University of Southern California Dissertations and Theses

Reconfigurable optical signal processing for efficient spectrum utilization in high-speed optical communication systems

(USC Thesis Other)

Reconfigurable optical signal processing for efficient spectrum utilization in high-speed optical communication systems

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content RECONFIGURABLE OPTICAL SIGNAL PROCESSING FOR EFFICIENT SPECTRUM UTILIZATION IN HIGH-SPEED OPTICAL COMMUNICATION SYSTEMS by Yinwen Cao A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING) August 2019 Copyright 2019 Yinwen Cao ii Dedication To my parents, my wife and my cute son To my advisor, OCLab colleagues, and other friends for their encouragement and support. iii Acknowledgments Still remember vividly that in August of 2013, how excited, nervous, and curious I was when I came to USC to start my PhD program in OCLab. Now, it has been more than five years, and I deeply enjoyed the most valuable period in my life till now. I have learned so much from my great advisor Prof. Alan E. Willner and my colleagues in OCLab. I will not make it without their encouragement and support. First and foremost, I would like to thank Prof. Alan E. Willner: for accepting me to his group; for being patient of my growth and progress; for directing me with novel topics; for providing me with constructive suggestions; and for encouraging me when I was down. He is a great advisor not only for my research but also for my life. I always gain a lot from his wisdom and vision during each short/long discussions. I would like to thank my wife, Ying Chen, who did not hesitate giving up her great job in SAP and to come with me to the U.S. With her F-2 visa in the beginning, she had to stay at home. However, she did not blame and always said: “Nothing is more important than being together.” Now, she has graduated from USC with a Master degree and I am so happy for her! Thank you and I love you! I would like to thank my 2-year old son, Ruichen (Ricky) Cao, who is the most precious gift from God. Ricky, you made my day and you made my life! Thank you and I love you! I would like to thank my great parents, who are always giving. Thank you and I love you! iv I would like to thank Prof. Stephan Haas and Prof. Wei Wu for being my defense committee member. I want to extend my appreciation to Prof. Alexander Sawchuk and Prof. Todd Brun, for their excellent questions and advice during my PhD qualification examination. I am grateful to Prof. Moshe Tur for the fruitful discussions during his visits to OCLab. I also would like to thank our collaborator, Dr. Youichi Akasaka, for being closely working together and providing insightful advice from the perspective of the industry. Last but not least, I would like to thank my OCLab colleagues: Dr. Hao Huang, Dr. Mohammad Reza Chitgarha, Dr. Yan Yan, Dr. Nisar Ahmed, Dr. Yongxiong Ren, Dr. Morteza Ziyadi, Dr. Guodong Xie, Dr. Changjing Bao, Dr. Amirhossein Mohajerin-Ariaei, Dr. Long Li, Zhe Zhao, Peicheng Liao, Cong Liu, Ahmad Fallahpour, Fatemeh Alishahi, Runzhou Zhang, Kai Pang, Haoqian Song, Kaiheng Zou, Hao Song, Huibin Zhou, and Karapet Manukyan, for all the valuable discussions and generious support. It is so lucky for me to meet you in my life and work with you for years! Thank you and I wish all the best for your future! v Table of Contents Dedication ................................................................................................................... ii Table of Content ......................................................................................................... v List of Figures ........................................................................................................... vii Abstract .................................................................................................................... xiv Chapter 1 Introduction ............................................................................................ 16 1.1 Nonlinear Wave Mixing Processes for OSP................................................. 16 1.1.1 Three-wave Mixing ............................................................................ 17 1.1.2 Four-Wave Mixing ............................................................................. 18 1.2 Basic Enabling Technologies ....................................................................... 19 1.2.1 Advanced Modulation Format ........................................................... 19 1.2.2 Coherent Detection............................................................................. 21 1.2.3 Optical Frequency Combs .................................................................. 23 1.3 Basic Enabling Operations ........................................................................... 23 1.3.1 Phase-Preserving Wavelength Conversion ........................................ 23 1.3.2 Wavelength Multicasting and Multiplexing ...................................... 24 1.4 Dissertation Outline ...................................................................................... 25 Chapter 2 Reconfigurable Optical Channel Slicing and Stitching for Fragmented Bandwidth Allocation .................................................................. 27 2.1 Introduction .................................................................................................. 27 2.2 Concept ......................................................................................................... 28 2.3 Experimental Setup for a Single Channel System ........................................ 29 2.4 Experimental Results for a Single Channel System ..................................... 31 2.5 Experimental Setup for a WDM System ...................................................... 36 2.6 Experimental Results for a WDM System ................................................... 37 2.7 Discussion and Conclusion ........................................................................... 39 Chapter 3 Reconfigurable Optical Inter-Channel Interference Mitigation for Spectrally Overlapped Signals .......................................................................... 40 3.1 Introduction .................................................................................................. 40 3.2 Concept ......................................................................................................... 41 3.3 Experimental Setup....................................................................................... 42 3.4 Results .......................................................................................................... 45 3.5 Discussion and Conclusion ........................................................................... 48 vi Chapter 4 Tunable Optical Single-Sideband Generation of OOK and PAM4 Data Channels .................................................................................................... 50 4.1 Introduction .................................................................................................. 50 4.2 Concept ......................................................................................................... 51 4.3 Experimental Setup....................................................................................... 52 4.4 Experimental Results .................................................................................... 54 Chapter 5 Performance Enhancement of an Optical High-Order QAM Channel by Adding Correlated Data to Robust Neighboring Channels ...................... 56 5.1 Introduction .................................................................................................. 56 5.2 Concept ......................................................................................................... 57 5.3 Simulation results ......................................................................................... 59 5.5 Experimental Setup....................................................................................... 62 5.6 Experimental Results .................................................................................... 64 5.7 Discussion and Conclusion ........................................................................... 65 Chapter 6 All Optical Signal Level Swapping and Multi-Level Amplitude Noise Mitigation Based on Optical Parametric Amplification ................................. 67 6.1 Introduction .................................................................................................. 67 6.2 Concept ......................................................................................................... 68 6.3 Experimental Setup of OPA Gain Regions Measurement............................ 70 6.4 Experimental Results for OPA Gain Profile ................................................. 70 6.5 Experimental Setup for 2-level Amplitude Noise Mitigation ...................... 73 6.6 Results for 2-Level Amplitude Noise Mitigation ......................................... 74 6.7 Discussion and Conclusion ........................................................................... 75 Chapter 7 Self-Homodyne Detection with a Low-Power Pilot Tone Using Brillouin Amplification and Phase-Preserving Wavelength Conversion ...... 77 7.1 Introduction .................................................................................................. 77 7.2 Concept ......................................................................................................... 78 7.3 Experimental Setup....................................................................................... 79 7.4 Experimental Results .................................................................................... 82 7.5 Discussion and Conclusion ........................................................................... 84 Chapter 8 Raman-Assisted Phase Sensitive Amplifier using Fiber Bragg Grating Based Tunable Phase Shifter .............................................................. 86 8.1 Introduction .................................................................................................. 86 8.2 Concept ......................................................................................................... 87 8.3 Experimental Setup....................................................................................... 88 8.4 Experimental Results .................................................................................... 90 8.5 Discussion and Conclusion ........................................................................... 94 References ................................................................................................................. 95 vii List of Figures Figure 1.1 Two types of cascaded three-wave mixing: (a) cascaded sum frequency generation (SFG) + difference frequency generations (DFG), and (b) cascaded second harmonic generation (SHG) + difference frequency generation (DFG). The frequencies (wavelengths) of the pump(s) and the signal are chosen near to the quasi-phase matching (QPM) frequency for an efficient wave mixing. ......................................................................................... 17 Figure 1.2 Two types of four-wave mixing (FWM) process: (a) degenerate FWM with a single pump to produce an idler with phase-conjugating of the input signal, and (b) non-degenerate FWM with two pumps to produce three idlers, one of which is the signal copy and the other two are conjugate signal copies. .................................................................................................................. 18 Figure 1.3 Different modulation formats with corresponding waveforms and constellations. ...................................................................................................... 21 Figure 1.4 The structure of a coherent detection system with DSP to compensate for various transmission distortions, such as chromatic dispersion, polarization-mode dispersion, and carrier phase noise. ....................................... 22 Figure 1.5 Concept of optical frequency combs. FSR: free spectrum range. ................ 23 Figure 1.6 Concept of phase-preserving wavelength conversion using a pair of comb lines............................................................................................................ 24 Figure 1.7 Concepts of (a) wavelength multicasting, and (b) wavelength multiplexing. ........................................................................................................ 25 Figure 2.1 Conceptual diagram of fragmented bandwidth allocation enabled by channel slicing and stitching. .............................................................................. 29 Figure 2.2 Experimental setup of channel slicing and stitching for a single optical channel. The optical frequency comb is employed to achieve phase- preserving channel copy generation and channel slices recombination (stitching)............................................................................................................. 30 Figure 2.3 Measured spectra: (a) before PPLN-1; (b) after PPLN-1; (c) after SLM- 2 filter; (d) before PPLN-2; (e) after PPLN-2 with constructive channel stitching; (f) after PPLN-2 with destructive channel stitching. ........................... 31 Figure 2.4 Constellation comparison of a 20 Gbaud QPSK channel under different scenarios: (a) B2B without channel slicing and stitching; (b) detection of only the left channel slice; (c) detection of only the right channel slice; (d) viii B2B with channel slicing and stitching; (e) after 10-km transmission with channel slicing and stitching. .............................................................................. 32 Figure 2.5 Constellation comparison of a 20 Gbaud QPSK by tuning the relative phase offset (Δφ) between two channel slices: (a) with channel equalization; (b) without channel equalization. (c) Measured EVMs with different Δφ. ......... 33 Figure 2.6 Constellation comparison of a 20 Gbaud QPSK by tuning the relative amplitude (Δα) between two channel slices: (a) with channel equalization; (b) without channel equalization. (c) Measured EVMs with different Δα. ......... 34 Figure 2.7 BER comparison for a 20 Gbaud QPSK system: (a) with and without digital equalization; (b) B2B and 10-km transmission. ....................................... 35 Figure 2.8 Channel slicing and stitching with three slices for a 28 Gbaud QPSK system: (a) optical spectrum before PPLN-1; (b) optical spectrum after SLM-2 filter; (c) optical spectrum after PPLN-2; (d) channel reconstruction by stitching three channel slices. ......................................................................... 35 Figure 2.9 BER comparison with different numbers of channel slices. ........................ 36 Figure 2.10 (a) EVM comparison between B2B and channel slicing and stitching for a 20 Gbaud 16QAM signal; (b) BER comparison. ........................................ 36 Figure 2.11 (a) Experimental setup for fragmented bandwidth allocation in 20 Gbaud QPSK WDM channels; (b) measured spectrum before fragmented bandwidth allocation; (c) measured spectrum after fragmented bandwidth allocation. ............................................................................................................ 37 Figure 2.12 Constellation comparison between direct channel insertion and fragmented bandwidth allocation enabled by channel slicing and stitching. ...... 38 Figure 2.13 BER comparison between direct channel insertion and fragmented bandwidth allocation enabled by channel slicing and stitching. ......................... 38 Figure 3.1 Conceptual diagram of the proposed optical ICI mitigation scheme for an overlapped channel system. There are three processing stages in the background with different colors: (1) channel copy generation; (2) complex coefficients adjustment; (3) channel addition. PPLN: periodically poled lithium niobate. .................................................................................................... 42 Figure 3.2 (a) Experimental setup for the ICI mitigation system and the corresponding optical spectra: (1) Channel-1 before overlapping, (2) Channel-2 before overlapping, (3) overlapped channels, (4) copy generation, (5) filtering and coefficient adjustment, (6) channel addiction. (b) Back-to- back (B2B) baseline configuration. (c) System implementation penalty ix compared with B2B baseline configuration under single channel transmission. ........................................................................................................ 43 Figure 3.3 Signal constellations comparison with and without ICI mitigation for (a) 20G-baud and (b) 25G-baud overlapped channel systems of different channel spacing (CS). .......................................................................................... 46 Figure 3.4 Q factor comparison with and without ICI mitigation under different channel spacing (CS) and baud rates. (a) 25G-baud with 20GHz CS and 20G-baud with 15GHz CS; (b) 25G-baud with 22.5GHz CS and 20G-baud with 17.5GHz CS; (c) 25G-baud with 25GHz CS and 20G-baud with 20GHz CS; (d) 25G-baud with 30GHz CS and 20G-baud with 25GHz CS. ...... 47 Figure 3.5 System Q factor varies with the phase change of the coefficient and the corresponding signal constellation. Different phases can change the system Q factor by more than 3dB. ................................................................................. 47 Figure 3.6 Simulated EVM comparisons by employing an adaptive intra-channel equalizer with different tap numbers for three scenarios: (1) with proposed ICI mitigation; (2) without ICI mitigation; (3) cutting the overlapped spectrum. ............................................................................................................. 48 Figure 4.1 Concept diagram of the proposed tunable optical single-sideband generation for a data channel using optical frequency combs and nonlinear wave-mixing. The tunability of this approach could be achieved by choosing a different number and spacing of comb lines. .................................................... 52 Figure 4.2 (a) Experimental setup of tunable optical SSB generation for different data rates. The LCoS filter is used to select comb lines and set SSB tap coefficients. The data is received by a photodiode (PD) without chromatic dispersion compensation after an 80-km SSMF transmission. (b) Input and output spectra of the PPLN, in which 3 taps are used for SSB generation of a 10-Gbit/s OOK channel. (c) Input spectra of the PPLN, in which 5 taps are used for SSB generation of a 10-Gbit/s or a 20-Gbit/s OOK channel. Note that since some of the taps in SSB impulse response are zeros, the corresponding comb lines are blocked. For a 20-Gbit/s OOK channel, the spacing of the selected comb lines is reduced by half compared to that of the 10-Gbit/s channel in order to let the same DCF introduce appropriate delays. ... 53 Figure 4.3 (a) Comparison of eye-diagrams and optical spectra between 10-Gbit/s OOK DSB and SSB channels. (b) Optical spectra using a different number of taps (combs). (c) BER comparison between DSB and SSB for a back-to- back scenario. ...................................................................................................... 54 Figure 4.4 (a) Eye-diagram and BER comparison between 10-Gbit/s OOK DSB and SSB channels after an 80km SSMF transmission. Because of the enhanced tolerance to dispersion-induced power fading of SSB, a >3dB x receiver sensitivity improvement is observed compared to the DSB channel. (b) Optical spectra and eye-diagrams for a 20-Gbit/s OOK DSB and SSB channels. (c) Optical spectra and eye-diagrams for a 20-Gbit/s PAM4 DSB and SSB channels. ............................................................................................... 54 Figure 5.1 Schematic diagrams of (a) directly inserting the high-order QAM channel (C) between S1 and S2, (b) enhancing the performance of the channel C using channel correlation, and (c) details for the channel enhancement approach. Channel correlation is achieved by modulating the same high-order QAM data both on a CW laser and on already data-carrying robust neighboring channels (S1 and S2). S1 and S2 can be BPSK or QPSK and this scheme could be extended to a different number of correlated channels. At the receiver, all channels are simultaneously detected and processed. ............................................................................................................ 58 Figure 5.2 Simulation results for the constellation of the target 10-Gbaud channel modulated with 4QAM (QPSK), 16QAM, 64QAM, and 256QAM, with and without channel correlation. Two neighboring 10-Gbaud BPSK channels are used as the correlated channels............................................................................ 60 Figure 5.3 Simulation results for (a) the BER of the target channel modulated with 4QAM (QPSK), 16QAM, 64QAM, and 256QAM, with and without channel correlation; (b) the BER of the neighboring BPSK channels that assist the target channel with different QAM orders. ......................................................... 61 Figure 5.4 Simulation results for (a) the BER of a 64QAM channel with a different number of correlated channels; (b) the BER of a different number of neighboring BPSK channels used as the correlated channels. ............................ 61 Figure 5.5 Simulation results for (a) the BER of a 64QAM channel using BPSK or QPSK correlated channels; (b) the BER of the BPSK or QPSK. ........................ 62 Figure 5.6 (a) A proof-of-concept experimental setup using two BPSKs as the correlated channels to recover a QPSK or 16QAM data channel. The wideband ASE source contributes statistically independent noise to all three channels; (b) offline signal processing in the experiment. .................................. 63 Figure 5.7 Experimental results for (a) a constellation comparison of a target 5- Gbaud QPSK or 16QAM data channel using double BPSK correlated channels; (b) a BER comparison of the target QPSK channel; (c) a BER comparison of the target 16QAM channel; (d) deterioration of the neighboring BPSK channels. ............................................................................... 65 Figure 5.8 Experimental results for the BER of a target 10-Gbaud (a) QPSK or (b) 16QAM channel, using a single BPSK correlated channel. ................................ 65 xi Figure 6.1 (a) Three gain regions (linear, saturation, inversion) of OPA. The gain inversion enables signal level swapping: the output amplitude levels are flipped compared with the input; (b) 2-level amplitude noise mitigation. Stage 1: amplitude squeezing for outer ring symbols based on linear and saturation regions; Stage 2: amplitude squeezing for inner ring symbols based on saturation and inversion regions. .......................................................... 68 Figure 6.2 Experimental setup to characterize OPA gain profile of the HNLF (without 2-ASK signal modulation) and realize signal level swapping (with 2-ASK signal modulation). 8-QAM modulation is achieved by cascading an intensity modulator with a QPSK modulator. Phase modulation is employed to suppress the SBS effect. .................................................................................. 70 Figure 6.3 (a) Measured input-output curves with different pump powers (1W, 1.5W and 2W). Both saturation and inversion OPA regions are observed with 1.5W and 2W pumps; (b) corresponding gain profiles. .............................. 71 Figure 6.4 Experimentally demonstration of signal level swapping for different baud rates: (a) waveform flipping for 10 Gbaud 8-QAM signal; (b) waveform flipping for 20 Gbaud 8-QAM signal. It is noted that both the input and output measurements are normalized. ................................................. 72 Figure 6.5 Constellation EVM comparison before and after all optical signal level swapping (a) 10 Gbaud 2-ASK/8-QAM; (b) 20 Gbaud 2-ASK/8-QAM. EVM degradation is less than 1%. ...................................................................... 73 Figure 6.6 (a) Experimental setup for amplitude noise mitigation of 10/20 Gbaud 8-QAM signals. Stage 1 is for upper level amplitude noise mitigation and Stage 2 is for lower level amplitude noise mitigation; (b) input and output power profiles for HNLF-1 (square symbols) and HNLF-2 (triangle symbols). ............................................................................................................. 74 Figure 6.7 (a) 10 Gbaud and (b) 20 Gbaud 8-QAM constellation after each stage for 2-level amplitude noise mitigation. (c) Measured amplitude error and EVM comparison after each stage. ...................................................................... 75 Figure 7.1 Conceptual diagram of the proposed self-homodyne detection (SHD) system with a low-power pilot tone. The scheme is composed of two stages: (i) Brillouin amplification for only the pilot tone, and (ii) phase preserving wavelength conversion using a pair of optical frequency comb lines. HNLF: highly nonlinear fiber; PPLN: periodically poled lithium niobate; SMF: single-mode fiber. ................................................................................................ 79 Figure 7.2 (a) Experimental setup for the proposed SHD system and the spectra when the slave laser is frequency locked or unlocked to the incoming pilot tone; (b) corresponding spectra measured at each node. ..................................... 81 xii Figure 7.3 (a) Bandwidth of Brillion amplification; (b) Brillouin amplification gain for the pilot tone versus pump power. ................................................................. 82 Figure 7.4 (a) The detected eye diagram varies with the relative phase among two comb lines for BPSK SHD; (b) detected eye diagrams for a 10G-baud system with different modulation formats. .......................................................... 83 Figure 7.5 BER versus received power for (a) 10G-baud BPSK and QPSK and (b) 20G-baud BPSK systems. ................................................................................... 83 Figure 7.6 (a) BER measurements for different levels of PSR and the corresponding Brillouin pump power with -8.4dBm received power; (b) eye diagram of 10-Gbaud Nyquist BPSK signal. ...................................................... 84 Figure 8.1 A schematic diagram of a Raman-assisted PSA using a FBG-based tunable phase shifter. The system includes four stages: (i) idler generation, (ii) phase adjustment, (iii) hybrid Raman/PSA, and (iv) pure PSA. Phase adjustment of the PSA pump is achieved by tuning the central wavelength of the FBG using a thermoelectric heater. Raman amplification is used to boost the power of the PSA pump and compensate for the power imbalance between the signal and idler, by placing the signal away from the effective Raman gain region. .............................................................................................. 88 Figure 8.2 (a) Experimental setup for a Raman-assisted PSA using a FBG-based tunable phase shifter. The link loss from Node-1 to Node-5 (including the 50/50 coupler, idler generation, phase adjustment, hybrid Raman/PSA, and pure PSA) is ~8 dB. The signal net gain is measured by comparing the signal power between Node-1 and Node-5; (b) Raman pump power vs. control voltage; (c) Raman gain profile, where the idler has more gain than the signal. This helps to decrease the power imbalance between the signal and the idler after idler generation; (d) Spectra at different positions. ................ 90 Figure 8.3 (a) The current of the thermoelectric heater vs. temperature and FBG central wavelength; (b) FBG transmittance under different temperature levels, in which a 60 o C temperature increase shifts the central wavelength by 0.66 nm; (c) Temperature change (ΔT) vs. phase-shift of FBG; (d) 1% PSA pump power after the FBG. The flat curve indicates the PSA pump power is maintained across different FBG central wavelengths. ....................................... 91 Figure 8.4 (a) The variation of output signal power vs. signal wavelength with different FBG temperature changes; (b) Gain improvement by tuning the FBG central wavelength for the signal at different wavelengths. ....................... 