1. Introduction

The use of cladding-pumped multicore amplifiers [1–7] is one of the most attractive technologies in space-division multiplexing (SDM). It offers potential energy and cost savings of optical fiber transmission systems by integrating multiple amplifiers into a single unit and using a single low-cost multimode pump laser diode (LD). A full in-line amplified multicore fiber (MCF) transmission line can be achieved by directly splicing erbium-doped multicore fibers and multicore integrated components to MCFs. Various types of cladding-pumped multicore amplifiers have been presented including a 7-core erbium-doped fiber amplifier (EDFA) [1,4], a 12-core erbium-ytterbium-doped fiber amplifier (EYDFA) [2], a 6-core EDFA [3], a 32-core EYDFA [5,7], and a 19-core EDFA [6]. The feasibility of cladding-pumped multicore amplifiers has been demonstrated in system experiments [8–10] and in long-distance transmission experiments over fully integrated in-line amplified 7-core [11] and 32-core systems [5,7]. Furthermore, a hybrid core and cladding-pumped EDFA was proposed for low power consumption transmission [12], and was demonstrated in a 7-core transmission experiment [13].

One major issue discussed regarding cladding-pumped multicore amplifiers is that signals in all cores will be amplified together, so we cannot adjust the input power and gain for each core individually. When there is significant loss variation between cores in a MCF transmission line, we need to spatially separate each channel, normalize the power spectrum and power level in each core, and spatially multiplex the wavelength division multiplexed (WDM) channels again into the MCF transmission line. Such operations will require additional equipment and will increase power consumption.

In this paper, we demonstrate hybrid amplification for SDM transmission systems using cladding-pumped multicore EDFA and core-pumped distributed Raman amplifiers. We have proposed a scheme that uses cladding-pumped amplification to efficiently amplify multiple spatial channels simultaneously, while distributed Raman amplification is used to perform spectral and spatial gain equalization for each spatial channel. In this work, we extend the previous work in [14] and employ a newly-developed 7-core WDM coupler and a 7-channel tap designed for incorporating the amplification scheme into a multicore transmission system. Moreover, we compare the maximum level of loss variation that can be compensated for by the proposed scheme when the source of loss is before or after the transmission line. We constructed a 7-core in-line amplified transmission line consisting of a 7-core MCF, a 7-core integrated WDM coupler, and a 7-core cladding-pumped EDFA and showed that the proposed core-by-core gain control scheme can compensate more than 5 dB loss variations between cores. Furthermore, we transmitted 100 GHz-spaced 46 WDM polarization-division multiplexed 16 quadrature amplitude modulation (PDM-16QAM) signals over a long distance of greater than 1000 km, and obtained good transmission performance with a Q-factor exceeded the 20% overhead forward error correction (FEC) limit of 5.7 dB.

2. Concept of hybrid cladding-pumped EDFA/Raman amplification

In SDM networks, the optical signal power in each core will differ because of loss variation between cores in MCFs and passive components, and gain difference between cores in optical amplifiers. Even with excellent multicore fibers, components, and amplifiers having uniform characteristics, power variation between cores can arise at splicing or connection points. Moreover, the power will change during network operation by the routing and allocation of WDM-SDM channels. Instead of performing spectral shaping to normalize the optical power level difference between cores, we propose active spatial gain equalization by using distributed Raman amplification.

The concept of MCF transmission system incorporating hybrid cladding-pumped EDFA/Raman amplification [14] is shown in Fig. 1. The system consists of a MCF, a multicore WDM coupler, a cladding-pumped multicore EDFA, optical monitors, and laser diodes for distributed Raman pumping. A multicore EDFA is used to amplify signals in multiple transmission lines, and the optical power in each core is monitored by an optical monitor. Low-power and low-noise distributed Raman amplification has been studied in conventional single-mode fiber transmission systems for active spectral gain equalization [15,16]. We apply this distributed Raman amplification in SDM systems to provide gain to WDM signals in cores having lower optical power relative to those in other cores. In previous studies, hybrid core-pumped EDFA/Raman amplification was used in high-capacity long-distance SDM experiments, where the Raman amplification was fully used in conjunction with core-pumped EDFAs for amplification of the DWDM channels in the full C-band through extended L-band regions [17,18]. On the other hand, in our proposed scheme, we use core-by-core Raman pumping for the purpose of equalizing the optical power and recovering the OSNR level of the WDM signals in each core. With this combined cladding-pumped EDFA and Raman amplification scheme, we can achieve low-power, high-capacity, and long-distance transmission systems.

