threshold of the 7-RCF would ensure that the inter-WDM-channel nonlinear effect has little impact on the system performance [9]. Due to limited equipment resources, only 40 wavelength channels were used in this experiment. Nonetheless, we do not anticipate a significant drop in performance beyond this spectral band. A total of 312 WDM channels spanning from 1538.19 nm to 1602.10 nm were evaluated in the high-capacity SDM systems using a 7-RCF with a similar design in Ref. [9], and remarkably consistent performance was sustained across all of the WDM channels.

After being amplified by a high-power erbium-doped fiber amplifier (EDFA), the generated WDM signals are divided into four branches using an optical power splitter. Three of these are amplified and further divided into six branches to implement mode channels with high OSNR in the fiber core under test, while the remaining branch is amplified by the high-power EDFAs and used to implement low OSNR channels in the dummy cores. The generated test and dummy mode/space channels are directed toward three different OAM spatial and modal MUX modules. Within each OAM space and mode MUX module, two hexagonally packed seven-core SMFs spliced with a fan-in device are included to convert the in-fiber fundamental modes to free-space Gaussian beams. Then two sets of seven-core Gaussian beams are collimated, linear-polarization filtered, and finally imaged onto spatial light modulators (SLMs). The SLMs are configured with phase masks (brown dashed box in Fig. 8 and detailed parameters can be found in Ref. [9]) to generate seven-core OAM beams with topological charge +l or −l. Subsequently, the six groups of seven-core OAM beams generated from the three MUX modules are power-combined and directed through a quarter-wave plate (QWP) to facilitate circular polarization conversion. After propagating through the polarization multiplexing module, which comprises a polarization beam splitter (PBS), an optical de-correlation path with a 4-f configuration, and a polarization beam combiner (PBC), four OAM modes ⟨+l,R⟩, ⟨+l,L⟩, ⟨−l,R⟩, and ⟨−l,L⟩ are generated within each MG, where L and R refer to the left- and right-handed circular polarization, respectively. Subsequently, the 84 space/mode channels (7cores×6OAMmodes×2orthogonal circular polarizations) each carrying 40 wavelengths are split into two branches equally via a beam splitter (BS), and they are coupled into the field-deployed 7-RCF from both ends, enabling bidirectional transmission. Here both forward and backward transmissions for each mode and wavelength channel share the same laser source, owing to the limited device resources, so the signal transmitted in both directions is the same. This could introduce coherent Rayleigh noise, manifested as the beating noise between the signal in one transmission direction and Rayleigh backscattering noise in the other direction [32–35]. However, this noise can be substantially mitigated through coherent optical detection (refer to Appendix B).

After going through the 5-km field-deployed 7-RCF bidirectional transmission, the intensity distribution of received OAM beams is illustrated in the inset figure (blue dashed box) of Fig. 8. Due to the coherent superposition of the four degenerate OAM modes within an MG, a linearly polarized (LP) mode-like intensity distribution [36] is observed in each fiber core. Subsequently, the OAM beams are collimated and split into two branches. After going through a VPP that matches the topological charge of the OAM mode being tested, each branch is converted into a Gaussian beam and coupled into an SMF-pigtailed dual-polarization integrated coherent receiver (ICR). Due to limitations in equipment resources at the receiver, only the four OAM beams within the same MG are simultaneously mode converted and then detected by the ICRs in only one direction. Subsequently, the eight electrical signals, generated by two ICRs, are sampled and stored using an eight-channel real-time oscilloscope operating at a sampling rate of 80 GSa/s for offline DSP, which includes timing phase recovery, 4×4 MIMO equalization based on the constant modulus algorithm, frequency offset estimation, and carrier phase estimation. The measurement is repeated until all the mode/space channels in both directions have been measured.

The power budget of the bidirectional OAM-SDM-WDM experimental system has been comprehensively evaluated, with the results presented in Table 2. The average power of the WDM signals at the input port of the fan-in device is standardized to 18–19 dBm, equivalent to approximately 2–3 dBm per WDM channel. This power level is subject to variations that depend on the OAM MG orders, primarily due to variations in insertion loss attributed to the seven-core OAM MUX module. The total insertion loss within the OAM MUX module includes optical element losses, such as those from the 3-dB beam combiner and the SLM, as well as coupling losses into the field-deployed 7-RCF. The observed coupling loss to the field-deployed 7-RCF for OAM MG |l|=4 is approximately 4 dB, which is 1 dB higher than that of OAM MG |l|=2 and 3. This MG-order dependent loss can be attributed to the radial mismatch between the generated free-space OAM beams and the OAM modes supported within the field-deployed 7-RCF. Additionally, alignment errors in the optical elements used for fiber mode coupling may have contributed to this discrepancy. Such challenges can be addressed through the implementation of precise local phase modulation for OAM modes with varying topological charges (|l|) in each fiber core and improved alignment of optical elements. Before being coupled into the field-deployed 7-RCF, the OAM beam is evenly divided into two branches and coupled from both ends of the field-deployed 7-RCF to achieve bidirectional transmission, which causes a power loss of 3 dB. After passing through the 5-km field-deployed 7-RCF and the OAM DEMUX module, ∼−24dBm of optical power is received at the input port of the optical pre-amplifier located before the ICR, higher than the sensitivity (∼−37dBm) of the pre-amplifier.Table 2.

