1. Advanced Optical Measurements in Next Generation Networks October 2007 Mike Harrop mike.harrop@exfo.com

2. Agenda Introduction Digital Transmission Dispersion in optical Networks. Dispersion challenges for 40G OSA challenges for 40G/ROADM’s

3. What is the fundamental of digital transmission…? 101010001001010101010101010000100101010011001010101001010 Tx Rx The Rx circuit is clocking at the system line rate and ‘simply’ needs to discern between a 1 and a 0 to recover the original signal.

4. The need for speed…

5. Eye diagram at Rx demonstrates signal quality Low BERT Intermediate BERT Unacceptable BERT

6. BERT causes a lot of pain to transmission groups Typical values for acceptable BERT levels:  >> 1 x10 -12  (or 1 bit error per 1,000,000,000,000 bits sent) In terms of QoS measurements:  single BIT error = 1 error second on the network Conclusion of high BERT:  Networks inability to operate at high speed  Poor QoS figures

7. What’s important in Optical Networks Source : British Telecom Laboratories Technical Journal 2003 (authors Sikora, Zhou and Lord), Advanced network parameters which have to be properly evaluated

8. What is Dispersion? Dispersion is the time domain spreading or broadening of the transmission signal light pulses - as they travel through the fibre OutRX In TX

9. Types of Dispersion Chromatic Dispersion: Different wavelengths travel at different velocities Polarization mode dispersion: Different polarization modes travel at different velocities Pulse Pulse Spreading Pulse Pulse Spreading

10. Types of Dispersion Chromatic Dispersion: Is deterministic Is linear Is not affected by environment Can be compensated Polarization mode dispersion: Is stochastic Is not linear Is affected by the environment Cannot be easily compensated

11. Chromatic Dispersion October 2007 Mike Harrop mike.harrop@exfo.com

12. Source wavelengths = do not propagate at the same speed, thus arrive at different times A pulse transmitted in such way suffers a spread, dispersion, limiting the transmission bandwidth. Chromatic Dispersion Issue 1 2 3 1 2 3 1 3 Pulse Pulse Spreading

13. Visualizing CD Let’s visualize a light pulse travelling into a fiber and segment it into 9 quadrants (easier to visualize, and to draw!!!)

14. Visualizing CD Fiber length: Light pulse: Pulse width

15. Effects of Dispersion

16. Why is Measuring Dispersion so important? As transmission speeds go up, the residual dispersion allowable at the receiver to give a fixed system penalty goes down. Receiver Tolerance for a 1dB power penalty 2.5 Gb/s16,000ps/nm 10 Gb/s1,000ps/nm 40 Gb/s 60ps/nm e.g. An 80km link at 1550nm will build up 17ps/(nm.km) x 80km = 1360ps/nm. Therefore at data rates at 10Gb/s and higher it is necessary to compensate for the chromatic dispersion. To compensate effectively you need to measure the dispersion of the link.

17. 16 times less CD, cause 1 Time slot 125 us Faster means less time between pulses

18. The chirp effect Pulse before modulation @ 2.5 Gb/s @ 10 Gb/s @ 40 Gb/s P P P P modulation 16 times less CD, cause 2 Faster means broader pulses

19. Dispersion Compensation Good News : CD is stable, predictable, and controllable. Dispersion compensating fiber (“DC fiber”) has large negative dispersion -85ps/(nm.km) DC fiber modules correct for chromatic dispersion in the link delay [ps] 0 d TxTx RxRx DC modules fiber span

20. Dispersion Compensation for DWDM Consider 3 channel SMF system Dispersion (ps/(nm.km)) Wavelength (nm) 1300 155015701530 18.5 17.0 16.2 0 -85 SMF-28 DCF Slope = 0ps/nm^2/km Dispersion -D +D Distance SMF after 80km 1296ps/nm @ 1530nm 1360ps/nm @ 1550nm 1480ps/nm @ 1570nm Using 16km of DCF @ 85 ps/(nm.km). Gives a residual dispersion of -64ps/nm @ 1530nm 0ps/nm @ 1550nm 120ps/nm @ 1570nm

21. Dispersion Compensation for DWDM Dispersion compensation modules can only compensate exactly for one wavelength DWDM system design requires knowledge of end-to-end CD as a function of wavelength… especially for long-haul Dispersion Transmission path 10 Gb/s Tolerance -D +D -D +D 40 Gb/s Tolerance For 40Gb/s transmission slope compensators will be required.