 

Fig. 1 Concept of hybrid cladding-pumped EDFA/Raman amplification.

Download Full Size | PDF

3. Compensation of loss variation by active core-by-core gain control scheme

The 7-core experimental setup shown in Fig. 2 was constructed to test the proposed amplification scheme. For the transmission fiber, a trench-assisted 7-core MCF [19] was used. The length was 46.5 km, the core pitch was 49 μm, and the cladding diameter was 192 μm. The effective area of each core was greater than 110 μm², and the loss and chromatic dispersion at 1550 nm were 0.205 dB/km and 21 ps/nm/km, respectively. Multi-functional free-space optics device has been demonstrated integrating a multicore isolator into fan-in/fan-out (FI/FO) devices [20]. In our system, we used a novel free-space optics based 7-core WDM coupler that integrates a multicore variable optical attenuator (VOA), a multicore WDM, and a multicore isolator. The excess loss ranged from 0.6 to 0.8 dB, and the worst inter-core crosstalk was less than −42 dB. The excess loss of the pump path was less than 1.5 dB. The output of the 7-core MCF was spliced to the input of the 7-core WDM coupler. In addition, a 7-core integrated 10-dB splitter with an excess loss of 0.8 dB (including the 10-dB tap loss of 0.46 dB) was spliced to the output of the 7-core WDM coupler for optical power monitoring. We used FI/FO devices, also based on free-space optics, with an excess loss of 0.3 to 0.4 dB/port and crosstalk less than −57 dB. The FI/FO devices were spliced to the input and output of the 7-core MCF and the 7-core integrated tap, respectively, to form a 7-core transmission line. Raman pump LDs at 1424 and 1452 nm for each core were connected to the WDM coupling port via another FI device. We used a 7-core cladding-pumped multicore EDFA designed to have an optimum gain spectrum when the input power was around −4 dBm/core. The gain was greater than 16 dB, the NF was less than 6 dB, and the total crosstalk was less than −36 dB. The cladding pumped EDFA was operated at a constant power of 15 W throughout the experiment. Each of the FO output ports of the transmission line was connected to the corresponding FI input ports of the EDFA.

 

Fig. 2 Seven-core (a) transmission line and (b) cladding-pumped EDFA.

Download Full Size | PDF

We used 200-GHz-spaced 23 continuous wave (CW) light sources covering the full C-band to characterize the loss and OSNR of the constructed 7-core span. The optical power input to each of the seven input ports was set at a constant value of + 7.5 dBm/core. First, the 7-ch VOA integrated in the 7-core WDM coupler was set at 0-dB attenuation and the span loss was measured. Figure 3(a) shows the measured loss of the 7-core transmission line. The loss of the 7-cores ranged from 12.6 to 13.7 dB/span over the whole C-band. Next, the 7-ch VOA was set to control the average optical power input to the 7-core cladding-pumped EDFA at an optimized value of −4 dBm/core. The spectra of the span were measured using an optical spectrum analyzer with 0.1 nm resolution. Figure 3(b) shows the measured OSNR over the C-band. The OSNR difference between the seven cores was less than 1 dB due to good loss uniformity.

 

Fig. 3 Measured (a) span loss and (b) OSNR as a function of wavelength.