91 Figure 8.5 Comparison of signal constellations between without (w/o) and with (w.) FBG central wavelength tuning. (a) Signal with a wavelength of 1569.8 nm benefits from FBG central wavelength tuning; (b) Signal with a wavelength of 1568.6 nm shows no improvement from FBG central wavelength tuning. .... 92 xiii Figure 8.6 Comparison of signal net gain (the signal power difference between Node-1 and Node-5 in Fig. 2(a)) under different scenarios: (1) with FBG tuning; (2) without FBG tuning; (3) with Raman amplification only (PSA pump is off); and (4) with waveshaper. ............................................................... 93 Figure 8.7 Comparison of measured BER vs. input power for 20 and 25 Gbaud QPSK signals under two scenarios: (1) without FBG central wavelength tuning; (2) with FBG central wavelength tuning. ................................................ 93 Figure 8.8 Comparison of (a) constellation and (b) BER of a 10 Gbaud 16-QAM signal with and without FBG central wavelength tuning. ................................... 94 xiv Abstract Optical fiber communication has the merit of high spectrum bandwidth (>20THz) and can transmit a significant amount of data information within a second. For a typical dual-polarization wavelength-division-multiplexing (WDM) system, this considerable data transmission bandwidth could support ~80 independent channels with a spacing of 50 GHz. However, this wide optical communication spectrum is being used up gradually with the rise of cloud computing, e-commerce and other internet services that require large amounts of the data transmission rate. In this regard, the manner of effectively and efficiently utilizing available optical communication spectrum resources accounts for a considerable part of research activities. In order to achieve efficient optical spectrum utilization, a common approach could be employing advanced data modulation format, enabling to encode more data information (bits) onto each symbol. In such a case, each 50-GHz bandwidth could support larger transmission capacity. A typical advanced modulation format could be quadrature-amplitude-modulation (QAM), which encodes the data information onto both the amplitude and phase of the optical carrier. However, as the QAM order increases, the deceased Euclidean distance of the symbol constellation makes the signal more vulnerable to various noises coming from both the transceivers and the transmission link. Another approach for efficient usage of the limited spectrum is to manipulate the bandwidth of the data channel. There are different manners such as using Nyquist pulse-shaping to reduce the data bandwidth to the baudrate; partially overlapping the data spectrum to include more channels within a given optical bandwidth; or transforming a double-sideband (DSB) channel to a single-sideband (SSB) channel to save half of the spectrum. Optical signal processing (OSP) could be an essential tool to enable efficient spectrum utilization from both aspects mentioned above. Compared to electrical xv approaches, OSP has potential advantages such as (i) being capable of high-speed signal processing; (ii) keeping the data in the optical domain without electrical- optical-electrical (O-E-O) conversion during transmission; and (iii) being relatively transparent to different modulation format and baudrate. In the first half of this dissertation, different methods based on bandwidth manipulation to enhance the efficient spectrum utilization are proposed and experimentally demonstrated. There are three topics, which are (i) reconfigurable optical channel slicing and stitching to enable fragmented bandwidth allocation onto discrete frequency slots; (ii) optical inter-channel-interference mitigation for spectrum overlapped channels; and (iii) tunable optical SSB generation for intensity modulation signals. In the second half, three noise mitigation methods and an optical amplification scheme are proposed and experimentally demonstrated for advanced modulation formats. These topics are: (i) channel performance enhancement by suppressing amplified-spontaneous-emission (ASE) noise through channel correlation and joint signal processing; (ii) optical multi-level amplitude-noise mitigation using optical parametric amplification (OPA); (iii) self-homodyne detection with a low-power pilot-tone to reduce the phase noise; and (iv) a low-loss phase sensitive amplification (PSA) to boost the signal power. 16 Chapter 1 Introduction In this chapter, the general idea of OSP based on the χ (2) and χ (3) nonlinear processes are first reviewed. Then, different basic enabling technologies, such as optical frequency combs, coherent detections, and advanced modulation formats, are discussed. Next, several basic enabling nonlinear operations, including phase- preserving wavelength conversion, wavelength multicasting, and wavelength multiplexing, are introduced. Finally, the organization and outline of the rest dissertation are explained. 1.1 Nonlinear Wave Mixing Processes for OSP Nonlinear wave mixing is the fundamental process for OSP [1][2]. In general, nonlinear wave mixing is the interaction among multiple optical waveforms at different wavelength (frequency) in a nonlinear media and the generation of new a waveform at another wavelength. A nonlinear device such as periodically poled lithium niobate (PPLN) has second-order susceptibility χ (2) , which enables three- wave mixing as sum frequency generation (SFG), difference frequency generation (DFG), second harmonic generation (SHG), and a cascading of differernt mixing processes [3][4]. A nonlinear device such as highly nonlinear fiber (HNLF) has third-order susceptibility χ (3) , which can result in self-phase modulation (SPM), cross-phase modulation (XPM), and four-wave mixing (FWM) [3][5]. The efficiency of all the process depends on the phase-matching conditions among the multiple interacting waveforms and the cumulative dispersion would decrease the process efficiency. Therefore, for OSP, the wavelengths of the interacting waveforms are typically chosen in a region of the low dispersion window. In the following part, χ (2) (three-wave mixing) and χ (3) (four-wave mixing) effects are explained in details. 17 1.1.1 Three-wave Mixing Three-wave mixing is a 2 nd order nonlinearity, which includes SHG, SFG, and DFG. For optical waveforms at two frequencies (f 1 and f 2 ), the processes of SHG, SFG, and DFG correspond to the generation of the new signal at 2f 1 /2f 2 , f 1 -f 2 , f 1 +f 2 , respectively. However, the frequency of the newly generated signals is usually out of the conventional wavelength (frequency) window for the signal transmission. Therefore, the cascaded SFG+DFG or SHG+DFG are usually employed to ensure the newly generated signal fall within the same frequency band. The two types of cascaded χ (2) effect are described in Fig. 1.1. In the case of cascaded SFG+DFG, the input signal firstly interact with continues wave (CW) pump, and an SFG component is generated at f signal +f pump . In the next stage, the SFG component is mixed with a dummy pump through DFG. The final output (idler) is the copy of the original signal. In the case of cascaded SHG+DFG, the CW pump interacts with itself and converts to the 2 nd harmonic. Then, the SHG component interacts with the signal through DFG, the final output (idler) is the conjugate copy of the original signal. Figure 1.1 Two types of cascaded three-wave mixing: (a) cascaded sum frequency generation (SFG) + difference frequency generations (DFG), and (b) cascaded second harmonic generation (SHG) + difference frequency generation (DFG). The frequencies (wavelengths) of the pump(s) and the signal are chosen near to the quasi-phase matching (QPM) frequency for an efficient wave mixing. SFG f idler f pump f SFG QPM DFG f signal f dummy Cascaded SFG and DFG with  (2) effect @QPM DFG * SHG f pump f SHG f signal f idler Cascaded SHG and DFG with  (2) effect (a) (b) QPM: Quasi-phase matching 18 1.1.2 Four-Wave Mixing Four-wave mixing (FWM) is a 3 rd order nonlinearity. In this process, a fourth waveform is generated from the nonlinear wave mixing of three waveforms. Depending on the number of CW pumps, FWM can be divided ito degenerate FWM (a single pump) and non-degenerate FWM (two pumps), which are shown in Fig. 1.2. Figure 1.2 Two types of four-wave mixing (FWM) process: (a) degenerate FWM with a single pump to produce an idler with phase-conjugating of the input signal, and (b) non-degenerate FWM with two pumps to produce three idlers, one of which is the signal copy and the other two are conjugate signal copies. In the scenario of degenerate FWM, a CW pump is mixed with the input signal and generate an idler at f idler , which is 2f pump -f signal . The phase of the idler is φ idler =2φ pump - φ signal . The idler with such a phase is called the conjugate signal copy. If the pump is placed at the zero-dispersion wavelength (ZDW), the wavelength (frequency) conversion efficiency would be maximized. In the scenario of non- degenerate FWM, there are two pumps being located around ZDW in symmetric. Through the nonlinear wave mixing, three idlers are generated with the following frequencies and phases 𝑓 𝑖𝑑𝑙𝑒𝑟 1 = 2𝑓 𝑝𝑢𝑚𝑝 1 − 𝑓 𝑠𝑖𝑔𝑛𝑎𝑙 (1.1) 𝑓 𝑖𝑑𝑙𝑒𝑟 2 = 𝑓 𝑠𝑖𝑔𝑛𝑎 𝑙 + 𝑓 𝑝𝑢𝑚𝑝 2 − 𝑓 𝑝𝑢𝑚𝑝 1 (1.2) 𝑓 𝑖𝑑𝑙𝑒𝑟 3 = 𝑓 𝑝𝑢𝑚𝑝 1 + 𝑓 𝑝𝑢𝑚𝑝 2 − 𝑓 𝑠𝑖𝑔𝑛𝑎𝑙 (1.3) * f signal f pump @ZDW f idler * * f pump1 f pump2 f idler3 f idler1 f idler2 f signal ZDW ZDW: Zero-dispersion wavelength Degenerate FWM with  (3) effect Non-degenerate FWM with  (3) effect (a) (b) 19 𝜑 𝑖𝑑𝑙𝑒𝑟 1 = 2𝜑 𝑝𝑢𝑚𝑝 1 − 𝜑 𝑠𝑖𝑔𝑛𝑎𝑙 (1.4) 𝜑 𝑖𝑑𝑙𝑒𝑟 2 = 𝜑 𝑠𝑖𝑔𝑛𝑎𝑙 + 𝜑 𝑝𝑢𝑚𝑝 2 − 𝜑 𝑝𝑢𝑚𝑝 1 (1.5) 𝜑 𝑖𝑑𝑙𝑒 𝑟 3 = 𝜑 𝑝𝑢𝑚𝑝 1 + 𝜑 𝑝𝑢𝑚𝑝 2 − 𝜑 𝑠𝑖𝑔𝑛𝑎𝑙 (1.6) It is worth mentioning that both degenerate and non-degenerate FWM involves two frequencies addition and one frequency subtraction. If all three frequencies are in the same frequency band, then the generated fourth waveform will be also in the same band. The result of the FWM is similar to the cascaded SHG+DFG or SFG+DFG. A typical device having FWM is HNLF, whose dispersion can be engineered to meet the phase-matching conditions within a specific frequency band. A typical device having three-wave mixing is PPLN, whose size could be much smaller than HNLF but still has a strong χ (2) effect. For both devices, the ZDW or QPM could be tuned by ~1 nm (~125 GHz) by adjusting the device temperature. 1.2 Basic Enabling Technologies Previous part explains the foundation of OSP, which is nonlinear wave mixing. This part addresses different necessary enabling technologies to advance OSP, which include advanced modulation format, coherent detection, and optical frequency combs. 1.2.1 Advanced Modulation Format In the digital communication system, the data information is usually represented by a sequence of alternative “0s” and “1s”. In optical communications, this sequence is encoded on the optical waveform. An optical waveform could be represented by 20 𝑥 (𝑡 ) = 𝐴 𝑒 𝑗𝜑 𝑒 𝑗 2𝜋𝑐 𝜆 𝑡 (1.7) Here, A is the amplitude. C is the speed of light, which is 3 × 10 8 m/s. Symbol φ denotes the phase, and λ represents the wavelength. Figure 1.3 shows different modulation formats and the corresponding signal constellation. Previously, only the amplitude of the waveform is encoded (modulated). By turning on and off the light, a single “1” or “0” can be transmitted. This modulation format is called on-off-keyed (OOK), which was widely deployed in the previous 10G or below optical communication systems. In such a case, each symbol carries only 1 bit, which might not be spectrum efficient. Instead of using the amplitude to carry the data information, the phase of the waveform could also be used. The data information can be represented by distinctive discrete phase values, and this encoding scheme is called phase-shift-keyed (PSK). For example, φ=0 denotes the information “0” while φ= π denotes the information “1”, which is called binary-phase-shift-keyed (BPSK). In such a case, each symbol still carries 1-bit information. However, we can increase the set size of the available phases for the data encoding. Therefore, each symbol could carry more information. For instance, the phase set could be {π/4, 3π/4, 5π/4, 7π/4}, and each phase could represent 2 bits as “00”, “01”, “10”, “11”, respectively. This encoding scheme is called quadrature- phase-shift-keyed (QPSK), which is an advanced modulation format being widely used for 100G communication systems nowadays. Furthermore, we can encode the data information onto both the amplitude and phase dimensions, and this modulation scheme is called quadrature-amplitude-modulation (QAM). A 16-QAM data symbol could carry 4-bits data information, which is a promising format candidate for the incoming 400G and above communication systems. 21 The signals with the advanced modulation format are typically received and demodulated using coherent detection, which will be discussed in the following section. Figure 1.3 Different modulation formats with corresponding waveforms and constellations. 1.2.2 Coherent Detection In the 80’s and 90’s, coherent detection in optical communications was extensively studies. Compared to the conventional direct detection, coherent detection enjoys higher receiver sensitivity by using local oscillator laser to amplify the received signal. However, with the invention and deployment of Erbium-doped optical fiber amplifier (EDFA), coherent detection lost the attention. After more than 0 1 1 0 1 0 0 1 01 10 00 11 0110 0011 0 1 1 0 0 0 1 1 -1 1 ( φ= π /4) ( φ=3π/4) ( φ=7 π/4) ( φ=5 π/4) BPSK QPSK OOK 16-QAM Waveform Constellation I Q I Q I Q I Q Time Time Time Time I: in-phase Q: quadrature 22 a decade, because of the fast development of digital signal processing (DSP), coherent detection has come back. The reason is that the coherent detection can capture both I and Q components, and therefore the whole information of the received optical signal can be preserved in the electrical domain. In such a case, DSP algorithms can be used to compensate various transmission distortions, which significantly increase the robustness of the optical communication systems. In addition, the advanced modulation formats as well as other state-of-art technologies could be explored. Figure 1.4 shows a single-polarization structure of a coherent detection with the post DSP algorithms. The received optical signal beats with a local oscillator (LO) laser in a 90° hybrid. By using the balanced detection consisting of two photodiodes (PDs), the direct current (DC) components of |𝐸 𝑠 (𝑡 )| 2 and |𝐸 𝐿𝑂 (𝑡 )| 2 from the precious beating process can be canceled, keeping only the I and Q component of the received signal. After being digitized by analog-to-digit converters (ADCs), different DSP algorithms are used to compensate channel distortions. After all the equalizations, the compensated signal is decoded to recover the original information. For dual-polarization signal transmission, an additional similar structure will be used to recover the data information on the other polarization. Figure 1.4 The structure of a coherent detection system with DSP to compensate for various transmission distortions, such as chromatic dispersion, polarization-mode dispersion, and carrier phase noise. 90 o Hybrid LO Laser Input Signal E S (t) E LO DSP I/Q Imbalance Comp. Clock Recovery CD Comp. PMD Comp. Frequency Offset Comp. Phase Recovery Data Decision Balanced Detection (2 PDs) Balanced Detection (2 PDs) Q(t) I(t) ADC ADC LO: Local oscillator CD: Chromatic dispersion PMD: Polarization-mode dispersion 23 1.2.3 Optical Frequency Combs Optical frequency combs can be generated using a mode-locked laser [6] or FWM [7]. Optical frequency combs consist of a series of “fingers” being located with equal distance in the frequency domain. The nature of frequency locking among the “fingers” means the frequency difference between adjacent comb lines stays the same even though the frequency of the original laser drifts with the time. The comb lines are also phase locked, which indicates the phase noise on each comb lines is the same and it only depends on the linewidth of the original laser. This unique characteristic will enable several basic OSP operations such as phase-preserving wavelength multiplexing, which might not be achievable using separate independent laser sources. Figure 1.5 Concept of optical frequency combs. FSR: free spectrum range. 1.3 Basic Enabling Operations This section describes several fundamental blocks (operations) for OSP, which includes: (i) comb-based phase-preserving wavelength conversion; (ii) optical wavelength multicasting; and (iii) optical wavelength multiplexing. 1.3.1 Phase-Preserving Wavelength Conversion As explained in Section 1.1, wavelength conversion (idler generation) could be achieved by either χ (3) effect or cascaded χ (2) effect. However, in both scenarios, the phase of the generated idler does not lock to that of the original signal. The phase unlocking comes from the phase noise of the independent pump(s). In some cases, FSR: Free spectral range f FSR  Equal distance  Frequency locking  Phase locking 24 we might want to coherently combine the idler and signal to achieve other advanced OSP functions or operations. However, the issue of phase unlocking would prevent the further process. Since comb lines are inherently phase locked, instead of using separate pump lasers, we could use comb lines to achieve the wavelength conversion as shown in Fig. 1.6. In such a case, the phases of the signal and the idler are the same, which enables an effective coherent combination between the signal and the idler for other potential advanced OSP operations. Figure 1.6 Concept of phase-preserving wavelength conversion using a pair of comb lines. 1.3.2 Wavelength Multicasting and Multiplexing In optical communication networks, wavelength multicasting is a basic function, which is to make multiple copies of the original signal at different wavelengths so that different customers can receive the broadcasting information at their operation frequency window. By using multiple pumps and nonlinear wave mixing in PPLN or HNLF, wavelength multicasting can be achieved as shown in Fig. 1.7(a). With N CW pumps, N-1 signal copies are generated. It is noted that if applications only require wavelength multicasting without further operations, separate laser sources could be used. However, if wavelength conversion is only an intermediate stage for further joint signal processing, such as adding all the weighted f Comb1 f Comb2 f Idler f Signal f 25 signal copies together with different delays to achieve finite-impulse-response (FIR) filtering, the coherence among different signal copies are requested. In such a case, optical frequency combs are required to ensure coherence. Another basic operation in OSP is wavelength multiplexing, which is to add different signal copies together coherently as depicted in Fig. 1.7(b). In such a case, optical frequency combs are required. Figure 1.7 Concepts of (a) wavelength multicasting, and (b) wavelength multiplexing. 1.4 Dissertation Outline This dissertation covers two approaches for efficient optical spectrum utilization. One approach is to manipulate the channel bandwidth while different methods will be discussed from Chapter 2 to Chapter 4. The other approach is to employ advanced modulation formats, which increases the number of bits that each symbol could carry. However, with increasing density of the signal constellation, the signal is more vulnerable to various channel noises such as ASE noise, amplitude noise, and phase noise. Chapters 5-8 focus on addressing the mitigation of different noises and the optical amplification of the signal with the advanced modulation format. Chapter 2 proposes a reconfigurable optical channel slicing and stitching to allocate an optical channel onto available discrete frequency slots. Therefore, the optical spectrum could be better utilized also in terms of the time efficiency. Compared to direct channel insertion which causes severe inter-channels-interference (ICI), the proposed method provides a ~6 dB OSNR benefit. Chapter 3 explains a scheme for spectrum saving by tightly allocating multiple channels with partially overlapping spectra. Then, an optical scheme to mitigate ICI f P f Idler1 f Signal f f Idler2 f Idler3 f P1 f P2 f P3 f comb1 f S2 f S1 f f S3 f S_mux f comb2 f comb3 f P Multicasting (comb is not required) Multiplexing (comb is required) S_Mux=S1+S2+S3 (a) (b) 26 distortion is introduced. Compared to conventional electrical approaches, the ICI effect could be mitigated before channel detection without receiving other adjacent contributing channels. Chapter 4 presents an optical SSB generation scheme using optical frequency combs. This scheme is demonstrated by different intensity-modulated signals such as OOK and PAM4. Compared to the DSB signal, the optical SSB signal has almost half of the spectrum with a ~1.5 dB system penalty. Chapter 5 introduces a method of high-order QAM channel enhancement by borrowing the extra system margins from adjacent robust BPSK/QPSK channels. After joint detection and post signal processing, ASE noise could be suppressed. A 3-dB OSNR improvement is observed by using two correlated robust channels. Chapter 6 proposes amplitude noise mitigation for a multi-level QAM signal using three different OPA regions. A ~20% decrease of the amplitude noise is observed afterward. Chapter 7 describes an optical self-homodyne detection using a low power pilot tone to compensate laser phase noise. The proposed detection scheme could work when the pilot to signal ratio (PSR) is as low as -30 dB. Chapter 8 proposes a low-loss Raman-assisted phase sensitive amplifier (PSA) to boost different advanced modulation formats such as QPSK and 16QAM. A >20 dB signal net gain is obtained in the experiment. 27 Chapter 2 Reconfigurable Optical Channel Slicing and Stitching for Fragmented Bandwidth Allocation 2.1 Introduction Elastic optical networks (EONs) are becoming of great interest owing to their ability to handle heterogeneous data channels with flexible bandwidths [1]-13]. Fragmented bandwidth allocation would be crucial in EONs to ensure the efficient use of the precious optical transmission spectrum. A key function of fragmented bandwidth allocation is the capability of adding a given data channel into any available fragmented frequency slots [14]. As an example, consider a situation in which an optical channel with a 200 GHz bandwidth is requested while the current optical network can provide only two separate spectral regions of 50 GHz and 150 GHz. A straightforward approach would be to (i) down-convert the 200 GHz optical data channel to the electrical/digital domain; (ii) decompose the original channel into two sub-channels with bandwidths of 50 GHz and 150 GHz separately; (iii) generate two optical channels that “fit” within the two available spectral regions by using two optical transmitters; (iv) reverse the sequence on reception to restore the 200 GHz channel. However, it might be valuable to achieve this goal using high-speed optical signal processing, in order to avoid inefficient O-E-O conversion. In this case, the original data channel is optically sliced into several partial frequency components, which could be assigned into the available fragmented frequency slots with the assistance of optical wavelength conversion. On reception, the sliced components are stitched together for channel reconstruction. In this chapter, we proposes the use of nonlinear wave mixing and a coherent optical frequency comb to implement reconfigurable channel slicing and stitching for an optical signal [15]. Feasibility of this approach is experimentally demonstrated using QPSK and 16QAM. The system performance is evaluated by tuning different parameters. In addition to 28 investigating the effect of relative phase offset, we further explore: (i) the influence of the relative amplitude and the number of channel slices on system performance; (ii) a 10-km transmission experiment compared to a back-to-back (B2B) scenario. One application of the channel slicing and stitching is to enable fragmented bandwidth allocation, which is experimentally demonstrated in a dense 6-channel wavelength-division-multiplexing (WDM) system. The experimental results show that the incoming 20 Gbaud optical QPSK channel is successfully reallocated into two available fragmented frequency slots and reconstructed at the receiver by using channel slicing and stitching. 2.2 Concept The conceptual diagram of fragmented bandwidth allocation enabled by channel slicing and stitching is illustrated in Fig. 2.1. Assume the current optical spectrum is occupied by multiple data channels with only two small frequency slots (Slot-1 and Slot-2) available. The incoming optical channel (S) has a large bandwidth that cannot be accommodated by any single frequency slot without introducing severe ICI from spectrum overlapping. However, the total bandwidth of the separate available frequency slots is larger than that of channel S. In this case, channel S can be sliced into two spectral fragments, which are then reallocated into the available frequency slots. The detail of this process is shown in Fig. 2.1. In the beginning, a coherent copy of channel S is generated at another wavelength by nonlinear wave mixing of channel S with an optical frequency comb [16][17]. After channel copy generation, an optical filter is employed to slice partial spectra of the two channels. It is noted that the combination of the two output channel slices (S1 and S2) should preserve all the information of the original channel S. Then, S1 and S2 are narrow enough to be inserted into the two frequency slots for transmission. To reconstruct the original channel S at the receiver, the two channel slices S1 and S2 are first selected from the current WDM system. Then, another stage of comb-based wavelength conversion is employed to recombine S1 and S2 in phase for channel recovery. Because of non-ideal filtering in both stages of spectrum 29 filtering and slice selection, S1 and S2 may have a partially overlapped spectrum, which can then produce inter-symbol interference (ISI), as shown in Fig. 2.1. However, the effect of ISI can be readily compensated by a digital linear equalizer afterwards and the original channel S can ultimately be recovered. Note that this channel slicing and stitching technique is scalable to more than two slices simply by generating more copies of the original data channel and by following the procedure shown in Fig. 2.1. Figure 2.1 Conceptual diagram of fragmented bandwidth allocation enabled by channel slicing and stitching. 