22. CD: Bad compensation

23. Dispersion Compensation for DWDM Note. In practise system vendors don’t compensate perfectly for CD at each stage. Usually a system will be pre-compensated and then not brought back to zero during transmission. This is to avoid additional non-linear penalties such as Four Wave Mixing and Cross Phase Modulation. Dispersion Z D Accumulated -D +D D Res +D -D Transmission path

24. Types of Dispersion Chromatic Dispersion: Different wavelengths travel at different velocities Pulse Pulse Spreading Chromatic Dispersion: Is deterministic Is linear Is not affected by environment Can be compensated

25. Chromatic Dispersion - Conclusion For 10Gbits/s and higher DWDM systems we need to measure both the dispersion and the slope accurately. Many ways to measure CD in fibre but with the tolerances required for accurate compensation – the only accepted method for making this measurement with this sort of accuracy is the Phase shift method

26. Measuring Chromatic Dispersion October 2007 Mike Harrop mike.harrop@exfo.com

27. Patented FTB-5800 method: Source Oscillator DUT or FUT Optical filtering Phasemeter Chromatic dispersion Measurement Method- Phase Shift FOTP-169

28. Ref Test 1 Few kms of fiber RGD 1

29. Chromatic dispersion Measurement Method- Phase Shift FOTP-169 Test 2 Few kms of fiber Ref RGD 2

30. Chromatic dispersion Measurement Method- Phase Shift FOTP-169 Test 3 RGD 3 Few kms of fiber ADVANTAGES: - More points: more resolution - Ideal for compensation - Ideal for complex networks Ref

31. Reference and Measured Spectral Regions The system compares spectral regions about 1 nm width (A,B,…) with a reference to find the relative group delay and compute CD

32. Measuring CD Delay points are acquired Lamdba Delay (ps) Lamdba Points are fitted according to models Delay (ps) Slope of Delay gives CD Lamdba 0 10 20 30 40 50 60 CD (ps/nm)

33. n The by-default or user selected mathematical model is fitted to the RGD point using the generalized least square method. u 3-term Sellmeier (Standard fiber) u 5-term Sellmeier u Lambda Log Lambda u Cubic (Unknown fiber, flattened fiber and amplified links) u Quadratic (Compensating, DSF and NZDSF fibers) u Linear RGD Fitting

34. Standard Fiber

35. Extrapolated 0 = 1320.14 nm CD at 1550nm = 16.641 ps/nm.km Standard Fiber

36. 0 = 1547.754 nm DSF Fiber

37. Example of NZDSF Analyzed with the help of the FTB-5800 NZDSF fiber (True Wave®)

38. Specifications Good repeatabilityGood accuracy

39. EXFO FTB-5800 l Industry leading accuracy on CD and Slope l Ideal for 10G-40G compensation l Source Shape insensitivity l EDFA testing l time saving l Component characterisation l Fast measurement l Powerful but simple software Measuring Chromatic Dispersion

40. Polarization Mode Dispersion October 2007 Mike Harrop mike.harrop@exfo.com

41. Reminder Polarization mode dispersion: Different polarization modes travel at different velocities Pulse Pulse Spreading Polarization mode dispersion: Is stochastic Is not linear Is affected by the environment Cannot be easily compensated

42. Visualizing PMD Let’s visualize a light pulse travelling into a fiber and segment it into 9 quadrants (easier to visualize, and to draw!!!)

43. Visualizing PMD Fiber section: Light pulse: Pulse width

44. PMD Impact If we transmit 1-0-1: 1 0 1 With PMD, this becomes: 1 0 1 The « 1 » is dimmer, the « 0 » can have light: BER

45. Asymmetries in fiber during fiber manufacturing and/or stress distribution during cabling, installation and/or servicing create fiber local birefringence. A "real" long fiber is a randomly distributed addition of these local birefringent portions. What causes PMD

46. What causes PMD? Fiber defects Environmental constraints GeometricInternal Stress Lateral Pressure Bend Heat Wind (aerial fibers)

47. Small Birefringence

54. Fast Slow

55. Large Birefringence

60. Fast Slow

61. Birefringence and mode coupling

77. Fast Slow

78. Causes of PMD Birefringence (Bad)  Introduced during manufacture  non uniform intrinsic fibre stresses ie core concentricity  non uniform extrinsic stresses ie pressure Mode coupling (Good)  fibre bend and twist  in-built stress in “spun” fibre  splices

