Optical Network Optimization based on Physical Layer Impairments

Transcription

1 Optical Network Optimization based on Physical Layer Impairments Joaquim F. Martins-Filho Photonics Group, Department of Electronics and Systems Federal University of Pernambuco - Recife Brazil jfmf@ufpe.br

2 Outlook Introduction Routing based on physical impairments Amplifiers for network optimization Considering four wave mixing (FWM) Wavelength assignment based on FWM Conclusions

3 Optical Network Scenario Dynamic, wavelength routed, transparent network MANAGEMENT PLANE A1 Connect A1-E1 CLIENT DOMAIN CONTROL PLANE A TRANSPORT PLANE B NNI (signaling) C NNI (transport) D TRANSPORT DOMAIN UNI (signaling) UNI (transport) E E1 CLIENT DOMAIN

4 Optical Network Scenario We have: Circuit-switched wavelength-routed networks Dispersion is compensated or managed Optical amplifiers to compensate for losses However, we have degradation of OSNR due to: Amplifier noise insertion Shot noise Amplifier gain saturation Noise induced by non-linear effects (FWM) Noise inserted in switching devices (OXC) The OSNR Degradation sets the limits of the transparent network, for a given QoS (BER).

5 Optical Network Scenario We need a Compensator for OSNR Degradation (3R-regeneration) And/Or we need: Low noise, high saturation-power amplifiers Low loss devices Intelligent RWA algorithms to mitigate the physical layer impairments (getting the best out of your infrastructure)

6 Basic Definitions Degradation of OSNR = Noise Factor The Noise Factor: The Noise Figure: F = F excess + F shot OSNR F = OSNR in out NF =10log F D. M. Baney, P. Gallion, e R. S. Tucker, Theory and measurement techniques for the noise figure of optical amplifiers, Optical Fiber Technology, vol. 6, pp , 000.

7 Cascade of Elements G F 1 = F excess1 + F = 1/L (L>1) shot1 F 3 = F excess3 + F shot3 F excess = 0 F shot1 = 1/G 1 F shot3 = 1/G F 3 = 1/G = L F F 0 c excess = excess1 + + G1 F c shot = G L 1 G 3 F excess3 G 1 L

8 Node Configuration In each node we have: EDFA Tx SWITCH WRS λ 1 WRS λ.. WRS λ N WRN (k) MUX DEMUX Rx TAP FIBERS F G LINKshot LINK = L = L G SW + MUX SW + MUX G PRE AMP L L PRE AMP G FIBER+ TAP FIBER+ TAP G BOOSTER L L BOOSTER DEMUX + SW DEMUX + SW PRE AMP FIBER+ TAP FLINKsig, sp = FBOOSTER + GBOOSTER F L

9 Noise Factor of a Lightpath FLink sig,sp1 FLink sig,sp FLink sig,sp3 FLink sig,sp4 FLink shot1 FLink shot FLink shot3 FLink shot4 GLink 1 F LIGHTPATH GLink GLink 3 GLink 4 = n F LINKsig, sp i + i 1 i= 1 i= 1 j= 0 G LINK n F LINKshot j i

10 Gain Saturation We also consider the effect of amplifier gain saturation: G = 1+ G P P sig sat A decrease in G causes an increase in F 0

11 Block Diagram of our RWA Algorithm Call Requests Select Wavelength Wavelength Available? Yes Select Lightpath Calculate BER BER < 10-1? Yes No No Block Call - Wavelength selection by first fit. - Lightpath selection by minimum F, using a modified Dijkstra algorithm, where F is the cost parameter. - BER calculation from OSNR and F OSNR out = OSNR F in For OSNR out > 3 db BER <10-1.

12 Simulated Network Mesh network. circuit-switched bidirectional connections in fibers. No wavelength conversion. Poisson distribution of call arrivals and call holding time. Uniform distribution of source-destination nodes calls simulated. 70km 1 70km 30km 3 70km 4 40km 80km 40km 80km 30km 5 50km 6

13 Simulation Parameters Parameter Typical Value Number of wavelengths 16 Number of fibers 1 Amplifier Gain Amplifier Noise Figure Amplifier Saturation Power OSNR in Switch loss MUX / DEMUX loss Tap loss Transmission fiber loss Compensating losses 5 db 16 dbm 39 db db 4 db 0.5 db 0. db/km

14 Simulation Results No. of wavelengths x No. of fibers Blocking Probability 0,1 0, Load (Erlangs) 16x1 08x 04x4

15 Simulation Results Blocking probability 0,1 0,01 1E-3 Number of available wavelengths Routing algorithm Solid symbols - Noise Figure Open symbols - Shortest path Load (Erlangs)

