coherent dwdm
TRANSCRIPT
-
8/10/2019 Coherent DWDM
1/7
Invited Papers
Coherent DWDM technology for high speed optical communications
Ross Saunders
Opnext Subsytems Inc., 151 Albright Way, Los Gatos, CA 95032, USA
a r t i c l e i n f o
Article history:
Available online 7 September 2011
Keywords:
DWDM
Coherent
Optical
DSP
QPSK
QAM
a b s t r a c t
The introduction of coherent digital optical transmission enables a new generation of high speed optical
data transport and fiber impairment mitigation. An initial implementation of 40 Gb/s coherent systems
using Dual Polarization Quadrature Phase Shift Keying (DP-QPSK) is already being installed in carrier
networks. New systems running at 100 Gb/s DP-QPSK data rate are in development and early technologylab and field trial phase. Significant investment in the 100 Gb/s ecosystem (optical components, ASICs,
transponders and systems) bodes well for commercial application in 2012 and beyond. Following in
the footsteps of other telecommunications fields such as wireless and DSL, we can expect coherent optical
transmission to evolve from QPSK to higher order modulations schemes such as Mary PSK and/or QAM.
This will be an interesting area of research in coming years and poses significant challenges in terms of
electro-optic, DSP, ADC/DAC design and fiber nonlinearity mitigation to reach practical implementation
ready for real network deployments.
2011 Published by Elsevier Inc.
1. Introduction
To keep pace with the rapidly growing volumes of data network
traffic drivenby the growthof the internet, service providers are al-
ways looking to increase thefiber capacity and wavelength spectral
efficiency in their networks [1]. Typical Dense Wavelength DivisionMultiplexing (DWDM) networks of today employ a 50 GHz channel
spacing, as per the international standard[2]. At 10 Gb/s data rate
spectral efficiency was not a major concern and simple On Off
Keying (OOK) modulation format was adequate for operation on
the 50 GHz DWDM grid. At 40 Gb/s, the spectral width of the signal
is 4x largerfor OOK, yielding a signalspectralwidththatiss toowide
to fit through 50 GHz channel spacing optical filters without induc-
ing excessive penalties. So at 40 Gb/s data rate system and tran-
sponder developers investigated alternate modulations schemes
to enable 40 Gb/s propagation over the same 50 GHz DWDM grid,
such as Phase Shaped Binary Transmission, PSBT [1], Differential
Phase Shift Keying (DPSK) [3] and DP-QPSK [4]. PSBT and DPSK
offered increased spectral efficiency over OOK, whilst still coding
1 bit per symbol. DP-QPSK, on theotherhand,codes 4 bits per sym-bol (in-phase and quadrature phase components of each polariza-
tion tributary). Coding more bits/symbol, enabled by the advent of
digital coherent transmission [5], reduces the spectral width of
the signal (to 1st order proportional to the baud rate). In fact,
DP-QPSK is so spectrally efficient that it canpropagate a higher data
rate of 127 Gb/s through many cascaded 50 GHzoptical filters, such
as Reconfigurable Optical Add/Drop Multiplexers (ROADMs) [6].
This higher 127 Gb/s data rate not only allows payload transport
of 100GE traffic[7], but also OTU4 link management overhead [8]
and 20% overhead, Soft Decision Forward Error Correction
(SD-FEC) [911] for high performance applications. Therefore,
100 Gb/s transmission using DP-QPSK offers the promise of a good
modulation format fit for DWDM networks operating on a 50 GHzgrid[12]. This was observed several years ago and was the reason
why this formatwas adopted by theOptical Internetworking Forum
(OIF) as a recommended modulation format for 100 Gb/s line
systems[13]. This industry forum has helped to focus investment
and multi-source agreements at the optical component and module
level to help foster an ecosystem that should accelerate network
adoption of 100G DP-QPSK transmission.
Looking to the future, as the internet growth continues with
expanding services such as High Definition (HD) video, mobile
broadband and telecommuting, the question is how will optical
transmission technology keep pace? Learning from other telecom-
munications fields such as wireless, satellite, radio and Digital Sub-
scriber Line (DSL) broadband access, we can say that all these
mediums utilize coherent transmission and all increase transmis-sion rates and spectral efficiency by coding more bits per symbol.