79. PMD - Lower Bit Rate fast axis z, t slow axis  t T0T0 T

80. PMD - Higher Bit Rate fast axis z, t slow axis  t

81. PMD vs Wavelength and Time Pradeep Kumar Kondamuri and Christopher Allen Information and Telecommunications Technology Center, The University of Kansas, Lawrence, Kansas, 66045 Douglas L. Richards Sprint Corporation, Overland Park, Kansas

82. 1dB Penalty probability: Very low Average PMD System Tolerance Low PMD average

83. 1dB Penalty probability: low Average PMD System Tolerance Limit PMD average

84. 1dB Penalty probability: very high Average PMD System Tolerance Too high PMD average

85. A PMD outage is when the instantaneous DGD exceeds a given threshold (Max DGD) A factor 3 between Max DGD and Average PMD is taken from a number of ITU ‑ T Recommendations (including G.959-1 OPTICAL TRANSPORT NETWORK PHYSICAL LAYER INTERFACES ) for 99.9954% of no PMD problems Once you know the system tolerance (Max DGD), aim at PMD < 1/3 of this value if you transmt Sonet/SDH PMD Power Penalty

86. PMD Pass-Fail criteria ITU-T G.959.1, version 7.6 defines Max DGD as 3* It also defines Max DGD as 30ps for OC-192 ITU-T G.650 places it at 25ps Max DGD, but this is based of FIBER, with no allowance to components. Good for Fiber Manufacturer, too tight for NSP IEEE-802.3ae has Max DGD at 19ps (10 GigE), and with a tolerance of 99.999987% (Corporation, Banks, etc need higher security) Max DGD is divided by 3.73 for this level

87. PMD vs Outage probability System vendors give Max DGD. You choose Outage probabliity, then calculate PMD to achieve

88. Maximum PMD value to ensure 99.9954% probability that the tolerable broadening will correspond to a mean power penalty of 1 dB. SONET-SDH Bit rate (Gbit/s) 2.5 10 40 Average PMD* (ps) 40 10 2.5 Digital Transmissions PMD Specifications

89. Maximum PMD value to ensure 99.999987% probability that the tolerable broadening will correspond to a mean power penalty of 1 dB 10 GigE Bit rate (Gbit/s) 10 Average PMD* (ps) 5 Digital Transmissions PMD Specifications

90. Total PMD vs PMD Coefficient Total link PMD (ps) 10ps over 400km 5ps over 50km Which is better? PMD Coefficient (ps/√km) used by fibre & cable manufacturers, based on ITU recommendations that a network will be 400km. For 10G Total limit is 10ps, using our network length of 400km gives: 10ps √400km = 0.5ps/ √km

91. Typical values for new fibre. G.652 Standard Single Mode <0.1ps/  km G.655 NZDSF <0.04ps/  km e.g. For a 80km SMF link you would expect to see 0.1 x sqrt(80km) = 1ps Delay For a 80km NZDSF link you would expect to see 0.04 x sqrt(80km) = 0.36ps Delay Installed base?

92. Installed Base Source: John Peters, Ariel Dori, and Felix Kapron, Bellcore 10G 40G

93. Reminder Polarization mode dispersion: Different polarization modes travel at different velocities Pulse Pulse Spreading Polarization mode dispersion: Is stochastic Is not linear Is affected by the environment Cannot be easily compensated

94. Pitfalls Chromatic Dispersion: Should be specified at the cable specs (install or rental of dark fiber) Should be tested/compensated on installation or ahead of system turn up Should be considered very deeply for DWDM systems Polarization mode dispersion: Should be specified at the cable spec level (install or rental of dark fiber) Fibers should be tested and classified for suitability of different lines speeds High levels could mean very costly re-engineering

95. Conclusions Uncontrolled fiber dispersion leads to increased BERT and lower QoS metrics Dispersion should be considered mission critical to any operator considering high speed digital transmission Accurate measurement and interpretation of those data are critical…

