Modulation Format Conversion in Future Optical Networks

Javier Cano Adalid

Master Thesis

Supervisors: Christophe Peucheret and Jorge Seoane

Research Center COM Technical University of Denmark

March, 2009

ii

Abstract

Thesignalgenerationandreceiverschemesofdifferentmodulationformatssuchas on off keying (OOK) in either nonreturn to zero (NRZ) or return to zero (RZ), correlative coding and phaseshift keying (PSK) are described and numerically simulated. Their receiver sensitivities for back to back structure are numerically simulatedat40Gb/sandcompared.Theresultsshowhowusingbalancedreception, DPSK has the benefit of approximately 3dB. We demonstrate that duobinary modulation is characterized by a higher tolerance to the chromatic dispersion in comparison with NRZ. Conversion techniques between modulation formats are described.Differentallopticalformatconversionbasedonthementionedmethodare discussed. Finally, format conversion from RZOOK to RZDPSK based on HNLF nonlinearities is investigated in great detail and numerically simulated at 40Gb/s. Someconverterrequirementssuchaspowers,wavelengths,bandwidths,pulsewidths andpulseshapesamongotherarestudied.Theresultsshowthatshorterpulsewidths incontrolsignalforthecaseofwalkoffpresenceprovideworseconversion.Sensitivity of converted RZDPSK is compared with a conventional 33RZDPSK signal showing a penalty close to 4 dB. Finally, the effect of the control wavelength variation at the input of the converter is investigated when the probe signal corresponds with conventional33%RZ,50%RZand67%RZdutycycle.

iii

Acknowledgements
IwouldliketothankmanypeoplewhohavesupportedmyTelecommunication degreestudiesandhelpedmetotheprocessofwritingthisthesis. Firstly, I would like to thank my supervisors Chritophe Peucheret and Jorge Seoaneforprovidingmethisgreatopportunityofrealizingthisthesisinthefieldofthe opticalcommunicationsandforhishelpandpatience. I extend many thanks to my friends and student roommates who made me morepleasanttherealizationofthisthesiswiththeircompany(Carlos,Javi,Pabloand Antonio).IwishtothankMichalPawlikforallhishelpmyfirstdaysandformakingme easiermyaccommodationinDepartmentofPhotonics. I am very grateful to Andrea, Antonio, Javi and Carlos for spending his time readingthethesisandtryingtogivemesomeadvises. IcouldnotforgetallpeopleImetduringmyperiodinDenmarkwhogaveme unforgettablemoments,especiallyMiguel,Maria,Josepin,Spanishinvasioningeneral, andallmyflatmatesduringmystayinLuntojtevejSkolaestraede. Lastly,butnotlessimportant,thankstomyfamily(myfatherJose,mymother MaraLuisaandmysisterLaura)whoalwayshassupportedmeinallmydecisionsand gavemethepossibilitytofinishmydegreeabroadinDenmark.Thankyoudadforall youradvices,supportandscoldingsduringallmystudentlife.Iknowthatifyouhad notencouragedmemaybeIwouldhavenoteverstudiedthisdegree.BesidesIwantto thanktherestofthefamilyaswell. ManythankstoallthepeopleIdidnotmentionintheseacknowledgements.It is not because you are not important for me but because of time reasons I have to finishhere.

iv

ListofAcronyms
AMI ASE BER BPF CD CSRZ CSRZDPSK CW CROW DFG DGD DPSK DQPSK DWDM EDFA ER FDI FWHM FWM FPLD GVD HNLF MAN MZI MZDI MZM NLSE NRZ NRZOOK OOK OSNR PMD PPLN PRBS PRZ AlternateMarkInversion AmplifiedSpontaneousEmission BitErrorRate BandPassFilter ChromaticDispersion CarrierSuppressedReturntoZero CarrierSuppressedReturntoZeroDifferentialPhaseShiftKeying ContinuousWave CoupledRingresonatorOpticalWaveguide DifferenceFrequencyGeneration DifferentialGroupDelay DifferentialPhaseShiftKeying DifferentialQuadraturePhaseShiftKeying DenseWavelengthDivisionMultiplexing ErbiumDopedFiberAmplifier ExtinctionRatio FiberDelayInterferometer FullWidthatHalfMaximum FourWaveMixing FabryPerotLaserDiodes GroupVelocityDispersion HighlyNonlinearFiber MetropolitanAreaNetwork MachZehnderInterferometer MachZehnderDelayInterferometer MachZehnderModulator NonLinearSchrdingerEquation NonReturntoZero NonReturntoZeroOnOffKeying OnOffKeying OpticalSignaltoNoiseRatio PolarizationModeDispersion PeriodicallyPoledLithiumNiobate PseudoRandomBitSequence PseudoReturntoZero

PSK RZ RZDPSK RZOOK QPM SHG SOA SOP SPM TE TM WAN WDM XGM XPM

PhaseShiftKeying ReturntoZero ReturntoZeroDifferentialPhaseShiftKeying ReturntoZeroOnOffKeying QuasiPhaseMatching SecondHarmonicGeneration SemiconductorOpticalAmplifier StateofPolarization SelfPhaseModulation TransverseElectric TransverseMagnetic WideAreaNetwork WavelengthDivisionMultiplexing CrossGainModulation CrossPhaseModulation

vi

Tableofcontents
Chapter1Introduction........................................................................................................................ 1 1.1.ProjectMotivation................................................................................................ 1 1.2.ThesisStructure .................................................................................................... 4 Chapter2AdvancedModulationFormats .................................................................................. 5 2.1.Classificationofmodulationformats ................................................................... 6 2.1.1.IntensityModulationFormats ........................................................................ 7 2.1.1.1.NRZOOK.................................................................................................. 7 2.1.1.2.RZOOK .................................................................................................... 8 2.1.1.3.CSRZ....................................................................................................... 11 2.1.2.CorrelativeCodingandPartialResponseFormats ...................................... 12 2.1.2.1.Duobinary .............................................................................................. 12 2.1.3.DifferentialPhaseModulation..................................................................... 14 2.1.3.1.BinaryDPSK ........................................................................................... 14 2.1.3.2.DQPSK.................................................................................................... 18 2.2.Numericalsimulations.Sensitivitycomparison.................................................. 21 2.3.Summary............................................................................................................. 25 Chapter3AllOpticalModulationFormatsConversion ..................................................... 26 3.1.GeneralOverview.Employedtechniques ......................................................... 27 3.1.1.SemiconductorOpticalAmplifier................................................................. 27 3.1.2.MachZehnderDelayInterferometer .......................................................... 29 3.1.3.RingResonatorOpticalwaveguide.............................................................. 29 3.1.4.NarrowBandSiliconMicroringResonator .................................................. 30 3.1.5.SpectralLinebyLinepulseshaping............................................................. 31 3.1.6.Highlynonlinearfiber .................................................................................. 32 3.1.7.PPLNwaveguide........................................................................................... 34 3.2.Stateoftheart................................................................................................... 37 3.3.OOKtoOOKformatconversion ......................................................................... 39 3.3.1.Motivation.................................................................................................... 39 3.3.2.Modulationformatconversions .................................................................. 40 3.4.OOKtoCorrelativeCodingformatconversion................................................... 54

vii

3.4.1.Motivation.................................................................................................... 54 3.4.2.Modulationformatconversion.................................................................... 55 3.5.OOKtoDPSKformatconversion ........................................................................ 55 3.5.1.Motivation.................................................................................................... 55 3.5.2.Modulationformatconversions .................................................................. 56 3.6.DPSKtoCorrelativeCodingformatconversion.................................................. 61 3.7.DPSKtoDPSKformatconversion ....................................................................... 62 3.7.1.Motivation.................................................................................................... 62 3.7.2.ModulationFormatconversions.................................................................. 62 3.8.FSKtoPSKformatconversion............................................................................. 64 3.8.1.Motivation.................................................................................................... 64 3.8.2.FormatConversion....................................................................................... 65 3.7.Summary............................................................................................................. 66 Chapter4AllOpticalmodulationformatconversionfromRZOOKtoRZDPSK basedonfibernonlinearityinopticalfiber .............................................................................. 67 4.1.Principleofoperation ......................................................................................... 67 4.2.ParametersChoice.............................................................................................. 68 4.2.1.Fiberparameters.......................................................................................... 68 4.2.2.Requirementsontheshapeandwidthofpulseinbacktobacksystems... 70 4.2.3.ControlandProbesignal.............................................................................. 73 4.2.4.BandPassFilter(BPF)effectbeforereceiver .............................................. 75 4.3.SystemSimulationModel................................................................................... 77 4.4.RequirementsinControlsignalforastandardRZpulsetrainClock.................. 84 Chapter5Conclusion ........................................................................................................................ 88 References............................................................................................................................................... 90 APPENDIXA:VPIScript .................................................................................................................... 94 APPENDIXB:MATLABcode..........................................................................................................102

viii

Chapter1Introduction
1.1.ProjectMotivation
Inthepastfewyearstherehasbeenawideimprovementinthefieldoffiber opticalcommunicationsmotivatedmainlybytworeasons:thenecessitytogradually get a lower cost in optical systems and the every time higher demand of bandwidth due to the increase of data exchange [1]. For these reasons, advanced modulation formats have received special interest recently in order to enhance the impairments and take advantage of the properties of each network. Thus, different modulation formats may be selectively employed in future optical networks depending on the systemfeatures,bitrateandthenetworksize[7][26]. Historically, onoff keying in either of nonreturn to zero (NRZ) and return to zero (RZ) have become the optical modulation format choice for the most of optical communications systems [4][5]. NRZOOK has been especially employed and is still used nowadays in longhaul 10Gb/s transmission systems [5]. Nonetheless, this modulation format, and binary modulation in general, presents an important disadvantageforhighspeedfuturenetworks.Thisisitslowspectralefficiency[5]. Recently, the growing requirement of achieving higher bit rates and consequently,theimprovementofthespectralefficiency,hasopenedthedoortonew advanced modulation format which not only carry the information in their optical amplitudes but also modulate their phases in order to enhance the robustness to chromatic dispersion, optical filtering and nonlinearities [1][7]. Modulation formats such as duobinary, alternate mark inversion (AMI), chirped returntozero (CRZ) and carriersuppressed returntozero (CSRZ) belong to this advanced group [4][7][8]. Nowadays, 40Gb/s transmission systems employ especially duobinary modulation format which is more robust against transmission impairments like chromatic dispersion(CD)andnarrowbandfiltering[5]. Deserve a separate mention those modulation formats which carry the information modulated in their own optical phase. They are known as phase shift keying(PSK)[2][4].Thisformatsusethephaseoftheprecedingbitasarelativephase referencefordemodulatingtheopticalsignalduetotheabsenceofabsolutephasein directdetection[4].Thisresultsindifferentialphaseshiftkeying(DPSK)whichcanbe considered as one of the most promising modulation format for longhaul networks[26].Abalancedreceiverisusedinsteadoftheconventionaldirectdetection in order to take advantage of the 3 dB sensitivity benefit [1] [4]. Both receivers are described in great detail in nextChapter. The only multilevel modulation (more than onebitpersymbol)discussedinthisthesisisdifferentialquadraturephaseshiftkeying (DQPS). It has been also attracted a wide interest recently experimentally demonstrated that achieve the best spectral efficiency compared with rest of mentionedmodulationformats[5].

Eventhoughnewadvancedmodulationformatpresentimportantadvantages respectconventionalOOKformats,NRZisstillwidelyemployedforterrestriallonghaul transmissionsystemsandforcurrent40Gb/stransmissionduobinaryischosenasthe suitablemodulationformat[5]. Theideaofachievinganendtoendopticaltransmissionisbecominggradually a realistic scenario since optical fiber has much higher bandwidth distance product compared with copper lines. It enables high speed connections along longer distances[5]. All optical communications systems will increase networks complexity due to different modulation formats that will be chosen depending on the requirements of each future network. Furthermore, all optical signal processing becomesanimportanttechnologytocarrytheinformationonlyintheopticaldomain [7]. This fact implies the necessity of a transparent all optical modulation format conversion between different communication networks. Always possible optical electrical and electricaloptical conversion leads to an increase of cost and time, but alsoaworseningofthesystemefficiency[12]. Thus, it will be important to analyze the features of a certain network before choosingthesuitablemodulationformat.Onesimplewaytocategorizethedifferent typesofnetworksdesignisbythescopeorthescale[5].Figure1showsaschemeof the basic telecommunication network hierarchy. One link in this longhaul network transportsthebiggestamountofdatatrafficoverthelongestdistancesandincludes transnational,transcontinentalandtransoceaniccommunications[5].
Figure1.Opticalnetworkhierarchyinspiredfrom[5]

As Figure 1 illustrates, in general, they can be classified in a three level hierarchy:backbonenetworks(longdistancetransmissions),wide/metronetworksand access networks [12]. Nevertheless, to find an exact classification for a certain real networkbecomesacomplextasksincetherearemanyfactorstoconsider.Obviously, themaincharacteristicsforanopticalnetworkarethesizeandthedistancesthatthey have to cover, but also it is important to focus to other features such as bandwidth requirement, number of exchange points, points where data is destined to or originates from etc. [5][7][34]. For these reasons, suitable choices of modulation format for each network and all optical converters between them will enhance

capacity, cost effectiveness and robustness of transmission links [5]. Hence, modulation format conversion will be required at the gateways in order to achieve transparently connection (for instance between Longhaul and WAN, WAN and MAN orbetweenMANandaccess)[26]. Inthepastfewyearsavarietyofschemesforallopticalformatconversionhave been proposed and some of them experimentally demonstrated. More methods between data conversions (OOK modulation formats) have been reported in comparison with OOK to PSK, OOK to Correlative Coding or PSK to PSK due to the traditional use of OOK formats in the most of the current networks. These methods use different types of techniques based on: single semiconductor optical amplifiers (SOAs) [22][24], injectionLocked Laser Diodes [12], SOAs in MachZehnder interferometer (MZI) structure [25], SOAloop mirror [17], MachZehnder delay interferometer(MZDI)[5],coupledringresonatoropticalwaveguide(CROW)[28]etc. All of them are discussed in Chapter 3. Recently, new promising technologies like highlynonlinearfiber(HNLF)orperiodicallypoledlithiumniobate(PPLN)waveguides aregraduallyattractingmoreinterestduetoitsfastresponse[5][27][31]. The table below shows the typical distances covered by transmission links in suchnetworksandtheproposedsuitablemodulationformat. Optical transmission system UltraLongHaul LongHaul WAN MAN Access Distance (Km) Proposed Modulation Formats
PSK PSK,CSRZ,AMI RZDQPSK NRZ,Duobinary NRZ

>3000 10003000 3001000 <300 <100

Table1.Classificationofopticaltransmissionsystemsinspiredfrom[5]

Inanycase,modulationformatconversionisapromisingallopticaltechnology which will be required at the gateway of different kind of future networks and will providefullflexibilityamongdifferent.

1.2.ThesisStructure
This thesis is structured in five chapters. Chapter 1 outlines the project motivation and describes the current state in optical modulation formats and the necessity of achieving transparent format conversion in future alloptical networks. Chapter 2 gives a classification of advanced modulation formats. The employed transmittersandreceiversaredescribedandsimulatedat40Gb/s.Finally,asensitivity comparisonispresented.Chapter3startswithafrom/tomatrixreviewingallformat conversion reported until nowadays and follows with a description of the different techniquesemployedinsuchmodulationconverters.Chapter4presentsthenumerical simulationsofRZOOKtoRZDPSKformatconversionbasedoncrossphasemodulation (XPM) in an optical fiber at 40 Gb/s. Different requirements are studied from the investigation.Chapter5summarizesanddiscussestheresultsobtainedinthethesis.

Chapter 2 Advanced Modulation Formats

Atthebeginning,asimplewaywasusedinopticalcommunicationsinorderto convertfromelectricaltoopticalbinaryinformation.Electricalbit1wasassociated withahigheropticalwhilebit0wasassociatedtoaloweropticalintensity[2].These employedonoffkeyed(OOK)signalscarriedtheinformationintheiramplitudes.Asit willbedescribedinthischapter,twoformatsareincludedinthismodulationgroup: Nonreturntozero(NRZ)andReturntozero(RZ)[3][4].However,graduallyamplitude modulationwillbecomenotasuitablechoiceforfuturetransmissionsystemsdueto largechannelspacing,highopticalsignaltonoiseratio(OSNR)requirementandsome nonlinearimpairment[5]. Inthepastfewyears,advancedmodulationformatshavereceivedconsiderable attentionintheareaof40Gb/sandabovedataratesperchannelinWDMnetworks [1][4] in order to increase the capacity and robustness of transmission links [5]. This multiplexing method is used today in the core of global telecommunication network [5].

Figure2.From[2],principleofWDMsystem

Some of this new advanced modulation formats do not carry only the information in the amplitude but also they also modulate the optical phase (without carrying information) in order to enhance the robustness to chromatic dispersion, optical filtering and nonlinearities [4][7][8]. Modulation formats such as duobinary, alternate mark inversion (AMI), chirped returntozero (CRZ) and carriersuppressed returntozero (CSRZ). Finally, the last important modulation group carries the informationintheopticalphase.Itisimportanttoremarkthatduetothedifficultyof receiving a absolute phase, this systems use phase shift between consecutive bits in order to carry and recover the information in reception [4][7]. For this reason this groupiscalleddifferentialphaseshiftkeyed(DPSK). Nonetheless,nowadaysopticalnetworksarenotcharacterizedbyemployinga wide variety of modulation formats. NRZOOK continues being used in the most terrestriallonghaultransmissionsystems[4][5]andduobinaryiswidelyemployedin currenttransmissionsystemsworkingat40Gb/s[5].

In future optical networks, different modulation formats will be chosen dependingonthenetworksizeandbitrate.Forexample,OOKwillbeasuitablechoice forMANsduetotheirpoorcosteffectiveness[7].Otherwise,DPSKwhichhasattracted significantattentioninresearchlastyears,itissupposedtobethechoiceinWANsas wellaslonghaultransmissionduetoitshighspectralefficiency[5][6][7].

2.1.Classificationofmodulationformats
The purpose of this subsection is to present a classification of the main modulation formats employed in different area networks paying special attention in thosewhichareemployedintheproposalformatconversionsdiscussedinChapter3. It is obvious that a suitable choice of modulation format enhances the features in a certain communication system depending on the necessities and limitation of the networktodesign. Backtoback systems are numerically simulated in VPI Transmission Maker WDM. These structures consist of one transmitter and one receiver without any intermediatestepwhichcangenerateanykindofdegradationinthemodulatedsignal. Hereby, the modulator technology chosen for carrying out numerical simulationisMachZehndermodulators(MZMs).Dependingonthewayofdrivingthe modulator, it will enable to get the whished modulation at the output of the transmitter.Figure3illustratesanoverviewofdifferentwaystodriveMZMinorderto get all modulation formats discussed in this thesis. For the following systems the pattern length is equal to 2048 bits and they are simulated at 40Gb/s with 128 samplesperbit.

Figure3.From[1]overviewofdifferentwaystodriveanMZMinordertogetdifferentmodulation formats

2.1.1.IntensityModulationFormats
2.1.1.1.NRZOOK

Figure4.TransmitterandreceptionstructureforNRZOOKinspiredfrom[5]

Non returntozero is the oldest and simplest modulation format [1][2]. This modulationisobtainedbyswitchingalasersourcebetweenONorOFF.Thisstepcan beachievedmodulatingtheoutputofthelaserusingaMZM.ForthecaseofOOKthe modulatorisbiasedat50%transmission(thispointisknownasquadraturepoint[5]) andisdriventomaximumtominimumtransmissionsuchasisillustratedinFigure3. Hereby, it is important to note that the MZM will be driven in all the modulation formatsbybothelectricalsignal(whichcarriestheinformation)andinvertedelectrical signal. Figure 4 illustrates a NRZOOK backtoback structure. The NRZOOK transmitter is build by a continuous wave (CW) ligthwave modulated in a MZM at 40Gb/s by a 211 1 pseudorandom bit sequence (PRBS). Once the optical light is modulated,theNRZOOKsignalpassesthroughavariableopticalattenuator(VOA)in order to be able to change the received power. It is interesting to note that the bit errorration(BER)calculatedinthereceiverdependsontheamountofpoweratthe input of the receiver. The function of the power meter is to record this received power. AsFigure4shows,thereceiverismadeupofapreamplifierwith 32 dB gain and 5dB noise figure, a fourth order Gaussian optical band pass filter (OBPF) with a bandwith(BW)equalto160GhzandaPINphotodiodewithathermalnoisedensityof 10pA/Hzandaresponsivityof1A/WfollowedbyanthirdorderBesselelectricallow passfilter(LPF).This last filter is used to reduce the bandwidthnoise. TheOBPFplaced after the preamplification remove the out of band amplified spontaneous emission (ASE)noisecomingfromtheamplifier.Weassumethatforthefollowingmodulation formatsetupswewillusethesamepreamplificationandbandfiltered.Thisdescribed

transmitter and receiver correspond to the typical structure for an intensity modulationdirectdetection(IMDD)system[7]. Thefunctionoftheclockrecoveryistoresynchronizethereceiveddatabefore passing through the BER estimator module. By using a script, the exact receiver sensitivity can be calculated changing the value of the optical attenuator. The sensitivityisdefinedastheamountofreceivedpowerneededtogetaBERof 109 .

