Optical Fiber Communication: evolutions before Our Eyes Guifang Li CEOL, The College of Optics & Photonics University of Central Florida CEOL Industrial Affiliates Day March 1, 015 1
009 Nobel Prize in Physics
Outline Introduction The Optics Happy Era Digital Coherent Optical Communication How, Why? Capacity Limits of SMF Space-Division Multiplexing Critical Challenges & Solutions Future Challenge & Prospects Conclusions 3
Channel Capacity Why Optical Shannon Capacity Formula: bits/s Hz C = W log (1 S / N) Bandwidth Spectral Efficiency bits/s/hz Signal-to Noise atio 4
Optical Communications Why Fiber Antenna Gain +10 db 1011001 Free Space Propagation Loss 300 db (Diffraction) Antenna Gain +10 db Link Loss =60 db 1011001 D=35 cm 80 km Fiber Loss (db/km) 1. 0.9 0.6 0.3 0 O wavelength E 80 km(nm) S C L SMF AllWave fiber 1300 1400 1500 1600 Wavelength (nm) Link Loss =16 db What have we been doing all those years?
Optical Communication Economy TX Cladding Core X Everyday Life E-Commerce Optic Communication System Tele-conf Intelli-Transp. Internet Internet Tele-Med Dist. learn 6
Introduction Traffic Demand Internet traffic will increase in the foreseeable future 10 years~100x! 10x 7
Introduction Optical Fiber Everywhere @Home HDMI Cable Active Optical Cables 8
Introduction Optical Fiber Everywhere Around the globe 9
Introduction Tech Evolution EDFA Tingye Li, Herwig Kogelnik, Alan Willner
Introduction Noises & Distortions Direct Detection 0 c Et () Ate () C =W log (1 S / N) j[ t ( t)] Photodetector L I t A t e A t [ 0 ( )] () () j t c t () Pre Amplified Direct Detection Optical Pre Amplifier Coherent Detection L Photodetector Data, E d (t) LO, E LO (t) * e ELO() t Ed () t / delay Optical Hybrid * Im ELO() t Ed () t 11
Introduction Noises & Distortions C =W log (1 S / N) Noises/Modulation Formats Before 1980 Intensity Modulation Direct Detection (IMDD) Thermal Noise Limited: Sensitivity N=1000s Photons/bit 1980 1990 TX X Phase Shift Keying (PSK) with Coherent (Homodyne) Detection Shot Noise Limited w/o amplifiers: Sensitivity N=9 Photons/bit for BPSK ASE LO Beat Noise Limited w/ amplifiers: Sensitivity N=18 Photons/bit for BPSK 1990 005 Intensity/Phase Modulation using Direct Detection with Optical Preamplification Signal ASE Beat Noise limited: Sensitivity N= 39 photons/bit (IM); N= 0 photons/bit (DPSK) Since 005 Digital Coherent Optical communication 1
eason 1 for Demise of Coherent Polarization and Phase Management Polarization after a few meters of fiber propagation is uncorrelated with the input polarization. Phase locking is very challenging! Laser Diode Phase Modulator 0 j 0 E() t A exp j () t e Tx :Tx Laser Phase Tx t j 0 Et () A0 exp j s() t Tx() t e Binary Phase-Shift Keying 0, for bit 1's and 0's s t Conventional Phase Locking Techniques: 1. Injection Locking. Phase Locked Loops
eason : Erbium-Doped Fiber Amplifier (EDFA) Ease and obustness of Pre amplified IMDD to achieve 39 photons/bit Pump Laser Er Doped Fiber WDM n = 0.7 1.8 1.6 1.4 g max Gain (db/m) 1. 1 0.8 0.6 0.4 0. g min 0 1510 1530 1550 1570 1590 Wavelength (nm)
The Wavelength-Division Multiplexing (WDM) evolution DS3 OC3/1 OC- 48OC- 48OC- 48OC- 48OC- 48OC- 48OC- 48OC- 48 Conventional Transmission - 0 Gb/s 40km 40km 40km 40km 40km 40km 40km 40km 40km 1310 1310 1310 1310 1310 1310 1310 1310 LTE PT 1310 PT 1310 PT 1310 PT 1310 PT 1310 PT 1310 PT 1310 PT 1310 LTE LTE 1310 1310 1310 1310 PT 1310 PT 1310 1310 PT 1310 LTE LTE 1310 1310 1310 1310 PT 1310 PT 1310 1310 PT 1310 LTE 1310 1310 1310 13106 x 101310Gb/sPT 1310 1310 PT 1310 LTE 1310 1310 1310 1310 1310 1310 1310 PT 1310 LTE 1310 1310 1310 1310 1310 1310 1310 PT 1310 LTE 1310 1310 1310 1310 1310 1310 1310 PT 1310 DS3 LTE 1310 1310 1310 fiber 1310 1310 1310 1310 LTE PT 1310 PT 1310 PT 1310 PT 1310 PT 1310 PT 1310 PT 1310 LTE 1310 1310 1310 1310 1310 PT1310 PT1310 LTE PT PT PT PT PT PT PT 1310 1310 1310 1310 1310 1310 1310 LTE PT PT PT PT PT PT PT STM-64... DS3 10 km mux O A 10 km 10 km Optical Amplifiers and WDM - 0 Gb/s In Each Direction: 1 fibers 1 fiber; 36 regenerators 1 optical amplifier WDM: Wavelength Division Multiplexing O A demux LTE DS3 LTE LTE LTE LTE 1310 LTE PT 1310 LTE 1310 LTE PT 1310 LTE PT LTE OC- 48OC- DS3 48OC- OC3/1 48OC- 48OC- 48OC- 48OC- 48OC- 48... STM-64 Circa 1994