Download Full Size | PDF

In a more complex MCF transmission system, a loss difference of around 5 dB can arise [21]. To produce a state of loss variation between multiple cores, a VOA (VOA_out) was inserted in core #1, between the output of the FO port #1 of the 7-core integrated tap and the input of the cladding-pumped 7-core EDFA FI port #1. The attenuation value of VOA_out was set at 5 dB to increase the transmission loss of core #1 by about 5 dB relative to those of other cores as shown in Fig. 3(a). The OSNR of core #1 degraded by approximately 4 dB over the whole C-band. Here, we applied our proposed core-by-core gain control scheme using distributed Raman amplification. The pump LDs at 1424 and 1452 nm were operated at an equivalent pump power to recover the original spectra. The total power required for compensating this 5 dB loss variation by the Raman gain was estimated to be 0.5 W per core [22]. In a worst case when six out of seven cores have a 5 dB higher loss than the core with the minimum loss, the required power for compensation would be 3 W per fiber. The resulting OSNR, also shown in Fig. 3(b), was almost identical to the original OSNR of core #1 observed without the additional 5-dB loss. Figures 4(a) and 4(b) show the measured Raman and EDFA gain, respectively. The Raman gain over the C-band was around 5 dB. This compensated for the excess loss produced by the additional 5-dB attenuation, and the EDFA gain of core #1 with the 5-dB attenuation and hybrid EDFA/Raman was almost the same as the EDFA gain of the seven cores without the additional attenuation.

 

Fig. 4 Measured (a) Raman and (b) cladding-pumped multicore EDFA gain.

Download Full Size | PDF

The OSNR values for different VOA settings in the 0-15 dB range were also measured. Figure 5(a) shows the measured OSNR values averaged over the C-band for different VOA_out settings. When we used only the cladding-pumped EDFA, the OSNR degraded along with larger loss deviation of core #1 from other cores, produced by increasing VOA_out. Using the hybrid EDFA/Raman amplification, the transmission line restored the original OSNR properties. As we increased the loss deviation of core #1 by increasing the attenuation value, the Raman gain needed for compensating the loss increased. The power as a function of Raman gain increases linearly as shown in [22]. Next, we tested the case when the main loss source was at the input of the MCF transmission line. A VOA (VOA_in) was inserted at the input of the FI port #1 of the 7-core MCF and the attenuation value of VOA_in was changed to increase the transmission loss of the WDM signals in core #1 relative to those in other cores. Figure 5(b) shows the measured OSNR values averaged over the C-band for different VOA_in settings. Using distributed Raman amplification, the OSNR level recovered up to around 7 dB additional loss. Comparing Figs. 5(a) and (b), we observed a difference in the amount of tolerable loss depending on the location of the loss source. When the loss variation mainly originates before the input or at the FI of the transmission line, we were able to compensate up to approximately 7 dB loss variation as shown in Fig. 5(b). For larger loss variation, the OSNR level became too low to be recovered by the Raman amplification, and some OSNR degradation remained. On the other hand, when the loss variation mainly originates after the Raman amplification, such as at the FO output or in the multicore EDFA, we were able to compensate loss variations of greater than 10 dB as shown in Fig. 5(a). The allowable loss variation will depend on the characteristics of the multicore amplifiers and system design, and so it must be taken into consideration.

 

Fig. 5 Measured average OSNR of core #1 with additional loss (a) after transmission, and (b) before transmission.

Download Full Size | PDF

4. Long-distance 7-core transmission experiment

The recirculating-loop system shown in Fig. 6 was constructed to test the long-distance transmission performance of the 7-core transmission line incorporating our proposed hybrid EDFA/Raman amplification scheme. The signal under test were generated at the transmitter (Tx) using a tunable external-cavity laser (ECL) with ~60-kHz linewidth and was modulated by an IQ modulator (IQM). Additional 45 channels from 191.7 to 196.2 THz (1527.994-1563.863 nm) covering the whole C-band region were generated using DFB lasers with 2-MHz linewidths. They were wavelength multiplexed, modulated by another IQM, and combined by an optical coupler. All 16QAM WDM signals were digitally generated, each driven at 12 Gbaud. A frame length of 32,880 was used. For each IQM, different pieces of pseudo-random-binary-sequence (PRBS) of length 2²³-1 were used to generate multi-level 16QAM signals.