Figure5.Opticalintensityeye(a)andopticalspectrum(b)diagramsofNRZOKKsimulatedinVPI forabitrateof40Gb/s

These results illustrated in Figure 5 correspond to the NRZOOK signal at the output of the transmitter. Figure 5 a) shows the optical intensity eye diagram after simulating the setup described previously. Figure 5 b) illustrates the power spectral density as a function of the optical frequency. This is relative to 193.1Ghz, called carrier frequency, where NRZ spectrum has a strong component. Clock tones are spaced at multiples of the bit rate (40Gb/s) but they are gradually more reduced comparedtothecentralcarrier.Thebandwidthisapproximatelytwicethebitrate. One of the most important advantages of OOK is its simplicity and low cost [1][5]. Otherwise, intensity modulation formats are characterized by a bad nonlinear toleranceduetothestrongopticalcarrier[5]. 2.1.1.2.RZOOK

Figure6.TransmitterandreceptionstructureforRZOOKinspiredfrom[5]

Figure6illustratestheexperimentalsetupfora40Gb/sRZOOsystem.Forthe implementationofRZsystemweemployaconventionalNRZtransmitterfollowedbya pulse carver. In this case, we implement it sinusoidally driven a second MZM since sinusoidal signals are easily generated at 40Gb/s. Depending on the carving method chosen(itmeansthewayMZMisdrivenbythesinusoidally)itispossibletoachieve differentdutycycles[1][5]. Figure7showsthepulseshapesofRZOOKsignalswith33%,50%and67%duty cycles. Three carver methods have been used to convert the optical NRZ into its RZ equivalent: 1) 33% duty cycle: the bias point of the second modulator is fixed at the transmissionmaximum.TheMZMisdrivenbya20Ghzsinusoidwithamplitude 2V . 2)50%dutycycle:inthiscase,themodulatorisbiasedinthequadraturepoint. Itrequiresa40Ghzelectricaldrivesignalwithapeaktopeakamplitudeof V . 3)67%dutycycle:theMZMisbiasedinthemiddleoftwomaximaanddriven by a 20Ghz sinusoid with amplitude 2V . This pulse carver is characterized by alternating phase between consecutives bits and corresponds with carriersuppress returntozeroCSRZOOKmodulation.

Figure7.Pulseshapefor33%,50%and67%dutycycle

Again,asingleendeddetectionisenoughtofullyrecovertheoriginaldata.The samereceptionschemedescribedintheprevioussubsectionisusedtodetecttheRZ OOKsignalsandcalculatetheBERfordifferentreceivedpowers.


Figure8.Opticalspectraof33RZ(yellow)and50RZ(red)simulatedinVPIforabitrateof40Gb/s

Figure8showstheopticalintensityeyediagramandtheopticalspectrumafter simulatingtheRZOOKsetupdescribedfor33%and50%dutycycle.Clocktonesare spaced again at multiples of the bit rate. Both RZ modulation formats have a strong spectralcomponentatthecarrierfrequencyaswell. RZOOKhasanarrowerpulseshapecomparedwithNRZOOK.Consequently,it impliesabroaderspectrum. It should be mentioned that we obtain different peak amplitudes for a same average power depending the pulse carver we use. Obviously, in thecase of 33% RZ the energy is compressed in a narrower pulse. Therefore, the peak amplitude is the highestcomparedwith50%and67%dutycycleasisdepictedinFigure9.

Figure 9. Optical intensity waveform for different pulse carvers generated with the same average power[4].

This fact is important when we try to reason the benefit or degradation in sensitivity reception. These comparisons among modulation formats are discussed in subsection2.2.

10

2.1.1.3.CSRZ For generating carriersuppressed returntozero (CSRZ) we use the same schemedescribedinFigure6withthethirdcarvermethod(67%dutycycle)explained intheprevioussubsection. Thismodulationformatischaracterizednotonlybyencodingtheinformation on both OOK intensity levels (0 or 1), but also the phase is change by in each bit transition such as is observed in Figure 10 c). Thus, this phase alternation is completely independent of the information carried in the intensity amplitude. This characteristicalsoimplies,onaverage,azeromeanopticalfieldenvelope[1][5].Asa consequence,payingattentiontotheopticalspectrumillustratedinFigure10a),the strong component in the carrier frequency is suppressed. For this reason, CSRZ becomesasuitablechoiceinordertoimprovethenonlineartolerance[5].

Figure 10. Simulated results for CSRZ at 40 Gb/s: a) Optical Spectrum. b) Pulses with 67% duty cycle. c)Alternatingphase

Figure 10 shows the optical spectrum and the pulse carver for generating a CSRZOOK.Aswecansee,thespectralcomponentatthecarrierfrequencyhasbeen suppressedduetothephasealternationineachbittransition.Sincetheopticalphase is periodic at half the bit rate (90ps) [1], we can observe characterized tones at 20GHz. The following clock tones are spaced at multiples of the bit rate from the firsttoneandtheyaregraduallymorereduced.

11

2.1.2.CorrelativeCodingandPartialResponseFormats
2.1.2.1.Duobinary

Figure11.TransmitterandreceptionstructureDuobinaryinspiredfrom[5]

Duobinary belongs to the correlative coding formats and introduces a correlationbetweenthephaseandthedatainformation.Inthismodulationformata phase change occurs every time there are an odd number of 0 bits between two successive 1 bits [1]. Figure 11 describes the simulation setup for a backtoback duobinary.Thetransmitterusesaprecodedintheelectricalsignalinordertoprevent possible error propagation [1]. This step enables to achieve a strong correlation betweenconsecutivebitsanditcanbeunderstoodliketheinversetransferfunctionof the enconder, modulation and decoder [5]. Thus, the precoder consists in logic exclusiveORoperationasFigure12shows:

Figure12.Precoderforaduobinarymodulation[7]

Theprecodeddataisconvertedintothreelevels(encoded)usingafourthorder BesselelectricalLPFwithabandwidthequalto10GHz(BitRate/4).Thisencodercan bedigitallyimplementedusingadelayandaddfilter[1]. AsisshowninFigure3,theelectricaldrivingsignalshouldhave 2V peakto peak amplitude and the MZM has to be biased in a minimum of its power transfer function(troughpoint)togettheduobinarymodulationusingaMZM.

12

Simulated signal constellation for Duobinary 90 120 0.03 150 0.02 0.01 30 0.04 60

180

210

330

240 270

300

Figure13.Simulatedresultsforduobinarysignalatthebitrateof40Gb/s.a)Duobinarysignal.b)Theoretical signalconstellationc)Simulatedsignalconstellationd)Opticalspectrum

13

Aconventionaldirectdetectionreceiveralreadydescribedinpreviousintensity modulationisusedtodemodulatetheduobinarysignal. Figure 13 depicted the main characteristics of the simulated results such as duobinary signal. From Figure 13 a), where simulated duobinary waveform and its phaseareplotted,wecanobservethatbothamplitudeandphasearemodulated(AM PSK). This modulation enhances a narrower spectrum than conventional intensity modulation signals described previously. In addition, not only the mainlobe in narrowerbutalsosidelobesaresignificantlysuppressed.Asaconsequence,wecould usenarroweropticalfiltersinordertoreducecrosstalkfromneighboringchannels,for exampleinaWDMcommunicationsystem[5][7][8]. By processing the transmitted duobinary signal with external software, in the caseMATLAB,wegetthesimulatedconstellationshowninFigure13.Firstofall,itis necessarytotakethesuitablesamplescorrespondingtoeachpeakamplitudeinorder to calculate their module and phase. Once we plot them, we get the simulated constellation as is illustrated in Figure 13 c). These samples plot themodule and the absolutephaseofthedifferentbits.Itisinterestingtonotethatweobservetwoclear symbols,whichcorrespondtological1,withanexactphaseshiftofaswewishedto obtain. Nonetheless, we can appreciate a module and phase variation in logical 0 symbol comparing theoretical and simulated constellation. The phase shift is generatedduetothephasechangeseverytimewehaveseveralconsecutive0since theyhavetocrossthetroughpointoftheMZMpowertransferfunction.Furthermore, thereiselectricalwaveformdegradationattheoutputoftheduobinaryBPFencoder. A possible alternative to avoid it is the use of a addanddelay instead of the BPF [1][7][8]. Tosumup,themostimportantadvantagesofduobinarycomparedwithOOK modulation formats are its higher tolerance to the chromatic dispersion (CD) and its narrowerbandwidth[1][5][7][8].

2.1.3.DifferentialPhaseModulation
2.1.3.1.BinaryDPSK Differentialphaseshiftkeying(DPSK)modulationformatusesthephaseofthe precedingbitasrelativephasereferenceinordertodemodulatethereceivedsignal. The optical power appears in each bit since the information is carried in the phase [1][4][5]. Figure 14 illustrates the simulated NRZDPSK signal at 40Gb/s before phase demodulation.Itisclearlyshownhowthedatainformationisencodedinthephase transmission instead of in the optical amplitude (OOK). In this subsection, simulated setupsforNRZandRZDPSKwillbedescribed. TransmitterisbasedonthesamestructureemployedinOOKsystemswiththe only difference of a precoding step in the original data sequence. Otherwise, the

14

receiver needs a MachZehnder Delay Interferometer (MZDI) with one bit delay in ordertorecovertheinformationencodedinthephase.

Figure14.SimulatedNRZDPSKsignalat40Gb/s

Asinthecaseofduobinary,DPSKdifferentiallyencodestheoriginaldatabefore modulationinordertoavoiderrorpropagationinthereceiver[1].Thus,theencoderis based again on logic XOR. The MZM employed in the transmitter is driven symmetricallyaroundzerotransmission(thebiaspointisfixedinthroughpoint).This isthereasonwhytheopticalpowerisnotconstantandpresentsfastpowerfallsaswe can see in the simulated DPSK signal. Thus, the electrical precoded signal with 2V peaktopeakamplitudemodulatesthephaseoftheCWlightwavebetween0and.It mustbenotedthatDPSK,likeOOK,canbeimplementedinNRZorRZformatbysimple useofaoneofthepulsecarversdescribedinsubsection2.1.1.2.Hence,thesimulation setupsusedforthismodulationformatsareillustratedFigure15(DPSK)andFigure16 (RZDPSK).

Figure15.TransmitterandreceptionstructureforDPSKinspiredfrom[5]

15


Figure16.TransmitterandreceptionstructureforRZDPSKinspiredfrom[5]

Unlike the already described modulation format such as OOK formats and duobinary,DPSKemploysabalancedreceivertorecovertheinformationcarriedinthe transmittedsignalsincethesignalcannottakeadvantageofthesquarelawpropertyof thephotodiode[1].ThesimulationsetupreceptionisshownFigure15andFigure16. OncetheopticalsignalhaspassedthroughthepreamplifierandtheOBPF,itreaches at the input of a MZDI. The differential delay time is equal to the bit time (1/BitRate=45ps). There are two ports at the output of the MZDI. It generates two kinds of optical interferences known as constructive and destructive interference. Thus,thetransferfunctionsaredefinedas[5]:

yA ( t ) = x ( t ) + x ( t T ) e j (2.1)
yB ( t ) = x ( t ) x ( t T ) e j
(2.2)

where yA, B ( t ) are the constructive and destructive ports respectively, r ( t ) is the sequence at the MZI input, T is the differential delay time and the phase interference between two arms. In this case, this last parameter is set to 0. Two photodiodes with a thermal noise density of 10pA/Hz and a responsivity of 1A/W convert both optical signals to electrical signals. Finally, the receiver signal passes throughthirdorderBesselelectricallowpassfilter(LPF). It is worth mentioning that both ports carry the full information (logically conjugated) as is shown in Figure 17. Therefore, only one of them is sufficient for recovering the original data. This reception is called singleended detection [4][5]. Nevertheless,inthiscaseweuseabalancedreceiverasisillustratedinFigure15and Figure16.Hence,thebalanceddetectionoutputisdefinedas[5]:
y ( t ) = y A ( t ) yB ( t ) =
2 2

1 1 1 1 = x ( t ) + x ( t T ) e j x ( t ) x ( t T ) e j 2 2 2 2

(2.3)

16

Figure17.Simulatedconstructivea)anddestructiveb)componentsattheMZDIoutput

Thankstothisbalancedreceiverweobtaina3dBbenefitcomparedtosingle ended detection [1][4][5][6]. This advantage can be understood from the signal constellation. In Figure 18 are described both theoretical OOK and DPSK signal constellation. In the case of OOK the difference between two symbols is ES with an
2 2 averagepowerequalto 1 20 + 1 2 Es = 1 2 Es .Ontheotherhand,wecanobserve inDPSKconstellationthatbothpointshavethesamesignalenergyandaphaseof0 andphaserespectively.Ifthedistancebetweentwosymbolsisequalto 2 Es then
2 theaveragepowercoincideswithOOKsince 1 2 Es + 1 2 Es = 1 Es . 2 2 2 2 2 2 2

Figure18.TheoreticalsignalconstellationforOOKandbinaryDPSKmodulationfrom[5]

17

Simulated signal constellation for DPSK 90 120 0.03 150 0.02 0.01 30 0.04 60

180

210

330

240 270

300

Figure19.SimulatedDPSKsignalconstellation

Figure 19 plots the signal constellation once the simulated NRZDPSK signal is processedwithMATLAB.Asitwasexpected,wecandifferentiatetwoclearpointswith aphaseshiftof180aswewishedtoobtain.Itisimportanttonotethattheimportant value is the relative phase between consecutive bits and not the absolute phase. It suffers a rotation of 90 if it is compared withthe expected theoretical constellation shownpreviously. 2.1.3.2.DQPSK Differential quadrature phaseshift keying (DQPSK) belongs to multilevel modulation formats which encode more than one bit per symbol. DQPSK transmits foursymbols(twobitspersymbol)associatedtofourdifferentphaseshiftsatahalfof thetotalbitrate[1][5].

Figure20.TransmitterandreceptionstructureforRZDQPSKinspiredfrom[5]

As is depicted in Figure 20, the DQPSK transmitter structure consists of two parallel DPSK modulators with one of the optical signals shifted /2. One splitter is

18

employed in order to divide the light from the laser source into two path of equal intensity [1][5]. Two independent PRBSs with a length of 211 1 generate the informationsequencesforelectricalu(t)andv(t)signals.Theseonespassthrougha precodedstepwhichisdescribedas:

I ( k ) = u ( k ) v ( k ) I ( k 1) + u ( k ) v ( k ) Q ( k 1) + u ( k ) v ( k ) I ( k 1) + u ( k ) v ( k ) Q ( k 1) Q ( k ) = u ( k ) v ( k ) Q ( k 1) + u ( k ) v ( k ) I ( k 1) + u ( k ) v ( k ) I ( k 1) + u ( k ) v ( k ) Q ( k 1)

(2.4)

(2.5)

RZDQPSK modulation can be obtained by means of adding whatever of the pulse carver described in subsection 2.1.1.2 after the MZM. This is the multilevel modulationformatthathasreceivedthemostinterestinthelastyears[5].Moreover, theuseofonepulsecarverenhancetheCDtolerancesinceRZpulsecarvinghelpsto reducetheresidualchirpinthemodulatedsignal[5]. AsfortheDQPSKreception,itconsistsbasicallyintwoDPSKreceiversalthough the phase shift between the two arms of both delay interferometers is set to /4 [4][5]. The remaining receiver parts such as OBPFs, four photodiodes (two for each MZI)andelectricalLPFsharethesamecharacteristicsthaninDPSKreception. Comparing DPSK and DQPSK signal constellations, is obvious that in this case we obtain the double number of points corresponding with both parallel DPSK modulators.However,thesignalbetweenpointsisreducedtothehalfcomparedwith DPSK constellation as is shown in Figure 21. We must point out that using the same symbolbitrate,20Gb/sforthecurrentsimulationsetup,thetotalbitratebecomes thedouble(40Gb/s).

Figure21.SignalconstellationforbinaryDPSKandquaternaryDQPSKmodulation[5]

Figure 22 a) shows the simulated optical intensity and phase waveforms at 40Gb/s. It is clear that the information is carried in the phase with three possible phaseshifts. The narrowband filtering is another of the most important advantages of DQPSKincomparisontobinarymodulationformat.AsisillustratedinFigure22b)the bandwidthoftheDQPSKspectrumisshorterthanDPSKspectrumforatotalbitrateof 40Gb/s.

19


Figure22.OpticalintensityandphasewaveformsforaDQPSKsignalsimulatedat40Gb/s

Figure 23 shows the relation between the constellation diagram and the simulatedDQPSKeyediagram.Twolevelsofintensitydipsexistineachbittransition dependingonthedifferentphaseshiftbetweensymbolswhichcanbe0,/2or.This isthecharacteristicofDQPSKtransmissionwithaparallelmodulator[5].

Figure23.RelationbetweenconstellationandsimulatedDQPSKeyediagraminspiredfrom[5]

Figure 24 illustrates the simulated signal constellation for DQPSK modulation format. We can observe a rotation in the plotted points corresponding to different samples. Although it does not correspond with the wished results we can explain it considering the existence of a slight variation in the absolute phase. Nonetheless, relative phase shift between consecutive bits is kept. This fact enables to success receptionofthetransmittedsignal.

20

Simulated signal constellation for DQPSK 90 120 0.02 0.015 150 0.01 0.005 180 0 30 0.025 60

210

330

240 270

300

Figure24.SimulatedDQPSKsignalconstellation

2.2.Numericalsimulations.Sensitivitycomparison
Thepurposeofthissubsectionistocomparethesensitivityinreceptionforall the modulation formats simulated in previous subsection. The procedure for getting the BER as a function of the received power is described as follows: the amount of power before reception is controlled by a VOA. Thus, an increase of attenuation impliesadecreaseofpowerandadifferentBER.Weplotthedifferentgraphsbyusing external software once all the samples (BER and received power) are calculated and savedthankstothepowermeterandBERestimatormodules. WebeginplottinginFigure25BERasafunctionofthereceivedpowerforall theOOKmodulationformatsdescribedpreviously.Asitwaswishedforshorterpulse widthsweachievebetterreceptionsensitivity.BytakingNRZsensitivityasreference, we can observe a benefit of 1.7dB, 1.2dB and 0.9dB for 33%, 50% and 67% RZ respectively.Itcanbeattributedtothecompressionofthesameamountofenergyin a shorter pulse width. This fact implies a higher peak power and higher distance between1bitand0bit.Obviously,itenablesanimprovementintheestimationof theBER. In Figure 26 and Figure 27 the same graphs are plotted for DPSK and DQPSK modulationformats.Astheyshow,thereisasensitivityimprovementagainduetothe compression of the same amount of energy in a shorterpulse width. For example, a benefit of 1.4 dB (33%RZDPSK), 1.2 dB (50%RZDPSK) and 0.7 dB (67%RZDPSK) are obvervedcomparingwithNRZDPSKsensitivity.

21

NRZ-OOK 33RZ-OOK 50RZ-OOK CSRZ-OOK

-log (BER)

6 7 8 9 10 11 12 -39 -38 -37 -36 -35 -34 -33 -32 -31

Received power (dBm)

Figure25.BERasafunctionofthereceivedpowerforOOKmodulationformats

NRZ-DPSK 33RZ-DPSK 50RZ-DPSK CSRZ-DPSK

-log (BER)

5 6 7 8 9 10 11 12 -42 -41 -40 -39 -38 -37 -36 -35

Received power (dBm)

Figure26.BERasafunctionofthereceivedpowerforDPSKmodulationformats

22

NRZ-DQPSK 33RZ-DQPSK 50RZ-DQPSK 67RZ-DQPSK

- log (BER)

5 6 7 8 9 10 11 12 -41 -40 -39 -38 -37 -36 -35 -34 -33

Received Power (dBm)

Figure27.BERasafunctionofthereceivedpowerforDQPSKmodulationformats

- log (BER)

~3dB
6 7

NRZ-OOK 50RZ-OOK DB NRZ-DPSK 50RZ-DPSK NRZ-DQPSK

~3dB
8 9 10 11 12 -41 -40 -39 -38 -37 -36 -35 -34 -33 -32 -31 -30 -29 -28

Received power (dBm)

Figure28.BERasafunctionofthereceivedpowerfordifferentmodulationformats

23

In Figure 28 BER differents modulation formats are plotted as a function of receivedpower.Aswediscussedinprevioussubsection,thefirstthingweshouldpoint out is 3 dB benefit achieved in balanced receiver in comparison to singleended detection [1][4][5][6]. The sensitivity degradation in NRZDQPSK respect NRZDPSK system (2 dB) can be attributed to two factors. On the one hand, the distance betweensomesymbolsislowerinDQPSKsignalconstellation(Figure23).Ontheother hand, the higher optical signal to noise ratio (OSNR) requierement due to the suboptimal phase demodulation [4][5]. It can be explained as follows: duobinary componentandAMIcomponentareattheconstructiveanddestructiveoutputsofthe MZDI . These components are orthogonal assuming ideal demodulation. Otherwise, forDQPSKmodulationbothconstructiveanddestructiveportsareneitherduobinary orAMImodulated[5]. We can observe also from Figure 28 a backtoback sensitivity difference of approximately 3dB between NRZOOK and duobinary system. In principle, both transmission systems should get the same sensitivity as is reported in [8]. As we discussedpreviouslybothmodulationformatsusethesamedirectdetectionreceiver. Theorically,itimpliesforduobinarythesamesymboldistancethanNRZOOKsincethe 2 squarelaw detection E of the photodiode converts 1 and 1 to 1 and 0 to 0. The reason of the duobinary penalty can be attributed to the electrical waveform degradationattheoutputoftheduobinaryencoderwhichimpliesresidualenergyin the0s.Thealternativeofusingonedelayandaddinsteadofalowpassfilterwould getabetterbacktobacksensitivity[5].

