Space-Division Multiplexing Over Multimode Fiber

Space-Division Multiplexing Over Multimode Fiber Nicolas K. Fontaine Bell Labs/Nokia, 791 Holmdel Rd, Holmdel NJ Why do we want to transmit over multiple modes? What are the new challenges? How is this better than just using the fundamental mode? What do we need to implement SDM? Component integration of Amplifiers, Switches, and Multiplexers 1 Nokia 2016

Bell Laboratories: Crawford Hill Laboratory Small building with about 75 Researchers covering: Optical transmission and sub-systems Wireless communication Data and networking Mostly a system laboratory One Nobel prize in Physics in 1978 The Horn Antenna Discovery of 3 o K cosmic background radiation in 1963. Nobel Prize in Physics in 1978. Crawford Hill Laboratory on a map Crawford Hill Laboratory from the air Murray Hill (headquarter) Crawford Hill Robert Wilson Arno Penzias 2 Nokia 2016 2 COPYRIGHT 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.

Acknowledgments Bell Labs Team: Roland Ryf, Haoshuo Chen, Jin Cang, Bin Huang, and Amado V. Benitez Hebrew University of Jerusalem: Dan Marom, Tom Hamarty, Miri Blau, and his students University of Laval: Sophie LaRochelle, Y. Messedaq, L. Wang CREOL Collaborators: Rodrigo Amezcua-Correa, Bin Huang, Guifang Li UC Davis Collaborators: S.J. Ben Yoo, Binbin Guan, Burcu Ercan, Ryan Scott Others: Bob Tkach, Peter Winzer, Sergio Leon-Saval, R. J. Essiambre OFS Labs Sumitomo Electric Prysmian Group Cailabs 3 Nokia 2016 COPYRIGHT 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.

MASSIVE TRAFFIC GROWTH FROM ULTRA-LONG-HAUL TO CHIP-TO-CHIP 4 Nokia 2016

Serial Interface Rates and WDM Capacities Gb/s Tb/s Capacity Scaling in Fiber Telecom Modes 100? [P. J. Winzer, IEEE Comm. Mag. 26-30 (2010)] WDM 10 1 100 10? Key Milestones: Optical fiber Erbium-doped fiber amplifier Coherent detection Next: Space-division multiplexing 1 1986 1990 1994 1998 2002 2006 2010 2014 2018 2022 WDM systems available today up to ~ 40-70 Tb/s SDM systems will be Petabit 5 Nokia 2016

FIBERS FOR SPACE-DIVISION MULTIPLEXING Uncoupled multi-core fibers B.Zhu et al., ECOC 2011 T. Hayashi et.al., ECOC 2011 K.Imamura et al., ECOC 2011 H. Takara et al., ECOC 2012 Coupled multi-core and few-mode fibers Sakaguchi et al., OFC 2012 R. Ryf et al., OFC 2012 R. Ryf et al., OFC 2013 N.K. Fontaine et.al., Sum.Top. 2013 M. N. Petrovich et.al., ECOC 2012 C. Cia et al., Sum. Top. 2012 Use modes to increase capacity of a single fiber! 6 Nokia 2016 Courtesy R. Ryf

A Few Record Results using SDM Fibers Uncoupled core: Super-Nyquist-WDM transmission over 7,326-km seven-core fiber with capacity-distance product of 1.03 Exabit/s km, K. Igarashi et al., OE 22, p 1220 (2014) Component Integration: 19-core MCF transmission system using EDFA with shared core pumping coupled via free-space optics, J. Sakaguchi, OE 22, p 90 (2014) Coupled core (up to six spatial modes): 1705-km Transmission over Coupled-Core Fibre Supporting 6 Spatial Modes, R. Ryf et al., ECOC 2014 PDP 3.2 Multi-Mode Fiber (up to 15 spatial modes): 305-km Combined Wavelength and Mode-Multiplexed Transmission over Conventional Graded-Index Multimode Fibre, R. Ryf et al., ECOC 2014 PDP 3.5 Multi-Core Multi-Mode (over 100 spatial modes): 2.05 Peta-bit/s Super Nyquist-WDM SDM Transmission Using 9.8km 6 mode 19-Core Fiber in Full C-Band, D. Soma et al., ECOC 2015 PDP.3.2 But, SDM is not only about the fiber... 7 Nokia 2016 COPYRIGHT 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.