For optical fiber technology development we should surely follow
the lead from these other telecommunications industries but we
have some fundamental and unique challenges that makes our life
difficult. Challenges such as:(i) operation at thebleedingedge high-
est electronics speed of >100 Gb/s for the key technologies such as
ADC/DAC/DSP/FEC/RF electronics/electro-optics; (ii) optical signal-
to-noise ratio (OSNR) requirements become tough as Shannons
Limit dictates that as we increase spectral efficiency via higher-or-
der modulation we need more OSNR and (iii) fiber nonlinearity
poses a major obstacle as high density signal constellations such
1068-5200/$ - see front matter 2011 Published by Elsevier Inc.
doi:10.1016/j.yofte.2011.06.016
Fax: +1 613 678 6707.
E-mail address: [email protected]
Optical Fiber Technology 17 (2011) 445451
Contents lists available at ScienceDirect
Optical Fiber Technology
www.elsev ier .com/locate /yofte
-
8/10/2019 Coherent DWDM
2/7
as M-ary Quadrature Amplitude Modulation are very sensitive to
phase errors due to nonlinear phase noise and Cross Phase Modula-
tion (XPM). Although these challenges appear daunting and formi-
dable it would be unwise to bet against theses problems being
solved by human engineering ingenuity given time and money, as
history has proven. This will be an extremely fertile area of optical
communication research over the next decade and beyond.
2. 100 Gb/s DP-QPSK implementation
2.1. Technology
The basic functional block diagrams for an optical coherent
detection modulation scheme, with control of the amplitude of
both in-phase, I and quadrature phase, Q, components of the mod-
ulated signal is shown in Fig. 1. Note that although the data
throughput is 100 Gb/s, extra overhead bytes are required for
64B/66B Ethernet PCS encoding, OTU4 framing overhead, training
sequence and FEC, which in combination adds 28% to the line rate.
This modulation format codes 4 bits per symbol (for the
In-phase, I and Quadrature-phase, Q components of the each polar-
ization multiplexed tributary), yielding a symbol rate of 32 Gbaud.
The transmit side consists of nested Mach Zehnder Modulators(MZM) structures. The coherent receiver requires mixing the
received signal light with a tunable laser local oscillator. Polariza-
tion Beam Splitters (PBS) and optical phase hybrids are included in
the receive structure to provide polarization and phase diversity. A
key advantage is that the Carrier Phase Estimator (CPE), polariza-
tion and I&Q demultiplexing is all achieved in the electronic do-
main using very fast Analog-to-Digital Converter (ADC) and
Digital Signal Processing (DSP). This alleviates the traditional prob-
lem with optical coherent technology in that a highly stable optical
Phase Locked Loop (PLL) is not required in this design.
Thecriticalenablingtechnologyin thisdesign is thedigitalcoher-
entreceiver,as shown in Fig.2. Thedistorted signalcoming from the
four balanced photodiodes is first quantized using quad 6 bit ADCs.
The adaptive equalizer in the DSP then provides the equalization of
CD, PMD, ROADM filtering distortion and unwanted S21 transferfunction imperfections in the Tx/Rx electro-optic drive chains.
Another critical enabling technology is next-generation soft
decision FEC, enabling up to 3 dB higher coding gain than current
state-of-the-art FEC. A FEC algorithm called Low Density Parity
Check (LDPC) is used, with increased overhead [9] and uses soft
decision decoding of the input. This is in contrast to all optical
transport systems today which use hard decision decoding. As the
digital coherent receiver already generates a quantized version of
the analog signal, this quantized signal can then be passed to a soft
decision FEC decoder.Fig. 3shows the LDPC FEC coding gain as a
function of soft decision bit resolution and overhead rate for an ide-
alized case with no implementation penalty. FEC design optimiza-
tion is a tradeoff between decoder complexity/chip size vs
performance in selection of bit resolution and electro-optic band-width/ROADM tolerance vs performance in selection of the over-
head rate.