NRZ-OOK NRZ-OOK (5 Km) DB DB (5 Km) DB (12.5 Km)

-log (BER)

5 6 7 8 9 10 11 12 -36 -35 -34 -33 -32 -31 -30 -29 -28 -27

Received power (dBm)

Figure 29. BER as a function of the received power for NRZOOK and duobinary modulations and differentfiberlength

24

 Finally, we demonstrated in Figure 29 how duobinary modulation is characterized by a higher tolerance to the chromatic dispersion. In this case, the simulation system consists in a transmitter and a receiver connected through a fiber instead of back to back structures used until now. In these simulations we employ a NRZ and DB transmitter, Higher NonLinear Fiber (HNLF) with a dispersion of

16x106 ps / nm/ km and direct detection receiver. We can observe how for 5 Km
length transmission intensity modulation presents a 4.5 dB sensitivity degradation comparedwithbacktobackOOKsystem.WecanobserveinFigure29anunexpected behaviorforduobinarymodulationtransmissionsinceaBERimprovementof 1dBin comparison with back to back system is achieved after 5 Km transmission. The eye diagram shown in Figure 30 presents a good eyeopening. So, we can attribute this benefittothefactthatthedispersioninthefiberforacertainlengthhasagoodeffect on duobinary transmitted signal. However, for a fiber length of 12.5 Km the signal reaches to the receiver noticeably distorted as is illustrated in Figure 30, but with a highertolerancetoCDsincelessthan0.5dBisobserved

Figure 30. Duobinary eye diagram for back to back, 5 Km transmission and 12.5 Km transmission simulatedat40Gb/sandusingaHNLFwithadispersionof16x106ps/nm/km

2.3.Summary
Advancedmodulationformatsareclassifieddependingthewaytheycarrythe information. Thus, we difference among intensity modulation formats, correlative coding and partialresponse formats and differential phase modulations. NRZOOK transmission is still widely employed in current optical networks. Nonetheless, duobinary modulation format which modulate both intensity and phase presents a better behavior in front of chromatic dispersion. DPSK transmission has attracted a speciallyinterestlastyearsduetoitshighspectralefficiency.Furthermore,balanced receptionprovidesasensitivitybenefitof3dBincomparisonwithdirectdetection.All this main modulation formats have been numerically simulated at 40Gb/s using VPI TransmissionMakerWDM.

25

Chapter 3 AllOptical Modulation FormatsConversion

Recently, the interest in alloptical formats conversion have increased noticeablysinceallopticalprocessingisconsideredoneofthemostimportantaspects in future optical networks. Thus, electrictooptical and opticaltoelectric converters willnotbenecessariesandconnectionbetweendifferentnetworkswillbecomemore cost effective [12]. As we mentioned in previous chapter, a considerable number of advancedmodulationformatsschemeshavebeenreportedinthepastfewyears.They maybeselectivelyemployedinfutureopticalnetworksdependingonthesize,bitrate andthesystemsettings[7][26]. The purpose of this Chapter is to describe the techniques employed in all opticalmodulationformatconversion.Tomybestknowledge,matrixfrom/toisfilled upwithallpossibleconversionsreporteduptonowinsubsection3.2.Subsequently, mostofthemareexplainedanddiscussedingreatdetail. OpticalNetwork
Ultralonghaulsystems Longhaulsystems OTDMlonghaulsystems WAN MAN DWDMaccessnetwork

Modulationformat
PSK[13][26][27][38] PSK[7],CSRZ[13],AMI[13] RZ[34] RZDQPSK[19],PSK[7] NRZ[19][34][38][7][23],Duobinary[13] NRZ[34]

Table2.Proposedmodulationformatsforfutureopticalnetworks

Table 2 try to present an overview of the different advanced modulation formatsproposedforfutureopticalnetworks.Aswediscussedintheintroductionof thisthesis,itisdifficulttomakeanexactclassificationofacertainrealnetworkdueto therearealotoffactorstoconsider.Obviously,themaincharacteristicsarethesize andthedistancesthattheyhavetocover,butalsoitisimportanttopayattentionto other features. For example bandwidth requirement and high data rate for future backbonenetworksorthewayoftransportingdatafrom/tousersinaccessnetworks [7][34]. So, suitable choices of modulation format for each network and all optical convertersbetweenthemwillenhancecapacity,costeffectivenessandrobustnessof transmissionlinks[5]. In the next section we present an overview of the different techniques employedinallopticalmodulationformatconversion.

26

3.1.GeneralOverview.Employedtechniques
3.1.1.SemiconductorOpticalAmplifier
The principal function of Semiconductor Optical Amplifiers (SOAs) in optical communication systems is amplifier an optical signal to increase the transmitted power,compensatelossesorimprovethesensitivityinreception.Nevertheless,SOAs canbealsoemployedtoperformsomeusefulfunctionsinopticalnetworks[40].Inthis thesis there are described different format conversion techniques based on a single saturated SOA, on optical switches, SOA MachZehnder Interferometer or on SOA nonlinearities such as Cross Gain Modulation (XGM), Cross Phase Modulation (XPM), Self Phase Modulation (SPM) and FourWave Mixing (FWM). The most important disadvantageofusingSOAinopticalapplicationslikemodulationformatconversionis the limitation of the operation speed due to recovery time, especially at higher operationbitrates[23][29][35][37]. CROSSGAINMODULATION

CarrierdensitychangeinaSOAaffectsalltheinputsignalssincematerialgain spectrum is homogenously broadened [40]. Thus, a strong signal at one wavelength ( ) can impose a conversion on a weak signal at other different wavelength ( 2 ) 1 when they are launched at the same time into a SOA. This mechanism is known as XGM.ByusinganopticalBPFitispossibletokeeptheconvertedsignal( 2 )attheSOA output. This configuration is illustrated in Figure 31 and implies a wavelength conversion:

Figure 31. Principle of XGM inspired from [40]. In this configuration the SOA acts as a wavelength converter

CROSSPHASEMODULATION

In SOA there is dependence between refractive index of active region and carrierdensity[40].Forthesereason,iftwosignalsareinjectedsimultaneouslyintoa SOA XPM will affect both signals as well. This nonlinearity effect is used in Mach Zehnder Interferometer (MZI) structures where constructive and destructive

27

interferencesenabletoconvertthephasechangesinintensitychanges.Dependingon theMZIconfigurationdifferentformatconversioncanbeachieved[19][20][40]. FOURWAVEMIXING FWM is a coherent nonlinear process which generates two sidebands. These new sidebands can become useful for multichannel applications. For example, a possible FWM generation occurs when a strong signal (pump) and a weak signal at o and o respectivelyarelaunchedintoSOAwiththesamepolarization.Anew spectral component are generated at o + . It corresponds with the conjugate of

o as shows Figure 32. Note that new spectral componentsarealsogeneratedat o 2 .

signal which is centered at

Figure32.From[40],FWMinaSOA

SOALOOPMIRROR SOAscanbeusedasopticalswitchesaswell.Inordertoenhancetheswitching time the amplifier can be placed in nonlinear loop mirror [11][17][40]. This configurationisemployedinmanyopticalapplicationssuchasallopticalregeneration, demultiplexing, switching and format converter [17]. Figure 33 described the optical switchprincipalofoperation.Wecanextendittounderstandwhateverofthecurrent SOA loop mirror applications. They are based on the phase difference between clockwise signal (called data pulse 1 in Figure 33) and counterclockwise signal (data pulse2).Theinputsignalsenterintothefiberloopthroughdifferentcouplers.

Figure33.From[40],schemeofanopticalswitchbasedSOAloopmirror

28

TheSOAisplacedinthefiberloopmirrorasymmetricallywithacertainlength displacement from the midpoint of the fiber. Obviously, it implies a delay time betweenonesignalanditsreplicaequalto Toffset 2 where Toffset isdefinedas[11][17]:

Toffset = 2

x c neff

(3.1)

where x isthelengthdisplacement, c isthespeedofthelightinvacuumand neff is theeffectiverefractiveindexinthefiber. Finally, at the output of the loop mirror the clockwise and counter clockwise signalsarerecombinedforobtainingtheconvertedsignal.

3.1.2.MachZehnderDelayInterferometer
MZDI is typically used in balanced receivers in order to demodulate the differentialphasebetweeneachbitanditssuccessor.Chapter2describesindetailthis structure and its main function. Thus, MZDI is able to extract the phase information from the DPSK input signal and convert it in intensity changes at both outputs. It is important to remark that fully information is present in both constructive and destructiveports (but inverted) [1]. It results in centrallobe with a spectralwidth of 1/T,whereTisthebittime,intheconstructiveport.Ontheotherhand,intheoptical spectrum of the destructive port the central carrier frequency is suppressed which impliesphasechangebetweeneachconsecutive1bit. The transfer function of a MZDI is defined in equations (2.1) and (2.2). Thus, and as it was brought forward in subsection 2.1.3.1 and numerically simulated at 40Gb/s, duobinary and AMI can be obtained from DPSK at the constructive and destructiveportsoftheMZDI,respectively(Figure17). AnotherpossibleformatconversionfromCSRZtoRZDuobinaryisreportedin [13]usingasimpleMZDI.

3.1.3.RingResonatorOpticalwaveguide
Acoupledringresonatoropticalwaveguide(CROW)onsiliconchipconsistofa chain of N coupled microring resonators which define the order. It has N microrings and N+1 coupling regions. For example, Figure 34 illustrated a 3Order CROW where ai ,0 and a i' ,0 aretheinputfieldsineachcouplingregionand bi,0 and bi',0 aretheoutput fields.

29


Figure34.3OrderCROWinspiredfrom[28]

Foraspecificcouplingregionoutputfieldscanbewrittenasafunctionofinput fieldsas:
bi ,0 ti b ' = i ,0 j i j i ai ,0 , x = 0,1 and i {0,1,..., N } ti ai' ,0

(3.2)

2 2 wheretheselfcouplingandcrosscouplingarerelatedby ti + i = 1[28].

As it will be shown in Figure 51, the CROW has a box like notch around its resonancewavelengthoncethecouplingcoefficientshavebeenchosen.Byincreasing the order the shape become more boxlike. It should be mentioned that in a real CROW it is not easy to control all these parameters and it induces a not desired modificationofthenotchshape[28].Themodulationformatconverterwillbebased onthedifferentinducedshiftsintheresonancewavelengths.

3.1.4.NarrowBandSiliconMicroringResonator
Several signal processes devices and circuits based on silicon microresonators suchasadddropfilters,carrierdispersionbasedmodulatorsorwavelengthconverters have been reported over the last years [29]. But also a proposed alloptical onchip NRZtoPseudoreturntozero(PRZ)formatconversionhasbeendescribedin[29]using narrowbandsiliconmicroringresonatorbasednotchfilters.Themainfeaturesofthis technique are its compatibility with complementary metal oxide semiconductor (CMOS)microelectronicprocesses,narrowbandwavelength,micrometerscaledevice footprintandaccessibilitywithintegratedwirewaveguides[29].

30


Figure 35. From [29], a) Optical micrograph of an electrically tunable racetrackshaped microring resonatorbasednotchfilterwithlaterallyintegratedpindiodeonaSOIsubstrateb)Transmission spectraofthenotchfilterwithvariousforwardbiasvoltages.

Figure 35 a) illustrates an optical micrograph of an electrically tunable racetrackshapedmicroringresonatornotchfilter.AswecanobservefromFigure35b) the function of the notch filter will be suppress a strong carrier component in the spectrum of the input signal with the null transmission. Te quality of the conversion dependsonthenulldepth.AsFigure35b)illustratesthefowardvoltagebiasenableto achievedifferenttransmissionspectra.

3.1.5.SpectralLinebyLinepulseshaping

Figure36.From[34],experimentalsetupofonelinebylinepulsesharper

Figure36showsthestructureofthelinebylinepulseshapingimplementedby a fiber coupled pulse sharper with a reflective geometry, a collimator, a telescope combination, a lens, a retroreflecting mirror and a circulator. This scheme is well describedin[34]. LinebylinepulsesharpergeneratesaRZpulsetrainatitsoutput.Theprinciple of operation consists on controlling the pulse width at the output with a shaped spectrum.Itispossibletocontroldependingonthenumberofspectrallineskeptat theoutputofharmonicallymodelockedfiberlaser.Forexample,inFigure37a)two spectral lines separated by 10GHz laser repetition rate enables to generate a cosine

31

functionwith50ps.Therestofspectrallinesgeneratedaresuppressedthankstothe highresolutionlinebylinesharper[34].Thus,itispossibletoachievewidthtunability aswecanobservefromFigure37.

Figure 37. From [34], linebyline controlled spectra and waveforms. The spectra are controlled to havefrom2lines(a)toalllines(h)achievingawidthtenabilityrangeof350ps[34]

Linebyline pulse shaping technique has been experimentally reported to performallopticalRZtoNRZformatconversion[34].

3.1.6.Highlynonlinearfiber
Highly nonlinear fibers (HNLF) have become in the past few years a promise technique in order to achieve nonlinear applications. Alloptical modulation format conversions using these nonlinearities have been studied in several reports and they are described in great detail in following subsections. Currently, XPM is the most studiednonlinearityinthefieldofmodulationformatconversion,speciallyfromOOK toPSK[26][27].TheChapter4trytostudyingreatdetailRZOOKtoRZDPSKformat conversion based on XPM. In addition, some requirements are discussed with numerical simulations. Nontransparency of FWM and its application as format converterarediscussedin[33]aswell. Ingeneral,theprincipleofoperationisbasedononeHNLFwheretwoormore signals are launched synchronously in order to obtain the desired all optical format conversionfrom.ThecalledcontrolsignalsinduceXPMontheopticalphaseoftheRZ clocksignal.Thus,theinformation carriedbysomeofthe signalsintheiramplitudes (OOK)canbemodulatedonthephaseofotherdifferentsignal. ThepropagationofthepulsesalongaHNLFfiberfollowsabehaviordescribed bythenonlinearSchrdingerequation(NLSE)whichequationisexpressedas[26]: i E 2 2 E 2 + s E E = i E 2 z 2 t (3.3)

32

wherezisthepropagationdistance,tthetimemovingwiththegroupvelocity,E(z,t) the complex envelope of the electrical field, 2 de group velocity dispersion, s the nonlinearityand thefiberloss.Inthistypeofformatmodulationwheremorethan onesignalislaunchedintothefiber,groupvelocitymismatchplaysanimportantrole in XPM interaction [33] since it induces walkoff (lost of the synchronism) between signalsalongthefiber.WalkofftogetherFWMinducedegradationontheconverted signalsuchaspowerdifferencesandunstablephasemodulationdependingonthebit pattern[26] CROSSPHASEMODULATION All materials behave nonlinearly intensities at high intensities and their refractive index increases with intensity. This nonlinear susceptibility has his physical originintheanharmonicresponseofelectronstoopticalfields[3].Inordertoinclude thenonlineareffectwecanwritethecoreandcladdingindicesofasilicafiberas[3]:
n i' = n i + n 2 P A eff , i = 1,2

(3.4)

where n 2 is the nonlinearindex coefficient, P is the optical power, and Aeff is the effective mode area introduced earlier. Even though the nonlinear part of the refractive index is quite small, modern lightwave systems are affected considerably duetothelongfiberlength.Specifically,thiseffectleadstothephenomenaofselfand crossphasemodulations[3]. In this case, we are going to focus in nonlinear phenomenon known as cross phasemodulation(XPM).Itoccursinsystemswheretwoomoreopticalsignalsfrom differentchannelsaretransmittedsimultaneouslyintoanopticalfiber[3][26][27].We can define XPM like a nonlinear effect where the optical intensity of one beam influencesthephasechangeofanotherbeam.Thus,thephaseforaspecificchannel becomes[3]:

iNL = Leff Pi + 2 Pj

j i

(3.5)

whereLeffistheeffectivelengthofthefiber.Itshouldbeclearfromlastequationthe dependencenotonlyonthepoweroftheownchannel(Pi)butalsoonthepowerof other channels (Pj). Compared with the corresponding equation for selfphase modulation(SPM)describedin[3]thereisanadditionalfactorof2.Thisfactormeans that XPM has, for the same amount of power, the double of influence in the total phase shift than SPM [3]. It implies that the effect of SPM can be practically not considered.

33

3.1.7.PPLNwaveguide
During the past few years, difference nonlinear applications by using second order nonlinearities and their cascading in periodically poled lithium niobate (PPLN) waveguides have attracted an increased attention [5][31]. Its cascaded second order nonlinearities have become promising techniques in the field of modulation format conversion due to their high nonlinear coefficient, ultrafast response, and negligible spontaneous noise[31]. On the contrary, PPLN has negligible third order susceptibility[5]. DependingonthenumberofsignalsarelaunchedinthePPLNwaveguideand which one is set at the secondharmonic generation (SHG) quasiphasematching (QPM) wavelength to take part of the SHG, we can differentiate between several cascaded second order processes [30][31][32]. In the following paragraphs it is described and theoretically analyzed two commonly cascaded second order nonlinearities in PPLN employed in telecommunications application. They are called cascaded secondharmonic generation and differencefrequency generation (cSHG/DFG)andcascadedsumanddifferencefrequencygeneration(cSFG/DFG)[2]. Figure 38 shows two kinds of cSHG/DFG processes. We must point out that there are two signals (the signal to convert and the pump) which are launched simultaneously atthePPLNinput(Figure38a)).Wegetanidlersignalattheoutputas aconsequenceofthenonlinearityeffectsinthewaveguide.Oneofthemostimportant advantagesisthatthecascadedSHG/DFGprocessispracticallyinstantaneous.Forthis reasonthistechniquecanbeconsidertransparenttobitrateandhashighconversion efficiency[5]. In the first scheme illustrated Figure 38 b), the pump is set to the SHG QPM wavelength. Thus, it participates in the SHG process converting itself to a second harmonicwavebySHGprocess.Then,thisconvertedwaveinteractswiththesignalin ordertoyieldanidlerwavebythesubsequentDFGprocess. AsFigure38c)shows,thesignalcanbesetattheSHGQPMwavelengthaswell. It is converted to a second harmonic wave by SHG process which simultaneously is mixedwiththepumptogenerateanidlerwavethroughtheDFGprocess. Itisworthnotingthatforthefirstdescribedcase,wherethepumpissettothe SHGQPMwavelength,itcannotachievetunableoperationforfixedinputsignaldueto theverysmallpumpwavelengthtolerance[30].Inordertoenhancethislimitationof tunability,wecanadjusttheinputsignalattheSHGQPMwavelengthinsteadofthe pumpsignal.Therefore,bychangingthepumpwavelength,theconvertedidlerwave willbeabletobetuned[30].

34


Figure 38. From [30], a) Schematic diagram for cSHG/DFG with the pump and the input signal launchedintothePPLNwaveguideb)OperationprinciplesofcSHG/DFGwiththepumpsetattheSHG QPM wavelength c) Operation principles of cSHG/DFG with the signal set at the SHG QPM wavelength

ThecSFG/DFGprocessesareillustratedinFigure39andneedsanextracontrol wave.ThesignalandthepumpgenerateasumfrequencywavebytheSFGprocess.At the same time, the sumfrequency wave and the control wave produce a converted idlerwavethankstoDFGprocess.Inthiscase,itshouldbementionedthatbothinput signal and output idler can be flexibly tuned by means of changing the pump and controlwiththecSFG/DFGprocesses[30].

Figure 39. From [30], a) Schematic diagram for cSFG/DFG with the pump, control and input signal launchedintothePPLNwaveguideb)OperationprincipleofcSFG/DFG.SFGprocessoccursbetween thesignalandthepump.

Inaddition,asFigure40illustrates,PPLNtechniquealsoprovidesmulticasting applications increasing the number of pumps signals since second harmonic wave

35

generated by SHG process is mixed simultaneously with the pumps to enable the multichannelidlers[21].

Figure40.From[31]multicastingoperationforcSHGDFG.Inthiscase,PPLNisemployedinCSRZtoRZ format conversion. By increasing the number of CW pumps it is possible to generate multichannel RZ idlers

Once we have described the different principle of operation for commonly cascaded second order nonlinearities in a PPLN waveguide, it is interesting to find a simple expression which allows understanding the format conversions proposed in subsections 3.3.2 and 3.7.2. Thus, under the nondepletion approximation, the normalizedcomplexamplitudeoftheidler(Ai)forcSHG/DFGprocesswiththepump setattheSHGQPMwavelengthisgivenby[30]: 1 2 * Ai iSH SHG DFG AP AS 2 (3.6) L cos ( L ) 1 sin ( L ) L sin ( L ) + cos ( L ) +i 2 2 whereAPandASarethenormalizedcomplexamplitudesoftheinputpumpandsignal, SHG and DFG aretheSHGandDFGcouplingcoefficients,iandSHaretheangular frequencies of the idler and the second harmonic wave, L is the length of the PPLN waveguideandreferstothephaseupsetfortheDFGprocess. From equation (3.6) it is possible to obtain a simpler approximation for the normalizedelectricfieldoftheidlerandopticalpowerexpressedas[30][32]:
2 * Ai AP AS

(3.7) (3.8)

P P2 P i P S

It is worth mentioning that two important useful conclusion for format conversions can be extracted from these equations. We can easily deduce that the normalizedcomplexamplitudeoftheidlerisproportionaltotheproductoftheinput signalandthesquareofthepump.Moreover,idlerremovethephaseinformationof thepumpsignalduetothesquarerelationshipbutkeepthephaseinformationfrom theinputsignal.

