Assessment On The Achievable Throughput of Multi-Band ITU-T G.652.D Fiber Transmission Systems

This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2020.2989620, Journal of
Lightwave Technology
JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. XXXX, NO. YYYY, WWWW HHH, ZZZZ 1
Assessment on the Achievable Throughput

of Multi-band ITU-T G.652.D
Fiber Transmission Systems
Alessio Ferrari, Antonio Napoli, Johannes K. Fischer, Nelson Costa, Andrea D’Amico, João Pedro
Wladek Forysiak, Erwan Pincemin, Andrew Lord, Alexandros Stavdas, Juan Pedro F.-P. Gimenez
Gunther Roelkens, Nicola Calabretta, Silvio Abrate, Bernd Sommerkorn-Krombholz, and Vittorio Curri
Abstract—Fiber-optic multi-band transmission (MBT) aims at carried out for other single-mode fiber types. We estimate a total
exploiting the low-loss spectral windows of single-mode fibers single-fiber throughput of 450 Tb/s over a distance of 50 km and
(SMFs) for data transport, expanding by ∼11× the available 220 Tb/s over regional distances of 600 km: ∼10× and 8× more
bandwidth of C-band line systems and by ∼5× C+L-band than C-band transmission respectively and ∼2.5× more than full
line systems’. MBT offers a high potential for cost-efficient C+L.
throughput upgrades of optical networks, even in absence of
available dark-fibers, as it utilizes more efficiently the existing Index Terms—Multi-band fiber transmission, high-capacity
infrastructures. This represents the main advantage compared systems, elastic optical networks.
to approaches such as multi-mode/-core fibers or spatial division
multiplexing. Furthermore, the industrial trend is clear: the I. I NTRODUCTION
first commercial C+L-band systems are entering the market
and research has moved toward the neighboring S-band. This
article discusses the potential and challenges of MBT covering
the ITU-T optical bands O → L. MBT performance is assessed
L EGACY optical networks operated on the C-band and
deploying direct-detection 10G optical WDM transmis-
sion technologies, could achieve a maximum throughput of
by addressing the generalized SNR (GSNR) including both the ∼1 Tb/s/fiber. Coherent optical technologies have dramati-
linear and non-linear fiber propagation effects. Non-linear fiber
propagation is taken into account by computing the generated cally improved the transmission scenario, and commercial
non-linear interference by using the generalized Gaussian-noise transponders can now convey over C-band line-systems up
(GGN) model, which takes into account the interaction of non- to 38.4 Tb/s/fiber [1], by implementing a 38× transmission
linear fiber propagation with stimulated Raman scattering (SRS), throughput increase, keeping the transmission infrastructure
and in general considers wavelength-dependent fiber parameters. unchanged. On the other hand, it is predicted that IP traffic
For linear effects, we hypothesize typical components’ figures
and discussion on components’ limitations, such as transceivers’, demand will continue to grow at a compound annual growth
amplifiers’ and filters’ are not part of this work. We focus rate (CAGR) as large as ∼26% [2] for back-bone networks,
on assessing the transmission throughput that is realistic to and it might be even larger for metro scenarios and data center
achieve by using feasible multi-band components without specific interconnect (DCI) [2]. In addition, 5G and high-capacity
optimizations and implementation discussion. So, results are access traffic as well as machine-to-machine communication,
meant to address the potential throughput scaling by turning-
on excess fiber transmission bands. As transmission fiber, we together with the expansion of cloud services will further load
focus exclusively on the ITU-T G.652.D, since it is the most the back-bone optical network infrastructure. Consequently,
widely deployed fiber type worldwide and the mostly suitable to the per-fiber throughput enabled by the coherent transmission
multi-band transmission, thanks to its ultra-wide low-loss single- technologies is not sufficient anymore to support the envi-
mode high-dispersion spectral region. Similar analyses could be sioned IP traffic explosion, and alternative solutions must be
found; with a firm request from Carriers to fully exploit the
This work was fundedby the H2020 Metro-Haul project, no. 761727; and installed transmission equipment in order to maximize returns
by the European Union Horizon 2020 research and innovation program under
the Marie Skłodowska-Curie ETN WON, grant agreements 814276. from investments [3].
Alessio Ferrari, Andrea D’Amico and Vittorio Curri are with Politecnico Two different approaches have been proposed to upgrade
di Torino, Torino, Italy, [email protected]. infrastructures to cope with these requirements: spatial division
Antonio Napoli and Bernd Sommerkorn-Krombholz are with Infinera,
Sankt-Martin-Str. 76, 81541, Munich, Germany, [email protected]. multiplexing (SDM) with multi-fiber (MF) or multi-core/mode
Johannes K. Fischer is with Fraunhofer Institute for Telecommunications fiber (MMF/MCF) transmission; and multi-band transmission
Heinrich Hertz Institute, Einsteinufer 37, 10587 Berlin, Germany. (MBT), which more efficiently exploits the available spec-
Nelson Costa and João Pedro are with Infinera, Lisbon, Portugal. João Pedro
is also with Instituto de Telecomunicaçoes, IST, UL, Lisbon, Portugal. trum of a single fiber. MBT and SDM are not mutually
Wladek Forysiak is with Aston Institute of Photonic Technologies, UK. exclusive, as MBT is a technique that maximizes the per-
Erwan Pincemin is with Orange, Lannion, France. fiber transmission, that can be eventually combined with SDM
Andrew Lord is with British Telecommunications, Ipswitch, UK.
Juan P. F.-P. Gimenez is with Telefonica, Madrid, Spain. by activating additional fibers, when needed. For both MF
Gunther Roelkens is with Gent University, Belgium. and MBT, transport platforms are commercially available,
Silvio Abrate is with Links Foundation, Torino, Italy. while MMF/MCF transmission needs complex multiple-input-
Nicola Calabretta is with TU/e, Eindhoven, The Netherlands.
Alexandros Stavdas is with University of Peloponnese, Greece. multiple-output (MIMO) transceivers not yet commercially
Manuscript received xxx 19, zzz; revised January 11, yyy. available. SDM can be implemented following two main
0733-8724 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on April 23,2020 at 14:42:52 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2020.2989620, Journal of
0.9 30
strategies: I) by using the available dark-fibers, i.e., MF ITU-T G.652A
ITU-T G.652D
transmission [4] or II) by following the much costlier op- 0.8 25
O-band E-band S-band C-band L-band
eration of rolling out new optical fiber infrastructures [5],
0.7 20
consisting of MF/MMF/MCF. The MMF/MCF approach leads
Dispersion [ps/nm/km]
Attenuation [dB/km]
to a potential capacity in the Petabit/s/fiber range [6], [7], 0.6 15
but it requires to establish a new transmission ecosystem with
0.5 10
huge CAPEX efforts. Moreover, the required technologies and
standardization lack of maturity for a near-term commercial 0.4 5
deployment. 0.3 0
On the other hand, SDM with MF transmission is the most
widely adopted solution in case of available dark-fibers as it 0.2 -5
PDFA BDFA TDFA EDFA EDFA
relies on mature and cost-effective technologies by replicating 0.1 -10
1260 1310 1360 1410 1460 1530 1565 1625
C-band line systems. Nevertheless, in absence of available Wavelength [nm]
dark-fibers, techno-economics address the operators towards
postponing such as solution that would require to install or Fig. 2: Measured attenuation (green and blue) and dispersion
lease new cables, with large CAPEX efforts and operational coefficient (black) for ITU-T G.652.A and ITU-T G.652.D
delays. If ducts are full, the cost for a new fiber roll-out fibers. DFA stands for doped fiber amplifiers, and the consid-
is ∼25 ke\km in rural areas, and up to ∼500 ke\km in ered doping materials are listed in Tab. II.
metropolitan areas. SDM with MF transmission is highly
costly also for network operators leasing fibers, that occurs bands will be considered for data transport, starting from the S-
when: I) they manage networks abroad1 ; II) they lease in- band [13]. In principle, all types of single-mode fibers could be
frastructures owned by incumbent operators; III) they cannot utilized for MBT, but the available bandwidth will depend on
afford to deploy the new fibers. Moreover, engineering works the actual fiber characteristics, mainly on the cut-off frequency
may imply large and unpredictable delays because of needs defining the single-mode spectral region. Fibers with hydroxyl
for local authorities approvals, and negotiations, in general. In ions (OH) causing an absorption peak in the E-band such as
summary, MBT is an attractive solution to support the continu- ITU-T G-652.A/B would see a significantly lower capacity
ous growth in fiber transmission demand, as it postpones fiber compared to modern, but already widely deployed ITU-T G-
roll-outs, and maximizes the return-on-investments on existing 652.D. On the other hand, pure silica core fiber (PSCF) which,
infrastructures [8], and is a seamless and complementary besides having a zero OH absorption peak, presents a very
solution to SDM. large effective area – small non-linear coefficient – and very-
low attenuation, will provide better performance than ITU-T
Band O E S C L G-652.D.
1530 1565 In 1990s, fiber manufacturer were able to purify silica
Wavelength
1260-1360 1360-1460 1460-1530 - -
(nm) [9] 1565 1625 from the OH ions, and the ITU-T G.652.D SMF fiber with
C-band 35 nm no absorption peak was proposed, standardized in 2003 and
C+L-band 95 nm commercialized to extend the transmission bandwidth of fiber
All bands 365
365 nmnm systems [14], [15]. Consequently, the market share for ITU-
T G-652.A/B have progressively decreased over the last 15
Fig. 1: ITU-T band definition for single-mode fiber [9]. years. According to [16], the cumulative percentile of ITU-T
G.652.D fiber – or compatible – deployed worldwide over the
MBT aims at enabling transmission beyond the C-band years 1993–2018 is ∼81% with an increase to up to ∼97%
by opening up the 2nd and 3rd low-loss windows of single- in 2013-2018. Similar data are reported in [15]. Note that
mode fibers (SMF) as depicted in Fig. 1 of [10] for ITU- these figures refer average value worldwide. For example, in
T G.652.A/D fibers. For the mostly and widely deployed the USA, a large part of the nationwide backbone is made
ITU-T G.652.D fiber [11], the ≤ 0.4 dB/km overall spectral of non-zero dispersion-shifted fiber (ITU-T G.655) with large
region ranges from 1260 nm to, at least, 1625 nm for a effective area, for which MBT is feasible only on C+L bands.
total bandwidth of ∼53.5 THz, i.e., ∼11× the C-band2 . The Among the fibers without the absorption peak, ITU-T G.652.D
optical bands {O, E, S, C, L} depicted in Fig. 1 are defined is the most deployed [11], [16] and is, therefore, the one
as per ITU [9]. Commercial MBT are targeting the upgrade we have considered in our analysis. Its measured physical
of existing C-band line systems, first, with commercially parameters are reported in Fig. 2: the black line displays
available C+L-band systems, by capitalizing on the re-use of the dispersion whereas the green and blue lines show the
Erbium amplifier technology also within the L-band [12]. This attenuation for legacy ITU-T G.652.A (from the 80’s) and
operation adds ∼7 THz to the ∼5 THz of the C-band for a ITU-T G.652.D optical fibers, respectively.
total bandwidth of ∼11.5 THz. In a second step, the remaining Fig. 2 shows that MBT has to cope with wavelength depen-
dent channel characteristics: the dispersion coefficient D(λ)
1 In this case, they are subjected to the host country regulations, which
ranges from –5 ps/nm/km to ≥20 ps/nm/km (from O→L-
usually do not allow foreign companies to freely deploy new fibers.
2 Note that commercial C-band systems occupy wider bandwidth than ITU band); whereas the fiber loss α(λ) ranges from ∼0.38 dB/km
standardized C-band (in O-band) to ∼0.18 dB/km (in C-band). Consequently, to
TABLE I: Comparison of strategies for network capacity expansion.

