A Study on Wavelength Converter Over WDM Network under Transmission Impairments

1. A Study on Wavelength Converter Over WDM Network under Transmission Impairments Hong-Ik Univ. Dept. of Electronic Engineering Ph.D Dissertation 2001.06.04 Song Jae Youn

2. Agenda DWDM Fundamentals Wavelength Converter Placement BER Formula Evolution New Routing Scheme 1 2 3 4

3. c . Df =- Dl l2 DWDM Fundamentals

4. The New Public Network Evolution of the Public NetworkA Changing World • Rapid pace of optical technology and new traffic demands are reshaping... • Huge bandwidth demand • Reliable transmission • New network architecture and standard

5. OC-12 OC-48 OC-192 Scaling Options Option 1: Overbuild Fiber Option 2: Upgrade SONET Option 3: Introduce WDM (Wavelength Division Multiplexing) OADM λ8 λ1λ2 ••• λ8 OADM : Optical Add/Drop Mux.

6. Optical Electrical Dense Wavelength Division Multiplexing (DWDM) Transmission: Optical TDM and WDM Optical Electrical Time Division Multiplexing (TDM)

7. The Impact TDM : series WDM : parallel

8. c . Df =- Dl l2 What made WDM enable?

9. Opening Up New Wavelength Bands C-band (conventional) 80 channels 1530 - 1562 nm Loss L-band (long) 80 channels 1570 - 1620 nm water-peak EDFAs C-band L-band # of waves @ 50 GHz 350 460 80 80 1st 2nd 5th 3rd 4th (nm) 850 1260-1360 1365-1525 1530-1562 1570-1604 Fiber loss graph

10. ITU Grid • First phase WDM • 1310 nm, 1550 nm windows • Second phase WDM • 2 ~ 4 channels ; 1550 nm window • Spacing : 400+ GHz • Current DWDM systems • 16, 32, 40 channel ; 1550 nm window • Spacing : 100 GHz • Next-generation systems • 40 ~ 80 channels ; 1550 nm window • Spacing : 50 to 100 GHz

11. WDM Network Evolution WDM/Point-to-Point Transport –High Capacity Transmission Fixed WDM/Multipoint Network –Fixed Sharing Between Multiple Nodes OXC and OADM Reconfigured WDM/Multipoint Network Fiber Amplifier Wavelength Multiplexer/ Demultiplexer OADM(Optical Add/Drop Mux.) OXC(Optical Cross-Connect)

12. WDM Components Optical Amplifiers • Booster: • - Amplifies individual l’s prior to being multiplexed • - For compensation component loss • (In)Line: • - Amplifies aggregated l’s within the span • - For longer transmission • Pre: • - Amplifies aggregated l’s prior to being demultiplexed • - For compensation fiber loss Tx Rx In-line Amplifier Pre- Amplifier In-line Amplifier Power/Booster Amplifier

13. Pump Laser (980 or 1480 nm) WDM Components Optical Amplifiers-EDFA Erbium Fiber Optical Signals (1550 nm window) Optical Isolator Input Output Residual Pump Wavelength-Selective coupler + Transparency : to bit rates, modulation formats + Amplify waveband : Simultaneous regeneration of multiple(1530 ~ 1565nm) WDM signals + Low delay : No O/E, E/O conversion - Noise Generator : Inherent spontaneous emission is also amplified (ASE) and serves as “noise”

14. WDM Components Wavelength Converter • Reduce Blocking in WDM networks • Increase channel numbers • - the narrower the channel spacing, • the more channel crosstalk …* • - the main crosstalk is • the linear in- band crosstalk, … • the large channel spacing is necessary…** • Place wavelength converters Figure : in [36] Italian Network *[3]fiber optic communication technology(PrenticeHall, 2001), p.523 **[36]Nonlinear Optical Communication Networks, (John Wiley & Sons, New York, 1998), p.415

15. WDM Components Wavelength Converter-FWM • Categories • Optical gating WCs • - Using semiconductor optical amplifiers (SOA) in XGM • Interferometric WCs • - Using cross phase modulation (XPM) in SOAs and Mach-Zehnder/Michelson interferometers • - Nonlinear optical loop mirrors (NOLM) employing fibre loops or SOAs • Wave mixing WCs • - Implemented using SOAs or passive waveguides • - Four wave mixing (FWM) • - Difference frequency generation (DFG) • FWM converter • Strict format transparency • Better for high rate • Better for limited-range wavelength conversion • Low conversion efficiency