36

In the case of cSHG/DFG process illustrates in Figure 38 c) the normalized electricfieldoftheidlerandopticalpowercanbeexpressedas:

2 * Ai AS AP

(3.9) (3.10)

P P2 P i S P

Following the same reasoning, in this case the normalizedcomplex amplitude oftheidlerisproportionaltotheproductofthesquareofinputsignalandthepump. Furthermore, idler preserves the phase information of the pump signal due to the linearrelationshipandremovesthephaseinformationfromtheinputsignal. Finally, in order to analyze the cSFG/DFG we can also derive the following expressions:
* Ai AS AP AC

(3.11) (3.12)

P PS PP P i C

For the current cascaded second order process idler signal can preserve the phaseinformationduetotheabsentofsquarerelationshipsinequation(3.11).

3.2.Stateoftheart
In this section matrix from/to is filled up with the main modulation format convertersreporteduntilnow.Thus,wecanobserveageneraloverviewaboutwhich conversionshaveattractedmoreattentionduringthepastfewyears.Obviously,more methods about data conversion (OOK modulation formats) have been reported in comparison with OOKPSK, OOKCorrelative Coding or PSKPSK since the most of opticalcommunicationsystemsprimarilyemployedconventionalOOK.Otherwise,new promisingtechnologieslikeHNLForPPLNaregraduallyattractingmoreinterestdueto its fast response and the necessity of achieving conversion from traditional OOK to newadvancedmodulationformatsorviceversa.

37

TO NRZOOK FROM NRZ OOK

RZOOK

CSRZOOK

Duobinary NRZ DPSK

Single saturated SOA(8Gb/s) [21] SOABased Nonlinear Polarization Switch (10Gb/s)[38]

RZDPSK

CSRZDPSK

RZDQPSK

RZ OOK

CSRZ OOK NRZ DPSK RZ DPSK

XPMinaMachZehnder photonicintegratedcircuit(PIC) [9] SOALoopMirror[11](10Gb/s) InjectionLockedFabryProt LaserDiodes[12] SOA/fiberBragggrating (10Gb/s)[15] MZIwavelengthconverter[10] SpectralLinebyLineControl (10Gb/s)[34] Delayinterferometerand narrowbandfilter (20Gb/s)[37]

Walkoffbalancednonlinearfiber loopmirror[10] SOALoopMirror[11] MachZehnderdelayinterferometer (40Gb/s)[14] PolarizationInsensitivespectral filteringofaXPMbroadenedsignal spectrum(10Gb/s)[16] XGMinaSOA[22] FWMinsingleSOA(40Gb/s)[23] XPMinasingleSOA[24] Coupledringresonatoroptical waveguide(40Gb/s)[28] NarrowBandSiliconMicroring ResonatorBasedNotchFilters (40Gb/s)[29] PPLNwaveguide(40Gb/s)[30]

SOAloopmirror (10Gb/s)[17] XPMandXGMina SOA(40Gb/s)[25] PPLNwaveguide (40Gb/s)[30][32]

SOAMZI wavelength Converter[19] Nonlinearity (XPM)inoptical fiber(40Gb/s) [26][27]

ParallelSOAMZI OOK/BPSK converters (10.7Gsymbol/s) [20] Nonlinearity (XPM)inoptical fiber (40Gsymbol/s) [26][27]

SOAloopmirror (10Gb/s)[17] PPLNwaveguide (40Gb/s)[30][32]

XPMin singleSOA (10Gb/s)[35]

Nonlinearity (XPM)inoptical fiber[27]

FWMinHNLF(40Gb/s)[33] PPLNwaveguide(40Gb/s)[36] SOAloopmirror(10Gb/s)[17]

MZDI (10Gb/s)[13] MZDI[13]

MZDI[13]

PPLN waveguide (40Gb/s) [30][32]

PPLNwaveguide (40Gb/s)[30][32]

Table3.Summaryofallopticalformatconversionstudieduptonow

3.3.OOKtoOOKformatconversion
3.3.1.Motivation
Thepurposeofthissubsectionistoarguethemotivationswhichleadtostudy possible conversions between OOK formats. Most of them are based in the own featuresofeachmodulationformatdiscussedinChapter2. All optical format conversion between NRZ and RZ has been reported using many methods including nonlinear optical loop mirrors, SOAs or PPLN among other techniques. The RZ format present more advantages for high speed OTDM transmission due to its robustness against the nonlinear effect in spite of the dispersioninducedeffect[11].Inaddition,RZcanreachtolongerdistancesduetoits highertolerancetomanyfiberstransmissionimpairments[34].Itisduetheenergyin RZsignalsismorecompressedinthecenterofeachbitslot,andthusmoredifferential groupdelay(DGD)isrequiredinordertocauseintersymbolinterference(ISI)[12].On the other hand NRZ is characterized by a narrower spectrum as it was numerically simulatedinsubsection2.1.1.1.Therefore,formatconversionbetweenNRZtoRZdata formats can be used as an optical gateway from the fast OTDM networks to slower DWDMaccessnetworkswhereismostimportantthespectralefficiency[12]. CSRZ modulation format has a good tolerance to some linear and nonlinear impairments generated along a optical fiber and is widely used in new reported 40Gb/s experiment systems [13]. Furthermore, CSRZ combined with narrow optical filtering enables high spectral efficiency without significant degradation of the pulse shape[13].Forthesereasons,CSRZbecomeapromisingformatforopticalnetworks wherethereisanimportantbandwidthdemandsandhighspectraldensityisrequired [25]. Thus, some robust conversion methods from/to CSRZ to NRZ or RZ are also describedinthefollowingsubsection. Finally, NRZ to PseudoReturn to Zero (PRZ) is described as well. This format conversionisessentialforclockrecoveryfromNRZformat,whichinprinciplecarries noclockcomponent.[29]

39

3.3.2.Modulationformatconversions
RZTONRZUSINGINJECTIONLOCKEDLASERDIODES Figure 41 shows the experimental setup of the RZ to NRZ modulation format conversion using InjectionLocked FabryPerot Laser Diodes (FPLD). This scheme not onlyprovidesaformatconversionbutalsoactsaswavelengthconverter.

Figure41.From[12],experimentalsetupoftheRZtoNRZformatconversion.GSDFB:Gainswitched distributed feedback laser. CWDFB: Continuouswave distributed feedback laser. BF: Birefringent fiber. PC: Polarization controller. EDFA: Erbiumdoped fiber amplifier. MOD: Modulator. VOA: Variableopticalattenuator.PD:Photodiode.BERT:Biterrorratetestset

Hereby, most of the proposed format conversions follow a similar simple principle of operation although they employed different techniques to obtain it. In general,twosignalsareinjectedintotheconverter.Oneofthemismodulatedinthe originalmodulationformat(inthiscaseRZ)andcarriestheinformation.Thissignalis usuallycalledcontrolsignal.Thefunctionoftheconverterismodulatethisinformation in the other signal (also called control) with a different modulation format using whatever of the techniques described in this thesis. It is important to note that in a realcase,controlsignalcomesfromatransmissionlineandprobesignalisgenerated intheownconverter.Inthissubsection,RZsignalactsascontrolsignalandaCWacts asprobesignal. Thus, a RZ control signal ( c ) is modulated at 10Gb/s from a GSDFB. One EDFA amplifiers the pulses train before passing through an OBPF which removes the outofbandamplifiedspontaneousemissionnoiseofthepreviousamplificationstep. AsFigure41illustrates,theRZsignalissplitinaBFintotwoorthogonalpolarizations with a delay equal to half bit. For this reason, the bit rate of the RZ pulses train becomestwicehigher(20Gb/s).Thissignal(bothorthogonalcomponents)iscoupled withaCWprobe ( p ) intotheFPLD.

40

Initially,theFPLDisinjectionlockedbytheprobesignalgeneratedintheCW DFB.TheFPLDhasacentraloutputwavelengthandcertainlongitudinalmodespacing [12]. Thus, if one of them lies in the injectionlocking range of the control signal wavelength ( c ) , FDLD becomes injectionlocked to [12]. This fact implies it c operateswithalowercarrierdensitycomparedwiththeinitialoneandhasahigher refractiveindex.Thus,itispossibletoachievetherequiredwavelengthchangeinthe longitudinalmodesoftheFPLD[12].Hence,thecontrolsignal canmodulatedthe c outputat p . AnopticalBPFfiltersouttheconvertedsignalat p beforereception. Figure 42 shows the eye diagrams and optical spectrum of the original and converted signal. It is worth noting that RZ spectrum is broader than NRZ spectrum. Experimentally,theconvertedNRZsignalpresentsadegradationof1.33dBcompared withabacktobackRZsignal[12].

Figure 42. From [12], a) Eye diagram of 10 Gb/s back to back RZ pump signal b) 10Gb/s modulationformatconvertedNRZsignalc)Opticalspectrumforthedualwavelengthinjection lockedFPLD

RZTONRZUSINGSOALOOPMIRROR Figure 43 depicts a proposed converter between NRZ and RZ modulation formats. For RZtoNRZ format conversion two signals are launched into SOAloop mirror. A continuous wave beam ( CW ), which follows two propagation ways with opposite direction in the loop, is injected into the SOAloopmirror through input B. TheoriginalRZsignalentersintoSOAloopmirrorthroughacoupler(inputA).

41


Figure43.From[11],SOAloopmirrorconfigurationforRZtoNRZandNRZtoRZformatconversion

The transmitted intensity for this configuration is given in [11] and can be writtenas:
I out = 1 I in Gcw ( t ) + Gccw ( t ) 2 Gcw ( t ) Gccw ( t ) cos diff ( t ) 4

(3.13)

where I in and I out aretheopticalintensityoftheoriginalsignal x ( t ) andconverted signal y ( t ) respectively. Gcw ( t ) and Gccw ( t ) are the SOA gain for the two possible propagationofthecontinuouswavebeaminSOAloopmirror.Finally, diff ( t ) defines the difference between cw ( t ) and ccw ( t ) signals as cw ccw due to asymmetric displacement. InordertounderstandeasilythefunctionSOAloopmirrorwecanconsiderthe effectsofXGMandXPMseparately[11].Hence,byconsideringonlyXGMeffectinSOA andassumingconstantsthedifferentphases,equation(3.13)becomes:
I out = 1 I in Gcw ( t ) + Gccw ( t ) 2 Gcw ( t ) Gccw ( t ) 4 1 = I in Gcw ( t ) + Gcw ( t + Toffset ) 2 Gcw ( t ) Gcw ( t + Toffset ) 4

(3.14)

whereitisimportanttonotethatbothSOAgainsarerelatedby Toffset thatmeansthe differentarrivaltimeofcwandccwbeams.Anyway,phasevariationispresentingain shapes of both beams [11]. In addition, it is necessary to satisfy the condition ToffsetT<<g which means that the gain is recovered when the ccw beam arrives to theSOA.InpreviousconditionTisthetime,g thegainrecoverytimeand Toffset isfixed dependingontheformatconversionwewanttoobtain[11].

42


Figure44.From [11], operationalprinciple of RZtoNRZ format conversion.The signalsx(t) andy(t) correspondwiththeelectricfieldoftheoriginalandconvertedsignal

RZtoNRZmodulationformatconversionisachievedfixing Toffset tobelessthan T.AsFigure44shown, x ( t ) isaRZsignalwhichinducesaphasechangeinSOAactive medium at its rising time. As a consequence, Gcw ( t ) and G ccw ( t + Toffset ) decrease in ordertobalancetherelativephasechanges.Thus,weobtainaNRZshapedwithred chirped components at both the rising and falling edge when both signals are combinedattheoutputoftheSOAloopmirror[11]. Ithasbeenreportedinatransmissionexperimentat10Gb/s[11]howtheeye diagramsoftheconvertedNRZsignaloverdifferentfiberlengthsisbetterthanthose ofNRZconventionallygeneratedusingaMZMoverthesamefiberlengths.Figure45 shows the experimental results. The reason of this improvement is that the red chirped frequencycomponents travel slower than the unchirped components due to thepulsetraveltroughthefiberintheanomalousdispersionregion[11].

Figure45.From[11],a)EyediagramsfortheproposedNRZconvertedsignalb)Eyediagramsforthe conventionalNRZsignalgeneratedbyMZMoverdifferentfiberlength

43

NRZTORZUSINGSOALOOPMIRROR For this modulation format conversion Toffset is set to T. In this case, the SOA loopmirror output y ( t ) can be understood as an exclusive OR logic gate. Thus,

y ( t ) becomeszerowhen Gcw ( t ) isequalto Gccw ( t ) andifbotharedifferentfromeach

other there is a certain amount of power in the power. Figure 46 depicted the waveforms of the signals which take part in the proposed NRZtoRZ format conversion.

Figure46.From[11],operationalprincipleofNRZtoRZformatconversion

The RZ converted signal has an alternate phase transition among consecutive pulses. This alternation is typical in modified duobinary signals [11]. In this case, experimental results also show improved fiber transmission compared with conventional RZ generated by a MZM. Moreover, it is proved that this transmission improvementisduetotheduobinarycodingeffectoftheSOAloopmirror.Finallyand experimentally, the duty cycle of the converted RZ signal can be modified adjusting Toffset [11]. NRZORRZTOCSRZUSINGSOALOOPMIRROR

Figure47.From[17],NRZtoCSRZschemeoftheconverterforNRZ/RZtoCSRZ

44

Figure 47 shows the schematic diagram of the NRZtoCSRZ converter. Toffset is fixed to one bit duration again. Two signals are injected into the SOAloopmirror in order to control the behavior of the SOA. A periodic control signal (optical clock) changestherefractiveindexintheactiveregion[17]andacontinuouswavelighthas the function of saturating the SOA. Moreover, it controls the depth of the phase modulationgeneratedbytheopticalclock.Thetunabledelaylineisusedtoadjustthe clockwiththetimeslotoftheoriginalNRZsignal. As is illustrated in Figure 47 control signal induced an inverse and periodical phasemodulationin cw ( t ) and ccw ( t ) signals.Thus,aphasechangebetween0and is added to consecutive bits in the input signal. This is one of the most important characteristicofCSRZsignals.Forthisreason,aNRZorRZsignalisconvertedtoaCSRZ signal.Wemustpointoutthatthemodulationformatconversionisalsopossiblefrom CSRZtoRZduetothesameprincipleofoperation.

Figure48From[17],a)Experimentalresultsforthedifferentmodulationformatconversionat10Gb/s b)BERmeasurementforthedemonstratedallopticalNRZtoCSRZ,RZtoCSRZ,andCSRZtoRZformat conversions

In Figure 48 are shown the eye diagram and spectrum experimentally got at 10Gb/s.WecanobservehowthealternationintheCSRZphasesuppressesthecentral carrier.Apowerpenaltyof0.8dB,+0.6dBand0.5dBareobtainedforNRZtoCSRZ, RZtoCSRZandCSRZtoRZrespectively[17].

45

RZTONRZUSINGTUNABLEALLFIBERDELAYINTERFEROMETER Format conversion from different duty cycles RZ to NRZ has been reported in[37]andexperimentallydemonstratedat20Gb/sbasedonallpassivesetup.

Figure49.From[37],experimentalsetupandprincipleofoperationforformatconversionfromRZto NRZ

Figure 49 shows the experimental set up used in [37] and the principle of operation for this format conversion. A RZ generator with different duty cycles transmits the original signal to the format converter. This consists of two filtering steps:aDIfilterandanopticalBPF. The tunable alloptical DI acts a comb filter (series of regularlyspaced spikes) suppressing some frequency components. In [37] and for 20 Gb/s RZ signals the frequencyspacinginthefilterisequalto40Gb/s.WhentheRZsignalwithacertain duty cycle is injected into the converter the carrier passes through the comblike DI whilesomeofitsspectralspikesaresuppressed.ItisworthnotingthatforRZsignal withlargedutycyclesuchas50%and67%thehigherorderspectralspikesareweak. Otherwise, for 33% duty cycle they are stronger. For this reason, another filter is necessary especially for narrower duty cycles in order to achieve the complete conversion[37]. Thus, narrow optical BPF filters out the desired spectral components for suppressing the residual components which induce power ripples to converted NRZ signal[37]. RZTONRZUSINGSPECTRALLINEBYLINECONTROL LinebylinecontrolisalineartechniquetoachieveallopticalRZtoNRZformat conversion. It is experimentally demonstrated at 10Gb/s in [34]. A possible experimental setup consists of a data modulator follows the by line by line sharper describedinsubsection3.1.5.Thus,thegenerateRZpulsesaremodulatedbecoming anRZformatdatastream.Thisfactimpliesthatthedifferentlinesofthespectraare broadened.

46


Figure50.From[34],RZtoNRZformatconversionusinglinetolinepulseshaping.Spectrumandeye diagramdetectedbyone50GHzphotodiodefora)datamodulatedRZformatwithfourspectrallines b)convertedNRZformatwithonlyonespectralline.Insetfiguresshowthespectrainlogscale

It is important to remark that depending on the number of spectral lines are allowedtopassthelinebylinestructure,itispossibleto controlthewidthrangeat theoutputoftheconverter. Figure 50 a) shows the spectrum for four spectral lines and we can observe paying attention in eye diagram that the waveforms signal corresponds with RZ modulationformat.Ontheotherhand,inFigure50b)onlyonespectrallineisallowed topassthelinebylineconverter.Inthiscase,approximateNRZlikeshapeisobtained after the photodiode. Thus, the RZtoNRZ format conversion is achieved using the proposed converter. Nonetheless,although higher spectral efficiency is shown in the conversion,convertedNRZsignalpresentnonidealitiesinitslevel1duetoimperfect suppressionofadjacentspectrallines[34]. NRZTORZBASEDONNONLINEARITYINACROWONASILICONCHIP ThisschemeforformatconversionfromadistortedNRZsignaltoahighquality RZsignalisbasedonnonlinearityinaCROWonasiliconchip.Twosignalscombinedin acouplerareinjectedintotheconverter.DistortedNRZ(controlorpumpsignalata wavelength 0 ) carriers the information and is amplified before feeding the CROW together with a RZ pulse train (probe signal at 1). Both signals are placed at the resonance wavelengths in order to maximize the power coupling to the microrings [28].Figure51a)depictstheregenerativeformatconverter. As Figure 51 illustrates, the principle of operation of this modulation format conversion is described as follows: we assume that control signal determines the resonanceshiftintheCROWduetoitspowerishigherthanprobepower.Whenthe distortedNRZsignalisa0bitbothwavelengthsareplacedinthebottomofthenotch (Figure51b)).So,theprobepowerisminimizedattheoutputoftheconverter.Note that there is a weak shift in this case due to a small power fluctuation [28]. Nevertheless,itisnotenoughtoinduceandimportantresonanceshift(Figure51c)). Ontheotherhand,whenthedistortedNRZsignalisa1bitbothwavelengthsareoff

47

resonancesincethehighcontrolpowerinducesaconsiderableresonanceshift(Figure 51 d)). In this case, the probe power is maximized at the output of the converter. Thus,thecontrolsignalcanmodulatetheprobepulsetrainbyusingtheboxlikenotch effect.AnopticalBPFsuppressthespectralcomponentsplacedaround 0 .

Figure51.From[28],a)RegenerativeNRZtoRZformatconverterusingasinglesiliconCROWb)the pump power is zero b) the pump power is sufficiently small c) the pump is sufficiently high d) a nonlinearfunctionfortheprobelight

Figure52showsthesimulatedresultsat10Gb/sforadistortedNRZsignalafter 40kmtransmissionandtwoRZprobesignalswithdifferentdutycycle.Wecanobserve howthesignalqualityisconsiderablyimprovedaftertheformatconversion.

Figure 52. From [28], 10Gb/s regenerative NRZtoRZ format conversion a) NRZ signal after 40 km transmissionb)20%dutycycleconvertedRZsignalc)10%dutycycleconvertedRZsignal.

Finally,itisworthmentioningthatthetwomostimportantadvantagesofthis method: the regenerative conversion and the ease for highdensity thanks to its compatibilitywiththematureCMOSfabricationprocess[28]. NRZTORZBASEDONFOURWAVEMIXINGINSINGLESOA Asisreportedin[23]asingletodualchannelNRZtoRZformatconversionis experimentallydemonstratedat40Gb/sand20Gb/susingFWMinasingleSOA. The experimental setup consists of the original NRZ signal (which carrier the information)atcandaprobepulseatp (whichcorrespondswithoneRZsignalwith

48

allitsbitsequalto1)combinedina3dBcouplerandlaunchedsynchronouslyintoa SOA[23].

rigth

Due to the FWM effects two new sidebands components are generated at = 2c p and left = 2p c alongbothsidesoftheprobeandNRZsignalasis

illustratedinFigure53:

Figure53.GenerationoftwonewsidebandsduetoFWMinasingleSOA

The SOA function can be understood as a logic AND since the output signal carriers the same information but in a different modulation format [23]. In addition, thismethodcanconsidersingletodualchannelthankstothegenerationofthetwo newsidebands.Nonetheless,from[23]isexperimentallydemonstratedat40Gb/sthat rightsidebandsuffershighersensitivitydegradationincomparisonwithleftsideband. NRZTORZBASEDONCROSSPHASEMODULATIONINSINGLESOA NRZ to RZ format conversion based on XPM in a single SOA and a delay interferometertocontrolthedutycycleisreportedin[24]. Two signals are launched at least into the SOA. One of them consists of an opticalsinewavesignal(probe)withahigherpowerwhichsaturatesdeeplytheSOA. Thus,theoriginalNRZsignalexperienceXPMinducedbytheprobesignalandavery weak XGM. In [24] the format conversion is experimentally demonstrated for 16 channelsatthesametime(10Gb/sforeachchannel).Whenmorethanonesignalis launchedintotheSOA,XPMbetweendifferentchannelsisverytinyduetotheirlower power.

