вход по аккаунту



код для вставкиСкачать
OFC/NFOEC Technical Digest © 2013 OSA
Impact of Pulse Shaping and Transceiver Electrical
Bandwidths on Nonlinear Compensated Transmission
Tomofumi Oyama1, Takeshi Hoshida2, Hisao Nakashima2, Chihiro Ohshima1,
Zhenning Tao3 and Jens C. Rasmussen2
Fujitsu Laboratories Ltd., (2)Fujitsu Limited, 1-1 Kamikodanaka 4-chome, Nakahara-ku, Kawasaki 211-8588, Japan.,
Fujitsu R&D Center, Ocean International Center, No. 56 Dong Si Huan Zhong Rd, Chaoyang District Beijing, 100025, China.
E-mail address:
Abstract: We investigate the impact of pulse shaping by numerical simulation of 127Gbit/s DPQPSK transmission with RZ, NRZ and Nyquist pulse formats. The transmitter analog bandwidth
is found critical for the latter two cases.
OCIS codes: (060.2360) Fiber optics links and subsystems; (060.1660) Coherent communications
1. Introduction
Intra-channel nonlinearity has been considered as one of the major obstacles to extend the transmission reach of
dual-polarization (DP) QPSK system. Non-linear mitigation techniques assisted by digital signal processing are
attracting interest as countermeasures. Several approaches have been proposed including digital pre-distortion at the
transmitter side[1-3], digital back-propagation or Volterra compensation at the receiver side[4-6]. Among them,
non-linear compensation algorithms based on perturbation analysis of Manakov equation[1,6] are the most practical
ways because they enjoy low complexity signal processing. So far, we have demonstrated that the pre-distortion
algorithm based on the perturbation analysis is tolerant to link uncertainties including chromatic dispersion,
polarization mode dispersion and polarization dependent loss in previous works[2,3]. However, the impact of pulse
formats and device imperfections have not been taken into account in the evaluation of the algorithms yet.
In this paper, we investigate the compatibility of perturbation pre-distortion (PPD)[1] and perturbation backpropagation (PBP)[6] with non-return-to-zero (NRZ), return-to-zero (RZ) and Nyquist pulse shaping formats on
127Gbit/s DP-QPSK transmission. In addition, we compare the impact of the transceiver electrical bandwidths for
the two non-linear compensation algorithms.
2. Simulation model
The evaluations were carried out by numerical simulation of 127Gbit/s DP-QPSK 11-channel WDM transmission.
The simulation model is shown in Fig. 1(a).
(a) Setup
(b) Transmitted optical spectra before mux.
Fig. 1. Numerical simulation model for 127Gbit/s DP-QPSK transmission.
DAC: digital-to-analog convertor. ADC: analog-to-digital convertor. LPF: low-pass filter. LD: laser. IQ: IQ modulator.
PBC: polarization beam combiner. ASE: ASE adder. CDC: chromatic dispersion compensator. NLC: nonlinear compensator.
978-1-55752-962-6/13/$31.00 ©2013 Optical Society of America
OFC/NFOEC Technical Digest © 2013 OSA
Fig. 2. Q-factor and transmission penalty as a function of (a) Tx-bandwidth and (b) Rx-bandwidth (OSNR: 15dB).
The transmitter-side (Tx-) DSP has the functions including PPD, root-raised cosine (RRC) filter for Nyquist
pulse shaping and linearizer for linearizing IQ modulation. The roll-off factor of RRC filter of 0.15 is chosen in this
work. The output signal from Tx-DSP is converted to analog signal by 63.5GSa/s digital-to-analog convertor and the
signal is input to IQ modulator through the low-pass filter (LPF). After polarization multiplexing, the DP-QPSK
signal is multiplexed with 10 neighboring signals in 50GHz channel spacing. The neighbors, being also 127Gbit/s
DP-QPSK format, are in the same pulse format as the measured channel. The transmission link consists of 25 spans
of single mode fiber (SMF) with 60km per span and optical amplifiers. ASE adders are placed respectively right
before and after the transmission link and the added ASE power is set equal to each other. After transmission, the
signal is received by the polarization diversity optical frontend and converted to baseband electrical signal. The
baseband signal is fed into 63.5GSa/s analog-to-digital convertor through the LPF. The receiver-side (Rx-) DSP has
functions including chromatic dispersion compensation (CDC), PBP, adaptive equalizer (AEQ) based on constant
modulus algorithm, frequency offset compensation[7] and the Viterbi-and-Viterbi carrier phase recovery[8]. The
number of the PBP stages was set to three and pre- and post-CDC are optimized to maximize the performance of
PBP[6]. The LPFs after DAC and before ADC are set to emulate the transceiver bandwidths. As the LPFs, we
employed 4th order Bessel filter. PPD and PBP are used selectively and the RRC filter and RZ pulse carvers are
activated corresponding to the pulse formats. Optical spectra in each pulse format before mux are depicted in Fig.
1(b), where a transmitter bandwidth of 22.3GHz and no PPD are adopted.