2.3 Experimental Setup for a Single Channel System The previous section indicates that the key function to achieving fragmented bandwidth allocation is the channel slicing and stitching. Fig. 2.2 shows a single channel experimental setup to demonstrate this function. … … f Inserted Channel Slices S1 S2 Incoming Optical Channel Copy Generation Spectrum Filtering f f Channel Slicing S1 S2 f S f f Slice Selection Slice Recombination f S1 S2 Channel Stitching Receiver with Channel Equalization ISI IQ Modulator Pre-Amp 20G Comb Source SLM Filter-1 PPLN-1 SLM Filter-2 PPLN-2 Coherent Receiver w. Channel Equalization 2nm 1nm 2nm 2nm ~1542.53nm 0.15W 0.4W 0.5W 1nm 0.06W 1nm SLM: Spatial Light Modulator PPLN: Periodically Poled Lithium Niobate Pre-Amp: Preamplifier SMF: Single Mode Fiber 20/28-Gbaud QPSK or 20-Gbaud 16QAM Channel Slicing Channel Stitching λ LO λ S 1 2 4 5 3 10km SMF slicing & phase offset tuning between S1/S2 30 Figure 2.2 Experimental setup of channel slicing and stitching for a single optical channel. The optical frequency comb is employed to achieve phase-preserving channel copy generation and channel slices recombination (stitching). At the transmitter, the data signal with the format of 20/28 Gbaud QPSK (and alternatively 20 Gbaud 16QAM) modulates a laser at the wavelength of 1542.53 nm in an optical IQ modulator. Note that the output optical signal is not pulse shaped. At the same time, an optical frequency comb with a 20 GHz repetition rate is generated by a MLL. In the case of slicing channel spectrum into two parts (S1 and S2), two comb lines at the wavelengths of 1538.90 nm and 1539.86 nm are selected by a spatial light modulator (SLM-1) filter. The wavelength difference between the two comb lines determines the wavelength shift of the channel copy, which is generated in the next step. In the following, both the signal (S) and the selected comb lines are amplified and injected into PPLN-1 waveguide with a QPM wavelength of 1541.00 nm. After PPLN-1, a channel copy is generated, albeit with 10 dB lower power due to the conversion efficiency. The input and output of PPLN-1 are shown in Fig. 2.3(a,b). Then, another SLM filter (SLM-2) is used to cut a left slice (S1) from the original channel and a right slice (S2) from the channel copy. The corresponding spectrum is shown in Fig. 2.3(c), in which the optical channel bandwidths of S1 and S2 are ~27 GHz and ~18 GHz, respectively. Afterwards, S1 and S2 are sent to either (i) channel stitching for the signal reconstruction, or (ii) a 10-km single mode fiber (SMF) for transmission. Wavelength (nm) 10dB/D 1537 1539 1541 1543 1545 10dB/D Wavelength (nm) 10dB/D Wavelength (nm) 10dB/D 10dB/D 10dB/D 1 2 3 4 5 5 Signal (S) 2 comb lines Signal Copy Signal (S) 2 comb lines Left Slice (S1) Right Slice (S2) 2 comb lines 2 comb lines Constructive Stitching 2 comb lines Destructive Stitching Before Channel Slicing 1537 1539 1541 1543 1545 1537 1539 1541 1543 1545 Wavelength (nm) 1537 1539 1541 1543 1545 Wavelength (nm) Wavelength (nm) 1537 1539 1541 1543 1545 1537 1539 1541 1543 1545 8 (a) (b) (c) (d) (e) (f) 31 Figure 2.3 Measured spectra: (a) before PPLN-1; (b) after PPLN-1; (c) after SLM-2 filter; (d) before PPLN-2; (e) after PPLN-2 with constructive channel stitching; (f) after PPLN-2 with destructive channel stitching. Before the receiver, S1 and S2 are amplified and mixed with two comb lines with the same wavelength difference in a second PPLN (PPLN-2) of the same QPM wavelength, as shown in Fig. 2.3(d). After PPLN-2, S1 is shifted to the right with a conversion efficiency of -10 dB and recombined with S2. Since S2 is about 10 dB lower than S1 as shown in Fig. 2.3(c), the two channel slices will now be combined with almost the same power. Because of non-ideal filtering, S1 and S2 have ~5 GHz partial spectrum overlap. As a result of this overlap, tuning the phase offset (Δφ) between S1 and S2 in SLM-2 can lead to constructive (Δφ=0) or destructive (Δφ=180°) channel stitching. As shown in Fig. 2.3(e,f), destructive stitching should be avoided due to the loss of partial channel spectrum. Finally, the stitched channel is filtered out and sent to an optical coherent receiver. Based on a conventional decision-directed algorithm [18], digital channel equalization with 11 taps is used to remove spectrum-overlapping-induced ISI. If the amount of spectrum overlap increases, more taps might be required to compensate for the increased ISI. Note that in Fig. 2.3(a), the input optical signal includes both the fundamental band and side lobes. The peak power difference between them is more than 15 dB, and we do not observe a significant contribution of the side lobes to the signal quality. Therefore, the output signal is allowed to primarily be composed of the fundamental band component, and we consider the channel bandwidth to be the frequency range of the fundamental band component. Moreover, in order to use channel slicing and stitching for the given WDM grid: (i) the bandwidth of the input signal should be larger than the spectral grid spacing; and (ii) the bandwidths of the sliced signals should be smaller than the spectral grid spacing. 2.4 Experimental Results for a Single Channel System 32 Fig. 2.4 shows constellation diagrams of a 20 Gbaud QPSK channel with 30 dB OSNR under different scenarios. Compared to a B2B baseline shown in Fig. 2.4(a), Fig. 2.4(b,c) indicates that the channel quality deteriorates if only a partial spectrum is detected. The different constellations between the left and right slices are the result of unequal bandwidths of the two slices. Fig. 2.4(d) shows that the channel is successfully recovered after channel stitching of the left and right slices. The signal quality is almost preserved after 10-km transmission as shown in Fig. 2.4(e), because the channel equalization at the receiver also compensates for chromatic dispersion [19-21]. Figure 2.4 Constellation comparison of a 20 Gbaud QPSK channel under different scenarios: (a) B2B without channel slicing and stitching; (b) detection of only the left channel slice; (c) detection of only the right channel slice; (d) B2B with channel slicing and stitching; (e) after 10-km transmission with channel slicing and stitching. The system performance is further evaluated by tuning the relative phase offset (Δφ) and relative amplitude (Δα) between S1 and S2 in SLM-2. The reason to investigate the impact of phase/amplitude imbalance is because the two parameters could affect the performance of channel stitching. In Fig. 2.5, when the phase is aligned (Δφ=0), channel equalization can compensate for the ISI effect due to the partial spectrum overlapping of the left and right channel slices. Channel equalization helps to decrease the error-vector- magnitude (EVM) from ~20% down to ~10%, as shown in Fig. 2.5(c). Additionally, the tolerance of Δφ can be as large as 150° with channel equalization, whereas the system without channel equalization fails if Δφ is 90°. Similarly, in terms of the relative amplitude Δα, the system with channel equalization can still work when Δα is 25 dB, whereas an equalization-free system cannot tolerate Δα of 20 dB. Left Slice Only B2B Slice & Stitch B2B Right Slice Only 10-km Transmission Slice & Stitch (a) (b) (c) (d) (e) 33 As a result, digital channel equalization not only enhances the performance of channel stitching by compensating for ISI, but it also increases the system tolerance for phase/amplitude imbalance. Additionally, a lower signal EVM can be obtained if the phase/amplitude imbalance is pre-compensated optically by SLM-2, as shown in Fig. 2.5(a,c) and Fig. 2.6(a,c). Therefore, in the remainder of the chapter: (i) channel equalization is included primarily to compensate for ISI; and (ii) the phase/amplitude imbalance is pre- compensated in the optical domain. Figure 2.5 Constellation comparison of a 20 Gbaud QPSK by tuning the relative phase offset (Δφ) between two channel slices: (a) with channel equalization; (b) without channel equalization. (c) Measured EVMs with different Δφ. w. Channel Equalization w/o Channel Equalization Δφ=0 deg Δφ=30 deg Δφ=60 deg Δφ=90 deg Δφ=120 deg Δφ=150 deg Δφ=180 deg Δφ=0 deg Δφ=30 deg Δφ=60 deg Δφ=90 deg Δφ (deg) EVM (%) (a) (b) (c) 0 10 20 30 40 50 60 0 50 100 150 200 w. Channel Equalization w/o Channel Equalization 10% 34 Figure 2.6 Constellation comparison of a 20 Gbaud QPSK by tuning the relative amplitude (Δα) between two channel slices: (a) with channel equalization; (b) without channel equalization. (c) Measured EVMs with different Δα. Fig. 2.7 shows the measured BER curves of the system. Compared to a B2B baseline system, the optical signal-to-noise ratio (OSNR) penalty of channel slicing and stitching with channel equalization is below 1 dB, as shown in Fig. 2.7(a). It is noted that for the B2B baseline system, the same channel equalization is used for comparison. In addition, the two nearly overlapping BER curves in Fig. 2.7(b) indicate the system penalty for the 10-km transmission is negligible. For longer- distance transmission, the increased chromatic dispersion, as well as the wavelength- dependent polarization rotation caused by higher order polarization mode dispersion (PMD), could affect the phase alignment among different channel slices. As a result, the performance of channel stitching might be degraded, which requires further investigation. Δα=0 dB EVM (%) Δα (dB) Δα=3 dB Δα=10 dB Δα=20 dB Δα=25 dB Δα=30 dB Δα=0 dB Δα=3 dB Δα=10 dB Δα=20 dB w. Channel Equalization w/o Channel Equalization (a) (b) (c) 0 10 20 30 40 50 60 0 10 20 30 40 w. Channel Equalization w/o Channel Equalization 35 Figure 2.7 BER comparison for a 20 Gbaud QPSK system: (a) with and without digital equalization; (b) B2B and 10-km transmission. The spectra of the channel slicing and stitching with 3 channel slices of a 28 Gbaud QPSK channel are shown in Fig. 2.8(a-c), while the corresponding separate channel slices and reconstructed channel constellations are shown in Fig. 2.8(d). Figure 2.8 Channel slicing and stitching with three slices for a 28 Gbaud QPSK system: (a) optical spectrum before PPLN-1; (b) optical spectrum after SLM-2 filter; (c) optical spectrum after PPLN-2; (d) channel reconstruction by stitching three channel slices. The BER curves of the 28 Gbaud QPSK system with two and three channel slices are presented in Fig. 2.9. It can be seen that the system performance does not strongly depend on the number of channel slices, and less than 1-dB OSNR penalty is observed compared to the B2B baseline. OSNR (dB) BER (a) (b) 1E-5 1E-4 1E-3 1E-2 1E-1 7 9 11 13 15 17 Slice & Stitch (20-Gbaud) Slice & Stitch w/o Equalizer (20-Gbaud) B2B (20-Gbaud) 3 dB OSNR (dB) BER 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 7 8 9 10 11 12 13 Slice & Stitch (20-Gbaud,B2B) Slice & Stitch (20-Gbaud,10 km) Wavelength (nm) Power (10 dB/D) 1537 1539 1541 1543 1545 Signal (S) 3 comb lines S1 S2 S3 3 comb lines Stitching S1, S2 and S3 Wavelength (nm) 1537 1539 1541 1543 1545 Wavelength (nm) 1537 1539 1541 1543 1545 Slice 1 Slice 2 Slice 3 Channel Stitching (b) (a) (c) (d) Power (10 dB/D) Power (10 dB/D) 36 Figure 2.9 BER comparison with different numbers of channel slices. The scheme is extended to a 20 Gbaud 16QAM signal. Less than 1.5% EVM deterioration with 30 dB OSNR is observed, as shown in Fig. 2.10(a). Compared to the QPSK scenario in Fig. 7(a), a larger OSNR penalty is observed for 16QAM in Fig. 2.10(b). A possible reason could be that high order QAM signals are more sensitive to any distortion introduced by nonlinear-wave-mixing-based wavelength conversion [22]. Figure 2.10 (a) EVM comparison between B2B and channel slicing and stitching for a 20 Gbaud 16QAM signal; (b) BER comparison. 2.5 Experimental Setup for a WDM System The application of channel slicing and stitching to enable fragmented bandwidth allocation is experimentally demonstrated in a WDM system with 6 QPSK channels of 20 Gbaud. The experimental setup is shown in Fig. 2.11(a) and the central wavelengths of the 6 channels are 1541.68, 1542.00, 1542.16, 1542.32, 1542.52, and 1542.87 nm. Compared to the single-channel experiment shown in Fig. 2, a stage of WDM channel generation is added, as shown by the dotted box. In this OSNR (dB) BER 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 7 8 9 10 11 12 13 14 B2B (28-Gbaud) Slice & Stitch (28-Gbaud 2 Slices) Slice & Stitch (28-Gbaud 3 Slices) 1E-5 1E-4 1E-3 1E-2 1E-1 15 17 19 21 23 25 Slice & Stitch (20-Gbaud 16QAM) B2B (20-Gbaud 16QAM) Slice & Stitch (EVM=7.9%) B2B (EVM=6.6%) OSNR (dB) BER (b) (a) <2dB 37 case, the attenuator is used to adjust the power of the WDM channels, in order to make it similar to that of the added optical channel S. After the attenuator, a polarization controller is used to align the polarization of the WDM channels with that of channel S in order to maximize the ICI effect. The optical spectra before and after fragmented bandwidth allocation are shown in Fig. 2.11(b,c), in which the two sliced channels each have ~22 GHz optical bandwidth. At the receiver, an extra SLM filter (SLM-3) is used for channel-slice selection and amplitude/phase adjustment. In order to allocate the same power into the two frequency slots, the channel slice with higher power is attenuated by 10 dB in SLM-2 to offset the effect of the -10 dB conversion efficiency in PPLN-1. Subsequently, the power difference is adjusted in SLM-3 before channel stitching in PPLN-2. Figure 2.11 (a) Experimental setup for fragmented bandwidth allocation in 20 Gbaud QPSK WDM channels; (b) measured spectrum before fragmented bandwidth allocation; (c) measured spectrum after fragmented bandwidth allocation. 2.6 Experimental Results for a WDM System The constellation comparison is shown in Fig. 2.12. Compared to direct channel insertion shown in Fig. 2.12 (c,d), where the entire channel S is inserted into IQ Modulator Pre-Amp 20G Comb Source SLM Filter-1 PPLN-1 SLM Filter-2 PPLN-2 Coherent Receiver w. Channel Equalization 2nm 1nm 2nm 2nm ~1542.53nm 0.15W 0.4W 0.5W 1nm 0.06W 1nm 20-Gbaud QPSK Channel Slicing Channel Stitching λ LO λ S WDM Laser Source QPSK Modulator 20-Gbuad QPSK Pre-Amp ATT SLM Filter-3 WDM Channels SLM: Spatial Light Modulator PPLN: Periodically Poled Lithium Niobate ATT: Attenuator Pre-Amp: Preamplifier 5dB/D 1540 1542 1544 1546 Slot-1 Wavelength (nm) Slot-2 5dB/D 1540 1542 1544 1546 Wavelength (nm) After Slicing & Insertion 38 either Slot-1 or Slot-2, fragmented bandwidth allocation can effectively avoid channel spectrum overlapping and therefore suffer much less ICI penalty. The reason for signal quality deterioration in Fig. 2.12(b) compared to the single-channel scenario in Fig. 2.12(a) might be attributed to non-ideal filtering for selecting channel slices, which includes the residual spectra from adjacent channels. Figure 2.12 Constellation comparison between direct channel insertion and fragmented bandwidth allocation enabled by channel slicing and stitching. Figure 2.13 BER comparison between direct channel insertion and fragmented bandwidth allocation enabled by channel slicing and stitching. For further system evaluation, BER measurements of the added channel S are presented in Fig. 2.13. Compared to direct channel insertion, fragmented bandwidth allocation has more than 6 dB OSNR improvement at a BER of 1e-3. There is an additional OSNR penalty of channel slicing and stitching compared to the single- channel scenario. Similarly, a possible reason for this penalty could be that the filter Single Channel Slice & Stitch in WDM Channels Slot-1 Direct Insertion Slot-2 Direct Insertion (a) (b) (c) (d) OSNR (dB) BER 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 6 8 10 12 14 16 18 20 22 Single Channel Slice & Stitch in WDM Slot-1 Direct Insertion Slot-2 Direct Insertion >6dB 39 for selecting a desired channel slice is not sharp enough to reject the adjacent channels. 2.7 Discussion and Conclusion This chapter experimentally demonstrates a reconfigurable channel slicing and stitching for an optical signal to enable fragmented bandwidth allocation without O-E-O conversion. In a 6-channel WDM system, a 20 Gbaud optical channel is successfully reallocated into two fragmented frequency slots and reconstructed at the receiver. Although this scheme is demonstrated for an optical channel that is not pulse shaped, we believe the scheme might also be applicable to channels that are pulse shaped, e.g., Nyquist shaping. In our experiment, there are different issues that could degrade the system performance, including: (i) the power of the optical signal is attenuated by the loss of different equipment, such as the PPLN (~5 dB insertion loss) and the SLM filter (~6 dB insertion loss); (ii) nonlinear wave mixing in both stages of the channel slicing and stitching requires sufficient signal power as provided by a 2W EDFA with a ~6 dB noise figure; and (iii) there are optical components with limited bandwidth. We note that there are other approaches that could reduce channel bandwidth to fit into the smaller frequency slot, such as narrow filtering [23] or higher-order QAM signal conversion [24]. They may not suffer the same degradations as we have, but could introduce other issues. The reason for using an optical frequency comb instead of independent continuous wave lasers is to ensure phase locking among different channel slices, which is generally required for successful signal recovery at the receiver. In our experiment, the selected comb lines (within a ~10-nm spectrum range) have a similar OSNR of ~30 dB. As the scheme is scaled to more channel slices with larger frequency spacing, the quality of the stitched signal might be affected by different OSNRs of different comb lines. In addition, we use the same comb source for both channel slicing and stitching for ease of experimentation; a more realistic implementation would likely use two independent comb sources for the transmitter and the receiver. 40 Chapter 3 Reconfigurable Optical Inter-Channel Interference Mitigation for Spectrally Overlapped Signals 3.1 Introduction More data channels can be accommodated in a given optical bandwidth by partially overlapping the optical spectra to increase the spectrum efficiency. On the other side, the use of DSP in coherent receivers is a major advance in optical communications of recent years [25-28]. One capability of DSP is to mitigate ICI, which is a typical crosstalk phenomenon arising from the allocation of different channels with channel spacing near to the baud rate for high spectral efficiency [29-31]. A common approach to mitigate ICI in the resultant spectrally overlapped channels is to use a digital multi-channel equalization based on different algorithms after photodetection, which usually requires receiving all the relevant channels that contribute to the ICI for the channel of interest [32-41]. This approach can be achieved using either one receiver with a large electrical bandwidth [33] or several synchronized typical receivers with limited bandwidth, each responsible for acquiring a different channel [34-37]. One motivation for exploring an optical-domain approach is to mitigate ICI before detection, so that only one typical receiver is needed for the target channel. Other motivations for implementing optical ICI mitigation might include: (i) the potential for high- speed signal processing [1]; (ii) avoiding accurate channel spacing estimation (10MHz estimation error could significantly degrade the performance of digital multi-channel equalization) [32]; (iii) freedom from optoelectronic conversion, which might be preferable for signal regeneration in the middle of the link. There have been several works which attempt to use optical signal processing to mitigate ISI caused by fiber link distortion such as chromatic dispersion [42][43]. However, 41 to the best of our knowledge, optical signal processing has not been used to mitigate the ICI effect in spectrally overlapped channels. In this chapter, an optical ICI mitigation scheme based on the nonlinear wave mixing in PPLN waveguides is proposed [44]. The implementation penalty of this scheme is about 0.5dB compared with a single channel B2B baseline configuration. The performance of the proposed method is evaluated experimentally by a dual-carrier QPSK overlapped system with both 20G-baud and 25G-baud under four different channel spacing conditions. 3.2 Concept Figure 3.1 describes the proposed optical ICI mitigation scheme as composed of three stages. At the input, two channels S 1 and S 2 are partially overlapped with a certain channel spacing Δf. The entire signal is expressed as follows: S = S 1 (f − ∆f/2) + S 2 (f + ∆f/2) (3.1) In the first stage of Fig. 3.1, a conjugate copy is generated by sending S and a pump P through a PPLN waveguide via the cascaded processes of SHG and DFG. The conjugate copy is also partially overlapped, which is denoted as: S ∗ = S 1 ∗ (2f p − f + ∆f/2) + S 2 ∗ (2f p − f − ∆f/2) (3.2) in which f p denotes the frequency of the pump P. In the second stage, an optical programmable filter is utilized to select desired channels (S 1 ′ , S 2 ∗′ ) or (S 2 ′ , S 1 ∗′ ) and modify the amplitudes and phases of S 1 ∗′ or S 2 ∗′ by multiplying a complex coefficient (c 1 or c 2 ) as in Fig. 3.1. Here S 1 ′ ≈ S 1 (f − ∆f/2) + γ 1 S 2 (f + ∆f/2), (3.3) S 2 ′ ≈ S 2 (f + ∆f/2) + γ 2 S 1 (f − ∆f/2), (3.4) S 1 ∗′ ≈ S 1 ∗ (2f p − f + ∆f/2) + γ 3 S 2 ∗ (2f p − f − ∆f/2), (3.5) S 2 ∗′ ≈ S 2 ∗ (2f p − f − ∆f/2) + γ 4 S 1 ∗ (2f p − f + ∆f/2). (3.6) The above equations are similar to [37][45] in which the second terms are ICI crosstalk and γ i < 1 (i = 1,2,3,4) depends on the frequency characteristics of the optical programmable filter. Due to channel overlap, the spectra envelopes of the four selected 42 channels at the second stage in Fig. 1 are asymmetric. For ICI mitigation of S 1 , at Port 1, S 1 ′ and S 2 ∗′ are selected while S 2 ∗′ is multiplied by c 2 , whose value is determined by the current ICI impact. Pump P is also retained in this stage to maintain coherence in the next stage process. In the third stage, through SHG and DFG in another PPLN waveguide, c 2 S 2 ∗′ is added back to S 1 ′ as S ̂ 1 = (1 + c 2 ∗ γ 4 ∗ )S 1 (f − ∆f/2) + (γ 1 + c 2 ∗ )S 2 (f + ∆f/2). (3.7) By tuning c 2 , the crosstalk term can be reduced. However, the ICI crosstalk cannot be fully compensated because γ i is not a constant in the frequency domain. Since the pump P is preserved, c 2 ∗ S 2 ′ is added to S 1 ′ with the exact channel spacing of Δf, which is why accurate estimation of Δf is unnecessary. The processed channel is then filtered as S ̂ 1 ′ = H f S ̂ 1 with residual ICI and sent to the detector (H f represents the filter transfer function). For ICI mitigation of S 2 , a similar set of operations are carried out at Port2, in which the pump P, S 2 ′ and c 1 S 1 ∗′ are selected. Figure 3.1 Conceptual diagram of the proposed optical ICI mitigation scheme for an overlapped channel system. There are three processing stages in the background with different colors: (1) channel copy generation; (2) complex coefficients adjustment; (3) channel addition. PPLN: periodically poled lithium niobate. 3.3 Experimental Setup Figure 3.2(a) illustrates the experimental setup of optical ICI mitigation of a QPSK modulation format in a two-overlapped-channel system. Two separate source Copy Generation (PPLN) Pump (P) S 1 S 2 f P S 2 Optical ICI Compensation for S 1 (PPLN) Optical ICI Compensation for S 2 (PPLN) Optical Programmable Amplitude/ Phase Filter S 1 P P P P Port 1 Port 2 ICI Δf f f f f f f f Reduced ICI Reduced ICI 43 lasers are modulated by independent QPSK modulators with uncorrelated data streams. A polarization controller (PC) is added after the QPSK Modulator-1 in Fig. 3.2(a) to ensure that the two channels have the same polarization which maximizes the ICI effect. Pre-amplifiers are placed in the link of both channels to maintain the same signal power for the channels which are then combined by a 50/50 coupler to produce the overlapped channels. Figure 3.2 (a) Experimental setup for the ICI mitigation system and the corresponding optical spectra: (1) Channel-1 before overlapping, (2) Channel-2 before overlapping, (3) overlapped channels, (4) copy generation, (5) filtering and coefficient adjustment, (6) channel addiction. (b) Back-to-back (B2B) baseline configuration. (c) System implementation penalty compared with B2B baseline configuration under single channel transmission. Data 1 Coherent Receiver λ Pump 1540.7nm λ LO 2nm 1nm 200mw 1nm QPSK Modulator 2 Data 2 λ Sig2 ~1539.4nm 1nm Pre-Amp Pre-Amp 90mw PPLN 1 160mw 9nm PPLN 2 1nm λ Sig1 ~1539.2nm QPSK Modulator 1 P P S 2 S 1 S 1 * S 2 * P Data 1 Coherent Receiver λ LO 1nm QPSK Modulator 2 Data 2 1nm Pre-Amp Pre-Amp 1nm QPSK Modulator 1 (a) (b) ATT Pre-Amp SLM Filter ATT 1nm 1nm Pre-Amp 1 2 3 4 3 4 5 6 Wavelength (nm) 10dB/D Wavelength (nm) Wavelength (nm) Wavelength (nm) 10dB/D 10dB/D Wavelength (nm) Wavelength (nm) 10dB/D 10dB/D 10dB/D 1 2 5 6 OSNR Adjustment OSNR Adjustment 1539 1540 1541 1542 λ Sig2 ~1539.4nm λ Sig1 ~1539.2nm (c) 6 7 8 9 10 11 12 13 7 9 11 13 B2B System OSNR (dB) Q (dB) 1539 1540 1541 1542 1539 1540 1541 1542 1539 1540 1541 1542 1539 1540 1541 1542 1539 1540 1541 1542 S 1 S 2 S 2 S 1 44 For optical ICI mitigation, the overlapped channels are amplified to 90mW and sent into the first PPLN waveguide along with 200mW pump at the wavelength of 1540.7nm. The QPM wavelength is temperature-tuned and stabilized at around 1541nm to produce the highest conversion efficiency. After that, the conjugated copies of the channels are generated and the spectra are shown in Fig. 3.2(a). Because the channels are partially overlapped, only about half of each channel spectrum is visible. In the next stage, all the components are sent through a SLM filter, based on liquid crystal on silicon (LCoS) technology, for channel selection and amplitude/phase adjustment. For ICI mitigation of S 2 , the outputs of the SLM filter are S 2 ′ , c 1 S 1 ∗′ and P, whose spectra are shown in Fig. 3.2(a). These three components are injected into the second PPLN waveguide, where c 1 S 1 ∗′ is added back onto S 2 ′ as S 2 ′ + c 1 ∗ S 1 ′ with the precise channel spacing of Δf. Finally, a filter and OSNR adjustment are added before the coherent receiver in which signal quality is evaluated and bit errors are counted. For OSNR adjustment in Fig. 3.2(a), the output power of the pre-amplifier prior to the coherent receiver is fixed. By attenuating the signal, different levels of ASE are loaded to vary the OSNR. Similar operations are implemented on the S 1 channel to generate S 1 ′ + c 2 ∗ S 2 ′ for the ICI mitigation. Figure 2(b) is the B2B baseline counterpart, in which the overlapped channels are generated at the transmitter in the same manner while the 1nm filter and electrical filter in the same coherent receiver are both centered on the target channel carrier for signal selection as in Fig. 3.2(a). The implementation penalty of the proposed ICI mitigation system is measured by comparing with a single channel B2B baseline configuration. In this scenario, there is only a single channel without ICI crosstalk. Because nonlinear noise is primarily generated during the nonlinear wave mixing stages, the signal quality is expected to degrade. Figure 3.2(c) shows that, for the same system Q factor, 0.5dB OSNR penalty is observed in the proposed optical ICI mitigation system. 45 3.4 Results The performance of the proposed ICI mitigation method is first assessed using a dual-carrier 20G-baud QPSK signal with channel spacing (CS) of 15GHz, 17.5GHz, 20GHz, and 25GHz. The signal constellation comparisons are shown in Fig. 3.3(a). To emphasize the ICI effect, there is no spectrum shaping or filtering at the transmitter. When CS is 25GHz, which is much larger than the baudrate of the signal, the ICI effect is insignificant and the proposed ICI mitigation method provides negligible benefit. The difference between the signal constellations of the two channels might attribute to the fact that the two channels are generated by separate modulators, which would produce different signal quality. When CS is equal to or less than the baudrate, the effect of ICI becomes significant. In the meantime, the improvement due to ICI mitigation turns to be noticeable. However, the ICI effect cannot be completely compensated, so even after ICI mitigation, the signal quality with a smaller CS is still worse than the signal quality with a larger CS, as shown in Fig. 3.3(a). To demonstrate the tunability of the proposed scheme, the baudrate of both QPSK channels is changed to 25G-baud and four different CS (20GHz, 22.5GHz, 25GHz and 30GHz) are selected accordingly. These additional constellation comparisons are depicted in Fig. 3.3(b). As was observed earlier, when CS is 30GHz, which is much larger than the signal baudrate, the ICI effect is insignificant and the benefit of ICI mitigation is negligible. However, when CS decreases to or below the baudrate, the effect of ICI becomes increasingly significant and the improvement from ICI mitigation grows accordingly. Still, the signal quality with smaller CS remains worse than the signal quality with larger CS as shown in Fig. 3.3(b). As a further evaluation, the Q factors under different CS conditions are calculated and compared. Figure 3.4 shows Q factor curves for both 20G-baud and 25G-baud systems under different CS conditions. The cross and square symbols represent Q factors for 20G-baud and 25G-baud systems with ICI mitigation. The triangle and diamond symbols denote Q factors for 20G-baud and 25G-baud systems 46 with B2B baseline configurations. In Fig. 3.4(a), CS is 5GHz smaller than the baudrate and the Q factor is below the FEC threshold of 8.5dB without ICI mitigation. Meanwhile, the proposed ICI mitigation makes the system Q factor exceed the FEC threshold. When CS increases but remains no larger than the baudrate, the ICI mitigation brings significant benefit, as shown in Figs. 3.4(b) and 4(c). An OSNR benefit of more than 4dB is observed when the system Q factor reaches the FEC threshold. When CS is much larger than the baudrate as in Fig. 3.4(d), the Q factor curves almost overlap, indicating that the proposed ICI mitigation brings little improvement since the ICI effect is already very small. Figure 3.3 Signal constellations comparison with and without ICI mitigation for (a) 20G-baud and (b) 25G-baud overlapped channel systems of different channel spacing (CS). CS 25GHz CS 17.5GHz CS 20GHz B2B ICI Comp. Ch1 Ch2 CS 15GHz B2B ICI Comp. B2 B ICI Comp. Ch1 Ch2 B2 B ICI Comp. CS 22.5GHz CS 25GHz CS 30GHz CS 20GHz 20G-baud 25G-baud (a) (b) 7 7.5 8 8.5 9 9.5 10 10.5 10 14 18 22 26 7 8 9 10 11 12 10 14 18 22 26 6 7 8 9 10 11 12 13 6 8 10 12 14 16 6.5 7.5 8.5 9.5 10.5 13 18 23 28 25G-baud (B2B) 25G-baud (ICI Comp.) 20G-baud (B2B) 20G-baud (ICI Comp.) OSNR OSNR OSNR OSNR Q (dB) Q (dB) Q (dB) Q (dB) (a) (b) (c) (d) 47 Figure 3.4 Q factor comparison with and without ICI mitigation under different channel spacing (CS) and baud rates. (a) 25G-baud with 20GHz CS and 20G-baud with 15GHz CS; (b) 25G-baud with 22.5GHz CS and 20G-baud with 17.5GHz CS; (c) 25G-baud with 25GHz CS and 20G-baud with 20GHz CS; (d) 25G-baud with 30GHz CS and 20G-baud with 25GHz CS. Because the coefficient ci (i=1, 2) is complex, not only the amplitude but also the phase will affect the system performance. Figure 3.5 shows the relationship between phase and system Q factor. The corresponding signal constellations are also depicted. It can be seen that there is an optimal phase which yields the highest Q factor. Figure 3.5 System Q factor varies with the phase change of the coefficient and the corresponding signal constellation. Different phases can change the system Q factor by more than 3dB. Figure 3.6 illustrates VPI simulated EVM comparisons with and without the proposed ICI mitigation when an adaptive intra-channel equalizer with different tap numbers is employed afterwards. The tap weights for the intra-channel equalizer are updated by constant modulus algorithm (CMA) [46]. As the tap number increases, both EVMs decease and the signal quality with the proposed ICI mitigation is always better than the one without ICI mitigation. In the ICI mitigation stage, the optimal amplitude of the coefficient is approximately 0.45. When the overlapped spectrum is completely cut, the signal’s EVM is increasing. The signal quality is degraded because part of the signal information has been erased. In the experiment, the original channels and the copies may experience different linear and nonlinear phase shift due to different wavelengths and power levels, and as a result, a non-zero 7 7.5 8 8.5 9 9.5 10 10.5 11 50 70 90 110 130 150 170 190 Q (dB) Phase (deg) 48 relative phase is expected and the coefficient should be complex. However, for the simulation result in Fig. 3.6, the optimal phase for coefficient is zero because different phase shifts are not considered. Figure 3.6 Simulated EVM comparisons by employing an adaptive intra-channel equalizer with different tap numbers for three scenarios: (1) with proposed ICI mitigation; (2) without ICI mitigation; (3) cutting the overlapped spectrum. 