Flexible and high data-rate coherent optical transceivers

(1)

Flexible and high data-rate coherent optical transceivers

Citation for published version (APA):

Rahman, T. (2017). Flexible and high data-rate coherent optical transceivers. Technische Universiteit Eindhoven.

Document status and date: Published: 13/03/2017

Document Version:

Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.

• The final author version and the galley proof are versions of the publication after peer review.

• The final published version features the final layout of the paper including the volume, issue and page numbers.

Link to publication

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:

www.tue.nl/taverne Take down policy

If you believe that this document breaches copyright please contact us at: openaccess@tue.nl

providing details and we will investigate your claim.

(2)

Optical Transceivers

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de rector magnificus, prof. dr. ir. Frank Baaijens, voor een commissie aangewezen door het College voor

Promoties in het openbaar te verdedigen op maandag 13 maart 2017 om 16.00 uur

door

Talha Rahman

(3)

Dit proefschrift is goedgekeurd door de promotoren en de samenstelling van de pro-motiecommissie is als volgt:

voorzitter: prof.dr.ir. A.B. Smolders 1e _promotor: _{prof.ir. A.M.J. Koonen}

copromotor: dr.ir. H. de Waardt

leden: prof.dr. P. Bayvel (University College London) Prof.Dr.-Ing. P. Krummrich (TU Dortmund) prof.dr.ir. F.M.J. Willems

dr. A. Napoli (Coriant GmbH)

prof.dr. C. Peucheret (Universit´e de Rennes1)

Het onderzoek of ontwerp dat in dit proefschrift wordt beschreven is uitgevoerd in overeenstemming met de TU/e Gedragscode Wetenschapsbeoefening.

(4)

This PhD dissertation has been approved by the committee of following members: voorzitter: prof.dr.ir. A.B. Smolders

1e _promotor: _{prof.ir. A.M.J. Koonen}

copromotor: dr.ir. H. de Waardt

leden: prof.dr. P. Bayvel (University College London) Prof.Dr.-Ing. P. Krummrich (TU Dortmund) prof.dr.ir. F.M.J. Willems

dr. A. Napoli (Coriant GmbH)

prof.dr. C. Peucheret (Universit´e de Rennes1)

A catalogue record is available from the Eindhoven University of Technology Library. Title: Flexible and High Data-rate Coherent Optical Transceivers

Author: Talha Rahman

Eindhoven University of Technology, 2017 ISBN: 978-90-386-4231-4

NUR 959

Keywords: Coherent optical transceivers, High capacity modulation formats, Flexible data-rate systems, Flexible modulation formats, flexible-grid Wavelength Division Mul-tiplexing

Copyright c 2017 by Talha Rahman

(5)

(6)

And say “My Lord, increase me in knowledge”.

(7)

(8)

Flexible and High Data-rate Coherent Optical Transceivers

In today’s world telecommunication has become an essential part of our daily life. With the recent advancement in smart devices, high definition multimedia, large volume of file transfer and the advent of internet of things, data traffic demands are growing ex-ponentially. In particular threefold increase in global IP traffic is estimated over a time span of five years (2015 - 2020). Furthermore, the busy hour internet traffic is increasing more rapidly than the average internet traffic which implies that the network capacity demands will evolve to be more and more dynamic. The maximum capacity of currently deployed optical transponder systems is limited only up to 10 Tb/s employing 50GHz spaced 96 wavelength division multiplexing (WDM) channels which are modulated by polarization multiplexed (PM-)quadrature phase shift keying (QPSK) format and co-herent detection. Although this is a large amount of capacity per filter, the exponential growth in data traffic is driving the deployed fiber networks towards a capacity crunch requiring immediate attention.

Deployment of new fibers to accommodate the growing traffic demands is a costly and time consuming process. As a near-term solution, the utilization of deployed fiber infrastructure with flexible-grid WDM architecture, in contrast to the fixed-grid 50 GHz WDM spacing, has been proposed. The flexible-grid architecture recommends the al-location of spectral resources with a finer resolution, minimizing the spectral gaps and guard bands, resulting in a potential 30% network throughput improvement. This archi-tecture requires deployment of re-configurable optical add-drop multiplexers (ROADMs) supporting access to finer spectral slices of ≤12.5 GHz and an associated control plane design. Advanced flexible-grid ROADMs consist of several wavelength selective switches (WSSs) and can induce severe filtering penalties when neighboring channels add-drop occurs. In order to mitigate these tight optical filtering penalties, a novel ROADM design including optical spectral shaping is proposed for coherent quadrature amplitude modulation (QAM) formats. The integrated optical spectral shaping block inside a ROADM enhances the higher frequency components in a Nyquist filtered signal mitig-ating the loss of spectral power due to tight optical filtering. Experimental investigation of the technique showed improvement in performance of 32 GBd Nyquist filtered QAM signals transmitted over a 37.5 GHz WDM transmission grid. In particular the signal Q-factor was improved by ∼1.5 dB employing the proposed ROADM design when periodic wavelength selective switchs (WSSs) filters were encountered.

(9)

and improve power efficiency, higher order modulation formats having spectral efficiency (SE) ≥4 bit/s/Hz have gained significant interest. These modulation formats utilize amplitude as well as quadrature dimension to deliver higher capacity and efficient spec-trum utilization. With the well established 40 GbE and 100 GbE Ethernet standards, the next generation of optical transponders will converge towards 400 GbE and 1.0 TbE. Due to bandwidth limitations of available state-of-the-art electrical components, such high data-rates cannot be achieved with a single optical carrier. A concept of superchan-nel has hence evolved which propose to place multiple Nyquist filtered optical chansuperchan-nels closely together; treated by network equipment as a single entity. The superchannel is generated by an array of transmitters and is received by an array of receivers to achieve the desirable net data-rates (400 Gb/s, 1.0 Tb/s, · · · ). In order to keep the cost per transmitted bit and power consumption minimum, the number of subcarri-ers in a superchannel should be minimized; this requires the utilization of a higher order QAM format with high symbol-rates. In this thesis a digital pre-emphasis has been utilized mitigating the bandwidth limiting penalties of digital to analog converters (DACs) and electro-optical modulators. The optimum transmission symbol-rate and forward error correction overhead for a PM-16-level quadrature amplitude modulation (16QAM) based quad subcarrier 1.0 Tb/s systems have been experimentally determined. At the optimum transmission symbol-rate, the long-haul transmission of data-rates up to 1.0 Tb/s has been experimentally investigated over several types of fibers which include standard single mode fiber, G.652 (SSMF), large area- pure silica core fiber (LA-PSCF) and large effective area fiber, G.655 (LEAF). Keeping the number of subcarriers fixed to four, higher order modulation formats of PM-32-level quadrature amplitude modulation (32QAM) and PM-64-level quadrature amplitude modulation (64QAM) based terabit superchannels were also considered in order to achieve higher capacity over the C-band of the SSMF. The terabit system design is then evaluated over a field deployed fiber having a total length of 762 km. This field trial marked the first-ever field demonstra-tion of the PM-64QAM modulademonstra-tion format with the highest per subcarrier symbol-rate over the largest distance. Furthermore, record potential C-band capacities over a single mode field deployed fiber of up to 38.4 Tb/s were demonstrated.

As the optical fiber is a nonlinear medium, its maximum achievable capacity is bound by the nonlinear Shannon limit; which puts a limitation on the maximum power that can be launched in an optical fiber. As a result a limit is put on the maximum achiev-able distance and capacity. In order to mitigate the distortions due to signal-signal nonlinear interaction, several digital signal processing based methods have been pro-posed. However, their practical implementation is restricted due to both limited gain and very high complexity in a WDM scenario. In context of this thesis digital subcarrier multiplexing was explored for M-ary QAM modulation formats in a realistic WDM scen-ario. As the number of digital subcarriers in an optical signal increases, the symbol-rate per subcarrier is reduced which induces performance penalties in a coherent receiver digital signal processing (DSP) because the classical phase and frequency estimation algorithms depends on modulated data samples. Here a distributed pilot tones based frequency and phase estimation algorithm for single and multiple digital subcarrier sig-nals was proposed. The proposed algorithm is independent of modulated data, allowing for frequency and phase estimation and correction prior to signal de-multiplexing by a 2×2 multiple input multiple output (MIMO) equalizer. In addition the algorithm is feed-forward in nature making it feasible for practical implementation. Employing the proposed digital pilot tones based carrier recovery scheme, transmission performance of

(10)

digital subcarrier multiplexed signals was evaluated. Compared to the single carrier case, digital subcarrier multiplexed QPSK, 8-level quadrature amplitude modulation (8QAM) and 16QAM signals achieved an increase in maximum transmission reach of 10%, 13% and 7%, respectively, over a non-dispersion managed long-haul (LH) SSMF link.

