I. Introduction

In recent years, optical transmission technology has advanced rapidly with the combination of multilevel modulation formats, coherent detection, and digital signal processing (DSP). 100 Gb/s optical transport systems have been widely deployed, and 400 Gb/s and 1 Tb/s superchannels have been demonstrated in laboratories and field trials. While transmission technology is necessary to transmit large capacities at higher data rates over longer distances, optical networking is now gradually evolving toward elastic optical network (EON) or flex-grid optical network architecture [1,2] in order to make better use of optical network resources to accommodate the ever-increasing traffic demand. In EON, the required minimum spectral resources are allocated adaptively based on traffic demand and network conditions, and the EON may also be able to trade-off between spectral efficiency (SE) and the all-optical reach. Currently, the EON can be achieved, as described in the literature, by relying on the elastic transponder techniques that are able to vary either symbol rate, code rate, or modulation formats. One example of such elastic transponders is to support several modulation formats such as polarization-division-multiplexed quadrature phase shift keying (PDM-QPSK) and PDM 16 quadrate amplitude modulation (PDM-16QAM). A combination of transmission and networking technologies may help to make effective use of available optical networking resources; however, the most fundamental improvement to the efficiency in utilization of optical network resources is to optimize the optical fiber cabling infrastructures to provide better capacity upgrade pathways and ensure sufficient transmission margins. Another effective way is to improve amplification technologies that are widely utilized in an optical network.

Historically, there is no doubt that transmission fibers have been critical to the enormous success of optical communication technology. Optical fibers, as one of the enabling technologies used for optical networks, have evolved for several decades, and their optical properties have been engineered for low system cost and high transmission performance. Recently, major optical fiber cable companies are focused on research and development (R&D) on a new class of transmission fibers [3–7] that have large effective areas ($A_{\text{eff}}$) and ultra-low attenuation for 400 G and beyond PDM coherent transport systems. In this paper, we will describe the recent development of new large-area low-loss fibers and advanced amplifier technologies for next-generation high-capacity terrestrial long-haul optical networks. Section II describes the key optical fiber properties of new class fibers and discusses their impacts on the transmission performance for 400 Gb/s multilevel modulated coherent transport systems; practical consideration of the large-area fibers, such as splicing and cabling for terrestrial optical networks will briefly be addressed; and the experimental results of transmission performance of new large-area low-loss fibers on a number of advanced high-order modulation format systems will also be presented. We then describe two advanced optical fiber amplifiers that will improve the efficiency in utilization of optical network resources and reduce total system cost. Section III will discuss the design and performance of an arrayed optical fiber amplifier using a compact Er-doped fiber (EDF) ribbon for next-generation reconfigurable optical add/drop multiplexer (ROADM) nodes. Section IV will describe the ultrawideband complementary Raman/EDFA, which has $+70\mathrm{nm}$ bandwidth for seamless C+L-band transmissions; the experimental demonstration of transmission 34.6 Tb/s capacity in single band over 2400 km fiber will be presented. A brief conclusion is presented in Section V.

