Optical physical-layer network coding

Optical Physical-Layer Network Coding –
another dimension to increase network capacity?
Lian-Kuan Chen, Ming Li, and Soung Chang Liew
Department of Information Engineering, The Chinese University of Hong Kong, Hong Kong SAR China
*email: [email protected]
Abstract: Network coding is a revolutionary technique that can enhance network throughput and
protection. This presentation introduces optical physical-layer network coding (OPNC), focusing
on “common-channel” OPNC that can fully utilize network resources.
OCIS codes: (060.2330) Fiber optics communications; (060.4259) Networks, packet-switched
Record-breaking demonstrations of high capacity optical transport networks have taken the center stage of optical
communications conferences for many years. The many-fold increase in capacity was enabled mainly by two
approaches. The first is to increase the number of channels per multiplexing dimension and/or the number of
multiplexing dimensions, e.g. TDM, WDM, and polarization-division multiplexing. Recently space-division
multiplexing, using multicore fibers and few-mode fibers [1], and orbital angular momentum multiplexing [2] have
been demonstrated. The second is to increase the spectral efficiency by, for instance, advanced modulation formats
[3], super channel [4], and elastic optical networks [5]. By now it seems that almost all possible dimensions in
multiplexing have been exploited and we are approaching the fiber “capacity crunch”, caused by nonlinear optical
effects, Shannon limits, fiber fuse and optical amplification/filter bandwidth [6][7]. Innovative technologies are
needed to further improve network efficiency and functionality. Network coding (NC) has been proposed, explored
and developed for various applications, including improving network efficiency [8] and security [9]. Especially, NC
implemented in the physical layer can boost network throughput significantly [10]. The first deployment of NC
concept in optical domain was demonstrated in [11], and not surprisingly, it was achieved by using optical XOR
gates that mimic the electronic counter parts to achieve the NC functionality. Later optical physical-layer network
coding (OPNC) was proposed and demonstrated with various features, including boosting network efficiency [12],
reducing resources needed for network protection [13], and increasing system throughput [14]. Recently
common-channel OPNC (CC-OPNC) was demonstrated [15], wherein the constituent signal components embedded
in the network-coded signal are carried on a common channel (i.e., they occupy the same signal space and cannot be
separated by conventional means of demultiplexing). This paper explains how the extra capacity is extracted by
OPNC and how it may help remove or bypass some of the aforementioned road blocks that lead to “capacity
Fig. 1 illustrates one application scenario of OPNC. Nodes A and B want to
broadcast packets N1 and N2 to nodes C and D, respectively. Node A transmits
signals containing packet N1 on both links A-E and A-C. Likewise, node B
transmits signals containing packet N2 on links B-E and B-D. Conventionally,
signals N1 and N2 will need to occupy two separate channels on link E-F.
N 1′
N=O{N1, N2} N2′
However, with OPNC, node E will network-code the signals N1 and N2, e.g. by
an XOR gate or passive signal combiner, into a new signal N, and forward
signal N to nodes C and D. At node C, packet N1 can be obtained from the
received signal N1′ on link A-C. Signal N1′ differs from signal N1 because they
undergo different transmission conditions in the paths A-C and A-E-F-C. Then
node C can recover packet N2 from signal N with the help of N1′. Similarly, at Fig. 1.An illustration of a network coding
node D, N1 can be recovered from signal N and the information from B-D link. scenario
In this way a channel in link E-F can be saved.
A. Optical PNC based on all-optical logic
Optical PNC can be realized by all-optical XOR logic gate. It has been demonstrated for NRZ-OOK signals based
on XPM and XGM in SOA [11], as well as for DPSK signals based on FWM in SOA [16]. The symbols of the
signals N1 and N2 need to be precisely aligned in the time domain. For network decoding, take network decoding N 2
from N and N1′ for example, the bits of N2 can be recovered by logical XOR operation on the bits of N and N 1′. With
reference to Fig. 1, the network-coded signal after the XOR gate has the same power level as the regular signal and
yet two bit streams are embedded within the network-coded signal at the same time. This increases network capacity
without increasing signal power, alleviating nonlinear optical effects and the fiber fuse problem.