Although higher-order modulation formats achieve a higher capacity, the achievable maximum reach for those is quite limited. For example, moving from 4 bit/s/Hz (PM-QPSK) to 8 bit/s/Hz (PM-16QAM), the capacity is doubled but the achievable reach is reduced by 75%. Hence, it is a very difficult decision for operators whether or not to upgrade to a newer transponder version. An ideal option is to have a transponder with a fine resolution in achievable reach and capacity to meet the specific link requirements. In this context a novel digital subcarrier multiplexed hybrid QAM scheme is presen-ted where different digital subcarriers are modulapresen-ted with a different order of QAM format depending on required capacity and reach. The proposed hybrid QAM scheme is an alternative to time domain hybrid QAM where different symbols in time frame are modulated with a different QAM format to achieve flexibility. The achievable flexibil-ity in data-rate and reach by subcarrier multiplexing hybrid QAM is further enhanced by employing phase conjugated subcarriers; where selected subcarriers are transmitted with their conjugated field on orthogonal polarization. The data signal and its conjug-ated counterpart suffer from the same nonlinear phase distortion if dispersion symmetry along the link is provided. At the receiver, digital coherent superposition is applied to mitigate nonlinear distortions and improve signal quality. The number of digital subcar-riers transmitted as conjugate pairs is controlled to enable capacity and reach flexibility leading to an ultra-flex transponder design. With this new transponder design employ-ing digital subcarrier multiplexemploy-ing, hybrid QAM and phase conjugated subcarriers an enhanced flexibility in spectral efficiency and reach is demonstrated with a fixed symbol-rate and a fixed forward error correction (FEC) code. In particular, data-symbol-rates ranging from 60 Gb/s to 245 Gb/s can be transmitted over distances of 15500 km to 2000 km, respectively. Consequently, a highly flexible data-rate and reach transponder design is experimentally demonstrated.

(11)

(12)

Contents xi

List of Figures xv

List of Tables xxiii

1 Introduction 1

1.1 Brief History . . . 1

1.2 Motivation . . . 2

1.3 Thesis structure. . . 4

1.4 Contributions . . . 5

2 Coherent Optical Transmission Systems 7 2.1 Fiber Optic Networks . . . 7

2.2 Switching in Optical Networks . . . 9

2.2.1 Arrayed Waveguide Gratings . . . 11

2.2.2 Optical switches . . . 12

2.2.3 Wavelength selective switches . . . 13

2.3 The Optical Fiber Channel . . . 19

2.4 Propagation Effects. . . 24

2.4.1 Power Loss in Optical Fiber . . . 25

2.4.2 Power loss compensation. . . 26

2.4.3 Linear Distortions . . . 35

2.4.4 Nonlinear Distortions . . . 41

2.4.5 Mitigation of Nonlinear Effects . . . 45

2.5 Transmitter and Receiver . . . 48

2.5.1 LASER . . . 48

2.5.2 Optical Modulator . . . 49

2.5.3 Spectral Shaping at Transmitter . . . 54

2.5.4 Modulation Formats . . . 58

2.5.5 Optical Receiver . . . 60

2.6 Summary . . . 73

3 Single Carrier High Capacity Long-haul Transmission Experiments 75 3.1 Digital Pre-emphasis . . . 76

3.2 100 Gb/s Transmission over Nonzero-Dispersion Shifterd Fiber (NZ-DSF) 82 3.2.1 Experimental Details. . . 83

(13)

3.2.2 Results and Discussion. . . 85

3.2.3 Conclusion . . . 89

3.3 200 Gb/s Transmission Employing QPSK and 16QAM Formats . . . 90

3.3.1 Experimental Details. . . 91

3.3.3 Conclusion . . . 95

3.4 Design of a 1.0 Tb/s Coherent Transceiver. . . 96

3.4.1 Experimental Setup . . . 97

3.4.2 Performance Improvement with DPE. . . 99

3.4.3 Optimization of Transmission Symbol-rate. . . 101

3.4.4 Spectral Spacing of Subcarriers . . . 105

3.4.5 Fiber Launch Power Optimization . . . 106

3.4.6 Transmission Performance over Various Fiber Types . . . 108

3.4.7 Transmission Performance with Hybrid Amplification . . . 110

3.4.8 Improvement with Digital Back Propagation . . . 112

3.4.9 Conclusion . . . 113

3.5 Power Allocation in Flex-grid Networks . . . 115

3.5.3 Conclusion . . . 118

3.6 High Capacity Transmission Employing PM-64QAM . . . 119

3.6.1 32×1.0 Tb/s Transmission over C-band . . . 119

3.6.2 96×400 Gb/s Transmission over C-band . . . 124

3.6.3 Conclusion . . . 125

3.7 Long-Haul Multi-Terabit Field Trial Employing PM-16QAM, 32QAM and 64QAM . . . 126

3.7.1 The Field Deployed Link . . . 127

3.7.2 Transmitter and Receiver Description . . . 128

3.7.3 Terabit Super-channels Configuration and Back-to-Back Charac-terization . . . 130

3.7.5 Conclusion . . . 138

3.8 Field Demonstration of 200 Gb/s and 400 Gb/s over Short Links . . . 138

3.9 Summary . . . 139

4 Digital Subcarrier Multiplexing and Flexible Data-rate Transmission141 4.1 Digital Subcarrier Multiplexing . . . 142

4.1.1 Carrier Recovery Employing Digital Pilot Tones . . . 144

4.1.2 Transmitter DSP and Insertion of Digital Pilots. . . 145

4.1.3 Receiver DSP and Pilot Tones based Carrier Recovery . . . 147

4.1.4 Simulation Analysis . . . 150

4.1.5 Experimental Evaluation of SCM Performance . . . 151

4.1.7 Conclusion . . . 162

4.2 Flexible Data-rate Transmission. . . 163

4.2.1 SCM Hybrid QAM . . . 164

(14)

4.2.3 DSP for phase conjugated subcarrier (PCSC) based Nonlinear

Mitigation. . . 166

4.2.4 Experimental Details. . . 168

4.2.6 Conclusion . . . 180

4.3 Summary . . . 180

5 Elastic Optical Networks 183 5.1 Penalties in Optical Nodes. . . 184

5.2 Optical Filtering penalties . . . 186

5.3 Mitigation of Optical Filtering Penalties . . . 189

5.3.1 Optical Spectral Shaping . . . 191

5.3.3 Conclusion . . . 195

5.4 Experimental Evaluation of Optical Spectral Shaping. . . 196

5.4.3 Conclusion . . . 199

5.5 SCM Hybrid QAM for ROADM Filtering Tolerance . . . 200

5.5.1 Transmitter and Receiver DSP Structure . . . 200

5.5.2 Experimental Details and Results. . . 201

5.5.3 Conclusion . . . 204

5.6 Summary . . . 204

6 Conclusions and Outlook 207

References 211

List of Publications 237

Acronyms 243

Acknowledgments 249

(15)

(16)

2.1 Hierarchy of Optical networks. Transoceanic links(>6000 km), ultra long-haul networks (>3000 km), Long-haul networks (>1000 km), Re-gional networks (>300 km), Metro networks (>100 km), Access networks

(<100 km). . . 8

2.2 A simple ROADM structure. . . 10

2.3 Distribution of different optical signals over fixed grid. . . 11

2.4 A 400 Gb/s superchannel composed of 4×100 Gb/s subcarriers over (a) Fixed grid occupying 200 GHz and (b) Flexible grid occupying 150 GHz.. 11

