On the Mitigation of Optical Filtering Penalties Originating from ROADM Cascade

Abstract

Wavelength selective switches (WSSs) that are integrated in reconfigurable optical add-drop multiplexers (ROADMs) induce penalties on the optical signal due to tight optical filtering, which increases as several ROADMs are cascaded in a meshed network. In this letter, we propose and analyze possible configurations for the mitigation of these penalties in optical domain using optical wave shaper (WS). Including one WS in every ROADM node allows transmission of 28 and 32 GBd signals, which are QPSK, 8-QAM, or 16-QAM modulated, through a cascade of 32 and 14 WSS filters, respectively. With an average bandwidth of 33 GHz per WSS, an optical signal to noise ratio penalty below 1 dB at ${rm BER}=1times 10^{-3}$ is observed.

Full text

154 IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 26, NO. 2, JANUARY 15, 2014
On the Mitigation of Optical Filtering Penalties
Originating from ROADM Cascade
Talha Rahman, Student Member, IEEE, Antonio Napoli, Danish Rafique, Bernhard Spinnler, Maxim Kuschnerov,
Iveth Lobato, Benoit Clouet, Marc Bohn, Chigo Okonkwo, Member, IEEE, and Huug de Waardt, Member, IEEE
Abstract— Wavelength selective switches (WSSs) that are
integrated in reconfigurable optical add-drop multiplexers
(ROADMs) induce penalties on the optical signal due to tight
optical filtering, which increases as several ROADMs are cas-
caded in a meshed network. In this letter, we propose and analyze
possible configurations for the mitigation of these penalties in
optical domain using optical wave shaper (WS). Including one
WS in every ROADM node allows transmission of 28 and 32 GBd
signals, which are QPSK, 8-QAM, or 16-QAM modulated,
through a cascade of 32 and 14 WSS filters, respectively. With
an average bandwidth of 33 GHz per WSS, an optical signal to
noise ratio penalty below 1 dB at BER =1×10−3is observed.
Index Terms—Optical communication, optical equalizers,
optical filters, optical pulse shaping.
I. INTRODUCTION
THE INCREASING data-rate requirements resulting from
the growing use of Internet applications including cloud-
computing, video streaming and VoIP has been driving the
research in flexible optical network systems [1]. Using the
available optical spectrum efficiently and hence reducing the
cost per transmitted bit has been a major focus of research
in this field. Today’s commercial optical transmission systems
employ both phase and polarization diversity and allow trans-
mission of 100 Gb/s. In order to further increase the spectral
efficiency, higher order modulation formats are being actively
investigated [2]–[5]. Furthermore, super-channel formed with
multiple sub-carriers are now being considered to increase the
bit rate to 400 Gb/s and 1 Tb/s [6]–[8].
In addition to the increased data-rate requirements, the
traffic demand has become more dynamic. The current WDM
infrastructure can support channels with fixed 50 GHz spac-
ing only. Interest is growing to adapt the optical network
infrastructure to cope with the dynamic nature of these
demands [9]. For the future grooming of the optical net-
works with data-rates ranging from 40 Gb/s up to 1 Tb/s on
Manuscript received September 5, 2013; revised November 4, 2013;
accepted November 8, 2013. Date of publication November 12, 2013; date of
current version December 26, 2013. This work was supported by the European
Union FP-7 IDEALIST Project under Grant 317999.
T. Rahman, C. Okonkwo, and H. de Waardt are with the Department
of Electrical Engineering, Eindhoven University of Technology, Eindhoven
5600 MB, The Netherlands (e-mail: t.rahman@tue.nl; cokonkwo@tue.nl;
h.d.waardt@tue.nl).
A. Napoli, D. Rafique, B. Spinnler, M. Kuschnerov, I. Lobato,
B. Clouet, and M. Bohn are with Coriant GmbH, Munich 81541,
Germany (e-mail: antonio.napoli@coriant.com; danish.rafique@coriant.com;
bernhard.spinnler@coriant.com; maxim.kuschnerov@coriant.com; iveth.
lobato@coriant.com; benoit.clouet@coriant.com; marc.bohn@coriant.com).