Strategy Main pros Main cons Maturity Capacity
1) Costly new fiber deployment Proportional to the
1) Mature technology
Multi-fiber 2) Complex management of SDM net- High number of available
2) Exploitation of available dark-fibers
works fibers
1) New deployment of optical fibers
2) Development of new components Proportional to the
Multi-core / - 3) Complex management due to large Significantly number of available
High capacity per optical fiber
mode fiber amount of WDM channels and dif- low fibers, number of cores
ferent cores and modes
1) Development of new components for

1) Exploits the full low-loss region of
Multi-band O, E, and S-bands. Up to ∼10× the current
SMF Low
transmission 2) Complex management due to large capacity per fiber
2) Maximizes previous investments
amount of WDM channels
achieve a full optimization of transmission over such a wide (NLI). The goal of this work is to assess the MBT potentialities
spectrum, adaptation to the fiber characteristics on a per- focusing on limitations introduced by fiber propagation, so,
channel and per-band basis is required. For instance, different we deliberately suppose to rely on ideal amplifying and
modulation formats might be assigned to different transmission filtering components as well as ideal rate-variable (flexible)
bands as proposed in [17]. Alternatively, probabilistic shaping transceivers, by neglecting impairments arising from non ideal
and rate-adaptive forward error correction (FEC) could be used components. Thus, we suppose amplifiers delivering a flat per-
to achieving a finer granularity [18]. band gain and noise-figure followed by ideal gain flattening
In the following paragraphs, we summarizes goals, filters. For per-band values of noise-figure we refer to feasible
assumptions and results of this article. amplifying technologies as reported in literature. As power
control strategy, we suppose the sub-optimal per-band flat
Goals: We aim at assessing the achievable throughput on a launch power at every fiber span and the per-band power
single mode fiber transmission system when relying on MBT spectral density is computed out by applying the local op-
on ITU-T G.652.D fiber. We assume the deployment of 32 timization for a global optimization (LOGO) strategy to each
Gbaud ideal rate-variable transceivers [18] on the 50 GHz band [30]. We are fully aware of the importance of analyzing
WDM fixed grid, so, results are not to meant as ultimate the components’ limitations and of the need to realize an
capacity assessment, but as reasonable figures achievable by ecosystem of devices enabling end-to-end transmission [31],
turning-on additional bands in MBT, with the aim to assess e.g., optical MBT amplifiers and filters, as reported in Sec. II.
the band-by-band throughput scaling-up factor referred to the But such analyses, as well as techno-economics, would be
C-band’s. We focus the analysis on limitations introduced by carried-out on top of fiber propagation potentialities in order to
fiber propagation, so we do to address specific components keep separated the potentialities of different design leverages.
and control-software optimizations. Some technologies are indicated merely with the objective of
Assumptions: As performance meter, we use the gener- providing a strategy to implement a pay-as-you-grow approach
alized signal-to-noise ratio (GSNR) [19], [20] that includes for system upgrade.
both the accumulated ASE noise introduced by amplifiers to Results: We derive, by estimating the GSNR, the total trought-
counteract fiber loss and SRS spectral tilt, and the non-linear put per band for the following scenarios: I) DCI-like, i.e.,
interference (NLI) generated by fiber non-linear propagation ≤ 80 km; II) metro networks ∈{150 km, 300 km}; and III)
by using the GNPy library [21] as it has been extensively extended metro & regional networks, i.e., up to 600 km. We
validated [22]. For the NLI calculation we follow a completely also assess the impact of using the O-band for Raman pumping
disaggregated approach where each fiber span introduces its instead of exploiting it for data transmission, to enhance
NLI contribution independently. For the NLI contribution the performance of the E-band, and potentially boosting the
calculation, we use the generalized Gaussian noise (GGN)- capacity of medium-long reach links.
model [23]–[27] that together with the proven conservative
accuracy of the GN-model [20] considers the effects of spectral The article is organized as follows. In Sec. II, we discuss the
and spatial variations of fiber loss and its interaction with NLI potentialities and challenges together with a possible road-map
generation [23]–[27]. As showed by results of this work, these for the development of MBT systems. Sec. III describes the
effects, mainly induced by the SRS, are dominant when the methodology used for the presented analyses. Next, Sec. III
transmission band approaches or exceeds the SRS peak, so presents the considered system configurations and scenarios,
when the transmission band occupancy approaches 10 THz. and highlights the main assumptions for throughput evaluation.
This article extends the work reported in [28], [29], precisely Afterwards, Sec. IV discusses on numerical analyses and
by including SRS effects while evaluating the evolution of presents results as maximum achievable data-rate per channel
signal power along the optical fiber and its interaction with and throughput per band. Finally, Sec. V draws conclusions
fiber non-linearities – the generation of non-linear interference and addresses future investigations.
O-band TX PDFA PDFA O-band OXC PDFA PDFA PDFA O-band RX
MB-DEMUX
MB-DEMUX
MB-DEMUX
MB-MUX
MB-MUX
MB-MUX
E-band TX BDFA BDFA E-band OXC BDFA BDFA BDFA E-band RX
S-band TX TDFA TDFA S-band OXC TDFA TDFA TDFA S-band RX
Fiber EDFA C-band OXC EDFA
Fiber Fiber
C-band TX EDFA EDFA EDFA C-band RX
L-band TX EDFA EDFA L-band OXC EDFA EDFA EDFA L-band RX
MB-Add TX MB-Drop RX
MB-TX MB-ROADM MB inline amp. MB-RX
Fig. 3: Generic multi-band transmission system as considered within the numerical analysis. Transmitter (MB-TX);optical
cross-connect (OXC); Receiver (MB-RX).
II. M ULTI - BAND OPTICAL SYSTEMS : POTENTIALITIES , MBT opens up to novel leverages for traffic and network
CHALLENGES , AND ROAD - MAP management. The broad spectrum enables to transmitting
Fig. 3 illustrates a generic MBT system: the multi-band hundreds of channels with quite different optical performance,
transmitter and receiver (MB-TX, MB-RX) might employ a and this allows the existence of different classes of traffic
comb of tunable lasers, specific for each transmission band or, among them. Another advantage of MBT is that it allows
alternatively, over the entire 54 THz low-loss spectrum. The operators to manage less fibers with respect to SDM. For
multi-band MUX / DEMUX filters are capable of selecting instance, while the O-band could carry short-reach traffic,
any band, while the band amplifiers are supposed to be imple- e.g., ≤ 80 km, long-haul (LH) traffic could be transported
mented by relying on ad-hoc doping glass materials, optimized on the better performing C- or L-bands. In this context, a first
for each band. For instance, we assume Praseodymium [32] assessment on the MBT potentialities, performing an iterative
for O-band; Bismuth [33] for E-band; Thulium [34] for S- power optimization scheme for C-, S- and L-band, has been
band; and Erbium for C- and L-bands. A possible structure proposed in [38], [39]. In [38], a 150 Tb/s capacity has been
for multi-band re-configurable optical add/drop multiplexers shown after 40 km by using S-, C- and L-bands. In this work,
(MB-ROADMs) and inline optical amplifiers is depicted in we show that for similar distances – 50 km, specifically – the
Fig. 3. System advantages may be obtained by using Raman throughput might triple by occupying the spectrum from O-
amplification [35], also in MBT, by sacrificing some spectral to L-band.
portions from data transmission, as a large-enough guard band
between the Raman pumps and channels must be guaranteed. B. Challenges
Alternatively, we could completely remove data transmission
from a band and use it to allocate the Raman pumps only. The main challenge for MBT is the low maturity of the
The latter is considered in ths work, where all O-band spe- key components. While C-band devices have achieved high
cific components are removed from the link and replaced by maturity, MB optical components are still at an early stage3 .
counter-propagating Raman pumps. Finally, Raman amplifiers For example, first prototypes of MB-TX / MB-RX have
could be also realized as discrete components [36]. been presented in [40], and multi-band amplifiers enabled
new transmission records with fiber capacity of 115 Tb/s
A. Potentialities in S-, C- and L-band as shown in [13] or achieving high
MBT is a realistic and practical approach to increase performance in the case of hybrid EDFA+Raman amplifier