Another important property of any metro, regional or long haul
technology is the capability to express through multiple cascaded
optical 50 GHz channel spacing ROADMs, without incurring a large
penalty. DP-QPSK offers excellent tolerance to narrowband optical
filtering due to the low baud rate, which reduces the spectral width
of the signal, and the adaptive equalizer in the Rx, which cleans-up
filtering distortion. Results comparing DP-QPSK with alternative
100G modulation formats are shown in Fig. 4. As can be seen,
DP-QPSK modulation format shows excellent tolerance to extre-
mely narrowband optical filtering. Depending on the ROADM band-
width and profile, 100G DP-QPSK is capable of expressing through
up to 10 cascaded ROADMs at 50 GHz channel spacing[6].
For 100G deployments to be cost-effective in Long Haul DWDM
networks, it is critical that the optical reach is sufficiently large that
the requirement for OEO regenerators is limited, ideally non-exis-
tent. The optical reach is maximized using two critical technologies:
(i) coherent detection (buys 2.5 dBQ) and(ii) LDPC soft decision FEC
(buys up to 3 dBQ). Optical reach up to 2000 km with EDFA only,
5000 km with EDFA + Raman and 7000 km for submarine links is
possible using 100G DP-QPSK technology, with deployable margin.
The propagation performance depends on fiber type and whether
in-line DCMs are deployed or not. Nonlinearity is mitigated by not
deploying in-line DCMsand performing all CD compensation within
the Opnext 100G module. Fig. 5 shows propagation results over dif-
ferent fiber types, with and without in-line DCMs. Note that the
optimum fiber type is SMF-28 fiber without in-line dispersion
compensation. This is also the lowest cost fiber plant for thecarrier.
2.2. Photonic and electronic integration
A 100 Gb/s DP-QPSK offers the ultimate in optical performance
and meets all the key marketing requirements dictated by large
CW laser
Tx Block Diagram
Rx Block DiagramBalanced
Photodiode
Iout
MZII
MZIQ
4x32Gb/s
[32Gbaud]inputs
Gray/2
PBS
TE
TM
MZII
MZIQ
/2
Gray
Gray
Gray
Balanced
Photodiode
BalancedPhotodiode
Balanced
Photodiode
4x32Gb/s
[32Gbaud]
outputsADC/DSP
Iin
Local
Oscillator
90 hybrid
(phase/polarizationdiversity)
Fig. 1. A 128 Gb/s DP-QPSK Tx/Rx implementation.
446 R. Saunders/ Optical Fiber Technology 17 (2011) 445451
-
8/10/2019 Coherent DWDM
3/7
carriers. The challenge is that this is a much more complex modu-
lation scheme than previous generations of optical transport
equipment. Both the electronic and photonic complexity has been
increased substantially. This creates a design challenge in terms of
cost, manufacturability, reliability and footprint. The industry
approach to tackling this significant challenge has been to develop
components with a high level of electronic and photonic integra-
tion. On the electronic side, SiGe MUX and CMOS modem chips
with large gate count and integrated system functionality. On the
photonics transmit side, a single optical assembly contains the
nested MZMs, PBS and splitters, as per OIF implementation
agreements [14]. On the photonics receive side, a single optical
assembly houses the PBS, phase hybrids, balanced photodetectors
and Trans-Impedance Amplifiers (TIAs). This integration is shown
in Figs. 6a and b. This high level of integration greatly eases the
manufacturability of the module.
2.3. Dynamic optical networking
Looking to the future it can be envisaged that future optical net-
works will utilize some level of all-optical switching to facilitate
fast restoration and protection of optical circuits. One limitation
of early 40 Gb/s systems is that Chromatic Dispersion (CD) tuningcan take a relatively long time, on the order of seconds or even
minutes. This is true for both direct detection systems that typi-
cally use a thermally tuned optical tunable dispersion compensator
and for 1st generation coherent systems where the CD equalizer
hunts for the optimum CD value of the link.
By designing coherent transmission systems with in-built train-
ing sequences, it is possible for the CD equalizer to know virtually
instantaneously the optimum tap weight coefficients in the equal-
izer to compensate for an unknown link CD, without requiring this
hunting algorithm. This enables very fast (ms) timeframe CD
acquisition and equalization. The Jones matrix equalizer inside
the MODEM can also rapidly adapt to the optimum State of
Polarization (SOP) and DGD compensation of the link. This rapid
Adaptive Equalizer
(n tap Digital FIR filter)A>D converters
Decision
Circuit
+
Adaptive
Algorithm
Input
distorted
data
Recovered
output
data
Error signal
ITE
QTE
ITM
QTM
Fig. 2. Digital coherent receiver.
hard
decision
soft
decision
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0%
FEC Overhead
NECG(
dB) 6 bits
4 bits
3 bits
2 bits
1 bit
Current gen EFEC
3to
4dBgain
Fig. 3. LDPC soft decision FEC performance.