Optical Power Budget Evaluation of the Bidirectional OAM-SDM-WDM Experiment System

The measured forward and reverse transmission bit-error-rate (BER) values of all 6720 channels (2direction×7cores×6OAMmodes×2polarizations×40 WDM channels) are shown in Figs. 9(a) and 9(b). For convenience, the BER evaluation for two orthogonal polarizations is performed together, resulting in the display of only two values within each MG. Notice that the BER performance of the OAM MG with topological charge |l|=3 is inferior compared to the other OAM MGs, which is primarily attributed to the relatively higher XT from the adjacent two OAM MGs (|l|=2 and |l|=4). However, the BER values of all bidirectional channels are below the 20% soft-decision FEC threshold of 2.4×10−2, only utilizing modular 4×4 MIMO equalization with a TDE tap number not exceeding 15, as shown in Figs. 9(c) and 9(d).

The SNR of the forward transmission signals is experimentally evaluated at varying levels of backward transmission signal power within the same fiber core, to assess the effect of Rayleigh backscattering (RB) noise in the bidirectional transmission system with coherent optical detection. As illustrated in Fig. 10(a), the SNRs of forward transmission signals decrease as backward transmission signal power increases. Figure 10(b) provides a comparison by illustrating the theoretically calculated signal-to-RB-noise power ratio after coherent detection at different levels of backward transmission power. It is noteworthy that the received signal and RB noise, post-coherent detection, are proportional to the photocurrents generated through the beating process between the optical signal or RB noise and the optical light from the local oscillator (LO) at the coherent optical receiver. As a result, the power ratio between the detected signal and RB noise is expressed as ⟨IS-lo⟩2/⟨IRB-lo⟩2 in Fig. 10(b). Further details regarding the theoretical RB analysis in the ring-core fiber together with coherent optical detection can be found in Appendix B.

An observation from Figs. 10(a) and 10(b) reveals that the experimentally evaluated SNR is lower than the theoretically predicted power ratio between received signals and RB noise. This discrepancy is primarily attributed to various factors beyond RB noise in the experimental system, such as Fresnel reflection of backward injected optical power at the fiber end facet, amplified spontaneous emission (ASE) noise of the EDFAs, inter-mode XT, electrical amplifier noise, quantization noise from the digital-to-analog converter (DAC) in the AWG and analog-to-digital converter (ADC) within the real-time oscilloscope.

Since the optical carriers of the forward and backward transmissions for each mode and wavelength channel share the same laser source due to equipment resource limitation (see Section 4.B), beating noise is anticipated to exist between RB noise from the backward transmission and signals transmitted in the forward direction, with its power theoretically being much higher than that of the RB noise itself [37]. However, coherent detection utilizing balanced detectors integrated into the ICR significantly mitigates such signal-RB beating noise, as explained in Appendix B. In the experimental system, the impact of Fresnel reflections of the backward transmission signals at the fiber end facet, typically characterized by a power considerably higher than that of the RB noise [38,39], is more pronounced on forward transmission signals at their receiving end. However, noise related to backward transmission, including both Fresnel reflection and RB noise, constitutes a small portion of total noise. Consequently, the experimentally evaluated SNR penalty due to the increasing backward transmission power, as depicted in Fig. 10(a), is smaller than that theoretically predicted in Fig. 10(b).

Figure 10(b) also illustrates that the signal-to-RB-noise power ratio is lower when forward and backward transmission channels share the same mode index (lF=lB, blue triangle line in Fig. 10) compared to cases with different mode indices (lF≠lB, orange square line in Fig. 10), given fixed forward and backward transmission powers. This is because the mode recapture factor, which represents the proportion of the total scattered power of the backward transmitting mode channel recaptured by the forward transmitting mode channel, is theoretically higher in the former case [40]. Detailed analysis can be found in Appendix B. Similar trends are observed in the experimentally evaluated results shown in Fig. 10(a). However, we attribute this phenomenon mainly to the further suppression of Fresnel reflection power by the OAM mode DEMUX module when the backward transmission mode index is different from the forward transmission one.

For mode-multiplexed transmission in both forward and backward directions, the theoretical analysis indicates a deterioration in the signal-to-RB-noise power ratio compared to single-mode transmission cases. This is due to the contribution of RB noise from two multiplexed backward transmission modes [e.g., lB=3 and 4 in Fig. 10(b)] to the target evaluated forward transmission mode [e.g., lF=3 in Fig. 10(b)]. In the experimental evaluation, the degradation in SNR performance in mode-multiplexed transmission results not only from increased backward-transmission-power-dependent noise but also from inter-mode XT.

