1. Introduction

Space-division-multiplexing (SDM) in multi-core fiber (MCF) has recently been demonstrated [1–3] to have the potential to dramatically increase the achievable spectral efficiency (SE) as well as the capacity of a single fiber for sustaining the capacity growth [4] of future optical transport systems. MCFs with seven uncoupled cores in a single glass fiber were the key enabler in demonstrating capacities of 109 Tb/s and 112 Tb/s by combining SDM with polarization division multiplexing (PDM) and dense wavelength division multiplexing (DWDM). However, transmission distances were limited to a single pass through the MCF, with a reach of 16.8 km in Ref [1]. and 76.8 km in Ref [2]. It is important to extend the transmission distance by using multi-span transmission to explore the potential of MCF-based SDM for practical long-haul applications. Multi-core optical coupling technology is crucial for concatenating several MCF spans. Since multi-core optical amplifiers for use in transmission systems are not yet available, tapered multi-core couplers (TMCs) [5] with low loss and low-crosstalk open the opportunity to perform multi-span transmission experiments based on multiple single-core amplifiers. We recently leveraged advances in MCF development, i.e., the suppression of core-to-core crosstalk through improved fiber design, to demonstrate a record aggregate per-fiber spectral efficiency (SE) distance product [6]. Two new techniques were used to enable this demonstration. First, an apparatus that consists of seven separate re-circulating loops running synchronously allowed ten 32-Gbaud PDM quadrature phase shift keying (QPSK) channels on a 50-GHz grid to be simultaneously launched and transmitted through the seven fiber cores over multiple spans. Second, a novel core-to-core signal rotation scheme was implemented to simultaneously equalize the optical characteristics of all cores in terms of loss, dispersion, and in-band crosstalk, leading to improved overall system performance. In this paper, we expand on our previous work [6] to detail the novel concepts and the transmission performance.

2. Seven-core MCF and tapered multicore fiber connector characteristics

The MCF [3,5] is designed for operation in C and L bands for applications in high SE and high capacity optical networks, and consists of seven cores arranged in a hexagonal array with 9-μm core diameter and 46.8-μm core pitch (shown as a photograph in the inset of Fig. 1 .). The cladding diameter is 186.5 μm and the polymer coating diameter is 315-μm. The cutoff wavelength of each core is ~1.44 μm, and the mode-field diameter (MFD) at 1.55 μm is 9.6 μm. The dispersion and dispersion slope of each core in the MCF at 1.55 μm is about 16.5 ps/km-nm and 0.06 ps/km-nm², respectively. The measured attenuation for the center core is about 0.23 dB/km and 0.37 dB/km at 1.55 μm and 1.3 μm respectively. The measured attenuation for each of the six outer cores is about 0.26 dB/km and 0.40 dB/km at 1.55 μm and 1.3 μm respectively. The attenuation characteristics of all cores are low across the entire transmission window [3] and the values are similar to that of a single-core standard single-mode fiber.

 

Fig. 1 MCF with the TMC and photograph of cross-section of the MCF. Also shown is the span loss table. The center core loss was increased with a 3-dB attenuator to match the loss of the other cores.

Download Full Size | PDF

It is crucial for multi-span MCF transmission to have an efficient coupling scheme between each core of the MCF and external single-core components such as optical amplifiers. A fiber-based tapered multicore coupler (TMC) was fabricated to match the core spacing and modefield properties of the MCF, as illustrated in Fig. 1. The TMC was designed to minimize power coupling between cores, eliminating the coupler as an additional source of crosstalk. Two TMCs were used, one at the input end and the other at the output end of the MCF. The measured insertion loss of each core of the two TMCs ranges between 0.5 dB and 2.8 dB, and the crosstalk between cores is less than −45 dB [5]. The high insertion loss for some of the connections is attributed to small core misalignment, modefield mismatch and core asymmetry.

3. Core-to-core signal rotation concept

The transmission performance of SDM systems using MCF must be maximized to reap the benefits associated with such a multiplexing scheme. In particular, performance variations among the signals traveling through the multiple cores of a multi-span MCF system need to be minimized. (As an example, the signals propagating in the center core of a hexagonally symmetric MCF, as shown in Fig. 1, experience crosstalk from signals propagating in all six outer cores, which is 3-dB higher than the crosstalk from the three neighboring cores experienced by signals propagating in the outer cores.) A potentially viable solution is to increase the uniformity of the transmission characteristics of the cores in the MCF, such as loss, dispersion, and crosstalk, through better fiber design and manufacturing. However, this approach has its limitations and also makes fabrication issues challenging.