Fig. 4. A 128 Gb/s DP-QPSK optical filtering tolerance.
6
7
8
9
10
11
12
13
-4 -3 -2 -1 0 1 2 3 4
Launch Power (dBm)
Q(
dB)
LEAF (no DCM)
SMF-28 (no DCM)
LEAF (with DCM)
SMF-28 (with DCM)
LDPC
FEC cliff
Min Q
w/ margin
Fig. 5. A 128 Gb/s DP-QPSK propagation over a 1520 km link.
R. Saunders/ Optical Fiber Technology 17 (2011) 445451 447
-
8/10/2019 Coherent DWDM
4/7
MODEM equalization of CD and PMD enables fast optical path res-
toration/protection, a key requirement for dynamic optical net-working architectures of the future.
Another advantage of coherent transmission is that the MODEM
itself provides real-time link parameter performance monitoring.
This allows in-skin monitoring of CD, PMD, SOP and SNR for link
troubleshooting and link quality metrics (e.g. could be used for
pre-emptive switching criteria) without the use of intrusive exter-
nal test equipment.
As next-generation DWDM line systems often advocate color-
less, directionless ROADMs to allow versatile optical wavelength
routing, another advantage of coherent is that a colorless receiver
can be used, such as optical splitters. The correct DWDM wave-
length is then selected by tuning the laser Local Oscillator (LO)
in the receive transponder. This eliminates the use of colored
DWDM demultiplexers (fixed single wavelength per fiber output)
and allows more flexible dynamic optical networkingarchitectures.
3. Mary QAM for Beyond 100Gb/s
3.1. Introduction
A high data rate can be achieved by coding multiple bits/symbol
and using coherent detection and Mary Quadrature Amplitude
Modulation (M-QAM) modulation format. Present day optical
transport speeds are limited by electronics speed, whereas this
technique allows data rates many multiples higher than the elec-
tronics speed. Although the use of QAM is well known in other
industries, such as satellite and wireless communications, it has
not been implemented to date for optical transmission.
A novel advantage of using M-QAM in DSP is that by enabling
programmable modulation (e.g. from QPSK to 256-QAM) the bit
rate transmitted can be traded for optical reach. This technique
will maximize the data rate for any given link length and distortion
properties of the channel. This capability is analogous to rate adap-tive DSL modems that maximize the data rate over local copper
90 deg
Hybrid
Mixer
XIADC
XI
XQADC
XQ
YIADC
YI
YQADC
YQ
I
Q
90 deg
Hybrid
Mixer
I
Q
SIGNAL
LOCAL
OSCILLATOR
DSP
PBS
PS
PBCPSSIGNAL
LASER
RZ PULSE CARVER
(OPTIONAL)
QPSK MODULATOR
QPSK MODULATOR
~CLOCK
MUX
I
Q
MUX
I
Q
Drive = 2V
BW > R
Fig. 6. A 128 Gb/s DP-QPSK photonic and electronic integration. (a) Transmit. (b) Receive.
CW laser
Tx Block Diagram
Rx Block Diagram
Balanced
Photodiode
Iout
Iin
MZII
MZIQQuad-phaseNRZ data
In-phase
NRZ data
/2
Balanced
Photodiode
Local
Oscillator
90 hybrid
(phase diversity)N.B. Needs
polarization
diversity ortracking for Iinand LO mixing
ADC/DSP
DSP/DAC
Baud rate, bn
Baud rate, bn
One polarization tributary shown for simplicity
Fig. 7. M-QAM Tx/Rx implementation.
448 R. Saunders/ Optical Fiber Technology 17 (2011) 445451
-
8/10/2019 Coherent DWDM
5/7
connections from local office to customer premise, using a training
sequence at installation. The increased data throughput without
any significant increase in the number of active electro-optic com-
ponents should yield a reduction in field failure rate and improved
reliability vs todays optical transponder technology.
