Towards 1TbE using Coherent WDM A.D.Ellis, F.C.G.Gunning, B.Cuenot2, T.C.Healy3 E.Pincemin4 Tyndall National Institute and University College Cork Department of Physics, Cork, Ireland Phone: +353 21 4904858 Email:
[email protected] (2now @ JDSU, St Etienne, France, 3Now @ Intune Networks, Dublin, Ireland, 4Orange Labs, Lannion Cedex, France) Abstract – In this paper we report the transmission properties of a 0.3 Tbit/s Coherent WDM signal and confirm the scalability of this signal to 0.6 Tbit/s using polarisation division multiplexing.
channels, which lies within the filter pass band. This high frequency content corresponds, in the time domain, to the transitions in their respective eye diagrams. Thus, the crosstalk is mainly confined in time to the vicinity of these transitions. In CoWDM the time delay of the data patterns is adjusted to align these transitions with the eye crossing of the channel of interest, whilst the relative optical phases of the channels are optimised to minimize the impact of the crosstalk at eye centre. This latter condition can be satisfied only if the channel spacing is precisely equal to a multiple of the bit rate, such that the crosstalk interference occurs periodically in exact synchronism with the data symbols [3].
1. Introduction The design of cost effective optical networks is increasingly constrained by the trade-off between increased impact of transmission impairments to achieve higher capacity and the corresponding number of laser sources required to achieve a given capacity. However, capacity demands continue to increase, with significant current research effort directed towards the transport of future 100 GbE signals for cost sensitive short haul and long haul links, and predicted demands for 1 TbE (Terabit Ethernet) [1]. Multi-carrier techniques [2] using a single laser source promise high aggregate capacities per source, with impairments dictated by the symbol rate of each subcarrier [5,6]. Such techniques employ either direct optical multiplexing (Coherent (CoWDM) [3]), or electronic multiplexing incorporating a cyclic prefix (Orthogonal Frequency Division Multiplexing [4]). In this paper, we consider the performance characteristics of CoWDM. We present numerical simulations confirming that non-linear effects and polarisation mode dispersion (PMD) are limited only to the same extent as a single carrier system of the same symbol rate (which is a small fraction of the overall data rate). These simulations are experimentally confirmed for a 0.3 Tbit/s CoWDM system. Finally, we experimentally demonstrate the compatibility of CoWDM with polarisation multiplexing, demonstrating a total capacity of 0.6Tbit/s with an information spectral density (ISD) approaching 2 bit/s/Hz.
3. Numerical Simulations The numerical simulations for CoWDM were carried out using VPI Transmission Maker v.7.0 with the configuration as shown in fig 1. The comb generator was based on two concatenated Mach-Zehnder modulators, driven with a sine wave of frequency 42.6GHz [10]. As a result, seven comb lines were generated at the precise spacing of 42.6GHz. This comb is launched into the data encoding block, where each line was demultiplexed and independently encoded with seven different 512 bit DBBS sequences, corresponding to seven sub-channels. Time delays and optical phase delays were optimised prior to transmission taking into account the phase delays of the receiver filters. Passive combination of the seven modulated sub-channels gave a total capacity of 0.3 Tbit/s , with an ISD of 1bit/s/Hz. The transmission block consisted of one span of 40km of standard singlemode fibre (SSMF with 17ps/nm/km dispersion and a slope of 0.08ps/nm²/km), perfectly compensated by slope-matched dispersion compensating fibre (DCF). Erbium-doped fibre amplifiers (EDFA) with noise figure of 4.5dB were employed, as illustrated in figure 1. The total power launched into the DCF was maintained at 0dBm. For 600km simulations, the transmission block was repeated 15 times. Each individual sub-channel was then measured after optical demultiplexing, using an asymmetric Mach-Zehnder dis-interleaver (85.2GHzFSR) (AMZI), and band-pass filtering