16. WDM Networks -broadcasting-and-select • All fiber technology • Inexpensive • Power split • Power limited Figure : Fiber optic communication technology(PrenticeHall, 2001)[3]

17. WDM Networks -wavelength-routed network • Optical path concept • Complex network • No power split • No receive broadcasted traffic • Wavelength reuse • More node connect Figure : Fiber optic communication technology(PrenticeHall, 2001)[3]

18. Agenda DWDM Fundamentals Wavelength Converter Placement BER Formula Evolution New Routing Scheme 1 2 3 4

19. Wavelength Converter in WDM Networks • Resolving Wavelength Conflicts • - Reduce # of wavelength • - Just local configuration • Network-Network Interfaces Between Subnets • - Connect WDM devices of different wavebands, multi-vendors • - Independent management and design

20. A Review of WC benefit-Analysis Studies(I) • Routing and wavelength assignment [11] • - 10~40% amount increase of reuse wavelength by using WC • Benefits of wavelength translation[12] • - In Mesh topology, WC is more efficient • WC is better in Mixing Traffic Networks[2] • - have shown not much difference in rings • - 30~40% reduction in capacity in common backbone mesh networks • Limited-range wavelength conversion[2] • -25% limited-range converter can make the same performance • with full-range converter • - limited conversion reduce hardware cost

21. A Review of WC benefit-Analysis Studies(II) • Sparse wavelength conversion[14] • - Not best All the node in network are convertible node • WC # above half of Node # are not efficient[26] Now, We Agree with using WC, What’s the Problem?

22. Problems of WCs in WDM network • Economic Problem • High cost • OXC complexity increase • Noise Problem • Makes ASE noise • Degrade signal quality • Effective blocking Problem • Hard fault : no free wavelength • Effective fault : signal quality is low WC reduce the hard fault, but increase the effective fault! How do we solve it?

23. Issues of WCs in WDM network • Where? - To place them along OXC where blocking may occur - NP-hard problem - Wavelength blocking reduction depends on network size, connectivity, available wavelength, traffic patterns and routing algorithm - Architecture complexity : Full conversion/Sparse conversion • Noise? - WC makes ASE noise as well as converting wavelength - Conversion efficiency is low - Cascade of WCs is an issue

24. A Review of WC Placement Studies(I) • Optimal placement based on dynamic programming[15] • + Optimal placement is better than random placement • - High computational complexity O( (node#)2 WC# ) : NP-Hard • Heuristic placement using ADD algorithm[24] • + Simple than exhaustive search • - At each stage, iteration of blocking probability for all nodes : still heavy calculation • - Based on the greedy algorithm, thus not optimal • Efficient algorithm for optimal converter placement[26] • +Auxiliary graphs is used, reduce calculation • - Not general rule for determining inner node/outer node • - Still heavy calculation • Efficient placement algorithm[28] • + Linear complexity • + If blocking probability of each path segment is same, blocking performance is best • - In case of the number of hop of the path is small(ex. 3,4 hops) • - In case of the blocking probability difference is large, cannot divide the segment

25. A Review of WC Placement Studies(II) • Heuristic placement of converter[2] • + Near optimal at non-uniform load network • - Need the statistics : • # of paths pass through • # of conversions • Nodal degree • How many lie on long path • If adjacent to the nodes with large transit traffic • - How we know the conversion # in a priori? • Impact of wavelength converters[29] • + A simple heuristic method, choose convertible node with high nodal degree • - Choose all node of nodal degree above a certain threshold, not realistic

26. A Review of WC Placement Studies(III) • Performance of FWM converter in SOAs[7] • + Deal with converter impairment, not ideal converter • - Use fixed routing • - Not use more general Q factor

27. My Objectives are to : • Calculation complexity : NP-Hard • Converter placement problem is MILP, Non-Convex * • Studies on calculation complexity[2][15][24][26][28][29] • Heuristic algorithm demands too much data[2], not general[28][29] • More realistic model • WC is not ideal, make ASE noise itself[7] • Different wavelength number link • Combination of amplifier noise, nonlinearities -> long link, few wave • Network interface links need more wavelengths than access links • As time goes, all links don’t upgrade simultaneously • After place WC, then? • Before/After place WC, routing algorithm needed a change • Dynamic routing > static routing ** * Ling li and Arun Somani,2000[28] ** Jaafar M. H. Elmirghani,2000[37]