the capacity of optical networks in near future, because it for C+L [41]. MBT opens up an enormous spectrum
efficiently uses the available optical fiber infrastructure, thus for transmission, enabling the co-propagation of a large
postponing new fiber roll-outs. Moreover, MBT enables a pay- number of channels. This is a critical aspect for the network
as-you-grow approach also for the in-line network elements, operators due to the considerably different transmission
e.g., filters, amplifiers. For example, some operators are in characteristics of each band, as shown in Fig. 2. Efficiently
the process of upgrading their C-band systems with the L- managing such a wide spectrum can be quite cumbersome,
band [37]. Although a techno-economic analyses is out-of- due to the high dependence of optical performance on
the-scope, we based our work on data received from the the selected wavelength. An effective exploitation of the
operators, which indicate MBT has high potential for capacity available capacity could be achieved by: I) deploying
increase. In order to deliver a simple quantitative projection, bandwidth variable transponders which can adapt their
we may assume {20, 20, 50}Tb/s throughput for C-, L-band, symbol rate/modulation format over the entire spectrum;
all remaining bands, respectively; and 30% of traffic growth. and II) using an advanced control plane embedding an
From this, one can find out that after hypothetical 20 years accurate performance model and a routing, modulation format
growth at 30%, {96, 48, 22} fibers would be necessary in and spectrum assignment algorithm that efficiently exploits
case of relying on C-only, C+L and all bands line systems, the performance variations across the entire 54 THz spectrum.
respectively. Therefore, it is clear that multiple fiber transmis-
sion is inevitable, but MBT is an effective technology to better 3 E.g., silicon photonics integrated circuits show high potential to realize
exploit the infrastructure by enhancing the spatial efficiency in cost-efficient MB-RX: a first discussion on the technology options and state-
bit/s/m2 in ducts. of-the-art of each multi-band component can be found in [31]
Tab. I list the potentialities and challenges of MBT in terms of and noise figures of amplifiers, together with power evolution
pros and cons when compared to transmission solutions such per-frequency in fibers. While, for the NLI calculation, we rely
as MF, MMF, and MCF. on the GGN-model that generalized the GN-model in case of
presence of substantial space and frequency variation of fiber
C. Road-map towards multi-band systems loss – and gain, in case of Raman amplification.
In [44], the GN-model was introduced by applying to optical
Fig. 4 displays a possible road-map towards the adoption transmission using coherent technologies the concept already
of MBT systems by network operators and service providers. presented in the original paper [45]. The GN-model has
Based on the currently available technology and, in particular, subsequently been extensively validated, confirming that it can
given the initial absence of mature MB amplifiers, we foresee be employed for fast but still accurate network performance
that the first application will involve mainly short-reach links. optimization [46]. Moreover, it was shown in [20] that it
The maximum achievable distance of these systems will can be used fro a quick yet accurate and conservative QoT
strictly depend on the channel power budget. At this stage, MB estimation in transmission over optical bandwidths as large
transceivers and fixed-filters will be the only pre-requisite. This as 3 THz. In case of a wider spectrum, the SRS becomes the
first use case is very relevant because it covers a fast growing most limiting effect as it induces a spectral tilt intra- and inter-
market sector, both in terms of demand and DCI applications4 . band towards lower frequencies that can be avoided only with
Once the technologies for MB amplifiers will be sufficiently a guard-band as large as 15 THz, which is approximately the
mature, longer reach links, such as long-distance DCI and SRS bandwidth. Indeed, it has been experimentally shown [47]
metro-aggregation networks, e.g., up to 300 km, may become that the SRS-induced spectral tilt is the main effect in a S+L
a reality. This will require additional CAPEX but, as MBT system.
enables a pay-as-you-grow approach, the increased cost can As the bandwidth for the inter-channel cross-talk including
be diluted over time. At this stage, it will be highly desirable the modulation effect is quite limited, the SRS-induced ampli-
that the amplification sites of the MB amplifiers coincide with tude disturbance is not a considerable impairment, as shown
the existing ones. At the final stage, regional and LH multi- experimentally in [48]. In [49], an extension of the GN-model
band networks may become viable, enabled by mature MB including the Raman amplification and assuming a flat spectral
amplifiers and ROADMs. Raman effect has been proposed, and in [50] design strategies
for optimal hybrid Raman/EDFA amplification have been
III. M ULTI - BAND TRANSMISSION ANALYSIS derived [51]. However, such assumptions are not sufficiently
In line systems using optical coherent technologies, the most general for MBT systems because, besides the variation of
limiting transmission impairments are the joint effect of the gain/loss with the space, also the variations in frequency are
accumulation of ASE noise – introduced by the amplifiers – considerable. Thus, the GN-model has been generalized to
and NLI disturbance generated within fiber propagation. Other the GGN-model [24], [26], [27] to accurately account for
effects, such as accumulated dispersion or polarization mode the frequency and space variations of gain/loss along the
dispersion (PMD) are mostly compensated by the DSP within fiber spans and across the transmission bands as well as the
the receiver. It is well accepted that the quality-of-transmission wavelength dependence of dispersion, so, fully considering the
(QoT) figure of merit for the deployed lightpaths is given SRS effects – cross-talk and amplification. In [52], [53], the
by the GSNR [19], [20] that includes both the effects of the GGN-model has been experimentally validated and, in [54],
accumulated ASE noise and NLI disturbance, defined as it has been proposed to estimate the QoT for point-to-point
PRX −1 and network performances, considering optical bandwidths as
GSNR = = OSNR−1 + SNR−1 NL , (2) large as 10 THz. Although GN-like models are usually quite
PASE + PNLI
accurate, especially in the C-band, their accuracy decreases
where OSNR = PRX /PASE is the optical SNR measurable in the presence of very-low fiber chromatic dispersion param-
on the optical spectrum analyzer, SNRNL = PRX /PNLI is eters. Nevertheless, in the case of ITU-T G.652 fibers, the
the non-linear SNR consider the effect of NLI only, PRX close-to-zero dispersion parameter occurs at high frequencies,
is the power of the channel under test (CUT), PASE is the where the power depletion due to SRS is large, making
power of the ASE noise and PNLI is the power of the NLI negligible with respect to ASE noise. The accuracy for
NLI. The GSNR corresponds to the error vector magnitude GN-like models is also reduced when used for short-reach
(EVM) [43] readable on the DSP-recovered constellation in links, such as single-span systems. However, the estimation is
case of absence of substantial phase-noise. Given the back-to- conservative in this case. Thus, it can still be used to depict a
back transceiver characterization, the GSNR well predicts the general trend on the system capacity.
BER, as extensively shown also in multi-vendor experiments Eq. 1 shows the GGN-model formula estimating the NLI
using commercial products [20]. We suppose the controller to power spectral density GNLI (Ls , f ) generated by a wavelength
set the line operating at the LOGO [30], so the GSNR will be division multiplexed (WDM) comb with a power spectral
dominated by the ASE noise being OSNR = 1 / 2 SNRNL . The density GTX (f ) after transmission along a fiber span with
OSNR calculation is analytic based on the knowledge of gain length Ls , non-linear coefficient γ, dispersion parameter β2
4 Solutions such as WDM 400ZR are now available on the market because
and dispersion slope β3 . The term ρ(z, f ) is the overall fre-
of the high demand. It is worth to mention that the largest number of DCI quency and space dependent fiber loss profile. This parameter
links is well below 80 km [42]. includes the non-flat fiber attenuation and the SRS: both
ZZ +∞
16 2
GNLI (Ls , f ) = γ GTX (f1 )GTX (f2 )GTX (f1 + f2 − f )
27 −∞
Z Ls 2
2 ρ(ζ, f1 )ρ(ζ, f1 + f2 − f )ρ(ζ, f2 )
exp(+j4π (f1 − f )(f2 − f )[β2 + πβ3 (f1 + f2 )]ζ) dζ df1 df2 (1)
0 ρ(ζ, f )
TABLE II: Considered parameters per band.