2.5 A 1×4 arrayed waveguide grating. FPR: free propagation region . . . 12

2.6 A planar lightwave circuit (PLC) based 4×4 switch architecture made from 1×2 switches. . . 13

2.7 A degree 4 ROADM with colorless add-drop ports. . . 14

2.8 Optical micro-electro-mechanical system (MEMS) based switch opera-tion. (a) Beam reflected by an optical micro-mirror, (b) Reflected beam is steered by tilting the micro-mirror.. . . 15

2.9 A liquid crystal on silicon (LCoS) cell or pixel. . . 15

2.10 Beam steering by LCoS pixels array achieved by applying differential phase difference along the light wave front. . . 16

2.11 Strudcutre and operation of a reflection grating. . . 16

2.12 Functional block diagram of a 1×4 WSS. . . 17

2.13 Experimentally measured transfer functions of 100 GHz wide optical multiplexers/de-multiplexers based on arrayed waveguide grating (AWG) and WSS (three adjacent channels). . . 18

2.14 Result of 50 GHz wide WSS filters’ cascade on effective passband bandwidth. 19 2.15 An optical fiber. . . 20

2.16 Electric field distribution along fiber corss section of radius a. . . 24

2.17 Attenuation (α) of a single mode fiber. . . 25

2.18 Structure of an erbium doped fiber amplifier (EDFA). . . 27

2.19 Energy levels of an Er3+ ion. . . 28

2.20 Emission and absorption cross sections of Er3+doped silica fiber co-doped with Al2O3and GeO2. . . 29

2.21 Variation in received OSNR with span length.. . . 31

2.22 Raman Scattering process in silica fiber. . . 32

2.23 Normalized Raman gain coefficient (gR/Aef f) for Silica fiber pumped with a single source at wavelength 1420 nm. . . 33

(17)

2.25 Signal power evolution along a single span in with (a) erbium doped fiber amplifier (EDFA) only, (b) EDFA and backward distributed raman amp-lifier (DRA), (c) Forward and backward DRA (d) Forward and backward

DRA plus remotely optically pumped amplifier (ROPA) . . . 36

2.26 Power evolution along a long-haul link with multiple spans employing EDFA only as well as EDFA and backward DRA. . . 36

2.27 Dispersion parameter (D) for a SSMF showing contributions of material dispersion, waveguide dispersion as well as the total dispersion. . . 37

2.28 Pulse spreading due to dispersion in a SSMF (D = 17_nm·kmps , Ts=100 ps). 37 2.29 Pulse delay due to birefringence in a short length of optical fiber. . . 39

2.30 Model of an optical fiber as a concatenation of several small birefringent elements with randomly aligned fast and slow axes. . . 40

2.31 Distribution of differential group delay (DGD) for a fiber with polarization mode dispersion (PMD) value of 0.5 √ps km. . . 40

2.32 Intra-channel XPM induces frequency shift in each pulse which leads to timing jitter after dispersion compensation. . . 43

2.33 Dispersion induces pulse spread in propagating pulses which interact and generate new frequency components which leads to generation of ghost pulses after dispersion compensation. . . 43

2.34 Power spectral density of laser output (Lorentzian line distribution). . . . 49

2.35 Structure of a Mach-Zehnder-Modulator (MZM). . . 50

2.36 Transfer function of a Mach-Zehnder-Modulator (MZM) with Vπ= 4.0V (Insertion loss is neglected). . . 52

2.37 Structure of an IQ-modulator made of two MZMs and a π/2 phase shifter. The MZMs are operated in push-pull mode. . . 53

2.38 A dual polarization IQ-modulator capable of modulating two orthogonal polarizations of laser light.. . . 53

2.39 Structure of an optical transmitter employing pulse shaping in digital domain and using DACs to to generate electrical drive signals for the IQ-modulators. . . 54

2.40 Comparison of different pulse shapes and their respective spectral widths. 56 2.41 Frequency response of a DAC employing zero-order-hold and its compar-ison to ideal low-pass response. . . 58

2.42 Frequency response of zero-order-hold and its pre-compensation filter. . . 58

2.43 Coherent QAM modulation formats and their achievable capacity. . . 59

2.44 Structure of a polarization and phase diversity coherent optical receiver. PBS: polarization beam splitter, LO: local oscillator, PD: photo-diode, ADC: analog to digital converter . . . 61

2.45 Different coherent optical detection schemes. . . 62

2.46 Baseband model of an optical transmission system. . . 63

2.47 Functional blocks of a DSP based optical receiver. . . 64

2.48 Butterfly filter structure used for cross-talk mitigation (polarization de-multiplexing) in coherent receiver DSP. . . 67

2.49 Impact of QPSK signal equalization based of different schemes. . . 69

2.50 A Digital phase locked loop structure used for frequency and phase es-timation. . . 70

2.51 Theoretical performance of 32 GBd PM-QPSK signal with differential as well as non-differential encoding. . . 71

(18)

2.52 QPSK performance for uncoded as well as coded case. NCG: net coding gain . . . 72

3.1 Architecture of effective sampling-rate and bandwidth doubling of DAC.

fsis the effective rate of DAC; individual DACs have

sampling-rate half of this value. . . 77

3.2 Experimentally measured DAC transfer function and effective number of

bits (ENoB). . . 78

3.3 DSP block diagram of an optical transmitter employing digital pre-emphasis

(DPE).. . . 78

3.4 System model considered to drive optimal DPE filter. . . 79

3.5 Signal spectra of a root-raised-cosine (RRC) pulse shaped signal without

and with DPE. . . 80

3.6 Simulation analysis of 8QAM and 16QAM signals with DPE. . . 81

3.7 Rs,max versus modulation format for ∆RSNR of 0.5 dB at bit error ratio

(BER) = 1 × 10−3. . . 81

3.8 Experimental setup with re-circulating optical loop. ECL: external

cav-ity laser, DA: driver amplifier, PME: polarization multiplexing emula-tion, ICR: integrated coherent receiver, WSS: wavelength selective switch,

LSPS: loop synchronous polarization scrambler. . . 83

3.9 power spectral density (PSD) of signal for varying RRC roll-off factor. . . 84

3.10 Back-to-back system analysis. . . 86

3.11 Back-to-back performance curves for 34 GBd PM-QPSK. . . 86

3.12 Optical launch power per channel optimization for EDFA only and

EDFA-Raman amplified WDM cases at a transmission distance of 2268 km. . . . 87

3.13 Pre-FEC BER of 34 GBd PM-QPSK signal for channel wavelength 1550.116 nm. 88

3.14 C-band performance with EDFA only amplification after 4500 km

trans-mission over nonzero- dispersion shifted fiber (NZ-DSF). . . 89

3.15 C-band performance with EDFA and backward Raman amplification after

6500 km transmission over NZ-DSF. . . 90

3.16 Experimental setup for 200 Gb/s WDM transmission. SSMF: standard

single mode fiber, G.652, RT scope: real-time scope. . . 91

3.17 Electrical eye diagrams for generating QPSK and 16QAM taken from a

sampling oscilloscope. . . 91

3.18 Post-FEC BER versus pre-FEC BER for different FEC overheads. . . 93

3.19 Back-to-Back characterization of 200 Gb/s signals with variable

symbol-rates and FEC-overheads (OHs). . . 94

3.20 Power per channel sweep for 200 Gb/s solutions employing PM-16QAM

format.. . . 95

3.21 Transmission of 200 Gb/s signals with different FEC-OHs employing

mod-ulation formats of PM-QPSK and PM-16QAM. . . 96

3.22 Experimental setup for 1.0 Tb/s super-channel WDM transmission. ECL: external cavity laser, DA: driver amplifier, PME: polarization multiplex-ing emulation, SSMF: standard smultiplex-ingle mode fiber, G.652, LA-PSCF: large area- pure silica core fiber, LEAF: large effective area fiber, G.655, WSS: wavelength selective switch, LSPS: loop synchronous polarization

(19)

