Leistungsfähigkeit von elektronischen Entzerrer in hochbitratigen optischen Übertragungsystemen. S. Otte, W. Rosenkranz. chair for communications,

Transcription

1 Leistungsfähigkeit von elektronischen Entzerrer in hochbitratigen optischen Übertragungsystemen S. Otte, W. Rosenkranz chair for communications, Sven Otte, DFG-Kolloquium, , 1

2 Outline 1. Motivation. Overview of Impairments: Chromatic Dispersion & SPM and PMD 3. Nonlinear Signal Components within electrical receiving Signal 4. Electronic equalizer approaches 5. Adaptive realization 6. Experimental results for Chromatic dispersion & SPM 7. Results for PMD 8. Summary Sven Otte, DFG-Kolloquium, ,

3 Why Electronic Equalization? chromatic dispersion PMD nonlinear effects application: compensate for dispersion and nonlinear effects - in WDM-links with DCF to compensate for residual dispersion - in the nonlinear power regime of the fiber compensate for PMD+GVD+SPM independent from physical origins standard adaptive algorithm available in general less cost intensive then optical compensator hardware implementation on receiver-ic combined with clock & data recovery LPF data recovery with feedforward & feed-back filter F(f) clock B(f) data Sven Otte, DFG-Kolloquium, , 3

4 Fiber Impairments chromatic dispersion (GVD) (Pulse broadening due to group velocity dispersion, square dependence) H( f) const e jπ Dλ L c f D=17ps/nm km for SSMF polarization mode dispersion (PMD) (differential group delay, mode coupling) α, δτ, ϕ α 1 1 1, δτ, ϕ α3, δτ3, ϕ3 wave-plate model, τ, γ self phase modulation (SPM) A z a b A + - =- t A j jg A A nonlinear Schrödinger-equation Sven Otte, DFG-Kolloquium, , 4

5 Optical Single Channel Transmission System laser modulator SSMF photodiode low-pass filter i(t) NRZ data stream B e clock & data recovery SPM GVD PMD... amount of nonlinear distortions? Sven Otte, DFG-Kolloquium, , 5

6 Nonlinear Signal components within the electrical receiving signal for GVD and SPM Volterra-series model: y () t = y () t + y () t receive linear nonlinear Ratio of linear signal components to nonlinear signal components vs. fiber length (17ps/nm/km) for linear fiber behavior Ratio of linear signal components to nonlinear signal components vs. mean optical input power at 100&150km km P L /P NL in db W/o receiving filter with receiving filter P L /P NL in db km L/km P in /dbm Sven Otte, DFG-Kolloquium, , 6

7 Electronic Equalizer Approaches DATA-DECISION OPTICAL FRONT-END FIR FILTER FORWARD EQUALIZER + - FEED-BACK EQUALIZER FIR FILTER FIR-DFE: feed-back & feedforward filter in order to compensate for precursor ISI DATA-DECISION NONLIN.FIR STRUCTURE NONLIN. FOR- WARD EQUALIZER + - NONLIN. FEED- BACK EQUALIZER NONLIN.FIR STRUCTURE NL-FIR-DFE: nonlinear processing based on Volterra theory within forward and backward branch in order to compensate for nonlinear distortions Sven Otte, DFG-Kolloquium, , 7

8 Adaptive Equalizer Implementation (LMS-algorithm) LPF accu f 0 (k i ) accu T b f 1 (k i ) mse mse vs. equal. LMS-algorithm coefficients reduction of numerical complexity by using the instantaneous square error f 0 (k i ) f 1 (k i ) as a figure to minimize resulting mse = Ein a dsimplified k y k expression for the gradient k i [ ( ) ( )] d se= dk y k {[ ( ) ( )] } d mse_min yd ( k ) dk ( ) weighting factor error f( k + 1) = fbk ( k ) + 1) µ = kbk ( ) + µ dk ( 1) [ dk ( ) yd ( t0 + k) ] eq.state b(k i ) T b accu Sven Otte, DFG-Kolloquium, , 8