3. Result and discussion
To begin with, we evaluated the properties of each pulse format without nonlinear compensation. Fig. 2 shows Qfactor and transmission penalty as a function of transceiver bandwidths, where one of the Tx- or Rx-bandwidth was
varied from 13GHz to 32GHz while the other bandwidth was fixed to 22.3GHz. The fiber launch power was set to
3dBm/channel and the OSNR was set to 15dB (in 0.1dB noise bandwidth). The Q-factor was calculated by counting
bit errors. The transmission penalty was defined as difference of Q-factor between back-to-back and after
transmission. RZ format showed the highest tolerance to nonlinearity and only RRC shaped signal was degraded due
to the shortage of the Tx-bandwidth. The Rx-bandwidth made no difference to Q-factor. It indicates that the AEQ in
Rx-DSP compensates the shortage of the Rx-bandwidth. It should be noted that the transmission penalty is
independent from transceiver bandwidths with every pulse formats.
(a) RZ
(b) NRZ
(c) RRC
Fig. 3. Q-factor after transmission as a function of Tx-bandwidth (Rx-bandwidth: 22.3GHz)
(a) RZ
(b) NRZ
(c) RRC
Fig. 4. Q-factor after transmission as a function of Rx-bandwidth (Tx-bandwidth: 22.3GHz)
(a) RZ
OFC/NFOEC Technical Digest © 2013 OSA
(b) NRZ
(c) RRC
Fig. 5. Relation between Q-improvement by PPD and Tx- and Rx-bandwidth.
Next, we evaluated the impact of the transceiver bandwidths on performance of PPD and PBP with each pulse
format. The fiber launch power was fixed to 3dBm/channel and the OSNR was set to 15dB in all simulation
conditions. The Tx-bandwidth dependency of the Q-factor after transmission is shown in Fig.3 and the Rxbandwidth dependency is shown in Fig.4. Here we again fixed the bandwidth of the one side to 22.3GHz while the
bandwidth of the other side was being varied. RZ pulse format was confirmed to give the highest Q-factor even in
the NLC assisted cases. In the RZ and RRC format cases, the performances of the both NLC algorithms are found to
be independent from the transceiver bandwidths. In the NRZ format cases, however, the performance of PPD was
declined as the Tx-bandwidth decreased. On the other hand, the performance of PBP was not affected with
transceiver bandwidths in all pulse formats.
Since we found the Tx-bandwidth dependency in the case of the combination with PPD and NRZ format, we
focused on PPD and evaluated the Q-improvement by PPD by varying the transceiver bandwidths for each pulse
formats. Fig.5 shows the dependency of the Q-improvement on the transceiver bandwidths. The Q-improvement was
defined as difference of the Q-factor between with and without PPD in each bandwidth setting. The Q-improvement
for the RZ and RRC format cases is flat and the values are consistently around 0.5dB and 0.7dB, respectively. In the
NRZ format case, however, the Q-improvement range from 0.3dB to 0.8dB with the Tx-bandwidth between 13GHz
and 32GHz while the Q-improvement is independent from the Rx-bandwidth. We think the reason of the Txbandwidth dependency on the performance of PPD as follows; In the perturbation analysis, signal pulse shape is
assumed as Gaussian[1]. So, the analyzed nonlinear noise has some deviation from the actual nonlinear noise,
especially for NRZ format due to the presence of optical power in data transition points. The deviation at the
transition points are pronounced by PPD with narrower Tx-bandwidth. But the effect from transition points is
reduced by RZ pulse carving. The performance of PPD, therefore, depends on Tx-bandwidth with NRZ but not with
RZ pulse format.
We found that the Tx-bandwidth is critical for NRZ and RRC formats. The shortage of the Tx-bandwidth causes
decline of the performance of PPD for NRZ format and degradation of back-to-back property for RRC format.
Therefore, it is desirable to equalize the Tx-bandwidth for NRZ and RRC format that can be realized by means of
4. Conclusion
We have investigated the impact of pulse shaping and transceiver electrical bandwidths on transmission assisted by
nonlinear compensation. We found that RZ format is not affected by the transceiver bandwidth regardless whether
nonlinear compensation is applied or not and can achieve the longest transmission reach. In the case of adopting
RRC format for a higher spectrum efficiency, careful engineering of the Tx-bandwidth becomes critical in order to
avoid degradation of back-to-back performance. The performance of NLC and transmission penalty with RRC
format, however, was found to be independent from transceiver bandwidths.
This work was partly supported by the National Institute of Information and Communications Technology
(NICT), Japan and by "The research and development project for the ultra-high speed and green photonic networks"
of the Ministry of Internal Affairs and Communications, Japan.
[1] Z. Tao et al., JLT vol.29, no. 17 pp.2570-2576, 2011.
[2] L. Dou et al., ECOC’12, Th.1.D.3.
[3] H. Nakashima et al., ECOC’12, We.3.C.5.
[4] E. Ip and J. M. Kahn, JLT, vol. 26, no. 20, pp.3416–3425, 2008.
[5] L. Liu et al., JLT, vol. 30, no. 3, pp. 310–316, 2012.
[6] W. Yan et al., ECOC’11, Tu.3.A.
[7] L. Li et al., OFC’08, OWT4.
[8] A. J. Viterbi and A. M. Viterbi., Transactions on Information
Theory, vol. 29, pp.543-551, 1983.
Без категории
Размер файла
790 Кб
ofc, oth3c, 2013
Пожаловаться на содержимое документа