Abstract
As traffic demands continue to grow,
supporting data rates beyond 100
Gb/s will be required to increase
optical channel capacity and support
higher-rate client interfaces. Video,
cloud, and data center interconnect
applications are driving significant
growth in both metro and long haul
traffic. Internet and over-the-top (OTT)
video is the biggest driver of bandwidth
to consumers, while enterprise cloud
applications, including Software as a
Service (SaaS), Platform as a Service
(PaaS), and Infrastructure as a Service
(IaaS), are delivering a similar impact on
enterprise bandwidth. Underlying both
of these trends is the need to provide
significant amounts of interconnects
bandwidth between data centers
housing cloud and video services.
Advanced modulation formats that
adapt to optimize spectral efficiency
over a range of channel signal-to noise
ratio conditions are required. Channels
are constructed by varying parameters
such as symbol rate, bits per symbol,
number of polarizations, and number
of optical and electrical subcarriers.
Channel capacity can also be increased
using advanced techniques such as
optical time-division multiplexing,
and fibers that support multiple cores
and modes. Many channel designs can
support higher data rates, but there are
trade-offs between complexity, spectral
efficiency, and optical reach.
Introduction
Driven by the escalating bandwidth
requirements of Internet video,
enterprise cloud, and data center
interconnect, service providers
worldwide have been migrating from
10G to 100G and beyond. The evolution
of networks to support higher data
rates is driven by market demand,
standardization activities, and the
availability of next generation optical
transceiver technology. Standardizing
a new data rate requires standards
for optical transmission and framing,
as well as the details of an optical
transceiver implementation. It is
also highly desirable to standardize
a new client rate at the same
time. For the 100 Gb/s data rate,
for example, the IEEE defined the
100 GbE client interface, while the
International Telecommunication
Union — Telecommunication
Sector (ITU-T) provided the optical
transport unit 4 (OTU4) framing, and
the Optical Internetworking Forum
(OIF) standardized the polarization
multiplexed quadrature phase
shift keying (PM-QPSK) transceiver
implementation. The transceiver
implementations use coherent
detection and digital signal processing
which allows amplitude, phase,
and polarization information to
be exploited. The OIF developed
implementation agreements for both
the transmitter and receiver that
define the functionality, interfaces,
and mechanical requirements. This
allows multiple sources for transceiver
components, even though digital signal
processor (DSP) design and algorithms are more commonly proprietary to each
supplier’s design.
The above figure shows the evolution of
optical and Ethernet standards. Optical
transport and Ethernet client standards
prior to 100 Gb/s were defined
separately and did not always interwork
well. For 100 Gb/s, the standardization
activities proceeded in parallel leading
to a cohesive set of standards that
has accelerated introduction of the
technology.
This paper focuses on technical
considerations for optical transport
with channel rates beyond 100 Gb/s.
The network considerations for rates
beyond 100 Gb/s are discussed, and
emerging technologies that will
further improve network capacity and
efficiency are covered.
Optical Technologies for
supporting channel rates 100
Gb/s and Beyond
In a traditional optical transmission
system, the optical spectrum is divided
into a fixed number of channels that
carry traffic using center frequencies
and channel spacings as defined by
the ITU-T G.694.1. There are many
techniques to modulate a signal for
transmission, but these techniques
have become increasingly complex
to support higher channel rates while
maintaining or improving spectral
efficiency. Traditionally, modulation
of an optical signal was accomplished
by turning the laser light on and off
to represent 1 and 0. This modulation
format, called on-off keying (OOK), is
the predominant modulation format
used for optical channel designs with
data rates up to 10 Gb/s. Constellation
charts that show symbol configurations
in the complex plane for several
modulation approaches are shown in
Figure 3.
With the introduction of 100G, the
industry shifted from very simple
modulation techniques (OOK) that
transported a single bit of data, to much
more advanced phase modulation
techniques (DP-QPSK) capable of
encoding and sending multiple bits at
once. Along with coherent receivers,
these more advanced modulation
techniques enable much higher data
rates and improved compensation for
optical impairments such as chromatic
dispersion (CD), polarization mode
dispersion (PMD), and optical loss.
The trade-off with these advanced
modulation techniques is they require
higher Optical to Signal Noise Ratios
(OSNR). OSNR translates directly into
the optical distances that can be
achieved prior to a regeneration node.
In other words, the more sophisticated
and powerful the modulation, the
shorter the optical reach. This tradeoff between modulation technique,
channel size, and OSNR requirements
are at the heart of current 400G
research efforts.
Modulation Schemes for 100G and Beyond
Transmission of optical signals beyond
100 Gb/s by increase of spectral
efficiency is currently of high interest
at research. The major focus is on
multi-level modulation format based
on MQAM (quadrature-amplitude
modulation) and coherent reception
applied at single carrier as well as at
multi-subcarrier modulations formats.