3.2. Mary QAM Implementation
The transmit/receive block diagrams are shown in Fig. 7. Thisshows the basic functional block diagrams for an optical coherent
detection modulation scheme, with control of the amplitude of
both in-phase, I and quadrature phase, Q, components of the mod-
ulated signal. In this case, the nominal baud rate, bn, is constant,
but the baud rate could also be rate adaptive to offer more contin-
uous rate adaption, rather than the discrete steps from moving
between M-QAM symbol spaces. Note that 2 polarizations can be
utilized (only 1 shown inFig 7) to double the traffic carrying capac-
ity, using a polarization beam combiner and separate quadrature
modulators for each orthogonal polarization state. The Rx consists
of a synchronous coherent detection scheme. The method shown in
Fig. 7 uses a free-running optical Local Oscillator (LO) and feed
forward carrier recovery, polarization demultiplexing, CD/PMD
compensation using analogue to digital conversion followed by
an adaptive digital Finite Impulse Response (FIR) filter.
As an example, the baud rate could be set to 25Gbaud (the de-
sign could also support variable baud rate to maximize capacity for
a given optical channel filter) and both polarizations modulated. To
maximize reach, the transponder could be configured to transmit
DP-BPSK, as shown inFig. 8. This would give the maximum OSNR
sensitivity and maximum launch power possible, therefore maxi-
mizing the distance that can be transmitted between signal 3R
regeneration points in the network. The capacity in this example
would be 50 Gb/s (1 bits/s/Hz). If the specific channel has excess
performance margin, the transponder could reconfigure to
PM-QPSK as shown inFig 8(b). Note that in this case the OSNR sen-
sitivity is the same (I and Q component noise is independent) butthat the launch power would be slightly lower than DP-PSK as it is
more sensitive to nonlinear phase noise (90 between symbol
states for QPSK, whereas 180 for PSK). This doubles the capacity
to 100 Gb/s. In a similar fashion, if the channel has still more mar-
gin (i.e. typically if it operates over a shorter reach) the transpon-
der can re-configure as DP-8QAM (Fig. 8c), DP-16QAM (Fig. 8d),
DP-32QAM (Fig. 8e), DP-64QAM (Fig. 8f), DP-128QAM (Fig. 8g)
and DP-256QAM (Fig. 8h). As can be seen from the constellation
diagrams inFig. 8, each time the M-QAM bits/symbol rate is incre-
mented, the channel carrying capacity increases, at the expense of
an increase in the required OSNR. This is all done in DSP/software.
The OSNR increase is due to the reduction in the minimum Euclid-
ean distance from symbol to symbol.
For denser M-QAM constellations, more SNR is required per
symbol for a given BER. This is shown in Fig. 9.
An analytic estimation of required OSNR sensitivity (assumes
ideal implementation) for different M-QAM bit rates in shown in
Fig. 10.
Fig. 8. Different M-ary QAM constellations, required DAC/ADC bit resolutions.
R. Saunders/ Optical Fiber Technology 17 (2011) 445451 449
-
8/10/2019 Coherent DWDM
6/7
In addition to the challenging OSNR levels required for M-QAM
optical transmission, M-QAM is more sensitive to nonlinear phasenoise and distortion, so fiber nonlinearity provides a major obstacle
to transmission distance. This is an active area for further study.
Reducing nonlinearity by using more distributed optical amplifica-
tion, such as Raman amplification or more frequent Erbium Doped
Fiber Amplifiers (EDFAs) will help to reduce peak power and hence
nonlinear distortion. The key optical component to help reduce the
nonlinearity is indeed theoptical fiber itself. New optical fibers with
reduced attenuation, reduced nonlinear coefficient n2 and higher
effective core area would all help to reduce nonlinearity and enable
higher optical launch powers and hence increased reach. SuchSuper
Large Effective Area (SLA) fibers already utilized in submarine net-
work deployments are already a good step in the right direction
[14,15].