Figure54.From[24],Principleofoperationofthemultichannelformatconversionsandspectraof16 DWDMchannela)beforeSOAb)afterSOAc)afterDId)convertedchannel

49

 Figure 54 illustrates the principle of operation of the multichannel format conversion.ItisimportantnotingthatthewholesetofNRZsignalsachievetheformat conversionbasedinthesameXPM.Thisnonlineareffect,inducedbytheprobesignal, impliesabroaderspectrumineachchannelinjectedintotheSOA(Figure54b)). DIfilter,whichhasacomblikefrequencyresponseenabletokeepthedesired spectral components of each channel at the same time. The duty cycle of the converted RZ signal can be controlled by changing different filter detuning [24]. One tunablefilterisusedtofilteroutoneoftheconvertedchannels. NRZTOCSRZBASEDONXGMANDXPMASINGLESOA a)b)

c)d)

e)f)

Figure 55. From [25], a) Proposed converter structure b) Temporal intensity after optical BPF (normalizedtoitsmaximumvalues)c)OpticalSpectrumof40Gb/sCSRZsignald)Temporalintensity andphaseof40Gb/sCSRZsignale)Phaseofsignalsd)Opticalspectra

50

 The proposed converter is illustrated in Figure 55 a) and consists of one NRZ signalcarryingtheinformationatstogetherwithanopticalclocksignalatplaunched at the same time into a SOA. Subsequently, an optical BPF filters out the converted signal (s) and removes spectral components at p (Figure 55 b). Note that the intensities of the input signals are optimized since the SOA nonlinearities (XGM and XPM) are controlled by the optical clock signal. In addition, in order to achieve the desiredformatconversionclocksignalshouldbealignedwiththebittransitionsofthe NRZsignal[25]. DuetoXGMeffect,alowpowerinclocksignalinducesahighamplificationin NRZsignalwhilehighpowerinclocksignalimpliesaverylowamplification[25].Thus, theRZlikeshapeisachievedatsasFigure55b)illustratesit. InordertounderstandtheXPMeffectforthisparticularformatconversion,itis interesting to analyze firstly the converted modulation format (CSRZ). Optical spectrum, temporal intensity and phase are shown in Figure 55 e) f). The different features of a CSRZ signal and a common technique to generate it are described in subsection2.1.1.3.Inthiscase,weanalyzetheCSRZphasebyconsideringareference frequencyfR.IffRisequalto0(itmeans0GHzinFigure55a))thenconstantphaseand phaseshiftareobservedforeachconsecutivebits.Nevertheless,iffRisequalto20 GHz we observe a phase with a linear slope equal to the bit time (T) within the bit periodandaphaseshiftforeachbittransitions[25]asitisillustratedinFigure55f). Theseslopesarecausedbythechangeofthereferencefrequencysinceitimpliesthe additionofexp(jtT)inthesignalphase[25]. Besides,clocksignalalsoinducesperiodicalphasevariationsinssignalsdueto XPM.Thus,afastphaseshiftisobservedwhenapulseisinjectedinSOA.Thephase recovers slowly its initial value when the pulse is absent. This value is fixed to by adjustingtheinputpoweroftheclocksignal[25].Figure55e)depictsthephaseofthe ssignalafterSOAandCSRZsignalphasebyconsideringfRequalto20GHz.Wecan observe the similarities between both phase waveforms. As a conclusion format conversionfromNRZsignaltoCSRZformatisobtainedwithadetuningof20GHz(the half of the bit rate)[25]. Figure 55 f) illustrates optical spectra before and after the SOA. As it can be read reported in [25], a conversion of multiple NRZ signals is possible using the proposed converter since the nonlinear SOA effects are always controlled by the clock signal. Furthermore, single channel conversion achieves experimentallya3dBsensitivityimprovementoftherequiredOSNRincomparison withtheinputNRZsignal. NRZTORZUSINGCSHG/DFGORCSFG/DFGINAPPLNWAVEGUIDE

NRZ to RZ is the simplest format conversion based on PPLN waveguide. Note that it is possible to achieve the proposed conversion from the different equations describedinthesubsection3.1.7ofthisthesis,usingwhatevercascadedsecondorder nonlinearityprocess.

51


Figure56.From[30],schematicillustrationofPPLNbasedallopticalNRZtoRZ

Figure 56 depicts a schematic illustration of NRZ to RZ PPLNbased. The most important characteristic is that the phase information is not involved in this format conversion.Forthisreason,onlythepresenceoftheNRZsignalandonepumpoptical clockisenoughtogeneratetheconvertedRZidler.ItcarriestheANDresultofinputof inputsignalandthepumpsignal[30]. NRZ/RZTOCSRZUSINGCSHG/DFGORCSFG/DFGINAPPLNWAVEGUIDE

AproposalforPPLNbasedallopticalformatconversionfromNRZorRZtoCSRZ has been proposed and numerically proved at 40 Gb/s using cascaded secondorder nonlinearitiesinaPPLNwaveguide[32]. Figure 57 illustrates the three different signals presented in the conversion system:theoriginalsignal(NRZorRZ),CSRZclock(CSPump)andtheconvertedsignal (CSRZIdler)

Figure57.From[32],schematicillustrationofPPLNbasedallopticala)NRZtoCSRZb)RZtoCSRZ

AswecanobservefromFigure57,inthecurrentmodulationformatconversion a CSRZ pump clock with a phase shift between adjacent bits is required. Hence, cSHG/DFGisusedtoconvertanNRZorRZsignalintoCSRZsignal.Wemustpointout thatbothoriginalandconvertedsignalcarrytheinformationintheiramplitudes. ForachievingtheproposedconversiontheNRZ/RZinputsignalissetattheSHG QPM wavelength where participates in SHG in order to generate a second harmonic wave.Simultaneously,lastgeneratedsignalismixedwiththeCSRZclockpumptoyield an idler wave thanks to DFG. The normalized electric field of the converted idler is

52

expressedas(3.9).Fromthisequationcanbededucedthattheidlerkeepthephase information of the pump signal (alternate phase in each consecutive bit) and the intensityamplitudeoftheNRZ/RZsignal.Forexample,a0bitintheinputimpliesno intensityinconvertedsignal.Inthesameway,bothconjugatephases0orareequal to0andrespectively.Fromequation(3.9)thephasecanbewrittenaswell:

i = + 2s p (3.15)
Thus, an NRZ or RZ input can generate a CSRZ converted signal at the PPLN output. It should be also mentioned that this modulation format conversion can be realizedusingcSFG/DFG.

Figure 58.From [30], cSHG/DFG based 40Gb/s NRZ/RZto CSRZ format conversion a) NRZ spectrum b)Converted CSRZ spectrum c) NRZ eye diagram d) Converted CSRZ eye diagram e) RZ spectrum f)ConvertedCSRZspectrumg)RZeyediagramh)ConvertedCSRZeyediagram

Figure58showsthesimulationsresultsat40Gb/s.Aswecanobservefromthe eyediagrams,theuseofPPLNtechniquesprovidesahighqualityinformatconversion. CSRZTORZUSINGCSHG/DFGORCSFG/DFGINAPPLNWAVEGUIDE

Following the same reasoning used in the previous two format conversion, CSRZ to RZ is possible to be achieved based on cSHGDFG processes with the CSRZ signalsetatSHGQPMwavelength.ApumpCWsignalisalsorequiredattheinputof the PPLN waveguide. From equations (3.9) and (3.10) is easy to understand how the opticalalternatingphaseoftheinputsignalisconvertedintoaconstantopticalphase inallthebitsduetothesquarerelationship.ItcorrespondstotheproposedCSRZtoRZ formatconversion[31].

53

NRZTOPRZUSINGNARROWBANDSILICONMICRORINGRESONATORBASEDNOTCHFILTERS The concept used in NRZ to PRZ format conversion using microring resonator based notch filter is similar to the rest filtering processes which employ the same technique. Thus, notch filter can suppress the NRZ carrier frequency if the carrier wavelengthofthesignallaunchedintothewaveguideisequaltothemicroringnotch filtercenterwavelength.

Figure59.From[29],principleusedinNRZtoPRZformatconversionbasedonamicroringresonator basednotchfiltera)Intensitywaveformsoforiginalandconvertedsignalsb)Wavelengthspectraof aninputopticalNRZ(top)andconvertedPRZsignal(bottom)

Assuminganidealnullinthedesiredcenterwavelength,Figure59depictsNRZ to PRZ format conversion suppressing the strong discrete carrier of the NRZ input signal. Once the signal passes through the narrow band notch filter the sideband remains and, thus, we achieve the PRZ spectrum. This converted signal that carries clock component can be relevant to high rate data NRZ clock recovery for optical communicationsnetworkapplications[29].

3.4.OOKtoCorrelativeCodingformatconversion
3.4.1.Motivation
NRZOOK continues being used in the most terrestrial longhaul transmission systems [4][5] and duobinary is widely employed in current transmission systems workingat40Gb/s[5]. DuobinaryprovidesnarrowspectralbandwidthandlargeCDtolerance.Thus,it is effective in dispersion uncompensated metro or local networks [13]. As we have alreadymentionedpreviously,oneofthemostimportantcharacteristicofCSRZformat

54

is its high tolerance to some linear filtering and nonlinear impairments accumulated alongalonghaultransmissionline[13]. Hence,CSRZtoDBwouldbedesirableatanintermediatenodebetweenalong haulbackboneandaMAN[13]

3.4.2.Modulationformatconversion
CSRZOOKTORZDUOBINARYUSINGMZDI CSRZOOK to RZDuobinary using a single MZDI can be easily understood passingCSRZsignalthroughadelayandaddfilter.Thus,theoutputcanbewrittenas follows:

( 1)

D ( n ) + ( 1)

( n +1)

D ( n + 1) = ( D ( n ) D ( n + 1) ) x ( 1) (3.16)
n n

whereCSRZhasbeenexpressedasaRZsignal(D(n))withphasealternation( ( 1 ) ). D(n)representsthepresentorabsenceofdepower {0,1} .Lookingattherightsideof the equation (3.16) we can observe how it describes a RZ signal passing through a delay and subtract with phase alternation. It corresponds with AMI modulation [13]. As is reported in [1] an AMI generator followed by a CSRZ pulses carver implies a conversion from AMI to Duobinary and vice versa. Thus, we can conclude by saying thatCSRZtoRZDuobinarycanbeachievedattheconstructiveportofaMZDI.

3.5.OOKtoDPSKformatconversion
3.5.1.Motivation
The main motivation of the current format conversions is to connect cost effective OOK based MANs to robust PSK based longhaul backbone networks [8][26][27]. Nowadays, OOK formats are still widely employed in optical communication systems of all scales [8]. Recently, PSK modulation formats have attracted an important interest since some studies have revealed that DPSK exhibits better performance that conventional OOK for long haul transmission [19]. Thus, it will becomeasuitablechoiceforlongdistanceinfutureopticalnetworksduetoitsbetter receiversensitivityandhighertolerancetoCDandPMD[13]. Becauseoftheaforementioned,formatconversionfromOOKtoPSKwouldbe desirableatanintermediatenodebetweenaMANandalonghaulbackbone[8].

55

3.5.2.Modulationformatconversions
NRZOOKTORZDPSKUSINGASOAMZIWAVELENGTHCONVERTER

Figure 60. From [19], schematic diagram of NRZOOK to RZDPSK based on a SOAMZI wavelength converter

Figure60showsthebasicschemediagramoftheproposalformatconversionin [19]. It is based on a semiconductor optical amplifierMach Zehnder interferometer (SOAMZI)wavelengthconverter[19].Threesignalsarelaunchedsynchronouslyatthe inputs of the modulation converter: a Control NRZOOK data which carries the information, a probe RZ clock and an assist continuous wave (CW). The first signalis launchedintotheupperarm,whiletherestofsignalsarelaunchedintobotharms. It is worth nothing that this configuration differs from conventional SOA wavelength converter because of the bias current of SOA#2, which is used for adjustingoutputpower,isslightlylowerthanthatofSOA#1.InthisSOAtheamplitude and phase corresponding to the probe signal are modified due to crossgain modulation(XGM)andcrossphasemodulation(XPM)inducedbyNRZcontrolsignal. Thus, E1 and E2 correspondwiththeamplitudeofprobetrainpulseattheoutputof SOA#1onceitischangedbyXGM(dependingonthepresenceof1or0inthecontrol data). Ontheotherhand,attheoutputofthelowerarm,amplitudeof ( E1 E2 ) 2 andphaseshiftofareachievedbyadjustingthecurrentlevelsofSOA#2. Consequently, once all the signals pass the SOAMZI wavelength converter it obtains the converted RZDPSK thanks to inphase interference and antiphase interference[19].Forthatreason,theequationsthatrelatetheconvertedsignalare:
E1 + E1 E2 j E E E + E2 e = E1 + 1 e j 2 e j = 1 2 2 2 2

(3.17) (3.18)

E2 e j +

E1 E2 j E1 j E E + E2 j e = e + E2 e j 2 e j = 1 e 2 2 2 2

56

Itisinterestingtoremarktheimportanceoflaunchinganassistlightintoboth armsofthesignalconversionconfiguration.Wegetsomedistortioninamplitudeand phasetheduetotheabsentofassistlight.Thereasonofthisdegradationistherapid change of carriers which induces amplitude fluctuations and frequency chirp, and dependsmainlyonthebitpatternoftheinputNRZOOKpulse[19]. Finally,aconventionalbalancedreceiverisenoughtorecovertheoriginaldata transmitted and to obtain the 3dB improvement in the receiver sensitivity in comparisonwiththesingleendedreception. NRZOOKTORZDQPSKUSINGPARALLELSOAMZIOOK/BPSKCONVERTERS

Figure61.From[20],schematicdiagramofNRZOOKtoRZDQPSKbasedonparallelSOAMZI

Previously,aNRZOOKtoRZDPSKconversionhasbeendescribedusingaSOA MZIwavelengthconverter.Inthiscase,theemployedstructureconsistsoftwoSOA MZIOOK/BPSKplacedinparallel[20],asisillustratedinFigure61. Two different NRZ data signals at 0 are launched in the modulation format converter,oneintheupperarm(Port1)andtheotheroneinthelowerarm(Port3). Thesesignalsbehaveascontrolpulses. A continuous wave (2) and a RZ clock (1) are injected inboth arms as assist lightandprobepulse,respectively.Again,assistlightisnecessarytosuppresstherapid changes of carrier density. It enables to enhance distortion in amplitude and phase [19][20]. Forthatreason,intheupperarm,thephaseoftheprobepulseaftertheSOA shifts between 0 and depending on NRZ data (1 and 0 respectively). In the lower arm,aphaseshifterfixedto/2impliesaphasesof/2or/2afterSOAMZI#2asis depictedinFigure61.Theseoutputsarecombinedbeforepassingthroughanoptical BPFwhichkeepstheconvertedsignalat1. It is important to note that optical power is present in every bit time at the outputsofSOAMZIsandcorrespondswithmodulatedRZsignalat1.Inaddition,the peakpowerattheconverteroutputhasthesameamplitudeastheincomingsignals

57

duetotheorthogonalinterferenceandthefourpossiblephases(3/4,/4,/4and 3/4) [20]. Thus, we get the converted RZDQPSK at the output of the proposed converter. Thismodulationformatconversionisnumericallysimulatedandexperimentally demonstratedin[20]at10.7Gsymbol/s,showingbothchannelsalmostthesameBER. NRZOOKTONRZDPSKUSINGASINGLESATURATEDSOA A simple method to get a NRZOOK to NRZDPSK format conversion is employing a single saturated SOA. The conversion is based on the gain and phase modulations of the SOA when it is launched at the input a NRZ signal of finite extinction ratio (ER) [8]. The close absence of power penalty and the absence of a wavelength conversion are the main advantages of this process compared to the previousone[8].Besides,duetotheformatconversionisbasedonaSOAthebitrate islimitedbytherecoverytime,whichdontshouldbehigherthanonebitduration[8]. TheprincipleofoperationisillustratedinFigure62,whereaNRZOOKwitha finiteERcanbeconvertedtoaNRZDPSKduetothenonlinearpropiertiesoftheSOA.

Figure62.From[8],principleofoperationoftheproposedformatconversion

Dependingonthedifferentpowersattheinputsignalitisgetacertaingainand phase shift between consecutive at the output. Thus, it is necessary to adjust the inputpowerwithafiniteERinordertoreachthedeepsaturationregionoftheSOA and to achive the wished conversion [8]. Thanks to the combined SPM (blue line in Figure62),gainsaturation(redline)andacorrectpowerinputchoiceaNRZsignalcan beconvertedtoaNRZDPSK.Itisworthnotingthattheamplitudediferencebetween 0bitand1bitisveryclose,butnotthesame.Nonetheless,weobtainaphaseshift ofattheoutputforthesameinputpower[8].

58

NRZOOKTONRZDPSKUSINGSOABASEDNONLINEARPOLARIZATIONSWITCH NRZOOK to NRZDPSK format conversion is based on XPM effect for phase encoding and nonlinear polarization rotation effect for amplitude equalization [38]. Figure63illustratestheexperimentalsetupemployedin[38]toperformtheproposed formatconversion.

Figure 63. Experimental setup used in [38]. AMZI: 1bitdelay, asymmetric MZI; PC: polarizer controller; OBF: optical bandpass filter, MZM: MachZehnder modulator, EDFA: erbiumdoped fiber amplifier;SOP:stateofpolarizations

A CW light (probe signal) at 0 is injected into a SOA after passing through a polarizercontroller.Simultaneously,anNRZsignalmodulatedinaMZM(pumpsignal) at1islaunchedinthesameSOAbutintheoppositedirectionthankstothecirculator. ThefunctionofPC1,asillustratesFigure63,istoadjustthepolarizationofthe probe signal to 45 respect to the orientation of the layers. This change in the SOP implies different gains in transverse electric (TE) and transverse magnetic (TM) components of the input signal due to the TE/TM asymmetric [38]. When a light polarizedlinearlyisinjectedintotheSOAitsSOProtatestowardthemainaxisofthe amplifier structure since its better gain. For this reason, if pump is OFF, it implies a smallerrotationintheSOPoftheprobesignalthanifthepumpisON[38]. Subsequently, an optical BPF filter out the probe signal at the output of the circulatorinordertokeeponlytheconvertedsignalat0. In order to get a constant power probe PC3 is adjusted to get the same projectionsontheoutputlinearpolarizer.Moreover,bothTEandTMcomponentsof theprobesignalsufferdifferentXPMeffectdependingonpumpswitchedONorOFF. Thus,thephaseoftheconvertedDPSKsignalvarieswiththepumppowerlevels[38]. Finally, a balanced receiver would be enough to recover the original information carried in the phase of the converted DPSK signal. The proposed format conversionisexperimentallydemonstratedat10Gb/sin[38]employingsingleended detection instead of balancer receiver. In this case, a sensitivity penalty of 1.3 dB is obtainedincomparisonwithNRZOOKbacktobacksystem.

59

NRZOOKTORZDPSKUSINGNONLINEARITYINOPTICALFIBER

Figure64.From[26],schematicdiagramofNRZOOKtoRZMaryPSKformatconversion

Figure64showstheprincipleofoperationforNRZOOKtoRZMaryPSK.Thus, it is possible to achieve different modulation formats depending on the number of controlsignalslaunchedintothefiberbymeansofusingthesameschematicdiagram. As we discussed in subsection 3.1.6 the converter is based on phase interaction between two or more signals in a fiber due to XPM effect. For this reason, one RZ probe signal and K number of NRZ channels (each one at its corresponding wavelength) are launched at the same into the HNLF. Depending on the number of channels,alsocalledcontrolsignals,differentmodulationformatsareachieved.Inthis case,inordertoconverttheoriginalNRZsignaltoRZDPSKKisequalto1[26][27]. Accordingwithequation(3.5)wecanexpressthephaseshiftinducedbyXPM ontheprobesignalas:

P = 2 Leff Pc

(3.19)

where Pc correspondswithpowerofthecontrolsignal.Hence,thesuitablechoiceof thisparameterwillprovidethedesiredphaseshift.ForthecaseofRZDPSK,0orif theleveloftheNRZcontrolcorrespondswith0bitor1bitrespectively. Alongthefiberitisimportanttokeepthesynchronismbetweenthepulsesof bothsignals.Moreover,takingadvantageofthenonlinearityeffectofthefiber,leads tothewishedphenomenaofformatconversionoftheRZprobesignal.Aswepointed outinthedescriptionoftheconversiontechnique,walkoffbetweenprobepulseand controlpulseswouldinducepowerdifferencesandunstablephasemodulationonthe convertedsignaldependingonthebitpattern[26].Forthisreason,thewavelengths choicebecomesanimportantkeyissue(discussedindetailinChapter4). An optical BPF centered at the probe wavelength enables to remove and attenuate not desirable spectral components, so only the converted signal is filtered

60

outinthefilter.Abalancedreceiverconvertstheopticalsignalreceivedattheoutput BPFbackintotheoriginalelectricalsignal. The proposed format conversion has been numerically simulated and experimentallydemonstratedin[26]. NRZOOKTORZDQPSKUSINGNONLINEARITYINOPTICALFIBER

NRZOOK to RZDQPK conversion follows the schematic diagram depicted in Figure 64 as well. In this case, two control signals should be launched into the fiber together the probe pulse train in order to achieve the proposed modulation format conversion.Thus,fromequation(3.5)canbededucedthenewphaseshiftinducedon theRZinputsignalduetoXPM:

P = 2 Leff Pck
K =1

(3.20)

where Pc1 and Pc2 correspondswiththepeakpowerofbothcontrolsignals.Again,they areadjustedsothatthephasechangeoftheprobepulseinducedbyonecontrolpulse becomes/2.Therefore,oncetheprobesignalisfilteredouttroughopticalBPF,the peakoftheconvertedpulsetheoreticallyhasaphaseof0,/2,or/4dependingon thecombinationoftwocontrolpulses. The proposed format conversion has been numerically simulated and experimentallydemonstratedin[26]and[27]. NRZOOKtoRZ8PSKformatconversionhasbeenalsonumericallysimulatedat 40Gsymbol/sin[26].Nevertheless,ithasbeenneverexperimentallydemonstratedto my best acknowledge. It should be mentioned that the possible number of phase levelsinconvertedsignalislimitedbyFWMandwalkoffeffects[27].