II. Large-Area Low-Loss Fibers

Next-generation optical networks will deploy 400 Gb/s and beyond transport systems, which employ coherent digital detection and PDM high-level modulation formats such as 16 QAM. In such systems, the chromatic dispersion and polarization-mode-dispersion (PMD) can be digitally compensated in the electrical domain; however, with wavelength division multiplexing (WDM), the cross nonlinearities make neighboring channels interact depending not only on their power but also on the state of polarization (SOP) of signals in such systems. The last is particularly problematic in polarization multiplexed coherent systems, as this is more sensitive to cross-polarization modulation (XPolM) [8], which is associated with nonlinear polarization rotation in fiber transmission links. This XPolM does not have an effect on NRZ noncoherent transmission systems, but it has an impact on a coherent system because the polarization tracking in digital receivers cannot follow fast (symbol-to-symbol) polarization changes. As a result, the polarization-multiplexed coherent systems, in particular for high-order QAM, are more susceptible to fiber nonlinear effects. Early research work showed at least 2–3 dB more nonlinear power penalties in a QPSK coherent system compared with that in direct detection systems [9]. This XPolX nonlinear effect is stochastic in nature from fiber transmission links; hence, it impairs the effectiveness of digital compensation. In addition, the high-level modulation formats require a much higher optical signal-to-noise ratio (OSNR). For example, in order to achieve the same performance at the same transmission distance, upgrading from QPSK-based 100 Gb/s to 16QAM-based 200 Gb/s and 400 Gb/s requires an OSNR improvement of about 6.5 and 10 dB, respectively. Hence, it is desirable to use new fiber types that have low nonlinear coefficients to retain the long-haul transmission capability when scaling up the spectral efficiency (SE) and the per-channel data rate. Recently, a new class of transmission fibers with ultra-large effective areas ($A_{\text{eff}}$) in a range of $120–155{\mathrm{\mu m}}^2$ and the attenuation loss as 0.1460 dB/km has been reported for terrestrial and submarine systems [3–7].

To reduce the fiber nonlinearity, which is proportional to $n_2/A_{\text{eff}}$, where $n_2$ is the nonlinear refractive index, it is necessary to increase the $A_{\text{eff}}$ of the fiber while not jeopardizing the bending performances. Step-index profiles with small index differences and large core diameters can be used to design large $A_{\text{eff}}$ fibers. However, the light confinement in the core becomes weaker, increasing the cutoff wavelength and degrading both macro- and micro-bending performance. Placing a depressed-index region in the cladding, or placing a trench, slightly apart from the core can improve the bending performance and reduce the cutoff wavelength [4]. A low Young’s modulus primary coating [6] can also be used to improve the micro- and macro-bending performance and to permit a larger fiber $A_{\text{eff}}$.

A simple figure of merit (FOM) was proposed to compare the new class fibers with standard single-mode fiber (SSMF) [10], which is described as follows:

 

Fig. 1. Relative FOM of large-area low-loss fiber normalized to SSMF for 80 km span transmission.

Download Full Size | PDF

We have recently developed large $A_{\text{eff}}$ TeraWave fiber for 100 G and beyond terrestrial long-haul optical networks. The TeraWave fiber is a single-mode fiber with a germanium-doped core and a depressed-index inner cladding region and is fabricated similarly as an AllWave zero-water-peak fiber. Considering the features of terrestrial cabling such as craft splicing, closures, and macro- and micro-bending, the $A_{\text{eff}}$ of TeraWave fiber is optimized to be $125{\mathrm{\mu m}}^2$. The average fiber loss, dispersion, and dispersion slope at 1550 nm are 0.185 dB/km, 20.0 ps/nm/km, and $0.06\mathrm{ps}/{\mathrm{nm}}^2/\mathrm{km}$, respectively. TeraWave fiber is ITU-T G.654.B [11] compliant; it uses the DLUX Ultra coating for excellent micro-bending performance, and it meets all macro-bending requirements in G.652.D and G.654.B. Volume splicing study using a commercially available splicer (for example, FiTel S178A) with a standard splice recipe shows that the averaged splice loss between TeraWave to TeraWave is 0.04 dB/splice, and TeraWave to SMF is about 0.15 dB/splice. With optimization of splicing programs using commercially available splicers, the splicing loss between TeraWave to SSMF is below 0.10 dB/splice.

We have also recently developed TeraWave ULL single-mode optical fiber, which is a $125{\mathrm{\mu m}}^2$ large area ultra-low-loss fiber, and the averaged fiber loss and dispersion at 1550 nm are 0.168 dB/km and 20.0 ps/nm/km, respectively. The fiber is optimized for long-haul transmission in the C+L-bands (1530–1625 nm) at 100 Gb/s, 400 Gb/s, and beyond for terrestrial optical networks, and it supports greater distances between regeneration and amplification sites, helping to lower the overall cost of deploying coherent systems.