The major target is to maximize their spectral efficiency. With respect to
potential future 400 Gb/s and 1 Tb/s
options, the need of a flexible grid has
been raised.
To achieve bitrates beyond 100 Gb/s on
a single carrier higher level modulation
schemes have to be applied. Recently
QAM scheme together with polarization
multiplexing is utilized to achieve a
channel rate of 200 Gb/s with 16 QAM.
In an M-QAM or 2m QAM signal, m bits
are transmitted in a single time slot or
symbol, where m is an integer value.
Adding polarization multiplexing to
make PM-2m-QAM format, 2 m bits are
transmitted per symbol. A PM-M-QAM
signals can be realized in principle by
parallel arrangements of PM-QPSK
modulators, where the modulators
are driven with binary data signals,
respectively. For example, two parallel
PM-QPSK modulators are required
to form a PM-16QAM modulator. A
more compact and generic approach
is based on the reuse of a PM-QPSK
modulator, for the generation of all PMM-QAM modulation formats, where the
modulators are driven with electrical
multilevel signals. Various constellations
can be applied for PM-QAM modulation
format, e.g. circular QAM symbol
constellations or quadratic constellation
with different sizes as depicted in Table
Networking Considerations
for Supporting Channel Rates
Beyond 100 Gb/s
One reason to develop optical channels
with capacities beyond 100 Gb/s is
to accommodate traffic flows from
switches or routers. Over the next few years, the IEEE will define higher-rate
Ethernet client interfaces that are likely
to be 400GbE and/or 1 TbE. A 400GbE
client could be mapped into a 400 Gb/s
optical channel, but the actual data rate
of the channel will be higher than 400
Gb/s since channel mapping overhead
and forward error correction (FEC)
must be included. The high-rate optical
channel, however, can also be used to
transport existing client data streams,
such as 10 GbE, 40 GbE, and 100GbE.
This multiplexing can be accomplished
electronically by mapping the clients to
containers that are then combined to
form the channel. OTN switching and
aggregation per the G.709 hierarchy is
well suited to this task, but containers
greater than 100Gb/s have not yet been
standardized. It is also possible to map
the clients directly to subcarriers.
Even though reconfigurable optical
add/drop multiplexers (ROADMs)are
an established technology for optical
transport networks, introducing optical
channels with rates higher than 100
Gb/s adds additional considerations
due to the variable bandwidth
requirements of these optical channels.
The ROADM is divided into two
sections: one for the express paths
and the other to support channel
add/drop. To support super-channels, both the express path and the add/
drop structure must support flexible
bandwidth assignment. In the express
path, channels from network directions
that bypass the node are switched
using a wavelength-selective switch
(WSS) to the desired network direction.
In the add/drop structure, channels
originating or terminating at the
node are switched from the network
through the add/drop structure to the
optical transceiver. Since each add/
drop port can support one or more
optical subcarriers, a super channel
can use either a single port or span
across several ports. To support superchannels with variable bandwidth a
ROADM with a colorless, directionless,
and contentionless (CDC) add/drop
structure is preferred. A colorless and
directionless add/drop structure allows
any add/drop port to support a superchannel with configurable wavelength
range that can then be switched to
any degree of the ROADM. Adding
the contentionless function simplifies
operation since there are no restrictions
on port assignments in the add/drop
structure.
The optical subcarriers belonging to
a super channel are required to travel
on the same lightpath with the same
endpoints. This allows the subcarriers
to be spaced closely together since the
typical WSS filter guard band required
for single carriers that are individually
be routed can be eliminated, thus
enhancing spectral efficiency. Note
that packing the optical subcarriers
tightly to form a super-channel may not allow the center frequency of
the subcarrier to be locked to the
traditional ITU-T grid. In principle,
the channel bandwidth should
be selected to maximize spectral
efficiency. However, in practice, there
are restrictions such as the bandwidth
granularity of the WSS, the frequency
stability of the laser, the optical
subcarrier spacing, and the WSS filter
guard band requirements (determined
by the worst case cascade of WSSs) that
establish the actual bandwidth. At the
receiver, coherent frequency selection
can be used to minimize optical
filtering requirements. We should note
that the ITU has reached agreement
on a center frequency granularity
of 6.25 GHz and full slot widths as a
multiple of 12.5 GHz. Furthermore,
any combination of frequency slots is
allowed as long as no two slots overlap.
The frequency stability for both lasers
and flexible grid WSS devices are within
1 GHz, but a channel with a 50 GHz
minimum spacing that supports 25
GHz or 12.5 GHz bandwidth increments
is practical today. In the near future
a 37.5 GHz minimum bandwidth
should be supportable, as WSSs with
higher resolution become available.
The ratio between carrier spacing and
symbol rate can be varied to optimize
spectral efficiency and channel reach
requirements.