Color versions of one or more of the figures in this letter are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/LPT.2013.2290745
Fig. 1. A simple switch and select ROADM node.
the same fiber, on-demand spectrum allocation is proposed.
This concept is named flex-grid and flex-rate optical transmis-
sion, and it is considered as the best candidate to cope with
estimated bandwidth demand in the next decade [10].
Reconfigurable Optical Add Drop Multiplexers (ROADMs)
are the key elements of a large number of deployed optical
networks. The properties of dynamic reconfigurability and
channel add/drop in the optical domain make them suitable
candidates for the evolving meshed optical networks. Several
implementations of ROADM are possible using optical devices
including MUX/DMUX, optical splitters/combiners, wave-
length blockers and Wavelength Selective Switches (WSSs).
The use of WSS provides the advantages of colorless add/drop
ports and simpler higher degree design which make them
feasible for the next generation ROADMs [11]. Fig. 1 shows a
switch and select ROADM node design using WSS for channel
add/drop and Erbium Doped Fiber Amplifiers (EDFAs) to
recover fiber and filtering losses.
Current flex-grid WSSs are capable of configuring the
bandwidth of each channel with a resolution of 12.5 GHz or
less and directing each of them to one of the N available output
ports. Due to the tight optical filtering of the WSSs inside the
ROADM, each node induces filtering penalty on the optical
signal. In long-haul and meshed regional transmission links,
the optical signal may pass through a cascade of ROADMs
before being detected. Due to the ROADM cascade, the net
3-dB bandwidth keeps decreasing and consequently a filtering
penalty arises, setting a limit on the maximum number of
ROADMs a signal can pass before regeneration is required.
In this letter we propose three different methods for the
mitigation of filtering penalty using optical pulse shaping.
In Section II, the measurement setup for the characterization
of the transfer function of the WSS is briefly illustrated.
Section III explains the simulation setup including the
evaluation of filtering penalty. In section IV different methods
for the mitigation of filtering penalty in the optical domain
are presented and their respective performance is analyzed
and discussed in section V. Finally, in section VI we conclude
our analysis.
1041-1135 © 2013 IEEE

RAHMAN et al.: MITIGATION OF OPTICAL FILTERING PENALTIES ORIGINATING FROM ROADM CASCADE 155
TAB L E I
CHARACTERISTICS OF TWO INVESTI GATED WSS DEVICES
Fig. 2. Simulation setup.
II. EXPERIMENTAL SETUP
In order to determine the actual transfer function of
the WSSs, an experiment was set up which included several
WSSs as device under test and a vector analyzer. In our
measurements we have used commercially available WSSs;
these are bidirectional devices, having four ports and support
flex-grid. The measured devices can support up to 96 WDM
channels and the bandwidth of each channel can be adjusted in
multiples of 12.5 GHz. Each of the channels can be directed
to any of the available ports through software configuration.
For our measurements, we have used a commercial vector
analyzer by Luna Technologies Inc., which works on the prin-
ciple of Mach-Zehnder interferometer. The device under test
(DUT) is placed in one of the arms of the interferometer which
allows the determination of amplitude and phase changes
induced by the DUT for two orthogonal polarization states,
giving a full birefringent description of the device. The output
is given in the form of elements of Jones matrices versus
frequency and different quantities including insertion loss,
group delay, group velocity dispersion are herein calculated.
The vector analyzer scans the whole C-band with a res-
olution of 320 KHz in frequency and gives insertion loss
with an accuracy of ±0.1 dB. The phase values are measured
with a precision of ±0.05 radians. The transfer function of
different C-band channels of WSS are determined by the com-
plex eigenvalues of the Jones matrices. In total 384 transfer
functions were saved to be used in the simulation. The charac-
teristics of different investigated WSS devices are summarized
in table I. Although the table reports the average bandwidths
of the WSS devices, in our analysis we used the individual
transfer functions.