We have systematically and experimentally investigated the system performance of TeraWave fiber in a 485 Gb/s coherent optical orthogonal frequency-division multiplexed (CO-OFDM) superchannel long-haul transmission system [3] and compared it with the SSMF. The CO-OFDM superchannel coherent system has advantages, including high SE, reduced guard band, and lower modulation baud rate, and it is a potential candidate for future high-capacity Tb/s-per-channel optical networks. However, the CO-OFDM superchannel systems are more susceptible to fiber nonlinearity when compared with other schemes. This is because the multisubcarrier configuration with small guard band in CO-OFDM transmission leads to the impairment of inter-subcarrier nonlinear interference. In the experiment, the 485 Gb/s CO-OFDM was generated with PDM-16QAM five-subcarrier modulation, and it was done with a dispersion uncompensated link with 80 km fiber span length using two amplification schemes, erbium-doped-fiber amplifiers (EDFAs) and hybrid EDFA/Raman amplifiers. The comparison experiment showed that the optimum signal launch power into fiber spans is about 2 dB higher in TeraWave fiber than in SSMF for transmitting of 485 Gb/s CO-OFDM signals over 1600 km ($20\times 80\mathrm{km}$) spans under both EDFA-only and hybrid Raman/EDFA amplification schemes. With EDFA-only amplification, the TeraWave fiber offers an $\sim 2\mathrm{dB}$ higher optimum $Q^2$ factor than the SSMF after 1600 km transmission. It is about 1 dB higher for the hybrid amplification scheme (Fig. 2). It was also found that TeraWave fiber allows transmission more than 60% longer than for a SSMF link at a similar $Q^2$ factor performance for 400 G OC-OFDM systems.

 

Fig. 2. Measured $Q^2$ versus launch powers of 485 Gb/s CO-OFDM signal transmission over 1600 km of ULAF and SSMF [3].

Download Full Size | PDF

We have further experimentally investigated the transmission performance of TeraWave fiber in a typical 256 Gb/s PDM-16QAM DWDM system with 37.5 and 50 GHz channel spacing [12]. The 256 Gb/s PDM-16QAM is a promising modulation format for two carrier 400 Gb/s systems due to a relatively simple scheme and maturity of opto-electronic components. The experiment was conducted with a dispersion uncompensated link with a 100 km fiber span length using three different amplification schemes, including EDFA-only, hybrid EDFA/Raman amplifiers, and all backward-pumped Raman amplifiers. It has been found that, when there is no ROADM, ten 256 Gb/s PDM-16QAM channels were transmitted over 4200, 3500, and 2000 km over TeraWave fibers within a 20% soft-decision FEC (SD-FEC) threshold, respectively, in the system with both 37.5 and 50 GHz channel spacing (see Fig. 3). The all backward-pumped Raman amplifiers and hybrid EDFA/Raman amplifiers can increase the transmission distances by about 100% and 70%, respectively, when compared with EDFA-only amplifiers. It was also found that cascaded ROADMs have a small impact on the system with the 50 GHz channel spacing.

 

Fig. 3. $Q^2$ factor versus transmission distance of the 10 channel 256 Gb/s PDM 16QAM system using the three different amplification schemes without (solid curves) and with ROADMs (dashed curves) with 50 GHz channel spacing. Optimum launch powers are used.

Download Full Size | PDF

The large-area low-loss fiber has also been studied in other advanced higher-order modulation format systems (for example, time-domain hybrid 32–64 QAM, for high SE 400 Gb/s ultra-long-haul transmission systems [13]), and it has been demonstrated that the 400 Gb/s-class DWDM signals on the standard 50 GHz ITU-T grid, which is 8.25 b/s/Hz net SE, were transmitted over 4000 km of large-area low-loss fiber with 100 km span length for a terrestrial optical network.

