MU2 (Contributed Oral)

1:45 PM - 2:00 PM

A New Cycle Slip Compensation Technique for

Ultra High Speed Coherent Optical Communications
Mario A. Castrillon*, Damian A. Morero*t and Mario R. Hueda*
*Laboratorio de Comunicaciones Digitales - Universidad Nacional de Cordoba - CONICET
Av. Velez Sarsfield 1611 - Cordoba (X5016GCA) - Argentina
tClariPhy Argentina S.A. - Humberto Primo 680 - Cordoba (5000) - Argentina.
Email: { acastrillon, dmorero, mhueda } @efn.uncor.edu

Abstract-We propose a novel cycle slip mitigation algorithm

suitable for next generation OTNs. Simulation results of a 100
PLL V&V
Gb/s DP-QPSK optical system show almost no degradation at a
post-FEe BER of 10-6•

Figure 1. Proposed CS compensation scheme.

I. INT RODUCTION

Coherent detection based receivers with electronic disper­

sion compensation (EDC) are being used on next generation & Viterbi (VV) based carrier phase estimator [3]. The low
optical transport networks (OTN) [1]. These receivers allow density parity check (LDPC) code proposed in [8] is used
using phase encoded modulation formats such as phase shift to investigate the effectiveness of the new CS compensa­
keying (PSK) or quadrature PSK (QPSK). tion algorithm. Simulation results of a 100 Gbls DP-QPSK
Carrier phase recovery (CPR) is a key function of coherent optical system will demonstrate that the new pilot symbol­
optical receivers [1], [2]. In these devices, CPR algorithms are based scheme with only 1% of overhead can eliminate the
required to compensate for effects such as laser phase noise performance degradation caused by cycle slips at a post-FEC
and carrier frequency fluctuations [3]. However, since QPSK BER of 10-6.
modulation has rotational symmetry, errors in the carrier phase
estimation may cause cycle slips (CS). After a CS occurs, all II. SYSTEM MODEL
detected symbols are erroneous and they cannot be corrected
Figure 1 depicts the application of the proposed CS detector
by forward error correction (FEC) codes [2]. To combat this
and corrector in combination with a carrier recovery scheme
catastrophic effect, differential modulation is typically used
as proposed in [3]. The latter is built upon a low latency
[2]. In differential modulation schemes the information is
phase locked loop (PLL) stage combined with a traditional
transmitted as the phase difference between two consecutive
VV carrier phase estimator. Let the discrete time PLL input
symbols. Therefore, the effects of a CS do not translate into
signal be modeled as
catastrophic bit errors. While this option provides a solution
to the CS problem, it is prone to introduce signal-to-noise (1)
ratio (SNR) penalty compared to non-differential schemes. For
instance, a penalty of l.2 dB (post soft-FEC) has been reported where Sk is the complex-valued transmitted symbol at the
for differential QPSK (DQPSK) modulation [4]. k-th time instant, ek is the cumulative phase effect due to
To avoid this penalty produced by differential modulation carrier offset, frequency fluctuations, and laser noise phase; nk
formats, the use of pilot symbols to prevent error propaga­ are independent identically-distributed (iid) complex Gaussian
tion has been proposed in previous works [5]-[7]. Although random variables with zero mean and variance No/2 per
they provide better performance than differential modulation, dimension. The PLL output is given by
their implementation in high-speed applications such as 100
Gigabits per second (Gb/s) OTN, is not straightforward. In par­ (2)
ticular, the need for feedback or iterative precessing requires a
where (h is the PLL phase correction, 'l/Jk ek (h is the
= -

careful design of parallel architectures for the implementation

residual phase error, and nk nk e-jih. Finally, the output
= .

of high speed receivers.

of the VV is
In this paper we present a novel pilot-assisted algorithm to
detect and correct errors caused by cycle slips. The proposed (3)
scheme has a low-complexity forward architecture suitable for
high speed parallel implementation. In particular, it can be where {f;k is the error phase estimated by the VV algorithm,
implemented as a simple post-processing stage after a Viterbi and nk = nk e-j,[;k.
.

978-1-4577-0733-9/12/$26.00 ©2012 IEEE 175

 III. PROPOSED CS COMPENSATION ALGORITHM

The proposed CS mitigation algorithm is composed of 3 o posFEC QPSK (ideal)

stages. In the first one, a coarse CS detection based on pilot
symbols is carried out. Then, a fine estimation of the CS
position is done based on the phase �k
estimated by Vv.
Finally, the errors due to the CS are corrected.
2
The transmitted information is divided into consecutive w
III
-0 -3
blocks of L = P + D complex symbols, where the first P 0;
.2
symbols are pilots and the last D symbols information. The
overhead of pilot symbols is defined as rp P/D. These pilot
=