III. SIMULATION SETUP
Fig. 2 shows the overall simulation setup. The baud-rate
of transmitter is set to 28 Gbaud and root raised cosine
(RRC) pulse shape with a roll-off factor of 0.3 are considered.
Using a roll-off factor lower than this value will reduce the
eye opening due to timing jitter and a value above that will
increase the signal bandwidth unnecessarily. The RRC-shaped
pulses are either QPSK, 8-QAM or 16-QAM modulated before
transmission over the channel. The channel consists of a
cascade of distributed noise sources and WSS filters with
average BW =33 GHz. The transfer function of each WSS
Fig. 3. Evolution of transfer function due to WSS cascade. (magnitude =left
axis, phase =right axis).
Fig. 4. OSNR penalty due to tight optical filtering (left axis) for
three different modulation formats, net 3-dB bandwidth (right axis).
Baud-rate =28 GBd.
is randomly selected only once from a set of 384 saved
measurement files. As a result of WSS filter cascade, the net
3-dB bandwidth decreases and the passband-ripple becomes
significant.
Fig. 3 shows the effects of WSS cascade on the net
bandwidth and shape of the passband after 1, 5, 15 and 30
WSSs. After 5 WSS filters only, the net 3-dB bandwidth is
decreased to about 26 GHz from an initial value of 33.3 GHz.
Moreover the passband-ripple increases 1.5 dB after 30 WSSs.
As a result of bandwidth narrowing, outer spectral
components of the transmitted signal are severely attenuated
resulting in significant intersymbol interference (ISI). This
leads to a higher required Optical Signal to Noise Ratio
(OSNR) for a target Bit Error Ratio (BER). The receiver
employs data-aided digital signal processing algorithms [12]
where training sequence are transmitted to approximate the
channel transfer function and partially compensates for the
tight optical filtering effects. For practical purposes we have
considered DAC and ADC to be Gaussian filters with a
bandwidth of 16.8 GHz and effective number of bits are
set to 6. Fig. 4 shows decrease in the net bandwidth due to

156 IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 26, NO. 2, JANUARY 15, 2014
Fig. 5. Proposed WS positions.
WSS cascade as well as the increase in OSNR penalty at
BER =10−3for QPSK, 8-QAM and 16-QAM. The OSNR
penalty for all three modulation formats is similar up to
a cascade of 15 WSSs but afterwards QPSK proves to be
the most robust due to larger Euclidean distance between
constellation points as compared to 8-QAM and 16-QAM.
IV. MITIGATION OF OPTICAL FILTERING PENALTY
The OSNR penalty due to tight optical filtering of a WSS
cascade can be compensated to some extent by using digital
pre-distortion and equalization techniques. But the achievable
gain is limited by dynamic range of DAC at the transmitter
and noise amplification at higher frequencies by the equalizer
at the receiver. To avoid these limitations, spectral shaping
in optical domain is a promising solution [4], [13]. Wave-
shaper (WS) devices are nowadays available with a frequency
resolution 1 GHz and magnitude resolution of 0.1dB.Due
to its potential benefits, we have investigated the use of WS
in three possible configurations. Fig. 5 shows the simulation
diagram with the possible locations where the WS can be
placed: at the beginning of the link, after every ROADM and
at the end of the link.
The shape of the WS filter is determined by the net transfer
function of the channel and the desired channel response. For
the current investigation, we have assumed the channel transfer
function is the effective WSS filter and desired response is the
5th order Gaussian filter with a bandwidth of 37.5 GHz. As
the accumulated phase remains flat even after the cascade of
30 WSSs (Fig. 3), we verified that a negligible degradation is
induced by it. Hence, in the calculation of WS filter, phase is
omitted in the final simulations. The WS filter is calculated
applying the following equation:
HWS(f)=Hdes(f)
Hch(f)+c0(1)
Where HWS(f)is the calculated WS function, Hdes(f)is
the desired channel response, Hch (f)is the current channel
response and c0is a constant. When the WS is applied at
the beginning and end of the link, equation 1 is used twice
only; once at the beginning and once at the end. However, in
case of WS applied after every 2 WSSs, equation 1 is utilized
after every 2 WSSs to determine the HWS(f).Fig.6shows
the comparison of Hch (f),Hdes(f)and HWS(f)for a single
scenario where Hch(f)is a cascade of 22 WSSs. The value
of c0can be tuned to adjust the maximum attenuation within
the passband. For practical purposes, in our investigation we
have limited the maximum passband attenuation to 3 dB.