Parameters / Band 0 E S C L
Wavelength range [nm] 1260 - 1360 1360 - 1460 1460 - 1530 1530 - 1565 1565 - 1625
Frequency range [THz] 220.59 - 238.10 205.48 - 220.59 196.08 - 205.48 191.69 - 196.08 184.62 - 191.69
Bandwidth [THz] 17.25 14.81 9.13 4.13 6.96
Number of channels 240 296 182 82 139
Central frequency [THz] 229.07 212.79 200.65 193.89 188.07
Type of amplifier PDFA [32] BDFA [33] TDFA [34] EDFA EDFA
Amplifier’s noise figure [dB] 7 6 7 5.5 6
Non-linear coefficient γ [1/W/km] 1.6 1.5 1.4 1.3 1.28
TABLE III: Transmitted power per channel per band.

O E S C L
LS = 50 km -4.0 dBm -3.4 dBm -2.7 dBm -3.1 dBm -2.5 dBm
Full MBT
LS = 75 km -0.9 dBm -1.1 dBm -1.0 dBm -1.7 dBm -1.0 dBm
LS = 50 km - -3.4 dBm -2.7 dBm -3.1 dBm -2.5 dBm
O-band off and no Raman amplification
Raman amplification in the O-band
Applications
A. The system under analysis
A fully loaded WDM comb on MBT, ranging from the O-
Metro, Regio &
Full Network
Long-Haul Networks to the L-band is investigated, for a total occupied bandwidth
of ∼53 THz. A 2 nm guard-band is assumed between adjacent
Long Reach DCI bands for filtering purposes5 . A non-transmission bandwidth
Amplified
Point-to-Point of 30 nm (5 THz) centered at the SMF zero-dispersion
Link
wavelength (D(λ) ≈ 0) is set and computed to maximize
Unamplified Short Reach DCI the O-band capacity as described in [28]. Tab. II reports the
Point-to-Point
Link
main parameters per band, showing that a total of ∼ 900
channels can be transmitted in a MBT system when employing
1. MB Transceivers 4. MB Spectrum Enabling a 50 GHz WDM grid. This amount of WDM channels is
3. MB Amplifiers
2. MB Fixed Filters Selective Switches Technologies
∼10 as large as the one of commercial systems, nevertheless
Fig. 4: Road-map towards the implementation of MBT sys- the maximum resulting total power we are considering is
tems. ≤ 1.5 W, for case with transmission over E→L with Raman
pumps. This power level is well below, for example, the
injected power in [56] or of the specification of commercial
amplifiers [57]. On the other hand, it is important to mention
that MBT requires major upgrades in the way optical systems
Raman cross-talk and Raman amplification, if any. Details will be certificated and standardized. In fact, current safety
on the evaluation of ρ(z, f ) are reported in appendix A. In standards, e.g., IEC 60825-2 consider a maximum optical
presence of Raman pumps, the SNR penalty introduced by power up to 500 mW and telecommunication operators are
the Rayleigh back-scattered pump-pump four-wave mixing obliged to respect this normative with their deployed system
(FWM) has been assessed according to [55]. Considering a using Raman cards. To conclude this topic, it is important to
Rayleigh backscattering coefficient (κ) of 10−7 m 1
and conser- mention that new standards will be required and new safety
vatively supposing large linewidth and possible variations in regulations and training for the workers installing the systems,
the pumps’ wavelegths, a 2 THz extra guard-band has been will be needed before a MBT can be deployed. Tab. III
adopted to keep the pump-pump FWM always negligible. So, reports the transmitted power per channel per band for each
in presence of Raman pumps, the number of channels in the considered system configuration and for each span length, at
E-band is reduced from 295 to 252. From Eq. 1, it is possible the input of the fiber span. The power per channel within each
to calculate the final GSNR, at the optical fiber output. Further band has been assumed as spectrally flat set as described in
details on the GSNR computation are reported in appendix B. 5 This is a benchmark value that is not yet feasible with mature filtering
technologies
25 TABLE IV: O-band Raman pump configuration.

O-band E-band S-band C-band L-band
20 Frequency [THz] 222.22 227.27 229.89 232.56 235.29
Power [mW] 200 200 350 350 350
Power level [dBm]
15 240 295 182 81 139

channels channels channels chs. channels
10
-5
-10
1260 1310 1360 1410 1460 1530 1565 1625
Wavelength [nm]
Fig. 6: Equivalent block diagram of a span.
of the next span. At the end of the optical line, the WDM
MBT comb is DEMUX’ed, amplified and, finally, detected.
Amplifiers and filters are kept as simple as possible, because
the aim of the work is to investigate whether the MBT is
feasible given the assumed noise figure values of the amplifiers
as present in the literature. For the target scenarios, from DCI
to regional networks, we considered two values for the span
length Ls : 50 km and 75 km.
B. The optical transmitted power

We use a LOGO strategy [30], which assumes flat per-
band power per channel, with the power level of each band
set accordingly. Each band is evaluated independently of the
others with the nominal values of the transmission band,
Fig. 5: Considered system configurations: a) Full MBT; b) i.e., the ones at the central frequency of each band, used as
O-band off, Raman in O-band, E → L-band. inputs for the power computation. We assume each lumped
amplifier has a constant gain equal to the highest loss within its
corresponding amplification band. In such a way, the lumped
Sec. III-B. In this work, root raised-cosine shaped signals with amplifiers plus GFFs act as MBT optical equalizers. The power
a roll-off factor of 0.15 and a symbol-rate of 32 GBaud are spectral density at the input of each fiber span, when signals
assumed. The modulation format is adaptable – i.e., a bit- are transmitted from O- to L-band, is illustrated in Fig. 5 a).
loading technique is supposed – and is determined by the
GSNR value6 The line system is considered periodic and
C. Throughput maximization: a trade-off between available
composed of Ns spans of equal length Ls and made of ITU-T
bands and physical effects
G.652.D fiber whose fiber loss coefficient and the dispersion
curve vs. wavelength are displayed in Fig. 2. The non-linear 1) Ideal flexible transceiver: We assume ideal flexible
coefficient γ is practically flat within each band, and the values transceivers, capable of completely exploiting the available
are reported in Tab. II. Fig. 6 depicts the optical system under GSNR, thus enabling the maximum feasible bit-rate, as shown
analysis. We assume to use a comb of transmitters specific in Sec. 3.2 of [58]. A total of 12% rate overhead is assumed,
for each transmission band. Then, a MB-MUX combines all and this value takes into account for the FEC overhead and
channels in a unique WDM MBT comb and the resulting DSP pilots.
WDM signal is transmitted through the optical link. Every 2) Transmission scenarios & Raman amplifiers: The MBT
MB-DEMUX introduces 3 dB loss. A MB-DEMUX is used at system is composed of 5 bands. The full MBT is first explored
the end of each fiber span to separate the different bands which (see Fig. 5 a)) and then the O-band off scenario is investigated
are then amplified by different lumped optical amplifiers. Since and exploited for amplification of the E-band with Raman am-
the aim of the study is to provide a benchmark, ideally flat plifiers. In this case, five Raman pumps are set over the entire
amplification, capable to sustaining the required power gain, is O-band as displayed in Fig. 5 b). The pump configurations are
assumed. A gain flattening filter (GFF) is used after each band reported in Tab. IV.
amplifier to restore the launched power profile at the input
IV. N UMERICAL RESULTS
6 Note
that this is only one of the possible pair of modulation formats
and grid. Clearly, in case of other modulation formats and symbol rates, the Fig. 7 shows the signal power at the beginning of the fiber
throughput will scale accordingly, but the trend will remain unchanged. span, i.e., for z = 0 (blue lines), and at the end, i.e. for z = LS
TABLE V: Average spectral efficiency per band [bps/Hz]