2. Operating Principle In order to understand the principle of CoWDM, consider a demultiplexing filter that predominantly passes the channel of interest, while substantially rejecting adjacent channels. Residual crosstalk will arise from the high frequency content of the two adjacent Comb Generator
Phase Preserving Data Encoding Mod.1
MZ
AWG 42.6 GHz
MZ
phase delays
Mod.7 42.6 GHz 42.6 Gbit/s sequences
optical delay
passive combiner
DFB
Receiver
Transmission
optical delay
40 km SSMF
ATT
FSR 85.2 GHz
DCF x N
ATT
0.64 nm
PD
CR Error Detector
ATT
0 ps/nm residual dispersion per span
data 1
Fig. 1. Schematic diagram of numerical simulations
PD
CR
19
-5
Q factor (in dB)
-10 17
-15
16
-20
15
-25 -30
14
-35 13
Backscattered power (dBm)
0
18
-40
12
-45 -10
-8
-6 -4 -2 0 2 4 Channel power injected in SSMF (in dBm)
and 0dBm respectively. Comparing the performance of CoWDM and the conventional WDM system (fig 3.) revealed a fixed 1.3dB difference due to the residual crosstalk of the CoWDM system. This penalty was conserved for cumulated Q factor probabilities lower than 0.1 implying that the impact of PMD was identical for the two systems [6]. In both cases, the inclusion of inclusion of the effects of the nonlinear refractive index resulted in an additional penalty of approximately 0.4dB, suggesting that the impact of nonlinearity for this particular case was almost the same for both systems. Finally, we observed that the PMD impact was considerably larger for a single NRZ channel at 0.3Tbit/s when compared to CoWDM, even in the case that the total PMD was halved. 1 Cumulated Probability
(0.64nm 3dB- bandwidth). Standard 40Gbit/s estimated the Q-factor for each of the seven independently modulated channels, and the results reported here are for the worst channel. Figure 2a shows the estimated Q-factor of the worst channel for each input power level in the absence of PMD. After transmission over 40km of SSMF, the performance was degraded by stimulated Brillouin scattering (SBS) for input powers per channel above 2dBm, as confirmed by the backscattered power with a clear threshold in the region of 2dBm per channel. For low injection power, we obtained relatively small variations of the Q-factor due to the 40dB OSNR set in the transmitter and the residual sub-channel crosstalk.
0.1
0.01
6
Fig. 2. Q-factor after 600km of SSMF with (closed diamonds) and without (open diamonds) pre-dispersion compensation, and after 40km of SSMF (triangles); including backscattered power for 40km of SSMF (empty circles) as a function of the injected power per channel.
For a total transmission distance of 600km of SSMF (closed diamonds), the Q-factor decreased, as expected, but remained above 9.1dB (FEC threshold) for a dynamic range of more than 14dB. SBS no longer dominated the performance, given an optimum input power per channel at around -6dBm, and therefore SBS was ignored in simulations for longer distances. As for standard 40Gbit/s systems, pre- and post-dispersion compensation may be used for CoWDM to optimise the dispersion map for long transmission reach, with zero residual dispersion at the receiver. Considering no residual dispersion per span for simplicity, an optimal pre-dispersion of -220ps/nm was found to enable 0.7dB Q-factor improvement. We note that this value is close to the optimal pre-dispersion value of a standard 40Gbit/s (~-180ps/nm/km for SSMF) system [12]. The impact of fibre PMD was simulated for this dispersion map, assuming typical values of 0.1ps/√km and 0.2ps/√km for SSMF and DCF respectively, and the transmission performance was evaluated for a total reach of 600km. More than 1000 simulated Q-factors, considering all simulated 7 channels, were used in order to correctly represent the statistical impact of PMD and allow the calculation of the probability distribution function of the Q factor. We considered the transmission of a 0.3 Tbit/s CoWDM and compared its performance with 7 x 42.6Gbit/s NRZ channels with 100GHz channel spacing as well as a single channel 0.3Tbit/s NRZ signal. The total input powers into SSMF and DCF were 2dBm
600 km 0.001 14
15
16 17 Q factor (in dB)
18
19
Fig. 3. Cumulated probability of Q factor after transmission over 600km of SSMF fibre for CoWDM with (closed triangles) and without (open triangles) nonlinearity; for standard NRZ system with (closed diamonds) and without (open diamonds) nonlinearity; and for single channel NRZ (closed squares) with half of the total PMD.