 

Fig. 6 (a) Recirculating loop for evaluating long-distance transmission performance of the 7-core in-line amplified transmission line with hybrid EDFA/Raman amplification, and (b) core allocation of the 7-core MCF.

Download Full Size | PDF

Each of the signals under test and the remaining wavelength channels were polarization-division multiplexed by a PDM emulator with a 348 ns delay, and were combined with an optical coupler. The 100-GHz-spaced 46 WDM PDM-16QAM signals were split into two branches. The first branch was used for the core under measurement to provide signal for the recirculating loop. The second branch was further split into six input signals, which were used as non-recirculating crosstalk signals loaded to all other cores. The seven signals were relatively delayed by 0, Δt, 2Δt, …, and 6Δt with a unit time delay Δt of 20 ns for decorrelation of signals between the 7 cores. A recirculating loop was formed for core #1 (center core), while non-recirculating signals were allocated to the input of the remaining cores #2-#7. The optimum input power for this transmission setup was determined by varying the optical power of the WDM signals input to core #1 for various distances and measuring the Q-factors. As a result of optimization, the best results were obtained at + 7.5 dBm/core, and the optical powers of the signals input to the 7 cores were all set at this value at the input of the FI device. The recirculating loop included a gain flattening filter (GFF), a loop-synchronous polarization scrambler (PS), and an amplifier to compensate for the insertion loss of the GFF and PS. After recirculation, the signals were further filtered by tunable optical filters and input to a coherent receiver. They were then digitized at 40 GS/s using a 4-ch digital storage oscilloscope and the stored data was post-processed offline.

We measured the Q-factors as a function of distance for core #1 at the center wavelength λ₂₃ = 1545.322 nm for the following five cases: (i) no additional loss and EDFA only, (ii) with 5 dB attenuation at the FI input (VOA_in = 5 dB) and EDFA only, (iii) with 5 dB attenuation at the FO output (VOA_out = 5 dB) and EDFA only, (iv) with 5 dB attenuation at the FI input (VOA_in = 5 dB) and hybrid EDFA/Raman amplification, and (v) with 5 dB attenuation at the FO output (VOA_out = 5 dB) and hybrid EDFA/Raman amplification. Figure 7 shows the measurement Q-factors. In a normal case when there is no additional loss (cases (i)), the Q-factor exceeded the FEC limit of 5.7 dB with 20% FEC overhead at a distance of 1395 km (30 loops). The worst crosstalk in the transmission line was −36 dB. Even after about 1500 km transmission, the Q penalty caused by inter-core crosstalk was estimated to be less than 1 dB [10], and so the maximum transmission distance was mainly determined by the received OSNR. For the case with a 5 dB additional loss and with only the EDFA (cases (ii) and (iii)), the Q-factor degraded rapidly as transmission distance increased. On the other hand, applying the proposed core-by-core Raman gain control scheme (cases (iv) and (v)) enhanced the transmission performance, and the Q factors were almost the same as the measurement without additional loss (case (i)).

 

Fig. 7 Measured Q-factors as a function of transmission distance.

Download Full Size | PDF

We thus confirmed that our proposed core-by-core gain control scheme using hybrid EDFA/Raman amplification can compensate for loss variation between cores and with it good long-distance transmission performance was obtained.

5. Conclusions

We proposed and demonstrated hybrid cladding-pumped multicore EDFA/Raman amplification for SDM transmission systems. Core-by-core gain control with distributed Raman amplification was utilized to compensate for the loss variation between cores. We transmitted PDM-16QAM signals over a long distance of greater than 1000 km using a 7-core recirculating loop system and obtained good transmission performance. The proposed scheme should prove to be useful for applications of cladding-pumped multicore amplifiers in SDM transmission systems.