O→L No Raman, E→ L Raman in O, E → L
O E S C L E S C L E S C L
DCI 8.7 10.0 10.0 9.9 9.4 10.0 10.3 9.9 9.6 11.2 11.1 10.8 10.6
Metro 6.9 8.2 8.3 8.1 7.7 8.2 8.5 8.1 7.8 9.4 9.3 9.0 8.8
Ls = 50 km Extended Metro 5.8 7.1 7.1 7.0 6.6 7.1 7.4 7.0 6.7 8.3 8.2 7.9 7.7
Regional Network 4.7 6.0 6.0 5.9 5.5 6.0 6.3 5.9 5.6 7.2 7.1 6.8 6.6
DCI 6.8 8.5 8.7 8.5 7.8 8.5 9.0 8.5 7.9 9.3 9.3 8.9 8.7
Metro 5.7 7.4 7.6 7.4 6.7 7.4 7.9 7.4 6.8 8.2 8.1 7.8 7.6
Ls = 75 km Extended Metro 4.6 6.3 6.5 6.3 5.6 6.3 6.7 6.3 5.7 7.1 7.0 6.7 6.5
Regional Network 3.6 5.2 5.4 5.2 4.5 5.2 5.6 5.2 4.6 6.0 5.9 5.6 5.4
(red lines), when all bands are used for data transmission. expected, due to the higher accumulated loss which decreases
As expected the low-wavelength signals experience larger the SNRASE . Meanwhile, the NLI and SRS cross-talk do not
attenuation as the fiber loss is larger and the SRS depletes considerably vary with the fiber length, provided that the
the channel power, on the contrary, high-wavelength channels effective length is exceeded, as in the analyzed scenario. So,
present less attenuation as the attenuation is lower and the the spectral dependence of the maximum throughput remains
signals receive power from the lower wavelengths through the practically unchanged. The discontinuities among transmission
SRS. The fiber attenuation and the SRS induce an overall tilt bands result from the different amplification technologies
of 8 dB after 50 km and of 10 dB after 75 km over the entire assumed for each band. These are summarized by different
bandwidth. noise figures, as reported in Tab. II. As previously stated,
Fig. 8 shows the GSNR (solid line), the SNRASE (dashed line) different power levels are set for each band according to the
and the SNRNL (dash-dotted line) after a 75 km span when all LOGO strategy, so depending on fiber parameters – length,
bands are exploited for data transmission. Most transmission α(λ), D(λ) and γ – and on the amplifier parameters gain and
bands present strong unbalancing between ASE noise and NLI noise figure [46]. The O-band shows the worst performance,
generation, despite the application of the LOGO strategy. This when compared to the remaining transmission bands as a
is because of the power transfer caused by the SRS towards the consequence of the higher fiber attenuation, higher noise
lower frequencies. Thus, an ASE limited transmission regime figure of the amplifiers, and high power depletion induced
is observed at the high-frequencies, especially in the O-band, by inter-channel SRS. Thus, whether to exploit the O-band
while the S-, C- and L-bands are mainly limited by the non- in MBT systems has to be analyzed from a techno-economic
linear Kerr effect. The crossing point between SNRASE and perspective for specific use cases. Another option for the use
SNRNL is ∼1450 nm, i.e., between E- and S-bands. For this of the O-band is to allocate Raman pumps to improving
reason, techniques such as the pre-emphasis of transmitted performance in the remaining highest-frequency band (see
power levels, optimized Raman pump power and wavelength Fig. 5) – the E-band – that is largely depleted by the SRS
selection could be explored to mitigate the impact of SRS [59], power transfer and consequently dominated by ASE noise.
[60]. The L-band has a slight upturn at the end of the band. Tab. VII and Fig. 10 report the maximum capacity per band
This is due to the absence of channels on one side of the and maximum total capacity, respectively, for each transmis-
band which decreases the NLI generated in that portion of the sion scenario. These results show that Raman amplification
band. The SNRASE presents some discontinuities among the enables a limited throughput increase, mainly over longer fiber
transmission bands – one is clearly observable at 1450 nm. spans, because of the large loss in the O-band that limits
Tab. V and Fig. 9 report the average spectral efficiency per the Raman efficiency. So, as for C-band only transmission
band and bitrate per channel, respectively, attainable in each systems, Raman amplification is beneficial only for very-long
of the transmission bands when considering four different fiber spans. An improvement of the maximum throughput due
transmission scenarios with span lengths of 50 km and 75 km: to Raman amplification limited to 10%, was observed in all
single span DCI, 150 km metro, 300 km extended metro and cases.
600 km regional optical networks. Three different transmission Fig. 10 shows that O- and S-band carry a comparable total
configurations are considered for each case: data transmission traffic, despite the higher number of wavelengths in O-band:
over the full MBT spectrum (”O → L”); data transmission 240 versus 182. This result is a consequence of the lower
from the E- to L-band only and without Raman amplification GSNR in the O-band, which provides a smaller bitrate per
(”No Raman, E → L”); and finally data transmission using wavelength. Despite the much worse optical performance, the
the E- to L-bands with Raman pumps placed in the O-band O-band can transport a higher total throughput with respect
(”Raman in O, E → L”). Further details on the computation to C-band, only due to the much wider bandwidth (∼4 times
of the average spectral efficiency per band are shown in wider), so ts actual usability must follow an accurate techno-
appendix C. economics analysis. Tab. VII shows that by increasing the
Results reported in Fig. 9 show that the best performing span length from 50 km to 75 km in the DCI scenario, the
channels are the ones in the middle of the MBT WDM comb, maximum throughput decreases by 75 Tb/s from 450 Tb/s
which correspond to the best balancing between SNRASE and to 375 Tb/s, when the full MBT from O- to L-band is
SNRNL , as reported in Fig. 8. Extending the fiber span leads assumed. This value corresponds to a decrease of the SMF
to a decrease of the maximum throughput per channel, as throughput of about 15%, highlighting the high impact of
(a) (b)
Fig. 7: Signal power at the beginning of the fiber (z = 0) and at the end (z = LS ), for LS =50 km (a) and LS =75 km (b)
for transmission.
Finally, Table VI reports the throughput increase with respect
to the use of full C-band and full C+L-band under the same
conditions. As expected, the throughput gain is always smaller
than the bandwidth enlargements as the GSNR decreases en-
larging the bandwidth. Furthermore, the gain always decreases
when propagation distances become larger and when the span
length is larger. In general a full O → L MBT can carry a
traffic 8 to 10 times larger than a full C-band transmission
and 2.5 to 3 times more than a full C+L-band MBT.
V. C ONCLUSION AND OUTLOOK
The currently exploited C-band is a minor portion of
Fig. 8: Single span GSNR, SNRASE and SNRNL in case all the available single-mode low-loss spectrum available for
bands are used for channel transmission with Ls = 75 km. transmission over ITU-T G.652 single-mode fibers. The vast
majority of optical fibers deployed in the installed network
infrastructures do not present anymore the OH absorption peak
in the E-band – as reported in the cited references – and so are
the DCI link length on performance. A smaller impact of ITU-T G.652.D fibers. Consequently, the entire single mode
the span length on the maximum throughput is observed for spectrum, from the O- to the L-band, could be exploited for
the remaining transmission scenarios, as expected, since the transmission once the technology become available relying on
optical performance / maximum throughput is already limited a fiber loss always smaller than 0.4 dB/km. High-bandwidth
for the reference scenario: span length of 50 km. Additionally, demand and short-distance transmission scenarios, such as
Tab. VII displays that using O-band for data transport can be DCI with distances ≤40 km, in case of absence of available
indeed interesting for the DCI transmission case with 50 km dark fibers, might represent the starting use cases where MBT
span length, as it enables almost 30% of additional throughput. could be first deployed. We carried out an extensive analysis,
Even for the longer fiber length of 75 km, when the O-band is by estimating the generalized SNR, to evaluate the multi-band
used for Raman pumping instead of data transport, the overall feasible throughput of the most widely deployed fiber, the
throughput is reduced roughly by 20%. Therefore, the use ITU-T G.652.D. We aimed at estimating the throughput limits
of the O-band for data transport enables a large additional set by fiber propagation to provide a reference assessment for
throughput in DCI scenarios, but with limited spectral effi- MBT techno-economics including different components’ op-
ciency, at the cost of a very large number of transceivers. tions, so we deliberately made simplifying hypotheses on the
So, the use of the O-band for intense data transport seems components to focus on the fiber transmission limitations. We
to be motivated only in case it is not possible to install considered four relevant scenarios, with transmission distances
new cables and low-cost transponders are available. Further ranging from 50 km up to 600 km, and two different span
techno-economic analysis is clearly required, in particular to lengths, 50 km and 75 km. Given these assumptions, our inves-
investigate if there exist solid use cases to open up the O-band tigations show that multi-band transmission may be a feasible
TABLE VI: Gain factor with respect to full C-band and full C+L.
Gain from C-band Gain from C+L-band
No Raman Raman in O No Raman Raman in O
O→L O→L
E→L E→L E→L E→L
DCI 10× 8× 8× 4× 3× 3×
Metro 10× 7× 8× 4× 3× 7×
Ls = 50 km
Extended Metro 9.5× 7× 8× 4× 3× 3×
Regional 9× 7× 8× 4× 3× 3×
DCI 9× 7× 7× 4× 3× 3×
Metro 9× 7× 7× 3× 2.7× 3×
Ls = 75 km
Extended Metro 9× 7× 7× 3× 3× 3×
Regional 8× 7× 7× 3× 3× 3×
(a) O → L, Ls = 50 km (b) No Raman, E → L, Ls = 50 km (c) Raman in O, E → L, Ls = 50 km
(d) O → L, Ls = 75 km (e) No Raman, E → L, Ls = 75 km (f) Raman in O, E → L, Ls = 75 km