4. Experimental Confirmation: The experimental set up used to test these numerical simulations was similar to the simulation set up of Fig 1, with the full modulator array replaced by a pair of independently driven modulators, one each for odd and even numbered channels, and the relative phase was optimised using a piezo ceramic fibre stretcher. The resultant experimental configuration is also shown in fig 7, and full details may be found in [3, 5]. All of the optical amplifiers were kept at a constant power and optical attenuators ensured the appropriate power level launched to each transmission section, where the total launch power was monitored. 4a. Self Phase Modulation In order to investigate the dominant non-linear effects for 42.6GHz spaced CoWDM, a transmission system comprising two links of 50km Vascade® R1000 fibre was assembled. An additional 12.1km of Corning® SMF-28e fibre was used to compensate the residual chromatic dispersion. Vascade® R1000 fibre was designed for ultra long distances and high bit rate (40Gbit/s) applications [7]. In this experiment, the received pulse patterns, after WDM demultiplexing (fig 5), showed remarkable similarity for both single channel and CoWDM signals, the main differences arising from
Post-compensation was added at the receiver to compensate for the residual dispersion, and each of the seven 40Gbit/s sub-channels was detected using a standard receiver configuration, after demultiplexing using the 85.2GHz FSR AMZI and 0.64nm filter. Typical results are shown in fig 6. Again, no evidence of four wave mixing was found, owing to the strength of the dispersion map for a sub-channel spacing of 40GHz. The results indicated an FEC limited transmission distance of 1,200km at 0.3Tbit/s and an OSNR limit of 26dB, which was a little more than 3dB away from the theoretical limit. This penalty was attributed to residual chromatic dispersion, and the pattern sensitivity of the modulators and drivers. 22
800 km Q-factor 1,200 km Q-factor
20
40 800 km OSNR 1,200 km OSNR
38 36
Q-factor (dB)
18
34
16
32
14
30
12
28
10 FEC Limit
26
8
24
6
22
4
Fig. 5. Non-linear signal distortion for single channel (top) and coherent WDM (bottom) transmission over 112km dispersion managed fibre with a launch power of 8dBm per channel.
Independent CoWDM experiments have also confirmed that the dispersion tolerance and OSNR sensitivity are governed by effects scaling with the modulation rate of the sub-carriers, rather than the total capacity [5, 14] 4b. Long Haul Transmission Having predicted numerically, and confirmed experimentally, that the performance of CoWDM is governed by the bit rate of each sub-channel, it remains to confirm that no unforeseen interactions between these impairments would further degrade the transmission performance. To this end a 1,200km transmission experiment was carried out. In this case, the subchannels were independently NRZ data (and delayed data-bar) encoded with a true 231-1 PRBS, and the sub channel rate was 40.0Gbit/s. The resultant 280Gbit/s signal (after allowing for FEC overhead) was then precompensated (-342ps/nm) before entering the recirculating loop, which comprised 4 spans, each of 100km of SSMF, followed by a slope matched dispersion compensating module. An EDFA (5.5dB noise figure) and the remaining span loss was compensated using forward (4.5dB gain) and backward (11.2dB gain) Raman amplification in each span of SSMF. After the 4 spans the signal was launched into a polarisation scrambler, dynamic gain equalizer (DGE) and a pair of EDFAs overcame loop specific losses. The residual chromatic dispersion and differential group delay of the loop were approximately 140ps/nm and 1.6ps per recirculation respectively.
OSNR/0.1nm (dB)
residual linear crosstalk (sinusoidal variation of data ones). In both cases, the pulse patterns were dominated by a reduction in amplitude of isolated ones, and significant distortion of consecutive ones, consistent with self phase modulation limited performance. No evidence of four wave mixing or SBS was observed in the optical spectrum. Within the limits of experimental error, the measured penalty as a function of per subchannel launch power for the 0.3Tbit/s CoWDM signal and the 42.6 Gbit/s NRZ signal were the same, indicating that both systems were limited by the same nonlinear effects.
-6
-4
-2
0
2
4
6
8
10
20 12
Total Launch Power (dBm) Fig. 6. Experimental Q-factor (left axis, closed symbols) and OSNR (right axis, open symbols) as functions of total launch power for 800 km (triangles) and 1,200 km (squares).
Similar results, using the same transmission link, were obtained for WDM transmission of NRZ signals with a channel spacing of 100 GHz [8] and similar overall transmission distances were also observed for other forms of CoWDM operating at lower aggregate capacities [5,9]. We may thus conclude that no additional degradations were found, and that CoWDM performance is determined by the sub-channel data rate. 4c. Polarisation Division Multiplexing The 0.3 Tbit/s transmission capacity reported above may be readily upgraded by adding additional channels. A 13 channel system (0.55Tbit/s) could be readily achieved by extending the modulator array and increasing the RF power delivered to the comb generating amplifiers [10]. However, in order to demonstrate 0.6Tbit/s capacity without increasing the occupied spectrum, polarisation division multiplexing (PolMUX) may be employed. This was achieved experimentally by adding a polarisation maintaining fibre based polarisation multiplexing stage to the output of the transmitter, and including a polarisation beam splitter within the receiver, as shown in Fig 7. The polarisation multiplexer comprised a polariser, to ensure that the energy is confined at the same state of polarisation, followed by a 45o-launch in respect to the principal axis of a polarisation maintaining fibre (PMF) to equally excite the fast and slow axes of the fibre. In
TRANSMITER
B
A
DFB
FSR= 85.2 GHz
MZ
MZ
clock
delay line
MZ
delay line
MZ
D
45°
delay line
PC
piezo
polariser
PC
PM fibre
C
PD
phase shifter
42.6 GHz
DAQ
data & data
42.6 Gbit/s
mixer
RECEIVER
Pattern Generator
Error Detector
CR
PD
42.6 Gbit/s
FSR=85.2 GHz
PC PBS
PD
0.64 nm
E
Power meter ATT
Fig. 7. Experimental configuration for CoWDM and PolMUX, where the transmitter is formed by (A) comb generator, (B) phase-preserving modulator array with (C) phase-stabilising control loop and (D) PolMUX, and (E) the receiver.