3.23 Variation in Q-factor with symbol-rate in back-to-back configuration for

PM-16QAM format. . . 99

3.24 Variation in Q-factor with symbol-rate employing DPE in back-to-back

configuration for PM-8QAM, 16QAM and 32QAM formats. . . 100

3.25 Electrical eye diagrams for generating 8QAM, 16QAM and 32QAM taken

from a sampling oscilloscope. . . 100

3.26 DPE performance with varying symbol-rate after transmission over SSMF

and LA-PSCF for QPSK, 8QAM and 16QAM. . . 101

3.27 Net gain in Q-factor for various fiber types and modulation formats. . . . 103

3.28 Back-to-back performance characterization of 40 GBd PM-16QAM signal. 105

3.29 Impact of channel spacing on the Q-factor of PM-QAM formats. . . 106

3.30 Q-factor performance versus launch power per channel for single chan-nel, super-channel and super-channel with WDM cases of SSMF,

LA-PSCF and LEAF (EDFA only amplification). Markers: experimental

data, Dashed lines: interpolation . . . 107

3.31 Transmission performance over EDFA only amplified three different single mode fibers (SMFs) for single channel, super-channel and super-channel with WDM scenarios. Markers: experimental data, Dashed lines:

inter-polation of experimental data, Solid line: FEC threshold . . . 109

3.32 Q-factor performance versus launch power per channel for 1.0 Tb/s super-channel with WDM for SSMF, LA-PSCF and LEAF (EDFA only as well as hybrid EDFA-Raman (HER) amplification). Markers: experimental

data, Dashed lines: interpolation . . . 111

3.33 Q-factor performance versus launch power per channel for 1.0 Tb/s super-channel with WDM for SSMF and LA-PSCF having length per span 121 km and 164 km, respectively. Markers: experimental data, Dashed

lines: interpolation . . . 112

3.34 Q-factor performance versus distance for 1.0 Tb/s super-channel with WDM for different amplification options and span lengths. Markers:

ex-perimental data, Dashed lines: interpolation, Solid lines: FEC threshold . 113

3.35 Q-factor performance versus distance for 1.0 Tb/s super-channel with WDM employing single channel digital back propagation (DBP). Markers:

experimental data, Dashed lines: interpolation, Solid lines: FEC threshold114

3.36 Experimental setup for the evaluation of nonlinear effects in dynamic

traffic scenarios.. . . 116

3.37 Q-factor versus launch power per subcarrier without neighboring traffic

after 760 km over EDFA only amplified SSMF span. . . 117

3.38 ∆Power sweep of the test channel relative to fixed neighboring channels

power of 0.0 dBm per channel. . . 118

3.39 Experimental setup for the transmission of 32×1.0 Tb/s super-channels

over LA-PSCF. . . 120

3.40 Transfer function of DACs and IQ-modulators cascade corresponding to

the XI, XQ, YI and YQ channels.. . . 120

3.41 Spectrum of 36.96 GBd PM-64QAM without and with DPE. . . 121

3.42 Back-to-back performance and power sweep of 37 GBd PM-64QAM format.122

3.43 Q-factor versus transmission distance for WDM transmission of 37 GBd

(20)

3.44 C-band transmission of PM-64QAM modulated signals after a transmis-sion distance of 984 km. Fiber type: LA-PSCF, span length: 82 km,

amplification: EDFA + backward Raman . . . 123

3.45 Back-to-back performance and power sweep of 43 GBd PM-64QAM format.125

3.46 Constellation diagrams of 64QAM signal corresponding to fc=193.4 THz.125

3.47 C-band transmission of 43 GBd PM-64QAM signals after transmission

over 328 km of LA-PSCF. . . 126

3.48 Schematic of the field deployed transmission link. Left: Map showing the link and the location of amplifier nodes. Right: Description of the link

with span lengths and corresponding losses. Link 3 is EDFA only amplified.127

3.49 Estimated DGD corresponding to a long-term measurements of ∼96 hours

with each measurement taking ∼20 sec. . . 128

3.50 Internal structure of the HER amplifiers used in this field demonstration. 128

3.51 The structure of transmitter. . . 129

3.52 The receiver structure. . . 129

3.53 Theoretically required optical signal-to-noise ratio (OSNR) versus symbol-rate for achieving data-symbol-rates of 250 Gb/s (a) and 300 Gb/s (b). Dashed lines: Shannon limit, Solid lines: Constellation constrained capacity,

Markers: Experimental points . . . 131

3.54 Back-to-back performance of 41.2 GBd PM-16QAM before and after FEC.132

3.57 Launch power sweep for PM--16QAM, 32QAM and 64QAM after 762 km

of field deployed transmission link. . . 134

3.58 C-band transmision of 16QAM modulated subcarriers. Right: Transmit

(top) and received (bottom) constellations . . . 135

3.59 Long-term BER measurements for PM-16QAM modulated super-channels

(fc= 193.4 T Hz). . . 135

3.62 Transmission results with field deployed link plus EDFA only

ampli-fied SSMF spans. Right top: Received constellation of 32QAM after

3 additional spans transmission, Right bottom: Received constellation of

16QAM after 8 additional spans transmission. . . 137

4.1 Digital spectra of a RRC filtered 34 GBd signal with single and multiple

digital subcarriers. sc: subcarrier . . . 143

4.2 Constellation diagrams without (top) and with (bottom) phase errors for

QPSK (left), 8QAM (center) and 16QAM (right) signals at the receiver.

An OSNR of 25 dB and symbol-rate of 34 GBd is considered. . . 144

4.3 Transmitter DSP steps for digital subcarrier multiplexing (SCM) signal

generation with digital pilot tones. PS: pulse shaping, DAC: digital to

analog converter, DA: driver amplifier, N: no. of digital subcarriers. . . . 145

4.4 Spectrum of a single carrier as well as SCM signal with 4 digital

(21)

4.5 Receiver DSP steps for SCM signals including pilot tone based carrier

recovery. MF: matched filtering, N: no. of digital subcarriers. . . 147

4.6 Extraction of digital pilot tones at receiver with different bandwidths of

first order Gaussian filter. . . 148

4.7 Estimation of phase error employing digital pilot tones for subcarrier i.. . 149

4.8 Estimated phase with filter bandwidths of 0.5 MHz and 20 MHz for a

linewidth of 1 MHz. . . 149

4.9 Simulation analysis of digital pilot tones assisted carrier recovery for SCM

PM-QPSK, 8QAM and 16QAM. . . 150

4.10 Experimental setup for the LH transmission evaluation of SCM QAM

formats. . . 151

4.11 Received OSNR evolution with increasing number of loop circulations. . . 152

4.12 OSNR at transmitter output with increasing number of digital subcarriers.152

4.13 pilot-to-signal ratio (PSR) optimization. Modulation: QPSK,

Symbol-rate: 40 GBd . . . 153

4.14 Back-to-back performance curves for single as well as multiple subcarrier

PM-QPSK signals at symbol-rates of 34 GBd and 40 GBd. sc: subcarriers154

PM-8QAM signals at symbol-rates of 34 GBd and 40 GBd. sc: subcarriers155

PM-16QAM signals at symbol-rates of 34 GBd and 40 GBd. sc: subcarriers155

4.17 WDM power per channel sweep after 3800 km transmission for single as well as multiple subcarrier PM-QPSK signals at symbol-rates of 34 GBd

and 40 GBd. sc: subcarriers . . . 156

4.18 WDM power per channel sweep after 1900 km transmission for single as well as multiple subcarrier PM-8QAM signals at symbol-rates of 34 GBd

4.19 WDM power per channel sweep after 1140 km transmission for single as well as multiple subcarrier PM-16QAM signals at symbol-rates of 34 GBd

4.20 WDM ∆power sweep for 40 GBd PM-QPSK, 8QAM and 16QAM signals

at symbol-rate of 40 GBd. sc: subcarriers . . . 158

4.21 WDM spectra before and after transmission. . . 159

4.22 Transmission performance of different subcarriers of QPSK with symbol-rate of 34 GBd and 40 GBd. 1st and the 4th subcarrier are at the edges

while 2nd and 3rd are in the middle of SCM signal.. . . 160

4.23 WDM transmission performance of single carrier as well as SCM

PM-QPSK, 8QAM and 16QAM formats. . . 161

4.24 Pure and hybrid QAM modulation schemes. (a) SCM pure modulation format with four subcarriers, (b) SCM hybrid modulation format with four subcarriers, (c) single carrier pure modulation, (d) time domain

hy-brid modulation (TDHM). sc: subcarrier. . . 165

4.25 PCSC: A SCM spectrum of four subcarriers with Y-polarization of 1st and 4th subcarrier carrying conjugate of corresponding X-polarization

subcarriers. . . 166

4.26 Receiver DSP steps for PCSC processing. . . 167

(22)

