Adaptive Modulation Schemes - IEEE Xplore

3 downloads 0 Views 280KB Size Report
2:Athens Information Technology Centre, Athens,. Phone: +353 21 4904858 Email: andrew.ellis@tyndall.ie. Abstract – In this presentation we will discuss recent.
TuD3.2 (Invited) 2:00 PM – 2:30 PM

Adaptive Modulation Schemes A.D.Ellis, I.Tomkos2, A.K.Mishra, J.Zhao, S.K.Ibrahim, P.Frascella, F.C.G.Gunning Tyndall National Institute and University College Cork Department of Physics, Cork, Ireland 2: Athens Information Technology Centre, Athens, Phone: +353 21 4904858 Email: [email protected] Abstract – In this presentation we will discuss recent advances in adaptive modulation schemes for ultra high bit capacity optical networks. This paper focuses on optimising the sub-carrier spacing for such high capacity networks..

accommodated with modest increases in network reconfiguration timescales [7]. These issues will be discussed in the presentation, but in this paper we consider the performance characteristics of ultra high capacity multi-carrier transmission limited by fibre non-linearity as a function of sub-carrier capacity, demonstrating a trade-off between reach and signal structure.

1. Introduction Wavelength routed optical networking have matured significantly since the introduction of the optical amplifier. Initially opaque point-to-point links to resolved bandwidth limitations. The first optical networking solutions were based on the establishment of pre-planned wavelength paths and static optical add/drop switching elements. Driven by higher bandwidth demands and the necessity for more flexible network connectivity, reconfigurable transmission and switching solutions were then deployed using reconfigurable optical add/drop multiplexers. Today, tuneable lasers, initially intended to reduce the cost of spare equipment, are expected to have a significant role in the future deployment of tuneable transmitters to maximise network resource utilisation. Similar trends in switched network capacity also suggest that future transponders should enable up to 1 Tbit/s transmission by 2015. There are three main system parameters where flexibility may be provided: a. modulation format, b. the number of efficiently packed sub-carriers, c. the signal wavelength, and appropriate combinations will maximise reach. In each case, maximising backwards compatibility with existing network equipment is essential. With respect to modulation format, there is a clear trend towards a common opto-electronic platform, namely an optical I-Q modulator (e.g. a dual parallel Mach Zehnder modulator) and a digital coherent receiver, which can support a multiplicity of modulation formats. However, the electronics remain specific to each format. Multi-carrier techniques using a single laser source, either directly optically multiplexed (coherent wavelength division multiplexing, CoWDM [1-3]), or generated electronically with the addition of a cyclic prefix (orthogonal frequency division multiplexing, OFDM [4-6]), are currently under development. In both implementations, flexibility in the occupied spectrum is controlled via the sub-carrier count, and impairments are dictated by the symbol rate of each sub-carrier [2,5]. Commercially available tuneable lasers offer either slow switching speeds or poor performance when used with higher order modulation formats. Recent research however, suggests that for appropriate laser characteristics, adequate performance my be

978-1-4244-3914-0/09/$25.00 ©2009 IEEE

2. Optimisation of Multi-Carrier Symbol Rate In order to achieve predicted 1 Tbit/s transponder capacities, it is necessary to implement some form of optical multiplexing, either in the frequency [1-6] or the time domain [8]. However, the optimum bandwidth allocation between sub-carriers has yet to be established. To investigate this issue, numerical simulations for a high-capacity multi-carrier transmission (CoWDM) system were carried out using VPI Transmission Maker v.7.5 (figure 1). The simulations were performed using “J” NRZ modulated sub-carriers, for a range of sub carrier spacings (which were equal to the bit rate per sub-carrier B) for a total capacity of 550 Gbit/s. This rate was chosen to allow for 1 TbE transmission with polarisation multiplexing [3]. jth source

PRBS B Laser

φ=j π/2, f=j B Overall Q

Modulator

BER/Q BER/Q Estimation BER/Q jth receiver Estimation BER/Q

+D fibre 1mW

-D fibre

Estimation Estimation

Receiver

AMZI

φ=(j-1/2) π

Figure 1

Filter f=j B

x40 100% compensation

- Schematic diagram of numerical simulations

The transmitter was simulated using an array of externally modulated lasers, spaced at precisely the symbol rate and with initial phases of each sub-carrier incremented by π/2 from its neighbour. The simulation conditions ensured a minimum of 16 samples per bit, 1150Gsample/s, with 1024 total simulated bits and 32 bits for each sub-carrier. The passively multiplexed signals were transmitted over three different links (Table 1), representing respectively an installed inland system [9] a submarine system [10] one without any dispersion compensation [4-6]. Each link had a length of 2,000km and an amplifier spacing of 50 km Ideal 100% residual dispersion compensation was performed in the optical

141

domain before photo-detection. This compensation may equally have been performed electronically.