3.6.DPSKtoCorrelativeCodingformatconversion
As we mentioned previously, duobinary is widely employed in current transmissionsystemsworkingat40Gb/s[5]andprovidesnarrowspectralbandwidth andlargeCDtolerance.TheuseAMIformatisattractiveforlonghaulsystemsdueto itsrobustnesstofibernonlinearimpairments[13]. AswenumericallysimulatedinChapter2,thesimplestmethodtoachieveDPSK to Correlative Coding format conversion is employing a MZDI. Figure 17 shows demodulated constructive and destructive component which corresponds with duobinary and AMI modulation. The generation of duobinary format by means of demodulatingaDPSKsignalhassomepracticalapplicationssinceitcanbeespecially interestingintransmissionssystemswithastrongnarrowbandfiltering[5].

61

3.7.DPSKtoDPSKformatconversion
3.7.1.Motivation
PSKcanadoptdifferentformatssuchasDPSK,RZDPSK,DQPSKorDQPSK.Allof them are described and numerically simulated in Chapter 2. Otherwise, each modulation format is characterized by different features. For example, NRZDPSK presents a higher spectral efficiency compared with RZDPSK formats. At the same time,theselastmodulationformatsallowanimprovementofthereceiversensitivity asitisshowninsubsection2.2. PPLN has become a promising technique in order to achieve conversions between DPSK and RZDPSK formats. These methods are based on second nonlinearitiesandarediscussedinthefollowsubsection. OneofthemostimportantadvantagesoftheDQPSKcomparedtoDPSKisthe increase in the tolerance regarding on CD and PMD approximately by a factor of 2. TheseDPSKandDQPSKwillbechosentolonghaultransmissionsystemsaccordingto type of network [19]. Nonetheless, any converter between these two modulation formatshasbeenreportedyet.

3.7.2.ModulationFormatconversions
NRZDPSKTORZDPSKUSINGCSHG/DFGORCSFG/DFGINAPPLNWAVEGUIDE

Figure65illustratesthethreedifferentsignalspresentedintheconverter:the original signal (NRZ DPSK), pump optical clock and the converted signal (RZDPSK idler). ForNRZDPSKtoRZDPSKformatconversion,asitcanbeobservedfromFigure 65,thephaseinformationofthesignalshouldbekept.Aswedescribedinsubsection 3.1.7suchfunctioncanberealizedbymeansofcSHG/DFGorcSFG/DFGprocesses[30] [32].Usingafiberdelayinterferometer(FDI)withadelaytimeof25ps(for40Gb/s) bothinputNRZDPSKsignalandRZDPSKidlercanbedemodulated.Ifconstructiveand destructive ports are used to recover the original information it corresponds with balancedreceptiondescribedinChapter1.

62


Figure65.From[30],schematicillustrationofNRZDPSKtoRZDPSKformatconversionbasedonPPLN

Again, in [30] simulations results at 40Gb/s are reported. Figure 66 shows spectra, waveforms and eye diagrams of all the signals that take part of the format conversion as well as a referenced RZDPSK. By comparing all of them, specially convertedandreferencedsignal,itispossibletoconcludethattheformatconversion isachievedsuccessfully.

Figure 66. From [30], simulated results for cSFG/DFG based 40Gb/s NRZDPSK to RZDPSK format conversiona)d)spectra,e)k)waveforms,l)q)eyediagramsa),e),f),l),m)40Gb/sNRZDPSKsignal b),g)40GHzpumpopticalclockc),h),i),n),o)ConvertedRZDPSKidlerd),j),k),p),q)ReferencedRZ DPSKsignalf),i),k),m),o),q)Constructivedemodulatione),h),j),l),n),p),usingaFDI

NRZDPSKTOCSRZDPSKUSINGCSHG/DFGORCSFG/DFGINAPPLNWAVEGUIDE

Figure65illustratesthethreedifferentsignalspresentinthecurrentconverter: the original signal (NRZ DPSK), CS pump optical clock and the converted signal (RZ DPSKidler).Notethatalthoughtheschematicillustrationisverysimilartotheprevious formatconversion,inthiscaseanopticalclockwithphasealternationisrequired. InthiscaseboththesignalphaseinformationandCSpumpopticalclockphase informationshouldbepreserved.ItispossibletogeneratecSFG/DFGprocessintothe PPLN[30][32]andaccordingtoequation(3.11).

63

Both input NRZDPSK signal and RZDPSK idler can be demodulated again by usingafiberdelayinterferometer(FDI)withadelaytimeof25ps(for40Gb/s).

Figure 67. From [30], schematic illustration of NRZDPSK to CSRZDPSK format conversion based on PPLN

Figure68showsthesimulationsresultsat40Gb/sreportedin[30].Bytakinga lookandcomparingspectra,waveformsandeyediagramswecanconfirmthatPPLN provideahighqualityformatconversionfromNRZDPSKtoCSRZDPSK.

Figure68.From[30],simulatedresultsforcSFG/DFGbased40Gb/sNRZDPSKtoCSRZDPSKformat conversiona)d)spectra,e)k)waveforms,l)q)eyediagramsa),e),f),l),m)40Gb/sNRZDPSKsignal b), g) 40GHz CS pump optical clock c), h), i), n), o) Converted CSRZDPSK idler d), j), k), p), q) ReferencedCSRZDPSKsignalf),i),k),m),o),q)Constructivedemodulatione),h),j),l),n),p),usinga FDI

3.8.FSKtoPSKformatconversion
3.8.1.Motivation
Few years ago, various kinds of optical modulations formats focused on the enhancing of the optical transmission capacity and achieving alloptical label

64

processing. FSK and PSK belonged to this group. Highspeed optical FSK transmission systems using direct or external modulation were reported. Furthermore, it is well knownthatPSKisapotentialreplacementforOOKinlonghaultransmissionsystems [39].

3.8.2.FormatConversion
FSKTOPSKUSINGOPTICALDOUBLESIDEBANDTECHNIQUE

Thecurrentsubsectiondescribesatechniqueforallopticalmodulationformat from FrequencyShiftKeying (FSK) to PhaseShiftKeying (PSK) based on optical DoubleSidebandSuppressedCarrier(DSBSC)modulationtechnique. FSK signal is generated in an external modulator. It consists of four optical phasemodulatorsinparallelasisillustratedinFigure69.Aswecanobserve,several radio frequency signals can be applied in the differentelectrodes. Thanks to rfa and

rfb it is possible to generate frequency shifted light wave at the output of the
and bottom arm. This signal controls the phase retardation from P to R ( f ( t ) ) and fromQtoR ( f ( t ) ) withsamevalueandoppositesign[39]. modulator.Thesesignalsareofthesamefrequencyandhaveaphaseshiftof90[39]. Furthermore, rfc enables to induce other phase difference in this case between top

Figure69.OpticalFSKmodulatorinspiredfrom[39]

TherearetwoseparatespectralcomponentsintheFSKsignalknownasupper sideband (USB) and lower sideband (LSB). We can keep one of the sidebands by controlling the rfc voltage. For example, if phase retardation f ( t ) is set to 4 , the opticalFSKwillgenerateLSB(0stateasisshowninFigure69).Otherwise,if f ( t ) is equalto 4 USBwillbegenerated[39]. Figure70illustratestheprincipalofmodulationformatconversionfromFSKto PSK. Once LSB or USB are generated, FSK signal passes through a DSBSC modulator withasinusoidalmodulatingsignal.Aswecanobserve,twospectralcomponentsare generated at the output of the modulator. One of them is always generated with a carrier frequency of f0 independently of the state switched in FSK modulator [39].

65

Thus, we obtain the PSK signal after filtering this spectral component by using an OBPF.

Figure 70. From [39], Principal of modulation format conversion from FSK to PSK and experimental resultsinspired

FSKtoPSKmodulationformatconversionhasbeenexperimentaldemonstrated andreportedat10Gb/s[39].

3.7.Summary
Different alloptical modulation format conversion has been reported up to now.MostofthemdescribeconversionsfromOOKformattoOOKformatsinceuntil fewyearsagoopticalcommunicationssystemsprimarilyemployedconventionalOOK signalsineitherNRZorRZ.Theseconvertedarebasedondifferentmethodssuchas SOAs, MZI, FPLD etc. Nonetheless, during the past few years new advanced modulation formats have attracted the interest of many studies. Therefore, new promising techniques like HNLFs and PPLN have receivedalso considerable attention since they enable new format conversion between OOK formats and PSK formats at highbitrates. The main disadvantage of techniques based on SOA is that they impose high inputsignalpowerandslowoperationspeedlimitedbySOArecoverytime[29]. To my best knowledge only one method based on CROW considers that the inputsignalistypicallydistortedduetothetransmissionoveracertaindistance.The restofproposalsmethodsarebasedonanassumptionthattheinputsignalisofhigh quality.Forthisreason,iftheinputsignalsuffersdistortionthepreviousmethodsmay notworkcorrectly.

66

Chapter 4 All Optical modulation format conversion from RZOOK to RZDPSK based on fiber nonlinearity inopticalfiber
Inthepreviouschapter,anoverviewofdifferenttechnologiesandmethodsfor achieving all optical modulation format conversion is given. The purpose of this chapteristostudyindepthRZOOKtoRZDPSKbasedonfibernonlinearities.Several numericallysimulationsusingthesoftwareVPITransmissionMakerWDMarecarried out in order to investigate some important requirements such as pulse shape, wavelengthorpowertransmissionamongothers. The use of HNLFs, together with PPLNs waveguides, has become a promising technologyinthefieldofallopticalprocessing.Modulationformatconversionsfrom OOKtoPSKusingnonlinearitiesintothefibershavebeenstudiedinseveralreportslike [26]and[27],buttomybestknowledgenorequirementhasbeenclearlyspecifiedand justified.Furthermore,inthischapteraRZcontrolsignalinsteadofNRZformatisused inordertoinducetheXPMontheprobepulsetrain. Theuseofsilicafiberasthenonlinearmediumismoreattractiveincomparison withthesemiconductordevicesduetoitsultrafastnonlinearresponse.However,we mustpointoutthatlongfibersandhighinputpowersarerequiredtoobtainenough nonlineareffect[37]. All the numerical simulations performed in this Chapter to investigate the proposedRZOOKtoRZDPSKarecarriedoutat40Gb/s.

4.1.Principleofoperation

Figure71.ConfigurationofallopticalRZOOKtoRZDPSKformatconversionbasedonnonlinearityin opticalfiber,inspiredfrom[26]

67

InordertoobtaintheproposedallopticalformatconversionfromRZOOKto RZDPSKweconsidertwodifferentRZsignalssynchronouslylaunchedintotheHNLF. The configuration follows the same principle of operation described in previous chapter for NRZOOK to RZDPSK format conversion. Strong control signal (which carriestheinformation),inducesaXPM,whichmodulatesinthiswaythephaseofthe weak RZ clock signal. Figure 71 illustrates the schematic drawing of this format conversion. Along the fiber it is important keep the synchronism between both signals. Thus, taking advantage of the nonlinearity effect of the fiber, it leads to the desired phenomena of format conversion of the RZ probe signal. Walkoff between probe pulse and control pulses would induce power differences and unstable phase modulationontheconvertedsignal. An optical BPF with a center wavelength at P (probe signal) enables to filter outtheconvertedsignalandremovenotdesirablespectralcomponents,soonlythe RZDPSK signal reaches the receiver. The balanced receiver converts the optical receivedsignalbackintotheoriginalelectricalsignal. In next subsection, weare going to discuss necessary parameters for carrying out the simulations of the proposed format conversion in VPI Transmission Maker WDM.

4.2.ParametersChoice
Inordertoachievethedesiredformatconversion,somefiberandtransmission parameters should be well known. For example, suitable choice of wavelengths and peak control power will enable to keep synchronize into the fiber and induce the desiredXPMrespectively.Forthisreason,wearegoingtodiscussalltheseparameters innextsubsections.

4.2.1.Fiberparameters
Parameter fo(zerodispersion) DispersionSlope Loss n2 Length Effectivearea Units [Hz] [ps/nm2/km] [dB/km] [m2/W] [m] [ m2]
Table4.ParametersoftheHNLF

Value 193.4x1012 0.019 0.3 3.33x1012 500 12

Table4summarizestheprincipalparametersthatcharacterizethenonlinearity effects in a certain HNLF fiber. Dispersion slope and zero dispersion frequency will

68

influence in wavelength choices. On the other hand, refractive index, length and effective area will allow calculating the suitable peak power for control signal which inducestheXPM. Dispersion becomes an important phenomenon to be considered. It occurs in optical fibers because the refractive index is wavelength dependent [3]. For this reason,eachwavelengthcomponentexperimentsdifferentspeedoverthefiberwhich produces a loss of synchronism between both signals launched in the HNLF. The dispersionparameterisexpressedas[3]:

D=

1 1 d vg

(4.1)

wherevgisthegroupvelocityinthefiber.TheparameterDhastheunitsofps/nm/km and is used to quantify dispersion. It should be mentioned that there is a specific wavelength in the fiber where the dispersion parameter is equal to zero. This wavelength is referred as zero dispersion wavelength (ref) [3]. Figure 72 illustrates graphicallydispersionandgroupvelocityparametersrelatedbyequation(4.1)[3].This figure helps to understand why the suitable wavelength choice enables to keep synchronize between different signals. As we can observe, the choice of C (control signal) and P (probe signal) wavelength should keep the symmetry with respect to 1551nm (zero dispersion wavelength) in order to decrease as much as possible the effectofawalkoffbetweenthetwosignalssynchronouslylaunchedintothefiber.

Figure72.Illustrativegraphfordispersionparameterandasafunctionofwavelengthinspiredfrom [3]whereSisthedispersionslopeand=1/vg

Otherwise,althoughbothsignalshavethesamegroupvelocity,itisimportant tonotethatdispersionwillaffectthemsincethepulsetrainshaveaspectrumwitha certainbandwidth. Forallthereasonsdescribedpreviously,thewavelengthchoseninsubsequent simulationsareC=1548.2nmandP=1555.0nm. Oncewehaveinvestigatedandfixedfiberparametersandwavelengthsforthe differentsignalswhichtakepartinproposedRZOOKandRZDPSKformatconversion, agoodpointtodiscussistheeffectoftheshapeandwidthofatransmittedsignal.

69

4.2.2.Requirementsontheshapeandwidthofpulseinbackto backsystems
AssumingthatintheproposedformatconversionanidealRZDPSKpulsetrain isdemodulatedinthereceiver,wewanttoobservetheeffectsoftheshapeandthe widthofthereceivedpulsesbeforecarryingouttheconvertersimulations. The systems used in the next simulations are very similar to the setup described in Chapter 2, specifically in subsection 2.1.1.2 and 2.1.3.1 for RZOOK and RZDPSK modulation formats, respectively. Theuse of a pulse sourcemodulated ina MZMinsteadofaconventionalpulsecarveristheonlydifferentbetweenbothsetups. OOK modulation format is received by direct detection while DPSK format is demodulatedinabalancerreceiver.Simulation results at 40Gb/s for different pulse shapeanddifferentpulsewidthareplottedinFigure73andFigure74forRZOOKand RZDPSKbacktobacksystems.

-23 -24 -25 -26

RZ-OOK

Sensitivity (dBm)

-27 -28 -29 -30 -31 -32 -33 -34 -35 0.0 0.1 0.2 0.3 0.4 0.5

1st order gaussian 2nd order gaussian 3rd order gaussian hyperbolic sechant

0.6

0.7

0.8

0.9

1.0

FWHM/Tb

Figure73.SensitivityasafunctionofthepulsewidthforRZOOKbacktobacksystem

70

RZ-DPSK
-24 -25 -26 -27 -28 -29 -30 -31 -32 -33 -34 -35 -36 -37 -38 -39 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Sensitivity (dBm)

1st order gaussian 2nd order gaussian 3rd order gaussian hyperbolic sechant

FWHM/Tb

Figure74.SensitivityasafunctionofthepulsewidthforRZDPSKbacktobacksystem

FromFigure73andFigure74wecanobservethatbothbacktobacksystems describe a very similar waveform. For all pulse shapes (1st 2nd 3rd and hyperbolic sechant),whilepulsewidthisincreasingitsvalue,thesensitivitycurvedecreasesuntil an optimum point. It corresponds in both OOK and DPSK transmissions to a pulse characterizedbyFullWidthatHalfMaximum(FWHM)equalto0.3xTb.Fromthispoint untilthemaximumpulsewidth,thecurvesstarttoconvergetoacertainvaluecloseto 32.5dBmand36dBmforOOKandPSKsystemsrespectively. Intheoptimumsensitivitypointweobtainanapproximatedifferenceof3.5dB between singleended detection (OOK reception) and balanced detection (DPSK reception).Thisdifferencecanbeattributedtothebenefitofapproximately3dBgiven bybalancedreceivers[1][4][5][6]. We can observe that the tendency from the optimum point previously mentionedistoobtainacertainsensitivitydegradationcomparedwithshorterpulse widths.Intuitively,thefactisthatifthepulseenergyisconcentratedinashortertime slot for a certain average power, the amplitude difference between 1 and 0 bits becomebigger.IcanbegraphicallyunderstoodinFigure7fromChapter2wherethree pulses with different duty cycle but same average power are plotted. This higher distancebetween1and0impliesalowerbiterrorprobabilityinthedecisionstep.

71

Otherwise,theprecedingreasoningcouldbecontradictoryifwepayattention toshortestpulsewidthvalues.Noticeablesensitivitydegradationisobservedinthese pointsincomparisonwithbroaderpulsewidth.Thiscontradiction(sinceforashorter pulsewidthandusingalwaysthesameaveragepower,thedistancebetween0and 1ishigher)canbeattributedtothesampletime.Itbecomesanimportantparameter toconsiderasitcanimplyquitedifferenceamplitudevaluesforalittlevariationinthe correct sample moment. Figure 75 depicts how for a short pulse width we obtain a higher amplitude variation when the sample time is not the optimum (it would correspondwiththeamplitudepeakofthepulse).

Figure 75 . Variation in the intensity amplitude due to not optimum sample time for two different pulses

In addition, the effect of using the same bandwidth in the band pass filter, mainly for shorter pulse widths, leads to filter uncorrectly these signals due to the broadspectrums.Forexample,wegetanimprovementofalmost3dBoncetheband pass filter bandwidth is increased from 160 GHz to 800 GHz when we carry out the simulationforatrainpulsecharacterizedbyaFWHM=0.025xTbinabacktobackDPSK system. In Table 5, simulated sensitivity results are expressed for different BPF bandwidth. Bandwidth(GHz) 4xBitRate 10xBitRate 20xBitRate 32xBitRate Sensitivity(dBm) 28.039 30.781 32.203 32.497

Table5.SimulatedreceiversensitivityfordifferentBWat40Gb/s

Theresultsobtainedinthesesensitivitygraphsbecomequiteinterestingsince, for the proposed modulation format conversion, a pulse train with the information modulated in its phase is demodulated in a balanced receiver. Thus, an optimum choiceforprobesignalinthetransmitterisapulsetraincharacterizedbyapulsewidth equalto0.3xTb(Tb=25ps)assumingidealfibertransmissionwithoutdispersioneffect.

72

4.2.3.ControlandProbesignal
Aswediscussedinsubsection4.2.1theprincipleofoperationofRZOOKtoRZ DPSKformatconversionisbasedonnonlinearitiesintoaHNLF.Hence,theprobepulse ismodulatedinitsphasethankstoXPMinducedbythecontrolpulses[26][27].With the purpose to find the suitable amplitude peak power for input RZ signal, the equation(4.2)shouldbesolvedinordertoachievethedesiredor0phaseshiftinthe convertedsignal.Thus,byusingthefiberparameterspreviouslydescribed,thephase changeoftheprobepulseduetoXPMcanbewrittenas[3][26][27]:
( p ) = 2 Leff P ( c )

(4.2)

where Leff istheeffectiveinteractionlengthdefinedas[3]:

Leff =

1 e L

= 464.307 m

(4.3)

and isanimportantnonlinearparameterexpressedasfollows[3]:

2 n2 = 3.615 x10 3 Aeff

(4.4)

Thus, in order to achieve a phase increase of or 0 depending on the information carried by the control signal (bit 0 or bit 1) equation (4.2) must be solvedas:

P ( c ) =

( p )

2 Leff

= 297.9 mW

(4.5)

ItisimportanttoremarkthatforobtainingafavorableXPMthismustbethe peakpowerattheinputofthefiber.Asaconsequence,inthetransmitterweneedthe doubleofpowerduetothe3dBcouplerlosses. Followingthesamereasoning,thepeakpoweroftheprobepulsetrainissetto 2mW. This power must be enough to avoid the degradation suffered due to coupler andfiberlosses.Notethatthe500mofHNLFinvolvealossof0.15dB.Incontrast,the increase of the transmitted probe power also implies a higher SPM which can be expressedas[3]:

NL = Leff ( Pin )

(4.6)

Therefore,averyhighpowerforprobesignalwouldimplyanadditionalphase generatedduetoSPMeffectanditwouldnotbecomeagoodchoicefortheproposed formatconversion.