Scaling Capacity using Orthogonal Dimensions Polarization Space Physical dimensions Frequency Time Quadrature Encode data in optical communications systems using different physical dimensions Can increase capacity by combining multiple dimensions Space has not been used! [P. J. Winzer, Nature Phot. 8, 345 348 (2014)] 8 Nokia 2016 8

Ultra-High Capacity Single-Mode Optical Fiber (SMF) Transverse mode (2D) Transverse y x z Core Cladding Two polarization modes 0.16 db/km loss at 1550 nm (194 THz) Erbium doped amplifiers conveniently provide 5 THz at this bandwidth 60 TB/s over 5000-km distances (capacity limit) 9 Nokia 2016

Maximizing Capacity with Modern Fiber Transmitter and Receiver x pol. Tx Rx y pol. 1) Produce a multi-level quadrature modulated waveform using a DAC 2) Use digital pulse shaping to produce rectangular waveforms 3) Use both polarizations 4) Pack as many channels within the amplifier bandwidth 5) Coherent detect the optical field at the Rx for both polarizations 6) Rely on power digital signal processing remove dispersion, unscramble polarizations, equalize, and sample (no optical signal processing) 7) Powerful Forward-Error-Correction 1.0 Tb/s (1 carrier, 3 spectral slices, 43 GBd each) 127.9 Gbd PDM-16QAM 3000 km transmission. [Mardoyan,.et.al.,OFC 15) PDM 512-QAM 3 GBaud (27 Gb/s) [Okamoto et al., ECOC 10] PDM 256-QAM 4 GBaud (32 Gb/s) [Nakazawa et al., OFC 10] PDM 64-QAM 21 GBaud (128 Gb/s) [Gnauck et al., OFC 11] PDM 16-QAM 80 GBaud (320 Gb/s) [Raybon et al., PTL 12] PDM QPSK 107 GBaud (224 Gb/s) [Raybon et al., PTL 12] 10 Nokia 2016

Digital Coherent Detection in the Lab Record data complex optical field on a real-time oscilloscope Employ power offline DSP to remove all linear impairments on the span. x pol. y pol. Agilent 90000 Q-Series 160 GS/s @ 63GHz 90 deg I x A/D Hybrid Signal PBS LO laser 90 deg PBS Hybrid Q x I y Q y A/D A/D A/D PBS Polarization beam splitter LO Local oscillator A/D Analog-to-digital conversion DSP Digital signal processing LeCroy LabMaster 10Zi 160 GS/s @ 65 GHz 11 Nokia 2016 11

Why can t we continue to increase capacity in SMF? Optical Nonlinearity Optical Processing: Great for generating light in difficult to access spectral regions. Telecommunications: Bad because the new frequencies become crosstalk for data signals. 12 Nokia 2016

Capacity C [bits/s] The Nonlinear Shannon Limit for Optical Fibers C = B log 2 (1 + SNR) Maximum capacity SNR [db] Signal launch power [dbm] Distributed Noise Tx Rx Nonlinearities distort signal at high optical powers Quantum mechanics dictates a lower bound on amplifier noise [R.-J. Essiambre et al., Phys. Rev. Lett. (2008) or J. Lightwave Technol. (2010)] 13 Nokia 2016 Courtesy P. Winzer

Better to Scale Capacity with Parallel Spatial Paths Tx Tx Rx Rx Three Paths C = B log 2 (1 + SNR) Two Paths One Path Single Path: Double capacity requires exponential increase in power Two Paths: Double capacity requires twice the power 14 Nokia 2016

Power Efficient Transmission (TE-Subcom) Rather than using 11.8 db more power to increase capacity by 4x in one mode, Submarine systems are power constrained 14350 km Tx over 12 core fiber with over 100 TB/s capacity Over 2x improvement in capacity at the same power level compared to using only 1 core 800mW pump power for all 12 cores. 15 Nokia 2016

Space-Division Multiplexed (SDM) System Fiber Span SDM is to transmit and receive N signals through a link supporting multiple spatial paths. Scrambling is inevitable in multi-span systems! Spatial integration will provide cost savings over parallel SMF systems! Multi-core amplifiers sharing isolators, gain-flattening filters, and pumps Switches that use 1 mirror (group of pixels) to switch many modes Can novel fibers provide unique applications, or better performance than SMF? Where to deploy SDM technology and interoperability? Long-haul, short spans, or just components? 16 Nokia 2016 16

Spatial Integration reduces costs and complexity of parallel systems A new level of integration not available in SMF!! N SEPARATED SPATIAL PATHS TX TX TX TX TX TX N separate paths can obtain same capacity as N modes. Costs scale linearly with N. N SPATIAL PATHS IN THE SAME FIBER RX RX RX RX RX RX Integration comes at expense of mode mixing. N N coupled paths through the link. 17 Nokia 2016 <Change information classification in footer>