Introduction Noises and Distortions Linear Distortions: Chromatic Dispersion: different wavelengths (colors, frequency) of light travel at different speeds in the fiber Dispersion ps nm km Increases with the distance and bit rate 101 Bit Pattern After Traversing A Length Of Optical Fiber 1 0? 1 Polarization-Mode Dispersion Bit Period Nonlinear Distortions: ' P Kerr Nonlinearity n n0 ni n0 n A kn 0 NL PL where A eff eff Increases w/ distance and power Decrease w/ A eff 16
Optical Fiber Polarization-Mode Dispersion (PMD) PMD Statistical Nature of PMD 1 3 4 n Mode-coupling at random locations with random strength PMD vector goes through a random walk. PMD is statistical due to environmental fluctuations. In 3D space, PDF of a random variable through random walk is Maxwellian. The mean square value of DGD scales with fiber length. Mean DGD scales w/ square root of fiber length, in units of ps / km. PMD is also frequency depend. Frequency-dependence of PMD is called high-order PMD. Frequency of occurrence Maxwellian distribution of the instantaneous DGD PMD 3.5PMD Instantaneous DGD (ps) Prob.(DGD>3xPMD) = 4x10-5 = 1 min/year Prob.(DGD>3.5xPMD) =10-6 = 3 sec/year
Outline Introduction The Optics Happy Era Digital Coherent Optical Communication How, Why? Capacity Limits of SMF Space-Division Multiplexing Critical Challenges & Solutions Future Challenge & Prospects Conclusions 18
Is there an Easy Solution to Fiber Dispersion? Dispersion (ps/nm.km) 0 10 0-10 O E S C L SMF DSF 1300 1400 1500 1600 Wavelength (nm) How About Dispersion- Shifted Fiber?
Fiber Is Tricky: Nonlinearity The index of optical fiber depends on the intensity of the light inside n j n j n L 0 P A eff dz L L 0 P z NL dz 16 n 3.10 cm / W The phase of optical signal after fiber propagation depends on its own intensity: Self Phase Modulation (SPM) k n P where 0 A eff Increases with distance and power SPM Dispersion Timing Jitter Time Frequency Time By the same token, in a WDM system, the phase of optical signal after fiber propagation depends on the intensity of all other channels: Cross Phase Modulation (XPM) NL, j Leff P j M m j P m
Four Wave Mixing in Fiber k j i ijk L k j i eff ijk ijk e P P P L D P 3 1 / sin 4 1 L L e L e d dd c D c k j i k j k i ijk k j i ) ( ) )( ( Three waves at different frequencies can mix to create a fourth wave: The power of the fourth wave: Is maximized when fiber dispersion is zero
Dispersion & Nonlinearity Management NZDSF DCF NZDSF DCF Dispersion 0 Dispersion (ps/nm.km) 0 10 0-10 100 km 0 db O Distance E 5 km 1 db DCM S SMF ~4 times lower dispersion than SMF Non-Zero Dispersion- Shifted Fiber (NZDSF) 1300 1400 1500 1600 Wavelength (nm) C DCF L 100 km 0 db 5 km 1 db Net zero dispersion for the span Non-zero local dispersion to suppress nonlinear effects Dispersion- Compensation Module
Non-Zero Dispersion Shifted/ Medium Dispersion Fibers +NZDSF -NZDSF +NZDSF -NZDSF 5 km 0.45 db/km 115 m 5 km 0.08 db/km 40 m +NZDSF -NZDSF 3 Mukasa et. al, J. Opt. Fiber. Commun. ep. 3, 9 339 (006) 3
Nonlinearity Tolerance and Intensity Waveforms: NZ vs Z Non eturn to Zero NZ NZ eturn to Zero Z LD LD NZ AM AM f c AM After transmission over 960 km SSMF NZ ( dbm) Z ( dbm)
Nonlinearity Tolerance and Phase Waveforms: Z vs CSZ NZ f c Z LD AM AM 0 000 00 0 000 NZ f c / Carrier Suppressed Z CSZ LD AM 0 0 π 0 π 0 0 0 π 0 After transmission over 960 km SSMF Z ( dbm) CSZ (5 dbm) Manipulating phase can improve nonlinear tolerance!