To improve fiber capacity, software
configurable transceivers can optimize
channel performance. The transmitter
and receiver can select the channel
modulation format to optimize
the channel transmission rate and
spectral efficiency. OSNR degradation
is normally proportional to the
transmission distance. Higher order
modulation requires higher OSNR at
the receiver to recover the signal, and
is also more sensitive to nonlinear
effects and crosstalk at ROADM
locations. Therefore, in principle,
longer transmission distances tend
to use lower order modulation, while
higher order modulation can be used
for shorter transmission distances.
The software configurable transceiver
simplifies deployment by using the
same hardware configuration to meet
various reach and spectral efficiency
requirements.
Emerging Technologies
New technologies are being developed
to further improve the performance of
high-rate optical channels. Digital signal
processing with coherent detection
has been used to compensate for linear
impairments in fiber, such as chromatic
dispersion and polarization mode
dispersion, but signal processing can
also be applied to improve nonlinear
impairments. Fiber nonlinearity is a
phenomenon that is dependent on
local optical intensity and is therefore
not easily compensated with traditional
linear approaches.
To improve transmission performance
for high-speed channels that use
phase modulated signals, optical
regeneration can be considered as an
approach to replace power hungry
and expensive optical-to-electrical-tooptical (OEO) regeneration. Practical
all-optical regeneration has been
a huge challenge; however, phase
sensitive amplification (PSA) is a
potential approach. Traditional optical
amplifiers, such as Erbium doped
fiber amplifiers (EDFAs), are phase
insensitive. When a phase modulated
signal enters the amplifier, both the inphase component and the quadrature
component will experience the same
amount of amplification. In a phase
sensitive amplifier, which is based on a
parametric amplification process, the
gain depends on the phase relationship
between the signal and the pumps.
The amplification can be tuned to favor
signal phase rather than noise phase by
adjusting the pumps. Therefore, a phase
modulated signal can be regenerated
by amplifying the signal and not
the phase noise in a phase sensitive
amplification process.
As channel rates have increased, optical
component integration and power
consumption have become significant
concerns. Photonic integrated circuits
(PICs) can provide improved optical
component integration, reduced
power consumption, and enhanced
reliability, while reducing overall
equipment cost. The current generation
of equipment is primarily built using
discrete optical components. The goal
of a PIC is to integrate the functions
provided by the individual components
into a photonic circuit thus reducing
the number of interconnections and
power consumption. The challenge
in implementing PIC technology,
however, is that active and passive
components are typically built using
different materials. Silicon is the best
material for passive waveguide related
functions, such as couplers, splitters,
and wavelength multiplexers, while
III-V materials (e.g., gallium arsenide or
indium phosphide) are best for active
component related functions such
as lasers, modulators, and receivers.
The research and development for PIC
technology has focused on the best
approach to seamlessly integrate the
passive and active functions into a
single design.
Conclusion
Selecting the preferred set of channel
parameters is a complex tradeoff between symbol rate, spectral
efficiency, optical reach, design
complexity, and the availability of
technology. Supporting bit rates
beyond 100 Gb/s can be achieved by
extending the technologies of today’s
PM-QPSK transceivers. Moving to
higher symbol rates has traditionally
been the approach to increase the bit
rate, but limitations in electrical and
optical components have made this
increasingly difficult. Today’s 100 Gb/s
transceivers support symbol rates
of 28–32 Gbaud, and the maximum
symbol rate is only improving slowly.
In the near term increasing the
bits per symbol and/or using more
optical carriers is the best approach
to supporting rates beyond 100 Gb/s.
More sophisticated transmitters that
include digital signal processing
and digital-to-analog converters will
support higher order modulation
and filtering of optical carriers to limit
bandwidth.
The channel rate can be doubled by
either implementing 16-QAM or using
two optical subcarriers, but each
approach has different trade-offs.
Moving to 16-QAM is more spectrally
efficient and only requires a single
transceiver, but there is a significant
reach penalty. Creating a super channel
with two optical carriers doubles the
implementation cost and provides
a smaller improvement in spectral efficiency, but with a minimal reduction
in reach.
Over time, many additional techniques
can be implemented to improve
channel capacity, performance, and
cost. Since these techniques are
more speculative, their timeframes
and benefits are still under review.
Algorithms to compensate for
nonlinear transmission impairments
can be implemented using digital signal
processing in the transceiver, although
reducing the algorithm complexity to
achieve a practical implementation
is challenging. Photonic integrated
circuits can be used to implement
arrays of transmitters and receivers
for super-channel applications and
should significantly reduce size,
power consumption, and cost. Optical
regeneration can be implemented
using a phase sensitive amplifier in
place of traditional OEO regeneration,
and fibers with multiple cores and
modes can provide an alternative to
using multiple fiber pairs.
Shweta Chaturvedi
Presales & Solution Team
Telecom Services Business
The necessary tools and techniques are available to implement rates beyond
100 Gb/s, and over time the implementations will be refined based on
continued development of both current and more speculative approaches.