3.5 Discussion and Conclusion In this chapter, as a proof of concept, the coefficients for ICI mitigation are optimized manually by monitoring the signal’s EVM on the coherent receiver. For more practical applications, further research might be required to design a feedback loop for adaptive coefficient adjustment. It should also be noted that after long distance transmission, the cumulated dispersion would introduce delay between different channels. This effect needs to be compensated before implementing the proposed ICI migration scheme. In addition, because only one tap is currently employed in the proposed optical ICI mitigation scheme, performance might be worse compared with conventional DSP-based ICI mitigation approaches where more taps are available. However, when more copies can be generated in the PPLN waveguide with sufficient efficiency, more taps could be employed in the optical solution. Together with the advantages such as free of multi-channel detection and channel spacing estimation, a quantitative comparison between the proposed all optical mitigation scheme and conventional DSP-based methods might be interesting. Although only two channels are demonstrated in this chapter, considering 0.05 0.15 0.25 0.35 0.45 0 10 20 30 40 50 60 w. ICI comp. w/o ICI Comp. Delete Overlapped Spec. Number of Taps in Adaptive Intra-channel Equalizer EVM 49 ±10nm QPM bandwidth in PPLN waveguides, this approach could be extended to WDM channel system. In order to mitigate the ICI effect in N channels simultaneously, the output of the SLM filter would need to be divided into N ports with one PPLN waveguide in each path. The following power consumption and cost problems might be alleviated by an implementation of this approach as a photonics integrated circuit [47]. 50 Chapter 4 Tunable Optical Single-Sideband Generation of OOK and PAM4 Data Channels 4.1 Introduction Spectrum and bandwidth are limited in optical communication systems [1][48]. The ability to generate and transmit data channels using less bandwidth has various advantages, such as accommodating more channels in a fixed spectrum [49][50]. For cost-sensitive applications, such as data center interconnections, there has been increased interest in IM/DD systems [51]. One approach to reducing bandwidth in these systems is to use higher- level formats than OOK, such as pulse-amplitude-modulation (PAM) [52]. Additionally, an old and common approach for reducing channel bandwidth is to simply generate and transmit SSB as opposed to DSB channels [53]. Another important advantage of the SSB technique is to alleviate dispersion-induced power fading in IM/DD optical communication systems without employing dispersion compensating fiber (DCF) [54][55]. SSB can be generated using either an in-phase and quadrature Mach–Zehnder modulator (IQ-MZM) [56] or a dual-drive Mach-Zehnder modulator (DD-MZM) [57], together with high-speed digital signal processing (DSP), typically including high-resolution digital-to-analog converters (DACs) [58]. There has been interest in using optical approaches to generate SSB data channels in order to potentially increase the operating speed/bandwidth and reduce the electronic processing complexity [59]. Reports have shown various optical approaches, including using: (i) multiple Mach-Zehnder interferometers with optical delay lines [60], (ii) a phase-shifted [61] or sampled fiber Bragg grating [62], and (iii) a micro-ring resonator [63]. However, these approaches might not be readily scalable or tunable. In this chapter, we demonstrate tunable optical SSB generation for up to 20-Gbit/s OOK and PAM4 data channels [64]. By assigning SSB filtering taps onto different comb lines and then multiplexing with delayed original DSB channel copies in a nonlinear device, an optical SSB data channel is generated. The sideband suppression becomes better by 51 choosing more comb lines. Moreover, by adjusting the comb spacing, the system could be tuned for different data rates. We compare the sideband suppression of a 10-Gbit/s OOK channel by varying the number of comb lines from 1 to 5. Compared to the original DSB data channel, the proposed optical SSB generation shows a ~1.5dB sensitivity penalty. The generated SSB channel is also transmitted in an 80-km standard single-mode fiber (SSMF) without chromatic dispersion compensation, and there is a >3dB receiver sensitivity improvement compared to the original DSB channel. This approach is further demonstrated for 20-Gbit/s OOK and PAM4 data channels. 4.2 Concept Figure 4.1 shows the concept of the proposed tunable optical SSB generation for a data channel using optical frequency combs and nonlinear wave-mixing. Two groups of comb lines (Group-A and Group-B) are selected from an optical comb source. Group-A is injected to an optical intensity modulator to carry the data generated from a bit pattern generator (BPG), and the output are multiple optical DSB data copies. Then, these data copies are transmitted in a dispersive device, such as DCF, to introduce an appropriate delay (N×T/2) to each data copy, where T denotes the symbol duration. Group-B is sent to a tunable SSB tap setting equipment, such as LCoS filter, to adjust the number and coefficient of the taps, which is a truncated version of the SSB impulse response. More comb lines means a better approximation of the SSB filter. The amplitude and phase profiles of the output Group-B are shown in Fig. 1. In the next, Group-B comb lines (with SSB tap encoding) and the delayed DSB copies are coupled with a single optical pump in a nonlinear optical device, such as PPLN, for optical channel multiplexing. By changing the wavelength of the single pump (λp), the wavelength of output SSB data channel (λSSB) could be tuned. 52 Figure 4.1 Concept diagram of the proposed tunable optical single-sideband generation for a data channel using optical frequency combs and nonlinear wave-mixing. The tunability of this approach could be achieved by choosing a different number and spacing of comb lines. 4.3 Experimental Setup Figure 4.2(a) illustrates the experimental setup to demonstrate the proposed optical SSB generation. A comb source with a 20 GHz repetition rate generated by a mode-locked laser is employed, and an LCoS optical filter selects two groups of comb lines (Group-A and Group-B). Group-A comb lines from Port-1 is pre- amplified and then sent to an intensity modulator to modulation either an OOK or a PAM4 data channel. The generated multiple DSB channel copies are sent to a ~500m DCF to introduce a half-symbol delay (T/2) between adjacent data copies. If a 10- Gbit/s OOK is modulated, the spacing in the selected comb group is set to four times of the repetition rate, which is 80GHz. After the DCF, a 50ps delay is introduced between adjacent data copies. If it is a 20-Gbit/s channel, the spacing in the selected comb group is changed to 40GHz, which leads to a 25ps delay after the same DCF. The Group-B comb lines from Port-2, whose amplitudes and phases are assigned with the truncated SSB filter coefficients, are amplified and coupled with the previously delayed multiple DSB channels as well as a single optical pump. These Comb source λ Copy generation Tap setting 1+j*Hilbert(.) AM RF signal Optical DSB copies Group-A (N comb lines) λ- dependent delay Group-A Group-B Delayed DSB copies 0 - π/2 π/2 Phase Amp. y(t): OSSB channel λ λp λ λ λ λ x(t)..x(t)..x(t) Nonlinear Multiplexing h-3 ..h0..h3 x(t-3T)..x(t)..x(t+3T) SSB taps λ SSB taps Group-B (N comb lines) y(t) = k= -N/2+1 N/2 Pump 53 three components are sent to a PPLN for an optical channel multiplexing. The generated SSB channel is selected by an optical sharp filter and then either sent to the PD or an 80km SSMF transmission link. Figure 4.2(b) shows the input and output spectra of the PPLN by selecting 3 comb lines as an SSB filter. The tunability of the proposed approach is explained in Fig. 4.2(c), in which the number and the spacing of comb lines could be adjusted to apply for different scenarios. Note that since some taps in SSB impulse response are zeros, the corresponding comb lines are blocked as shown in Fig. 4.2(c). That is the reason why the comb lines for 5 taps do not have equal spacing. Figure 4.2 (a) Experimental setup of tunable optical SSB generation for different data rates. The LCoS filter is used to select comb lines and set SSB tap coefficients. The data is received by a photodiode (PD) without chromatic dispersion compensation after an 80-km SSMF transmission. (b) Input and output spectra of the PPLN, in which 3 taps are used for SSB generation of a 10-Gbit/s OOK channel. (c) Input spectra of the PPLN, in which 5 taps are used for SSB generation of a 10-Gbit/s or a 20- Gbit/s OOK channel. Note that since some of the taps in SSB impulse response are zeros, the corresponding comb lines are blocked. For a 20-Gbit/s OOK channel, the spacing of the selected comb lines is reduced by half compared to that of the 10-Gbit/s channel in order to let the same DCF introduce appropriate delays. 1 Port 1 Port 2 ~500m DCF 20 GHz MLL Comb PPLN LCoS Filter Intensity Modulator Bit Pattern Generator 9 nm 9 nm 1 nm 50% 50% 50% 50% 0.4 nm 80km SSMF 1 nm 9 nm 2 3 taps 5 taps 5 taps for 20G 3 DSB copies 1542 1546 1550 1554 1558 10 dB/div 1 1542 1546 1550 1554 1558 10 dB/div Output 2 1542 1546 1550 1554 1558 10 dB/div 10 dB/div 1 1 (a) (b) 1542 1546 1550 1554 1558 5 DSB copies (c) Wavelength (nm) Wavelength (nm) Pre-Amp Pre-Amp MLL: mode locked laser LCoS: liquid crystal on silicon SSMF: standard single-mode fiber DCF: dispersion compensating fiber PPLN: periodically poled lithium niobate Wavelength (nm) Wavelength (nm) 54 4.4 Experimental Results Figure 4.3(a) shows the optical spectra (100MHz resolution) and eye-diagram comparison of a 10-Gbit/s OOK channel with DSB and 3-tap SSB. It can be seen that the left sideband is suppressed with a correct tap setting. If the taps are set incorrectly, severe ISI is introduced, which distorts the eye-diagram. Figure 4.3(b) shows the sideband suppression with a different number of comb lines (taps), which indicates that more comb lines provide a higher suppression ratio of the sideband. The inset presents the output spectrum of the PPLN with correct setting of the 5 taps. The B2B BER comparison is illustrated in Fig. 4.3(c). The implementation penalty (both input and output are DSB) is ~0.5dB by comparing the triangle symbols and cross symbols. By comparing the triangle symbols and diamond symbols, a ~1.5dB penalty is observed after converting DSB to SSB. Figure 4.3 (a) Comparison of eye-diagrams and optical spectra between 10-Gbit/s OOK DSB and SSB channels. (b) Optical spectra using a different number of taps (combs). (c) BER comparison between DSB and SSB for a back-to-back scenario. Figure 4.4 (a) Eye-diagram and BER comparison between 10-Gbit/s OOK DSB and SSB channels after an 80km SSMF transmission. Because of the enhanced tolerance to dispersion-induced power fading of SSB, a >3dB receiver sensitivity improvement is observed compared to the DSB channel. Frequency (GHz) 1542 1546 1550 1554 1558 10 dB/div Adjusted 5 SSB taps Output Received power (dBm) BER Power (dBm) Power (dBm) Frequency (GHz) 300 mv/div DSB 300 mv/div With incorrect SSB taps 300 mv/div With correct SSB taps (a) (b) (c) ~1.5 dB Wavelength (nm) 300 mv/div 300 mv/div DSB (after an 80km SSMF) SSB (after an 80km SSMF) Received power (dBm) BER DSB (20-Gbit/s OOK) SSB (20-Gbit/s OOK) DSB (20-Gbit/s PAM4) SSB (20-Gbit/s PAM4) Frequency (GHz) Frequency (GHz) 20dB/div 20dB/div 300 mv/div 300 mv/div 300 mv/div 300 mv/div (a) (b) (c) >3dB 55 (b) Optical spectra and eye-diagrams for a 20-Gbit/s OOK DSB and SSB channels. (c) Optical spectra and eye-diagrams for a 20-Gbit/s PAM4 DSB and SSB channels. Figure 4.4(a) shows the eye-diagrams for a 10-Gbit/s OOK DSB and SSB channels after an 80km SSMF transmission. Because of the dispersion-induced power fading, the eye-diagram of the DSB channel becomes worse than that of the SSB channel. By a BER comparison, there is a >3dB receiver sensitivity benefit for the SSB channel. Figure 4.4(b) shows the spectra of the SSB generation for a 20- Gbit/s OOK DSB channel, in which the left sideband is kept. There is small channel degradation from the eye-diagram comparison. The 20-Gbit/s SSB PAM4 generation is depicted in Fig. 4.4(c), which displays both the optical spectra and eye-diagrams for DSB and SSB. 56 Chapter 5 Performance Enhancement of an Optical High-Order QAM Channel by Adding Correlated Data to Robust Neighboring Channels 5.1 Introduction Future optical communication networks could have different data modulation formats with different OSNR requirements [65]. In such optical networks, there might be a desire to send a high-order QAM data channel into an available frequency slot. However, a potential problem could be that the current slot has a high-level ASE noise; therefore, the slot does not have enough OSNR to support the spectrum-efficient QAM data channel. Instead of sending a single data channel, one solution could be to co-transmit multiple data copies on different channels. At the receiver, the multiple copies are combined to enhance the channel performance using digital [66][67] or optical [68][69] signal processing. Another potential solution could be to encode the high-order QAM data information onto one or more existing robust neighboring channels so that: (i) the performance of the high-order QAM channel is improved after using joint signal processing for the target and neighboring channels; (ii) the performance of the robust neighboring channels is decreased but still could be sufficient for the system operation; and (iii) transmitting the high-order QAM channel does not consume other users’ spectra. In this chapter, we numerically simulate and experimentally demonstrate performance enhancement of an optical high-order QAM channel by adding correlated data to robust neighboring BPSK or QPSK channels [70]. Our approach allows transmitting the multiple QAM data copies on channels that are already utilized by other network users. The performance enhancement of the target QAM channel comes from borrowing the extra system margin from the robust neighboring channels. In this scheme, we send one or more optical BPSK or QPSK channels that already exist in the network, together with a CW laser to modulate the same electrical high-order QAM data. At the output, besides the target 57 optical QAM channel, the other BPSK or QPSK channels also become QAM channels, which are generated by multiplying the target electrical QAM data to the BPSK or QPSK channels. After joint detection and signal processing of the multiple correlated channels, the performance of the target QAM channel is improved at the cost of degrading the robust neighboring channels. In the simulation, a ~3-dB OSNR improvement is observed for a channel modulated with formats from 4QAM (QPSK) to 256QAM using two BPSK channels as correlated channels. This scheme is further experimentally demonstrated for (i) a 5-Gbaud QPSK or 16QAM channel with double 5-Gbaud BPSK channels as the correlated channels, and (ii) a 10-Gbaud QPSK or 16QAM channel with a single 10-Gbaud BPSK channel as the correlated channel. A ≥2 dB OSNR improvement is experimentally observed. 5.2 Concept The concept of performance enhancement for a high-order QAM channel using channel correlation is illustrated in Fig. 5.1. Suppose there are two optical BPSK channels (S1 and S2) and one available frequency slot in a system with a limited OSNR condition. In order to transmit a new high-order QAM signal (16QAM for example), one option is to send it only through the available frequency slot as channel C, which results in low signal quality as shown in Fig. 5.1(a). Another option is that besides sending the 16QAM signal through channel C, the existing robust BPSK channels (S1 and S2) are recruited to also carry the 16QAM signal by channel correlation. After joint detection and signal processing, the performance of the 16QAM signal is enhanced at the cost of degrading the robust BPSK channels, which is shown in Fig. 5.1(b). The details of the proposed scheme are illustrated in Fig. 5.1(c). Channel correlation is achieved by feeding the QAM-driven optical modulator with three optical sources, which are a CW, S1, and S2. Since the constellation of a typical QAM signal is quadrant symmetric [70], the resulting multiplication of the QAM signal with the BPSK channels (S1 and S2), having the same baud rate, basically adds a phase rotation without altering the QAM constellation layout, as depicted in Fig. 5.1(c). This also applies to S1 and S2 with 58 QPSK modulation. There are three channels at the output of the optical modulator: one is the target 16QAM channel (C), with a CW as its optical source; the other two output channels (S1’ and S2’) are the BPSK channels (S1 and S2) multiplied by channel C. Depending on the phase of Si (i=1 or 2), Si’ is either C or C rotated by a phase of π. As a result, C, S1’ and S2’ carry the same 16QAM information and become mutually correlated. Note, however, since channel C, S1’ and S2’ are at different wavelengths, the ASE noise at these three channel bandwidths generally is uncorrelated. Figure 5.1 Schematic diagrams of (a) directly inserting the high-order QAM channel (C) between S1 and S2, (b) enhancing the performance of the channel C using channel correlation, and (c) details for the channel enhancement approach. Channel correlation is achieved by modulating the same high- order QAM data both on a CW laser and on already data-carrying robust neighboring channels (S1 and S2). S1 and S2 can be BPSK or QPSK and this scheme could be extended to a different number of correlated channels. At the receiver, all channels are simultaneously detected and processed. At the receiver, the three channels (C, S1’, and S2’) are simultaneously detected and processed. There are three steps for the signal processing. In Step 1, the BPSK information is extracted by comparing the phase of Si’ to the phase of C [72]. Since the BPSK decoding relies on the phase comparison of two high-order QAM constellations, the bit-error-rate (BER) of BPSK is affected by both the QAM order Optical modulator S1’ (S1xC) C Joint detection and signal processing ASE noise S1 (Optical) Si’(t) C(t) Step 1: Phase comparison to extract Si Step 3: Coherent channel combination C C1 0 Si(t) Si(t) π CW laser Flow chart for joint signal processing Step 2: Erase Si from Si’ Si’ Conj. (Si) Ci BPSK S2 (Optical) BPSK S2’ (S2xC) If Si(t)= 0 If Si(t)= π then Si’(t)= then Si’(t)= Encoding 16QAM to Si (i=1, 2) i=1, 2 Δθ If Δθ<π/2, decode Si as else, decode Si as i=1, 2 i=1, 2 C2 Correlated Signal Uncorrelated Noise ASE noise S1 (BPSK) S2 (BPSK) High-order QAM data, e.g., 16QAM (Electrical) Optical transmitter S1 (BPSK) C (16QAM) S2 (BPSK) Detection and signal processing S1 C S2 ASE noise S1 (BPSK) S2 (BPSK) Channel correlation S1’ (16QAM) encoded by C C (16QAM) S2’ (16QAM) encoded by C Joint detection and signal processing S1 C S2 (a) (b) (c) C High-order QAM data, e.g., 16QAM (Electrical) High-order QAM data, e.g., 16QAM (Electrical) and C(t)= and C(t)= 59 and the QAM OSNR. The extracted BPSK information is denoted as Si ̅ . In Step 2, the BPSK information is erased from Si’ by multiplying it to the conjugation of Si ̅ , and the output is denoted as Ci (i=1 or 2). Because of the decoding error, Si ̅ is not precisely equal to Si and Ci is not an ideal copy of C. In Step 3, C, C1, and C2 are added constructively. Due to the uncorrelated nature of the ASE noise, the constructive channel combination leads to a noise reduction of the 16QAM channel. This scheme can be extended to a different number of correlated channels and, as mentioned above, could also use QPSK as the correlated channels. It is noted that the baud rate of correlated BPSK/QPSK channels should be equal to that of the target QAM channel. 5.3 Simulation results The proposed scheme is first demonstrated in a simulation of a 10-Gbaud system with a symbol length of 5×10 5 . In the beginning, two BPSK channels are used as the correlated channels, and the target QAM channel is varied from 4QAM (QPSK) to 256QAM. The three channels, modulated by independent laser sources with a 100 kHz linewidth, have the same OSNR and the power is 0 dBm. For the carrier recovery at the receiver, the 4QAM (QPSK) channel uses the Viterbi-Viterbi algorithm [73], and all other higher-order QAM channels use blind phase search (BPS) algorithm [74]. The block size for the carrier recovery is 29. In our approach, cycle slips (i.e., phase discontinuities after the phase unwrapping) could occur after independent carrier recoveries for each channel. In the joint signal processing, the cycle slips are transferred to the decoded BPSK channel (Si) through phase comparison. After erasing Si from Si’, the cycle slips between C, C1, and C2 could be mitigated. Fig. 5.2 shows the constellations of the target QAM channel with and without using channel correlation. As a result, channel correlation improves the target QAM channel of all tested orders. The performance enhancement of the target QAM channel is further demonstrated by a BER comparison in Fig. 5.3(a). There is a ~3-dB OSNR 60 improvement for the QAM channel with different orders. However, the channel improvement comes at the expense of the BPSK channels, whose performance deteriorates with increasing QAM order, as shown in Fig. 5.3(b). The reason is that, as the QAM constellation becomes denser, higher OSNR is required to extract the BPSK information by the phase comparison in Fig. 5.1(c). Figure 5.2 Simulation results for the constellation of the target 10-Gbaud channel modulated with 4QAM (QPSK), 16QAM, 64QAM, and 256QAM, with and without channel correlation. Two neighboring 10-Gbaud BPSK channels are used as the correlated channels. w. channel corr. w/o channel corr. QPSK 16QAM 64QAM 256QAM 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 0 2 4 6 8 10 12 14 16 18 20 22 24 QPSK (w/o corr) QPSK (w. corr) 16QAM (w/o corr) 16QAM (w. corr) 64QAM (w/o corr) 64QAM (w. corr) 256QAM (w/o corr) 256QAM (w. corr) (a) BER (Target Channel) QPSK 16QAM 64QAM 256QAM FEC 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 0 2 4 6 8 10 12 14 16 18 20 22 24 BPSK (w. 256QAM corr) BPSK (w. 64QAM corr) BPSK (w. 16QAM corr) BPSK (w. QPSK corr) BPSK (w/o corr) BER OSNR (dB) (b) ~3dB OSNR (dB) FEC 61 Figure 5.3 Simulation results for (a) the BER of the target channel modulated with 4QAM (QPSK), 16QAM, 64QAM, and 256QAM, with and without channel correlation; (b) the BER of the neighboring BPSK channels that assist the target channel with different QAM orders. The benefit of increasing the number of correlated channels is illustrated in Fig. 5.4(a). It can be seen that the more harnessed correlated channels, the higher the improvement to the 64QAM channel. Regarding the performance of the BPSK channels, since it only relies on the phase comparison between a single correlated channel and the target channel, Fig. 5.4(b) shows almost the same BER as the number of correlated channels increases. Figure 5.4 Simulation results for (a) the BER of a 64QAM channel with a different number of correlated channels; (b) the BER of a different number of neighboring BPSK channels used as the correlated channels. Fig. 5.5(a) shows the performance enhancement of a target 64QAM channel using QPSK as the correlated channels, including a comparison to the BPSK case. Although the QPSK channel correlation brings an OSNR benefit for the 64QAM channel, the amount of improvement is less than the BPSK channel correlation. A potential reason is that for the same OSNR, QPSK has a higher symbol error rate than BPSK. After erasing the QPSK information, it brings a larger penalty to Ci in Fig. 5.1(c). Fig. 5.5(b) displays the BER comparison of the BPSK and QPSK used as correlated channels. The degradation of correlated BPSK/QPSK channels depends on their modulation format, and QPSK will likely suffer more degradation than BPSK due to the denser constellation. In addition, if original BPSK/QPSK data of an existing channel has some degree of mutual information with the QAM data, then it 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 13 14 15 16 17 18 64QAM (w/o corr) 64QAM (w. 1BPSK corr) 64QAM (w. 2BPSK corr) 64QAM (w. 4BPSK corr) 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 13 14 15 16 17 18 BPSK (1 channel corr) BPSK (2 channels corr) BPSK (3 channels corr) OSNR (dB) BER (Target Channel) BER OSNR (dB) (a) (b) 4 62 might be possible to decrease the BPSK/QPSK symbol error when performing the channel decoding in the joint signal processing. Moreover, the decreased BPSK/QPSK symbol error may further improve the target QAM channel by producing more accurate Ci in Fig. 1(c). Figure 5.5 Simulation results for (a) the BER of a 64QAM channel using BPSK or QPSK correlated channels; (b) the BER of the BPSK or QPSK. It is noted that there is no fiber transmission in our simulation. If the transmission is considered, however, the dispersion-induced inter-channel walk-off should be compensated before the joint signal processing [44]. Moreover, when the BPSK/QPSK channels are converted to the correlated high-order QAM channels, they will likely become more sensitive to the channel degradation due to fiber nonlinear impairments [75]. In such a case, the OSNR enhancement for the target QAM channel might be limited. 