Fibre Nonlinearities in WDM-Systems with Reduced ... - IEEE Xplore

5 downloads 0 Views 668KB Size Report
Abstract We investigate the dependence of nonlinear penalty on symbol-rate in polarisation- multiplexed coherent DQPSK WDM-transmission systems over an ...
ECOC 2010, 19-23 September, 2010, Torino, Italy

P4.20

Fibre Nonlinearities in WDM-Systems with Reduced ChannelSpacing and Symbol-Rate (1)

(1)

(1)

C. Behrens , R. I. Killey , S.J. Savory , M. Chen

(2)

(1)

and P. Bayvel

(1)

Optical Networks Group, Dept. of Electronic & Electrical Engineering, University College London, UK, [email protected] (2) European Research Centre, Huawei Technologies, Riesstr. 25, Munich, Germany Abstract We investigate the dependence of nonlinear penalty on symbol-rate in polarisationmultiplexed coherent DQPSK WDM-transmission systems over an uncompensated SSMF link with a fixed spectral efficiency of 2 bit/s/Hz. The performance is found to be similar, irrespective of the symbol-rate. Introduction Higher-order modulation formats have attracted 1-3 since they allow increased much interest spectral efficiency and, therefore, increased capacity, avoiding the need to install new fibre. Coherent detection which gives access to the full information of the optical field in both polarisations, has enabled powerful DSP 4 algorithms to compensate for linear and 5,6 nonlinear impairments. However, the current ITU-standard limits spectral efficiency when reducing the spectral width per WDM-channel below 50GHz. This and its implications on the transmission performance which have been subject to ongoing discussion within the 7-8 research community , are addressed here for coherent PDM-DQPSK for the first time in a detailed investigation. In this paper we describe simulations comparing transmission performance of WDM systems when reducing the channel spacing below the ITU grid, while keeping the spectral efficiency constant. Polarisation-multiplexed DQPSK is considered as a convenient format for investigation at symbol-rates ranging from 56Gbd to 7Gbd for transmission over 1040km

Simulation Setup The DQPSK signal is generated by an IQmodulator, modulating the in-phase and quadrature components of the optical field with binary signals (Fig. 1 (a)). The limited transmitter th bandwidth is emulated with a 5 order electrical Bessel filter with a 3dB bandwidth of 0.8 x symbol-rate. We simulated WDM-transmission of 9, 18, 36 and 72 channels, over a total bandwidth of 900 16 GHz. Each channel was modulated with a 2 bit long De-Bruijn-sequence, decorrelating the symbol-patterns in both polarisations by half of the pattern length. Symbol-rates of 56, 28, 14 and 7Gbd have been investigated with a channel spacing of 100, 50, 25 and 12.5GHz, resulting in a fixed spectral efficiency of 2 nd bit/s/Hz. The optical interleavers had a 2 order Gaussian transfer function with a 3dB bandwidth equivalent to the channel spacing. The transmission line consists of 13 x 80km SSMF without any inline dispersion compensation. EDFAs with a noise figure of

(b)

(a)

A/D

PIN

A/D

MZI

PIN

A/D

PIN

A/D

PC

π/2 3dB

MZI

x

Signal Preparation and Resampling

PIN PC

y

CD comp

CD comp

CMA - equalizer

PBS

Demux Differential Precoder

Data

uncompensated standard single mode fibre (SSMF).

Carrier recovery

Decision circuitry

Carrier recovery

Decision circuitry

LO

Fig. 1 (a) Transmitter structure of DQPSK with (b) coherent receiver and subsequent DSP

978-1-4244-8535-2/10/$26.00 ©2010 IEEE

Ch1 Ch2

Ch3 Ch4

P4.20

transmitter- and LO-laser was set to be 200 kHz with a 0 GHz frequency offset between the two. To suppress direct detection terms the LOsignal ratio was set to 24dB. The limited receiver th bandwidth was modelled with 5 order Bessel filters, with a 3dB bandwidth of 0.8 x symbolrate. After resampling the signal to 2 samples/symbol and normalisation, the chromatic dispersion was compensated in the 4 frequency domain. A standard CMA –equaliser was implemented to compensate for PMD and invert the linear impulse response of the 9 channel. The Viterbi & Viterbi algorithm is used to recover the carrier phase of the signal. After decoding, Monte-Carlo error counting was performed to determine the BER.