New Challenges in SDM Handling 10x increase in capacity Mode-Mixing fix with Multiple-Input Multiple Output (MIMO) DSP Similar to polarization mixing in SMF Fibers are not perfect waveguides over long distances Perturbations, splices, stress, acoustic effects Mode Dependent Gain and Loss Modes have different shapes and propagation constants Mode Dependent Group Delay Affects MIMO equalizer complexity Requires additional components beyond SMF systems Spatial multiplexers modal gain equalizers DGD equalizers MIMO 18 Nokia 2016 18

Few-mode-fiber (FMF) with 6 total modes Six spatial and polarization modes to place information. 19 Nokia 2016

Linear-Polarized (LP) Modes (Spatial Modes) LP 01 LP 11a LP 11b Few-Mode Fiber Y-pol X-pol Scalar approximation to vector modes. Six spatial and polarization paths. Boxes indicate near-degenerate modes expect strong coupling. 20 Nokia 2016

Mode-Mixing Eigenmodes do not exist over long distances Input Patterns Ideal Fiber Propagation Modes at Output I 1 I 2 I 3 LP Launch Fiber Propagation with Mixing Scrambled Outputs I 1 I 2 I 3 21 Nokia 2016

What does Mode Mixing do to Data? Input Output 2 x 2 MIMO DSP engine 1 nm OF gr EDFA 22 Nokia 2016 PC Y-polarization Mode 1 X-polarization Mode 2 Y-polarization Pol-div. 90 deg Hybrid PC LO offline rocessing Transmission Fiber Mode Mixing X-polarization I x Mode 1 LeCroy LabMaster 9 Zi Mode 2 offline processing Y-polarization X-polarization Need to know the complex optical field!

Multiple-Input Multiple-Output Processing I 1 = h 11 O 1 + h 12 O 2 + h 13 O 3 + h 14 O 4 + + h 1N O N Take orthogonal combinations of all outputs to recover the input. Matrix is adapted using feedback from a training algorithm or a blind algorithm. 23 Nokia 2016

Unscrambling Crosstalk with Multiple-Input Multiple-Output Processing I 1 is QPSK, I 2 is BPSK O 1 contains both I 1 and I 2 and O 2 contains both I 1 and I 2 Find correct amplitude (scaling) and phase (rotation). Inputs I 1 QPSK Received Signals O 1 Constant O 2 I 1 Recovered Signals MIXING + = I 2 BPSK O 2 Constant O 1 I 2 + = 24 Nokia 2016

Transmission (10 db/div) Transmission (10 db/div) Increasing Fiber Length Transmission (10 db/div) Differential Group Delay + Mode-Mixing No Coupling (uncoupled core fiber, parallel SMF systems) Output Mode 1 Mode 2 Weak Coupling (few-mode fiber) Mode 1 Mode 2 Strong Coupling (Square-root growth of DGD and MDL) Mode 1 Mode 2 25 Nokia 2016 Time

22.8-km, 15 Spatial Mode Multi-Mode Fiber 5 Mode Groups, Weak and Strong Coupling Graded index to manage DGD Mode Areas 120um 2 to 292um 2 26 Nokia 2016 26

Impulse Response after 5 km 10-mode fiber (4 groups) 15-mode fiber (5 groups) -20 db Crosstalk between groups Strong Mixing within groups 27 Nokia 2016 27

Ring Core Fiber (RCF) Strong Mode Mixing Example 2 4 RCF supports only the first radial mode - LP n1 modes. RCF has strong coupling between modegroups. Lower effective index difference between modegroups LP 01 mode has similar spatial profile to the LP 11 mode Two surfaces for scattering. 28 Nokia 2016

Transfer Function Evolution vs. Fiber Length LP 01 -LP 01 LP 11 -LP 11 LP 01 -LP 11 15 m 190 m 1.7 km Square-root growth 29 Nokia 2016

Increasing Separation Coupled Core Fiber The best transmission fiber low MDL and low DGD - All modes look the same Strong Coupling - TX impairments grow at square root rather than linearly Much more compact than uncoupled core fiber Improved tolerance to fiber nonlinearities due to strong mixing 30 Nokia 2016