28. A New Heuristic WC Placement Algorithm • A node transit much Traffic • - Maximum link utilization node • - Link utilization algorithm NO Multiple candidate? choose YES • A node have a longer path • - Longer distance, more efficient WC * NO Multiple candidate? YES NO • A node have a large node degree • - More node degree, transit more traffic Multiple candidate? random choose END * K. R Venugopal and M. Shivakumar,1999[25]

29. A New Algorithm : Rule 1 • The ground • The higher link utilization, the more blocking occur • In case of nodal degree, cannot put in node order efficiently • If blocking probability of converter sets are same, blocking performance is best * • Steps • Set path bet. Node pair based on Min. hop algorithm • Choose candidate nodes according Link utilization algorithm • Link utilization definition : • - CASE I :(Uniform link) number of link used of all paths • - CASE II :(Non-uniform link) link assigned wavelength # • * Suresh Subramaniam and Murat Azizoglu,1999[19], Ling li and Arun Somani,2000[28]

30. A New Algorithm : Rule 1 -link utilization algorithm • CASE I : Uniform link • CASE II : Non-uniform link : wavelength # of link (i, j)

31. A New Algorithm : Rule 1 -link utilization algorithm If node number is n in the path, Then choose the node c satisfied If cannot find the optimal c, find Where Also find satisfied Where Calculate Then choose the node satisfied, Else choose node

32. A New Algorithm- Results(I) • CASE I : Uniform links(Mesh network in[24]*) *Hiroaki Harai and Hideo Miyahara,1999[24]

33. A New Algorithm- Results(II) • CASE I : Uniform links(NSFNET in [24]*) Average blocking probability Suggested algorithm : 0.006005 ADD algorithm : 0.006465

34. A New Algorithm- Results(III) • CASE I : Uniform links(Mesh network in [25]*) *K.R Venugopal and M. Shivakumar,1999[25]

35. A New Algorithm- Results(IV) • CASE I : Uniform links(Mesh network in [26]*) *Sashisekaran and Arun,1999[26]

36. A New Algorithm- Results(V) • CASE I : Uniform links(NSFNET in [26]*)

37. A New Algorithm- Results(VI) • CASE II: Non-Uniform links (Korea topology in [27]*) Network modeling using OPNET** *오호일, 송재연, 김장복 대한전자공학회 추계학술대회논문집, 2000[27] **MIL3, Network Simulator

38. A New Algorithm- Results(VII) • CASE II: Non-uniform links (Korea topology in [27])

39. Algorithm Performance -Complexity (Node # = N, WC # = M)

40. Algorithm Performance -Complexity

41. Algorithm Performance -Heuristic

42. Algorithm Performance -Heuristic

43. A New Algorithm- Results • Low computation complexity • A few information a priori placement • More general placement algorithm • Arbitrary topology satisfied

44. Agenda DWDM Fundamentals Wavelength Converter Placement BER Formula Evolution New Routing Scheme 1 2 3 4

45. hn hn hn hn Transmission Impairments Optical Amplifier Noise • The Optical Amplifier (OA) amplifies signal irregularities • Amplifies noise(Amplified Spontaneous Emission) as well Spontaneous Emission Stimulated Emission • It turns out the dominant contribution to the receiver noise comes from beating of spontaneous emission with the signal[8]

46. Transmission Impairments Wavelength Converter noise • FWM WC also make ASE noise itself • - noise power density value followed [7]* • Low conversion efficiency • - channel spacing(200GHz, 100GHz) [35]** : • upconversion {-7.5, -5.5}dB • down conversion {-15, -11}dB Figure : the typical input and output spectra of the converter[38] * conversion efficiency data 500μm-long SOA(BT&D 1100-1550)을 SOA FWM conversion efficiency model 에 적용[7] **Hideyuki Sotobayashi, Ken-ichi Kitayama, 1999[35]

47. BER formula evolution in WDM networks where

48. BER formula evolution in WDM networks The approximated noise current (thermal noise, shot noise neglected) and

49. BER formula evolution in WDM networks where photodetector responsivity :

50. BER formula evolution in WDM networks where