Here, we propose and demonstrate a novel concept, termed core-to-core signal rotation (CCR) that addresses the above concerns and maximizes transmission performance. The concept is depicted schematically in Fig. 2 . Every MCF span has single-core to multi-core connectors (TMCs in our demonstration) that couple WDM channels in and out of the MCF. The WDM channels launched into cores (C₁, C₂, …, C_M) of a MCF span are routed to a spatially different core in the next MCF span and CCR continues along the MCF transmission link. The locations where CCR is applied can be optical add/drop multiplexer (OADM) sites and/or optical amplifier sites. Within some limitations, CCR can also be used at splice points within a MCF span. In Fig. 2, the signal rotation is performed at the optical amplifier site, where the signals from cores C₁, C₂, …, C_M, after amplification by erbium doped fiber amplifiers (EDFA), are launched into cores C₂, C₃, …, C₁ of the next MCF span. This CCR approach not only reduces the worst-case crosstalk, but also has the effect of statistically averaging out variations of span loss and dispersion among the cores, and any other component imperfections along the transmission path of each core. This is particularly true in a single-span recirculating loop using the same span elements repeatedly in the loop. In a deployed straight-line MCF system, connecting individual MCF spans using CCR is expected to have a similarly beneficial effect.

 

Fig. 2 Core-to-core signal rotation concept

Download Full Size | PDF

4. 2688-km SDM-DWDM transmission experiment

Figure 3 shows the schematic of the experimental setup.

 

Fig. 3 Schematic of the experimental setup: (a) transmitter; (b) multiple recirculating loops with core-to-core rotation every round trip and (c) coherent receiver. Inset shows transmitter eye diagram.

Download Full Size | PDF

Ten C-band external cavity lasers (ECLs), each with a nominal linewidth of 100 kHz, were operated on a 50-GHz frequency grid from 193.200 THz to 193.650 THz (1548.15 nm to 1551.72 nm). The odd and even channels were separately combined using passive power combiners, and each group of five channels was separately modulated using integrated double-nested Mach-Zehnder modulators with 35-GHz 3-dB bandwidth and Vπ of 2.5 V. The in-phase (I) and quadrature (Q) branches of the modulators were driven by 32-Gb/s binary electrical signals from a pulse pattern generator with a pseudo-random bit sequence (PRBS) of length 2¹⁵-1, with several tens of bits decorrelation between the I- and Q- waveforms. Following modulation, polarization multiplexing was achieved by 3-dB splitting the WDM signals, delaying one copy by several hundred symbols, and recombining them in a polarization beam splitter (PBS) using manual polarization controllers (PCs). The odd and even channels were then combined with a 100-GHz to 50-GHz interleaver (IL).

For SDM transmission, the ten DWDM channels were launched into a specially configured 7-fold re-circulating loop apparatus that is running all the loops synchronously, as shown in Fig. 3(b). The seven re-circulating loops shared a common load switch that launched identical copies of the DWDM channels into each of the loops via a 1x8 power splitter. The signals at each of the seven loop inputs were first amplified by erbium-doped fiber amplifiers (EDFAs) before launching into a 76.8-km seven-core-fiber [2] through a TMC. The variations in the fiber lengths from the splitter to each core input (which includes EDFAs with length differences of several meters) were sufficient to provide multiple symbol decorrelation between cores at the point of launch. A second TMC was used to couple out the signals after transmission. The signals were amplified to compensate for the fiber loss, and channel powers were equalized using a wavelength blocker (WB) array module that incorporated eight independent 96-channel 50-GHz WBs. One WB was used per core of the MCF and the eighth WB was used at the receiver as described later. This way, a two-dimensional wavelength/space gain equalizing filter was constructed. After each WB, the DWDM signals from one core were sent to the re-circulating loop input of the next core, in a cyclic fashion. This ensured that the DWDM signals traversed each of the seven cores once every seven round trips. The table in the inset of Fig. 1 shows the measured loss of each span, where a span includes the input TMC, the core, and the output TMC [2]. The center span had about 3 dB less loss than the other spans due to variations in TMC losses and core losses [2]. A 3-dB attenuator was inserted at the input to the center span, so that all seven cores had a nominal span loss of about 23-24 dB.

A 1x7 optical switch (SW) was used to direct the received signal from the outputs of each of the seven re-circulating loops to the eighth WB, where the WDM channel to be detected was selected. This selected channel was then sent to a coherent receiver consisting of a polarization-diversity 90-degree hybrid, followed by four balanced detectors. The signal was combined with another ECL serving as the local oscillator (LO), tuned to within ± 100 MHz of the signal carrier. The four signal components (Ix, Qx, Iy, Qy) were digitized asynchronously using two synchronized 2-channel 80-GS/s real-time oscilloscopes. Digitized waveforms of 1-million samples each were processed offline in a computer to perform electronic dispersion compensation, polarization de-multiplexing, frequency/phase recovery, and bit error ratio (BER) measurement using efficient PDM-QPSK algorithms [7] without any crosstalk-compensating multiple-input multiple-output (MIMO) processing.