16 Simulation Results Blocking Probability 0,1 0,01 1E-3 1E Load (Erlangs) Amplifier Noise Figure NF4dB NF5dB NF6dB NF7dB NF8dB

17 Simulation Results 1 Blocking Probability 0,1 0,01 1E-3 1E-4 1E Load (Erlangs) Saturation Power 13dBm 15dBm 17dBm 19dBm 1dBm 3dBm

18 All-Optical gain clamping Laser cavity Attenuator Defines the laser threshold Power (dbm) Filter Should be tuned in the ASE range Wavelength (nm)

19 Gain-clamping implications Gain medium moderately saturated Population inversion: constant Lower gain dependency with signal power

20 Typical all-optical gain clamped configurations

21 Experimental setup (Circulator) Clamping laser copropagates with the signal Optical filter Lower losses for signal lightpath (lower NF) Suppressed clamping laser in the output Optical filter

22 Experimental Results: Gain and Noise Figure Gain (db) ( a ) Copropagating pump Pump power 43 mw 65 mw 75 mw Feedback atenuator losses (db) NF penalty is about 1 db Noise figure (db) Gain (db) Counterpropagating pump ( b ) Pump power 43 mw 65 mw 75 mw Feedback atenuator losses (db) Noise figure (db)

23 Experimental Results: Saturation Output Power Gain (db) Feedback loss: 18dB (clamped) 34dB (unclamped) Output Power (dbm) Unclamped amplifier: P sat = 9,5 dbm Clamped amplifier: P sat = 1 dbm 3 db 3 db

24 Network Simulation Results Blocking Probability 0,1 0, Load (Erlangs) Unclamped amplifiers Clamped amplifiers Unclamped amplifier: P sat = 9,5 dbm, NF = 5 db Camped amplifier: P sat = 1 dbm, NF = 6 db

25 How to take FWM into account? FWM generated power is already well modeled: S. Song et al, Intensity-Dependent Phase-Matching Effects on Four-Wave Mixing in Optical Fibers, IEEE J. of Lightwave Technology, vol. 17, no. 11, pp , Nov FWM generated power can act as noise One can not evaluate the impact on transmission performance directly from FWM generated power Example: Channel FWM - Noise

26 Proposed model Our model: Considers FWM as fiber noise components (FWM is the source of excess noise in fiber) Evaluate all the beating processes involved in the transmission Includes FWM and other sources of excess noise in the noise budget calculation of the system Features the QoS degradation due to FWM Can be used as a design tool

27 Model - Formulation Noise factor F = SNR SNR in out Signal to noise ratio < i signal < Δi Noise > > SNR < i Signal = < ΔiNoise Photocurrent generated by the optical signal of power level P Mean-square value of a single-sided noise power spectrum > >

28 Optoelectronic model < i signal > + < Δ i >, in noise, in No medium SNR entrada laser laser (a) Optical fiber (b) < Photodetector i signal > + < Δ i, out noise, out photodetector > SNR saída

29 Model - Formulation Electrical signal < Δi < i >= R P signal signal Noise power spectrum noise >= R B 0 S noise ( f ) df

30 Model - Formulation In the fiber input SNR in = < isignal, in > < isignal, in > = < Δi > < Δi > + < Δi noise, in noise, shot noise, SSEin > < Δ < Δi SNR i noise,shot noise,ssein in > > Shot noise component Beat noise between signal and source spontaneous emission R P signal = 0 qrp B + < Δi signal noise SSEin, >

31 Model - Formulation In the fiber output SNR out = qrb < Δi noise FWM + 0 P signal, SSEout R > P exp signal exp ( αl) ( ) αl + < Δi > noisefwm, + SSEout Square mean value of the noise due to additive noise components: FWM + SSE attenuated

32 Model - Formulation F = F Noise factor = qrb ( αl) 0 qrp P signal R That is equal to exp qrb 0 signal P exp P signal R B 0 signal P + < signal exp Δi noise, SSEin ( αl) ( ) αl + < Δi > qrp exp signal ( αl) B 0 > noise, FWM + SSEout + < Δi + < Δi noise, FWM noise, SSEin > + SSEout >

33 Model Evaluation of additive noise components Considering a square law photodetector k = r E i na ε = i 0 c E i det e j = Rk ( i ω it+φ ) Constant i r E i Electromagnetic fields involved