73

Finally, it is interesting to investigate which is the response of the proposed converterfordifferentsignalpowersattheinputoftheHNLFandfordifferentlengths of fiber. As we described in this subsection, it is possible to obtain the correct converted signal for a certain length by using the equation (4.2) depending on the inputpower.Wehaveshownmathematicallythatapowerof298mWmakespossible aphaseshiftbetweenconsecutivebitsrequiredforaDPSKmodulation.Thus,ifwe keep on increasing the input power of the control signal, it will lead to achieve the neededphaseshiftagain.Forexample,fora500meterslengthfiber:
( p ) = 2 Leff ( 298mW ) = ( p ) = 2 Leff (894 mW ) = 3

(4.7) (4.8)

BothpowersenableacorrectRZtoRZDPSKformatconversion.Figure76showsthe receiver sensitivity as a function of the power in the fiber input. We can observe a parabola with its minimum placed exactly in the power calculated previously. This pointcorrespondstoaphaseshift.Forthisreason,thebalancedreceiverobtainsthe bestsensitivity.InFigure76areillustratedalsothesameresultsfor250mand1000m fiberlengths.

-10 -12 -14 -16

500m length fiber 250m length fiber 1000m length fiber

Sensitivity (dBm)

-18 -20 -22 -24 -26 -28 -30 -32 -34 0 100 200 300 400 500 600 700 800 900 1000 1100

Control power before fiber (mW)

Figure76.Receiversensitivityasafunctionofthecontrolpowerattheinputofthefiber

74

4.2.4.BandPassFilter(BPF)effectbeforereceiver
In the previous subsection has been discussed the fiber parameters, requirementsontheshapeandwidthofthetransmittedpulsesandthesignalpowers in order to obtain the proposed format conversion. Two signals are launched synchronouslyintothefiber.Controlpulsesinteractwithprobepulsesandviceversa generatingdifferentphasechanges.Theprobepulseismodulatedinphaseassuming idealformatconversionbasedonXPMinducedbycontrolsignal. A filtering step is needed to extract the phase modulated probe. For this reason, we want to investigate the effect of this filter. Figure 77 depicts the setup employedinfollowingsimulations.

Figure77.SetupemployedforinvestigatinginopticalBPFeffectinspiredfrom[5]

As we can observe, the system is build by the same DPSK transmitter and receiver used in subsection 4.2.2. The only difference between the current RZDPSK transmitterandthatonedescribedindetailinChapter2istheuseofanopticalpulse sourceinsteadoftheconventionalpulsecarver(33%,50%and67%dutycycle).Note thatbothwaysenableRZformat.Furthermore,a2ndorderopticalBPFisaddedinthe middleofthesysteminordertostudyitseffect.Thepulsewidthinthetransmitterhas beenfixedtothesupposedidealFWHMvaluefromtheprevioussubsectionandequal to0.3xTb. Byusingascript,thepowerinreceptionisprogressivelymodifiedinorderto obtainthereceivedpowerneededforaBERequalto109. Figure 78 plots the sensitivity of the receiver as a function of the filter bandwidth simulated at 40Gb/s. The horizontal red line indicates the sensitivity achievedwithouttheeffectofthefilter. Asitwaswished,whiletheBWisincreasingthereceiversensitivityapproaches progressivelytoacertainvalue.Thisvalueisclosetotheresultobtainedsimulatingthe sameschemewithoutanyopticalBPFbeforethereceiver. We can justify the curve illustrated in Figure 78 as follows: obviously, the RZ DPSKsignalwillkeepconstantitsbandwidthinfrequencydomainifthepulsewidthis fixedtoacertainvalue.Thus,assumingthattherearenotspectralinterferencesfrom othersignals,thesensitivitybecomesconstantoncethesignalissufficientlyfiltered.

75

-32.5 -33.0 -33.5 -34.0

Sensistivity W/O filter

Sensitivity (dBm)

-34.5 -35.0 -35.5 -36.0 -36.5 -37.0 -37.5 -38.0 -38.5 0 1 2 3 4 5 6 7 8 9 10

Bw/R

Figure78.Sensitivityasafunctionofthefilterbandwidthforthesetupdescribedpreviously

Figure 79. RZDPSK optical spectrum for FWHM= 0.1xTb ps(red), FWHM= 0.2xTb (yellow), FWHM=0.3xTb(green)andFWHM=0.5xTb(blue)

76

The spectra illustrated in Figure 79, correspond to RZDPSK signals with different pulse widths. Obviously, it is well known that not all the pulse trains have associated the same spectrum in frequency domain because of the duration of the pulsesintimedomainareinverselyproportionaltothebroadopticalspectrum.Green linecorrespondswiththeidealRZDPSKpulsetrainusedinthepreviousstudyofthe filtereffects. Thus,wecanconcludethatforafixedpulsewidth,afterfilteringwithacertain BW and recovered successfully the transmitted signal, we can increase the OBPF bandwidthwithoutachievinganimportantsensitivityimprovement.Nonetheless,ifa neighboursignalissetinaclosewavelengthitcouldprovokeinterference.

4.3.SystemSimulationModel
Thissectionisdevotedtomakeafirstapproximationoftheformatconversion proposedandprobeforsomenumericallysimulationsthecorrectlyoperation.Upto now, based on the different studies made in previous subsection, we know some parameterswhichcanoptimizetheproposedformatconversion.Thus,thepulsewidth of the probe signal is fixed to 0.3xTb (extracted from subsection 4.2.2) and the bandwidth of the optical BPF at the output of the fiber is set to three times the Bit Rate (extracted from subsection 4.2.4). We have calculated the peak power of the controlsignalinordertoinducethetargetphaseshift.Thepeakpowerofthecontrol and clock signal are 27.75 dBm (596 mW) and 3 dBm (2 mW), respectively. As we discussedandsetinthesubsection4.2.1,thewavelengthofthecontrolpulseandthe probepulseare1548.2nmand1555.0nm,respectively.

Figure80.SimulationsetupforRZOOKtoRZDPSKformatconversioninspiredfrom[5]

ThesystemsetupisbuiltusingthemodulesprovidedbyVPI.Figure80shows thesetupconsistingontwoRZtransmitterslaunchingatthesametimeaclocksignal (calledprobeaswell)andacontrolsignal,whoseopticalpulsesaremodulatedusinga MZM.Anoptical3dBcouplercombinesbothsignalsattheHNLFinput.Alongthefiber theprobepulseismodulatedinphaseduetotheeffectofXPMinducedbythecontrol pulse[26][27].OpticalBPFfiltersouttheconvertedsignal,whichisdemodulatedina balancedreceiver.ItisdescribedingreatdetailinChapter2.Forthesetupconsidered

77

weanalyzedthesimulatedwaveformandphaseattheoutputofHNLFatthebitrate of40Gb/s(Tb=45ps). Datasignal(controlpulse)isgeneratedbymodulatingatrainpulseinaMZM with 40 Gb/s PRBS of length 211 1 . In contrast, the clock signal (probe pulse) is generatedusinganopticalpulsesourcewhichguaranteestheexistenceofoneoptical pulseforeachtimeslot. Someparameterssuchastheshapeofbothsignalsandthepulsewidthofthe controlsignalhavetobedecidedyet.Inprinciple,wecansupposethatalongerpulse width and more rectangularlike shape will enhance the modulation format conversion.Furthermore,assumingthepresenceofwalkoff,itispossibletoachieve theXPMsincetheshorterprobepulseskeepalignedwiththemaximumpeakpowerof thecontrolsignal.ThisideacanbeeasierunderstoodbylookingFigure81:

Figure 81. From [16], schematic illustration of XPM assuming presence and absence of walkoff a)IncomingNRZcontrolsignalb)Probepulse

Then,inordertodiscussoneofthesefreeparametersitisnecessarytosetthe restofthem.Byfollowingthepreviousreasoning,wewanttoinvestigatethecarrying out of the proposed format conversion for different pulse shapes. Thus, we fix the pulse width of the control and probe signal to 0.7xTb (17.5 ps) and 0.3xTb (7.5 ps) respectively. The Gaussian order of the control signal is set to 2. Simulating for the absenceandthepresenceofwalkoff,Figure82plotsthesensitivityofthereceiveras a function of the order Gaussian of the control signal. Note that for higher Gaussian ordersthepulsebecomesmorerectangular. When the control signal is placed at 191.4 THz (C =1548.2 nm) both signals launched in the fiber are characterized by the same group velocity because of both prove and control wavelengths keep the symmetry with zero dispersion wavelength [3].Inthiscase,Figure82illustratesanapproximatelyconstantcurve(blueline)since thesensitivityofthereceiverispracticallythesameforeachGaussianorder.Itmeans thattheinfluenceofthecontrolpulseshapeintheformatconversionisveryweakfor thecaseofwalkoffabsence. We can generate walkoff by changing the central frequency of the control signal to 185THz (C =1548.2 nm). In this case, Figure 82 shows a little sensitivity improvementforhigherorderGaussian.Nonetheless,wecanobservethatfromfifth orderGaussianthereceivedsensitivitybecomesconstant.

78

-16 -18 -20

fc=191.4 THz fc=185 THz

Sensitivity (dBm)

-22 -24 -26 -28 -30 -32 -34 0 1 2 3 4 5 6 7 8 9 10

~ 12.5 dB

Order Gaussian
Figure82.ReceiversensitivityasafunctionofthecontrolGaussianorder

-7 -8 -9 -10 -11

fc=185 THz Fifth order Gaussian

Sensitivity (dBm)

-12 -13 -14 -15 -16 -17 -18 -19 -20 -21 -22 -23 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

FHWM/Tb

Figure 83. Sensitivity as a function of the control pulse width for different control and prove group velocities

79

Oncewehaveobtainedtheresponseoftheconverterfordifferentcontroland probeparameters,wewanttoinvestigatetheeffectofpulsewidthonthemodulation formatconversion.Thus,theGaussianorderofthecontrolpulsesissettofive(lower Gaussianorderwheresensitivitystartstobeapproximatelyconstantfromtheresults previouslyobtained)anditspulsewidthbecomeavariableparameter. Figure 83 shows the sensitivity as a function of the control pulse width for different control and prove group velocities (Note again that control and probe wavelengths do not keep the symmetry with respect to the zero dispersion wavelength).Wecanobservethatforshorterpulsewidththeconvertedsignalsuffers noticeable sensitivity degradation. Unexpectedly, we obtain the best sensitivity for a FWHMequalto0.6xTb.Thisfactcanbeattributedtothenotoptimumgenerationof theRZcontrolsignal.Wemustpointoutthattheinterferencebetweenconsecutive1 bits increases for lower Gaussian orders and higher pulse widths. It can cause importantdistortionsinthegeneratedcontrolpulse.Figure84showsgraphicallythe effectofthisinterference.

Figure84.Residualpowerbetweenconsecutive1bitsinthecontrolsignaltransmitter

TheresidualpowerbetweenconsecutivebitsalsoinducesaXPMintheprobe signal and, depending on the walk off generated over the fiber, it can imply a bad interactionbetweenbothcontrolandprobesignals. Up to now, we have discussed and extracted some conclusions about which requirements are needed, such us power signals, shape and size of the pulses, wavelength choice etc. to enhance the proposed RZOOK to RZDPSK format conversion. In order to prove its suitable operation, set up described in Figure 80 is numerically simulated at the bit rate of 40 Gb/s setting first all the parameters discussed in this Chapter. Now, the wavelength of the control pulse and the probe pulse are set to C =1548.2 nm and P =1555.0 nm, respectively. Thus, both signals have the same group velocity. The peak power and FWHM of the probe pulse are 2mW and 7.5 ps (0.3xTb). The peak power of the control pulse is set to 596mW so thatthephaseshiftoftheprobepulseinducedbyXPMis.Thispulseischaracterized by a FWHM equal to 17.5 ps (0.7xTb). The optical pulse sources employed in both transmitters generate second and fifth order Gaussian for probe and control pulse, respectively. The simulated eye diagram of both signals is shown in Figure 85. The

80

fiberparametersoftheHNLFaredescribedinTable4.Theconvertedsignalisfiltered using an optical BPF with a bandwidth equal to three times the Bit Rate. Thus, it is possibletoremovenotdesiredspectralcomponents.Finally,aconventionalbalanced receiver(describedinChapter2)demodulatestheopticalsignalinordertorecoverthe originalinformation.

Figure 85. Eye diagrams of control and probe signal employed in the proposed format conversion

Figure86showstheopticalspectrumattheoutputoftheHNLF(redline)and the optical spectrum of the probe signal at the input of the fiber (yellow line). New spectral components at FWM wavelengths have been generated due to the nonlinearities of the fiber. We observe that the signal have been successfully converted from RZOOK to RZDPSK by comparison of the optical spectrum of the probesignalattheinputandtheoutputofthefiber.

Figure86.OpticalspectrumattheoutputoftheHNLF(redline).OpticalspectrumoftheRZprobesignal (yellowline)

81

Onlytheconvertedsignalispresentedinthebalancedreceiverinputoncethe signalattheoutputofthefiberisfiltered.Figure87a)andb)illustratestheintensity andphasewaveformandopticalspectrumoftheconvertedsignal.Theopticalpower appearsineachbitandtheinformationiscarriedinthephase[1][4][5].Figure87c) plots the signal constellation once the converted RZDPSK signal is processed with MATLAB.Asitwasexpected,wecandifferentiatetwoclearpointswithaphaseshiftof approximately 180. It is important to note that the important value is the relative phasebetweenconsecutivebitsandnottheabsolutephase. Inprinciple,thephaseofthesignalshouldonlypresenttwovaluesdepending ontheinformationcarried.Nevertheless,fromthesimulationresults,thephaseofthe convertedsignalpresentpeaksbetweenconsecutivepulses.Anyway,thesepeaksdo not have any influence in the correct reception due to these samples are not consideredforrecoveringtheoriginalbitsinthereceiver. Figure 88 shows the residual power ( 7 W ) existing between consecutive pulses. This is the reason of the continual existence of phase peaks which are characterizedbyrelativephaseamplitudeof180.

Absolut Phase Constellation 90 120 0.02 150 0.01 30 0.03 60

180

210

330

240 270

300

Figure87.ConvertedRZDPSKsignala)Intensityandphasewaveformsb)OpticalSpectrumc)Absolutephase constellation

82


Figure88.Residualpowerbetweenconsecutivepulses

Finally, Figure 89 illustrates the BER as a function of the received power for 33RZDPSK and converted RZDPSK signals. In this case, 33RZDPSK generated employing a conventional NRZ transmitter followed by a pulse carver is taken as a referencesignal.Thus,theconversionprocessrepresentsareceiversensitivitypenalty of4dB.

Converted RZ-DPSK 33RZ-DPSK

-log (BER)

5 6 7 8 9 10 11 12 -42 -41 -40 -39 -38 -37 -36 -35 -34 -33 -32 -31 -30

Received power (dBm)

Figure89.BERasafunctionofthereceivedpowerfor33RZDPSKandconvertedRZDPSKsignals

83

4.4.RequirementsinControlsignalforastandardRZpulsetrain Clock
Thepurposeofthissubsectionistostudytherequirementsontheshapeand width of the control pulse respect to the RZ clock. As we observed in previous subsection, the change of Gaussian order in control signal did not imply a clear sensitivityimprovementoftheconvertedsignal.BychangingthecontrolsignalFHWM withthepulseshapefixed(1storderGaussian)wearegoingtoobservethedifferent converter responses for each type of probe signal. We assume again the same parameters for HNLF fiber we set in previous studies and a length of 500 m. As we haveanalyzedpreviously,inordertopreventawrongconversionfromRZtoRZDPSK based on nonlinearities in optical fiber due to a possible walkoff, a longer control pulsewidthissupposedtobeusefulagainstthedifferentgroupdelayvelocities. The clock signal is standard RZ pulse train with 33, 50 or 67% duty cycle, as generated in a MachZehnder modulator. It could be higher order Gaussian, but a possiblerealimplementationoftheschemewouldrequiretheuseofanexpensive40 GHzmodelockedlaser. Figure 90 shows the new set up diagram for the proposed modulation converterwherethepulsesourcefortheprobesignalhasbeenreplacedbyastandard pulsetrainwith33%,50%and67%dutycycle.Itissimulatedagainat40Gb/s.

Figure 90. Proposed modulation format conversion based on HNLF for a standard RZ probe signals inspiredfrom[5]

Figure91,Figure92andFigure93plotthereceiversensitivityasafunctionof the pulse width for 33% 50% and 67% RZ probe signal respectively. We can observe how the curves shown in the figures decrease (better sensitivity) for higher pulse widths. Nonetheless, only converted 50% RZ probe signal achieve the optimum sensitivity for maximum pulse width (FWHM=1xTb). Even obtaining sensitivity values very close for longer pulse width, one must wonder why the optimum point not correspondwithFWHM=1xTb.Itcanbeattributedtothedifficultyofsettingacertain peakpowerinthecontrolsignalduetotheinterferencesbetweenconsecutivepulses. Figure94illustratestheeyediagramsforthecontrolsignaldependingontheFWHM parameter. Lower interferences are observed for shorter pulse widths. For FWHM=1xTbweobservethatthecontrolsignalisalmostaNRZformat.

84

-23 -24 -25 -26

Sensitivity (dBm)

-27 -28 -29 -30 -31 -32 -33 -34 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

FWHM/Tb

Figure91.Sensitivityasafunctionofthepulsewidthofthecontrolsignal(1storderGaussian)for33% RZprobesignal

-19 -20 -21 -22

Sensitivity (dBm)

-23 -24 -25 -26 -27 -28 -29 -30 -31 -32 -33 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

FWHM/Tb

Figure92.Sensitivityasafunctionofthepulsewidthofthecontrolsignal(1storderGaussian)for50% RZprobesignal

-16 -17 -18 -19 -20 -21 -22 -23 -24 -25 -26 -27 -28 -29 -30 -31 -32 -33 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Sensitivity (dBm)

FWHM/Tb

Figure93.Sensitivityasafunctionofthepulsewidthofthecontrolsignal(1storderGaussian)for67% RZprobesignal

85

a)

b)

c)

d)

e)

f)

g)

h)

i)

j)

Figure94.Interferencebetweenadjacentpulsesforthecontroltransmitter.1storderGaussian. a)FWHM=0.1xTb,b)FWHM=0.2xTb,c)FWHM=0.3xTb,d)FWHM=0.4xTb,e)FWHM=0.5xTb,f) FWHM=0.6xTb,g)FWHM=0.7xTb,h)FWHM=0.8xTb,i)FWHM=0.9xTb,j)FWHM=1xTb

86

 Finally,wemustpointoutthatinarealscenariocontrolsignaldoesnotbelong to the converter, but it comes from an external network. Assuming that the signal reacheswithhighquality,wearegoingtostudywhatisthewavelengthlimitationin the proposed converter for the control signal. The detuning of the wavelength from theoptimumvaluecauseapenaltyonthesensitivity.From189to193THzthepenalty is less than 3 dB, specially constant for 33% RZ probe signal, as walkoff is more probabletobecoherentwithlowerFWHM.

-18 -19 -20 -21 -22

33% RZ probe signal 50% RZ probe signal 67% RZ probe signal

Sensitivity (dBm)

-23 -24 -25 -26 -27 -28 -29 -30 -31 -32 -33 -34 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196

Control Frequency (THz)

Figure95.Receiversensitivityasafunctionofthecontrolcenterfrequency

87

Chapter5Conclusion
Inthisproject,advancedmodulationformatsandallopticalformatconversion have been investigated. The study was motivated by the fact that in future optical networks different modulation formats may be selective depending on characteristic such us scope, size and bit rate among others. Thus, a transparently interconnection betweennetworkswillbedesiredforfuturecommunicationssystems. Firstofall,opticalmodulationformatswereclassifiesin:Intensitymodulation formats (NRZOOK, RZOOK and CSRZOOK), Correlative Coding and PartialResponse Formats (Duobinary and AMI) and Differential Phase Modulation (Binary DPSK and DQPSK).Theyaremainlycharacterizedbythewaythattheycarrytheinformation.The first and oldest type of modulation format uses its optical intensity to modulate the information. Unlike CSRZ where there is an alternation in each bit transition, correlative coding introduces a correlation between the optical phase and the data information. Differential phase modulation encodes the information on the phase changebetweenconsecutivebitsduetotheabsenceofanabsolutephase.Finally,a multilevelphasemodulation(DQPSK)withtwobitspersymbolwasdiscussedaswell. In order to investigate the advantages and limitations for all modulation formats,theirtransmittersandreceiversweredescribedbeforesimulatingthemina backtobackstructureusingVPITransmissionMakerWDM.Themodulatortechnology chosen for carrying out numerical simulation is MachZehnder modulators (MZMs) whichwillenabletoobtainthewhishedmodulationattheoutputofthetransmitter dependingonthewayofdriving. OOK modulation format have a very low spectral efficiency and are characterizedbyapoornonlineartoleranceduetothestrongopticalcarrierpresentin optical spectrum. CSRZ attenuates it thanks to the signal alternation in is phase. DuobinarycomparedwithOOKmodulationformatshasahighertolerancetotheCD anditsnarrowerbandwidth.DPSKandDQPSKpresentaveryhighspectralefficiency comparedwithOOKanduseabalancedreceiverinordertodemodulatedereceived signal. Theresultsofthenumericallysimulationsshowedthatforthesameaverageof power RZOOK formats have a better sensitivity than NRZOOK, which is due to the higherdifferencebetween1and0forshorterpulsewidth.Theresultsalsoillustrate the approximately 3 dB sensitivity benefit provided by the use of balanced receiver insteadofsingleenddetection.Byinterconnectingtransmitterandreceiverthrougha fiber,iseasytoobservethehighertolerancetoCDofduobinaryincomparisontoNRZ OOK. Inthisthesisdifferenttechniquesusedinmodulationformatconversionwere described.