The above results demonstrated that the large-area low-loss fibers are indeed beneficial for system reach and margin irrespective of amplifier configuration; they increase system SE and have much better efficiency in utilization of optical network resources for future EON. Further enlargement of $A_{\text{eff}}$ (e.g., $>200{\mathrm{\mu m}}^2$) would inevitably deteriorate bending and cutoff behavior and cause high splicing loss. Hence, fiber design compromises must be made between improved transmission performance and limitations from careful handling and splicing of new fibers.

III. Arrayed EDFA Using Ribbonized EDF

The EON will not only be able to trade-off between SE and system reach but also provide more networking flexibility in terms of wavelength path routing, switching, and assignments. In such networks, multiple-degree ROADM with colorless, directionless, contentionless (CDC) functions is expected to play an important role for realizing dynamic capacity allocation [14]. However, the insertion loss of the ROADM nodes with these sophisticated functions is generally higher than that for the basic 2-deg node. As a result, an optical amplifier array with $1\times 8$ or $1\times 16$ is often required in order to compensate additional insertion losses in $M\times 1$ wavelength selective switches (WSSs) and $1\times N$ optical couplers [14]. Currently, a number of discrete EDFA modules using separately pumped diodes are usually employed. As the degree of ROADM increases, the ROADM becomes bulky, costly, and inefficient due to many such individual EDFAs. For example, if we assume that a commercially available module is installed between a $1\times 4$ WSS and $4\times 1$ optical coupler in the CDC ROADM node, a few hundred discrete EDFA modules are required in a single node [15]. This approach would become undesirable in terms of equipment size and cost when the degree of ROADM is increased. As a result, a compact arrayed EDFA can be beneficial in reducing the array module size and manufacturing cost for future ROADM nodes.

The concept of an arrayed optical amplifier by shared pump lights has been proposed to reduce amplifier array module size; however, these approaches use either a discrete Er-doped fiber (EDF) [15], which occupies a large space, or an erbium-doped waveguide amplifier (EDWA) [16], which is still not widely commercially available due to its poor performance. We have recently proposed and designed an arrayed fiber amplifier using ribbonized EDF to reduce the EDFA array module size and manufacturing cost (e.g., with one ribbon EDF coil instead of making eight EDF coils). A schematic of a compact arrayed EDFA using an EDF ribbon is shown in Fig. 4; it is composed of optical isolators, WDM combiners, and an EDF ribbon and configured in a co-propagation pumping scheme. Other functions such as gain flattening and input/output signal monitors can also be implemented.

 

Fig. 4. Diagram of a $1\times 8$ arrayed EDFA. Inset: schematic of an 8-EDF ribbon.

Download Full Size | PDF

A diagram of an 8-EDF ribbon is shown in the inset of Fig. 4. The EDF ribbon includes a plurality of individually coated EDFs arranged in a close-packed round shape; a colored ink layer is applied over each individual EDF coating for easy identification. The colored EDFs are encased in a two-layer structure, in which a soft UV-cured inner layer is used to cushion fibers, and a hard UV-cured outer layer is used to protect fibers. In order to obtain the same amplification performance of each EDFA in the array, each individual EDF within the ribbon has the same optical properties such as Er absorption, MFD, cutoff wavelength, and background loss. Table I shows the peak absorption of each EDF in the ribbon with the average peak absorption of 6.49 dB/m and standard deviation of 0.04 dB/m, showing good uniformity of the EDF. The EDF ribbon peels for easy fiber access, and it is made in a round shape for easy packaging. The EDF ribbon has a diameter of 1.25 mm and can be coined in a compact form. The ribbon diameter can be further reduced by using 125/200 μm or 80/165 μm cladding/coating diameter EDF, respectively.