B. Optical PNC based on wavelength division multiplexing
For optical PNC, the exact symbol-level alignment between N1 and N2 can be relaxed if the network decoding is
realized by waveform subtraction instead of logical XOR operation. An optical PNC scheme based on combing the
NRZ-OOK signals N1 and N2 with a coupler was proposed in [12]. The center wavelengths of N1 and N2 need to be
sufficiently different, to preserve the linear relationship between the network-coded signal N and the signals N1 and
N2 in direct detection. Network decoding N2 from N and N1′ can be realized by subtracting the waveform of N1 from
that of N in the analog domain.
C. Optical PNC based on polarization multiplexing
Another way to preserve the linear relationship in direct detection is to combine N 1 and N2 through polarization
multiplexing [13][14]. When the center wavelengths of N1 and N2 are exactly the same, the orthogonality between
the signal components N1 and N2 in the network-coded signal N is maintained during transmission. After direct
detection the electrical waveform of N is the addition of those of N1 and N2. The corresponding network decoding
has been successfully demonstrated for NRZ-OOK signals [13] and OFDM signals [14].
D. Coherent common-channel optical PNC
For optical PNC schemes based on multiplexing (i.e., methods B and C above), the signal components of N1 and
N2 in the network-coded signal N can also be obtained by demultiplexing. To fully utilize network resources,
CC-OPNC for PM-DQPSK signals was proposed [15]. Each of N1 and N2 exploits polarization-multiplexing for its
own signal (i.e., each of them uses both polarizations). Furthermore, both N1 and N2 are carried on the same
wavelength. After network coding, the signal components of N1 and N2 in N cannot be separated by means of
demultiplexing. The linear relationship is preserved by coherent detection.
For CC-OPNC, a major challenge in network decoding N2 from N and N1′ (which carries the same information as
N1) is estimating the carrier phase of the signal component N1 embedded in N. The estimation difficulty arises from
that, within N, the signal component N1 is interfered by the other signal component N2. The property that equalized
and polarization-demultiplexed DQPSK signal has constant amplitude at sampling time can be utilized to
circumvent the difficulty [15]. In particular, the carrier phase of the signal component of N1 can be estimated as the
phase that minimizes the moving sum of the squared error in the network-decoded signal [15].
We have presented the recent development of OPNC and its applications. We posit that OPNC is a promising
scheme to fully utilize network resources and to circumvent the optical power limit that hinders further increase in
network capacity. In particular, CC-OPNC is shown to reduce the number of channels used as two signals are in the
same signal space. There are still many challenges ahead for OPNC, including 1) how to incorporate more than two
signals on a common channel; 2) how to circumvent the higher signal processing complexity when higher-order
modulations are adopted; and 3) how to exploit OPNC in more sophisticated multi-channel systems. This project is
supported in part by the University Grants Committee, Hong Kong SAR, under Project AoE/E-02/08.
D. J. Richardson et al., Nature Photonics, no. 7, pp. 354, 2013.
N. Bozinovic et al., ECOC 2012, paper Th.3.C.6.
P. J. Winzer et al., ECOC 2008, paper Th.3.E.5.
D. Hillerkuss et al., Nature Photonics, no. 5, pp. 364, 2011.
M. Jinno et al., IEEE Communications Magazine, vol. 47, pp. 66, 2009.
R.-J. Essiambre et al., JLT, vol.28, pp. 662, 2010.
T. Morioka et al., NTT Technical Review, vol. 9, no. 8, 2011.
S.-Y. R. Li et al., IEEE Trans. Inf. Theory, vol. 49, pp. 371, 2003.
N. Cai et al., IEEE Int. Symp. Information Theory 2002, pp. 323.
S. C. Liew et al., Elsevier Physical Communication, vol. 6, pp. 4, 2013.
T. Kono et al., 17th Microoptics Conference (MOC’11), 2011, paper H-17.
Q. Wang et al., IEEE PTL, vol. 24, pp. 2166, 2012.
Z. X. Liu et al., IEEE PTL, vol. 24, pp. 1424, 2012.
Z. X. Liu et al., ECOC 2013, paper Mo.3.F.3.
M. Li et al., IEEE PTL, vol. 26, pp. 1340, 2014.
Y. An et al., OFC 2013, paper JW2A.60.