this experiment, a 100m-long PMF was used, with an arbitrary net differential delay of 137ps, and therefore decorrelating the patterns by 5.8bits. The output of this 596.4Gbit/s transmitter was then launched into the preamplified receiver, which enabled each individual tributary channel to be selected using a 0.64nm tuneable band-pass filter followed by a polarisation demultiplexer consisting of a polarisation controller and a polarisation beam splitter (PBS). The two orthogonal polarisations were measured from different ports of the PBS, without re-adjusting the polarisation controller. 4
5
-log(BER)
6 7 8 9 10 11 12 13 -35
-30
-25
-20
-15
-10
Total Received Power (dBm)
Fig. 8 BER curves for single channel NRZ (closed squares), 300 Gbit/s CoWDM (closed circles), and 598 Gbit/s PolMUX CoWDM (up and down triangles).
Bit-error rate (BER) measurements as a function of total receivdr power were performed for: single-channel NRZ, CoWDM at 0.3Tbit/s, and finally 0.6Tbit/s PolMUX CoWDM, as shown in Fig 8. In the single channel case, receiver sensitivity of -28.4dBm at a BER of 10-9 was obtained. The average receiver sensitivity over 7 CoWDM channels at 0.3Tbit/s capacity was approximately 7.1dB higher, due to the increase of the number of channels (~8.4dB); offset by the reduction in spontaneous-spontaneous beat noise (~0.45dB) due to the inclusion of the PBS at this stage. A further increase of 4.2dB in the average receiver sensitivity was observed for PolMUX CoWDM with 3dB coming from doubling the channel count, leaving a PolMUX penalty of only 1.2dB taking into account the benefits of the PBS, and with negligible overall penalty (< 0.5dB) if this is neglected.
5. Conclusions In this paper we have demonstrated that the Coherent WDM technique, with multiple sub channels modulated at 42.6 Gbit/s, offers a tolerance to transmission impairments similar to those of individual 42.6 Gbit/s channels. Readily practical all- optical multiplexing and a symbol rate of 42.6Gsymbol/s should enable the development of CoWDM systems with only minor modifications to existing commercial technologies [e.g. 11]. This approach enabled a transmission reach of 1,200km at a total throughput of 280Gbit/s after FEC, representing a record bit rate distance product of 0.3Tbit/s-Mm per laser, using a transmission link designed for operating with current systems at 10 and 40Gbit/s. Polarisation multiplexing enabled a total capacity of 0.6 Tbit/s, demonstrating excellent potential for future systems based on 1TbE. 6. Acknowledgements The authors would like to acknowledge IRCSET for a PhD scholarship. This material is based upon work supported by the Science Foundation Ireland under Grant 06/IN/I969. The authors would also like to thank M.Rukosueva of Corning Inc for provision of the Vascade R1000 fibre used in this paper. . 7. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
R.Metcalfe, OFC’08, Plenary. Talk (2008) R.W.Chang et.al.,, IEEE Trans. Comm. Tech. 16-4, pp529-540 (1968) A.D.Ellis et.al., OFC’06, OThR4 (2006) S.L.Jansen et.al., OFC 2008, PDP2 (2008) F.C.G.Gunning et.al., PTL 18-12, pp1338-1340 (2006) A.J.Lowery, OFC’06 Paper PDP39 (2006) W.D.Cornwell, et.al., OFC 2002,WP4 (2002) E.Pincemin et.al., Optics Express, 14-25, pp1204912062, (2006) A.Sano et.al, ECOC’07, PD1.7, (2007) Healy et.al., Optics Express, 15-6, pp2981-,,(2007) S.Corzine et.al., OFC’08, PDP18 (2008) Y. Frignac et al, OFC 2002, ThFF5 (2002) F. Bruyere et.al, PTL, 6-3, pp443-445 (1994)
13. 14. T.Healy et.al., ECOC’05, Tu3.2.5, (2005)