4.28 QPSK performance with launch power per channel sweep for single carrier as well as SCM signals. In case of SCM QPSK, the gain in Q-factor can

be scaled in finer steps. . . 170

4.29 Transmission performance of 39 GBd PM-QPSK signals with varying number of digital subcarriers employing phase conjugation for nonlin-ear mitigation. PCSC: phase conjugated subcarrier, Right: Constellation

without (bottom) and with (top) digital coherent superposition (DCS) . . 171

4.30 Transmission performance of 39 GBd PM-8QAM signals with varying number of digital subcarriers employing phase conjugation for nonlinear

mitigation. PCSC: phase conjugated subcarrier, Right: Constellation

without (bottom) and with (top) DCS . . . 172

4.31 Transmission performance of 39 GBd PM-16QAM signals with varying number of digital subcarriers employing phase conjugation for nonlin-ear mitigation (total number of subcarriers = 2), Right: Constellation

without (bottom) and with (top) DCS . . . 173

4.32 ∆Power optimization for 4-8 hybrid format. Transmission distance =

2280 km . . . 174

4.33 ∆Power optimization for 8-16 hybrid format. Transmission distance =

1140 km . . . 174

4.34 Constellation constrained capacity for non-hybrid as well as hybrid QAM

formats. . . 175

4.35 Back-to-back performance of 4-8 hybrid modulation and theoretical curves

with ∆Power = 3 dB. . . 175

4.36 Back-to-back performance of 8-16 hybrid modulation and theoretical curves

with ∆Power = 1 dB. . . 175

4.37 Transmission performance of 39 GBd PM- 4-8 hybrid QAM signals without as well as with phase conjugation for nonlinear mitigation. PCSC: phase

conjugated subcarrier . . . 177

4.38 Transmission performance of 39 GBd PM- 8-16 hybrid QAM signals without as well as with phase conjugation for nonlinear mitigation. PCSC:

phase conjugated subcarrier . . . 178

4.39 Flexible capacity and reach employed at a fixed symbol-rate of 39 GBd. SCM non-hybrid and hybrid modulation formats are employed in com-bination of PCSC to achieve fine granularity in data-rate and maximum

reach. . . 179

5.1 Experimental setup for the evaluation of in-band cross-talk due to

non-ideal extinction ratio (ER) of add-drop multiplexers (ADMs). . . 184

5.2 OSNR penalty due to in-band cross-talk occuring due to add-drop of

WDM channels.. . . 185

5.3 Experimental setup for the evaluation of OSNR penalties due to reduction

in effective channel bandwidth. . . 186

5.4 The relation between the adjusted bandwidth and the measured

band-width of WSS (both the bandband-widths are dual-sided). . . 187

5.5 OSNR versus BER performance with changing optical filter bandwidth. . 188

5.6 OSNR penalty due to reduction in filter bandwidth. . . 189

5.7 Simulation setup employed in combination with experimentally measured

(23)

5.8 Evolution of transfer function with WSS cascade. . . 190

5.9 Number of cascaded WSS filters. . . 191

5.10 Proposed positions of wave-shaper (WS) for the mitigation of optical

fil-tering effects in ROADMs.. . . 192

5.11 A comparison of Hch(f ) (a cascade of 22 WSSs), HW S(f ) and Hdes(f ). . 192

5.12 OSNR penalty with increasing number of WSSs filters (the assigned chan-nel bandwidth is 37.5 GHz and the measured 3-dB bandwidth is ∼33 GHz). Performance without spectral shaping as well as with three cases of

spec-tral shaping are plotted. . . 193

5.13 Performance variations with varying random selection of cascaded WSS

filters. . . 194

5.14 OSNR penalty with increasing number of WSSs filters for a test channel symbol-rate of 32 GBd (the assigned channel bandwidth is 37.5 GHz and the measured 3-dB bandwidth is ∼33 GHz). Performance without

spectral shaping as well as with three cases of spectral shaping are plotted.195

5.15 Experimental setup for the evaluation of optical filtering penalties due to ROADMs as well as their compensation with a WS. WSS: wavelength selective switch, WS: wave-shaper, ROADM: re-configurable optical

add-drop multiplexer, ∆f: 37.5 GHz . . . 197

5.16 Evolution of effective channel filter after 1–10 ROADM passes. . . 198

5.17 Transfer function of the WS employed in each ROADM. . . 198

5.18 Variation in Q-factor improvement compared to no spectral shaping case

with varying pass-band attenuation. . . 198

5.19 Q-factor performance with increasing number of WSSs (2 WSSs per loop,

37.5 GHz per WSS). . . 199

5.20 Transmitter and receiver DSP structure for generating and detecting SCM

hybrid QAM signals. LPF: low-pass filter . . . 201

5.21 Experimental loop setup used for the evaluation of filtering penalties with

a 37.5 GHz wide WSS filter in every loop. . . 202

5.22 Back-to-back performance of SCM hybrid as well as non-hybrid 150 Gb/s

signal. . . 202

5.23 SCM signal spectra with increasing number of WSS passes. . . 202

5.24 Q-factor performance with increasing number of WSS filters for

(24)

2.1 Typical distances in an optical network hierarchy. . . 7

2.2 Different wavelength bands of optical spectrum. . . 26

3.1 Parameters of SSMF and LEAF. . . 82

3.2 Modulation formats, Symbol-rates and SE for net 200 Gb/s line-rate.. . . 92

3.3 OSNR penalty compared to theoretical limit at the FEC threshold for the

various investigated cases. . . 94

3.4 Characteristics of Transmission Fiber at λ=1550 nm. . . 98

3.5 Req. pre-FEC BER and Q-factor corresponding to different FEC-OHs.. . 102

3.6 Transmission length used corresponding to each modulation format and

fiber type for the optimization of FEC-OH. . . 102

3.7 No. of subcarriers and net symbol-rate per subcarrier required to achieve

1.0 Tb/s employing PM-16QAM format. . . 104

3.8 Summary of transmission performance of 1.0 Tb/s super-channel with

WDM. . . 115

3.9 Summary of field trial transmission experiment and demonstrated

poten-tial C-band capacity. . . 133

3.10 Characteristics of additional SSMF spans used for reach extension of PM-16QAM and PM-32QAM super-channels’ WDM transmission after

762 km field deployed link.. . . 137

4.1 Summary of QPSK transmission results with varying number of phase

conjugated subcarrier. The net data-rate, SE and maximum reach can be

finely tuned.. . . 171

4.2 Summary of 8QAM transmission results with varying number of phase

conjugated subcarrier. Symbol-rate = 39 GBd, No. of digital subcarriers = 6. . . 172

5.1 Various SCM hybrid QAM formats and achievable data-rates and SE for

(25)

(26)

Introduction

This dissertation describes the contributions of the author in the field of ultra-long haul and long haul optical communication. In this chapter, first a brief history of telecommunication is provided followed by the motivation of contributed work and in the last part the structure of this thesis is summarized.

1.1 Brief History

The modern day communication can be traced back to the advent of telegraph in mid nineteenth century which used electrical signals for transmission of messages over long distance and saw a large scale deployment over US and Europe [1]. Electrical transmis-sion dominated the mode of communication ever since. Later on telephone was invented which communicated voice over long distances by converting it to electrical signal. In addition to limited bandwidth, the electrical cables used as transmission medium had losses of upto 10 dB/km requiring closely spaced repeaters making the system expensive to operate [2]. Wireless communication, employing electromagnetic waves which did not require a transmission medium, also evolved however majorly for the broadcast of radio and television signals. Long distance communication utilizing electromagnetic waves require line-of-sight condition and hence the achievable distance is limited by curvature of earth (roughly 30 km) if not by other hinderences (e.g. buildings, mountains, etc.). It was envisioned at the end of 1950s that satellites could be used for transoceanic wireless communication [3]. Satellite communication hence gained attention; a geostationary satellite can provide coverage (receive and transmit signals) within half of the phase of earth while potentially maintaining a stationary footprint on the globe. The most recent commercial satellites of today provide a thoughput of 140 Gb/s [4].