Positive Dispersion Fibre

Negative Dispersion Fibre

L

α

D

γ

L

α

D

γ

1

50

.2

17

1.3

6.6

.38

-127

5

2

33

.187

18.5

1.04

17

.24

-37

3.9

3

50

.2

17

1.3

n/a

n/a

n/a

n/a

The overall results of these simulations are shown in Figure 4 which illustrates the variation in Q factor with channel spacing from 1 GHz to 550 GHz (single carrier). 16 Maximum Overall Q (dB)

Map

5 Gbit/s. For the OSNR limited case (red squares) the observed BER varies randomly with channel number. However, if the performance is degraded by nonlinearity (blue diamonds and line) the impact of quasi phase matched four wave mixing process [5, 13] (~6 GHz bandwidth for this map) is apparent.

Table 1: Fibre parameters used for numerical simulation showing length L (km), loss α (dB/km), dispersion D (ps/nm/km) and non-linear coefficient γ (/W/km). rd

At the receiver, each channel was demultiplexed by a 3 order Gaussian filter with a bandwidth equal to 1.8 times the symbol rate [1] and a half-bit delay asymmetric Mach Zehnder interferometer. The overall estimated Q factor was obtained by bit error rate averaging all channels after individual Q estimation [11]. Whilst these simulations were performed for NRZ CoWDM, the conclusions will also applicable to OFDM implementations due to the similarity in the signal structure and to high order formats due to the dominance of peak to average power ratio considerations. Overall Estimated Q (dB)

Map 3

8 6 4 2 0 0

5

10

Map 1

2 1

10 100 Sub-Carrier Spacing (GHz)

1000

3. Conclusions To enable a seamless migration from current networks to adaptive high capacity networks with capacities in the region of 1 Tbit/s per transponder, a multi-carrier symbol rate in the region of 10 to 40Gsymbol/s should be adopted to maximise reach over existing dispersion maps. To maximise flexibility, this should be implemented in wavelength and format adaptive manner.

Figure 2: Typical simulated Q factor variation with launch power, for 14 sub-carriers at 40 Gbit/s (560 Gbit/s). Sub-channel estimated BER

Map 2

4

15

Total Launch Power (dBm)

1 0.1 0.01

Map 3

6

All maps are degraded by intra-channel four wave mixing for sub carrier spacing above 100GHz with dispersion management tending to restore peak powers and exacerbate the penalty. Below 10 GHz sub carrier spacing, quasi-phase matched four wave mixing becomes dominant for maps 2 and 3, while map 3 offers the best performance, consistent with the reduction of penalties arising from quasi phase matching. It is evident from figure 5 that current dispersion maps show clear optimum symbol rates, corresponding closely to the data rates for which they were originally designed for (10 Gbit/s for map 1 and 40 Gbit/s for map 2).

10

-5

8

Figure 4: Variation of maximum Q factor with sub-carrier spacing for all three dispersion maps.

Map 2

-10

10

0.1

Map 1

12

12

0

16 14

14

non-linearity OSNR

4. Acknowledgements

0.001

This material is based upon work supported by Science Foundation Ireland under Grant 06/IN/I969.

0.0001 0

20

40

60

80

100

120

Sub-channel number

5. References

Figure 3: Example of simulated bit error ratio measurements as a function of channel number for 110 5GHz spaced sub-carriers at 5 Gbit/s transmitted over map 2 for launch powers of -8dBm (red) and +8dBm (blue), both with overall Q factors of 3.1dB.

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

Typical results are shown in Figure 2 and 3. In figure 2, we observe that the nonlinear performance of dispersion maps 2 and 3 is somewhat different to maps 1, where it is likely that the additional non-linearity in the dispersion compensative fibre has adversely impacted the performance [12]. In Figure 3, we illustrate typical measured raw BER data for map 2, with a symbol rate of

978-1-4244-3914-0/09/$25.00 ©2009 IEEE

142

A.D.Ellis et.al, OECC’08, Paper WeA1, (2008) T.Healy, et.al., OFC’06, Paper JThB10, (2006) F.C.G.Gunning et.al., CLEO Europe, Paper CI8-5, (2007) S.L.Jansen et.al., OFC 2008, PDP2 (2008) L.B.Du et.al., Optics Express, pp19920 (2008) Y.Ma et.al., OFC’09, PDPC1 (2009) A. K. Mishra, et.al., OFC’08, Paper OTuC4, (2008) C.Schmidt-Langhorst, OFC 2009, PDPC6 (2009) E.Pincemin et.al., Optics Express, pp12049, (2006) S.Ten, LEOS’04, pp543- (2004) SignalAnalyserEl, VPI TransmissionMaker 7.5 (2008). N.J.Smith, Optics Letters, pp570- (1996). D.A.Cleland et.al., Elect Lett, pp-307- (1992).