TABLE I. Peak Absorption of Individual EDF in Ribbon EDF

View Table

The input/output of individual EDFs can be separated by peeling off from the ribbon EDF and splicing to other fiber components. Pump power can be shared by multiple EDFs, for example, by splitting power from one diode into multiple portions, with each portion used to pump individual EDFs in the ribbon. Power can be managed using individual variable optical attenuators (VOAs). Thus, the arrayed EDFA can be made compact, efficient, and low cost.

We developed a solid 8-EDF ribbon in a round shape with a diameter of 1.25 mm and experimentally investigated the performance of the arrayed EDFA by using a $1\times 8$ ribbonized EDF. Each EDFA in the $1\times 8$ arrayed amplifier was pumped by 976 nm LD 405 mW pump power in a co-propagation pumping scheme, and a 12 m length of ribbon EDF was employed in the experiment. Figure 5(a) shows the measured gain and the gain deviation from the averaged gain curve for each EDFA in the $1\times 8$ arrayed EDFA. The average gain of 19.3 dB is obtained with average output power of 21.9 dBm; and the gain deviations were $<\pm 0.8\%$, exhibiting excellent gain shape uniformity among the EDFA within the array. This gain deviation can be further reduced by slightly adjusting the length of individual EDF and the pump power. Figure 5(b) plots the measured noise figure (NF) of eight EDFAs from the arrayed EDFA, and the NFs range from 3.5 to 4.8 dB across the C-band for eight EDFAs, showing good performance of the arrayed EDFA.

 

Fig. 5. (a) Measured gain and gain deviation of individual EDFAs in an arrayed EDFA. (b) Measured NF of the eight EDFAs in a $1\times 8$ arrayed amplifier.

Download Full Size | PDF

IV. Ultrawide Complementary Raman/EDFA for Seamless C+L-Band Transmission

One of the important features for EON is to improve the efficiency in utilization of optical network resources. It is well known that optical fiber has tremendous transmission bandwidth (e.g., $+9\mathrm{THz}$ bandwidth in the C+L-band). However, currently major network operators deploy optical fiber transport systems only in either the C-band or the L-band with the bandwidth around 4 THz using EDFA. Multiband transmission systems have been considered to increase capacity in a single fiber. Typical C+L-band systems with separate C- and L-band EDFAs require guard bands of 8 nm or more near 1565 nm. A seamless C+L transmission band avoids the band splitters and combiners required by systems with C+L EDFAs, thus simplifying the system configuration and potentially lowering overall system cost. In addition, the reduced number of loss elements due to the C+L-band splitter/combiners along the transmission link may result in higher received OSNR, thus improving system performance. One of the technical challenges for such ultrawideband transport systems is the high-efficiency seamless wideband optical amplifier. In this section, we describe the design and performance characteristics of ultrawideband complementary Raman/EDFAs that have the flat bandwidth of $+70\mathrm{nm}$ and present the experimental results of 34.6 Gb/s single-band transmission over 2400 km of TeraWave fiber using this complementary Raman/EDFA.

A. Complementary Raman/EDFA

The simple scheme of the complementary Raman/EDFA is shown in Fig. 6. Each of the EDFA and Raman amplifiers covers a different portion of gains for the C+L-bands, and they are combined in a serial cascade. In this example, the EDFA section is a standard single-stage 980 nm forward-pumped scheme using commercially available EDF. The EDFA section mainly provides the gain for the C-band, whereas backward-pumped distributed Raman amplification from a transmission line using the pump lasers at the pump wavelength around 1490 nm provides the L-band gain. Here, a 100 km TeraWave fiber span was used for the transmission line. The TeraWave fiber used in this experiment has a 0.182 dB/km attenuation at 1550 nm, and its peak Raman gain efficiency at 1550 nm is 0.25/(W.km). A gain flattening filter (GFF) is designed and inserted between Raman and EDFA to equalize the combined Raman–EDFA gain shape; here, a fixed GFF was used in this experiment (see Fig. 7). Compare the case for use of the C+L band with two separate EDFAs, where a C/L wavelength band coupler/splitter is used; in this simple structure, only a few optical components are needed, so it is cost effective. The gain bandwidth of the complementary Raman/EDFA can be as large as about 80 nm (1530 to 1610 nm). A wideband flat gain of this amplifier can be obtained; however, the NF may not be flat over the bandwidth.