The bit-rate distance product (B·L) of transmission systems in 1960s was limited merely to 100 Mb/s·km and it was realized that B·L could be increased several folds by using optical waves [2]. In 1960, laser was invented which would serve as a coherent optical source (counterpart of oscillator in RF domain) for transmission [5]. In 1966, K. C. Kao et al. showed that a dielectric fiber with 1% higher refractive index than its surrounding can serve as an optical waveguide however the losses of the investigated materials were very high [6]. In 1970 scientists at Corning Glass Works developed the first low-loss single mode fiber having attenuation of <20 dB/km by doping titanium

(27)

in silica glass core. It had the potential to transmit 65000 times as much information as copper wire [7]. The continued research for improved fiber manufacturing resulted in reduction of fiber attenuation and in 1978 NTT made a SMF with attenuation factor of only 0.2 dB/km at wavelength of 1550 nm [8]. The first non-experimental fiber link was installed in 1975 in Britain. After two years, first live telephone call was communicated through optical fiber in US. The low loss of optical fibers resulted in increased repeater spacing (60 - 70 km) and use of fiber optics for communicating telephone signals grew extensively in early 1980s. The first trans-atlantic optical cable (TAT-8) was deployed in 1988 which carried upto 40000 telephone channels between US, Britain and France [9]. It should be noted that in those systems, electrical signal regeneration was required by periodically spaced repeaters since the option of signal amplification in optical domain was not available. The regeneration capability of electrical repeaters was limited to single channel only.

In order to improve receiver sensitivity and increase repeater spacing, coherent op-tical receivers employing homodyne and heterodyne techniques gained interest in late 1980s [10,11]. In 1986 however, the first optical amplifier for a wide wavelength window around 1550 nm was demonstrated which avoided the limitation of receiver sensitivity and optical communication research took a new direction. In the following years, WDM systems were introduced, carrying multiple data channels over a single fiber, increas-ing the fiber throughput several folds. In early 1990s, the internet was launched which showed a rapid growth resulting in exponential increase in capacity demands [12]. In the following years, systems with data-rate 10 Gb/s per channel saw a wide scale deploy-ment. Those systems are based on intensity modulation coupled with direct-detection technology. Transponder systems providing data-rate of 40 Gb/s per channel were sub-sequently developed also based on direct-detection however those systems utilized phase shift keying. It was however realized that moving to data-rates >40 Gb/s, coherent detection would be beneficial [13]. As a result, coherent optical communication was re-introduced however this time as a mean to enhance capacity [14–16]. In contrast to the direct-detection systems, coherent transmission systems utilize digital signal processing at the receiver to mitigate linear transmission impairments. In 2007, first coherent op-tical 100 Gb/s single channel transmission over a live link was demonstrated followed by first commercial deployment in 2009 [17–19]. Coherent optical transmission has ever since gained increasing interest commercially as well as in research because of its prom-ising potential for high capacity transmission which would lead to even reduced cost per transmitted bit.

1.2 Motivation

In the early 1990s when internet was introduced voice traffic dominated the global net-works however this trend was changed already in the following few years [20]. The growth rate of internet is comparatively much higher which is also associated to emergence of a wide range of services totally dependent on internet. In addition the wide scale usage of mobile devices are resulting in ever increasing data-traffic which has a direct impact on the capacity demands faced by the backbone optical networks. Furthermore, global adaptation of internet based services, such as social networking, video conferencing and multimedia services to name a few [21–25], are driving the data-rate demands to new heights. As a result, a three fold increase in global internet traffic over a time period of five years (2015 - 2020) is expected [26]. The compound annual growth rate (CAGR) of

(28)

internet traffic is expected to saturate at ∼22% [26] which will drive the existing optical fiber infrastructure to its limits in the near future; leading to a capacity crunch.

In order to accommodate the growing capacity demands, new optical fibers could be deployed however this is a very costly solution and will not be able to provide the desired reduction in cost per transmitted bit. It is hence desired to use the already deployed fiber optic with minimum changes in network infrastructure. This leads to the conclusion that newer transponder systems must be developed and deployed utilizing the optical fiber bandwidth in an efficient way thereby increasing the SE and overall network throughput. Furthermore, the available optical fiber spectrum can be efficiently utilized by avoiding spectral gaps in WDM channels by utilizing a flexible grid architecture instead of ITU-T standard 50 GHz grid [27,28]. Utilization of flexible-grid architecture can lead to enhanced optical filtering penalties and in this thesis a novel optical spectral shaping method is proposed for mitigation of these effects in optical domain.

In context of next generation transponders, data-rates of 400 Gb/s and 1.0 Tb/s have gained particular interest; which are expected to be the next Ethernet standards. To achieve these high data-rates, coherent optical transmission of higher order mod-ulation formats must be utilized. Furthermore, because of inability to achieve such high data-rates with a single channel due to limitations of electrical components, su-perchannel technology has been introduced. In this thesis, experimental evaluation of dual-subcarrier 400 Gb/s as well as quad-subcarrier 1.0 Tb/s superchannels transmis-sion over long haul and ultra-long haul distances has been presented. Digital signal pro-cessing techniques both at transmitter and receiver were used to limit signal bandwidth and mitigate nonlinear phase distortions in order to achieve high capacity transmission. These results are achieved by first optimizing the per channel transmission symbol-rate keeping in view advanced forward error correction options likely to be used with higher order modulation formats. In order to reduce cost per transmitted bit for 400 Gb/s and 1.0 Tb/s signals, it is desired to increase the per subcarrier data-rate thereby re-ducing the total number of subcarriers in a superchannel. By increasing the order of modulation, single-subcarrier 400 Gb/s and tri-subcarrier 1.0 Tb/s solutions were also experimentally demonstrated. Various terabit superchannel designs were also evaluated through a field trial with a tier1 operator Orange in France.

The achievable capacity of an optical fiber is limited by the nonlinear shannon limit. In order to surpass this limit, several digital signal processing based methods have evolved which mitigate self phase modulation effects. These techniques usually suf-fer from high computational complexity and hence limit their practical application. A low complexity nonlinear mitigation scheme is hence desirable to enhance capacity. In this context a digital subcarrier multiplexing scheme was investigated for QAM formats which relies on symbol-rate optimization of subcarrier for nonlinear tolerance. Optimum selection of subcarrier symbol-rate provides tolerance to self phase modulation effects resulting in improved transmission performance in WDM scenario. As the symbol-rate of a signal is reduced, tolerance to frequency and phase offset in a coherent optical receiver is reduced; resulting in performance penalties ultimately limiting achievable reach. In order to estimate and correct the frequency and phase offset for single and multiple digital subcarriers, a distributed digital pilot tones based method is introduced which is well suited for single as well as multiple digital subcarrier systems. The pro-posed algorithm can be applied regardless of modulation format. The algorithm is a feed-forward algorithm which makes it suitable for practical implementation without additional penalties.

(29)

Increase in capacity by utilizing higher order modulation formats significantly limits the maximum achievable reach due to their reduced tolerance to linear and nonlinear noise encountered along transmission link. Furthermore, the optical networks increas-ingly face dynamic capacity demands depending on changing customer requirements. Hence, a flexible transceiver adaptive to network demands is highly desirable. In this thesis, a novel digital subcarrier multiplexed hybrid QAM is proposed to provide adapt-ive capacity and reach. Furthermore, phase conjugated twin waves have been used for digital subcarrier multiplexing signals enabling enhanced resolution in capacity versus reach selection.

1.3 Thesis structure

This dissertation is divided in total six chapters. The basic concepts of coherent optical transmission systems are explained in chapter2which are essential to understand results presented in the subsequent chapters.