Differential Phase-Shift Keying (DPSK) Sensitivity E Im DPSK: 0 photons/bit OOK: 39 photons/bit Less power needed for DPSK reduced NL. E e E Im E e Constant Amplitude Every bit experiences the same deterministic SPM. On-off keying (OOK) Binary phase shift keying (BPSK) LO Free 1 Bit DPSK Encoded Data (Optical) Delay +- Demodulated Data cos( )
Introduction Coherent: nd Coming Coherent +D N=18 +D, -D N=0 D=0 N=1000s EDFA +/-D=0 N=39 N=9 015 Tingye Li, Herwig Kogelnik, Alan Willner
Outline Introduction The Optics Happy Era Coherent Optical Communication How, Why? Capacity Limits of SMF Space-Division Multiplexing Critical Challenges & Solutions Future Challenge & Prospects Conclusions 8
Coherent Communication Encoding: information is encoded on the electric field of the lightwave. Decoding: measure the electric field. Data, E d (t) LO, E LO (t) (0,0) * e ELO () t Ed () t (1,0) E Im (0,1) (1,1) E e QPSK 16 QAM / delay 90 0 Optical Hybrid * Im ELO () t Ed () t The local oscillator has to be matched in phase and polarization with the incoming data. Difficulties in optical phase locking and polarization tracking were the main obstacles for coherent, the 1 st coming. ecent successes in coherent transmission relied on DSP-based phase and polarization management 9
Digital Coherent Comm. 1. DSP-Based Phase Management: Data, E d (t) LO, E LO (t) Free-running / delay Et () Aexp j d() t n() t 3 d 0,,, :Data Phase 4 : Phase Noise of eceived Signal * e Ed() telo() t * Im Ed() telo() t Example: DSP Algorithm for QPSK n 4 A j d t n t exp 4 ( ) 4 ( ) with 4 ( t) m 4 A exp j 4 n( t) d arg(.) A D C A D C () t () t D S P (.) 4 Phase Estimation (.)/4 d n 4 ( t n ) + n () t + How? Output d () t - Output. Noe, J. Lightwave Technol. 3, 80 (005). 30
Digital Coherent Comm.. DSP-Based Polarization Demultiplexing How? Optical Communication E x ' E x Tx PBC PBS x E y E ' y andom polarization rotation fiber E E ' x ' y J J 11 1 J J 1 E E x y Antenna ' ' Ex E x Ex i i1 i1 E y E y E y i i i Tx Ex J J J, : A Set of Training Symbols ' Wireless Communication Ey Multiple-Input-Multiple-Output (MIMO) Han & Li, Optics Express 005 x 31
Digital Coherent Comm. Why? (8) Coherent +D N=18 D=0 N=1000s DSP can perform a number of other EDFA functionalities D=0 better than or impossible N=39 for optics in WDM systems N=9 +D, -D N=0 015 Tingye Li, Herwig Kogelnik, Alan Willner 3
Digital Coherent Comm. Why? 1. Digital coherent communication enables electronic dispersion compensation Dispersion governed by the linear Schrodinger equation: A j A 1 A Time Domain: 3 0, z 6 3 t t 3 Frequency Domain: j A z, A j 1 3 z 6, 0, A z A e 3 j j 6 3 3 Dispersion is an all-pass complex filter, on the E-field of light, with a transfer function given by H( ) e z j j 6 3 3 z
Digital Coherent Comm. Why? 1. Digital coherent communication enables electronic dispersion compensation (EDC) Transfer function of dispersion : H( ) e j j 6 3 3 z H 00 150 100 50 0-50 -100 To reverse the effect of dispersion, 1. Detect E-field of received signal (thus coherent detection) * H H -150-00 -10-5 0 5 10. Apply a filter with M. G. Taylor, IEEE Photon. Technol. Lett. 16, 674 (004). ht () 1 H Tap-Delay Line Filter 34
Digital Coherent Comm. Why? Benefits of electronic dispersion compensation (EDC) SMF DCF SMF DCF +D D +D D EDC Electronic Dispersion Compensation: Eliminate the need for DCFs (small effective area)less nonlinearity improved performance Fewer amplifiers reduced noiseimproved performance No DCF and fewer amplifiers reduced cost 35