34 General formula >= R < Δ = = = + = = = = + = = n i n j i j j n i n i j n j i k k k n k k j i n i n i j j i n i i noise P n u P PP P n P P PP PP P i 0, , ] [ 4 3] [ δ P0 signal power Pi noise power components in the signal wavelength δ[n] delta function u[n] discrete step function

35 Model - In the fiber input P = 0 P signal < Δi noise, SSEin P = P 1 SSE >= R P + SSE 4 + P 6P signal signal P SSE P SSE ( P + P ) signal SSE

36 Simulation Channel FWM - Noise

37 Simulation - Fiber parameters Length 5.6 km Linear attenuation coefficient km -1 Modal area 50 μm non-linear refractive index.68x10-0 m /W Wavelength of the first signal in the grid Dispersion Dispersion Coefficient nm 0.84 ps/nm.km ps/nm km

38 Simulation Results: 100 GHz channel spacing FWM power (dbm) (a) Linksim S. Formulation Song et.al Power per channel (mw)

39 Simulation Results: 100 GHz channel spacing 0 (a) BER = 1 erfc SNR in F Log(BER) Linksim Formulation Power per channel (mw)

40 Simulation Results: 50 GHz channel spacing FWM power (dbm) (b) Linksim Formulation S. Song et.al Power per channel (mw)

41 Simulation Results: 50 GHz channel spacing 0 (b) Log(BER) Linksim Formulation Power per channel (mw)

42 Applications Wavelength assignment in optical networks. Include FWM in our routing algorithm by Noise Factor

43 Application: Optimization of Wavelength Assignment SNR (db) 3 Wavelength BER = 10-1

44 Conclusions (1/) For network optimization (from the performance point of view): Low noise figure, high saturation-power amplifiers Gain-clamped amplifiers are good candidates, but transient effects should be considered. Low loss devices Intelligent RWA algorithms to mitigate the physical layer impairments. It can increase the network capacity without changing infrastructure It can provide QoS at the physical layer It can provide Service Differentiation (some users at BER of 10-1, others at 10-9, and so on )

45 Conclusions (/) Future work: Addition of FWM effect in the routing algorithm Consideration of FWM in wavelength assignment Development of a network design tool, for physical layer optimization (use of Genetic Algorithm).

46 DWDM Network Laboratory: Experiment

47 DWDM Network Laboratory: Simulation

48 Acknowledgments Team: - Dr. Carmelo J. A. Bastos Filho - Daniel A. R. Chaves - Eric A. J. Arantes - Dr. Farshad Yazdani -HelderAlves Pereira - Dr. Isnaldo J. de Souza Coelho - Luciana Pedrosa Salles - Marcela Martins Melo - Rafaelli Neves de Alencar Vidal - Dr. Raul Camelo de A. Almeida Jr. - Sergio Campello Oliveira

49 Acknowledgements Financial Support: SIEMENS Support: U F P E

50 References D. M. Baney, P. Gallion, e R. S. Tucker, Theory and measurement techniques for the noise figure of optical amplifiers, Optical Fiber Technology, vol. 6, pp , 000. S. Song et al, Intensity-Dependent Phase-Matching Effects on Four-Wave Mixing in Optical Fibers, IEEE J. of Lightwave Technology, vol. 17, no. 11, pp , Nov J. F. MARTINS-FILHO, et. al., Novel Routing Algorithm for Optical Networks Based on Noise Figure and Physical Impairments. In: ECOC European Conference on Optical Communications, 003, Rimini - Itália. Proceeding of ECOC 003, 003. v. 3. p J. F. MARTINS-FILHO, et. al. Impact of device characteristics on network performance from a physical-impairment-based routing algorithm. In: OFC Optical Fiber Communication Conference and Exhibit, 004, Los Angeles - EUA. Proceedings of OFC 004 on CD-ROM, 004. p C. J. A. BASTOS FILHO, J. F. MARTINS-FILHO, Noise Figure Model for Transmission Performance Evaluation Considering Four Wave Mixing and Source Spontaneous Emission. In: SBMO/IEEE MTT-S IMOC International Microwave and Optoelectronics Conference, 005, Brasília. Proceedings of IMOC 005. IEEE / SBMO, 005. v. 1. p C. J. A. BASTOS FILHO, M. M. MELO, J. F. MARTINS-FILHO, Influence of Pump Direction in All-Optical Gain-Clamped Erbium Doped Fiber Amplifier. In: SBMO/IEEE MTT-S IMOC International Microwave and Optoelectronics Conference, 005, Brasília. Proceedings of IMOC 005. IEEE / SBMO, 005. v. 1. p. 1-5.

51 Thank you! Joaquim F. Martins Filho