Fig. 9: Net bitrate per wavelength for each scenario when considering two different span lengths. Each row reports different
span lengths – Ls = 50 km the first and Ls = 75 km the second one. Each column refers to different bandwidth uses: full
O→L the first, E→L without Raman amplification the second and E→L with Raman amplification the last one.
TABLE VII: Maximum capacity [Tb/s]. low-loss spectrum of SMF, in case it is needed. To afford the
No Raman Raman in O ever growing traffic demand, MBT will not be sufficient and
O→L
E→L E→L multiple fiber transmission will be a firm request. But, the use
DCI 450 348 359
Metro 367 268 302
of MBT in cables including multiple fibers will enlarge the
Ls = 50 km throughput per fiber, consequently enlarging the transmission
Extended Metro 314 229 265
Regional 263 208 229 spatial efficiency in bit/s/m2 , so enabling a better exploitation
DCI 375 296 298
Metro 323 258 261
of the precious space available in fiber ducts that requires large
Ls = 75 km CAPEX investments.
Extended Metro 272 219 225
Regional 222 181 189 MBT has the potential to guarantee a throughput per
fiber above 450 Tb/s/fiber over DCI distances of ∼ 50 km,
which is approximately a 10× increase compared to the
alternative to widely researched approaches based on multi- best commercial C-band systems [1]. In case of regional /
mode and/or multi-core fibers, because it does not require the long-haul distances, MBT may achieve a throughput well
deployment of new cables. Therefore, we can conclude that the above 200 Tb/s/fiber. In general, a full O → L MBT can
most efficient option, once dark-fibers and space within ducts accommodate ∼8/10× more traffic than the only C-band and
is exhausted, is to install multi-band transmission covering the ∼2.5/3× more than C+L-band.
entire low-loss spectrum of SMF. This process is starting with Further improvements such as interleaving signals and Raman
the enabling of the already commercial available L-band, and pumps can be exploited to further enhance the propagation
we envision it will possibly continue by populating the entire performance as shown in [59], [60].
O⭢L No Raman, E⭢L Raman in O, E⭢L O⭢L No Raman, E⭢L Raman in O, E⭢L
180 150
150 120
120
90
Tbps
Tbps
90
60
60
30 30
0 0
O-Band E-Band S-Band C-Band L-Band O-Band E-Band S-Band C-Band L-Band
(a) DCI (Single Span), Ls = 50 km (b) DCI (Single Span), Ls = 75 km

180 180
150 150
120 120
Tbps
Tbps
90 90
60 60
30 30
0 0
(c) Metro (150 km), Ls = 50 km (d) Metro (150 km), Ls = 75 km

180 180
150 150
120 120
Tbps
Tbps
90 90
60 60
30 30
0 0
(e) Extended Metro (300 km), Ls = 50 km (f) Extended Metro (300 km), Ls = 75 km
180 180
150 150
120 120
Tbps
Tbps
90 90
60 60
30 30
0 0
(g) Regional (600 km), Ls = 50 km (h) Regional (600 km), Ls = 75 km

Fig. 10: Maximum capacity per transmission band. Each row refers to a different reach: DCI (75 km) the first, metro (150
km) the second, extended metro (300 km) the third and regional (600 km) the last one. Left-hand side: 50 km spans; and
right-hand side: 75 km spans.
output. Afterwards, the channels are demultiplexed using a

MBT demultiplexer. Each band is then amplified via lumped
A PPENDIX A amplification and a GFF is applied to the signal. Finally,
ASE noise is added. For each WDM channel, the GSNR is
The effect of SRS is modelled according to [61] and the
computed as:
Rayleigh back-scattering is not taken into account as it has PS
been demonstrated that it is negligible in Silica-Core Fibers GSNR = (5)
PASE + PNLI
(SCF) with pumps having power lower than 800 mW [62].
where, PS is the signal power, PASE is the ASE noise,
The power evolution P (z, fi ) of each signal and Raman
defined as
pump is described by the following system of N ordinary
PASE = h Geq Feq f Bref (6)
differential equations (ODE):
 where h is the Planck constant, f is the frequency, Bref is the


 ± dP (z,f
dz
1)
= − α(f1 )P (z, f1 ) + reference noise bandwidth. The reference bandwidth is equal
PN
j=2 CR (f1 , fn )P (z, fj )P (z, f1 )
to the signal symbol rate RS . Geq and Feq are the equivalent




Raman-EDFA gain and noise figure, respectively. In case

. . .



Raman amplification is not used, Geq and Feq correspond to

dP (z,fi )

± dz = − α(fi )P (z, fi ) +


the gain and the noise figure of lumped amplifiers, respectively.

PN
CR (fi , fn )P (z, fj )P (z, fi ) − Finally, PNLI is
 Pj=i+1
i−1 fi
j=1 fj CR (fi , fn )P (z, fj )P (z, fi )


PNLI = Bref GNLI (Ls , f ), (7)





 ...