30x30 MIMO Transmission Experiment 1) Nyquist shaped 30 GBd QPSK on 33-GHz grid 2) 30x delay decorrelation to emulate 30 independent spatial-polarization channels 3) 22.8-km MMF 4) x3 Time-multiplexed coherent receiver to measure phase and amplitude of 30 signals simultaneously = 2x30x40 GS/s = 2.4 TS/s 31 Nokia 2016 31

Characterization of 50m MMF span using a sweptwavelength interferometer 15-17 db MDL across C-Band 32 Nokia 2016 32

Equalizer Coefficients After 22.8-km 30 Modes by 30 Modes by 1000 delay taps = 1 Million equalizer coefficients!!! 33 Nokia 2016 33

Bit-Error-Rates After 23km 30 modes at 2 bit/s/hz = 60 bit/s/hz 34 Nokia 2016 34

Spatial Multiplexers Types: - Mode Selective - Mode-Group Selective - Scrambling Physics: - Adiabatic (Photonic Lantern) - Resonant (Directional Coupler) - Discrete/Bulk - Broadcast and Select (Phase Mask) Technologies: - Fiber, 3D Waveguide, Silicon Photonics, Planar Waveguide 35 Nokia 2016

Scrambling Transfer Matrix Excited Modes Input Modes Output Field = 37 Nokia 2016 Since the fiber scrambles modes, mode-mux is not always necessary! Must be Unitary orthogonal columns no mode dependent loss (MDL). COPYRIGHT 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.

Fiber Photonic Lantern Spatial Multiplexer Uncoupled Cores Lower index capillary New multi-mode core Isolated Fiber Cores Fiber Cladding Adiabatic taper from singlemode cores into multi-mode core. Lossless/Unitary. Splice directly to MMF. Scalable to a large number of modes. Can directly excite modes if dissimilar fibers are used. 38 Nokia 2016 COPYRIGHT 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.

3 Mode Lantern Modal Analysis Adiabatic transition from coupled core modes to fiber modes Lantern Modes Coupled core modes to fiber modes Cladding Modes 3 refractive indices Core Cladding Capillary 40 Nokia 2016

3 Mode Lantern Mode Analysis Exciting the Cladding Modes causes loss Lantern Modes Cladding Modes 41 Nokia 2016

Dissimilar Fiber Mode Selective Lantern Adiabatic transition from groups of cores to modes. Lantern Modes Cladding Modes 42 Nokia 2016 42

Photonic Lanterns are Scalable to Support Many Modes 3 Modes 6 Modes 15 Modes Requires the correct geometrical arrangement of cores to match the symetry of the modes. Fabricated by CREOL/UCF 43 Nokia 2016

Special preforms to hold the fibers Fabricated by CREOL/UCF 44 Nokia 2016

Output Patterns of 15 Mode Fiber Lanterns 2-3 db Insertion Loss Amado Velázquez-Benítez ECOC2015 Tu.3.3.2 45 Nokia 2016 45

Components: Erbium Doped Fiber Amplifiers Types: Multimode Multicore Multimode/Multicore Pumping Schemes Core Cladding Raman Challenges Mode dependent gain and noise 47 Nokia 2016 47

19-Core Amplifier Integration 48 Nokia 2016 48 J. Sakaguchi, OE 22, p 90 (2014) 19 cores can be processed simultaneously by a single free-space element Isolator Pump-combiner 10 pump laser diodes for 19 cores. Gain = 19.6-23.3 db, Noise Figure = 6-9 db Could replace 19 SMF amplifiers.

Cladding-pumped Six Mode-EDFA GI6MF MM Pump (105/125mm MMF) splice Supplied by ORC, Univ. Southampton 6M-EDF (2.5m) High index polymer (for pump dump) GI6MF 30 20 IN Intermediate fiber LP01 LP11 LP21 LP02 Low index polymer 25 20 Intermediate fiber OUT Y. Jung et al., Opt. Express 22, 29008 (2014). Gain [db] 10 15 10 NF [db] 0 1530 1540 1550 1560 1570 1580 49 Nokia 2016 Wavelength [nm] 5 0 Two fully-fiberized boxed optical amplifiers ~20dB average modal gain and ~3dB MDL Roland Ryf, et al., ECOC2015 Tu.3.2.2

Cladding Pumped Multi-Core EDFA pump combiner 6-core EDF 6-core CC- MCF Measured at 0dB input and 8W pump Haoshuo Chen et al., ECOC2015 We.1.4.2 50 Nokia 2016