symbols are used to estimate the phase based on the minimum

iL -I
euclidean distance criteria as

<Pi arg min { f lxk' ej<p - dkl2}

k=i L
(4) 7.0
IdB]
=

<p
SNR

where r.p E {a, -7r/2, 7r/2, 7r} is the possible offset produced Figure 2. BER vs SNR with laser linewidth of 500 KHz at Tx and Rx
by a CS, dk is the kth pilot symbol value, and <Pi indicates the
estimated phase offset corresponding to the ith block. When
<Pi #- <PHI, a CS in ith block is detected. V. CONC LUSIONS
Once a CS is detected, its exact position inside the block is In this paper we have introduced a novel algorithm to
estimated. This is done by comparing the VV phase estimation
�k �k-N'
combat CS in optical coherent communication systems. The
and a delayed version of itself, The position of the proposed scheme can be easily combined with traditional car­
CS is estimated as rier phase estimation algorithms in order to virtually eliminate
gi k i arg min
E{ L,. ,(HIlL-I} I (D.�k+1L - D.<Pi) mod (±7r) I (5)
=
the impact of the CS on the performance of QPSK receivers.
2
In particular, the new low complexity approach is able to

where D.�k �k-N - �b D.<Pi <Pi - <PHI. Parameter N is operate at very low OSNR regimes with almost no degradation.
= =

Furthermore, the described CS compensation architecture is

optimized by computer simulations. The idea behind equation
suitable for parallel implementation in high speed receivers.
5 is based on the following observations. A CS caused by
These features make the proposed technique a very attractive
additive noise generates a fast jump of the phase estimated
by the VV in the direction of CS. The phase difference D.�k alternative to be used with powerful FECs as required in next­
generation optical networks.
between the end and the start of the jump largely cancels the
CS jump D.<Pi
estimated by pilot symbols. On the other hand, REFERENCES
a CS caused by a fast change in the laser phase noise (which [ 1] D. Crivelli et aI., "Adaptive digital equalization in the presence of
cannot be tracked by the VV) generates a peak value in the chromatic dispersion, PMD, and phase noise in coherent fiber optic
direction of CS. This peak value partially cancels the CS phase systems," in Global Telecommunications Conference, 2004. GLOBECOM
'04. IEEE, vol. 4, Dec. 2004, pp. 2545-2551.
jump estimated by pilot symbols generating a minimum value [2] M. Taylor, "Phase estimation methods for optical coherent detection using
at the CS position. Finally, after the CS position is estimated, digital signal processing," Lightwave Technology, Journal of, vol. 27,
no. 7, pp. 901-9 14, Apr. 2009.
the CS effect is canceled.
[3] P. Gianni et aI., "A new parallel carrier recovery architecture for intradyne
IV. SIMULATION RESULT S coherent optical receivers in the presence of laser frequency Ouctuations,"
in Global Telecommunications Conference (GLOBECOM 2011), 2011
Numerical results derived from computer simulations are IEEE, Dec. 20 1 1, pp. 1-6.

presented and analyzed. Dual polarization (DP) QPSK modu­ [4] A. Bisplinghoff et al., "Soft decision metrics for differentially encoded
QPSK," in Optical Communication (ECOC), 2011 37th European Con­
lation with a transmission rate of 100 Gb/s is analyzed. We use ference and Exhibition on, Sept. 20 1 1, pp. 1-3.
500 KHz of laser Iinewidth at Tx and Rx in a non-dispersive [5] S. Zhang et aI., "Pilot-assisted decision-aided maximum-likelihood phase
optical channel. The adopted carrier recovery scheme is able to estimation in coherent optical phase-modulated systems with nonlinear
phase noise," Photonics Technology Letters, IEEE, vol. 22, no. 6, pp.
efficiently compensate carrier frequency offset, jitter, and laser 380-382, Mar. 2010.
phase noise [3]. Fig. 2 shows the BER versus the SNR for (i) [6] X. Wu et aI., "Iterative carrier recovery in turbo receivers with distributed
proposed algorithm, (ii) an ideal QPSK (i.e., without CS), and pilots," in Consumer Electronics, Communications and Networks (CEC­
Net), 2011 International Conference on, Apr. 20 1 1, pp. 5024-5026.
(iii) DQPSK modulation. The overhead of pilot symbols is [7] H. Zhang et aI., "Cycle slip mitigation in POLMUX-QPSK modu­
l.01% (this overhead has been considered in the definition lation," in Optical Fiber Communication Conference and Exposition
of the SNR). An interleaver of 20 codewords has been also (OFCINFOEC), 2011 and the National Fiber Optic Engineers Confer­
ence, Mar. 20 1 1, pp. 1-3.
incorporated. In Fig. 2 a gain of 1.16 dB compared to DQPSK [8] D. Morero et al., "Non-concatenated FEC codes for ultra-high speed
at post FEC BER of 10-6 can be observed. Furthermore, it is optical transport networks," in Global Telecommunications Conference
(GLOBECOM 2011), 2011 IEEE, Dec. 20 1 1, pp. 1-5.
important to note that performance is only 0.07 dB from the
ideal QPSK ones.

176