 dP (z,fN ) where Ls is the span length and GNLI (Ls , f ) is the NLI power

 ± dz = − α(fi )P (z, fN ) −


 PN −1 fN spectral density defined as reported in equation 1.
 C (f , f )P (z, f )P (z, f )
j=1 fj R i n j N
(3)
Each equation describes the power evolution along the spatial A PPENDIX C
variable z of the signal or the pump at the i-th frequency
The average spectral efficiency per band is the average
fi , with fi < fj for i < j. Thus, f1 is the lower frequency,
performance of the occupied band B and it has been computed
while fN is the larger. N is the number of signal and Raman
as
pumps. The ± sign in each equation indicates co-propagation PNc h P
i=1 Rbnet,i SEn RS
if ”+” or counter-propagation if ”−” of the i-th signal or = n (8)
B Nch ∆f
Raman pump. α(fi ) is the fiber attenuation coefficient at
the frequency fi , CR (fi , fn ) is the Raman gain efficiency where, Rbnet,i is the bitrate of the n-th channel, RS is the
between the frequencies fi and fn . The boundary conditions symbol rate, Nch is the number of channels, ∆f is the
of the problem are given by the transmitted power P (z0 , fi ) WDM grid size and SE is the channel net spectral efficiency
of each signal and Raman pump at the transmission point computed as
2 log2 (1 + SNR)
z0 . z0 is equal to 0 for all co-propagating Raman pumps and SE = , (9)
for all signals (being co-propagating by definition), while z0 1 + OH
is equal to the span length LS for all counter-propagating where OH is the overhead.
Raman pumps. Thus, the problem is described by a system
of coupled ordinary differential equations (ODE) with two R EFERENCES
points boundary conditions (z0 = 0 and z0 = LS ). Since, [1] “Infinera Groove G30,” https://www.infinera.com/
infinera-sets-highest-performance-600g-transmission-record/.
in general, there is not an analytic solution, a numerical [2] “Cisco Visual Networking Index: Forecast and Methodology, June
ODE solver is used to compute P (z, fi ). Finally, the overall 2017,” https://www.cisco.com/c/en/us/solutions/service-provider/
frequency and space dependent fiber loss profile ρ(z, f ) is visual-networking-index-vni/index.html.
[3] G. Wellbrock and T. J. Xia, “How will optical transport deal with future
computed as: network traffic growth?” in 2014 The European Conference on Optical
s Communication (ECOC). IEEE, sep 2014.
P (z, fi ) [4] P. J. Winzer, “Scaling Optical Networking Technologies for Next Gen-
ρ(z, fi ) = (4) eration SDM,” in 2018 Optical Fiber Communications Conference and
P (z0 , fi ) Exposition (OFC), 2018, p. Th1F.4.
[5] T. Jimenez, V. López, F. Jimenez, O. Gonzalez, and J. Fernandez,
“Techno-economic analysis of transmission technologies in low aggrega-
tion rings of metropolitan networks,” in Optical Fiber Communications
A PPENDIX B Conference and Exhibition (OFC), 2017. IEEE, 2017, pp. 1–3.
[6] R. Van Uden, R. A. Correa, E. A. Lopez, F. Huijskens, C. Xia, G. Li,
Each span of the transmission line is modeled according A. Schülzgen, H. De Waardt, A. Koonen, and C. Okonkwo, “Ultra-high-
to the equivalent block diagram reported in Fig. 6. The density spatial division multiplexing with a few-mode multicore fibre,”
Nature Photonics, vol. 8, no. 11, p. 865, 2014.
WDM signal in each band is multiplexed using a multiband [7] T. Kobayashi, M. Nakamura, F. Hamaoka, K. Shibahara, T. Mizuno,
multiplexer and the resulting signal is propagated through the A. Sano, H. Kawakami, A. Isoda, M. Nagatani, H. Yamazaki et al., “1-
fiber which is characterized by a transfer function ρ(LS , f ). Pb/s (32 SDM/46 WDM/768 Gb/s) C-band dense SDM transmission
over 205.6-km of single-mode heterogeneous multi-core fiber using
This transfer function includes both the fiber attenuation and 96-Gbaud PDM-16QAM channels,” in Optical Fiber Communication
the SRS. The impact of NLI is added to the signal at fiber Conference. OSA, 2017, pp. Th5B–1.
[8] A. Napoli, N. Costa, J. K. Fischer, J. Pedro, S. Abrate, N. Calabretta, capacity of installed optical fiber infrastructure via wideband transmis-
W. Forysiak, E. Pincemin, J. P.-P. Gimenez, C. Matrakidis et al., “To- sion,” in 2018 20th International Conference on Transparent Optical
wards multiband optical systems,” in Photonic Networks and Devices. Networks (ICTON). IEEE, 2018, pp. 1–4.
OSA, 2018, pp. NeTu3E–1. [29] A. Ferrari et al., “Upgrade capacity scenarios enabled by multi-band
[9] “Handbook – Optical fibres, cables and systems - ITU, optical systems,” in ICTON 2019, 2019.
2009,” https://www.itu.int/dms pub/itu-t/opb/hdb/T-HDB-OUT. [30] P. Poggiolini, G. Bosco, A. Carena, R. Cigliutti, V. Curri, F. Forghieri,
10-2009-1-PDF-E.pdf. R. Pastorelli, and S. Piciaccia, “The logon strategy for low-complexity
[10] T. Miya, Y. Terunuma, T. Hosaka, and T. Miyashita, “Ultimate low-loss control plane implementation in new-generation flexible networks,” in
single-mode fibre at 1.55 µm,” Electronics Letters, vol. 15, no. 4, pp. Optical Fiber Communication Conference (OFC), 2013. IEEE, 2013,
106–108, 1979. pp. 1–3.
[11] “Prysmian, Broadband boost for Rural New Zealand,” [31] A. Napoli, N. Calabretta, J. K. Fischer, N. Costa, S. Abrate, J. Pedro,
https://www.prysmiangroup.com/staticres/Nexst-2015-2/ V. Lopez, V. Curri, D. Zibar, E. Pincemin et al., “Perspectives of multi-
broadband-boost-for-rural-new-zealand.html. band optical communication systems,” in OECC, 2018.
[12] J.-X. Cai, H. Batshon, M. Mazurczyk, O. Sinkin, D. Wang, M. Paskov, [32] Fiberlabs-Inc, “Praseodymium fluoride fiber glass doped amplifier,”
W. Patterson, C. Davidson, P. Corbett, G. Wolter et al., “70.4 tb/s www.fiberlabs.com. [Online]. Available: www.fiberlabs.com
capacity over 7,600 km in c+ l band using coded modulation with hybrid [33] E. Dianov, “Bismuth-doped optical fibers: a challenging active medium
constellation shaping and nonlinearity compensation,” in Optical Fiber for near-IR lasers and optical amplifiers,” Light: Science & Applications,
Communication Conference. OSA, 2017, pp. Th5B–2. vol. 1, 2012.
[13] J. Renaudier, “100nm ultra-wideband optical fiber transmission systems [34] S. Aozasa et al., “Tm-doped fiber amplifiers for 1470-nm-band WDM
using semiconductor optical amplifiers,” in 2018 European Conference signals,” PTL, vol. 12, no. 10, pp. 1331–1333, 2000.
on Optical Communication (ECOC). IEEE, 2018, pp. 1–3. [35] V. Curri and A. Carena, “Merit of raman pumping in uniform and
[14] “The Importance of International Standards in the Evolution of uncompensated links supporting nywdm transmission,” Journal of Light-
Telecommunications Networks, white paper,” https://www.corning. wave Technology, vol. 34, no. 2, pp. 554–565, 2015.
com/content/dam/corning/media/worldwide/coc/documents/Fiber/RC-% [36] L. Krzczanowicz, M. A. Iqbal, I. Phillips, M. Tan, P. Skvortcov,
20White%20Papers/WP-General/WP5160.pdf. P. Harper, and W. Forysiak, “Low transmission penalty dual-stage
[15] “Increasing data traffic requires full spectral window broadband discrete raman amplifier,” Optics express, vol. 26, no. 6, pp.
usage in optical single-mode fiber cables, white paper,” 7091–7097, 2018.
http://telecoms.com/wp-content/blogs.dir/1/files/2015/02/White-paper [37] “MPB-Communications, technical data sheets,” https://www.
Increasing-Data-Traffic Final.pdf. globenewswire.com/news-release/2018/10/03/1600666/0/en/
[16] “Cru worldwide telecom cables market report,” 2018. Windstream-Deploys-Infinera-C-L-Solution-Sets-Foundation-to-Double-Fiber-Capacit
html.
[17] S. Okamoto, K. Horikoshi, F. Hamaoka, K. Minoguchi, and A. Hirano,
[38] F. Hamaoka, K. Minoguchi, T. Sasai, A. Matsushita, M. Nakamura,
“5-band (o, e, s, c, and l) wdm transmission with wavelength adaptive
S. Okamoto, E. Yamazaki, and Y. Kisaka, “150.3-tb/s ultra-wideband
modulation format allocation,” in ECOC 2016; 42nd European Con-
(s, c, and l bands) single-mode fibre transmission over 40-km using¿
ference on Optical Communication; Proceedings of. VDE, 2016, pp.
519gb/s/a pdm-128qam signals,” in 2018 European Conference on
1–3.
Optical Communication (ECOC). IEEE, 2018, pp. 1–3.
[18] G.-H. Gho, L. Klak, and J. M. Kahn, “Rate-adaptive coding for optical
[39] F. Hamaoka, M. Nakamura, S. Okamoto, K. Minoguchi, T. Sasai,
fiber transmission systems,” Journal of Lightwave Technology, vol. 29,
A. Matsushita, E. Yamazaki, and Y. Kisaka, “Ultra-wideband wdm
no. 2, pp. 222–233, 2011.
transmission in s, c, and l-bands using signal power optimization
[19] V. K. et al., “The subsea fiber as a shannon channel,” in Suboptic, 2019,
scheme,” Journal of Lightwave Technology, 2019.
pp. 1–7.
[40] C. Doerr, L. Chen, T. Nielsen, R. Aroca, L. Chen, M. Banaee, S. Aze-
[20] M. Filer, M. Cantono, A. Ferrari, G. Grammel, G. Galimberti, and mati, G. McBrien, S. Y. Park, J. Geyer et al., “O, e, s, c, and l
V. Curri, “Multi-vendor experimental validation of an open source band silicon photonics coherent modulator/receiver,” in Optical Fiber
qot estimator for optical networks,” Journal of Lightwave Technology, Communications Conference and Exhibition (OFC), 2016. IEEE, 2016,
vol. 36, no. 15, pp. 3073–3082, 2018. pp. 1–3.
[21] A. Ferrari, M. Filer, K. Balasubramanian, Y. Yin, E. Le Rouzic, [41] M. Ionescu et al., “91 nm c+l hybrid distributed raman/erbium-doped
J. Kundrát, G. Grammel, G. Galimberti, and V. Curri, “Gnpy: an open fibre amplifier for high capacity subsea transmission,,” in 2018 European
source application for physical layer aware open optical networks,” Conference on Optical Communication (ECOC), Sep. 2018, pp. 1–3.
Journal of Optical Communications and Networking, vol. 12, no. 6, [42] M. Filer, J. Gaudette, Y. Yin, D. Billor, Z. Bakhtiari, and J. L. Cox,
pp. C31–C40, 2020. “Low-margin optical networking at cloud scale,” IEEE/OSA Journal of
[22] A. Ferrari, M. Filer, K. Balasubramanian, Y. Yin, E. L. Rouzic, Optical Communications and Networking, vol. 11, no. 10, pp. C94–
J. Kundrát, G. Grammel, G. Galimberti, and V. Curri, “Experimental C108, 2019.
validation of an open source quality of transmission estimator for open [43] R. S. et al, “Error vector magnitude as a performance measure for
optical networks,” in Optical Fiber Communication Conference (OFC) advanced modulation formats,” Photonics Technology Letters, vol. 24,
2020. Optical Society of America, 2020, p. W3C.2. [Online]. Available: pp. 61–63, 2012.
http://www.osapublishing.org/abstract.cfm?URI=OFC-2020-W3C.2 [44] A. Carena, V. Curri, G. Bosco, P. Poggiolini, and F. Forghieri, “Modeling
[23] M. Cantono, J. L. Auge, and V. Curri, “Modelling the impact of srs on nli of the impact of nonlinear propagation effects in uncompensated optical
generation in commercial equipment: an experimental investigation,” in coherent transmission links,” J. Lightwave Technol., vol. 30, no. 10, pp.
Optical Fiber Communication Conference. Optical Society of America, 1524–1539, May 2012.
2018, pp. M1D–2. [45] A. Splett, C. Kurtzke, and K. Petermann, “Ultimate transmission ca-
[24] M. Cantono, D. Pilori, A. Ferrari, and V. Curri, “Introducing pacity of amplified optical fiber communication systems taking into
the Generalized GN-model for Nonlinear Interference Generation account fiber nonlinearities,” in 19th European conference on optical
including space/frequency variations of loss/gain,” arXiv preprint communication, ECOC. IEEE, 1993, pp. 41 – 44.
arXiv:1710.02225, 2017. [46] V. Curri, A. Carena, A. Arduino, G. Bosco, P. Poggiolini, A. Nespola,
[25] M. Cantono, D. Pilori, A. Ferrari, C. Catanese, J. Thouras, J.-L. Augé, and F. Forghieri, “Design strategies and merit of system parameters
and V. Curri, “On the interplay of nonlinear interference generation with for uniform uncompensated links supporting nyquist-wdm transmission,”
stimulated raman scattering for qot estimation,” Journal of Lightwave Journal of Lightwave Technology, vol. 33, no. 18, pp. 3921–3932, 2015.
Technology, 2018. [47] K. Minoguchi, S. Okamoto, F. Hamaoka, A. Matsushita, M. Nakamura,
[26] I. Roberts, J. M. Kahn, J. Harley, and D. W. Boertjes, “Channel Power E. Yamazaki, and Y. Kisaka, “Experiments on stimulated raman scatter-
Optimization of WDM Systems Following Gaussian Noise Nonlinearity ing in s-and l-bands 16-qam signals for ultra-wideband coherent wdm
Model in Presence of Stimulated Raman Scattering,” Journal of Light- systems,” in Optical Fiber Communication Conference. Optical Society
wave Technology, vol. 35, no. 23, pp. 5237–5249, 2017. of America, 2018, pp. Th1C–4.
[27] D. Semrau, R. I. Killey, and P. Bayvel, “The gaussian noise model in [48] G. Saavedra, D. Semrau, M. Tan, M. A. Iqbal, D. Elson, L. Galdino,
the presence of inter-channel stimulated raman scattering,” Journal of P. Harper, R. Killey, and P. Bayvel, “Inter-channel stimulated raman scat-
Lightwave Technology, vol. 36, no. 14, pp. 3046–3055, July 2018. tering and its impact in wideband transmission systems,” in 2018 Optical
[28] J. K. Fischer, M. Cantono, V. Curri, R.-P. Braun, N. Costa, J. Pedro, Fiber Communications Conference and Exposition (OFC). IEEE, 2018,
E. Pincemin, P. Doaré, C. Le Bouëtté, and A. Napoli, “Maximizing the pp. 1–3.
[49] V. Curri, A. Carena, P. Poggiolini, G. Bosco, and F. Forghieri, “Extension