5.5 Experimental Setup A proof-of-concept experimental setup is shown in Fig. 5.6(a). In the experiment, two BPSK channels, which are generated from separate laser sources, are used for the correlated channels. Then, an optical delay adjustment is employed to align the timing of the optical BPSK channels to modulate the target QPSK or 16QAM data channel. A third laser is used as a carrier for the target QPSK or 16QAM channel. At the output of the optical IQ modulator, there are three optical channels, one of which is the target channel and the other two of which are the target channel multiplied by the two BPSK channels (correlated channels). By tuning each 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 13 14 15 16 17 18 64QAM (w. BPSK corr) 64QAM (w. QPSK corr) 64QAM (w/o corr) OSNR (dB) BER (Target Channel) 1E-5 1E-4 1E-3 1E-2 1E-1 13 14 15 16 17 18 BPSK (w. 64QAM corr) QPSK (w. 64QAM corr) OSNR (dB) BER (a) (b) 63 laser’s power and polarization, all three channels are adjusted to have the same power and polarization. In the next, a wideband optical ASE source is employed to introduce the OSNR degradation, the level of which is controlled by an attenuator. Finally, all three channels are sent to a single coherent receiver to record the waveform for the offline signal processing. In the experiment, we use a coherent receiver with balanced detection, which could theoretically minimize the common signal-signal beat noise term [76]. However, there might be some residual signal- signal beat noise in our receiver, which could degrade the system performance. Figure 5.6 (a) A proof-of-concept experimental setup using two BPSKs as the correlated channels to recover a QPSK or 16QAM data channel. The wideband ASE source contributes statistically independent noise to all three channels; (b) offline signal processing in the experiment. Fig. 5.6(b) shows the offline signal processing. All three channels are first separated and downshifted to the baseband. Then, the IQ imbalance and delay compensations are implemented. Afterward, channel equalization is used to compensate for the system distortion [77][78]. Next, the frequency offset between the transmitter and receiver lasers is compensated by a fast Fourier transform based algorithm [79]. The BPS-based carrier recovery uses 32 test phases. After the pre- processing, the three channels are jointly processed in the three steps described in Fig. 1(c). In the scenario of three channels, the speed of each channel is 5 Gbaud and IQ Modulator Pre-Amp Joint Detection & Offline Processing BPSK data Phase Modulator Pre-Amp ASE Source QPSK or 16QAM data Delay Adjust ATT λ 1 λ 2 λ 3 Channel separation & frequency downshift IQ imbalance & delay compensation Channel equalization Frequency offset estimation Carrier recovery Extract and recover BPSK channel(s) Erase BPSK from correlated QAM channel(s) Coherent combination of multiple QAM channels Pre-processing (a) (b) Pre-Amp: pre-amplifier ATT: attenuator 64 the channel spacing is 10 GHz. We also investigate the scenario of two channels, each with 10 Gbaud separated by a 15 GHz channel spacing. In our experiment, there is no fiber transmission and all the channels are generated at the same node. In practical optical networks, channels could come from different nodes such as add/drop nodes. In that case, the effects of link loss, polarization rotation, and chromatic dispersion should be considered before performing the channel correlation. 5.6 Experimental Results Fig. 5.7 shows the experimental results for the three-channel scenario. The constellations with and without using channel correlation are displayed in Fig. 5.7(a). Fig. 5.7(b) and (c) show the BER comparison and indicate a >2-dB OSNR improvement for the target QAM channel. The deterioration of the BPSK channels is illustrated in Fig. 5.7(d), which confirms that the BPSK channels degrade with the increase of the QAM order. Finally, a BER comparison of the scenario for two channels with 10 Gbaud is shown in Fig. 5.8. Similar to the simulation results, the OSNR improvement decreases with fewer correlated channels. 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 4 5 6 7 8 9 10 QPSK Exp (w/o Corr) QPSK Exp(w. Corr) 1E-5 1E-4 1E-3 1E-2 1E-1 4 5 6 7 8 9 10 11 12 13 14 15 16 BPSK Exp(w/o Corr) BPSK Exp (w. QPSK Corr) BPSK Exp (w. 16QAM corr) OSNR (dB) BER (Target Channel) OSNR (dB) (b) (c) OSNR (dB) BER w. channel corr. w/o channel corr. (d) (a) FEC > 2dB 1E-5 1E-4 1E-3 1E-2 1E-1 6 8 10 12 14 16 16QAM Exp (w/o Corr) 16QAM Exp (w. Corr) > 3dB FEC C BER (Target Channel) 65 Figure 5.7 Experimental results for (a) a constellation comparison of a target 5-Gbaud QPSK or 16QAM data channel using double BPSK correlated channels; (b) a BER comparison of the target QPSK channel; (c) a BER comparison of the target 16QAM channel; (d) deterioration of the neighboring BPSK channels. Figure 5.8 Experimental results for the BER of a target 10-Gbaud (a) QPSK or (b) 16QAM channel, using a single BPSK correlated channel. 5.7 Discussion and Conclusion It is noted that in our three-channel experiment, we use a single 32-GHz bandwidth receiver to detect all three 5-Gbaud channels simultaneously. If the channel baud rate increases to 25 Gbaud for example, the required bandwidth will exceed 75 GHz, which is quite large and beyond the capability of our receiver. In such a case, the OSNR enhancement of the target QAM channel would be limited. One solution could be using three synchronized low-bandwidth receivers to detect three channels separately, and then perform the joint signal processing electronically [44]. Alternatively, one might be able to use the single-input-multiple-output (SIMO) approach to gain a ~3-dB improvement by replicating the data on another channel [80]. Our scheme encodes the same data on different channels, which could still carry their independent information. A ~3-dB system improvement would likely need the use of two correlated channels in our approach. 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 5 6 7 8 9 10 QPSK Exp (w/o Corr) QPSK Exp(w. Corr) 1E-5 1E-4 1E-3 1E-2 1E-1 8 9 10 11 12 13 14 15 16 16QAM Exp (w/o Corr) 16QAM Exp(w. Corr) (a) OSNR (dB) BER (Target Channel) OSNR (dB) (b) ~2dB ~ 3dB FEC FEC BER (Target Channel) 66 In a real system, the high-order QAM channel and correlated BPSK/QPSK channels could have different launch powers. When the correlated channels have higher power, the target QAM channel could borrow more system margin to improve the performance. However, the neighboring channels having lower power may not be able to provide an extra system margin, and might not be capable of enhancing the target channel performance. 67 Chapter 6 All Optical Signal Level Swapping and Multi-Level Amplitude Noise Mitigation Based on Optical Parametric Amplification 6.1 Introduction Optical signal processing has the potential advantage of fast signal operation which avoids optoelectronic conversion [1][81][82]. Typical signal operational functions could be achieved by using the physical properties of optical elements such as HNLF and PPLN waveguide. In the past, nonlinear processes have realized several functions including amplification [83][84], multicasting [85], switching [86][87] and regeneration [88][89]. In particular, nonlinear wave mixing has the merits of high speed and low additive noise. There are numerous research achievements on different types of nonlinear materials [90][91]. Optical parametric amplification (OPA) is a type of nonlinear signal processing [92][93]. Generally, in OPA, an input signal experiences linear gain by power transferred from the pump. However, through a mechanism known as pump depletion [94], further increase of input signal power would decrease the amplification gain, leading to saturation of the output power. This second gain region (saturation) has been utilized to squeeze amplitude noise for constant-amplitude modulation formats such as QPSK [95-97]. If the input signal power continues to increase, the gain drops further and the power is transferred back from the signal to the pump. This results in a third OPA gain inversion region. By using these three different gain regions (linear gain, saturation gain and gain inversion), specific optical signal processing functions could be achieved. In this chapter, we demonstrate all optical signal level swapping (inversion of amplitude levels) and amplitude noise mitigation for a 2-level-amplitude signal [98]. The existence of three OPA gain regions is verified by measuring the gain profile of 68 the OPA in an HNLF. Using only the gain inversion region, signal level swapping is realized for 2-amplitude-shift-keyed (2-ASK) and 8-QAM with both 10 Gbaud and 20 GBaud. Less than 1% EVM penalty is observed. In the following, all three gain regions are employed to realize 2-level amplitude noise mitigation for 10/20 Gbaud 8-QAM signals within two OPA stages. Furthermore, we experimentally investigate OPA gain profiles under different pump power levels. 6.2 Concept Figure 6.1 (a) Three gain regions (linear, saturation, inversion) of OPA. The gain inversion enables signal level swapping: the output amplitude levels are flipped compared with the input; (b) 2-level amplitude noise mitigation. Stage 1: amplitude squeezing for outer ring symbols based on linear and saturation regions; Stage 2: amplitude squeezing for inner ring symbols based on saturation and inversion regions. Figure 6.1(a) explains the three OPA gain regions. Taking a 2-ASK signal as an example, the constellation has two points (A and B) with different amplitude levels. In the linear region, A and B experience the linear OPA gain while the output constellation and relative waveform remain unchanged. In the saturation region, the input A and B have the same output power. Hence, two constellation points merge Out (mw) In (mw) Linear Gain Saturation Inversion (Swap) A B A B A B A(B) A B B A Out A In Out B A B A B A B A B B A B A In Before HNLF After HNLF Stage 1 Stage 2 Swap Input Output 69 into one and the output waveform has only one level. When the input power level of A and B is even higher, they reach to the inversion region. In this case, the output power is inversely related to the input power; therefore, the output constellation points of A and B are exchanged and the waveform is flipped. This signal level swapping can be effectively employed in multi-level amplitude noise mitigation. It is noted that conventional all optical amplitude noise mitigation, which is usually achieved by OPA saturation, is applicable for constant-amplitude modulation format such as QPSK signal. This is because amplitude squeezing using OPA gain saturation can be performed only on the highest amplitude level. However, with the help of signal level swapping, the constellation points at low amplitude level can be flipped to the high level and conventional amplitude squeezing can be utilized. Applying this idea to 2-level modulation format such as 8-QAM, all optical amplitude noise mitigation can be achieved. Figure 6.1(b) shows the concept of 2-level amplitude noise mitigation by using all three OPA gain regions. The input is an 8-QAM signal degraded by ASE noise. At Stage 1, by changing the signal power, the noisy symbols at the upper level reach to the OPA saturation region; while the noisy symbols at the lower level fall into the OPA linear region. In this case, the noise on the upper level symbols is squeezed while the noise on the lower level symbol stays the same as shown in Fig. 6.1(b). At Stage 2, by tuning the signal power, the upper level symbols reach the inversion gain region and the lower level symbols fall into the saturation region. At the output, not only the symbol levels are swapped but the noise on the lower level symbols is also squeezed. Therefore, by utilizing the three OPA gain regions, the amplitude of the noise on both 2 levels is squeezed. It is noted that since the amplitude levels of the final output are flipped compared with the original input signal, another swapping stage might be needed to recover the original constellation. 70 6.3 Experimental Setup of OPA Gain Regions Measurement Figure 6.2 illustrates the experimental setup for OPA gain regions measurement and how to realize all optical signal level swapping. In order to measure the gain profile, a CW signal at 1560 nm is coupled with a pump at 1556.3 nm. To suppress the stimulated Brillouin scattering (SBS) effect, both the CW and pump are phase modulated. For the gain profile measurement, the module of ASK modulation is bypassed and the CW signal is directly phase modulated by QPSK for SBS suppression. Then, the signal and pump are coupled in a 700m HNLF with nonlinear coefficient of 21.4 W -1 km -1 , ZDW of 1551.5nm and a dispersion slope of 0.043ps/km/nm 2 . The corresponding gain profile of the HNLF is characterized by varying the power of the input signal. Figure 6.2 Experimental setup to characterize OPA gain profile of the HNLF (without 2-ASK signal modulation) and realize signal level swapping (with 2-ASK signal modulation). 8-QAM modulation is achieved by cascading an intensity modulator with a QPSK modulator. Phase modulation is employed to suppress the SBS effect. 6.4 Experimental Results for OPA Gain Profile Figure 6.3(a) depicts the relationship between signal input power and signal output power after OPA. When the pump power is 1W, with increasing signal input power, the signal output power grows linearly with 9dB gain. As input power exceeds 12dBm, OPA gain decreases; but the output power still grows. Therefore, this gain profile cannot be utilized for signal level swapping. However, when the pump power is increased to be at least 1.5W as shown in Fig. 6.3(a), the signal Intensity Modulator QPSK Modulator Data ATT Coherent Receiver Phase Modulator 3Gbaud PRBS (2 15 -1) Data HNLF λ Sig 1560nm λ Pump 1556.3nm λ LO 1nm 1nm 1nm ASK Modulation (only for signal level swapping) QPSK Modulation PD 700m 71 output power first similarly experiences constant OPA gain; then grows slower and becomes saturated. Furthermore, when signal input power continues to increase, the output power begins to drop. The gain profiles corresponding to different pump powers are displayed in Fig. 6.3(b). Besides the sharp drop of the gain curves for pump power of 1.5W and 2W, the gain can also be negative as shown within the dashed circle, which is a unique property of the HNLF compared with EDFA. It is due to the power of the signal being transferred to the pump as well as to the other high order harmonics. Additionally, in the region shown between the two dashed lines in Fig. 6.3(a), amplitude squeezing and signal swapping can be realized simultaneously when the lower level input symbols reach to the flat region (gain saturation) and upper level input symbols reach to the descending region (gain inversion). Figure 6.3 (a) Measured input-output curves with different pump powers (1W, 1.5W and 2W). Both saturation and inversion OPA regions are observed with 1.5W and 2W pumps; (b) corresponding gain profiles. 2-amplitude-level modulation format such as 2-ASK and 8-QAM are employed to evaluate all optical signal level swapping. In the experimental setup shown in Fig. 2, the 8-QAM signal is emulated by modulating data with 2-ASK signal followed by a QPSK modulator. The signal power and pump power are set to be 0.15W and 1.5W respectively. At the receiver side, the signal is split in two paths. In one path, a PD is used to record the waveform; while in the other path, coherent detection is utilized to obtain the constellation and evaluate the overall equality of the signal. -5 0 5 10 15 20 25 30 -12 -8 -4 0 4 8 12 16 20 2W 1.5W 1W -5 0 5 10 15 20 -12 -8 -4 0 4 8 12 16 20 2W 1.5W 1W Input (dBm) Input (dBm) Output (dBm) Gain (dB) 72 Figure 6.4(a) shows the waveform of 10 Gbaud 8-QAM signal at the input and output. The inverted waveform indicates the successful signal level swapping. Similar waveform inversion has also been observed for a 20 Gbaud 8-QAM signal as shown in Fig. 6.4(b). Figure 6.4 Experimentally demonstration of signal level swapping for different baud rates: (a) waveform flipping for 10 Gbaud 8-QAM signal; (b) waveform flipping for 20 Gbaud 8-QAM signal. It is noted that both the input and output measurements are normalized. The EVM of the input signal and output signal are compared in Fig. 6.5. EVM increases by less than 1% at the output, which indicates that the signal quality is almost preserved after the swapping operation. It is noted that the EDFA power for the signal should be carefully tuned to ensure its two distinct power levels reach to the suitable OPA regions to have the optimal swapping performance. Time (ns) Time (ns) 10Gbaud 20Gbaud 0.25 1 Amplitude (a.u.) Amplitude (a.u.) 0.25 1 73 Figure 6.5 Constellation EVM comparison before and after all optical signal level swapping (a) 10 Gbaud 2-ASK/8-QAM; (b) 20 Gbaud 2-ASK/8-QAM. EVM degradation is less than 1%. 6.5 Experimental Setup for 2-level Amplitude Noise Mitigation Figure 6.6(a) is the experimental setup for 2-level amplitude noise mitigation of an 8- QAM signal using signal level swapping. The signal is degraded by an ASE source at the transmitter. A single pump laser is split and used for the following two OPA stages. Stage 1 is used for the amplitude noise squeezing of the high amplitude level symbols, where the 8-QAM signal with the power of 0.12W is coupled with a 2W pump in a 900m HNLF-1 with ZDW of 1556nm and nonlinear coefficient of 9.2 W -1 km -1 . The characteristics of HNLF-1 are measured with the same method described in Fig. 2. The input and output power profile is shown in Fig. 6.6(b) with the curve of square symbols. The saturation region occurs when the power level exceeds 16dBm. In Stage 2, the signal is coupled with the divided pump source of 1.5W in a 700m HNLF-2 with the same characteristic shown in Fig. 6.3. The input and output power profile is also depicted in Fig. 6.6(b) with triangle symbols. Here, the signal level swapping and amplitude squeezing are simultaneously achieved. Compared with the power profile of HNLF-1, for the same input power, the gain of HNLF-2 is nearly 6dB higher and both saturation and inversion regions are observed. This is due to the fact that HNLF-2 has a much higher nonlinear coefficient. 10Gbaud 20Gbaud EVM 10.0% EVM 10.1% EVM 8.7% EVM 9.1% EVM 11.3% EVM 11.5% EVM 9.3% EVM 9.9% (a) (b) 74 Figure 6.6 (a) Experimental setup for amplitude noise mitigation of 10/20 Gbaud 8-QAM signals. Stage 1 is for upper level amplitude noise mitigation and Stage 2 is for lower level amplitude noise mitigation; (b) input and output power profiles for HNLF-1 (square symbols) and HNLF-2 (triangle symbols). 6.6 Results for 2-Level Amplitude Noise Mitigation By adding ASE noise on the original signal, the constellation diagram is scattered, which is shown in Fig. 6.7(a, b). Due to gain saturation in Stage 1, the upper level symbols are squeezed. As a result, the amplitude error, defined as the root mean square (RMS) of the amplitude difference between the received and the ideal constellations, is deceased from 8.26% to 7.07% and 9.7% to 8.2% for 10 Gbaud and 20 Gbaud signals, respectively in Fig. 6.7(c). Stage 2 swaps the two amplitude levels and squeezes the original lower level symbols. It can be seen in Fig. 7(c) that the amplitude error is decreased further to 5.98% and 7.74% respectively. The total decrease of amplitude error is more than 20%. However, due to the high power of the system, the additional nonlinear phase noise leads to EVM degradation Intensity Modulator QPSK Modulator Data ATT ATT Coherent Receiver Phase Modulator 3Gbaud PRBS (2 15 -1) Data HNLF 1 HNLF 2 λ Sig 1560nm λ Pump 1556.3nm λ LO 1nm 1nm 1nm 1nm 1nm 1nm ASE Source 2nm 50% 50% 900m 700m Stage 1 Stage 2 2W 1.5W 0.12W 0.15W -10 0 10 20 30 -12 -8 -4 0 4 8 12 16 20 Stage 1 Stage 2 Input (dBm) Output (dBm) 75 shown in Fig. 6.7(c). For the 10 Gbaud signal, EVM degradation is 1% while for 20 Gbaud scenario, EVM degradation is 1.4%. Although the degraded EVM would increase bit error rate (BER), this system penalty might be compensated by other phase squeezing approaches [17, 18, 20, 21]. Figure 6.7 (a) 10 Gbaud and (b) 20 Gbaud 8-QAM constellation after each stage for 2-level amplitude noise mitigation. (c) Measured amplitude error and EVM comparison after each stage. 6.7 Discussion and Conclusion The current approach is based on 2-level-amplitude modulation formats. If more cascading stages are available, amplitude noise mitigation for higher-level modulations such as 16QAM might be achieved. On the other hand, if a 3-level staircase amplitude quantization similar to [88] could be potentially constructed, the amplitude squeezing of 16QAM might be simplified with one stage. The problem of high power consumption in this approach might be typically alleviated by using HNLFs with higher nonlinearity or longer length. Meanwhile, the additional nonlinear phase noise could be mitigated by several approaches [96-100]. Also, the extension of WDM application can be supported by the recent dispersion stable HNLF with ~50nm flat OPA gain region [101]. Although the crosstalk from FWM between different channels degrades system performance, it could be mitigated by other methods such as unequal channel spacing [102] and polarization interleaving [103]. Furthermore, if WDM is implemented, smaller input power on a 11.5 12 12.5 13 13.5 14 5 6 7 8 9 10 0 1 2 10G Amp Error 20G Amp Error 10G EVM 20G EVM 10Gbaud 20Gbaud Input Stage 1 Stage 2 Stage Num. EVM (%) Amplitude error (%) 76 single channel might be enough for OPA saturation and inversion, which is potentially helpful to reduce the power consumption per channel and the nonlinear phase noise. 77 Chapter 7 Self-Homodyne Detection with a Low- Power Pilot Tone Using Brillouin Amplification and Phase-Preserving Wavelength Conversion 7.1 Introduction Optical coherent detection is a critical component of optical communication systems [104][105]. Coherent detection notably requires a LO laser aligned to stably beat with the received signal especially for the advanced modulation format such as QPSK and QAM. One approach for achieving this stable beating is to use an optical phase locked loop (OPLL) to track and compensate for the phase difference between the transmitter laser and the LO laser [106][107]. The other typical approach applies digital signal processing for phase recovery, using either feedback [108-110] or feedforward [74][111-113] manners. Self-homodyne detection (SHD) is a different approach that relies on the transmitter to emit a pilot tone along with the data channel. At the receiver, the pilot is extracted as an LO laser to beat with the data channel, using either an optical hybrid or a digital hybrid schemes, which can potentially simplify the signal processing at the receiver and relax the laser linewidth requirement [114][115]. However, the optical power of the pilot tone needs to be sufficiently high (high pilot-to-signal ratio: PSR) to ensure necessary SHD performance [116], which usually requires a separate amplification by splitting the path before the receiver. In this case, the resulting loss of the phase locking between the pilot tone and data channel requires accurate feed-back compensation. One solution to this problem might be to transmit a low-power pilot tone, which will be selectively amplified by an in-line amplification at the receiver to increase the PSR for better SHD performance. The Brillouin amplification (BA) process is known for its low threshold pump power requirement and narrow-band amplification [117]. One study employed BA in an SHD system, which used an 8B-10B data encoding scheme for spectrum reshaping and inserted the pilot tone in the center frequency, even though BA is known for the high noise figure (NF) [118]. Potential 78 problems of this approach might be: (i) it may not be easy to extend the method to IQ modulation format signals, such as QPSK or QAM; (ii) the additional interfering Rayleigh back-scattering might fall into the data channel spectrum beyond 10GHz, which would degrade signal quality; (iii) the 8B-10B encoding introduces extra complexity and information redundancy. In this chapter, we propose an SHD scheme that places a low-power pilot tone adjacent to the data channel [17]. Both I and Q components of the standard data channel without prior encoding can be detected directly by the PD; meanwhile, the data channel spectrum is outside of Rayleigh back-scattering, which would not affect the quality of the data channel. 7.2 Concept The proposed scheme comprises two stages: (i) use stimulated Brillouin scattering to selectively amplify only the pilot tone, and (ii) relocate the amplified pilot tone to the center of the data channel spectrum using nonlinear wave mixing with phase-locked optical frequency combs. By tuning the relative phase among different comb lines, both I and Q components can be directly detected. With -30dB PSR, the proposed SHD scheme is experimentally demonstrated in a single channel B2B SHD system of 10/20G-baud BPSK or 10G-baud QPSK as well as the 100-km transmission of 10G-baud BPSK signals. We also experimentally demonstrate that if multiple pilot tones can be amplified simultaneously, a similar scheme could receive multiple independent channels. The concept of the proposed SHD system is shown in Fig. 7.1, in which the input includes a data channel (S) and a pilot tone (P) with a PSR of -30dB. In Stage-1, a portion of the pilot tone is extracted to be upshifted by Δλ in order to serve as the Brillouin pump. After amplification by an EDFA, the Brillouin pump propagates in a SMF in the opposite direction to the input channel. Due to the narrow-band nature of BA, only the pilot tone is amplified, with a gain coefficient of g. In Stage-2, the data channel (S) and the amplified pilot tone (gP) are coupled with a pair of coherent comb lines in a genetic nonlinear element which 79 could be an HNLF or a PPLN. By choosing the appropriate wavelength difference (the wavelength difference between S and P) and relative phase offset between the two comb lines, the pilot is shifted to the center of the data channel, and both I and Q components can be recovered after the nonlinear elements using direct detection at the PD. Figure 7.1 Conceptual diagram of the proposed self-homodyne detection (SHD) system with a low- power pilot tone. The scheme is composed of two stages: (i) Brillouin amplification for only the pilot tone, and (ii) phase preserving wavelength conversion using a pair of optical frequency comb lines. HNLF: highly nonlinear fiber; PPLN: periodically poled lithium niobate; SMF: single-mode fiber. 7.3 Experimental Setup The experimental setup of the proposed SHD scheme is shown in Fig. 7.2(a). At the transmitter, the data is generated using a high-speed arbitrary waveform generator (AWG), while the I component is added with a low-power sinusoidal pilot tone which is generated by a 40GHz clock synthesizer. The clock in the AWG is frequency locked with the pilot tone clock by a 10MHz reference. Then, the data signal modulates a laser at the wavelength of 1550.7nm in an IQ modulator. The spectrum of the signal with pilot tone is shown in Fig. 7.1 (b1), where a PSR of - 30dB is observed. The signal with the pilot tone is sent to either (i) the proposed SHD structure directly for B2B evaluation, or (ii) a 100-km link compromising an 80-km SMF and a 20-km DCF, as shown in Fig. 7.1(a). In the detection system, the incoming signal is divided into two paths. In the upper path, the left pilot tone is extracted by a narrowband optical filter and sent to a slave laser which is frequency-locked to the input pilot tone. The temperature and the λ 2 λ 1 Injection Locking λ 1 λ 1 -Δλ λ 2 λ 1 SMF Two Comb Lines λ 4 λ 3 λ 2 λ 1 λ 4 λ 3 λ 2 Brillouin Amplification for Pilot Tone Phase-Preserving Wavelength Conversion Phase Filter λ 2 λ 1 λ 4 λ 3 λ 2 Phase Filter S P S gP ~|S+gP| 2 ~|S+jgP| 2 λ 2 λ 1 S P -30dB Δθ=0 Δθ=π/2 Nonlinear Element (HNLF or PPLN) λ 1 Wavelength Shifter Nonlinear Element (HNLF or PPLN) Brillouin Pump 80 current of the slave laser need to be tuned to establish this locking. The insets in Fig. 7.2(a) illustrate the scenarios in which the slave laser is frequency locked or unlocked to the pilot tone. The output is then modulated by a 10.810GHz sinusoidal tone in an intensity modulator biased at the null point. The generated lower sideband (higher frequency) CW is selected and amplified to 150mW to act as the BA pump, as shown in Fig. 7.2(a). Afterwards, the BA pump propagates in the opposite direction to the incoming signal in a 500m SMF, in which only the pilot tone is amplified by ~40dB by the narrowband Brillouin interaction, as shown in Fig. 7.2(b2). The signal with pilot tone is then amplified to 186mW in an EDFA and sent into a narrowband optical filter, where the Rayleigh back-scattering component is suppressed. In the next stage, the signal with the pilot tone is mixed with a pair of comb lines in a 460m HNLF. The pair of comb lines is generated by a MLL-based optical frequency comb with a free spectral range (FSR) of 20GHz, as shown in Fig. 7.2(b3). After a SLM filter, two comb lines with a 40GHz frequency difference are selected with appropriate phase adjustment, as shown in Fig. 7.2(b4). Then, the two comb lines are amplified to 300mW before sending them to the HNLF. The complete spectra before and after HNLF are shown in Fig. 7.2(b5) and Fig. 7.2(b6) respectively, where the amplified pilot tone is shifted to the middle of the data spectrum. Finally, the data channel is selected in a filter and sent to a high-speed PD for real-time eye-diagram capture and BER measurement. Because the available experimental setup had access to only a single HNLF and SLM filter (in contrast to the concept in Fig. 7.1), the two output paths (I and Q) of Fig. 1 had to be separately measured, each following proper adjustment of the relative phase between the comb lines in the SLM filter. For the BER calculation of IQ modulation format such as QPSK, the reported BER is the average of separate I and Q measurements. This scheme can be extended to simultaneous I/Q measurement, given additional equipment. 