Based on the above analysis, it is found that in short-distance bidirectional transmission systems, the effect of Fresnel reflection at the fiber end facet surpasses that of RB noise. To enhance the SNR of the received signal, one effective strategy involves mitigating Fresnel reflection, such as by applying an anti-reflection film or index-matching medium to the fiber end facet. Furthermore, in contrast to SMF-based bidirectional transmission systems, the received signal performance is affected not only by RB noise but also by inter-mode XT. In future investigations, we aim to conduct a comprehensive assessment of the collective impact of these two noise sources on the performance of transmission systems at varying distances. This evaluation will contribute to establishing the appropriate transmission distance range for bidirectional SDM transmissions with weakly coupled mode channels.

Bidirectional SDM transmission based on a 5-km field-deployed 7-RCF with a cladding diameter of 178 μm is experimentally demonstrated in this paper. The 7-RCF within the fiber cable supports a total of 35 OAM MGs, encompassing seven cores with five OAM MGs each. The measurements of attenuation, mode coupling, and transmitted power stability on the field-deployed 7-RCF were conducted, which showed no significant differences compared to the 7-RCF before cabling. These results confirm the fiber’s adaptability to cable deployment and its enhanced resilience to environmental disturbances. Implemented over the 7-RCF with 84 OAM mode channels each carrying 40 wavelengths, the SDM-WDM system achieves a raw (net) SE of 2×241.92(2×201.6) bit/(s Hz) and a capacity of 241.92 (201.6) Tbit/s by utilizing a bidirectional transmission scheme. This represents an approximately 10× increase compared to the SE of reported field-deployed optical fiber cable transmission systems, solely applying modular 4×4 MIMO equalization with a TDE tap number not exceeding 15.

Due to constraints in device resources, the experiment was not extended to test SDM transmission over the field-deployed fiber cables at longer distances using the recirculating loop scheme. Nonetheless, the 7-RCF employed in our demonstration exhibits the potential to enable longer-distance SDM transmission with high SE and capacity, while keeping low MIMO complexity, and consequently, low cost and low power consumption. Previously we successfully demonstrated high SE/high capacity SDM transmissions over 7-RCF spools covering distances of 14 km [17], 34 km [9], and 60 km [6], utilizing the same 4×4 MIMO equalization with low TDE tap numbers, thus ensuring low MIMO complexity. We also demonstrated a 100-km transmission supported by a single-core RCF, employing 4×4 MIMO processing with a low TDE tap number to address XT among the degenerate intra-MGs [41].

The good performance of the SDM transmission for up-scaling of SE and capacity per optical fiber while keeping low MIMO complexity within distances of tens of or even 100 km is realized by exploiting the uniquely excellent characteristics of OAM modes in RCFs. The radial confinement of the ring-shape core of the RCF makes the number of degenerate (or near degenerate) modes in the OAM MGs fixed at four when their topological charge |l|≥1 (OAM modes with topological charge value +l and −l each carrying two orthogonal polarizations [42], whose differential mode delays are very low due to their good degeneracy). The differential effective refractive index Δneff between adjacent MGs increases with |l| [36], which leads to reduced inter-MG coupling therefore good scalability to high-order mode space. Therefore, MIMO processing with a small 4×4 scale and a low TDE tap number and thus low complexity [43–47] can be sustained in the OAM-RCF-based systems when more SDM channels are involved to realize high SE/capacity. In our 7-RCF OAM MUX module [9], mode multiplexing currently relies on a straightforward power-combining strategy, leading to insertion losses that increase with the number of MGs utilized. While this method is sufficient for a small set of modes, it becomes less efficient as the number of mode channels grows, with power combining losses becoming more pronounced. Exploring advanced techniques such as multi-core spiral transformation [48] or multi-plane light conversion [49] could offer better power efficiency. Additionally, there is a clear advantage in developing MUX/DEMUX modules that are more integrated. However, such advancements require tackling the significant challenge of converting densely packed input Gaussian beams into a complex 2D input fiber array, a task that presents notable optical engineering challenges.

While the inter-MG XT of the 5-km 7-RCF in field-deployed cable remains lower than that of space and mode MUX/DEMUX modules, we anticipate a degree of performance degradation for longer field-deployed fiber cables compared to lab-deployed optical fiber spool before cabling. For instance, fiber fusion errors during the laying process may elevate the inter-MG XT of the RCF [50]. However, judicious control of the fiber fusion error range (e.g., no more than 0.4-μm radial misalignment for the RCF [51]) can effectively manage the increase in inter-MG XT, keeping it to a reasonably low level. By overcoming additional challenges in field-deployed fiber cables for practical application in the future, SDM schemes utilizing OAM modes in RCFs stand as a promising candidate for next-generation high SE/capacity transmission over optical fibers with distances of tens of kilometers (e.g., metro areas, inter-data center links), where weak coupling among the non-degenerate modes within each fiber core is achievable, and in-line optical amplification toward FM-MCFs is unnecessary.