and validation of the gn model for non-linear interference to uncompen-
sated links using raman amplification,” Optics express, vol. 21, no. 3,
pp. 3308–3317, 2013.
[50] V. Curri and A. Carena, “Merit of raman pumping in uniform and
uncompensated links supporting nywdm transmission,” Journal of Light-
wave Technology, vol. 34, no. 2, pp. 554–565, 2015.
[51] S. Kado, Y. Emori, S. Namiki, N. Tsukiji, J. Yoshida, and T. Kimura,
“Broadband flat-noise raman amplifier using low-noise bidirectionally
pumping sources,” in Proceedings 27th European Conference on Optical
Communication (Cat. No. 01TH8551), vol. 6. IEEE, 2001, pp. 38–39.
[52] M. Cantono, J. L. Auge, and V. Curri, “Modelling the Impact of
SRS on NLI Generation in Commercial Equipment: an Experimental
Investigation,” in Optical Fiber Communication Conference. OSA,
2018, p. M1D.2.
[53] M. Cantono, D. Pilori, A. Ferrari, C. Catanese, J. Thouras, J. Auge,
and V. Curri, “On the Interplay of Nonlinear Interference Generation
with Stimulated Raman Scattering,” Journal of Lightwave Technology,
vol. 36, no. 15, pp. 3131–3141, 2018.
[54] D. Semrau, E. Sillekens, R. I. Killey, and P. Bayvel, “The isrs gn model,
an efficient tool in modeling ultra-wideband transmission in point-to-
point and network scenarios,” in 2018 European Conference on Optical
Communication (ECOC), Sep. 2018, pp. 1–3.
[55] J.-C. Bouteiller, L. Leng, and C. Headley, “Pump-pump four-wave
mixing in distributed raman amplified systems,” Journal of lightwave
technology, vol. 22, no. 3, pp. 723–732, 2004.
[56] I. Sackey, T. Richter, C. Meuer, R. Elschner, M. Jazayerifar, K. Pe-
termann, and C. Schubert, “Characterization of polarization insensitive
fopa using a polarization-maintaining highly non-linear fiber,” in Optical
Fiber Communication Conference. Optical Society of America, 2013,
pp. OTu2D–2.
[57] “MPB-Communications, technical data sheets,” http:
//mpbcommunications.com/en/site/products/network ready telecom
solutions/2RU-Series/raman pumps/index.html.
[58] P. J. Winzer, D. T. Neilson, and A. R. Chraplyvy, “Fiber-optic trans-
mission and networking: the previous 20 and the next 20 years,” Optics
express, vol. 26, no. 18, pp. 24 190–24 239, 2018.
[59] T. Naito, T. Tanaka, K. Torii, N. Shimojoh, H. Nakamoto, and
M. Suyama, “A broadband distributed raman amplifier for bandwidths
beyond 100 nm,” in Optical Fiber Communication Conference and
Exhibit. IEEE, 2002, pp. 116–117.
[60] A. Gnauck, R. Jopson, and P. Winzer, “Demonstration of counter-
propagating raman pump placed near signal-channel wavelengths,” IEEE
Photonics Technology Letters, vol. 29, no. 1, pp. 154–157, 2017.
[61] J. Bromage, “Raman amplification for fiber communications systems,”
Journal of Lightwave Technology, vol. 22, no. 1, p. 79, 2004.
[62] P. Hansen, L. Eskildsen, J. Stentz, T. Strasser, J. Judkins, J. DeMarco,
R. Pedrazzani, and D. DiGiovanni, “Rayleigh scattering limitations in
distributed raman pre-amplifiers,” IEEE Photonics Technology Letters,
vol. 10, no. 1, pp. 159–161, 1998.

Assessment On The Achievable Throughput of Multi-Band ITU-T G.652.D Fiber Transmission Systems

Uploaded by

Copyright:

Available Formats

Assessment On The Achievable Throughput of Multi-Band ITU-T G.652.D Fiber Transmission Systems

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Assessment On The Achievable Throughput of Multi-Band ITU-T G.652.D Fiber Transmission Systems

Uploaded by

Copyright:

Available Formats

This article has been accepted for publication in a future issue of this journal, but has not been

Assessment on the Achievable Throughput

TABLE I: Comparison of strategies for network capacity expansion.

1) Development of new components for

O-band TX PDFA PDFA O-band OXC PDFA PDFA PDFA O-band RX

TABLE II: Considered parameters per band.

TABLE III: Transmitted power per channel per band.

25 TABLE IV: O-band Raman pump configuration.

15 240 295 182 81 139

B. The optical transmitted power

TABLE V: Average spectral efficiency per band [bps/Hz]

(a) O → L, Ls = 50 km (b) No Raman, E → L, Ls = 50 km (c) Raman in O, E → L, Ls = 50 km

(d) O → L, Ls = 75 km (e) No Raman, E → L, Ls = 75 km (f) Raman in O, E → L, Ls = 75 km

(a) DCI (Single Span), Ls = 50 km (b) DCI (Single Span), Ls = 75 km

(c) Metro (150 km), Ls = 50 km (d) Metro (150 km), Ls = 75 km

(g) Regional (600 km), Ls = 50 km (h) Regional (600 km), Ls = 75 km

output. Afterwards, the channels are demultiplexed using a

[49] V. Curri, A. Carena, P. Poggiolini, G. Bosco, and F. Forghieri, “Extension

You might also like