Chapter3explains the results of single carrier high capacity transmission experiments targeting data-rates of 100 Gb/s, 200 Gb/s, 400 Gb/s and 1.0 Tb/s. Experimental eval-uation of a simple digital pre-emphasis algorithm to mitigate bandwidth limitations of electrical components is explained. The algorithm is then used to design high capacity transceiver employing high symbol-rate operation. Results of transmission over several kinds of optical fibers are discussed assessing tolerance to linear and nonlinear noise for variable span lengths and amplification options. The chapter also includes results of a high capacity field trial which assessed performance of multiple terabit superchannels employing coherent transmission of high symbol-rate PM-16QAM, 32QAM and 64QAM formats. The field trial link consisted of 762 km long SSMF and the losses were com-pensated by hybrid EDFA Raman amplification for most of the spans.

Chapter4 describes the investigation on digital subcarrier multiplexing systems and its applications as a low complexity solution for nonlinear mitigation. In order to avoid limitations due to laser linewidth induced phase noise, a distributed pilot tones based carrier recovery scheme is proposed and its performance is also evaluated. The transmis-sion performance of various coherent modulation formats is assessed in a WDM scenario thereby resulting in realistic improvement estimates employing this technique. Further-more, a novel hybrid QAM scheme employing subcarrier multiplexing is presented which provides a viable solution for capacity/ reach adaptive coherent transceiver. In addi-tion, a transceiver architecture employing phase conjugation on selected subcarriers is discussed and enhanced resolution in achievable capacity/ reach is presented.

Chapter5explains the concept of elastic optical networks from physical layer’s per-spective. Major limitations in those networks originating from optical nodes are experi-mentally analyzed and methods for the mitigation of those penalties are also discussed. In particular optical spectral shaping integrated in each ROADM is shown to provide an improved transmission performance over a flexible grid architecture. The use of a proposed hybrid QAM scheme is shown to possess better tolerance towards optical fil-tering effects. Finally conclusions and outlook from the results of this dissertation are drawn in chapter 6.

(30)

1.4 Contributions

The author is solely responsible for carrying out all the described experiments in this thesis. The first step of the experimental work was to optimize the performance of optical transmission loop by selecting suitable components as well as characterization of different kinds of fibers. The author was supported by Erik de Man specially during that phase as well as periodically afterwards. In order to support robust high symbol-rate transmission a digital pre-emphasis scheme is essential. In this context, the use of experimentally measured response of the electrical components and its integration to the digital pre-emphasis algorithm is essential, which was successfully accomplished by the author. As a result, reduced subcarrier count 16QAM, 32QAM and 64QAM based 1.0 Tb/s super-channels could be transmitted over long-haul distances. The system design proposed by the author were also tested over a field deployed fiber in collaboration with Orange telecom. The field trial link was set up by project partners from Orange, Ekinops and Keopsys however, the author was responsible for setting up and optimizing the transmitter and receiver as well as carrying out all the measurements. Main contributors to the baseline coherent receiver DSP are Dr. Bernhard Spinnler and Dr. Maxim Kuschnerov. The FEC code used in the field trial was designed by Dr. Stefano Calabr`o. Realizing the need for low complexity nonlinear mitigation methods the concept of digital subcarrier multiplexing recently gained momentum. Starting from the baseline DSP, the author independently designed and programmed the transmitter and receiver DSP algorithms related to digital subcarrier multiplexing. In this context a distributed digital pilot tones based carrier recovery scheme was ingeniously designed, programmed and validated both by simulations as well as experiments by the author. The proposed digital pilot tones based carrier recovery method is a feed-forward one and is also suit-able for applicaiton specific integrated circuit (ASIC) implementation. For the next generation of transceivers, a main challenge is to support capacity and reach selection with a finer granularity. In this context, a novel digital subcarrier multiplexed hybrid QAM schemes is designed and experimentally tested by the author. The proposed design also employs phase conjugated twin waves concept to enhance nonlinear tolerance and further enhance flexibility. The performance of flexible capacity and reach with the proposed scheme is also experimentally validated by the author.

In the recent past, the physical layer building blocks of flexible grid networks re-cently became available and this architecture could be deployed in the near future. One major limitation that degrades signal performance in those networks is cascaded optical filtering. In order to mitigate those limitations, the author proposed integration of op-tical spectral shaping in each ROADM. The proposed scheme was validated through simulation analysis as well as experiments by the author.

The author worked under direct supervision of Dr. Danish Rafique (first 3 years), Dr. Bernhard Spinnler and Dr. Antonio Napoli from Coriant as well as Prof. A. M. J. (Ton) Koonen, Dr. Huug de Waardt and Dr. Chigo M. Okonkwo from the Eindhoven Univeristy of Technology.

(31)

(32)

Coherent Optical Transmission

Systems

Internet has deeply impacted the way of human life by finding implications in all its aspects. As a result of public introduction of the world wide web in 1990s data networks have seen increasing traffic growth ever since. The growth rate of internet traffic is much greater than the voice traffic previously dominating the global networks [20]. Within the first decade of its public introduction, internet based traffic became the driving force for grooming of data networks. In this chapter, a hierarchical description of global optical networks is provided (Sec. 2.1). All the deployed optical networks of today employ SMF as a transmission medium. The properties of SMF will be described in Sec. 2.3which includes both linear and nonlinear effects and their implications for telecommunication. The most important components of optical transmitter and receiver as well the digital signal processing relevant for them are described in Sec. 2.5.

2.1 Fiber Optic Networks

The structure of deployed optical networks can be roughly divided in to three groups i.e. (i) long-haul, ultra long-haul and transoceanic, (ii) Metro/ Regional and (iii) Access Networks [20,29]. The schematic representation of today’s optical networks is shown in Fig. 2.1. The division of networks level is based mainly on the targeted distance which also attributes to the amount of data traffic transported through them. The maximum length of links in different level of network hierarchy are summarized in table2.1. As for

Table 2.1: Typical distances in an optical network hierarchy. Network level Distance

Access <100 km Metro 100 km - 300 km Regional 300 km - 1000 km Long-haul 1000 km - 3000 km Ultra long-haul 3000 km - 6000 km Transoceanic >6000 km

(33)

Long-haul, Ultra Long-haul and Transoceanic links

Metro Networks Access Networks

Figure 2.1: Hierarchy of Optical networks. Transoceanic links(>6000 km), ultra long-haul net-works (>3000 km), Long-haul netnet-works (>1000 km), Regional netnet-works (>300 km), Metro netnet-works (>100 km), Access networks (<100 km).

any network, routing elements or routers are the main component. They are responsible for routing data signals or packets in a specific direction in order it to reach the desired destinaiton. Transmission networks are evolving towards all optical solutions in order to improve flexibility and reduce power consumption caused by electrical to optical conversion as well as avoid the electronic bottlenecks [30].

The access networks are comoposed of typically short links and are at the lowest level of hierarchy. These networks represent the subscribers connection to their immediate service providers. They are also called the last-mile networks and the link lengths can be up to 100 km. These networks are currently dominated by copper wires and radio links however this trend is quickly changing as the service providers realize that data traffic will continue to increase and fiber networks can adapt to increasing demands economically; owing to their wide bandwidth and lower loss. The analogy of fiber to the ‘x’ (FTTx) has hence recently evolved, where ’x’ can be home (H), premise (P), building (B), node (N) or curb/ cabinet (C) [31]. Typical data-rates in access networks are several hundred Mb/s and are evolving towards 1 Gb/s [32]. Recent evolution in cloud computing and data centers has introduced a new domain in networks i.e. data center interconnects (DCI). Connection among geographically diverse data centers is required for quick service provisioning, resilience and disaster recovery. Depending on the nature of services provided by data centers, the generated data traffic can be much higher requiring novel networks design enhancing software based control and adaptability to vendor independent network equipment. DCI networks are typically point-to-point links having networking architecture different than that of access or metro networks [33]. The aggregated data from access networks is routed to the next level of hierarchy termed as metro/ regional networks. These networks are responsible for connecting different cities or possibly countries together. Fiber optics has already dominated this part of networks for data transmission because of large amount of data traffic need to be routed among different regions. Due to increased usage of data-rate per user in an access network fueled by internet and cloud based applications, data-rate demands for metro/ regional networks are growing much faster. Traditional circuit switched network architecture designed for voice transport is no longer able to cope with growing demands

(34)

resulting in evolution of metro networks towards all optical switching [34–36]. Further-more, metro networks are evolving from ring or star topology towards meshed architec-ture in order to have improved capacity and reliability as well as increased redundancy in the network. The channel data-rates of metro/ regional networks are 10 Gb/s which are rapidly evolving to 40 Gb/s; both based on direct-detection transceiver technology. However, 40 Gb/s deployment is expected to be surpassed by 100 Gb/s signals based on coherent detection technology [33,37].