 

Fig. 6. Diagram of a complementary Raman/EDFA for a 100 km TeraWave fiber span.

Download Full Size | PDF

 

Fig. 7. Designed GFF used in the complementary Raman/EDFA.

Download Full Size | PDF

In order to improve the flatness of the NF, which will be important for a relatively flat OSNR over the wide C+L-band, additional Raman pump lasers at multiple different wavelengths can be employed. We fabricated a 73 nm bandwidth complementary Raman/EDFA using two pump wavelengths (1425/1489 nm). In this design, the EDFA section used 9.5 m of OFS MP980 EDF pumped by a 980 nm diode with 460 mW pump power. Commercially available semiconductor diodes were used as the Raman pumps after de-polarization using a polarization beam combiner. The measured EDFA and Raman on–off gain are plotted in Fig. 8(a) when the pump powers at 1425 and 1489 nm are 390 and 580 mW, respectively. It can be seen that about 7 dB Raman gain in a short wavelength range in the C-band was obtained by the 1425 nm pump. The measured total gains after GFF and the effective NF are shown in Fig. 8(b). The output power of the Raman/EDFA is 22.4 dBm and the gain flatness is 0.82 dB. Gain values can be adjusted slightly by setting the pump powers in order to accommodate different span losses in real terrestrial systems; however, the gain flatness will be affected if a fixed GFF is used and an adaptive GFF can be implemented to further improve the gain flatness. It also can be seen that the effective NF of the L-band is lower than that in the C-band, which is due to the relatively high gain from the distributed Raman amplifications. It is a result of the distributed nature of the gain along the transmission fiber, compared with the lumped gain of the EDFA at the end of the fiber. The large $A_{\text{eff}}$ TeraWave fiber provides a good balance between avoiding nonlinear impairments while allowing sufficiently high Raman efficiency so that high distributed gain can be achieved with $<1\mathrm{W}$ pump power using diodes at only two wavelengths.

 

Fig. 8. (a) Measured EDFA gain and Raman gain from the complementary Raman/EDFA. (b) Measured total gain w/GFF and effective NF.

Download Full Size | PDF

B. 34.6 Tb/s Single-Band Transmissions

Several system demonstrations of ultrawide, seamless-band transmission were reported more than 10 years ago [17–19]. Recently, a 54 Tb/s transmission with a seamless bandwidth of $\sim 73\mathrm{nm}$ has been demonstrated for submarine systems [20], and ultrawide seamless band transmissions of 9 Tb/s [21] and 17.3 Tb/s [22] have been reported for terrestrial network applications. Both all-Raman distributed amplifiers and hybrid Raman/EDFA have been reported for the ultrawide seamless band long-haul transmission, compared with all-Raman schemes; the complementary Raman/EDFA offers low pump consumption and possibly low overall cost for the long-haul network. Here, we describe an experimental demonstration of transmission of 34.6 Tb/s capacity in a 70 nm seamless band over 2400 km of a TeraWave fiber.