Annular Clad Three-Mode EDF 18 Spatial Channels Small Cladding increases the pump intensity in the cores! 51 Nokia 2016 Parameter Value Ø Core 16.0~17.0 µm Λ core-core 62 µm Ø Inner-Clad 85 µm Ø Ring-Clad 170 µm NA Core 0.104 NA Inner-Clad 0.11 NA Ploy-Clad 0.4 A eff,lp01 168 µm 2 A eff,lp11 179 µm 2 ρ Er 2.8 10 25 m -3 Γ s,lp01 0.92 Γ s,lp11 0.8 [Cang Jin ECOC 2015 PDP, Haoshuo Chen Nature Photonics]

Multimode Optical Amplifier Supporting Many Modes Low-Index Polymer Design Optimizations: Output power Mode-dependent gain Noise figure Low-splice loss to transmission fiber Pumping efficiency # of modes (10 and 15) Excited Modes (approximately 28) 52 Nokia 2016 <Change information classification in footer>

Modal Gain Equalization Core Modes doping Doping Avoid complicated mode-equalization schemes: Pump control Complex doping structures (ring, multi ring, structure) External mode equalizers All modes have nearly the same overlap with the gain. Uniformly pumped bulk material Purposely design the core to support more modes than needed Amplified modes confined well inside the core 53 Nokia 2016

Gain Vs. Number of Modes Supported 54 Nokia 2016

Improving Cladding Pumped Amplifiers Noise Figure Pump Intensity >> Large enough to invert the gain Output Power Signal intensity at saturation << Pump intensity Reduce Cladding Area Increases pump intensity Improves noise figure and gain tilt Enlarge Core Area Reduces signal intensity Increases saturated output power Optimized for Core Pumping Optimized for Cladding Pumping 55 Nokia 2016

Transfer Matrix Measurements 20x20 matrix with complex coefficients across all of frequency Use singular value decomposition to calculate orthogonality of matrix rows Background MDL from the multiplexers is around 8 db. 56 Nokia 2016

Multi-Span 10 Mode Transmission Full C-band 121 km To be presented at ECOC 2016 57 Nokia 2016

Wavelength Selective Switches (WSS) Output Ports Input Ports Switching States: #ports #wavelengths. 100 wavelength channels, 20 ports = 2000 Throughput: #wavelengths spectral efficiency per port/fiber Treat modes as inseparable super-channel (joint switching). How to achieve a throughput or switching state improvements over SMF WSS? How to connect WSS to the many different types of SDM fibers? How to remove spatial mode dependence? 58 Nokia 2016 58

Steering Dimension 1 M Switch Fourier Transform f f ommon Fiber Array Microlens Array Fourier Transform Lens Mirror or LCoS Fourier transform converts angle steering (linear phase) into positional shifts. A big beam on the LCoS reduces the minimum port-to-port steering angle. 59 Nokia 2016 COPYRIGHT 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.

1x10 for 3 Mode Fiber with FMF Interfaces Ryf et.al, ECOC 2013, PDPD1.C.4 Pros: Low loss from direct interface to FMF. 10 ports with 3 spatial modes. Throughput equivalent to 30 port SMF WSS Cons: Modes have different passband shapes, and steering induced xtalk. 60 Nokia 2016 60

Passbands with Multiple Modes LP 01 LP 11 Passband effects Amplitude Phase Mask [M1(x)M2(x)] convolved with the Mask(x) Frequency (a.u.) 62 Nokia 2016

Joint WSS with SMF Interfaces (OFC 2015) SMUX 3x19 Grating Converging Lens Anamorphic Optics and Polarization Diversity LCoS Fibers in rows are switched identically. Switch throughput can be increased by factor of 3, without additional steering. All modes see the same filtering function! 63 Nokia 2016 63

WSS on the Optical Bench 3x19 Fiber Collimator Array 64 Nokia 2016 64

Spectral Performance 100 GHz Interleaver 5 THz spectral coverage 0.5 db bandwidth of 78 GHz 1-2 db variation in insertion loss between jointly switched channels Throughput equivalent to a 57 port WSS (3x19) 65 Nokia 2016 65

SDM Heterogeneous Network with Spatial Multiplicity of 6 3 Types of SDM Fibers 2 Types of Amplifiers 2 Joint WSSs 66 Nokia 2016 66

Conclusions SDM is the next generation for optical communications! SDM has new challenges: Unknown fibers 10x capacity Modes have different shapes MIMO New components SDM components need to provide cost advantage over SMF systems: WSS should use smart joint switching to increase switch throughput. Amplifiers that share components to reduce costs. 67 Nokia 2016 67