The long-haul, ultra long-haul and transoceanic links are at the top of optical net-works hierarchy as depicted in Fig. 2.1. They are also called backbone or national net-works by the operators and are responsible for transporting data aggregated by several metro/ regional networks. Hence, these links have stringent requirements both on capa-city and reach. Coherent optical transceivers with per channel data-rates of 100 Gb/s are used by these networks which provide a total capacity of 9.6 Tb/s employing 50 GHz spaced 96 WDM grid following international telecommunication union standardization (ITU-T) recommendations [27]. All the deployed optical networks of today employ SMF as a transmission medium and its properties will be discussed in a following section. In context of this thesis ultra long-haul, long-haul, metro/ regional networks, their lim-itations and counterfeiting those limlim-itations from physical layer’s perspective will be focused.

2.2 Switching in Optical Networks

An optical network is made up of two or more nodes connected in ring, star or mesh topology. In addition to switching and routing of multiplexed input signals, a node of an optical network might perform several other functions e.g. optical signal amplification, wavelength conversion, mitigation of linear channel distortions, signal regeneration etc. Switching and routing elements in a node are responsible to direct input signals to the appropriate output(s) in order them to reach the desired destination. Depending on the available network resources, a node might convert the wavelength of incoming signal to avoid blocking in the following link in case that particular wavelength is already used. Conversion of wavelength of a particular channel requires optical optical-to-electrical (OE) conversion and re-transmission at a different wavelength by electrical-to-optical (EO) conversion. The elements responsible for switching optical signals are called ADMs, optical add-drop multiplexers (OADMs) or re-configurable optical add-drop multiplexers (ROADMs); depending on their functional capability. As the name suggests, ADMs are responsible for adding some data channels to the fiber link and dropping some other channels from the fiber link.

Any kind of ADM is composed of three basic components i.e. (i) wavelength de-multiplexer, (ii) switches and (iii) wavelength multiplexer. Conventionally the input optical signal was converted to an electrical signal, switching was performed in the elec-trical domain followed by a EO conversion and transmission. This trend is however changed by OADMs which allow switching in optical domain without requiring unne-cessary optical-electrical-optical (OEO) conversions. Switching in optical domain allows for modulation format as well as data-rate independent operation which is desirable in order to relax the requirements in designing optical networks. Reconfigurability in an OADM is a very useful feature. ROADMs can be remotely reconfigured hence minimize the requirements of physical replacement of components for network reconfiguration. Consequently, agile optical networks are based on ROADMs which can be

(35)

dynamic-ally configured as might be required for hitless service provisioning. Fig. 2.2 shows a simple ROADM structure and its functionality is as follows: a WDM input signal is first de-multiplexed in to N channels of which M can be dropped using correct switch-ing configurations. Up to M channels can be added to the output WDM signal which

D E M U X M U X ... ... ... ... ... ... ... ... Drop NxM MxN Add WDM IN WDM OUT 1xN Nx1 1x2 2x1 λ1 λ3 λ2 λN λ1 λ3 λ2 λN

Figure 2.2: A simple ROADM structure.

are multiplexed and output to a single port. The number of directions form which input-output ports can be accessed defines the degree of a ROADM. A multiple degree ROADM, as might be required in a meshed network, can be built by the interconnec-tions of same basic block as shown in Fig. 2.2. Today most of metro, regional and access networks employ ring network topology requiring a degree 2 ROADM wherever add-drop of some WDM channels is required. An interconnection of two ring networks requires a 4 degree ROADM. Long-haul and ultra long-haul networks typically employ mesh network topology requiring multiple degree ROADMs. As the data-rate demands are constantly increasing, in order to increase network throughput, metro and regional networks are also evolving towards mesh networks topology.

ROADMs are classified as colorless, directionless and/or contentionless, depending on its supported functions. A colorless ROADM provides colorless add-drop ports mean-ing each add-drop port supports operation independent of wavelength. If each add-drop port of a ROADM can switch any wavelength to any direction, it is termed as direction-less. A contentionless ROADM supports multiple channels of same wavelength to be added-dropped from each single direction. Each functionality increases cost and power consumption of the device resulting in colorless-directionless (C-D) ROADM to be a more favored option by the service providers rather than colorless-directionless-contentionless (C-D-C) one.

Fixed grid and Flexible grid networks

The spectral width or channel bandwidth allocated for each optical signal in WDM is specified by ITU-T grid [27]. Current generation transceivers with data-rates of 10 Gb/s, 40 Gb/s or 100 Gb/s are allocated a fixed 50 GHz channel bandwidth for transmission over an optical fiber. This standard channel bandwidth allocation is also termed as fixed grid allocation. The fixed grid architecture has served well for the past many years however the resulting SE in fixed grid architecture is not optimized because of the unusable spectral gaps in the spectrum. Current transceivers with 10 Gb/s, 40 Gb/s and 100 Gb/s data-rates achieve SE of 0.2 bit/s/Hz, 0.8 bit/s/Hz and 2.0 bit/s/Hz over the fixed grid architecture. The distribution of channels over a fixed grid architecture is shown in Fig. 2.3 which shows unusable spectral gaps between different signals. Currently commercial optical transceivers employ a single optical carrier for data transmission. Next generation transceivers aim at supporting data-rates of 400 Gb/s

(36)

50 GHz 50 GHz 50 GHz 10 Gb/s 40 Gb/s 100 Gb/s

Figure 2.3: Distribution of different optical signals over fixed grid. 400 Gb/s superchannel 200 GHz (a) 400 Gb/s superchannel 150 GHz (b)

Figure 2.4: A 400 Gb/s superchannel composed of 4×100 Gb/s subcarriers over (a) Fixed grid oc-cupying 200 GHz and (b) Flexible grid ococ-cupying 150 GHz.

and 1.0 Tb/s following a factor 10 increase from existing systems. Since these desired data-rates are very hard to transmit with a single optical carrier, multiple optical carriers must be employed. A set of multiple optical subcarriers carrying aggregate data-rates of 400 Gb/s or more is termed as a superchannel. Unlike the commercially available single optical carrier systems, a superchannel require channel bandwidth larger than 50 GHz. In a fixed grid architecture, multiple 50 GHz slots would be assigned to a superchannel transmitted through the fiber which still possess spectral gaps as shown in Fig. 2.4a for a 400 Gb/s superchannel made up of 4×100Gb/s subcarriers. A superchannel is supposed to traverse the optical network between transmitter and receiver as a single entity. Spectral gaps within a superchannel can be avoided by placing subcarriers over non-standard central frequencies as highlighted in Fig. 2.4b resulting in a reduced spectral footprint. Consequently, an improved SE can be achieved employing flexible grid architecture. Flexible grid architecture is highly desirable for future networks in order to maximize bandwidth efficiency as well as optical fiber network throughput.

2.2.1 Arrayed Waveguide Gratings

Wavelength multiplexing and de-multiplexing in optical domain can be performed using an AWG [38,39] whose structure is shown in Fig. 2.5. A WDM signal fed in to the AWG goes through a free propagation region (FPR) where it diverges and is coupled to an array of optical waveguides at the input aperture. The waveguides in the array have different optical path lengths resulting in a wavelength dependent phase difference between the optical signals propagating through different waveguides at the output aper-ture. The difference in optical path length between adjacent waveguides is fixed and is designed according to the desired wavelength of operation. Optical signal from the out-put aperture converges in the following FPR. As a result of phase difference induced by the arrayed waveguides, different wavelength signals are converged at different positions on the image plane. By positioning multiple output waveguides along the image plane, optical signal corresponding to different wavelengths can be tapped out [40]. As a result, wavelength de-multiplexing operation is achieved. The channel bandwidth of an AWG