4.5dB are used to offset the loss of the optical fibre. These EDFAs were set to operate in saturation with a fixed output power of 17dBm with attenuators used to achieve the required power levels. Signal propagation in the fibre is modelled with the symmetrical split-step Fourier method including the effects of chromatic dispersion, dispersion slope, first order polarization mode dispersion, power dependence of the refractive index (Kerr effect) and nonlinear polarization scattering. Tab. 1 summarizes the link parameters used throughout the simulations. Tab. 1: Link parameters α [dB/km] D [ps/km/nm] 2 S [ps/km/nm ] γ [1/W/km] 0.5 PMD coefficient [ps/km ] Span length [km] Number of spans EDFA noise figure [dB] EDFA output power [dBm]

0.2 16 0.06 1.2 0.1 80 13 4.5 17

Results Fig 2 (a) shows the BER as functions of power spectral density (PSD) for single-channel and WDM transmission. The curves for different symbol-rates overlap at low power spectral density (PSD), which is the expected result in the linear part of the plot for signals with a constant spectral efficiency. WDM transmission shows a penalty on the linear part of the curves with respect to singlechannel transmission, which is due to additional coherent crosstalk introduced by the optical Mux- and Demux-filters at the transmitter and receiver.

nd

At the receiver, the signal was filtered with a 2 order Gaussian filter of a 3dB bandwidth corresponding to the channel-spacing and detected with a single ended coherent receiver (see Fig. 1 (b)). The combined line-width of (a)

(b)

BER

1x10

-1

1x10

-3

1x10

-4

-3

-2

single channel WDM

56Gbd - 100GHz 28Gbd - 50GHz 14Gbd - 25GHz 7Gbd - 12.5GHz

maximum PSD @ BER = 3x10

1x10

-13

3x10

-13

2x10

-13

1x10

-5

1x10 -15 1x10

1x10

-14

1x10

power spectral density (W/Hz)

-13

7 14

28

56

symbolrate (Gbd)

Fig. 2: (a) BER as a function of power spectral density for single-channel (open symbols) and WDMtransmission (filled symbols). On the right hand side (b) the maximum transmittable power spectral density at -3 BER=1x10 is plotted.

P4.20

The nonlinear part in the plots of Fig. 2 (a) shows a decreased performance for higher symbol-rates in case of single channel transmission due to a higher impact of intrachannel four wave mixing (IFWM). A larger pulse overlap results in higher distortion, because more symbols contribute to the IFWM 10 process . In the case of WDM-transmission, additional inter-channel cross-phase modulation and FWM is present, further reducing the maximum PSD at a given BER. All symbol-rates show comparable behaviour, a similar finding to that found for on-off-keying over a compensated 8 transmission link described by Hodžić . 7Gbd only performs slightly worse than 56Gbd, 28Gbd and 14Gbd, which might be due to the very small channel spacing, increasing the penalty from crosstalk from the closest WDMneighbours, compared to the case of a larger channel spacing. Fig. 2 (b) shows the maximum transmittable -3 PSD at BER ≤ 3x10 for single channel and WDM transmission. In the case of single channel transmission 7Gbd leads to better transmission performance, allowing to increase the maximum transmittable PSD by a factor of 2.5 compared to 56Gbd. However, in case of WDM transmission, all symbol-rates show comparable performance. Hence, the optimum symbol-rate is determined only by the most cost effective implementation (trade-off between number of transceivers required and the operating speed of the electronics) and the most convenient channel spacing for network routing. Choosing a higher symbol-rate has an additional benefit, since most of the recently proposed nonlinear compensation schemes compensate only for intrachannel nonlinearities, the higher relative contribution of SPM to the overall nonlinear penalty at the higher symbol-rates might increase the efficiency of these schemes. Currently, coherently detected higher order modulation formats, such as D8PSK, D8QAM and D16QAM are under investigation.

Conclusions We have investigated the impact of varying the symbol-rate for single-channel and WDM transmission of PDM-DQPSK for a fixed spectral efficiency of 2 bit/s/Hz. We considered an uncompensated 1040 km SSMF link. In the case of single-channel transmission, the lowest symbol-rate of 7Gbd performs best due to decreased influence of intra-channel four wave mixing. However, in the case of WDM transmission, comparable performance for all symbol-rates was observed. Consequently, the optimum symbol-rate is determined only by the most cost effective implementation (governed by the trade-off between number of transceivers required and the operating speed of the electronics) and the most convenient WDM channel spacing for network routing. Acknowledgements: The work described in this paper was carried out with the support of the BONE-project ("Building the Future Optical Network in Europe”), a Network of Excellence funded by the European Commission through the 7th ICT-Framework Programme, Huawei Technologies and the Royal Society. References 1 P.J. Winzer et al., Proc. ECOC’09 , PD 2.7 (2009). 2 X. Zhou and J. Yu, J. Lightwave Technol. 27, 3641-3653 (2009). 3 A.H. Gnauck et al., Proc. OFC’09 , PDPB8 (2009). 4 S.J. Savory, Optics Express 16, 804-817 (2008) 5 G. Golfarb and G. Li, Optics Express 17, 8815-8821 (2009) 6 K. Kikuchi, Optics Express 16, 889-896 (2008) 7 A. Hodžić et al., Proc. ECOC’03, We4.P.106 (2003). 8 P.J. Winzer et al., J. Lightwave Technol. 28, 547-556 (2010) 9 A. Viterbi and A. Viterbi, Transactions on Information Theory 29, 543-551 (1983) 10 C. Behrens et al., submitted to PTL