The 34.6 Tb/s single-band transmission experiment was done by using $173\times 256\mathrm{Gb}/\mathrm{s}$ PDM-16QAM DWDM channels with a dispersion uncompensated fiber link [23], and the diagram for this experiment is shown in Fig. 9. The transmitters consisted of 173 distributed feedback (DFB) lasers and a tunable external cavity laser (ECL) at wavelengths ranging from 1530.31 to 1600.60 nm on a 50-GHz-spaced ITU frequency grid. Commercially available four-channel 64 GS/s digital-to-analog converters (DAC) were employed to generate 32 Gbaud 16 QAM signals and an offline DSP coherent receiver by 80 GS/s digital oscilloscope as the analog-to-digital converter (ADC) was used to evaluate system performance (BER counting and $Q^2$ factor calculation) [23]. Four units of the complementary Raman/EDFA aforementioned were employed to compensate for the loss of 100 km fiber spans in the recirculating loop experiment. The optimization of the transmission performance was carried out by pre-emphasis for the output power of ECL for selected channels from the C+L-bands, and the total launched power into fiber spans was about 20.1 dBm, which was limited by the output power from DRAs. Figure 10(a) plots the performance of four selected channels versus the transmitter pre-emphasis power after six loop (2400 km) transmissions, and it indicates that the signal launched power in the L-band is slightly lower than these in the C-band. This is due to the fact that the signal path average power in the L-band is generally higher than that in the C-band because the large portion of gain in the L-band is contributed from distributed Raman gains. The $Q$-factor of the four selected channels as a function of transmission distance is shown in Fig. 10(b), showing that the L-band channels can potentially have a reach of $\sim 3200\mathrm{km}$. The recovered signal constellation for $x$ and $y$ polarization of channel 1554.94 nm after 2400 km transmission is shown in the inset of Fig. 10(b). Finally, the channel powers in the C+L seamless band were slightly pre-emphasized; the received optical spectrum after transmission is plotted in Fig. 11(a), showing the negative tilt of signal power about 3 dB across the entire C+L-band. The received OSNRs for 173 channels after 2400 km transmission ranged from 18.9 to 22.8 dB (in 0.1 nm RBW), with an average of 20.7 dB [Fig. 11(b)]. The average $Q$-factor was 6.83 dB after 2400 km transmission, with the worst channel $Q$-factor of 6.08 dB, which is above the SD-FEC threshold of 5.92 dB $Q$-factor (BER of $2.4\times {10}^{-2}$), and would yield a BER below ${10}^{-15}$ after correction by the SD-FEC [24].

 

Fig. 9. Diagram for seamless band transmission loop experiment.

Download Full Size | PDF

 

Fig. 10. Performance versus transmission distance. Inset shows the recovered constellations for $x$ and $y$ polarization of the channel 1554.94 nm after 2400 km.

Download Full Size | PDF

 

Fig. 11. (a) Received optical spectrum. (b) Received OSNR and measured -$Q$factors after $24\times 100\mathrm{km}$ transmission.

Download Full Size | PDF

This demonstration indicates the potential high-capacity seamless C+L-band transmission by using the wideband complementary Raman/EDFA plus large-area low-loss TeraWave fiber for a long-haul terrestrial optical network. The seamless band transmission permits the traffic to upgrade smoothly and to be more flexible in routing and assignment of wavelength channels across the C+L-band for EON. As this experiment was done at 50 GHz channel spacing, the total capacity can be easily scaled up to $+50\mathrm{Tb}/\mathrm{s}$ for a single-band transport system by using 37.5 GHz channel spacing.

V. Conclusion

This paper reviewed new developments of ultra-large-area low-loss fibers for next-generation high-capacity terrestrial long-haul optical networks. The key optical fiber properties of new class fibers have been described, and their impacts on the transmission performance for 400 Gb/s PDM multiple-level modulation coherent transmissions have been discussed. In addition, we have described two advanced optical fiber amplifier technologies that improve the efficiency in utilization of optical networking and reduce total system costs. The design and performance of an arrayed optical fiber amplifier using a compact ribbonized EDF for next-generation ROADM nodes have been discussed; the design and characteristics of complementary Raman/EDFA, which has $+70\mathrm{nm}$ bandwidth for seamless C+L-band transmissions, have also been described. Finally, the experimental demonstration of 34.6 Tb/s transmission capacity in a single band over 2400 km fiber was presented.