3.3. Advanced coding techniques
The two front battle between diminishing OSNR sensitivity and
increasing tolerance to nonlinear distortion for higher order modu-
lation schemes has some other weapons at its disposal. Inevitably
these techniques come at the expense of increased implementation
complexity so they areboundedby what is practicalto fit into MOD-
EM ASICs andthermal management limitations. CMOS migration to
lower geometries such as 28 nm will help open-up the possibilities
here, both in terms of transistor speed and gate count density.
Some potential areas of future study:
1. Soft Decision Forward Error Correction (SD-FEC). Need further
study here to improve net coding gain and reduce implementa-
tion penalty through more efficient algorithms, more decoder
iterations/bit resolution, code puncturing, etc.
Fig. 10. M-QAM bit rate vs OSNR sensitivity tradeoff.
Fig. 9. M-QAM symbol error rate probability.
450 R. Saunders/ Optical Fiber Technology 17 (2011) 445451
-
8/10/2019 Coherent DWDM
7/7
2. Trellis/Block Coded Modulation (TCM/BCM). With coded modula-
tion additional coded bits can be used to provide redundancy
rather than send extra symbols (e.g. mapping raw 16-QAM into
Trellis-coded 32-QAM, 1 extra bit coding redundancy). This
effectively expands the signal constellation and increases the
minimum Euclidean distance between adjacent symbols, relax-
ing the OSNR sensitivity requirements[16].
3. Optimized constellation geometry. Rectangular QAM constella-
tions as shown inFig. 8are not the most efficient but are easierto realize. A more optimized M-QAM constellation, such as cir-
cular, increases the minimum Euclidean distance and OSNR
sensitivity. In addition, nonlinear phase noise is intensity
dependent and should be considered in the constellation
design. This optimization of the M-QAM constellation will come
at the expense of added complexity, likely higher DAC resolu-
tion and RF drive chain linearity/S-parameter performance.
4. Optimized symbol mapping. The M-QAM symbol mapping should
be carefully designed and optimized holistically combined with
the SD-FEC and TCM/BCM code designs. Symbol mapping diver-
sity minimizes bit errors and optimum combinations of M-QAM
symbol mapping with SD-FEC/TCM design should be sought.
5. Nonlinear compensation/mitigation. Coherent MODEMs can be
designed with some level of nonlinear compensation using
techniques such as digital backpropagation [17]. In addition
the carrier phase estimation filter shape and bandwidth profile
can be optimized to mitigate against effects such as XPM, pos-
sibly in a dynamic or programmable manner. These techniques
should allow increased optical launch power and hence higher
received OSNR but once again at the expense of increased elec-
tronic DSP complexity.
6. Use of multiple carrier Orthogonal Frequency Division Multiplexing
(OFDM).It has been claimed that the use of OFDM can reduce
nonlinearity vs single carrier transmission, at least in certain
applications such as highly periodic dispersion managed sys-
tems [18]. This needs further study to weigh-up the pros and
cons. Whether single carrier or multiple carriers are used in
coherent systems, if spectral efficiency is to increase then
higher-order modulation such as M-QAM is needed in eithercase.
4. Conclusions
The advent of coherent DWDM technology is enabling 100GE
optical transport over backbone optical networks with link engi-
neering rules similar to 10 Gb/s OOK channels. This enables a
10scaling of network/fiber capacity and is possible without any
change in DWDM channel spacing or DWDM common equipment
design. The formation of a 100G DWDM ecosystem in the OIF in the
infancy of this technology has helped focus R&D capital investment
and should act as a catalyst driving early technology adoption by
system vendors and service providers. Standardization by the IEEE
on 100GE and ITU on OTU4 encapsulation has also been critical in
laying the foundation for this technology. Moreover, the collabora-tion between IEEE and ITU on 100GE encapsulation into OTU4
frame format and commonality in such things as the electrical
interface PMD has really helped to focus engineers and minimize
time wasted reinventing the wheel. The stage is now set for
service providers to start certification testing and initial field appli-
cations of 100 Gb/s DWDM wavelengths. As always for new tech-
nology introduction there will be a period of frantic bug-fixing,
ASIC re-spins and hardware design fixes as system vendors and
service providers run thorough verification lab testing and robust-
ness is built into 100G transponder designs. First office field roll-
outs are expected in 2012 timeframe and as volume begins to ramp
optical vendors will already be working on cost-reduced, footprint-
reduced, performance enhanced next-gen 100G solutions to meet
the very high volumes expected in 2013/14 as volumes ramp [19].