81 Figure 7.2 (a) Experimental setup for the proposed SHD system and the spectra when the slave laser is frequency locked or unlocked to the incoming pilot tone; (b) corresponding spectra measured at each node. λ~1550.7nm 10G/20G PRBS 2 15 -1 IQ Modulator ~ 10MHz Reference I Q 40GHz Slave Laser Intensity Modulator SMF~500m ~ 10.810GHz 20G Comb Source SLM Filter HNLF~450m 1 6 Brillouin Amplification for Pilot Tone Phase-Preserving Wavelength Conversion 2 3 4 5 Transmitter Detector - Δλ + Δλ - Δλ Left Pilot Tone SMF ~80km DCF ~20km 100km Link ATT Wavelength (nm) 1550.96 1550.97 1550.98 1550.99 Wavelength (nm) 10dB/D 10dB/D Slave laser locked Slave laser not locked (a) 1550.96 1550.97 1550.98 1550.99 AWG 1549.5 1552 1554.5 Wavelength (nm) 1547 1551 1555 1550 1550.5 1551 1551.5 Wavelength (nm) 1 Wavelength (nm) 1550 1550.5 1551 1551.5 2 3 1552.5 1553 1553.5 1554 Wavelength (nm) 4 1550 1552 1554 Wavelength (nm) 5 Wavelength (nm) 6 S P P S gP comb lines P 2 comb lines S gP comb S+ηgP Rayleigh 10dB/D 10dB/D 10dB/D 10dB/D 10dB/D 10dB/D (b) 82 7.4 Experimental Results The narrowband characteristic of the BA is illustrated in Fig. 7.3(a). The frequency offset denotes the frequency drift away from the optimal Brillouin frequency shift in the SMF of the experiment, which is 10.810GHz. The BA gain drops to zero when the frequency offset is ~45MHz. Figure 7.3(b) shows the relationship between Brillouin pump power and BA gain, where a 150mw pump produces a 40dB gain on the pilot tone with a power of -34dBm at the input of the 500m SMF. Figure 7.3 (a) Bandwidth of Brillion amplification; (b) Brillouin amplification gain for the pilot tone versus pump power. The system performance is first evaluated using 10G-baud BPSK and QPSK data channels, each with a pilot tone of -30 dB PSR. Figure 7.4(a) shows that the eye-diagram opens and closes with the relative phase between the two comb lines. Thus, the relative phase between the comb lines needs to be tuned for optimal performance. Figure 7.4(b) presents the eye-diagrams for different modulation formats and scenarios. As a baseline for comparison, a 10G-baud BPSK data signal with a residual carrier in the center is sent directly to the PD. Compared to the baseline system, the proposed SHD has a relatively higher noise level which is attributed to the extra spontaneous Brillouin noise during pilot tone boosting by BA. After a 100km transmission, the signal quality degrades a little further, which might be attributed to residual chromatic dispersion after the DCF compensation. By adjusting the relative phase between the two comb lines in Fig. 7.2(b5), I and Q components of the QPSK signal are obtained, as shown in Fig. 7.4(b). Pump Power (mW) Gain (dB) 5 15 25 35 45 25 75 125 175 225 Frequency Offset (MHz) Gain (dB) 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 (a) (b) 83 Figure 7.4 (a) The detected eye diagram varies with the relative phase among two comb lines for BPSK SHD; (b) detected eye diagrams for a 10G-baud system with different modulation formats. For quantitative evaluation of the proposed SHD scheme, a 10G-baud channel BER is measured in real-time, as shown in Fig. 7.5(a). Compared to the ideal case, the BPSK without transmission has a similar BER curve. After the 100km link, a power penalty of less than 1dB is observed with a BER of 1e-3. For a QPSK signal, a power penalty of 4dB is observed with a BER of 1e-3. A BER comparison with a 20G-baud BPSK data signal is shown in Fig. 7.5(b) to demonstrate the tunability of the scheme. Compared to the 20G-baud BPSK baseline, a power penalty of approximately 0.5dB is obtained at the BER of 1e-3. Figure 7.5 BER versus received power for (a) 10G-baud BPSK and QPSK and (b) 20G-baud BPSK systems. B2B system performance is also investigated for different PSRs, while the received power is kept at -8.4dBm. Figure 7.6(a) shows that as PSR decreases, the power of the Brillouin pump needs to be increased to maintain the same output 0 deg 15 deg 30 deg 60 deg 90 deg BPSK w. residual carrier BPSK SHD BPSK (100km) SHD QPSK (I) SHD QPSK (Q) SHD (a) (b) BPSK (w. residual carrier) BPSK BPSK (w. residual carrier) BPSK QPSK BPSK (100km) Received Power (dBm) -11 -9 -7 -5 -3 -1 1 2 3 4 5 6 7 8 1 -Log 10 (BER) Received Power (dBm) 2 3 4 5 6 7 1 -Log 10 (BER) -11 -9 -7 -5 -3 -1 8 FEC FEC (a) (b) 84 power of the amplified pilot tone. Therefore, output PSR remains the same, which produces a relatively constant BER. The proposed scheme can even work with a PSR as low as -40dB. When the PSR further decreases to -45dB, the slave laser fails to lock to the faint input pilot tone, and the BER dramatically increases. The proposed scheme is also verified with Nyquist pulse shaping of 0.01 rolling off factor, as shown in Fig. 7.6(b). Figure 7.6 (a) BER measurements for different levels of PSR and the corresponding Brillouin pump power with -8.4dBm received power; (b) eye diagram of 10-Gbaud Nyquist BPSK signal. 7.5 Discussion and Conclusion This scheme could potentially be extended to WDM system with low PSR, which requires multiple BA pumps to be generated accordingly [119]. It is noted that the pilot tone is separated from the data channel by a 40GHz gap. For WDM applications, this placement may affect the spectrum efficiency. Moving the pilot tone closer to the data spectrum would alleviate this problem. However, the resultant interference from the data spectrum could degrade the quality of the pilot tone. Therefore, optimizing the location of the pilot tone requires further investigation. In this chapter, BPSK/QPSK are employed only to demonstrate the concept of the proposed SHD system, which is based on high-speed optical nonlinear wave mixing. Compared to conventional differential detection [120] and digital phase tracking [121] for BPSK/QPSK, the SHD complexity might be higher. However, SHD is expected to relax encoder and decoder design restrictions in deferential detection schemes for MPSK signals [122]. Also, SHD could be important for high order QAM with increased data rate, in which 120 140 160 180 200 220 240 1.0E-3 6.0E-3 1.1E-2 1.6E-2 2.1E-2 -45 -40 -35 -30 -25 -20 PSR (dB) BER Brillouin Pump (mW) Nyquist BPSK (a) (b) 85 high-speed digital processing for phase tracking might be a challenge. In addition, the ability of this approach to support multiple-channel SHD realization could be of interest in the future. 86 Chapter 8 Raman-Assisted Phase Sensitive Amplifier using Fiber Bragg Grating Based Tunable Phase Shifter 8.1 Introduction Next generation of enhanced-capacity optical communication systems employing higher-order QAM have received much attention [123]. These advanced modulation formats can typically benefit from the use of low-noise optical amplifiers to support long-haul transmission. Indeed, the phase sensitive amplifier (PSA) is a potential candidate because it can achieve a lower noise figure than conventional phase insensitive amplifiers [124-127]. PSA relies on a FWM-based parametric nonlinear interaction between a signal, a pump and a so-called idler (copier). By tuning the relative phase between these three components, the correlated signal and idler will undergo constructive coherent addition; concurrently, the uncorrelated noises on the signal and idler will undergo incoherent addition, which results in low-noise amplification. To date, this copier-PSA scheme has been experimentally demonstrated in a variety of different configurations [128-131]. The copier-PSA system typically requires the signal and idler are of similar amplitude, and the phase is matched between the signal, the idler, and the pump. The amplitude and phase adjustment is often achieved by a LCoS-based programmable filter [128-131], which normally has an insertion loss of ~5 dB [132][133]. This high insertion loss attenuates the power of the signal, the idler and the pump, which decreases the signal net gain (counting the idler generation). Therefore, it might be valuable to have a low-loss copier-PSA system that would provide a high signal net gain. In this chapter, we demonstrate a low-loss Raman-assisted PSA using a fiber Bragg grating (FBG) as a tunable phase shifter to provide ~20 dB signal net gain [134]. Instead of using an LCoS programmable filter, the amplitude and phase adjustment is realized by non- uniform distributed Raman amplification and a tunable low-loss FBG (0.4 dB insertion loss), 87 respectively. The total component loss of the proposed copier-PSA system is measured to be ~8 dB. Compared to a Raman-assisted PSA using a programmable filter with 6.5 dB insertion loss (total component loss is ~14 dB), the proposed approach improves the signal gain by >20 dB because of a reduced link loss for the signal and the PSA pump. Moreover, compared to a system with the FBG of a fixed central wavelength, by tuning the FBG central wavelength, (1) an up-to-5.6 dB signal gain improvement is obtained; and (2) a ~4 dB receiver sensitivity enhancement is demonstrated for 20 and 25 Gbaud QPSK signals as well as a 10 Gbaud 16-QAM signal. 8.2 Concept The concept of the proposed Raman-assisted PSA using FBG-based tunable phase shifter is depicted in Fig. 8.1. This system comprises four stages: (i) idler generation, (ii) phase adjustment, (iii) hybrid Raman/PSA, and (4) pure PSA. In the stage of idler generation, the PSA pump power is set comparatively low to avoid a significant phase insensitive amplification for the signal itself. Then, a phase adjustment is realized by (i) placing the PSA pump wavelength close to FBG central wavelength; and (ii) tuning the FBG central wavelength by varying the FBG temperature, which adds temperature-dependent phase-shift to the pump [134]. Regarding the FBG with a narrow bandwidth (<1 nm), tuning the FBG central wavelength should not affect the signal (S) and idler (I) phase. In the stage of hybrid Raman/PSA, the power imbalance between the signal and idler is compensated by the higher boosting on the idler than the signal, which is achieved by placing the signal wavelength away from the effective Raman gain region. Meanwhile, the PSA pump is also boosted in-line by Raman amplification to enhance the signal gain [136]. Finally, another stage of pure PSA is cascaded for a further signal amplification. PSA Pump S I HNLF PSA Pump S I Phase Adjustment of PSA Pump PSA Pump (θp) S I Raman Pump FBG Thermoelectric Heater Current Source Idler Generation Hybrid Raman/PSA PSA PSA Pump S HNLF HNLF: highly nonlinear fiber FBG: fiber Bragg grating Input S PSA Pump PSA Pump (θp+ΔθFBG) S I HNLF S Output 88 Figure 8.1 A schematic diagram of a Raman-assisted PSA using a FBG-based tunable phase shifter. The system includes four stages: (i) idler generation, (ii) phase adjustment, (iii) hybrid Raman/PSA, and (iv) pure PSA. Phase adjustment of the PSA pump is achieved by tuning the central wavelength of the FBG using a thermoelectric heater. Raman amplification is used to boost the power of the PSA pump and compensate for the power imbalance between the signal and idler, by placing the signal away from the effective Raman gain region. 8.3 Experimental Setup The experimental setup is shown in Fig. 8.2(a). At the transmitter, a laser is modulated by an optical I/Q modulator driven by an RF signal (QPSK or 16-QAM). The signal input power is adjusted by an attenuator (ATT-1). Meanwhile, a PSA pump at 1566.8 nm is amplified and combined with the signal through a 50/50 coupler. The PSA pump is phase modulated by an 800 MHz pseudo-random binary sequence (PRBS) to suppress stimulated Brillouin scattering (SBS). An optical circulator is used before HNLF-1 to monitor the potential pump power reflection by the FBG during tuning the central wavelength. An idler is generated with ~−10 dB conversion efficiency in the first stage, in which the signal and the PSA pump (21 dBm) are mixed in a 200 m highly nonlinear fiber (HNLF-1). The nonlinear coefficient, zero dispersion wavelength, and dispersion slope of HNLF-1 are 21.4/W/km, 1551.5 nm, and 0.043 ps/km/nm2, respectively (The other two HNLFs in the following stages have similar parameters). The input and output spectra of the first stage are shown in Fig. 8.2(d1–d2), in which the signal is barely amplified. In the next stage, the PSA pump, the signal and the idler are sent through a 15- mm-long FBG with a 0.4 dB insertion loss and maximum reflectivity >99%. The FBG is placed on a thermoelectric heater, the surface temperature of which is controlled by a current source. By changing the temperature, the FBG central wavelength moves, adding a phase-shift to the PSA pump. For example, a 0.8 A current increases the FBG temperature by 60 o C above room temperature (25 o C) and moves the FBG wavelength by 0.66 nm. 89 In the third stage, the signal, the idler and the PSA pump are amplified in a 500 m HNLF (HNLF-2) with backward Raman amplification provided by a ~32 dBm Raman pump at 1455 nm. The relationship between Raman pump power and control voltage is shown in Fig. 8.2(b). The 1460 nm isolator is used to block the output Raman pump; while the 1550 nm isolator is employed to block the potential unwanted SBS. By choosing a signal wavelength longer than 1568 nm (at the edge of the Raman gain region), the signal obtains much less gain from Raman amplification than the idler, as shown in Fig. 8.2(c). The gain difference between the signal and idler helps to compensate for the power imbalance coming from the first stage (idler generation). Meanwhile, the PSA pump is also boosted by Raman amplification. The spectrum after the hybrid Raman/PSA stage is shown in Fig. 8.2(d3). The final stage is a pure PSA process, in which the signal, the pump and the idler are sent to a 300 m HNLF (HNLF-3), where the signal receives additional gain, as shown in Fig. 8.2(d4). After HNLF-3, the signal is selected and sent through another attenuator (ATT-2) to adjust the input power for the pre-amplifier before the coherent receiver. ATT-2 is used to fix the input power to the pre-amplifier after each system adjustment/tuning, such as changing Raman pump power, or varying the input signal power. HNLF-2 ~500m QPSK/16-QAM Data I/Q Modulator I Q HNLF-1 ~200m ATT-1 1 1561 1563 1565 1567 1569 1571 Wavelength (nm) Phase Modulator 800MHz PRBS 2 31 -1 Raman Pump (1455nm) Coherent Receiver HNLF-3 ~300m 20dB/Div 3 4 5 2 3 4 5 (a) EDFA 24dBm LO PSA Pump λp: 1566.8nm Signal Laser 50/50 ATT-2 FBG Thermoelectric Heater 1550nm 1460nm Pre-Amp 4 6 8 10 12 14 16 1556 1559 1562 1565 1568 1571 1574 Raman Gain (dB) Signal Wavelength 20dB/Div 20dB/Div 20dB/Div 1561 1563 1565 1567 1569 1571 Wavelength (nm) 1561 1563 1565 1567 1569 1571 Wavelength (nm) 1561 1563 1565 1567 1569 1571 Wavelength (nm) (d) Current Source PRBS: pseudorandom binary sequence HNLF: highly nonlinear fiber ATT: attenuator FBG: fiber Bragg grating LO: local oscillator Signal region Idler region Voltage Control 28 29 30 31 32 33 10 10.5 11 11.5 12 12.5 Voltage (V) Raman Pump (dBm) (c) (b) Raman-Assisted PSA 2 1 2 3 1 2 3 90 Figure 8.2 (a) Experimental setup for a Raman-assisted PSA using a FBG-based tunable phase shifter. The link loss from Node-1 to Node-5 (including the 50/50 coupler, idler generation, phase adjustment, hybrid Raman/PSA, and pure PSA) is ~8 dB. The signal net gain is measured by comparing the signal power between Node-1 and Node-5; (b) Raman pump power vs. control voltage; (c) Raman gain profile, where the idler has more gain than the signal. This helps to decrease the power imbalance between the signal and the idler after idler generation; (d) Spectra at different positions. 8.4 Experimental Results The relationship between current supply, temperature, and the FBG central wavelength is illustrated in Fig. 8.3(a). With increasing the injected current, the FBG temperature climbs and its central wavelength is red-shifted. Fig. 8.3(b) shows that the FBG central wavelength is tuned by changing the temperature and a 60 o C temperature increase moves the FBG central wavelength by 0.66 nm. Fig. 8.3(c) characterizes the phase-shift induced by the thermal tuning of the FBG. A temperature change (ΔT) of 30 o C provides a phase-shift of ~15 degree while a ΔT of 60 o C introduces a phase-shift of ~60 degree. The PSA pump power after the FBG with different central wavelengths is shown is Fig. 8.3(d). The flat curve indicates the PSA pump power is not significantly decreased by reflection within the wavelength tuning range of 0.66 nm. Current (A) Temperature ( o C ) Central Wavelength (nm) 1565.7 1565.8 1565.9 1566 1566.1 1566.2 1566.3 1566.4 0 10 20 30 40 50 60 70 80 90 100 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 FBG Temperature FBG Central Wavelength Wavelength (nm) Transmittance (dB) 0.66nm 1 1.2 1.4 1.6 1.8 2 1565.7 1565.9 1566.1 1566.3 (a) (b) (d) Wavelength (nm) 1% Pump Power after FBG (dBm) Phase Shift (deg) (c) ΔT ( o C) 0 20 40 60 0 10 20 30 40 50 60 70 91 Figure 8.3 (a) The current of the thermoelectric heater vs. temperature and FBG central wavelength; (b) FBG transmittance under different temperature levels, in which a 60 o C temperature increase shifts the central wavelength by 0.66 nm; (c) Temperature change (ΔT) vs. phase-shift of FBG; (d) 1% PSA pump power after the FBG. The flat curve indicates the PSA pump power is maintained across different FBG central wavelengths. Tuning the FBG central wavelength could affect the phase matching condition of the PSA, which is represented by the output power variation of the signal, as shown in Fig. 8.4(a). It can be seen that a 10 o C temperature change, corresponding to a ~5 degree phase-shift for the PSA pump, causes no more than a 2-dB power variation for different input signal wavelengths. However, a 60 o C temperature change, corresponding to a ~60 degree phase-shift for the PSA pump, produces a >6 dB power variation. For a range of wavelengths, Fig. 8.4(b) shows the gain improvement that can be achieved by phase tuning within a ~60 degree range via the FBG tuning. Due to the system dispersion, the optimal phase for each wavelength is different and the signal might already have optimized phase condition for some wavelengths at room temperature. For example, no gain improvement is observed at 1568.6 nm. However, the gain improvement could be 5.6 dB when the signal wavelength is 1569.8 nm. Figure 8.4 (a) The variation of output signal power vs. signal wavelength with different FBG temperature changes; (b) Gain improvement by tuning the FBG central wavelength for the signal at different wavelengths. -6 -4 -2 0 2 4 6 8 10 12 1568 1570 1572 1574 Signal Power Variation (dB) Signal Wavelength (nm) ΔT=10 o C ΔT=60 o C 0 1 2 3 4 5 6 1568 1569 1570 1571 1572 1573 1574 Gain Improvement (dB) Signal Wavelength (nm) (a) (b) 92 The wavelength-dependent system optimization is further illustrated by comparing the constellations of a 20 Gbaud QPSK signal in Fig. 8.5. Since the signal gain varies under different system configurations, the signal power is adjusted by ATT-2 to ensure the same input power to the pre-amplifier. Fig. 8.5(a) shows that for a signal wavelength of 1569.8 nm, the EVM is improved by ~6% through tuning the FBG central wavelength. Fig. 5(b) shows that FBG temperature tuning cannot bring extra benefit at a wavelength of 1568.6 nm, because the FBG under the room temperature already provides the optimal phase. Figure 8.5 Comparison of signal constellations between without (w/o) and with (w.) FBG central wavelength tuning. (a) Signal with a wavelength of 1569.8 nm benefits from FBG central wavelength tuning; (b) Signal with a wavelength of 1568.6 nm shows no improvement from FBG central wavelength tuning. Figure 8.6 shows the signal net gain (the signal power difference between Node-1 and Node-5 in Fig. 8.2(a)) by changing the Raman pump power under different scenarios. The signal wavelength is fixed at 1569.8 nm for the remaining part. It can be seen that the system with the FBG central wavelength tuning always provides highest gain under different power levels of the Raman pump. The signal net gain using only Raman amplification is also included, which is <5 dB even the Raman pump power is at its maximum. In addition, the signal net gain is negative if we instead use a programmable filter (waveshaper) with a 6.5 dB insertion loss. This high insertion loss attenuates both the signal and the PSA pump, leading to a significant decrease of the PSA efficiency. w/o FBG Central WL Tuning w. FBG Central WL Tuning EVM=19.69% EVM=13.21% WL=1569.8 nm WL=1568.6 nm EVM=13.94% EVM=13.90% (a) (b) w/o FBG Central WL Tuning w. FBG Central WL Tuning 93 Figure 8.6 Comparison of signal net gain (the signal power difference between Node-1 and Node-5 in Fig. 2(a)) under different scenarios: (1) with FBG tuning; (2) without FBG tuning; (3) with Raman amplification only (PSA pump is off); and (4) with waveshaper. The comparison of bit error rate (BER) performance with and without FBG central wavelength tuning is shown in Fig. 8.7, including both 20 and 25 Gbaud QPSK signals. It can be seen that FBG central wavelength tuning achieves a similar sensitivity benefit of ~4 dB across the different signal baud rates. Figure 8.7 Comparison of measured BER vs. input power for 20 and 25 Gbaud QPSK signals under two scenarios: (1) without FBG central wavelength tuning; (2) with FBG central wavelength tuning. Finally, a 10 Gbaud 16-QAM signal is also demonstrated in the system, as shown in Fig. 8.8. The benefit of tuning the FBG central wavelength is verified by measuring both EVM and BER. Fig. 8.8(a) shows that the EVM of the 16-QAM signal is improved by ~2% through tuning FBG central wavelength. This EVM improvement leads to a ~4 dB receiver sensitivity enhancement in Fig. 8.8(b). -10 -5 0 5 10 15 20 25 28 29 30 31 32 33 Signal Net Gain (dB) Raman Pump Power (dBm) w.FBG Central WL Tuning w/o FBG Central WL Tuning Raman Only w. Waveshaper 1E-5 1E-4 1E-3 1E-2 1E-1 -42 -40 -38 -36 -34 -32 BER Signal Input Power (dBm) 20G w.FBG Central WL Tuning 20G w/o FBG Central WL Tuning 1E-5 1E-4 1E-3 1E-2 1E-1 -41 -39 -37 -35 -33 -31 BER Signal Input Power (dBm) 25G w.FBG Central WL Tuning 25G w/o FBG Central WL Tuning ~4dB ~4dB (a) (b) -39 -37 -35 -33 -31 -29 -38 -36 -34 -32 -30 -28 94 Figure 8.8 Comparison of (a) constellation and (b) BER of a 10 Gbaud 16-QAM signal with and without FBG central wavelength tuning. 8.5 Discussion and Conclusion In conclusion, a Raman-assisted PSA with FBG for phase tuning is experimentally demonstrated. The system benefits from the low-loss of the FBG and improves the signal net gain by >20 dB as compared to a system with a programmable filter (waveshaper). A ~4 dB receiver sensitivity enhancement is observed by comparing the systems with and without tuning the FBG central wavelength. It is noted that the current experimental configuration shows a wavelength-dependent signal performance in Fig. 4 and Fig. 5. This phenomenon might be caused by the system dispersion from the discrepancy between zero dispersion wavelength of the HNLFs (~1551 nm) and the PSA pump wavelength (~1567 nm), whose location is determined by the characteristic of the FBG. In future, HNLFs with zero dispersion wavelength close to 1567 nm might enable the extension of the scheme to multi-channel PSA operation. w/o FBG Central WL Tuning w. FBG Central WL Tuning 1E-4 1E-3 1E-2 1E-1 -38 -37 -36 -35 -34 -33 -32 -31 -30 -29 BER Signal Input Power (dBm) 10G 16QAM w.FBG Central WL Tuning 10G 16QAM w/o FBG Central WL Tuning (a) (b) ~4dB EVM=10.8% EVM=8.6% -35 -34 -33 -32 -31 -30 -29 -28 -27 -26 95 References [1] A. E. Willner, S. Khaleghi, M. Chitgarha, and O. Yilmaz, "All-Optical Signal Processing," Invited Paper, IEEE/OSA Journal of Lightwave Technology, vol. 32, no. 4, pp. 660-680, 2014. [2] S. Wabnitz and B. J. Eggleton, All-Optical Signal Processing: Data Communication and Storage Applications, vol. 194, Springer, 2015. [3] A. E. Willner, A. Fallahpour, F. Alishahi, Y. Cao, A. Mohajerin-Ariaei, A. Almaiman, P. Liao, K. Zou, A. N. Willner, M. Tur, "All-Optical Signal Processing Techniques for Flexible Networks," in Journal of Lightwave Technology, vol. 37, no. 1, pp. 21-35, 2019. [4] C. Langrock, S. Kumar, J. E. McGeehan, A. E. Willner, and M. M. Fejer, “All-optical signal processing using χ2 nonlinearities in guided-wave devices,” Journal of Lightwave Technology, vol. 24, no. 7, pp. 2579–2592, Jul. 2006. [5] M. Marhic, Fiber Optical Parametric Amplifiers, Oscillators and Related Devices. New York, NY, USA: Cambridge Univ. Press, 2007. [6] V. R. Supradeepa, C. M. Long, R. Wu, F. Ferdous, E. Hamidi, D. E. Leaird, and A. M. Weiner, “Comb-based radio-frequency photonic filters with rapid tunability and high selectivity,” Nature Photonics, vol. 6, no. 3, pp. 186–194, 2012. [7] P. Del’Haye, A. Schliesser, O. Arcizet, T. Wilken, R. Holzwarth, and T. Kippenberg, “Optical frequency comb generation from a monolithic microresonator,” Nature, vol. 450, pp. 1214-1217, 2007. [8] M. Jinno, B. Kozicki, H. Takara, A. Watanabe, Y. Sone, T. Tanaka, and A. Hirano, "Distance-adaptive spectrum resource allocation in spectrum-sliced elastic optical path network [Topics in Optical Communications]," IEEE Communication Magazine., vol. 48, no. 8, pp. 138-145, 2010. [9] O. Gerstel, M. Jinno, A. Lord, and S. J. B. Yoo, "Elastic optical networking: a new dawn for the optical layer?," IEEE Communication Magazine, vol. 50, no. 2, pp. 12-20, 2012. [10] K. Christodoulopoulos, I. Tomkos, and E. A. Varvarigos, "Elastic Bandwidth Allocation in Flexible OFDM-Based Optical Networks," Journal of Lightwave Technology, vol. 29, no. 9, pp.1354-1366, 2011. [11] B. Guo, Y. Xu, P. Zhu, Y. Zhong, Y. Chen, J. Li, Z. Chen, and Y. He, "Multicasting based optical inverse multiplexing in elastic optical network," Optics Express, vol. 22, no. 12, pp. 15133-15142, 2014. 