96. Measuring Polarization Mode Dispersion October 2007 Mike Harrop mike.harrop@exfo.com

97. TIA/EIA FOTP 124 : Polarisation Mode Dispersion for Single-mode fibres by Interferometry. Traditional Interferometric Method (TINTY) Limitations Gaussian Interferogram Smooth ripple free, Gaussian like source Ideal random coupling DUT Autocorrelation Peak Cross correlation Half width Gaussian fit Broadband Source Polarizer FUT Interferometer Mirror Analyzer Detector

98. FOTP-124: Are these Gaussian??? Saudi Arabia: South Africa:

99. FOTP-124: Are these Gaussian??? USA: UK:

100. FOTP-124: Are these Gaussian??? UK:

101. ? Source Shape Auto-correlation Infinitely broad sourceInfinitely thin line Broad uniformVery thin peak Odd-looking spectrum Broad peak, humps, ripple, etc… Add Autocorrelation to Crosscorrelation Autocorrelation: source shape

102. TIA/EIA FOTP 124a : Polarisation Mode Dispersion for Single-mode fibres by Interferometry. Generalised Interferometic Method (GINTY) No Limitations No reliance on Gaussian Interferogram Any fibre or component can be measured Any source shape acceptable Broadband Source Polarizer Interferometer Mirror Analyzer PBS Detectors FUT

103. FOTP-124 6.1.2 PMD Calculation for Fibers with Strong Mode Coupling The PMD delay,, is determined from the half width parameter,  , of the Gaussian curve fitting applied to the interferogram according to: Where   is the RMS width of the Gaussian calculated from the interferogram … 6.2 Accuracy Accuracy is related to the capability to precisly fit the interferogram with the Gaussian function …

104. What do the standards say? Ref. IEC 61282 Fibre Optic communication system design guides – Part 9: Guidance on PMD measurements and theory

105. Measuring PMD FTB-5500B: l Highest accuracy and resolution l Ideal for 10G-40G compliance & certification l Source Shape insensitivity l Test the whole link l EDFA, OADM testing l Fast measurement time l Powerful but simple software l Same source as FTB-5800 CD Analyzer

106. Polarization Optical Time Domain Reflectometer October 2007 Mike Harrop mike.harrop@exfo.com

107. What to do with a link with high PMD? Frequent PMD problems (not measured when built) Need a way to find high PMD sections: PMD TOT =  N (PMD N ) 2 Example: 15ps, 2ps, 1ps, 6ps 225ps 2 + 4ps 2 + 1ps 2 + 36ps 2 = 266ps 2 266 1/2 = 16.31ps Find the 15ps section, replace it, problem solved…

108. Birefringence & Mode Coupling fast slow h fast slow fast slow fast slow Fibres with short (h) where Fast & Slow axis change frequently, tend to have low PMD Fibres with long (h) where Fast & Slow axis Change infrequently, tend to have high PMD

109. DOP Polarization-OTDR Quantitative  =  not measured  PMD value not measured DOP SOP1, DOP SOP2, h and L = all measured  Tendency for High PMD fiber under test Pulsed DFB Laser Detector Polarimeter /4 Polarizer /4 SOP 1 /SOP 2 4x2 OTDR acquisitions for characterizing SOP(z)

110. Example of Measurement and Validation (1) Link Length ~ 41 km PMD = 9.8 ps PMD coefficient ~ 3 ps/  km Cable opened and PMD measured with EXFO FTB-5500B PMD test set: 29 km, PMD = 4.3 ps 5 km, PMD = 17.4 ps 7 km, PMD = 6.9 ps 29 km 5 km 7 km High Contrast

111. Example of Measurement and Validation (2) Link Length ~ 41 km PMD = 9.8 ps PMD coefficient ~ 1.53 ps/  km Cable opened and PMD measured with EXFO FTB-5500B PMD test set: 6 km, PMD = 9.25 ps 35 km 6 km High Contrast

112. Bi-directional Measurements Quite similar results

113. Fiber Mapping in a Cable km0369121518212427303336394245485154576063 Fiber # PMD (ps) 1 17.6 2 219.4 3 312.4 4 43.7 5 58.4 6 68.8 7 78.2 8 815.7 9 92.5 10 28.1 11 9.5 Open and test Source: Connibear, A.B. and Leitch, A.W.R., Uni. Port Elizabeth, “Locating High PMD Sections of an Overhead Cable Unsing Polarization OTDR” Fiber # PMD (ps) 40.6-49.6km Replace and retest fiber# 17.61.7 219.418.52.9 312.47.2 43.72.7

114. Questions?