Fig. 6. Comparison of Hch (f)(cascade of 22 WSSs), Hdes(f)and HWS(f)
transfer functions.
V. RESULTS AND DISCUSSION
In this section we discuss the performance improvements
for different configurations of WS for QPSK, 8-QAM and
16-QAM. Fig. 7 shows the performance curves for the differ-
ent positions of WS in the link namely: (I) at the beginning,
(II) at the beginning and the end and (III) after every 2 WSS
(∼after every ROADM). The simulations were performed for
a maximum of 32 WSS passes (i.e. equivalent to 16 nodes).
The signal is transmitted at a line-rate of 28 Gbaud, the
WSS filters have a average bandwidth of 33 GHz and the
performance are shown in terms of OSNR penalty versus the
number of passed WSSs. Regardless of the position of WS
and the number of WSSs in the link we chose the constant in
equation 1 so that the maximum passband attenuation of the
WS is 3 dB.
At the beginning of the link, the signal has maximum
OSNR. When a WS is placed at the beginning of the link,
a slight improvement can already be noted for the case of all
the investigated modulation formats.
Essentially, a WS reshapes the spectrum to reduce ISI
induced by tight filtering of WSS. Placing a WS only at
the end enhances the noise in addition to recovering the
signal. Consequently, if the improvement in ISI achieved
by wave-shaping is overridden by degradation due to noise
enhancement, a net performance degradation is observed as
in case of placing WS both at start and the end for QPSK
(Fig. 7a) compared to WS only at the start. In case of 8-QAM
and 16-QAM the ISI is dominant and the use of an additional
WS at the end brings benefit, which can be observed in Fig. 7b
and Fig. 7c after 30 and 16 WSSs respectively; comparing to
the case with WS only at the start.
Finally, placing a WS after every two WSSs (i.e. after
every ROADM) leads to the best performance improve-
ment and enables 32 WSS passes with a maximum penalty
below 1 dB for all the investigated modulation formats.
Selecting a different set of WSS transfer functions, the
OSNR penalty without the WS varies in the range of
±0.5 dB but with WS placed after every 2 WSSs, the
OSNR penalty after 32 WSSs still remains below 1 dB.
Based on our analysis, we conclude that the integration of
a WS inside a ROADM node would allow optical coher-
ent transmission over a meshed European network, which

RAHMAN et al.: MITIGATION OF OPTICAL FILTERING PENALTIES ORIGINATING FROM ROADM CASCADE 157
Fig. 7. OSNR penalty for QPSK, 8-QAM and 16-QAM. Line-rate =28 GBd,
WSS Avg. BW =33 GHz. (a) Performance curves for QPSK. (b) Performance
curves for 8-QAM. (c) Performance curves for 16-QAM.
typically consists of up to 16 ROADMs, without significant
OSNR penalty. A similar performance trend was observed
for the transmission of signal with line-rate of 32 GBd and
WSS bandwidth of 33 GHz. This analysis is summarized in
Table II.
