L. Billès, D. Leguen, S. Lobo, L. Bramerie, S. Del Burgo, F. Merlaud, T Georges
Corvis Algety, Lannion, France (e-mail : lbilles@corvis.com)
Abstract: Spectral efficiency will be discussed as a key issue for long reach systems. Ready to deploy, 0.4 bit/s/Hz, Nx10Gb/s experiments will be highlighted, as well as next generation 40Gbit/s systems.
High spectral efficiency and long transmission distance is required for WDM systems to cope with rapid increase of internet traffic. Moreover, capacity has to be coupled with span loss constraint of terrestrial networks, ranging between 18 and 28 dB. There is a trade-off between these three parameters. In Fig. 1, 10-9 uncorrected error rate distance and the spectral efficiency of laboratory transmission systems that could foresee industrial systems (with span loss greater than 18 dB) are reported. 0.4 bit/s/Hz was demonstrated at 10, 20 and 40 Gbit/s per channel /1-11/. It is also a realistic goal for 10 Gbit/s-based DWDM ultra-long-range terrestrial systems with large span loss. Key enabling technologies are FEC, which can provide system margins of laboratory transmission systems, Raman amplification /12/, PMD compensation, especially for 40 Gbit/s bit rate and modulation format. For short distances, when signal spectrum is clean and spectral efficiency is limited by demultiplexing capability, duobinary and polarisation multiplexing with large bit rate have proven to be the most efficient, while, for long distance transmission, non-linear robust RZ format, like dispersion managed soliton (DMS) is required. It allows a good control of temporal and spectral pulse characteristics along path, at 10 and 40 Gbit/s, which also limits intra-channel interaction. Examples are given below of compromises between reach, spectral efficiency and span loss for different fibre types using this technique, combined with alternating polarisation of adjacent channels for XPM reduction. At 2.5 Gbit/s, thanks to Raman amplification and FEC, ULH reach has been installed making available for the first time all-optical networking /13/.
|
Figure 1 : Comparison of spectral efficiencies of laboratory DWDM systems (circle: 10 Gbit/s, triangle: 20 Gbit/s, square: 40 Gbit/s and light symbols: smaller span loss or corrected error rate). |
 |
Figure 2: iso-contours for log10(BER) on input-fibre power per channel against distance with all-EDFA amplification (up) and hybrid Raman-EDFA amplification (down). (SMF28 25 dB/ 100 km span ; 25 GHz channel-spacing) |
At 10 Gbit/s per channel, spectral efficiency can be improved allowing to increase capacity. The price to pay is to refine dispersion management (with a particular care to dispersion slope compensation) and shorter reach with lossy spans. A compromise must be found between reach, capacity and loss. In order to illustrate this compromise, main results of dispersion-managed soliton loop experiments at Nx10.66 Gbit/s on different types of fibre are presented.
9x10.66 Gbit/s RZ transmission was achieved with 25 GHz channel-spacing. Two different types of 2-stage amplifiers for dispersion and loss compensation have been tested: one is all-EDFA, the other one is an hybrid Raman-EDFA amplifier. Fig. 2 compares BER measurements on input fibre power against distance for the two amplification schemes. At a given distance, postchirp was fixed for all powers and optimised for best performance case. Non-linearities limit maximum transmission distances (@10-9 BER) to 1500 km at +3 dBm/ch with EDFA and 2100 km at -1 dBm/ch with hybrid amplification. Noise figure improvement by about 6 dB due to Raman amplification in SMF explains differences between low power limits of both experiments. No FWM component was found after 2100 km, as expected, thanks to high local dispersion of SMF-28 fibre. As XPM is greatly reduced thanks to alternate polarisations between adjacent channels, the main cause of non-linearity impairments is the intra-channel interaction jitter.
The 10x10.66 Gbit/s RZ transmission with 25 GHz channel-spacing with alternate polarisations was demonstrated on TWC 18 dB/75 km span with Raman amplification. The TWC dispersion at 1532 nm was 2.16 ps/nm/km. The dispersion was partially compensated every 300 km. Maximum transmission distance is mainly limited to 2800 km @10-9 BER, by intra-channel interaction (Fig.3). At 1564 nm, the distance is limited to 2600 km for D = 3.3 ps/nm/km at -6 dBm. Here, long distance, high spectral efficiency could be maintained over this fibre thanks to span loss reduction and channel power decrease.
 |
Figure 3: iso-contours for log10(BER) on input-fibre power per channel against distance (TWC 18 dB/ 75 km span and 25 GHz channel-spacing). |
 |
Figure 4: iso-contours for log10(BER) on input-fibre power per channel against distance (TW-RS 25 dB/ 100 km span and 50 GHz channel-spacing) |
 |
Figure 5: iso-contours for log10(BER) on input-fibre power per channel against distance (TW-RS 23 dB/ 100 km span and 100 GHz channel-spacing) |
With high, 25 dB span loss, long distance transmission can be optimised on TW-RS, with lower spectral efficiency. The 10x10.66 Gbit/s RZ transmission with 50 GHz channel-spacing and alternate polarisations was demonstrated with Raman amplification. The maximum transmission distance is 3200 km @10-9 BER.
Dispersion managed soliton (DMS) transmission was tested on recirculating loop at 40 Gbit/s with alternate polarisation between successive pulses of 5 ps FWHM. The 100 km/ 21 dB span of DSF with D = 0.63 ps/nm/km was compensated at 90% using DCF. Q2 = 16.5 dB (BER = 10-10) was measured at 1800 km. Such a distance can be achieved while decreasing spectral efficiency down to 0.1 bit/s/Hz.
On the other hand, one can optimise spectral efficiency. The price to pay is a reduction in transmission reach, but trade-off can be found minimizing together TDM and WDM impairments. 10x42.6 Gbit/s DMS transmission was tested in a 6x100 km straight line experiment over 23 dB/100 km spans of TW-RS, with a 100 GHz spacing and Erbium amplification, without any in-line dispersion compensation. At 600 km, there are still comfortable margins that give a good potential for long reach using Raman amplification.
Following already installed WDM systems at 2.5 Gbit/s, the deployment of the next generations at 10 Gbit/s and later at 40 Gbit/s will improve spectral efficiency over most of fibre types at the price of finding a compromise between reach, span loss and capacity. The key techniques to be controlled are the DMS, Raman amplification and FEC.