Digital Coherent Comm. Why?. Enables electronic compensation of other linear distortions/impairments PMD compensation using a matrix of tap-delay line filters, instead of simple Jones Matrix. x CD comp h xx h xy + Frequency offset Carrier recovery Decision circuitry eal Imag h yx y CD comp h yy + Frequency offset Carrier recovery Decision circuitry eal Imag Non-ideal frequency responses of all components in the transmitter/receiver Courtesy: Seb Savory, Optics Express Feb 007. 36
Digital Coherent Comm. A Milestone Nortel Coherent DSP Chip 10 Gsymbol/s QPSK with Pol. Multiplexing => 40 Gb/s eal-time eceiver 193 nm CMOS 10 Million Gates 1 trillion operations per second 100 Engineers; 3 year effort Courtesy: Kim oberts, Optics Express January 008.
Digital Coherent Comm. 3. Electronic nonlinearity compensation: Digital Back Propagation (DBP) Why? z A Nˆ Dˆ A ˆ j 1 D 3 3 t 6 t Nˆ zstep Nˆ z step 3 Nˆ j A Nˆ z step NL Propagation in eal Fiber A 0, t Dˆ z step Dˆ z step Dˆ z step Dˆ z step.. Dˆ z step Dˆ z step A z, t zstep zstep z step Nzstep N ˆ z step ˆ N ˆ z step A 0, t z ˆ step D z ˆ step D z ˆ step D z ˆ step D.. z ˆ step D z ˆ step D A z, t zstep z zstep A Nˆ Dˆ A A Nˆ Dˆ A z z step Propagation in Virtual Fiber 38
DBP Experimental Details 3 WDM Channels Loop Length: 160 km Modulation Format: BPSK Symbol ate: 6 Gsymbols/s Transmitter Channel Spacing: ~6.5 GHz All 3 channels can fit into the 13 GHz analog bandwidth of the realtime scope 3 channels are orthogonal, i.e., minimal linear cross talk Coherent eceiver 39
DBP Experimental esults: 760km NZ-DSF, P L =6dBm One-Step DBP Lnl Lwo = 80km = 750km Lfwm =100km Goldfarb& Li, PTL 008 40
Outline Introduction The Optics Happy Era Coherent Optical Communication How, Why? Capacity Limits of SMF Space-Division Multiplexing Critical Challenges & Solutions Future Challenge & Prospects Conclusions 41
Coherent Comm. Multiplexing Methods Capacity Limits SMF Capacity Limits Pol PolMux Mux DoF Nonlinear limit 5 b/s/hz /f 10 5 Spectral Efficiency (b/s/hz) Y-Pol SMF Capacity Limit: ~10THz 10b/s/Hz =100Tb/s (Nonlinear Shannon Limit) Capacity=BWSE X-Pol =500Gb/s 100 GHz Electronic Bottleneck 1 3 4 5 6 7 N WDM 10 THz EDFA Bandwidth Limit Bandwidth 4
Coherent Comm. Tech. vs Demand D. ichardson, et. al. Filling the Light Pipe SCIENCE 330(15):37,010 Demand Increase: 10x/4Yr SMF Capacity Saturation NL Shannon Limit 100Tb/s SMF Optical Communication will experience a Capacity Crunch ~ 00 Single Fiber Capacity equirement: 10x/7yr 43
Coherent Comm. Price of SE Shannon Capacity Formula: C =W log (1 S / N) Coherent increase S/N: logarithmic growth C log ( S / N) Increasing signal power in a single channel by a factor of M C log ( MS / N) = log ( M) log ( S / N) Transmitting same total power (M S) in M channels C Mlog ( S / N) 44
Outline Introduction The Optics Happy Era Coherent Optical Communication How, Why? Capacity Limits of SMF Beyond: Space-Division Multiplexing Critical Challenges & Solutions Future Challenge & Prospects Conclusions 45
Beyond Coherent A New Frontier Searching for a New DoF for Multiplexed Transmission. 1.SMF. FMF: Few-Mode Fiber Mux DoF 3. Multicore Fiber Cladding Core 1 Core Core N Spatial DoF: Mode Core 46