Migrating to data rates beyond 100 Gb/s faces some real chal-
lenges in terms of OSNR sensitivity and nonlinearity. Perhaps wewill just utilize more wavelengths and fibers without increasing
spectral efficiency but that method will also run into scaling issues
as fibers run-out and managing too many DWDM overbuilds be-
comes unwieldy for carriers. Coherent transmission certainly
opens-up the capability of moving to higher-order modulation for-
mats and increased spectral efficiency but to meet the optical reach
requirements we may need a fundamental improvement in optical
fiber and/or optical amplification technology. This will be a fertile
area of optical research in coming years as engineers tackle how to
scale optical transport data-carrying capability whilst staying
within the fundamental constraints of Shannons Limit[20].
References
[1] M. Birk et al., Field trial of a 40 Gbit/s PSBT channel upgrade to an installed1700 km 10 Gbit/s system, in: Proc. OFC 2005, Paper OTuH3, Los Angeles, CA.
[2] ITU-T Rec. G.694.1, Spectral Grids for WDM Applications: DWDM FrequencyGrid, 06/2002.
[3] A.H. Gnauck et al., IEEE Photon. Technol. Lett. 15 (3) (2003).[4] C. Laperle et al., Wavelength Division Multiplexing (WDM) and Polarization
Mode Dispersion (PMD) Performance of a Coherent 40 Gbit/s Dual PolarizationQuadrature Phase Shift Keying (DP-QPSK) Transceiver, Paper PDP16, NFOEC2007.
[5] M. Taylor, Coherent detection method using DSP for demodulation of signaland subsequent equalization of propagation impairments, IEEE Photon.Technol. Lett. 16 (2) (2004) 674676.
[6] B. Zhang et al., Penalty Free Transmission of 127 Gb/s Coherent PM-QPSK over1500 km of NDSF with 10 Cascaded 50 GHz ROADMs, Paper NTuC5, NFOEC/OFC 2010.
[7] IEEE P802.3ba, 40 Gb/s and 100 Gb/s Ethernet Task Force. .
[8] ITU-T Rec. G.709, Interfaces for the Optical Transport Network, Edition 3.0, 12/2009.
[9] I. Djordjevic et al., Generalized low-density parity-check codes for opticalcommunication systems, J. Lightw. Technol. 23 (5) (2005) 19391946.
[10] M. Scholten et al., Enhanced FEC for 40G/100G, ECOC 2009, WorkshopPresentation WS1-06.
[11] F. Chang, K. Onohara, T. Mizuochi, Forward error correction for 100G transportnetworks, IEEE Commun. Mag. 48 (3) (2010) S48S55.
[12] M. Birk et al., Field trial of a real-time, single wavelength, coherent 100 Gbit/sPM-QPSK channel upgrade of an installed 1800 km link, in: Proc. OFC/NFOEC,2010, Paper PDPD1.
[13] OIF, 100G Ultra Long Haul DWDM Framework Document. .
[14] Vascade Optical Fiber Data Sheet, Corning Inc. .
[15] UltraWaveOcean Fibers, OFSFitel Inc..
[16] D. Lin, D.J. Costello, Error Control Coding: Fundamentals and Applications,second ed., ISBN:0-13-042672-5, 2004.
[17] E. Ip, J.M. Kahn, Compensation of dispersion and nonlinear impairments usingdigital backpropagation, J. Lightw. Technol. 26 (20) (2008) 34163425.
[18] L. Du, A. Lowery, Fiber Nonlinearity Compensation for CO-OFDM Systems withPeriodic Dispersion Maps, Paper OTu01, OFC/NFOEC 2009.
[19] D. Innis et al., OVUM Market Research: The 40G and 100G Optical Modules,Components, IC, Actives and Passives Forecast: Revenues, Unit Volumes, andASPs, November 29, 2010.
[20] C.E. Shannon, A mathematical theory of communication, Bell Syst. Tech. J. 27(July, October) (1948). 379423, 623656. .
R. Saunders/ Optical Fiber Technology 17 (2011) 445451 451