5. Measurement results

Figure 4(a) shows the measured back-to-back BER performance of the 128-Gb/s PDM-QPSK channel at 193.400 THz, while Fig. 4(b) shows the DWDM spectra before and after transmission. The required optical signal-to-noise ratio (OSNR), defined with a 0.1-nm noise bandwidth, is 15.5 dB at BER = 1 × 10⁻³, and 12.8 dB at BER = 1.5x10⁻², the threshold for 19.5% overhead hard-decision forward error correction (FEC) that is assumed in this investigation following [8], where a BER of 1.5x10⁻² (a Q² of 6.7 dB) was shown to be correctable to a BER better than 1x10⁻¹⁵. Compared to the theoretical performance (red curve in Fig. 4(a)), the implementation penalty was 1.6 dB at BER = 1 × 10⁻³. With the overhead assumed in this investigation, the net data rate of each wavelength channel is 107.3-Gb/s, yielding a net DWDM SE of 2.15-b/s/Hz in each core, and a net aggregate per-fiber SE of 15-b/s/Hz.

 

Fig. 4 (a) Measured back-to-back BER performance of the 128-Gb/s PDM-QPSK channel at 193.400 THz. Inset: typical recovered signal constellation at an OSNR of 40 dB. (b) Spectra of the 10x128-Gb/s DWDM channels before and after transmission.

Download Full Size | PDF

Figure 5(a) shows the received BER as a function of launch power per core at a distance of 2688 km (35 round trips) for the channel at 193.400 THz, measured at each of the seven loop outputs. At this distance, the variations in BER from core to core are uncorrelated and mostly determined by the uncertainty in equalizing the DWDM channel powers. The optimum launch power for this reach is around 0 ± 1 dBm, where the BER from all the cores is below the FEC threshold of 1.5x10⁻². It is instructive to mention here that the spread in the BER for a given power from the seven loop outputs is small, as expected for the CCR scheme implemented here. In the absence of such a “scrambling”, the spread is generally too large from signal power variations originating in the finite granularity of the WB attenuator setting used for channel equalization. In our experiment, we found the BER varied between 3x10⁻³ to 5x10⁻² at a given “nominal” launch power of 0 dBm.

 

Fig. 5 (a) BER as a function of launch power per core at a distance of 2688 km of the channel at 193.400 THz, measured at the output of each of the seven loops. Inset: typical recovered signal constellation at 2688 km from the output of loop 1. (b) BER performances of the channel at 193.400 THz as a function of distance, measured at the output of each of the sevens loops, with 0-dBm launch power into each core.

Download Full Size | PDF

In an independent experiment, a single span of 80-km single-core standard single mode fiber (SSMF) was incorporated in a recirculating loop with one EDFA at the launch, one EDFA following the span transmission, and one WB for channel equalization. The span loss of this setup was adjusted to match that of the center core of the MCF with attenuators placed both at the input (to emulate the input TMC loss) and at the output (to emulate the output TMC loss). The optimum launch power for transmission over 35 spans was found to be between −2 dBm and 0 dBm, limited by the accuracy of setting the channel powers. The received BER was between 5x10⁻³ and 2.0x10⁻² for the channel at 193.400 THz, similar to that measured with the MCF. These results confirm that SDM transmission performance with the MCF matches that of single-core SSMF spans, providing further evidence of the low crosstalk characteristics of the MCF as well as the benefits of CCR for averaging out the impairments.

In all subsequent experiments, the power of each channel was set to a nominal value of 0 dBm in each core. Figure 5(b) shows the measured BER as a function of distance traversed in the MCF for the channel at 193.400 THz. Measurements were made on this channel at all seven loop outputs. The spread in the BER from the different cores is low, demonstrating the benefit of using CCR at every span to equalize the performance of all the cores.

Finally, the BER performance of each of the ten DWDM channels is shown in Fig. 6 , at the transmission distance of 2688 km. Measurements were made at the output of each loop and WB8 was set to select each of the ten channels. The measurements indicate that the worst and best BER values are 1.3x10⁻² (Q² value of 7 dB) and 2.7x10⁻³ (Q² value of 9 dB), respectively, below the FEC threshold of 1.5x10⁻². The delivered OSNR at the nominal launch power of 0 dBm was between 14 and 16 dB among the ten channels at the seven loop outputs, indicating a transmission penalty of about 1dB.

 

Fig. 6 BER performance of all the ten DWDM channels after transmission over 2688 km (35 spans), measured at the outputs of the seven loops, with 0-dBm launch power into each core.

Download Full Size | PDF

6. Summary

We have successfully transmitted ten 50-GHz spaced 128-Gb/s PDM-QPSK channels over 35 spans of a low-crosstalk MCF, achieving a net aggregate per-fiber spectral efficiency of 15 b/s/Hz and a net aggregate per-fiber spectral efficiency times distance-product of 40,320 km⋅b/s/Hz. A novel core-to-core rotation scheme was implemented that allowed for statistical averaging of the transmission performance over multi-span MCF. The demonstration highlights the promise of using the newly available low-crosstalk MCF to enable transmission over long distances with mature and well-established coherent reception techniques.