88

SomeofthemuseaSOAindifferentstructuresinordertoobtainthedesired format conversion. These methods are based on a single saturated SOA, on optical switches, SOA MZI or on SOA nonlinearities such as XGM, XPM, SPM and FWM. Besides,SOAhasanimportantlimitationinoperationspeedduetotherecoverytime, especiallyathighoperationbitrates. MZDI, which is typically used in balanced receivers, can also be used as converter since when a DPSK signal is injected in its input both constructive and destructiveportspresentingtwomodulatedsignalswiththetransmittedinformation. CROW on silicon chip is a promise technique since using its box like notch arounditsresonancewavelengthisabletoconvertfromdistortedNRZtoaRZsignal. Thisisthefirstreportedandexperimentallydemonstratedmethodthatnotconsiders highqualityintheinputsignal. PPLN waveguides and HNLFs have become in the past few years a promise techniqueinordertoachievenonlinearapplications.InfibersXPMisthemoststudied nonlinearityinthefieldofmodulationformatconversion,especiallyfromOOKtoPSK. ByusingPPLNwaveguide,itispossibletoachievemodulationformatconversionbased on its cascaded second order nonlinearities: cSHG/DFG and cSFG/DFG. They have become promising techniques due to their high nonlinear coefficient, ultrafast response,andnegligiblespontaneousnoise. Besides, matrix from/to has been filled up with the main modulation format converters reported until now. It was shown in order to allow a general overview aboutwhichconversionshaveattractedmoreattention.OOKtoOOKconversionhas beenwidelystudied. Ingeneral,theprincipleofoperationofproposedformatconversionsisbased ontwoormoresignalslaunchedsynchronouslyintotheconverter.Oneofthesignals, calledcontrol,carriestheinformationintheoriginalmodulationformat.Thefunction of the converter is modulated this information on the rest of signals injected in the converteraswell.Thesemethodsalsoimplyawavelengthconversion. Different requirements for all optical format conversion from RZOOK to RZ DPSKwerediscussedandsubsequentlytheconversionwasnumericallysimulated.The convertedsignalpresentedasensitivitydegradationofalmost4dBincomparisonwith a33RZDPSKsignalreceivedinbacktoback.Thesuitablechoiceofwavelengthenables tominimizethedispersioneffectkeepingsynchronizedbothcontrolandprobesignal alongthefiber. Finally,standards33%,50%and67%RZwasemployedasaprobesignal.The modulation format conversion was studied varying the central frequency of the control. 33% duty cycle has the best response to these frequency variations. The sensitivitycurvesasafunctionofthewavelengthadopttheshapeofparaboliccurves withitsminimumatthewavelengthchosenforminimizingdispersioneffect.

89

References
[1] J. Winzer and RenJean Essiambre, Advanced Modulation Formats for High Capacity Optical Transport Networks, Journal of Lightwave Tecnology, vol. 24, no.12,2006. [2] Gabriel Charlet, Progress in Optical Modulation Formats for HighBit Rate WDM Transmission, IEEE Journal of selected topics in Quantum Electronics, vol. 12, No. 4, 2006. [3] Govind P. Agrawal, FiberOptic Communications Systems, third edition, 2002, WileySeriesinmicrowaveandopticalengineering. [4]A.H.GnauckandP.J.Winzer,OpticalPhaseShiftKeyedTransmission,Journalof LightwaveTecnology,vol.23,no.1,2005. [5]DirkVanDenBorne,Robustopticaltransmissionsystems,TechnischeUniversiteit Eindhoven,2008. [6] Chris Xu, Xiang Liu, Xing Wei, Differential PhaseShift Keying for High Spectral Efficiency Optical Transmissions, IEEE Journal of selected topics in Quantium electronics,vol.10,no.2,2004. [7] Takashi Ono, Yutaka Yano, Kiyoshi Fukuchi, Toshiharu Ito, Hiroyuki Yamakazi, MasayukiYamaguchiandKatsumiEmura,CharacteristicsofOpticalDuobinarySignals in Terabit/s Capacity, HighSpectral Efficiency WDM Systems, Journal of Lightwave Technology,vol.16,No.5,1998. [8] Kazushige Yonenaga and Shigeru Kuwano DispersionTolerant Optical Transmission System Using Duobinary Transmitter and Binary Receiver, Journal of lightwavetechnology,Vol.15,No.8,1997. [9]S.G.Park,L.H.Spiekman,M.EiseltandJ.M.Wiensenfeld,Chirpconsequenceof alloptical RZtoNRZ conversion using crossphase modulation in an active semiconductorphotonicintegratedcircuitIEEEPhotonicsTechnologyLetters,vol.12, no.3,pp.233235,2000. [10]H.K.Lee,K.H.Kim,J.T.Ahn,M.Y.Jeon,andE.H.Lee,Allopticalformatconversion fromNRZtoRZsignalsusingawalkoffbalancednonlinearfiberloopmirrorElectron. Lett.,vol.32,pp.23352336,1996. [11]GhungGhiuLee,YunJongKim,ChulSooPark,HyuekJaeLeeandChangSooPark, ExperimentalDemonstrationof10Gb/sDataFormatConversionsBetweenNRZtoRZ UsingSOALoopMirrror,JournalofLightwaveTechnology,vol.23,no.2,2005.

90

[12] C.W. Chow, C. S. Wong and H. K. Tsang, AllOptical Modulation Format Conversion and Multicasting Using Injection Locked Laser Diodes, Journal of LightwaveTechnology,vol.22,no.11,February2004. [13] Yuanyuan Lu and Yikai Su, Conversions among binary optical modulation formats,OpticsExpress,vol.16,no.6,2008. [14]J.Yu,G.K.Chang,J.BarryandY.Su,40Gb/ssignalformatconversionfromNRZto RZusinga MachZehnderdelayinterferometerOpticalCommunication,pp419422, 2005. [15]P.S.Cho,D.Mahgerefteh,andJ.Goldhar,10Gb/sRZtoNRZformatconversion usingasemiconductoropticalamplifier/fiberBragggratingwavelengthconverter,in Proc.Eur.Conf.OpticalCommunication(ECOC98),vol.1,pp.353354,1998. [16] C.H. Kwok, Chinlon Lin, PolarizationInsensitive AllOptical NRZtoRZ Format Conversion by Spectral Filtering of a Cross Phase Modulation Broadened Signal Spectrum,IEEEJournalofSelectedTopicsinQuantumElectronics,vol.12,no.3,2006. [17] Wendi Li, Minghua Chen, Yi Dong and Shizhong Xie, AllOptical Format ConversionFromNRZtoCSRZandbetweenRZandCSRZusingSOABasedFiberLoop Mirror,IEEEPhotonicsTechnologyLetters,vol.16,no.1,2004. [18] L. Xu, B.C. Wang, V. Baby, I. Glesk and P.R. Prucnal, Alloptical data format conversionbetweenRZandNZRbasedonaMachZehnderinterferometricwavelength converter,IEEEPhotonicsTechnologyLetters,vol.15,pp308310,2003. [19] Ken Mishina, Akihiro Maruta, Shunsuke Mitani, Toshiharu Miyahara, Kazuyuki Ishida, Katsuhiro Shimizu, Tatsuo Hatta, Kuniaki Motoshima and Kenichi Kitayama, NRZOOK to RZBPSK Modulation Format Conversion Using SOAMZI Wavelength Converter,JournalofLightwaveTechnology,vol.24,no.10,2006. [20] Ken Mishina, Suresh M. Nissanka, Akihiro Maruta, Shunsuke Mitani, Kazuyuki Ishida,KatsuhiroShimizu,TatsuoHattaandKenichiKitayama,Allopticalmodulation format conversion from NRZOOK to RZQPSK using parallel SOAMZI OOK/BPSK convertersOpticsExpress,vol.15,no.12,2007 [21]CishuoYan,YikaiSu,LilinYi,LufengLeng,XiangqingTian,XinyuXuandYueTian, AllOpticalFormatConversionfromNRZtoBPSKusingasinglesaturatedSOA,IEEE PhotonicsTechnologyLetters,vol.18,no.22,2006. [22] Hung, Yu, Chang and Lin, 10Gb/s all optical NRZtoRZ data format in a dark opticalcombinjectedsemiconductoropticalamplifierOpticsExppres,vol.15,no.6, 2006. [23]J.Dong,X.Zhang,F.Wang,Y.YuandD.Huang,SingletodualchannelNRZtoRZ format conversion by fourwave mixing in single semiconductor optical amplifier, Electronicsletters,vol.44,no.12,2008.

91

[24] Yu Yu, Xinliang Zhang, J.B. RosasFernndez, Dexiu Huang, R.V. Penty and I. H. White,SingleSOAbased16DWDMchannelsallopticalNRZtoRZformatconversions withdifferentdutycyclesOpticsExpress,vol.16,no.20,2008. [25] Tiago Silveira, Ana Ferreira, Antonio Teixeira and Paulo Monteiro, 40Gb/s MultichannelNRZtoCSRZFormatConversionUsinganSOA,IEEEJournalofLightwave Technology,vol.16,no.1,pp.203205,2004. [26]KenMishina,SatoruKitagawaandAkihiroMaruta,Allopticalmodulationformat conversion from onoffkeying to multiplelevel phaseshiftkeying based on nonlinearityinopticalfiberOpticsExpress,vol.15,no.13,2007. [27]SatoruKitagawa,SureshM.NissankaandAkihiroMaruta,Allopticalmodulation formatconversionfromNRZOktoRZMaryPSKbasedonfibernonlinearity,Optics Express,2006. [28] Tong Ye, Cishuo Yan, Yuanyuan Lu, Fangfei Liu and Yikai Su, Alloptical regenerative NRZtoRZ format conversion using coupled ringresonator optical waveguide,OpticsExpress,vol.16,no.20,2008. [29] Linjie Zhou, Hui Chen and Andrew W. Poon, Onchip NRZtoPRZ Format Conversion Using NarrowBand Silicon Microring Resonator Based Notch Filters, JournalofLightwaveTechnology,vol.24,no.10,2006. [30]JianWang,JunqiangSun,XinliangZhang,DexiuhuangandMartinM.Fejer,All OpticalFormatConversionsUsingPeriodicallyPoledLithiumNiobateWaveguides,, IEEEJournalofQuantumElectronics,vol.45,no.2,2009. [31]JianWang,JunqiangSun,XinliangZhang,DexiuHuang,andM.M.Fejer,Optical phaseerasureanditsapplicationtoformatconversionthroughcascadedsecondorder processesinperiodicallypoledlithiumniobate,OpticsLetters.,vol.33,no.16,2008. [32] Jian Wang, Junqiang Sun, Xinliang Zhang, and Dexiu Huang, Proposal for PPLN Based AllOptical NRZtoCSRZ, RZtoCSRZ, NRZDPSKtoCSRZDPSK, and RZDPSKto CSRZDPSK Format Conversions, IEEE Photonics Technology Letters, vol. 20, no. 12, 2008. [33] Jian Wang, Junqiang Sun, J. B. RosasFernndez, G. Huang, Xinliang Zhang and Dexiu Huang, First Demonstration on the Nontransparency of FWM and its applicationof40Gb/sAllOpticalCSRZtoRZFormatConversion,OFC,2008. [34] Zhi Jiang, Daniel E. Leaird, Andrew M. Weiner, Width and WavelengthTunable Optical RZ Pulse Generation and RZtoNRZ Format Conversion at 10 Ghz Using SpectralLinebyLineControl,IEEEPhotonicsTechnologyLetters,vol.17,no.12,2005. [35]W.AstarandG.M.Carter,10Gb/sRZOOKtoBPSKformatconversionbycross phase modulation in single semiconductor optical amplifier, Electronics Letters, vol.42,no.25,2006.

92

[36] Jian Wang, Junqiang Sun, Xinliang Zhang and Dexiu huang, First Demonstration on the Nontransparency of PPLN and its Potential Application of CSRZtoRZ Format Conversion,OFC,2008. [37] Yu Yu, Xinliang Zhang, Dexiu Huang, Lijun Li and Wei Fu, 20Gb/s AllOptical FormatConversionsFromRZSignalsWithDifferentDutyCyclestoNRZSignals,IEEE PhotonicsTechnologyLetters,vol.19,no.14,2007. [38]HuanJiang,HeWen,LiuyanHan,YiliGuoandHanyiZhang,AllopticalNRZOOK to BPSK Format Conversion in an SOABased Nonlinear Polarization Switch, IEEE PhotonicsTechnologyLetters,vol.19,no.24,2007. [39] Tetsuya Kawanishi, Takahide Sakamoto and Masayuki Izutsu, AllOptical Modulation Format conversion from FrequencyShiftKeying to PhaseShiftKeying by Using Optical DoubleSideband Modulation Technique, Conference on Lasers & ElectroOptics,2005. [40] Michael Connelly, Semiconductor Optical Amplifiers and their Applications, Dept.ElectronicandComputerEngineeringUniversityofLimerick.

93

APPENDIXA:VPIScript
setDoDebug1 ##################################################################### ##################################################################### #######SENSITIVITYSEARCHALGORITHM #######INPUTS: ####### ####### ####### ####### #######

#######WantedBERBERtofind

#######WantedBERErrorMaximumallowedBERValue #######EXAMPLE:1.05e9

#######MaxAttMaximumattenuationindBtoconsider ####### #######MaxIterationsMaximum#iterationsbeforegivingup ####### #######OUTPUTS: #######BER #######Att #######SENS #######SENSdBm #######OK #######ErrorCode ####### ####### ####### ####### ####### ####### ####### ####### ####### #######

BitErrorRate

AttenuationusedtogetBER SensitivityinW SensitivityindBm

Errorindication(OK=1=>noerror) Descriptionoferror(ifany)

#######DESCRIPTION:

#######ThisalgorithmvariestheattenuationinAttenuator1to #######reachanBERclostoWantedBER.Theallowederroris

#######specifiedinWantedBERError(WantedBERError=1.05e9allows####### #######forupto5%error).Themaximumattenuationisspecified####### #######bytheuserthroughtheparameterMaxAtt.Thenumberof#######

94

#######iterationsdependsheavilyonMaxAtt,don'tsetittoohigh!####### #######IfWantedBERisfound,theBER,theAttenuation,andthe ####### #######isreturnedinvariablesspecifiedabove. ####### ####### ####### ####### #######

#######IfWantedBERisnotfound,anerrorcodeisreturned #######

#######ThisscriptwaswrittenbyHenrikNrskovPoulsenin #######FortranandconvertedtoTCL/TKbyTorgerTokle.

###################################################################### #######IMPORTANT!!! #######

#######MakesurePostValue1isthepowerandPostValue2istheBER####### ###################################################################### if{$B2B==1}{ setAttLow0 setAttHigh100 }else{ setAttLow[expr$B2BAtt13] setAttHigh[expr$B2BAtt+3] } ###Parametersusedinthesensitivityiterationprocess setfound0 setcount0 setBER0 setAtt0 setSENS"" setSENSdBm"200"

95

setErrorCode"" setTargetBER$WantedBER setTargetBERError[exprlog10($WantedBERError)log10($WantedBER)] ###DEBUGGINGFILE if{$DoDebug==1}{ setdebugpath"C:/DocumentsandSettings/s080726/Desktop/debug2.txt" setdebug[open$debugpatha] puts$debug"STARTING" close$debug } #################################################################### ######FINDSTARTPOINTSATZEROANDMAXATTENUATION######## #################################################################### ###CheckifpossibletoreachtargetBER setstateAttenuator1Attenuation$AttLow run1 wrapup incrcount setBER[statevaluePostValue2InputValue] setPWR[statevaluePostValue1InputValue] ###WRITETOLOGFILE setPWRLogFile[expr10*log10($PWR*1000)] setdebug[open$debugpatha] puts$debug"" puts$debug"Iteration:$count Power:$PWRLogFile BER:$BER"

puts$debug""

96

close$debug if{$BER<1e40}{setBER1e40} setErrorLowAtt[exprlog10($BER)log10($TargetBER)] if{$ErrorLowAtt>0}{ setexit1 setErrorCode"BERhigherthanWantedBERatminimumattenuation($AttLowdB)" setOK0 }else{ setexit0 } if{$DoDebug==1}{ setdebug[open$debugpatha] puts $debug "LOW: Att: $AttLow BER: $BER Error: $ErrorLowAtt EXIT: $exit $ErrorCode" close$debug } ###FindBERforlowestlimit if{$exit==0}{ setstateAttenuator1Attenuation$AttHigh run1 wrapup incrcount setBER[statevaluePostValue2InputValue] setPWR[statevaluePostValue1InputValue] ###WRITETOLOGFILE setPWRLogFile[expr10*log10($PWR*1000)] setdebug[open$debugpatha]

97

puts$debug"" puts$debug"Iteration:$count Power:$PWRLogFile BER:$BER"

puts$debug"" close$debug if{$BER<1e40}{setBER1e40} if{$BER<$TargetBER}{ setexit1 setErrorCode"MAXIMUMATTENUATIONISTOOLOW" setOK0 setBER1 }else{ setexit0 } setBER[statevaluePostValue2InputValue] setErrorHighAtt[exprlog10($BER)log10($TargetBER)] if{$DoDebug==1}{ setdebug[open$debugpatha] puts $debug "HIGH: Att: $AttHigh BER: $BER Error: $ErrorHighAtt EXIT: $exit $ErrorCode" close$debug } } ############################################################# ######STARTITERATIONS######## ############################################################# setOldError$ErrorLowAtt while{$exit==0}{

98

###SETNEWATTENUATION set Att [expr ($ErrorLowAtt*$AttHigh $ErrorHighAtt*$AttLow) / ($ErrorLowAtt $ErrorHighAtt)] setstateAttenuator1Attenuation$Att run1 wrapup incrcount setBER[statevaluePostValue2InputValue] setPWR[statevaluePostValue1InputValue] setError[exprlog10($BER)log10($TargetBER)] ###WRITETOLOGFILE setPWRLogFile[expr10*log10($PWR*1000)] setdebug[open$debugpatha] puts$debug"" puts$debug"Iteration:$count Power:$PWRLogFile BER:$BER"

puts$debug"" close$debug ###WRITETODEBUG if{$DoDebug==1}{ setdebug[open$debugpatha] puts$debug"$Att$BER$Error$count" close$debug} ###SETNEWLOWORHIGHLIMITS if{[expr$Error*$ErrorLowAtt]>0}{ setAttLow$Att setErrorLowAtt$Error }else{

99

setAttHigh$Att setErrorHighAtt$Error } ###CHECKIFSUCCESSFUL if{[exprabs($Error)]<$TargetBERError}{ setexit1 setOK1 setErrorCode"" setSENS$PWR setSENSdBm[expr10*log10($PWR*1000)] if{$DoDebug==1}{ setdebug[open$debugpatha] puts$debug"SUCCESSFUL,ERRORis$Error" close$debug } }else{ if{[expr$Error*$OldError]>0}{ setb[expr$OldError/($OldError+$Error)] if{[expr$Error*$ErrorLowAtt]>0}{ setErrorHighAtt[expr$ErrorHighAtt*$b] }else{ setErrorLowAtt[expr$ErrorLowAtt*$b] } setOldError$Error } }

100

###CHECHIFTOOMANYITERATIONS if{$count>$MaxIterations}{ setexit1 setOK0 if{$DoDebug==1}{ setdebug[open$debugpatha] puts$debug"UNSUCCESSFUL,DONE$countITERATIONSANDERRORIS$Error" close$debug}}} if{$DoDebug==1}{ setdebug[open$debugpatha] puts$debug"EXIT" puts$debug"" puts$debug"" close$debug }

101

APPENDIXB:MATLABcode
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % MODULE AND PHASE WAVEFORMS FOR CONVERTED DPSK SIGNAL %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % FHWM = X ps %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% fieldX = fopen('FieldX01.txt', 'r'); fieldY = fopen('FieldY01.txt', 'r'); x1 = fscanf(fieldX,'%f'); y1 = fscanf(fieldY,'%f'); x=x1(1:1:end); y=y1(1:1:end); x = x'; y = y'; fclose(fieldX); fclose(fieldY); %Module Calculation z=x+y.*i; m=abs(z); %Phase calculation phase_rad= angle(z); phase=(180/pi).*phase_rad; phase_norm=phase/180; figure(1) subplot(2,1,1),plot(m) subplot(2,1,2),plot(phase_norm)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % ABSOLUT AND RELATIVE PHASE FOR CONVERTED DPSK SIGNAL %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % FHWM = X ps %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% clear all fieldX = fopen('fieldX.txt', 'r'); fieldY = fopen('fieldY.txt', 'r'); x1 = fscanf(fieldX,'%f'); y1 = fscanf(fieldY,'%f'); x=x1(794:128:end); y=y1(794:128:end); x = x'; y = y'; fclose(fieldX); fclose(fieldY);

102

%Module Calculation z=x+y.*i; m=abs(z); %Phase Calculation phase_rad= angle(z); phase=(180/pi).*phase_rad; for (k=1:2046) incr_phase_rad(k)=phase_rad(k+1)-phase_rad(k); end incr_phase=(180/pi).*incr_phase_rad; figure(1) repre=polar(phase_rad,m); title('Phase constellation') set(repre,'LineStyle','none') set(repre,'Marker','*') figure(2) m2=ones(1,2046); repre2=polar(incr_phase_rad,m2); set(repre2,'LineStyle','none') set(repre2,'Marker','*')

103