Impact of the Transmitted Signal Initial Dispersion Transient on the Accuracy of the GN-Model of Non-Linear Propagation

Transcription

1 Impact o the Transmitted Signal Initial Dispersion Transient on the Accuracy o the GN-Model o Non-Linear Propagation A. Carena (), G. Bosco (), V. Curri (), P. Poggiolini (), F. Forghieri () () DET, Politecnico di Torino, Corso Duca degli Abruzzi, 4, 09, Torino, Italy () Cisco Photonics Italy srl, Via Philips, 0900, Monza, Italy.

2 Motivation Non-linear propagation in uncompensated links can be studied using the ingredients: Signal is Gaussian distributed Nonlinear Intererence is Gaussian distributed and additive Nonlinear Intererence is perturbative First ingredient is not veriied at system input: it takes some accumulated dispersion to turn the signal into Gaussian noise This work investigates the error introduced by the Initial Dispersion Transient (IDT) with respect to prediction o the

3 Outline A quick recap o the NLI estimation technique Simulation setup Reerence system description Results Impact on system perormance prediction Conclusions 3

4 4 A. Carena et. al, "Modeling the impact o nonlinear propagation eects in uncompensated optical coherent transmission links", IEEE/OSA Journal o Lightwave Technology, vol., no. 0, 5 May 0, pp ) ( s span N NLI NLI Coherent NLI accumulation s span N NLI NLI ) ( Incoherent NLI accumulation 3 NLI P ch ) )( ( 4 ) )( ( sin ) )( ( sin ) )( ( 4 ) ( ) ( ) ( 7 6 ) ( d d L L N j e e G G G G S S S L j L Tx Tx Tx NLI S S N s

5 NLI estimation technique NLI variance was estimated directly on the scattering diagram by averaging o all points Noiseless simulations with: non-linearity turned on tot non-linearity turned o lin The NLI variance was ound as: NLI tot lin and as NLI 3 P ch 5

6 Reerence system: Tx & Rx TRANSMITTER R S =3 Gbaud 8G PM-QPSK 56G PM-6QAM Nyquist-WDM DSP spectral shaping WDM roll-o=0.0 D=33.6 GHz 9 channels RECEIVER Coherent receiver Electrical bandwidth B elt =0.5 R S =6.0 GHz ADC SpS DSP LMS with training sequence 5 taps 6

7 Reerence system: Link 00 km Fiber span x50 spans EDFA EDFA SMF Attenuation =0. [db/km] Non-linearity =.3 [/W/km] Dispersion D=6.7 [ps/nm/km] NZDSF Attenuation =0. [db/km] Non-linearity =.5 [/W/km] Dispersion D=3.8 [ps/nm/km] 7

8 SMF 60 PM-QPSK [db(/w )] N span

9 SMF PM-QPSK Black dashed: Coherent 50 [db(/w )] N span

10 SMF PM-QPSK Black dashed: Coherent Blue solid: Simulation with PD (+00,000 ps/nm) [db(/w )] N span

11 SMF PM-QPSK Black dashed: Coherent Blue solid: Simulation with PD (+00,000 ps/nm) [db(/w )] Red solid: Simulation NOPD N span

12 SMF PM-QPSK Black dashed: Coherent Blue solid: Simulation with PD (+00,000 ps/nm) [db(/w )] Red solid: Simulation NOPD Black dash-dotted: Incoherent N span

13 SMF PM-6QAM Black dashed: Coherent Blue solid: Simulation with PD (+00,000 ps/nm) [db(/w )] Red solid: Simulation NOPD Black dash-dotted: Incoherent N span

14 NZDSF PM-QPSK Black dashed: Coherent Blue solid: Simulation with PD (+00,000 ps/nm) [db(/w )] Red solid: Simulation NOPD Black dash-dotted: Incoherent N span

15 System impact: Max Reach N span, MAX 3 DN span, db 3 D db Inaccuracies in estimation are mitigated by /3 5

16 System impact BER target = PM-QPSK L span =0 km 5 N span [dbm] P TX PM-6QAM L span =80 km 6

17 System impact BER target = 0-3 PM-QPSK L span =0 km Black dashed: Coherent Black dash-dotted: Incoherent N span [dbm] P TX PM-6QAM L span =80 km 7

18 System impact 5 0 BER target = 0-3 PM-QPSK L span =0 km Black dashed: Coherent Black dash-dotted: Incoherent N span 5 0 Red solid: Simulation NOPD [dbm] P TX PM-6QAM L span =80 km 8

19 System impact 5 0 BER target = 0-3 PM-QPSK L span =0 km Black dashed: Coherent Black dash-dotted: Incoherent N span 5 0 Red solid: Simulation NOPD Blue solid: Simulation with PD (+00,000 ps/nm) [dbm] P TX PM-6QAM L span =80 km 9

20 Conclusions The Initial Dispersion Transient does have some impact on the accuracy o the With QAM constellations, the Coherent always provides a lower bound to system perormances High-order constellations show better accuracy because they are closer to Gaussian distribution already at transmitter (higher PAPR) The Incoherent typically delivers good prediction It is not a more aithul modeling, two approximations tend to cancel each other out 0

21 Acknowledgments This work was supported by CISCO Systems within a SRA contract. andrea.carena@polito.it