Beyond Coherent SDM Mode-Division Multiplexing: MDM Capacity # Modes FMF Space-Division Multiplexing (SDM) =MDM+Core Mux 1. MDM:. Core Mux: # of Channel # of Channels D=# of Modes D=# of Cores 3. SDM: # of Channel D=#of Mode Core MC-FMF SDM Goal: 100x single-fiber capacity increase 47
Outline Introduction The Optics Happy Era Coherent Optical Communication How, Why? Capacity Limits of SMF Space-Division Multiplexing Critical Challenges & Solutions Future Challenge & Prospects Conclusions 48
SDM Critical Challenge Core-Mux Core Mux: increasing multiplexed channels Due to mechanical properties of SiO,fiber cladding diameter is limited to ~50um; any larger will limit flexibility/deployability. Within the limited fiber cross-section, increasing # of cores leads to Small core-to-core distance Increased crosstalk Small core diameter Increased Nonlinearity Therefore # of cores is limited to ~0. Core-multiplexing cannot increase capacity 100x 49
SDM Critical Challenge Mode-Mux Mode Crosstalk in MDM 011011 010110 110010 11001 010110 011011 011011 110010 010110 011011 110010 010110 011011 Coupling Point Digital Signal Processing (DSP) 11001 010110 011011 110010 010110 011011 110010 010110 011011 Delay Distance 3 9 Operations DSP computational complexity Delay/Distance # of Channel 50
SDM Critical Challenge Mode-Mux Mode Mux: DSP Complexity Example 000km TX 30 modes X Compared with SMF system: Capacity increased: 30x DSP complexity increased:1,000,000x compared to EDC, unrealistic power consumption DSP complexity is a critical challenge,which determines the feasibility of MDM. 51
SDM Critical Challenge Solution 1 Increase Mode Crosstalk LP 01 LP 1 Fast Mode Slow Mode Low Crosstalk:DSP Complexity Distance=000 LP 01 Fast Mode LP 1 Slow Mode High Crosstalk: DSP Complexity Distance K. P. Ho and J. M. Kahn, J. Lightwave Technol., 01 =45 5
SDM Critical Challenge Solution Frequency-Domain Equalization Time-Domain: Crosstalk at different times Between different modes D computation Fourier Transform Frequency Domain: No crosstalk between different frequencies Crosstalk only between modes 1D computation Frequency-domain equalization can achieve orders of magnitude savings in computation N. Bai and G. Li, Photonics Techology Letters Vol.4 Issue 1. 1918 191(01). 53
SDM Critical Challenge Solution 3 Divide-and-Conquer FMF:1 core x6 modes MC-FMF: 3 cores x Modes DSP Complexity # of Channel Cladding DSP Complexity Channel /# of Cores For a 10-core fiber,dsp complexity can reduce by a factor of 10. 54
SDM Critical Challenge Solutions Optics + electronics 1. Strong coupling among MDM dimensions : wc L sc L. Multi core few mode fibers h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h h 1 core x 6 modes 3 core x modes 3. Frequency Domain Equalization Technologies to overcome mode crosstalk: combination of optics and electronics. Page 55
Outline Conclusions Digital coherent technology has brought the SMF capacity to the nonlinearity Shannon limit. Single mode fiber capacity crunch is coming. Space division multiplexing (for which coherent and DSP provides the foundation) can potentially be the disruptive technology. Fundamental research opportunities in SDM abound, both in terms of optics and electronics. Future applications in Fundamental mode transmission and SDM will make the next few years very exciting for optical communications research. 56