96 [12] N. Sambo, F. Paolucci, G. Meloni, F. Fresi, L. Potì, and P. Castoldi, "Control of Frequency Conversion and Defragmentation for Super-Channels [Invited]," Journal of Optical Communication and Networking, vol. 7, no. 1, pp. A126- A134, 2015. [13] J. K. Fischer, S. Alreesh, R. Elschner, F. Frey, M. Nölle, C. Schmidt- Langhorst, and C. Schubert, "Bandwidth-Variable Transceivers based on Four-Dimensional Modulation Formats," Journal of Lightwave Technology, vol. 32, no. 16, pp. 2886-2895, 2014. [14] M. Jinno, H. Takara, B. Kozicki, Y. Tsukishima, Y. Sone, and S. Matsuoka, "Spectrum-efficient and scalable elastic optical path network: architecture, benefits, and enabling technologies," IEEE Communication Magazine., vol. 47, no. 11, pp. 66-73, 2009. [15] Y. Cao, A. Almaiman, M. Ziyadi, A. Mohajerin-Ariaei, C. Bao, P. Liao, F. Alishahi, A. Fallahpour, Y. Akasaka, C. Langrock, M. Fejer, J. Touch, M. Tur, and A. E. Willner, "Reconfigurable channel slicing and stitching for an optical signal to enable fragmented bandwidth allocation using nonlinear wave mixing and an optical frequency comb," Journal of Lightwave Technology, vol. 36, no. 2, pp. 440-446, 2018. [16] G. Lu, T. Bo, T. Sakamoto, N. Yamamoto, and C. Chan, "Flexible and scalable wavelength multicast of coherent optical OFDM with tolerance against pump phase-noise using reconfigurable coherent multi-carrier pumping," Optics Express, vol. 24, no. 20, pp. 22573-22580, 2016. [17] Y. Cao, M. Ziyadi, A. Almaiman, A. Mohajerin-Ariaei, P. Liao, C. Bao, F. Alishahi, A. Fallahpour, B. Shamee, A. J. Willner, Y. Akasaka, T. Ikeuchi, S. Wilkinson, C. Langrock, M. M. Fejer, J. Touch, M. Tur, and A. E. Willner, "Pilot-tone-based self-homodyne detection using optical nonlinear wave mixing," Optics Letters, vol. 42, no. 9, pp. 1840-1843, 2017. [18] S. J. Savory, "Digital filters for coherent optical receivers," Optics Express, vol. 16, no. 2, pp. 804-817, 2008. [19] S. J. Savory, G. Gavioli, R. I. Killey, and P. Bayvel, "Electronic compensation of chromatic dispersion using a digital coherent receiver," Optics Express, vol. 15, no. 5, pp. 2120-2126, 2007. [20] J. McNicol, M. O'Sullivan, K. Roberts, A. Comeau, D. McGhan, and L. Strawczynski, "Electrical domain compensation of optical dispersion [optical fibre communication applications]," in Optical Fiber Communication Conference, Anaheim, USA, 2005, paper OThJ3. [21] B. J. C. Schmidt, A. J. Lowery, and J. Armstrong, "Experimental Demonstrations of Electronic Dispersion Compensation for Long-Haul 97 Transmission Using Direct-Detection Optical OFDM," Journal of Lightwave Technology, vol. 26, no. 1, pp. 196-203, 2008. [22] G. Lu, A. Albuquerque, B. J. Puttnam, T. Sakamoto, M. Drummond, R. Nogueira, A. Kanno, S. Shinada, N. Wada, and T. Kawanishi, "Pump-phase- noise-free optical wavelength data exchange between QAM signals with 50- GHz channel-spacing using coherent DFB pump," Optics Express, vol. 24, no. 4, pp. 3702-3712, 2016. [23] F. Fresi, M. Secondini, G. Berrettini, G. Meloni, and L. Potì, "Impact of Optical and Electrical Narrowband Spectral Shaping in Faster Than Nyquist Tb Superchannel," Photonics Technology Letters, vol. 25, no. 23, pp. 2301- 2303, 2013. [24] G. Huang, Y. Miyoshi, A. Maruta, Y. Yoshida, and K. I. Kitayama, "All- Optical OOK to 16-QAM Modulation Format Conversion Employing Nonlinear Optical Loop Mirror," Journal of Lightwave Technology, vol. 30, no. 9, pp. 1342-1350, 2012. [25] S. J. Savory, “Digital optical coherent receivers: Algorithms and subsystems,” IEEE Journal of Selected Topics in Quantum Electronics, vol. 16, no. 5, pp. 1164–1179, 2010. [26] E. Ip, P. Ji, E. Mateo, Y.-K. Huang, L. Xu, D. Qian, N. Bai, and T. Wang, “100G and beyond transmission technologies for evolving optical networks and relevant physical-layer issues,” Proc. IEEE, vol. 100, no. 5, pp. 1065– 1078, 2012. [27] D. Lavery, R. Maher, D. S. Millar, B. C. Thomsen, P. Bayvel and S. J. Savory, "Digital Coherent Receivers for Long-Reach Optical Access Networks," in Journal of Lightwave Technology, vol. 31, no. 4, pp. 609-620, 2013. [28] K. Roberts, D. Beckett, D. Boertjes, J. Berthold and C. Laperle, "100G and beyond with digital coherent signal processing," in IEEE Communications Magazine, vol. 48, no. 7, pp. 62-69, 2010. [29] J. Wang, C. Xie and Z. Pan, "Generation of Spectrally Efficient Nyquist- WDM QPSK Signals Using Digital FIR or FDE Filters at Transmitters," in Journal of Lightwave Technology, vol. 30, no. 23, pp. 3679-3686, 2012. [30] E. Connolly, A. Kaszubowska-Anandarajah and L. P. Barry, "Cross Channel Interference due to Wavelength Drift of Tunable Lasers in DWDM Networks," in IEEE Photonics Technology Letters, vol. 19, no. 8, pp. 616-618, 2007. [31] T. Koike-Akino, D. Millar, K. Kojima and K. Parsons, "Laser Frequency Drift Compensation with Han–Kobayashi Coding in Superchannel Nonlinear 98 Optical Communications," European Conference on Optical Communication, paper Th.1.6.5, Valencia, 2015, pp. 1-3. [32] J. Pan, A. Stark, C. Liu and S. E. Ralph, "Fractionally-spaced frequency domain linear crosstalk cancellation with spectral alignment techniques for coherent superchannel optical systems," Optical Fiber Communication Conference, paper OW4B.6, Anaheim, CA, 2013, pp. 1-3. [33] N. Kaneda, T. Pfau and J. Lee, "Frequency diversity MIMO detection for coherent optical transmission," European Conference on Optical Communication, paper Th.2.C.1, London, 2013, pp. 1-3. [34] M. Sato, R. Maher, D. Lavery, K. Shi, B. C. Thomsen and P. Bayvel, "Frequency Diversity MIMO Detection for DP- QAM Transmission," in Journal of Lightwave Technology, vol. 33, no. 7, pp. 1388-1394, 2015. [35] C. Liu, J. Pan, T. Detwiler, A. Stark, Y. T. Hsueh, G. K. Chang and S. E. Ralph, "Joint ICI cancellation for superchannel coherent optical systems in nonlinear transmission regimes," Optical Fiber Communication Conference, paper OTu2I.4, Anaheim, CA, 2013, pp. 1-3. [36] T. Zeng, "Superchannel transmission system based on multi-channel equalization," Optics Express, vol. 21, no. 12, pp. 14799-14807, 2013. [37] F. Hamaoka, K. Saito, T. Matsuda and A. Naka, "Super high density multi- carrier transmission system by MIMO processing," European Conference on Optical Communication, paper Mo.3.5.5, Cannes, 2014, pp. 1-3. [38] S. Chandrasekhar and Xiang Liu, "Experimental investigation on the performance of closely spaced multi-carrier PDM-QPSK with digital coherent detection," Optics Express vol. 17, no. 24, pp. 21350-21361, 2009. [39] K. Shibahara, A. Masuda, S. Kawai and M. Fukutoku, "Multi-stage successive interference cancellation for spectrally-efficient super-Nyquist transmission," European Conference on Optical Communication, paper Th.1.6.4, Valencia, 2015, pp. 1-3. [40] S. Yamamoto, K. Saito, A. Naka and H. Maeda, "Compatibility between nonlinear compensation and crosstalk compensation using MIMO processing in super-high-density multi-carrier transmission system," European Conference on Optical Communication, paper Th.2.5.3, Valencia, 2015, pp. 1- 3. [41] F. Lehmann, P. Ramantanis and Y. Frignac, "Joint channel estimation, interference mitigation, and decoding for WDM coherent optical communications," Journal of Optical Communications and Networking, vol. 6, no. 3, pp. 315-325, 2014. 99 [42] P. J. Winzer, G. Raybon, C. R. Doerr, M. Duelk and C. Dorrer, "107-gb/s optical signal generation using electronic time-division multiplexing," Journal of Lightwave Technology, vol. 24, no. 8, pp. 3107-3113, 2006. [43] S. Khaleghi, O. F. Yilmaz, M. R. Chitgarha, M. Tur, N. Ahmed, S. R. Nuccio, I. M. Fazal, X. Wu, M. W. Haney, C. Langrock, M. M. Fejer and A. E. Willner, "High-Speed Correlation and Equalization Using a Continuously Tunable All-Optical Tapped Delay Line," IEEE Photonics Journal, vol. 4, no. 4, pp. 1220-1235, 2012. [44] Y. Cao, M. Ziyadi, A. Mohajerin-Ariaei, A. Almaiman, P. Liao, C. Bao, F. Alishahi, A. Fallahpour, B. Shamee, J. -Y. Yang, Y. Akasaka, M. Sekiya, M. Tur, C. Langrock, M. Fejer, J. Touch, and A. Willner, "Reconfigurable optical inter-channel interference mitigation for spectrally overlapped QPSK signals using nonlinear wave mixing in cascaded PPLN waveguides," Optics Letters, vol. 41, no. 14, pp. 3233-3236, 2016. [45] S. Yamamoto, K. Saito, F. Hamaoka, T. Matsuda, A. Naka and H. Maeda, "Characteristics Investigation of High-Speed Multi-Carrier Transmission Using MIMO-Based Crosstalk Compensation in Homodyne Detection Scheme," Journal of Lightwave Technology, vol. 34, no. 11, pp. 2824-2832, 2016. [46] J. Treichler and M. Larimore, "New processing techniques based on the constant modulus adaptive algorithm," IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 33, no. 2, pp. 420-431, Apr 1985. [47] D. Moss, R. Morandotti, A. Gaeta, and M. Lipson, “New CMOS-compatible platforms based on silicon nitride and Hydex for nonlinear optics,” Nature Photonics, vol. 7, pp. 597–607, 2013. [48] T. Rahman, D. Rafique, B. Spinnler, S. Calabro, E. Man, U. Feiste, A. Napoli, M. Bohn, G. Khanna, N. Hanik, E. Pincemin, C. Bouette, J. Jauffrit, S. Bordais, C. Andre, C. Dourthe, B. Raguenes, C. Okonkwo, A. Koonen, and H. Waardt, "Long-Haul Transmission of PM-16QAM-, PM-32QAM-, and PM- 64QAM-Based Terabit Superchannels Over a Field Deployed Legacy Fiber," Journal of Lightwave Technology, vol. 34, no. 13, pp. 3071-3079, 2016. [49] L. Zhuang, C. Zhu, B. Corcoran, M. Burla, G. Roeloffzen, A. Leinse, J. Schröder, and A. Lowery, "Sub-GHz-resolution C-band Nyquist-filtering interleaver on a high-index-contrast photonic integrated circuit," Optics Express, vol. 24, no. 6, pp. 5715-5727, 2016. [50] T. Richter, C. Schmidt-Langhorst, R. Elschner, T. Kato, S. Watanabe, and C. Schubert, "Distributed Generation of a 400-Gb/s Nyquist 16QAM Dense Superchannel by Fiber-Frequency Conversion," Optical Fiber Communication Conference, paper Th1D.3, 2014. 100 [51] H. Mardoyan, M. Mestre, J. Estaran, F. Jorge, F. Blache, P. Angelini, A. Konczykowska, M. Riet, V. Nodji-adjim, J. Dupuy, and S. Bigo, "84-, 100-, and 107-GBd PAM-4 intensity modulation direct-detection transceiver for datacenter interconnects", Journal of Lightwave Technology, vol. 35, no. 6, pp. 1253-1259, 2017. [52] D. Plant, M. Morsy-Osman, and M. Chagnon, "Optical Communication Systems for Datacenter Networks," Optical Fiber Communication Conference, paper W3B.1, 2017 [53] Z. Li, M. Erkılınç, K. Shi, E. Sillekens, L. Galdino, B. Thomsen, P. Bayvel, and R. Killey, "SSBI Mitigation and the Kramers–Kronig Scheme in Single- Sideband Direct-Detection Transmission With Receiver-Based Electronic Dispersion Compensation," Journal of Lightwave Technology, vol. 35, no. 10, pp. 1887-1893, 2017. [54] H. Murata, K. Kaneda, and Y. Okamura, "38GHz optical single-sideband modulation by using guided-wave electrooptic modulators with periodic polarization-reversal," Conference on Lasers and Electro-Optics/International Quantum Electronics Conference and Photonic Applications Systems Technologies, paper CThT10, 2004. [55] H. Chen, J. Lee, N. Kaneda, J. Chen, and Y. Chen, "Comparison of VSB PAM4 and OOK Signal in an EML-Based 80-km Transmission System," Photonics Technology Letters, vol. 29, no. 23, pp. 2063-2066, 2017. [56] C. Wong, S. Zhang, Y. Fang, L. Liu, T. Wang, Q. Zhang, S. Deng, G. Liu, and X. Xu, "Silicon IQ Modulator for Next-Generation Metro Network," Journal of Lightwave Technology, vol. 34, no. 2, pp. 730-736, 2016. [57] X. Ruan, K. Li, D. Thomson, C. Lacava, F. Meng, I. Demirtzioglou, P. Petropoulos, Y. Zhu, G. Reed, and F. Zhang, "Experimental comparison of direct detection Nyquist SSB transmission based on silicon dual-drive and IQ Mach-Zehnder modulators with electrical packaging," Optics Express, vol. 25, no. 16, pp. 19332-19342, 2017. [58] H. Chen, N. Kaneda, J. Lee, J. Chen, and Y. Chen, "Optical filter requirements in an EML-based single-sideband PAM4 intensity-modulation and direct- detection transmission system," Optics Express, vol. 25, no. 6, pp. 5852-5860, 2017. [59] M. Asghari and J. Azaña, "All-optical Hilbert transformer based on a single phase-shifted fiber Bragg grating: design and analysis," Optics Letters, vol. 34, no. 3, pp. 334-336, 2009. [60] K. Takano, N. Hanzawa, S. Tanji, and K. Nakagawa, "Experimental Demonstration of Optically Phase-Shifted SSB Modulation with Fiber-Based 101 Optical Hilbert Transformers," Optical Fiber Communication Conference, paper JThA48, 2007. [61] C. Sima, J. Gates, H. Rogers, P. Mennea, C. Holmes, M. Zervas, and P. Smith, "Phase controlled integrated interferometric single-sideband filter based on planar Bragg gratings implementing photonic Hilbert transform," Optics Letters, vol.38, no. 5, pp. 727-729, 2013. [62] X. Wang, M. Hanawa, K. Nakamura, K. Takano, and K. Nakagawa, "Sideband suppression characteristics of optical SSB generation filter with sampled FBG based 4-taps optical Hilbert transformer," Asia-Pacific Conference on Communications, pp. 622-625, 2009. [63] X. Xu, J. Wu, M. Tan, T. Nguyen, S. Chu, B. Little, R. Morandotti, A. Mitchell, and D. Moss, "Orthogonally Polarized RF Optical Single Sideband Generation and Dual-Channel Equalization Based on an Integrated Microring Resonator," Journal of Lightwave Technology, vol. 36, no. 20, pp. 4808-4818, 2018. [64] Y. Cao, K. Zou, H. Zhou, A. Almaiman, P. Liao, F. Alishahi, A. Fallahpour, K. Manukyan, and A. E. Willner, "Demonstration of Tunable Optical Single- Sideband Generation of 20-Gbit/s OOK and PAM4 Data Channels," Optical Fiber Communications Conference, paper M1B.6, San Diego, CA, 2019. [65] A. Nag, M. Tornatore, and B. Mukherjee, "Optical Network Design With Mixed Line Rates and Multiple Modulation Formats," Journal of Lightwave Technology, vol. 28, no. 4, pp. 466–475, 2010. [66] Y. Tian, Y. Huang, S. Zhang, P. Prucnal, and T. Wang, "Demonstration of digital phase-sensitive boosting to extend signal reach for long-haul WDM systems using optical phase-conjugated copy," Optics Express, vol. 21, no. 4, pp. 5099–5106, 2013. [67] X. Yi, X. Chen, D. Sharma, C. Li, M. Luo, Q. Yang, Z. Li, and K. Qiu, "Digital coherent superposition of optical OFDM subcarrier pairs with Hermitian symmetry for phase noise mitigation," Optics Express, vol. 22, no. 11, pp. 13454–13459, 2014. [68] T. Umeki, O. Tadanaga, M. Asobe, Y. Miyamoto, and H. Takenouchi, "First demonstration of high-order QAM signal amplification in PPLN-based phase sensitive amplifier," Optics Express, vol. 22, no. 3, pp. 2473–2482, 2014. [69] A. Albuquerque, B. Puttnam, J. Mendinueta, M. Drummond, S. Shinada, R. Nogueira, and N. Wada, "Experimental investigation of phase-sensitive amplification of data signals in a four-mode fiber-based PSA," Optics Letters, vol. 40, no. 2, pp. 288–291, 2015. 102 [70] Y. Cao, K. Zou, P. Liao, A. Almaiman, A. Fallahpour, F. Alishahi, A. Mohajerin-Ariaei, C. Bao, A. N. Willner, D. Starodubov, M. Tur, and A. E. Willner, "Performance Enhancement of an Optical High-Order QAM Channel by Adding Correlated Data to Robust Neighboring BPSK or QPSK Channels," Optics Letters, vol. 43, no. 23, pp. 5697-5700, 2018. [71] G. Karam and H. Sari, "A data predistortion technique with memory for QAM radio systems," IEEE Transactions on Communications, vol. 39, no. 2, pp. 336–344, 1991. [72] Y. Yu and J. Zhao, "Modified phase-conjugate twin wave schemes for fiber nonlinearity mitigation," Optics Express, vol. 23, no. 23, pp. 30399–30413, 2015. [73] A. J. Viterbi and A. M. Viterbi, "Nonlinear estimation of PSK-modulated carrier phase with application to burst digital transmission," IEEE Transactions on Information Theory, vol. 29, no. 4, pp. 543–551, 1983. [74] T. Pfau, S. Hoffmann, and R. Noé, "Hardware-efficient coherent digital receiver concept with feedforward carrier recovery for M-QAM constellations," Journal of Lightwave Technology, vol. 27, no. 8, pp. 989–999, 2009. [75] K. Toyoda, Y. Koizumi, T. Omiya, M. Yoshida, T. Hirooka, and M. Nakazawa, "Marked performance improvement of 256 QAM transmission using a digital back-propagation method," Optics Express, vol. 20, no. 18, pp. 19815–19821, 2012. [76] B. Corcoran, B. Foo, and A. Lowery, "Single-photodiode per polarization receiver with signal-signal beat interference suppression through heterodyne detection," Optics Express, vol. 26, no. 3, pp. 3075-3086, 2018. [77] J. R. Treichler and B. G. Agee, "A new approach to multipath correction 3f constant modulus signals," IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 31, no. 2, pp. 459-472, 1983. [78] X. Zhou and J. Yu, "Multi-Level, Multi-Dimensional Coding for High-Speed and High-Spectral-Efficiency Optical Transmission," Journal of Lightwave Technology, vol. 27, no. 16, pp. 3641–3653, 2009. [79] M. Selmi, C. Gosset, M. Noelle, P. Ciblat, and Y. Jaouén, "Blockwise digital signal processing for PolMux QAM/PSK optical coherent systems," Journal of Lightwave Technology, vol. 29, no. 20, pp. 3070–3082, 2011. [80] B. Clerckx and C. Oestges, MIMO Wireless Networks: Channels, Techniques and Standards for Multi-Antenna, Multi-User and Multi-Cell Systems, Academic Press (Elsevier), 2013. 103 [81] M. Saruwatari, "All-optical signal processing for terabit/second optical transmission," Journal of Selected Topics in Quantum Electronics, vol. 6, no. 6, pp. 1363–1374, 2000. [82] R. Slavík, F. Parmigiani, J. Kakande, C. Lundström, M. Sjödin, P. Andrekson, R. Weerasuriya, S. Sygletos, A. Ellis, L. Nielsen, D. Jakobsen, S. Herstrøm, R. Phelan, J. O'Gorman, A. Bogris, D. Syvridis, S. Dasgupta, P. Petropoulos, and D. Richardson, "All-optical phase and amplitude regeneration for next- generation telecommunication systems," Nature Photonics, vol. 4, pp. 690– 695, 2010. [83] T. Umeki, H. Takara, Y. Miyamoto, and M. Asobe, "3-dB signal-ASE beat noise reduction of coherent multi-carrier signal utilizing phase sensitive amplification," Opt. Exp. Vol. 20, no. 22, pp. 24727-24734, 2012. [84] K. J. Lee, F. Parmigiani, S. Liu, J. Kakande, P. Petropoulos, K. Gallo, and D. Richardson, "Phase sensitive amplification based on quadratic cascading in a periodically poled lithium niobate waveguide," Optics Express, vol. 17, no. 22, pp. 20393–20400, 2009. [85] G. Lu, T. Sakamoto, and T. Kawanishi, "Pump-Phase-Noise-Tolerant Wavelength Multicasting for QAM Signals using Flexible Coherent Multi- Carrier Pump," Optical Fiber Communication Conference, paper M2E.2, 2015. [86] H. Hu, E. Palushani, M. Galili, H. Mulvad, A. Clausen, L. Oxenløwe, and P. Jeppesen, "640 Gbit/s and 1.28 Tbit/s polarisation insensitive all optical wavelength conversion," Optics Express, vol. 18, no. 10, pp. 9961-9966, 2010. [87] J. Leuthold, L. Moller, J. Jaques, S. Cabot, L. Zhang, P. Bernasconi, M. Cappuzzo, L. Gomez, E. Laskowski, E. Chen, A. Wong-Foy, and A. Griffin, "160 Gbit/s SOA all-optical wavelength converter and assessment of its regenerative properties," Electronics Letters, vol. 40, no. 9, pp. 554–555, 2004. [88] J. Kakande, R. Slavík, F. Parmigiani, A. Bogris, D. Syvridis, L. Nielsen, R. Phelan, P. Petropoulos, and D. Richardson, "Multilevel quantization of optical phase in a novel coherent parametric mixer architecture," Nature Photonics, vol. 5, pp. 748–752, 2011. [89] B. J. Puttnam, D. Mazroa, S. Shinada, and N. Wada, "Phase-squeezing properties of non-degenerate PSAs using PPLN waveguides," Optics Express, vol. 19, no. 26, pp. B131–B139, 2011. [90] J. Levy, M. Foster, A. Gaeta, and M. Lipson, "Harmonic generation in silicon nitride ring resonators," Optics Express, vol. 19, no. 12, pp. 11415-11421, 2011. [91] D. Logan, A. Dow, D. Stepanov, P. Abolghasem, N. Kherani and A. Helmy, "Harnessing second-order optical nonlinearities at interfaces in multilayer 104 silicon-oxy-nitride waveguides," Applied Physics Letters, vol. 102, no.6, pp. 061106-061106, 2013. [92] J. Hansryd, and P. Andrekson, "Broad-band continuous-wave-pumped fiber optical parametric amplifier with 49dB gain and wavelength-conversion efficiency," Photonics Technology Letters 13, no. 3, pp. 194-196, 2001. [93] P. Kylemark, P. Hedekvist, H. Sunnerud, M. Karlsson, and P. Andrekson, "Noise characteristics of fiber optical parametric amplifiers," Journal of Lightwave Technology, vol. 22, no. 2, pp. 409–416, 2004. [94] M. Marhic, K. Wong, M. Ho, and L. Kazovsky, "92% pump depletion in a continuous-wave one-pump fiber optical parametric amplifier." Optics letters vo. 26, no. 9, pp. 620-622, 2001. [95] X. Guo, K. Lei, X. Fu, H. Tsang, and C. Shu, "Amplitude Regeneration of 80- Gb/s Polarization-Multiplexed RZ-DPSK Signals in a Dual-Orthogonal-Pump Fiber Optical Parametric Amplifier," Optical Fiber Communication Conference, paper OTu2D, 2013. [96] A. Mohajerin-Ariaei, M. Ziyadi, M. Chitgarha, A. Almaiman, Y. Cao, B. Shamee, J. Yang, Y. Akasaka, M. Sekiya, S. Takasaka, R. Sugizaki, J. Touch, M. Tur, C. Langrock, M. Fejer, and A. Willner, "Phase noise mitigation of QPSK signal utilizing phase-locked multiplexing of signal harmonics and amplitude saturation," Optics Letters, vol. 40, no. 14, pp. 3328-3331, 2015. [97] F. Parmigiani, K.R.H. Bottrill, L. Jones, G. Hesketh, D.J. Richardson, and P. Petropoulos, "Phase and Amplitude Regeneration through Sequential PSA and FWM Saturation in HNLF," European Conference on Optical Communication, paper We.3.6.3, 2015. [98] Y. Cao, M. Ziyadi, Y. Akasaka, A. Mohajerin-Ariaei, J. -Y. Yang, A. Almaiman, P. Liao, F. Alishahi, N. Ahmed, A. J. Willner, S. Takasaka, R. Sugizaki, J. Touch, M. Sekiya, M. Tur, and A. Willner, "All Optical Signal Level Swapping and Multilevel Amplitude Noise Mitigation based on Different Regions of Optical Parametric Amplification," Optics Letters, vol. 41, no. 4, pp. 677-680, 2016. [99] B. Stiller, G. Onishchukov, B. Schmauss, and G. Leuchs, "Phase regeneration of a star-8-QAM signal in a phase-sensitive amplifier with conjugated pumps," Optics Express, vol. 22, no. 1, pp. 1028-1035, 2014. [100] J. Yang, Y. Akasaka, and M. Sekiya , "Optical Phase Regeneration of Multi- Level PSK Using Dual-Conjugate Pump Degenerate Phase-Sensitive Amplification," European Conference on Optical Communication, paper P3.07, 2012. 105 [101] S. Takasaka, Y. Taniguchi, M. Takahashi, J. Hiroshi, M. Tadakuma, H. Matsuura, K. Doi, and R. Sugizaki, “Quasi Phase-Matched FOPA with 50nm Gain Bandwidth Using Dispersion Stable Highly Nonlinear Fiber,” Optical Fiber Communication Conference, paper W3E.2, 2014. [102] F. Forghieri, R. W. Tkach, A. R. Chraplyvy, and D. Marcuse, “Reduction of four-wave mixing crosstalk in WDM systems using unequally spaced channels,” Photonics Technology Letters, vol. 6, no. 6, pp. 754–756, 1994. [103] K. Wong, G. Lu, and L. Chen, "Polarization-interleaved WDM signals in a fiber optical parametric amplifier with orthogonal pumps," Optics Express, vol. 15, no. 1, pp. 56-61, 2007. [104] D. Lavery, R. Maher, D. S. Millar, B. C. Thomsen, P. Bayvel and S. J. Savory, "Digital Coherent Receivers for Long-Reach Optical Access Networks," Journal of Lightwave Technology, vol. 31, no. 4, pp. 609-620, 2013. [105] M. G. Taylor, "Coherent detection method using DSP for demodulation of signal and subsequent equalization of propagation impairments," Photonics Technology Letters, vol. 16, no. 2, pp. 674-676, 2004. [106] L. Kazovsky, "Balanced phase-locked loops for optical homodyne receivers: Performance analysis, design considerations, and laser linewidth requirements," Journal of Lightwave Technology, vol. 4, no. 2, pp. 182-195, 1986. [107] S. Ristic, A. Bhardwaj, M. J. Rodwell, L. A. Coldren and L. A. Johansson, "An Optical Phase-Locked Loop Photonic Integrated Circuit," Journal of Lightwave Technology, vol. 28, no. 4, pp. 526-538, 2010. [108] A. Tarighat, R. Hsu, A. Sayed, and B. Jalali, “Digital adaptive phase noise reduction in coherent optical links,” Journal of Lightwave Technology, vol. 24, no. 3, pp. 1269–1276, 2006. [109] H. Louchet, K. Kuzmin, and A. Richter, “Improved DSP algorithms for coherent 16-QAM transmission,” European Conference on Optical Communication, Brussels, Belgium, Paper Tu.1.E.6, 2008. [110] T. Sakamoto, G. Lu, A. Chiba, and T. Kawanishi, "Digital Optical Phase Locked Loop for Real-Time Coherent Demodulation of Multilevel PSK/QAM," Optical Fiber Communication Conference, paper OMS5, 2010. [111] X. Zhou, “An improved feed-forward carrier recovery algorithm for coherent receivers with M-QAM modulation format,” Photonics Technology Letters, IEEE, vol. 22, no. 14, pp. 1051 –1053, july15, 2010. [112] J. Li, L. Li, Z. Tao, T. Hoshida, and J. Rasmussen, “Laser-linewidthtolerant feed-forward carrier phase estimator with reduced complexity for QAM,” Journal of Lightwave Technology, vol. 29, no. 16, pp. 2358– 2364, 2011. 106 [113] I. Fatadin, D. Ives and S. J. Savory, "Laser Linewidth Tolerance for 16-QAM Coherent Optical Systems Using QPSK Partitioning," Photonics Technology Letters, vol. 22, no. 9, pp. 631-633, 2010. [114] T. Kobayashi, A. Sano, A. Matsuura, Y. Miyamoto and K. Ishihara, "Nonlinear Tolerant Spectrally-Efficient Transmission Using PDM 64-QAM Single Carrier FDM With Digital Pilot-Tone," Journal of Lightwave Technology, vol. 30, no. 24, pp. 3805-3815, 2012. [115] M. Morsy-Osman, Q. Zhuge, L. R. Chen, and D. V. Plant, "Feedforward carrier recovery via pilot-aided transmission for single-carrier systems with arbitrary M-QAM constellations," Optics Express, vol. 19, no. 24, pp. 24331- 24343, 2011. [116] T. N. Huynh, L. Nguyen, V. Vujicic and L. P. Barry, "Pilot-tone-aided transmission of high-order QAM for optical packet switched networks," Journal of Optical Communications and Networking, vol. 6, no. 2, pp. 152- 158, 2014. [117] N. A. Olsson and J. van der Ziel, “Cancellation of fiber loss by semiconductor laser pumped Brillouin amplification at 1.5 μm,” Applied Physics Letters, vol. 48, pp. 1329-1330, 1986. [118] L. Banchi, M. Presi, R. Proietti, and E. Ciaramella, "System feasibility of using stimulated Brillouin scattering in self coherent detection schemes," Optics Express, vol. 18, no. 12, pp. 12702-12707, 2010. [119] A. Lorences-Riesgo, M. Mazur, T. Eriksson, P. Andrekson, and M. Karlsson, "Self-homodyne 24×32-QAM superchannel receiver enabled by all-optical comb regeneration using brillouin amplification," Optics Express, vol. 24, no. 26, pp. 29714-29723, 2016. [120] R. Griffin and A. Carter, “Optical differential quadrature phase-shift key (oDQPSK) for high capacity optical transmission,” Optical Fiber Communication Conferencw, paper WX6, 2002. [121] R. Noé, “Phase noise tolerant synchronous QPSK/BPSK baseband type intradyne receiver concept with feedforward carrier recovery,” Journal of Lightwave Technology, vol. 23, no. 2, pp. 802–808, 2005. [122] T. Miyazaki, "Linewidth-tolerant QPSK homodyne transmission using a polarization-multiplexed pilot carrier," Photonics Technology Letters, vol. 18, no. 2, pp. 388-390, 2006. [123] S. Beppu, K. Kasai, M. Yoshida, and M. Nakazawa, "2048 QAM (66 Gbit/s) single-carrier coherent optical transmission over 150 km with a potential SE of 15.3 bit/s/Hz," Optics Express, vol. 23, no. 4, pp. 4960-4969, 2015. 107 [124] C. M. Caves, "Quantum limits on noise in linear amplifiers," Physical Review D, vol. 26, no. 8, pp. 1817-1839, 1982. [125] M. Vasilyev, "Distributed phase-sensitive amplification," Optics Express, vol. 13, no. 19, pp. 7563-7571, 2005. [126] Z. Tong, C. Lundstrom, P. A. Andrekson, C. J. McKinstrie, M. Karlsson, D. J. Blessing, E. Tipsuwannakul, B. J. Puttnam, H. Toda, and L. Gruner-Nielsen, "Towards ultrasensitive optical links enabled by low-noise phase-sensitive amplifiers," Nature Photonics, vol. 5, pp. 430-436, 2011. [127] T. Kazama, T. Umeki, M. Abe, K. Enbutsu, Y. Miyamoto, and H. Takenouchi, "Low-Parametric-Crosstalk Phase-Sensitive Amplifier for Guard-Band-Less DWDM Signal Using PPLN Waveguides," Journal of Lightwave Technology, vol. 35, no. 4, pp. 755-761, 2017. [128] J. Kakande, C. Lundström, P. A. Andrekson, Z. Tong, M. Karlsson, P. Petropoulos, F. Parmigiani, and D. Richardson, "Detailed characterization of a fiber-optic parametric amplifier in phase-sensitive and phase-insensitive operation," Optics Express, vol. 18, no. 5, pp. 4130-4137, 2010. [129] K. J. Lee, F. Parmigiani, S. Liu, J. Kakande, P. Petropoulos, K. Gallo, and D. Richardson, "Phase sensitive amplification based on quadratic cascading in a periodically poled lithium niobate waveguide," Optics Express, vol. 17, no. 12, pp. 20393-20400, 2009. [130] R. Neo, J. Schröder, Y. Paquot, D-Y Choi, S. Madden, B. Luther-Davies, and B. J. Eggleton, "Phase-sensitive amplification of light in a χ(3) photonic chip using a dispersion engineered chalcogenide ridge waveguide," Optics Express, vol. 21, no. 7, pp. 7926-7933, 2013. [131] Y. Zhang, C. Husko, J. Schröder, S. Lefrancois, I. H. Rey, T. F. Krauss, and B. J. Eggleton, "Phase-sensitive amplification in silicon photonic crystal waveguides," Optics Letters, vol. 39, no. 2, pp. 363-366, 2014. [132] J. Hou, L. Chen, W. Dong, and X. Zhang, "40 Gb/s reconfigurable optical logic gates based on FWM in silicon waveguide," Optics Express, vol. 24, no. 3, pp. 2701-2711, 2016. [133] H. Zhang, M. Tang, Y. Xie, H. Liao, S. Fu, P. Shum, and D. Liu, "Programmable all-fiber structured waveshaper based on linearly chirped fiber Bragg grating and digital thermal controller," Applied Physics B, vol. 112, no. 4, pp. 479-484, 2013. [134] Y. Cao, H. Song, Y. Akasaka, P. Liao, A. Almaiman, F. Alishahi, A. Fallahpour, C. Bao, A. Mohajerin-Ariaei, T. Ikeuchi, D. Starodubov, J. Touch, M. Tur, and A. E. Willner, "Raman-Assisted Phase Sensitive Amplifier Using 108 a Fiber Bragg Grating-Based Tunable Phase Shifter," Optics Letters, vol. 43, no. 16, pp. 3949-3952, 2018. [135] S. Y. Li, N. Q. Ngo, S. C. Tjin, P. Shum, and J. Zhang, "Thermally tunable narrow-bandpass filter based on a linearly chirped fiber Bragg grating," Optics Letters, vol. 29, no. 1, pp. 29-31, 2004. [136] X. Fu, X. Guo, and C. Shu, "Raman-Enhanced Phase-Sensitive Fibre Optical Parametric Amplifier," Scientific Report, vol. 6, pp. 1-8, 2016.