TAB L E I I
PERFORMANCE SUMMARY AT 32 GBd: NUMBER OF WSS PASSES AT
1dBOSNRPENALTY FOR A TARGET BER =1×10−3ACHIEVED
w/o WS, WS AT T H E STA RT,WSAT START +END AND
WS AFTER EVERY 2WSSS
VI. CONCLUSION
In this letter we discussed possible optical wave-shaping
solutions for the mitigation of filtering penalties originating
from cascade of WSS. Placement of WS at the beginning of
the link utilizes the advantage of highest available OSNR and
gives some improvements for all the investigated modulation
formats. By placing an additional WS before the receiver, the
performance of the QPSK signals degrade because the impact
of ISI mitigation is not as significant as the OSNR degradation
resulting from the WS. Placing one WS after every ROADM
brings the best performance improvement and enables 28 GBd
QPSK, 8-QAM and 16-QAM modulated signals to pass
through a cascade of 32 WSS filters of bandwidth 33 GHz
with OSNR penalty below 1-dB. Keeping the bandwidth of
WSSs same and increasing the line-rate to 32 GBd, a cascade
of 14 WSS can be passed with OSNR penalty of 1-dB. The
current results show huge performance gains and suggest to
integrate optical wave-shaping inside flex-grid ROADMs to
efficiently mitigate the inherent optical filtering penalty.
REFERENCES
[1] (2013, May) The Zettabyte Era–Trends and Analysis [Online]. Available:
http://www.cisco.com/
[2] H. Buelow, T. Rahman, F. Buchali, W. Idler, and W. Kuebart, “Transmis-
sion of 4-D modulation formats at 28-Gbaud,” in Proc. OFC/NFOEC,
2013, pp. 1–3.
[3] S. Yamanaka, et al.,“11×171 Gb/s PDM 16-QAM transmission over
1440 km with a spectral efficiency of 6.4 b/s/Hz using high-speed DAC,”
in Proc. ECOC, 2010, pp. 5–10.
[4] X. Zhou, et al., “PDM-Nyquist-32 QAM for 450-Gb/s per-channel
WDM transmission on the 50 GHz ITU-T grid,” J. Lightw. Technol.,
vol. 30, no. 4, pp. 553–559, Feb. 15, 2012.
[5] D. Rafique and A. D. Ellis, “Nonlinear and ROADM induced penalties
in 28 Gbaud dynamic optical mesh networks employing electronic signal
processing,” Opt. Express, vol. 19, no. 18, pp. 16739–16748, Aug. 2011.
[6] V. Sleiffer, et al., “Transmission of 448-Gb/s dual-carrier POLMUX-
16QAM over 1230 km with 5 flexi-grid ROADM passes,” in Proc.
OFC/NFOEC, Mar. 2012, pp. 1–3, paper OW4C.3.
[7] F. Buchali, K. Schuh, L. Schmalen, W. Idler, E. Lach, and A. Leven, “1-
Tbit/s dual-carrier DP 64QAM transmission at 64Gbaud with 40% over-
head soft-FEC over 320km SSMF,” in Proc. OFC/NFOEC, Mar. 2013,
pp. 1–8, paper OTh4E.3.
[8] X. Liu and S. Chandrasekhar, “Beyond 1-Tb/s superchannel transmis-
sion,” in Proc. IEEE PHO, Oct. 2011, pp. 893–894.
[9] X. Zhang, et al., “Bridge-and-roll demonstration in GRIPhoN (globally
reconfigurable intelligent photonic network),” in Proc. OFC/NFOEC,
2010, pp. 1–3.
[10] O. Gerstel, M. Jinno, A. Lord, and S. J. B. Yoo, “Elastic optical
networking: A new dawn for the optical layer?” IEEE Commun. Mag.,
vol. 50, no. 2, pp. 12–20, Feb. 2012.
[11] R. Shankar, M. Florja´nczyk, T. J. Hall, A. Vukovic, and H. Hua, “Multi-
degree ROADM based on wavelength selective switches: Architectures
and scalability,” Opt. Commun., vol. 279, no. 1, pp. 94–100, 2007.
[12] M. Kuschnerov, et al., “Data-aided versus blind single-carrier coherent
receivers,” IEEE Photon. J., vol. 2, no. 3, pp. 387–403, Jun. 2010.
[13] X. Zhou and L. Nelson, “400G WDM transmission on the 50 GHz
grid for future optical networks,” J. Lightw. Technol., vol. 30, no. 24,
pp. 3779–3792, Dec. 15, 2012.