9 Properties of LMS-Adaptation Process Comparison between theoretical and simulated acquisition behavior of a tap synchronous equalizer under first order PMD distortions Residual error vs. number of iterations at τ=80ps Convergence rate vs. DGD (mse(k)-mse_min)/mse_min/db linear fit theoretical curve iteration Simulation result mean steady state error rate of convergence/µ theoretical curve simulation &linear fit τ/ps Sven Otte, DFG-Kolloquium, , 9

10 Acquisition behavior of LMS algorithm Monte-Carlo simulation T b DGD=0.8T b, γ = 0.5 T f 0 (k i ) T f 1 (k i ) equalizer: FIRDFE {6,}, f 3 =1, f i =0, b 1, =0 e0 e-1 e- w/o equalization FIR-DFE {6,} training sequence FIR-DFE {6,} decision directed mode ε(k i ) - µ + - T b b(k i) T e-3 BER e-4 e-5 e-6 e-7 e-8 e SNR/[dB] 5dB acquisition requirements: EO>15% BER<10 - Sven Otte, DFG-Kolloquium, , 10

11 Measurements Equalizer: Prototype implementation (cooperation partner) of a Fractionally spaced equalizer with 10 coefficients and LMS-based adaptive coefficient setting Performance evaluation Experiments were carried out in the lab at the chair for communications in Kiel (10Gb/s) Sven Otte, DFG-Kolloquium, , 11

12 Measurement Results for GVD uncompensated link with linear fiber behavior and D=17ps/nm/km 8 Decreased sensitivity against residual PP: 350ps/nm improvement in maximal transmission 3dB PP: 0km Power Penalty 3.5dB Penalty/dB dB w/o equalizer with equalizer 350 ps/nm 3.5dB total dispersion/ [ps/nm] Sven Otte, DFG-Kolloquium, , 1

13 Measurement Results for SPM uncompensated link with high optical input power, 1700ps/nm (L=100km) 8 SPM effect causes additional nonlinear signal components within electrical receiving signal Power Penalty reduction decreases from 3.5dB w/o SPM to 1.5dB with an considerable impact of SPM Penalty/dB dB with equalizer w/o equalizer 1.5dB P input /dbm Sven Otte, DFG-Kolloquium, , 13

14 Simulation Results for PMD PMD parameter: mean DGD value = PMD = 0.5T b # wave plates =100 Power Penalty /[db] w/o equalization 1st order worst case assuming an outage probability of 10-5 (5min/year) a tolerable DGD of τ corresponds to PMD= τ /3. DGD PMD 0.3T b 0.1T b w/o 0.6T b 0.T b FIR-DFE 0.9T b 0.3T b NL-FIR-DFE Power Penalty /[db] FIR-DFE DGD /[t/tb] DGD /[t/tb] Power Penalty /[db] NL-FIR-DFE DGD /[t/tb] Sven Otte, DFG-Kolloquium, , 14

15 Summary of Results and Conclusions PMD PMD/ L Maximum transmission distance in km at 10Gb/s Maximum transmission distance in km at 40Gb/s w/o equalization FFIRDFE NLFIRDFE w/o equalization FFIRDFE NLFIRDFE GVD/SPM Pinput dbm in Dispersion tolerance in ps/nm at 10Gb/s Dispersion tolerance in ps/nm at 40Gb/s w/o equalization FFIRDFE NLFIRDFE w/o equalization FFIRDFE NLFIRDFE Sven Otte, DFG-Kolloquium, , 15

16 Project summary so far A Volterra-Theory model has been developed which enables a theoretical performance analysis of electronic equalizer in optical transmission systems A complete performance evaluation including adaptation algorithm properties were performed on a simulation basis at 10Gb/s and 40Gb/s for GVD, SPM and PMD Experimental Performance evaluation at 10Gb/s in cooperation with an industry partner Thanks to the Deutsche Forschungsgemeinschaft for supporting this work!! Sven Otte, DFG-Kolloquium, , 16

17 Nonlinear Signal components within the electrical receiving signal for a section PMD-model φ 75ps 75ps P lin /P nl / [db] DGD/ [ps] Ω / [ps ] Ω P lin /P nl DGD Nonlinear components vanish if nd order coefficient vanishes. That is either if fast/slow axis coincide with fast/slow axis or fast/slow axis coincide with slow/fast axis φ/π Sven Otte, DFG-Kolloquium, , 17