Abstract (if available)

Abstract Optical fiber communication has the merit of high spectrum bandwidth (>20THz) and can transmit a significant amount of data information within a second. For a typical dual-polarization wavelength-division-multiplexing (WDM) system, this considerable data transmission bandwidth could support ∼80 independent channels with a spacing of 50 GHz. However, this wide optical communication spectrum is being used up gradually with the rise of cloud computing, e-commerce and other internet services that require large amounts of the data transmission rate. In this regard, the manner of effectively and efficiently utilizing available optical communication spectrum resources accounts for a considerable part of research activities. ❧ In order to achieve efficient optical spectrum utilization, a common approach could be employing advanced data modulation format, enabling to encode more data information (bits) onto each symbol. In such a case, each 50-GHz bandwidth could support larger transmission capacity. A typical advanced modulation format could be quadrature-amplitude-modulation (QAM), which encodes the data information onto both the amplitude and phase of the optical carrier. However, as the QAM order increases, the deceased Euclidean distance of the symbol constellation makes the signal more vulnerable to various noises coming from both the transceivers and the transmission link. ❧ Another approach for efficient usage of the limited spectrum is to manipulate the bandwidth of the data channel. There are different manners such as using Nyquist pulse-shaping to reduce the data bandwidth to the baudrate

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

All-optical signal processing toward reconfigurable optical networks

PDF

Reconfigurable high‐speed optical signal processing and high‐capacity optical transmitter

PDF

Reconfigurable and flexible high-speed optical signal processing and spectrally shared optical subsystems

PDF

Reconfigurable high-speed processing and noise mitigation of optical data

PDF

High-speed and reconfigurable all-optical signal processing for phase and amplitude modulated signals

PDF

Optical signal processing for enabling high-speed, highly spectrally efficient and high capacity optical systems

PDF

Optical wave mixing for tunable delays and high‐speed signal processing

PDF

Microresonator-based Kerr frequency comb for high-speed optical communications and signal processing

PDF

Nonlinear optical signal processing for high-speed, spectrally efficient fiber optic systems and networks

PDF

Spectrally efficient waveforms for coherent optical fiber transmission

PDF

Reconfigurable high speed optical signal processing for optical communication and modulation format manipulation

PDF

Optical signal processing for high-speed, reconfigurable fiber optic networks

PDF

Applications in optical communications: quantum communication systems and optical nonlinear device

PDF

optical communication systems and space-time wave packet generation using frequency combs and spatial modes

PDF

Free-space optical communications and probing through turbulent links using structured light

PDF

Multiplexing techniques and reconfigurable networking functions for optical communications using orbital angular momentum beams

PDF

Integrated silicon waveguides and specialty optical fibers for optical communications system applications

PDF

Applications of all optical signal processing for advanced optical modulation formats

PDF

High-capacity optical communications using structured light in random and disturbed media

PDF

Detection and optical signal processing using phase based optical modulation formats

Asset Metadata

Creator Cao, Yinwen (author)

Core Title Reconfigurable optical signal processing for efficient spectrum utilization in high-speed optical communication systems

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Publication Date 06/13/2019

Defense Date 04/15/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag fiber optics links and subsystems,modulation,nonlinear optics,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Willner, Alan (committee chair), Haas, Stephan (committee member), Wu, Wei (committee member)

Creator Email yinwenca@usc.edu,yinwencao@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-173541

Unique identifier UC11659797

Identifier etd-CaoYinwen-7478.pdf (filename),usctheses-c89-173541 (legacy record id)

Legacy Identifier etd-CaoYinwen-7478.pdf

Dmrecord 173541

Document Type Dissertation

Format application/pdf (imt)

Rights Cao, Yinwen

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA