Abstract. Key words: DPSK, DQPSK, OOK, FWM, WDM, BER, limitation factors

Transcription

1 Abstract Sabol, Dušan: WDM systémy s DPSK kľúčovaním [Diplomová práca]. Žilinská univerzita v Žiline. Elektrotechnická fakulta; Katedra telekomunikácií. Školiteľ: Ing. Jozef Dubovan. Stupeň odbornej kvalifikácie: inžinier (Ing.). Žilina: EF ŽU, s. Complex approach to optical communication is proposed in my work, where limitation factors due to significant physical phenomena with cancellation techniques and their suitability investigated. Economical fundamental factors are also involved following trends in optical communication are proposed. Frequently used DPSK format is theoretically described and impact of OSNR level and laser linewidth on its performance is investigated by simulation technique provided in VPI simulation software. Behavior of used modulation format within WDM system examined at four channel s WDM and affect of Kerr effect is qualitatively assessed. Key words: DPSK, DQPSK, OOK, FWM, WDM, BER, limitation factors

2 ANOTAČNÝ ZÁZNAM - DIPLOMOVÁ PRÁCA Priezvisko, meno: Sabol, Dušan školský rok: 2005/2006 Názov práce: WDM systémy s DPSK kĺúčovaním Počet strán: 51 Počet obrázkov: 46 Počet tabuliek: 1 Počet grafov: 0 Počet príloh: 0 Použitá lit.: 17 Anotácia (slov. resp. český jazyk): Táto diplomová práca skúma limitujúce faktory dnešných WDM sytémov. Popis DPSK modulácie a jej aplikácii, ako budúceho modulačného formátu pre optické komunikácie ďalšej generácie. Práca obsahuje simulácie popisujúce chybovosť jednotlivých typov modulácii a následne ich poškodenia pôsobením Kerrovho efektu v štovkanálovom WDM systéme. Anotácia v cudzom jazyku (anglický resp. nemecký): This diploma thesis investigates limitation factors of current WDM systems. Description of DPSK modulation and its applications as modulation format for optical communication of next generation. Thesis contains simulation that describes BER of selected modulations and their distortions due to Kerr effect at four channel WDM system. Kľúčové slová: DPSK, DQPSK, OOK, FWM, WDM, BER, Limitujúce faktory Vedúci práce: Ing. Jozef Dubovan Recenzent práce: Prof. Ing. Milan Dado, PhD. Dátum odovzdania práce:

3 Content List of Figures and Tables...III List of Abbreviations...VI 1 Introduction Complex approach to optical communication Physical characteristic Mitigation of impairments and distortions Nonlinear effects Kerr effect Stimulated scatterings Amplification Shaping Jitter Modulation format Economical aspects Catalysts of OT Inhibitors of OT Differential phase-shift keyed format DPSK format DPSK transmitter Transmitter evaluation Pulse carver DPSK Receiver Balanced versus single-ended detection Tolerance to Optical Filtering DPSK transmission at 10 Gb/s DPSK transmission at 40 Gb/s DQPSK application Trends in the optical communication Transmission bands and spectral efficiency Multiplexing techniques Time division multiplex (OTDM)...34 I

4 4.2.2 Wavelength Division Multiplex Simulation DQPSK format DPSK format OOK format Impact of Kerr effect Description of simulation scheme OOK performance DPSK performance DQPSK performance Potential of best setup Conclusion...51 II

5 LIST OF FIGURES AND TABLES FIGURE 1: SPECTRAL VARIATION OF ATTENUATION; SOURCE [1]...2 FIGURE 2: PRODUCTS OF KERR EFFECT; [6]....4 FIGURE 3: A) TEMPORAL VARIATION OF SPM-INDUCED PHASE SHIFT AND FREQUENCY CHIRP FOR GAUSSIAN (DASHED CURVE) AND SUPER-GAUSSIAN (SOLID CURVE) PULSE. B) EXPERIMENTALLY OBSERVED SPECTRA FOR A NEARLY GAUSSIAN PULSE AT THE OUTPUT OF A 99 M LONG FIBRE. SPECTRA ARE LABELED BY MAXIMUM PHASE SHIFT RELATED LINEARLY TO THE PEAK POWER; [4]....5 FIGURE 4: EYE DIAGRAM OF COMPLETE AND INCOMPLETE COLLISIONS; [5]...5 FIGURE 5: A) NONLINEAR CROSSTALK DUE TO FWM, B) EYE DIAGRAM OF AMPLITUDE NOISE DUE TO FWM; [5]....6 FIGURE 6: EXAMPLES OF INTRACHANNEL EFFECTS; [7]....7 FIGURE 7: RAMAN GAIN COEFFICIENT FOR PURE SILICA AND 1550 NM PUMP WAVELENGTH; [8]...8 FIGURE 8: THE OUTCOME OF SRS AT WDM SYSTEM; [5]....8 FIGURE 9: BEHAVIOR OF INPUT AND OUTPUT POWER DUE TO SBS; [5]....9 FIGURE 10: CHARACTERISTIC OF ERBIUM AMPLIFIER (LEFT) AND RAMAN AMPLIFIER (RIGHT); [7]...11 FIGURE 11: VARIOUS HYBRID AMPLIFICATION SCHEMES; [7]...12 FIGURE 12: THE DISPERSION S INFLUENCE; [7] FIGURE 13: MATERIAL, WAVEGUIDE AND CHROMATIC DISPERSION FOR CURRENT FIBRES;.13 FIGURE 14: DISPERSION COMPENSATION BY FBG; [9] FIGURE 15: A), C) SYMMETRICAL DISPERSION MANAGEMENT; B), D) NON SYMMETRICAL DISPERSION MANAGEMENT; [5]...15 FIGURE 16: THE EFFECT OF DISPERSION SLOPE FOR BORDER AND CENTRAL FREQUENCIES; [5]...15 FIGURE 17: PMD DUE TO BIREFRINGENCE EFFECT; [11] FIGURE 18: CANCELLATION OF TIMING JITTER WITH SYMMETRIC DISPERSION PROFILE; [5] FIGURE 19: COMPARISON OF IP TRAFFIC PREDICTIONS WITH INSTALLED CAPACITY; [1]...19 FIGURE 20: SIGNAL CONSTELLATION OF OOK AND DPSK MODULATION; [16]...22 FIGURE 21: TYPICAL DPSK TRANSMITTERS: A) IMPLEMENTATION WITH PM, B) IMPLEMENTATION WITH MZM; [16]...24 III

6 FIGURE 22: A) A TYPICAL RZ-DPSK TRANSMITTER. (B) OPTICAL INTENSITY AND PHASE WAVEFORMS GENERATED BY AN IMPERFECT PULSE CARVER; [16] FIGURE 23: DPSK RECEIVER; [16]...26 FIGURE 24: NUMERICAL CALCULATIONS FOR THE REQUIRED OSNR AT BER = FOR 33% RZ-DPSK AND OOK AS A FUNCTION OF RECEIVER AMPLITUDE IMBALANCE Β; [16]...28 FIGURE 25: PENALTIES IN NON IDEAL RZ-DPSK RECEIVERS: (A) AMPLITUDE IMBALANCE IN THE BALANCED DETECTOR (DASHED CURVE IS FOR A DELAY INTERFEROMETER WITH AN EXTINCTION RATIO OF ONLY 10 DB), (B) PHASE IMBALANCE IN THE BALANCED DETECTOR, (C) DELAY-TO-BIT RATE MISMATCH IN THE DELAY INTERFEROMETER, (D) LASER FREQUENCY OFFSET FROM THE IDEAL AS SET BY THE INTERFEROMETER PHASE DIFFERENCE (DASHED CURVE IS FOR 33% RZ-DPSK). CIRCLES ARE EXPERIMENTAL RESULTS; [16] FIGURE 26: EXPERIMENTAL RESULTS OF PENALTIES FROM NARROW OPTICAL AND ELECTRICAL FILTERING; [16]...29 FIGURE 27: DQPSK TRANSMITTER AND RECEIVER; [16] FIGURE 28: EXPERIMENTAL SETUP OF; [18] FIGURE 29: SIMULATION SCHEME OF BER = ƒ (OSNR) FIGURE 30: COMPARISON OF EXPERIMENTAL [18] AND SIMULATION OUTPUT...37 FIGURE 31: BER VS OSNR FOR 10 GB/S AND 40 GB/S DQPSK FORMAT FIGURE 32: SIMULATION SCHEME OF BER = ƒ (OSNR) FIGURE 33: BER VS OSNR FOR 10 GBS AND 40 GB/S DPSK FORMAT FIGURE 34: SIMULATION SCHEME OF FWM GENERATION...41 FIGURE 35: EYE DIAGRAM OF OOK TRANSMITTER, A) 2 MHZ LINEWIDTH, B) 500 KHZ LINEWIDTH...42 FIGURE 36: EYE DIAGRAM AFTER FIRST LOOP FIGURE 37: SPECTRUM AFTER FIRST LOOP FIGURE 38: DPSK BACK-TO-BACK EYE DIAGRAM OF A) 2 MHZ AND B) 500 KHZ LINEWIDTH...44 FIGURE 39: DPSK EYE DIAGRAM AFTER THE FIRST LOOP...45 FIGURE 40: DPSK SPECTRUM AFTER THE FIRST LOOP...46 FIGURE 41: DQPSK EYE DIAGRAMS IN THE BACK-TO-BACK CONDITION A) 2 MHZ, B) 500 KHZ FIGURE 42: DQPSK EYE DIAGRAM AFTER THE FIRST LOOP...47 IV

7 FIGURE 43: DQPSK SPECTRUM AFTER THE FIRST LOOP...48 FIGURE 44: DPSK EYE DIAGRAM AFTER FIRST LOOP...49 FIGURE 45: DPSK SPECTRUM AFTER FIRRST LOOP...49 FIGURE 46: SIGNAL POWER OF SELECTED CHANNELS RELATE ON THE NUMBER OF LOOP...50 TABLE 1: PARAMETERS OF ULTRA HIGH SPEED ELECTRONICS; [16]...35 V

8 List of Abbreviations AMI ARMA ASE ATM AWGN B BB BER BoD CSRZ CT CW D DBPSK DC DCF DI DPSK DWDM EDFA FBG FEC FTTx FWM HBT HEMT ifwm IPoWDM IT ixpm MA MZM NF NRZ NZ-DSF OBS OOK OPS OT OT OTDM P 0 PLC PM Alternate-Mark Inversion Autoregresive Moving Average Amplified Spontaneuos Emission Asynchronous Transfer Mode Additive White Gaussian Noise Channel bit rate Broad Band Bit Error Rate Bandwidth on Demand Carrier-Suppressed Return to Zero Communication Technologies Continuous Wave Fibre Dispersion Differential Binary Phase-Shift-Keyed Direct current Dispersion Compensating Fibre Delay Interferometer Differential-Phase-Shift-Keyed Dense Wavelength Division Multiplex Erbium Doped Fibre Amplifier Fibre Bragg Grating Forwad Error Correction Fibre to the various structure Four Wave Mixing Heterojunction Bipolar Transistor High Electron Mobility Transistor Intra Four Wave Mixing IP over WDM Information Technologies Intra Cross Phase Modulation Moving Average Mach-Zehnder Modulator Noise Figure Non Return to Zero Non Zero - Dispersion Shifted Fibre Optical Burst Switching On-Off Keying Optical Packet Switching Optical Communication Optical Technologies Optical Time Division Multiplex Threshold Power Planar-Lightwave-Circuit Phase Modulator VI

9 PMD QoS RZ RZ-AMI RZ-DPSK RZ-Duobinary SBS SDH SMF SNR SPM SRS WDM xdsl XPM xpon Polarization Mode Dispersion Quality of Service Return to Zero Return to Zero Alternate-Mark Inversion Return to zero Differential-Phase-Shift-Keying Return to Zero Duobinary Stimulated Brillouin Scattering Synchronous Digital Hierarchy Single Mode Fibre Signal to Noise Ratio Self Phase Modulation Stimulated Raman Scattering Wavelength Division Multiplex various Digital Subscriber Line Cross Phase Modulatio various Passive Optical Network A eff E 2 g B g R Effective Core Area Optical Intensity Inside the Fibre Peak Value of Brillouin-Gain Coefficient Raman-Gain Coefficient k 0 k 0i L L eff n (ω) n 2 β Δn NL (i) δω Φ NL (i) Φ NL ω Wave Number Wave Number of i th Carrier Fibre Length Effective Length Constant Part Given by Sellmeier Equation Nonlinear Refractive Index Coefficient Related to 3 rd Order Susceptibility Propagation Constant Nonlinear Refractive Index of i th Wavelength in XPM Frequency Chirp Nonlinear Phase Shift Nonlinear Refractive Index of i th Wavelength in XPM Anglular Frequency VII

10 1 Introduction The demand for higher transmission capacity and higher efficiency push optical communication (OC) to create more complex designs and solutions. In this process, a lot of various effects occur. Their impact on the transmission depends from particularly conditions. Optimal setting is a trade-off between introduced distortions due to phenomena and their relations in the optical fibre and our capabilities to suppress them. The possibility of pure physical experiments is unacceptable from more points of views. The response on the current situation is the application of simulation methods as one part of developing and implementation process. Simulation roughly estimates optimal parameters and then follows the tuning of physical device(s) or system(s). The simulation accuracy and therefore success of whole project depends from simulation models and conditions. At the beginning, credibility of simulation s outputs should be verified with basic experiments and then predicts results of desired design. Second chapter regards the physical phenomena in the optical fibre and within devices are briefly described. Their relative impact on the total transmission impairments is evaluated and possibilities of mitigation are considered as well. The events and factors with positive and also negative consequences on the OC are presumed Current massive trend in the transmission systems reflects the demand of longer distance and higher bit rates per channel. Substitution of modulation format has been following as one step to fulfillment of higher requirements. Primary used on-off keying (OOK) is compared with differential phase-shift keying (DPSK). Throughout diploma thesis, term DPSK refers to differential binary PSK sometimes referred to as DBPSK. Properties and applications of DPSK format consider chapter 3. PSK detection sensitivity differs from the sensitivity of intensity detection. The dominant limitation creates phase noise. Its origin results from the finite laser linewidth and nonlinear phase induced by Kerr effect. The specification of major contributor gives us system potential. This is the objective of chapter 4, where I want to specify OOK, DPSK and DQPSK format performance dependents from varying values of amplitude and phase noise represent by OSNR and laser linewidth for 10 Gb/s and 40 Gb/s. Then, impact of Kerr effect within NZ-DSF and SMF is evaluated. Assessment will be realized by simulation software for optical applications VPI Photonics, allowing wide range of WDM and component design. 1

11 2 Complex approach to optical communication 2.1 Physical characteristic Today s optical communication uses transmission medium, that is ultra lossless in the comparison to other technologies. It offers a bandwidth of 400 nm (equivalent of 54.5 THz), which is defined by loss <0.35 db/km (see Figure 1) [1]. It is equivalent to billions B channels or millions of TV channels. Limitations due to losses have got to do three mechanisms: Absorption: on the impurities and at IR area (Fig. 1) Scattering: predominantly formed by the Raleigh scattering Radiating losses: occur at the deep IR area and photonic crystal fibres» Bending loss: energy emission from the fibre on the turnings, micro and macro bends at very high wavelengths (Figure 1)» Seepage loss: energy release through slim cladding, which creates photonic band gap. Figure 1: Spectral variation of attenuation; Source [1]. Today s state of the art fibres are very close to their physical limit and next enhancement will be able at hollow core fibre based on the band gap photonic crystal fibre. It has got high losses due to simple manufacturing process. Water content in the preform and the drawn fibres is too high [2]. Another problem is a manufacturing of the finesse core surface. Conventional fibres improved over the past 20 years to their current 2

12 stage. First successful demonstration of hollow core fibre was in 1999 [3]. The question that arises is how many times do the developers need to overcome processing problems? 2.2 Mitigation of impairments and distortions Optical pulse is influenced by deterministic and stochastic processes as well during the whole transmission and which impact the signal features (shape, amplitude and time). The transfer functions describe deterministic part of the influence. Manufacturing and also installing processes and their imperfections can be just excesses from calculated value but they also introduce the stochastic behavior. The source of degradations has got linear and nonlinear base. Linear processes have an effect on the time domain characteristics. On other hand, nonlinearities processes influence the spectrum of the signal and they occur in the case of high power in fibre, which might define more points of view. Resulting effects have to be evaluated as a compact issue. The level of inevitable regeneration depends on the working conditions of transmission system, especially on the number of channels, bit rates and distance Nonlinear effects The response of any dielectric to light becomes nonlinear and intense electromagnetic fields and optical fibres are no exceptions. On a fundamental level, the origin of nonlinear response is related to anharmonic motion of bound electrons under the influence of applied field [4]. Nonlinearities are undesirable during transmission and have the highest impact at the beginning of amplifier span. They are a limiting factor to the long distance and high capacity transport systems. The source of nonlinearity is transmission medium (fibre) and not amplifier [5]. Next, two groups of elastic and inelastic nonlinear effects, which exist at fibre, will be described. The simplest way how to cope with nonlinear effects is to prevent them. This can be achieved via design of new networks, where proper fibre with large effective area can be applied. Otherwise other methods for specific issue at hand have to be used Kerr effect It is elastic in the sense that no energy is exchanged between electromagnetic field and the dielectric medium. It describes the intensity dependence of the refractive index. In the simplest form, the refractive index can be written as 2 2 n( ω, E ) = n( ω) + n ( E ) (2.1) 2 3

13 The intensity dependence of the refractive index leads to a large number of nonlinear effects and the two most widely studied are self-phase modulation (SPM), cross-phase modulation (XPM) and four-wave mixing (FWM) [4]. Figure 2: Products of Kerr effect; [6]. Self-phase modulation SPM refers to the self-induced phase shift experienced by an optical field during its propagation. It is responsible for spectral broadening of pulses as a consequence of the time dependence on nonlinear phase shift Φ NL. NL φ = nkle (2.2) The time dependence of its time derivation (marked δω) refers to frequency chirping, Figure 3 a). The Chirp induced by SPM increases in magnitude with the propagated distance Figure 3 b). The extent of spectral broadening depends on the pulse shape. 4

14 a) b) Figure 3: a) Temporal variation of SPM-induced phase shift and frequency chirp for Gaussian (dashed curve) and super-gaussian (solid curve) pulse. b) Experimentally observed spectra for a nearly Gaussian pulse at the output of a 99 m long fibre. Spectra are labeled by maximum phase shift related linearly to the peak power; [4]. Cross-phase modulation XPM is always accompanied by SPM when at least two optical frequencies are propagated simultaneously in the fibre. XPM stands for equally intense optical fields of different wavelengths the contribution of XPM to the nonlinear phase shift is twice that SPM. Nonlinear phase shift for the field at ω i is given by φ = nk Δ n (2.3) () i () i NL 2 0i NL Δ n = n E + E N () i 2 2 NL 2 i 2 j (2.4) j i Figure 4: Eye diagram of complete and incomplete collisions; [5]. 5

15 XPM induces timing jitter shown in Figure 4 [5] and it can lead to modulation instability, asymmetric spectral and temporal changes of copropagating optical pulses [4]. Four wave mixing FWM transfers energy from strong pump waves to new waves describes following equations: ω4 = ω1+ ω2 ω3 (2.5) β4 = β1+ β2 β3 (2.6) FWM induces amplitude noise, which is a source of serious degradation shown by Figure 5. However its efficiency is very sensitive on the phase matching [4]. a) b) Figure 5: a) nonlinear crosstalk due to FWM, b) eye diagram of amplitude noise due to FWM; [5]. Limitations due to Kerr effect have got different nature: Interchannel effects: involve phenomena mentioned above and dominate at 10 the Gb/s channel. These effects have got counterpart at anomalous dispersion regime and trade-off between both features is strong instrument of how to get optimal conditions of complex transmission system. Soliton transmission is possible and therefore very robust pulse over long distance can be achieved. But it has got a lot of issues that have to be overcome. More modulation techniques are available with the influence on the final result. Intrachannel effects: pulse spreading is high and nonlinear intersymbol interference becomes the major single-channel penalty, Figure 6. They dominate at 40 Gb/s and higher bit rate channels [7]. Dispersion induces intrachannel effects, so it is the limiting factor no more. Different design of dispersion map, which employs fibre 6

16 spans consisting of three or higher number of concatenated sections is useful. Another possibility supposes different modulation scheme based on the DPSK format [vsetky o DPSK] Figure 6: examples of intrachannel effects; [7] Stimulated scatterings Optical field transfers part of its energy to the nonlinear medium. Stimulated Raman scattering (SRS) and stimulated Brillouin scattering (SBS) belong to this group. They are related to vibrational excitation modes of silica. The main difference is that optical photons participate in SRS while acoustic photons participate in SBS. In a simple quantum-mechanical model applicable to both SRS and SBS, a photon of incident field is annihilated to create a photon at a lower frequency and a photon with the right energy and momentum to converse the energy and the momentum. Even SRS and SBS are very similar in their origin, different dispersion relations for acoustic and optical photons lead to some basic differences. Stimulated Raman scattering In any molecular medium, spontaneous Raman scattering can transfer a small fraction of power (~ 10-6 ) from one optical field to another, whose frequency is downshifted by an amount determined by the vibrational modes of the medium. Raman gain spectrum is specific for the material and for fused silica as shown in Figure 4. This behavior is caused by amorphous nature of the silica glass and it can be tuned by dopands. 7

17 Figure 7: Raman gain coefficient for pure silica and 1550 nm pump wavelength; [8]. SRS occurs when the pump power exceeds a threshold value and then builds up almost exponentially. The threshold power is explained as follows [4]: Aeff P0 16 (2.7) L g eff R Typical value of Raman threshold is approximately 500 mw of whole optical power. The influence of SRS on the WDM system is shown by Figure 5. Impact of SRS is not able to be reduced by dispersion as not Kerr effect by itself but the dispersion changes result due to intense presence of Kerr effect [4]. The way to suppress SRS is through power management in optical fibre. First decision is set the fibre type, mainly its A eff. Then variation of number of channels and therefore power level is available. SRS is a challenge and offers next dimension at WDM design. Figure 8: The outcome of SRS at WDM system; [5]. 8

18 Stimulated Brillouin scattering The pump field generates an acoustic wave through the process of electrostriction. The acoustic wave in turn modulates the refractive index of the medium. This pump-induced index grating scatters the pump light through Bragg diffraction. Scattered light is downshifted in frequency due to Doppler shift and propagates in the backward direction. SBS saturates maximum power per channel in the optical fibre (Figure 6). Figure 9: Behavior of input and output power due to SBS; [5]. SBS is a very narrow-band process and its spectral width of gain spectrum is ~ 10 MHz because it is related to the damping time of acoustic waves. Thus SBS occurs efficiently for CW pump or pump pulses whose spectral width is smaller than the gain bandwidth. The threshold power is explained similarly like that of the SRS threshold Aeff P0 21 (2.8) L g The threshold level predicted by (2.8) is only approximate. The effective Brillouin gain can be reduced by many factors and therefore SBS can be as low as ~1 mw for a CW pump and nearly ceases to occur for short pump pulses (width <10 ns) [4]. SBS can be suppressed by dithering laser frequency [5], but is not a limiting factor of today s transmission systems. eff B Amplification A signal after the passing of specific distance or device has to be amplified on the required level. Amplifiers set up the real part of device s transfer function and they are employed due to compensating losses or setting required power level. Qualitative characteristics of the amplifier determine the features of transmission system: 9

19 Noise figure: describes an inherent process of every amplifier. The noise generated during amplification has got broadband characteristic and its presence in the signal band decrease OSNR and transmission capabilities. The noise figure is defined: NF SNR SNR IN = (2.9) The maximum number of amplifier spans determines beginning value OSNR, noise figure of used amplifiers and required OSNR at receiver. Total transmission distance can be tuned by varying mentioned parameters. Bandwidth: is defined by frequencies that belong to 3 db fall from the peak gain. It limits the number of channels at WDM system. Gain: specifies the level of amplification for the current frequency and therefore amplifier span or repeater spacing. Longer repeater spans usually mean shorter total distance due to higher level of ASE. So long haul systems can have a half or less amplifier span against short haul. Gain flatness: it is spectral characteristic of the gain and describes difference of the lowest and the highest gain from the mean gain. High value induces strong selffiltering effect, which limits the using of concatenated amplifiers. It will lead to extremely high dynamic of received signals and drowning of low amplified signals in the noise. This parameter can be suppressed also after manufacturing by optical gain flatness filters which can be integrated at amplifier or connected on the amplifier s output. OUT The two amplification s schemes are suitable to use: 1. Lumped amplification: the amplifiers footprint is small comparing to the transmission size. It is up to 1 km length of fibre. Amplifying is based on the 3 or 4 level model. 2. Distributed amplification: the transmission fibre is also an amplifier simultaneously. Raman Effect and 3 or 4 levels model is considered as well. Doped fibre amplifiers The stimulated emission of the proper quantum jump, which has to be similar to the transmitted pulses, conducts to the amplification. Rare earth elements are necessary 10

20 part of structure. It is tuned by appropriate dopants. Today s knowledge of dopants and designs reduces their applications on the C, L, O and S band (see Figure 1, Figure 7). These amplifiers have got high gain and power conversion and the number of operating windows is sufficient for traffic demand. On the other hand, the minimum level of noise figure is 3 db, what can be limit for transoceanic and transcontinental applications. It influences the length of amplifier s span. Thus, more amplifiers have to be deployed with the impact on the cost and reliability. The issue of distributed doped amplifiers still has been only at theoretical models. Raman amplifiers Raman amplifier employs SRS and consists from any transmission fibre and group of pump lasers lasing on the proper frequencies, which are shifted to upper values to match maximum Raman gain, 14 THz for silica glass (Figure 8). It can be created by one laser as well, what depends on the desired parameters. Its operating band is set up by pump s wavelength and spectral characteristics are influenced by fibre properties (dopants and refractive index profile). Raman amplifier is able to amplify whatever band limited only available laser wavelengths and it is not any barrier. Figure 10: Characteristic of Erbium amplifier (left) and Raman amplifier (right); [7]. If Raman amplifiers want to reach sufficient gain, they will have to use very high pump power. Therefore pump efficiency is lower against doped fibre amplifiers moreover strong Kerr effect occurs with high powers and WDM crosstalk is another issue. On the other hand they allow to provide negative effective NF in the case of amplification within the transmission fibre [8]. 11

21 The latest trends in optical amplification have tended to focus on Raman amplifiers and doped fibre amplifiers are on the edge of interest. Both types have got unique features, which don t exclude co-operative realization. Hybrid schemes don t have adequate attention. They are able to offer outstanding results (Figure 4) and their potential should be explored deeper. It will probably happen later, when the requirements on the transmission will reach higher levels. Figure 11: Various hybrid amplification schemes; [7] Shaping The shape describes modulated envelope applied on the optical carrier. Pulses used at OC have got finite linewidth and therefore compose from the band of frequencies. Optical part of spectrum is affected by the dispersion at any material. The dispersion hits only time domain characteristic of the pulse that assigns individual velocity to every frequency. It broadens the pulse width during its transmission while the spectrum has been unaffected It would result to intersymbol interference. It is metaphorically demonstrated by Figure 5. 12

22 Figure 12: The dispersion s influence; [7]. The sequence of frequencies depends from the dispersion regime: Anomalous dispersion regime: blue part of pulse goes faster than red part. Normal dispersion regime: red part of pulse goes faster than blue part. The dispersive influence of the optical fibre is called Chromatic dispersion that consists from two parts: Material dispersion: characterizes dispersion properties of the current material. Waveguide dispersion: is caused by different refractive index in the core and cladding due to pulse propagation in the core and also in the cladding, where it has different velocities. It is variable and permits the tuning by the refractive index profile. It describes Figure 13. Figure 13: Material, waveguide and chromatic dispersion for current fibres; The issue of chromatic dispersion is more important as the bit rates per channel increase. For a system penalty of 1 db, the bit rate, dispersion and distance are related as follows [10]: The dependence of system B DL 10 ps / nm( Gb / s) (2.10) 13

23 It means that four times higher bit rate allows sixteen times smaller tolerance than earlier value of tolerance. The need of dispersion compensation can accomplish more technologies: Dispersion compensating fibres: use fibre with negative chromatic dispersion joined after typical fibre and creates part of route. Dispersion is tuned by waveguide dispersion and modern DCF offer similar value of attenuation and high negative chromatic dispersion but effective core area is smaller due to refractive index profile. Recently DCF has overcome other way of dispersion compensation and it is the most frequent technique. Fibre Bragg grating: short piece of fibre up to one meter with concatenated refractive index pattern, which is designed to refract the incoming frequencies in the aim to clear their group velocity dispersion. FBG imposes degradation, which can express NF and it has got high thermal sensitivity and small effective core area [red book]. It has to use the circulator for proper work. circulator Spread pulse Chirped FBG Figure 14: Dispersion compensation by FBG; [9]. Optical filters: contain several filter approaches. One of them is already mentioned FBG, which represents ARMA response and other possibilities is planar waveguide that meet MA and optical all-pass filter with constant magnitude response. Those alternatives are used mainly at dispersion equalization [10]. Soliton transmission: use optical pulses with special shape and power level, where SPM and dispersion are at equilibrium. Technical issues haven t allowed commercial deployment. 14

24 Dispersion compensation can perform symmetrical or non symmetrical management. The first possibility pre-compensates on the inverse value of dispersion maximum. On the other hand, non symmetrical technique works only at positive interval. On the Figure 10, there are the results of simulation (left one) and experiment (right one) of symmetrical and non symmetrical dispersion compensation in the centre (upper row) and in the end (lower row) of the span [5]. It is clearly seen that symmetrical compensation gives better conditions for transmission.... a) b) c) d) Figure 15: a), c) symmetrical dispersion management; b), d) non symmetrical dispersion management; [5]. Dispersion slope: is third member in the Tayler expansion of the propagation constant. It assigns different value of the dispersion for the various frequencies Figure 11. It introduces next parameter, which characterizes property optical fibre and it has to be compensated as well. The dispersion compensation doesn t mean dispersion slope compensation and it is equalized at the end of the transmission. Figure 16: The effect of dispersion slope for border and central frequencies; [5]. 15

25 Polarization mode dispersion: PMD is has got different principle against phenomena mentioned above. Its stochastic character gives atypical features compare with previous effects. Whole impact is described by:» 1 st order PMD: wavelength independent, time variant and increases with root of the length.» 2 nd order PMD: wavelength dependent and occurs after suppression of 1 st order PMD. The source of PMD is random birefringence along the link and in the components. It causes various velocities of both parts of linearly polarized mode. Consequently the pulse spreading occurs (Figure 12). It becomes serious issue from 10 Gb/s and it can induce temporally high increase of BER. Thus, PMD is important limitation factor at high speed networks. Figure 17: PMD due to birefringence effect; [11]. The best way of PMD suppression is to prevent induction of birefringence by the using of precise technological and installation techniques. Then PMD equalizers can be employed at the end of link for single channel [4] or low speed polarization-scramblers at the beginning of the fibre, which should decrease the PMD on the tolerant level [12], [13]. The last possibility is more efficient than PMD equalizers due to its application on the whole WDM band Jitter Jitter is stochastic process of every component within transmission system. This section considers just optical contribution to the jitter. PMD is the strong contributor of jitter and its origin and suppression was described earlier. Frequency jitter of laser 16

26 linewidth is transferred to the timing jitter by dispersion of the fibre. Thus, precise and stable lasers and filters will produce less timing jitter [5]. Another component is fluctuations of refractive index due to high powers and interaction with dispersion. Symmetrical dispersion compensation is suitable to remove this effect showed by Figure 13. But this dispersion scheme has to design with amplification scheme and consider amplifier position due to maintenance of low Kerr effect. Bit-pattern-dependent effects can remove modulation format with constant power level at every state. Figure 18: Cancellation of timing jitter with symmetric dispersion profile; [5]. 2.3 Modulation format The information can transmit signal s amplitude, phase, frequency or combination of those features. Huge bandwidth offered by optical fibre and high value of laser s linewidth have established OOK like gold standard. It offers three variations, NRZ, RZ and CSRZ, which are more resilient to some effects and also more sensitive to some of them. NRZ OOK: the most tolerant on the chromatic dispersion and optical filtering. RZ OOK: the best performance against intra-channel nonlinear effects after transmission due to the best performance in back-to-back. CSRZ OOK: the superior resilience to nonlinearities and represents satisfactory opinion to meet all key requirements for future all-optical networks [14]. It seems that basic features of the OOK don t match demands for future all-optical networks. More precise lasers with narrower linewidth have opened possibilities of phase shift keying. Today s situation is suitable for the developing of DPSK transmitters with 17

27 balanced receiver due to its economical efficiency. DPSK overcomes OOK at following points: 3 db lower required OSNR for the same BER (theoretically) 3 db lower peak power Sensitive on the phase noise More resistant against nonlinear effects More robust at higher bit rates 2.4 Economical aspects If we don t want to do research for research, optimal implementation of every technology will have to meet following rule, where optical communication (OC) and its basic, pillar optical technologies (OT), is not an exception: Solution = ƒ (service) OC is a part of global market and tries to meet its demand. The predictions of development are very volatile and whole process of the creating of estimation is difficult. I specify main factors affects on the OT s boom and recession Catalysts of OT Traffic growth Since the year 2000, global traffic has been dominated by internet data. Studies of traffic of the year 2005 indicated a global traffic growth of 115% per a year. Estimation of Atlantic traffic through 2025 is illustrated by the Figure 19. Other driving factors would be globalization of FTTx-xPON access for Mb/s BB solutions and installation and services cost reduction to the current level of xdsl as we are witnesses nowadays [1]. 18

28 Figure 19: Comparison of IP traffic predictions with installed capacity; [1]. The open issue has been the conditions of IP traffic growth, which stimulates demand for higher quality multimedia. There will probably be the ultimate limit that average user will be able to comprehend, but this has yet to be reached [15]. Service convergence and their adaptability to NGN Sectors of IT and CT have converged to the common platform due to removing of interworking restrictions and creating open platform environment by the way. This process has developed with success and disappointment as well. The cheapest alternative based on IP has been chosen as the basic transport protocol that is not optimal solution for switching circuit services requiring guaranteed parameters of transmission. Therefore upgrades to the connection oriented services have to be implemented and they need approximately up to twice bandwidth compared to legacy systems and adequate margin in the congestion protection. Those factors have brought better managing and flexibility of services, unclear savings and on other hand higher requirement on the capacity of core network. Introduction of new services This area includes the roll-out of new services or improving their existing quality. The main objects of interest are multimedia and real time applications. The latest demonstration is Triple Play. Its strategy is to offer voice, Internet and video through one data access. The potential is huge and there will be a lot of modifications. Dimension of that service is not able to realize without OT. There is also opportunity for radio and metallic connections but only as final section between customer and multiplexer. 19

29 2.4.2 Inhibitors of OT Providers concentration on the short-time objectives Low-cost solution with fast market returns have become the key driving concerns of OC innovation. This emphasis prevents industrial and academic research from exploring technologies that are deemed to immature for short-term deployment and that have no immediately business value. The phenomenal growth of capacity in recent wireless BB services has provided a perception that bandwidth is infinite. But new driving factors to appear in the next 3-7 years, should steer OC industry, namely: exploding bandwidth demand, lightwave capacity exhaustion and facing ultimate technology limits [1]. Establishing of the new layer model One way how to satisfy the demand for higher capacity is revolutionist change of interconnection s model and direct mapping from 3 rd layer onto physical medium with new non redundant management either IP or optical layer. This approach represents IPoWDM and it would replace mapping through ATM and SDH to the WDM layer. This technique crashes on the deep penetration of the current technology and it would be alternative solution in the extreme cases. Technological development at cooperative sectors OC is a one ring in chain, which brings the service from the server to the customer s application. The reliability of service providing determinates the crucial process as the chain s strength depends from the weakest ring. This idea includes negative and also positive development of the similar technologies. Huge progress on the field of DSP at 1990s meant stop for commercial deployment of ultra high speed optical systems due to enormous video compression. Insufficient development of server performance would restrict data traffic and it will decrease progress of whole ICT sector. Other reasons This section should describe phenomena that haven t got rational base or I haven t recognized. One example of my ideas is technological deflation at the end of 20 th Century. It still has been finding the lost goodwill of the investors. 20

30 3 Differential phase-shift keyed format Phase-shift-keyed (PSK) formats carry the information in the optical phase itself. The receiver has to compare detected signal with reference signal and extract information. Due to the lack of an absolute phase reference in direct-detection receivers, the phase of the preceding bit is used as a relative phase reference for demodulation. This results in DPSK formats, which carry the information in optical phase changes between bits. DPSK has got several advantages against ASK modulation thus it is not surprising, that many of the recent long-haul WDM transmission records at per-channel rates of 10 and 40 Gb/s are now held by systems based on DPSK. Optical systems based on DPSK are not new. DPSK was extensively studied in the late 1980s and early 1990s for use mainly in single-span fiber-optic systems employing coherent receivers as well as in the context of free-space optical communications, where the 3-dB sensitivity advantage over OOK could be exploited. When erbium-doped fiber amplifiers (EDFAs) were introduced, interest in coherent systems declined. For about a decade, OOK-based WDM systems using optical-amplifier repeaters dominated the research in long-haul optical communications. Interest in DPSK reemerged several years ago, as WDM systems were pushed to ever-higher levels of performance [16]. 3.1 DPSK format In the DPSK format, optical power appears in each bit slot, with the binary data encoded as either a 0 or π optical phase shift between adjacent bits. The optical power in each bit can occupy the entire bit slot (NRZ-DPSK) or can appear as an optical pulse (RZ-DPSK). The most obvious benefit of DPSK when compared to OOK is the ~ 3-dB lower OSNR required to reach a given BER. This can be understood by comparing the signal constellations for DPSK and OOK, as shown in Figure 20. For the same average optical power, the symbol distance in DPSK (expressed in terms of the optical field) is increased by 2. Therefore, only half the average optical power should be needed for DPSK as compared to OOK to achieve the same symbol distance. This ~ 3 db benefit of 21

31 Figure 20: Signal constellation of OOK and DPSK modulation; [16] DPSK modulation can be only extracted by using balanced detection. In practice, and neglecting the loss of the optical preamplifier s input optical isolator, a receiver sensitivity of 60 photons/bit has been reported for a 10-Gb/s RZ-OOK signal. Using an RZ-DPSK signal and a balanced-photodiode detection scheme, the sensitivity was improved to 30 photons/bit. At 42.7 Gb/s, a sensitivity of about 38 photons/bit has been reported using RZ-DPSK. Again, this is approximately 3 db better than the best OOK results of 78 photons/bit. The lower OSNR requirement of DPSK can be used to extend transmission distance, reduce optical power requirements or relax component specifications. DPSK with balanced detection has been demonstrated to offer large tolerance to signal power fluctuations in the receiver decision circuit because the decision threshold is independent of the input power. DPSK is more robust to narrow-band optical filtering than OOK, especially when balanced detection is employed. Numerical simulations and experiments have shown DPSK to be more resilient than OOK to some nonlinear effects. This results from the fact that: i) the optical power is more evenly distributed than in OOK (power is present in every bit slot for DPSK, which reduces bit-pattern-dependent nonlinear effects) and ii) the optical peak power is 3 db lower for DPSK than for OOK for the same average optical power. Finally, an extension to differential quadrature phaseshift keying (DQPSK) and other multilevel formats should enable higher spectral efficiency and greater tolerance to chromatic- and polarization-mode dispersion [16]. 22

32 3.2 DPSK transmitter Two commonly used RZ-DPSK transmitter setups are shown in Figure 21. The transmitters consist of a continuously oscillating laser followed by one or two external modulators, typically based on technology. Phase modulation can either be performed by a straight-line phase modulator [PM, Figure 21 (a)] or by a Mach Zehnder modulator [MZM, Figure 21 (b)]. A PM only modulates the phase of the optical field, resulting in aconstant-envelope optical signal [see measured sampling scope traces in Figure 21 (a)]. Since phase modulation does not occur instantaneously, a PM inevitably introduces chirp across bit transitions [see symbol diagram in Figure 21 (a)]. A sinusoidally driven second modulator ( pulse carver ) may be used to carve pulses out of the phase-modulated signal, thus generating RZ-DPSK. The inset to Figure 21 (a) shows the resulting optical power waveform. (The seemingly limited extinction ratio of the RZ-DPSK pulses is a measurement artifact, caused by detecting a 40-Gb/s signal using a 32-GHz-bandwidth photodiode.) When using a MZM for phase modulation, the modulator is biased at its transmission null, and is driven at twice the switching voltage required for OOK modulation. If a z-cut MZM is used, it is driven in push-pull configuration to minimize chirp, whereas an x-cut modulator requires only a single electrical drive. Since the phase of the optical field changes its sign upon transitioning through a minimum in the MZMs power transmission curve, two neighboring intensity transmission maxima have opposite optical phase, and a near-perfect 180 phase shift is obtained, independent of the drive voltage swing. As can be seen from the symbol diagram in Figure 21 (b), the benefit of highly accurate phase modulation comes at the expense of some residual amplitude modulation at the transition of two bits, with the width of the resulting intensity dips depending on the drive signal s bandwidth and voltage. However, since DPSK encodes information in the optical phase rather than in the intensity, these dips are of reduced importance, especially for RZ-DPSK, where the pulse carver cuts out the amplitudemodulation-free center portions of the bits only, and thus largely eliminates any residual dips [16]. 23

33 Figure 21: Typical DPSK transmitters: a) implementation with PM, b) implementation with MZM; [16] Transmitter evaluation Transient chirp for PM-based DPSK transmitters and intensity dips for MZMbased transmitters, the effects of drive waveform imperfections are worth mentioning. The nonlinear (cosine) transmission curve of the MZM ameliorates the impact of drivewaveform overshoots or of limited drive-signal rise times. Any remaining imperfections are only translated into optical intensity variations, but the information-bearing optical phase is left intact. On the other hand, using a PM for phase modulation, any drivewaveform imperfections get directly mapped onto the optical phase, thus potentially degrading performance. Difference of two modulators will arrive, if the driver output power and the combined driver-plus-modulator bandwidth cannot be chosen arbitrarily high, which is the case in practice, especially for high-data-rate systems. [16] Pulse carver Since DPSK carries information in the phase of the optical signal, optical phase distortions (such as chirp) will have a severe impact on DPSK receiver performance. At the transmitter, phase distortions may be caused by imperfect pulse carvers. In order to operate chirp-free, a dual-drive MZM pulse carver has to have infinite DC extinction, and 24

34 has to work in perfect push-pull operation, i.e., the sinusoidal drive amplitudes have to be of the same amplitude and of opposite phase. Any deviation from this ideal condition inevitably produces chirp. Figure 22 (a) shows three commonly used ways of pulse carving by applying a sinusoidal drive signal to a MZM-based pulse carver. Figure 22: a) A typical RZ-DPSK transmitter. (b) Optical intensity and phase waveforms generated by an imperfect pulse carver; [16]. Three important facts are evident from the optical intensity and phase waveforms shown in Figure 22 (b): First, when sinusoidally carving at the data rate (50% RZ), the residual optical phase variations are identical for each bit, while they are different for adjacent bits when carving at half the data rate (33% and 67% RZ). Since it is the difference between the optical phase of two adjacent bits that is used to decode DPSK signals at the receiver, higher degradations due to pulse carver chirp are found for 33% and 67% duty cycle RZ-DPSK than for 50% RZ-DPSK. Second, we see from the opposite phase curvatures (50% and 67%) or slope (33%) that chirp due to finite DC extinction ratios of the MZM can partially be compensated by imbalancing the drive amplitudes. Third, we notice that for 33% RZ a drive-signal amplitude imbalance leads to linear phase transitions (i.e., to optical frequency shifts) at pulse center, while a drivesignal phase error produces a phase offset at pulse center. Since pure bit-alternating frequency offsets do not disturb the phase difference between adjacent bits at pulse center (where the intensity is highest, and thus the contribution to the demodulated signal is largest), a higher tolerance is found for drive amplitude imbalance than for drive phase errors in the case of 33% RZ. For 67% RZ, the situation is opposite, and we find a higher 25

35 tolerance to drive phase errors than to drive amplitude imbalance. Experimental as well as numerical quantifications of pulse carver tolerances for RZ-DPSK can be found in [16]. 3.3 DPSK Receiver A typical balanced DPSK receiver is shown in Figure 23. The optical signal is first passed through a Mach-Zehnder delay-interferometer (DI), whose differential delay is equal to the bit period. This optical preprocessing is necessary in direct-detection receivers to accomplish demodulation, since photodetection is inherently insensitive to the optical phase; a detector only converts the optical signal power into an electrical signal. In a direct-detection DPSK receiver, the DI lets two adjacent bits interfere with each other its output ports. This interference leads to the presence (absence) of power at a DI output port if two adjacent bits interfere constructively (destructively) with each other. Thus, the preceding bit in a DPSK-encoded bit stream acts as the phase reference for demodulating the current bit. Ideally, one of the DI output ports is adjusted for destructive interference in the absence of phase modulation ( destructive port ), while the other output port then automatically exhibits constructive interference due to energy conservation ( constructive port ). For the same reason, the two DI output ports will carry identical, but logically inverted data streams under DPSK modulation. Figure 23: DPSK receiver; [16]. 26

36 Careful analysis of the optically demodulated signals at the DI output reveals that the constructive port carries duobinary modulation, whereas the destructive port carries alternate-mark inversion (AMI). Today, technical difficulties in implementing stable delay interferometers have been overcome, and DIs have been demonstrated both in fiberbased and in planar-lightwave-circuit (PLC) technologies. Fine-tuning of the differential delay to match the laser center frequency and achieve good interference quality is typically achieved using a heating element on one of the interferometer arms. Also, polarization-dependent phase shifts within the DI have to be avoided. Since both DI outputs ports carry the full (logically conjugated) information, they can be either detected by themselves ( single-ended detection ), or connected to two photodiodes using a balanced receiver (see Figure 23). Identical path lengths between the output coupler of the DI and the point of subtraction within the balanced receiver can be achieved using variable optical delay units or photonic integration of the detectors with the DI. Alternatively, separate detection of both output ports in combination with joint digital signal processing can be applied. The 42.7-Gb/s eye pattern at the output of the balanced-detector circuit obtained in our experiments is shown in Figure 23 [16] Balanced versus single-ended detection Introduction of a detector amplitude imbalance, β, is defined as S β = S A A S + S B B (3.1) where S A and S B are the overall opto-electronic conversion factors for the destructive (A) and constructive (B) DI output ports, respectively. Balanced detection is achieved for S A =S B, while detection of the constructive (destructive) port alone is found for β=1(β=-1). Figure 24 (solid curve) shows numerical calculations for the required OSNR at BER = for 33% RZ-DPSK as a function of receiver amplitude imbalance β. It can be seen that a balanced DPSK receiver performs about 2.7 db better than its single-ended counterpart. Also shown (dashed curve) is the required OSNR at BER = for OOK, which is (by definition) independent of β, and comparable to the OSNR needed for single-ended detection of DPSK. This shows that the frequently cited 3-dB benefit of DPSK over OOK, neither is exactly 3 db, nor is a property of DPSK alone; it is a property of the modulation format in combination with the detection scheme. 27

37 Figure 24: numerical calculations for the required OSNR at BER = for 33% RZ-DPSK and OOK as a function of receiver amplitude imbalance β; [16] Balanced detection in a beat-noise-limited scenario has to be numerically modeled using the exact probability density functions (PDFs) of detection noise rather than Gaussian approximations to these PDFs, as is common practice and works well for single-ended OOK receivers. If Gaussian PDFs are used to represent the noise statistics at the decision gate, we obtain the dotted curve in Figure 24. While being reasonably accurate for singleended DPSK detection, the Gaussian approximation (as well as all simulation techniques based on the Gaussian noise assumption, such as a standard -factor analysis) is bound to fail in predicting balanced DPSK receiver performance. The reason for this important simulation aspect rests in the fact that the tails of the exact (chi-square-like) PDFs differ significantly from the tails of the Gaussian distributions. For single-ended detection of DPSK as well as for OOK, this difference in the PDFs, by pure numerical coincidence, cancels to a high degree of accuracy when calculating BER. In contrast, this beneficial cancellation is not found for balanced receivers, owing to the different nature of detection: a single-ended receiver compares a single, noisy signal against a deterministic (non noisy) threshold to retrieve the digital data, while a balanced receiver essentially compares two noisy signals against each other [16]. Qualitative physical aspects affected on the detector imbalance include amplitude imbalance, temporal receiver imbalance, interferometer extinction, delay, phase error, frequency offsets. Their quantitative impact is shown by Figure

38 Figure 25: Penalties in non ideal RZ-DPSK receivers: (a) Amplitude imbalance in the balanced detector (dashed curve is for a delay interferometer with an extinction ratio of only 10 db), (b) Phase imbalance in the balanced detector, (c) Delay-to-bit rate mismatch in the delay interferometer, (d) Laser frequency offset from the ideal as set by the interferometer phase difference (dashed curve is for 33% RZ-DPSK). Circles are experimental results; [16] Tolerance to Optical Filtering As mentioned previously, DPSK is more tolerant of tight optical filtering than OOK. The reason for the good performance can be attributed to the use of ISI-tolerant RZ coding and higher robustness of balanced DPSK to reduced optical filter bandwidths. Figure 26: Experimental results of penalties from narrow optical and electrical filtering; [16] 29

39 Figure 26 shows 40-Gbit/s measurement results for 33% RZ-DPSK. The BER target was The gain of balanced DPSK reception over single-ended detection is seen to be some 4 db, and increases to over 5 db at low optical bandwidths, where both OOK and the destructive DI output port show severe penalties [16]. 3.4 DPSK transmission at 10 Gb/s In a linear system employing optical amplifiers, the 3-dB DPSK advantage over OOK would double the achievable distance by allowing the accumulation of twice as much amplified-spontaneous-emission (ASE) noise. Limitations from chromatic dispersion (CD) and polarization-mode dispersion (PMD) are similar for DPSK and OOK signaling, some chromatic dispersion advantage has been reported for NRZ-DPSK. However, transmission performance in fiber is also affected by the Kerr nonlinearity. This is exhibited as four-wave mixing (FWM), self-phase modulation (SPM) and cross-phase modulation (XPM).The extent of these effects depends on several system design factors, including average optical power, peak optical power, modulation format, transmission regime (pulse-preserved or pulse-overlapped), and the nonlinear interaction of signal with ASE noise. As in OOK systems, dispersion management can be used in DPSK systems to reduce the FWM efficiency among WDM channels to low levels. Therefore, interchannel FWM is generally not a concern. SPM and XPM affect DPSK signals somewhat differently than OOK signals. In SPM, the intensity variations of an optical signal modulate the signal s optical phase via the nonlinear refractive index, causing a red shift on the rising edges of pulses, and a blue shift on the falling edges. The effect is to broaden the signal spectrum. The broadened signal spectrum, combined with dispersion, then broadens the received pulses, introducing a transmission penalty (although we note that solitons balance SPM and dispersion to maintain pulse shape). For DPSK signals, an additional effect is important, because the information is carried by the optical phase. Noise-induced power fluctuations are converted into phase fluctuations by SPM, and become a source of transmission penalty. This nonlinear interaction of signal and noise is referred to as the Gordon Mollenauer effect. In the nonlinear regime, performance should be substantially different in the two cases for a DPSK signal, while remaining essentially unchanged for an OOK signal. After nonlinear transmission, the Q -factor of the DPSK signal depended strongly on the transmitter OSNR, whereas the Q -factor for the OOK signal did not. 30

40 In XPM, the intensity of one signal modulates the phase of another. In OOK systems, the collisions of WDM signals passing through each other impart phase variations that, when combined with dispersion, result in pattern-dependent timing jitter of the received pulses. Complete collisions at near-constant power cause less harm, as the phase variations caused during the first half of the collision are largely undone in the second half. However, as WDM channels are placed closer together, the difference in propagation speed between adjacent channels becomes lower, the pulses move through each other more slowly, and partial collisions increase the XPM penalty. DPSK signals, however, exhibit power in every bit slot. Therefore, all pulses in a given WDM channel experience similar collisions, mitigating the XPM effect. Of course, a second-order effect is expected, as noise-induced amplitude fluctuations on pulses in one channel cause phase fluctuations in another. Generally speaking, it appears that long-haul 10-Gb/s singlechannel OOK systems can outperform DPSK systems, which are limited by the Gordon- Mollenauer effect, although we stress that single-channel performance is highly systemdependent, and that there can be cases in which DPSK will outperform OOK. In 10-Gb/s WDM systems, both experimental measurements and computer simulations indicate that DPSK and OOK perform similarly at a spectral efficiency of 0.2 b/s/hz. At a spectral efficiency of 0.4 b/s/hz and higher, DPSK, due to its increased robustness to XPM, can outperform OOK. However, we again stress that system performance is dependent on many factors, including channel power and dispersion map [16]. 3.5 DPSK transmission at 40 Gb/s At 40-Gb/s, single-channel effects in the pulse-overlapped (pseudolinear) regime mainly limit signal transmission. In particular, intrachannel FWM (ifwm) transfers power between bit slots as pulses disperse into each other and mix due to fiber nonlinearity. The effect in OOK is amplitude fluctuations on the 1s, and ghost pulses (residual power) on 0s. In DPSK systems, the phase fluctuations from this mixing are more detrimental than amplitude fluctuations. In intra-channel XPM (ixpm), intensity fluctuations of the dispersed, overlapped pulses modulate the optical phase. The effect in OOK is timing jitter when combined with dispersion, while in DPSK, both the timing jitter and the phase fluctuations are detrimental. As mentioned earlier, undistorted DPSK has 3-dB lower peak power than OOK for a given average power, due to having power in each bit slot. In the pulse-overlapped regime, nonlinear DPSK penalties can be reduced because of this more smoothly distributed power. Also, correlation between the nonlinear 31

41 phase shifts experienced by adjacent bits (due to experiencing a similar environment in transmission), combined with differential detection and should reduce nonlinear DPSK penalties. Experimental results have consistently shown better performance for DPSK than OOK in 40-Gb/s single-channel and WDM systems [16]. 3.6 DQPSK application There have been a number of applications that have been proposed and demonstrated for PSK. One of these is to increase spectral efficiency through the use of multilevel signaling. In particular, DQPSK has recently received intense study. The most widely used implementation of a DQPSK transmitter and receiver is shown in Figure 27. The transmitter consists of two parallel DPSK modulators that are integrated together in order to achieve phase stability (a serial arrangement is also possible, and has been used Figure 27: DQPSK transmitter and receiver; [16]. in experimental demonstrations). The receiver essentially consists of two DPSK receivers, although the phase difference in the arms of the delay interferometers is now set to +π/4 and -π/4. The benefit of DQPSK is that, for the same data rate, the symbol rate is reduced by a factor of two. Consequently, the spectral occupancy is reduced, the transmitter and receiver bandwidth requirements are reduced, and the chromatic dispersion and PMD limitations are extended. As compared to DPSK, the required OSNR to reach a given BER is increased by about 1 2 db, depending on the BER. Also, the frequency offset tolerance between the laser and the delay interferometer is about six times less than for DPSK, making the DI design and stabilization somewhat challenging. Even higher spectral efficiency can be achieved using various combinations of phase- and amplitudeshift keying. Such multilevel modulation can also improve system tolerance to chromatic 32

42 dispersion and PMD. However, these schemes quickly become quite complicated to implement, require higher OSNR, and are sensitive to nonlinear phase noise [16]. 33

43 4 Trends in the optical communication Development of OT has got good drive although it doesn t reach its maximum from the last decade of 20 th Century. History, the newest requires and economical background have formed design of OC. One clearly defined direction is price cut and rising bit rates per channel up to 40 Gb/s. Those hints can realize more ways and their short description follows. 4.1 Transmission bands and spectral efficiency The choice of transmission band is not very difficult. Physical properties of existing infrastructure and compensation technique lead to the C and L band exploitation. They are feasible in the case of required capacity, availability and reliability. Higher requirements can handle two alternatives that represent the using of another band or improving spectral efficiency. Second variant has applied till now. It includes narrower channel spacing and multilevel modulation. Its potential is finite and occupying of next band will come sooner or later. 4.2 Multiplexing techniques The growing demand for higher bit rates can accommodate two different multiplexing techniques, which have got different approaches of the exploitation of offered bandwidth. Another fact is enhanced characteristics of available sources including narrower linewidth Time division multiplex (OTDM) OTDM is serial transmission of number time delayed channels with lower bit rates to the common channel at one wavelength. Key parameters are at time domain. Its huge bandwidth is very sensitive on the value of total dispersion and timing jitter. This one channel hasn t got enough power to induce nonlinear effects. OTDM allows dynamic allocation of offering transmission capacity, BoD service. Then, it supports QoS and optical burst switching (OBS) and optical packet switching (OPS) are supposed but every device has to work at one data format and their number may be high. 34

44 Optimal deployment of OTDM is metropolitan and access networks, where bit rate flexibility is required. Four times higher speed WDM requires four times higher number of link terminated devices against the same number of link terminated devices at OTDM. On the other hand, OTDM requires precise dispersion management with low PMD and higher OSNR. The transmission capacity of OTDM reaches only fragment of DWDM performance due to maximum frequency values and lack of capabilities for processing in optical domain. The parameters of the state of the art electronics is shown by Table 1 [17]. HEMT HBT GaAs InP SiGe InP Mux/Demux Gb/s G V PP Fujitsu, 2 IBM, 3 NTT, 4 Lucent Table 1: Parameters of ultra high speed electronics; [17] OTDM hasn t reached required feasibility and the metro and access networks use others technological solutions Wavelength Division Multiplex WDM is parallel transmission of number of channels within one fibre. The key parameters are at spectral domain. Thus, nonlinear effects and linewidth of sources are cardinal issues. The chromatic dispersion and PMD become crucial at long haul systems. Basically, it is circuit switching technique, which forms its characteristics such as higher employed bandwidth of the fibre, the channel independence, data format transparency. The optimal application for WDM deployment is at core and backbone networks. 4.3 Simulation DQPSK format BER curve of PSK modulation describes Figure 30 a). There are experimental results of DQPSK format [18], and the object of investigation was impact of various values of OSNR and laser linewidth on the narrowband filtered DQPSK modulation in 35

45 the back-to-back configuration. It represents BER of the transmitter. I have used it as verification of my simulation model. Its experimental setup is shown by Figure 28. Figure 28: Experimental setup of; [18]. I created equivalent setup to its scheme and it is shown by Figure 29. I used generator of additive white Gaussian noise (AWGN) instead of optical amplifier. Figure 29: Simulation scheme of BER = ƒ (OSNR). NRZ-DQPSK source consists from CW laser followed by two dual stage Mach- Zehnder modulators. The used bit rate was 10 Gb/s. The source output power and power density of AWGN was set on the values, that corresponded to OSNR = 5 db. Attenuator modified level of OSNR in the range from 5 to 20 db. Signal passed through 1. order Gaussian bandpass filter with 6 GHz linewidth (spectral efficiency equals to 1.67 b/s/hz), what was identical with experiment. Signal was demodulated in the optical delay line demodulator and detected at balanced receiver. At the end, BER was assigned to every value of OSNR. It was done for 0, 10, 20, 50 and 100 MHz laser linewidth. The simulation results are shown by Figure 30 b). 36

46 a) b) Figure 30: Comparison of experimental [18] and simulation output. Maximum deviation was in the range of one order, what gave sufficient likelihood. Next step involved increasing of bit rate on the 40 Gb/s and filter bandwidth expansion on the equal value of spectral efficiency. The verification was based on the two assumptions: 1. theoretical prediction that the same BER value for 40 Gb/s transmission requires 6 db higher OSNR level compared to 10 Gb/s transmission due to noise power 2. application of previous prediction and obtained results from first simulation. The results for 10 and 40 Gb/s are shown by Figure

47 Figure 31: BER vs OSNR for 10 Gb/s and 40 Gb/s DQPSK format. The previously mentioned assumptions were not accomplished completely. Only ideal source met prediction completely. Other curves haven t got typical inflexed point and don t converge to typical value for current phase noise (corresponds to linewidth). It is partially satisfied approximately until 15 db or BER = The variant with 40 Gb/s bit rate was not estimated with desired likelihood. The reason of that act may be various. One possibility could be small time window and only small part of frequencies had been generated and processed in the BER estimator. Another variant comprises problem(s) at simulation model. Incorrect conditions, impertinent algorithm or mistake in the algorithm, which can skip noise at calculation, because proper dependence occurs in the convex part of curve (until 15 db) and the absence of inflexed point concave part of curve. The first one couldn t be tested due to lack of available memory. Second one is the good question to designers of software. Finally, the 40 Gb/s case hasn t been considered DPSK format Another simulation investigated DPSK modulation. The setup was modified by DPSK components. DPSK source consisted from CW laser and one dual stage Mach- 38

48 Zehnder modulator. Laser power was set on one half of the previous value. Filter has got double bandwidth due to the same spectral efficiency. One delay line interferometer with single balanced receiver was employed and BER estimation provided the same module. The scheme shows Figure 32. Figure 32: Simulation scheme of BER = ƒ (OSNR). I haven t got experimental data from DPSK measurements. Thus I used theoretical predictions and previously verified DQPSK simulation. DPSK should have identical behavior like DQPSK (shape of BER curve). Two signal s states of DPSK against four signal s states of DQPSK are able to reach the same BER level at worse conditions, because the linewidth has to be much lower than modulation bandwidth. The transmission bit rate was 10 Gb/s. The output data corresponds to theoretical predictions that DPSK is more resistant against phase noise. The divergence of particularly curves is not rapid, but typical flat waveform occurs for 1 GHz linewidth. The OSNR difference between DQPSK and DPSK shows better performance of DPSK. Its benefit increases with lower BER. Identical trends but different setup is reported too [18]. 39

49 Figure 33: BER vs OSNR for 10 Gbs and 40 Gb/s DPSK format. The DPSK source generated 40 Gb/s stream didn t repeat the specific behavior of coherent systems with phase noise. The Flat waveform is missing but convex part is assessed very well, but only one curve with inflexed point and it fused with others. This case excludes possibility of low value of time window, because source with 1 GHz linewidth was described correctly at 10 Gb/s bit rate. It indicates problem at simulation model. The simulation variant 40 Gb/s DPSK will not be joined at next considerations OOK format The simulation of OOK format didn t give reasonable BER values. It has displayed the same number for different values of OSNR. So, OOK is not evaluated. 4.4 Impact of Kerr effect Description of simulation scheme Kerr effect occurs in the presence of high power. Its influence is significant until effective fibre length. Therefore my scheme is adapted on the investigation of Kerr effect (Figure 34). Four channels WDM system was set to catch the FWM influence. This 40

50 simulation is the extension of previously investigated DPSK, DQPSK and OOK formats. It will comprise the BER of transmitter affected by Kerr effect within part of fibre span. Then, the resilient of current format can be derived. The main part of parameters is identical with previous simulations. Channel spacing is set on the 50 GHz. The optical fibre is arranged in the 4 km long fibre loop. DCF is insert between transmission fibre and measurement devices. Demultiplexing provides the identical filter like in the previous case. Optical pulse passes five times through a loop and after every circle is created set of measurements. It includes optical spectrum, eye diagrams of signal s channels and power levels of selected signal channel and two lower, two upper induced bands. Simulation set comprises two common types of optical fibres (SMF, NZ-DSF), three modulation formats examined previously and two various sources. First one demonstrates common telecommunication (DFB) laser type with 2 MHz linewidth and second one represents today s high-end (DFB) laser but ordinary one after certain time (it is hard deal to specify that time, I guess until two years) with 500 khz linewidth. Better DFB lasers are available as well. Figure 34: Simulation scheme of FWM generation. Computing resources were not adequate to software requirements. It led to the crashes of whole simulation. Lower global parameters had to be used and consequently BER estimation is not available. Therefore quality assessment has to use only eye diagrams, which don t reflect real affects on the BER. On the pictures, there will be cases that deploy NZ-DSF, what means worse alternative. 41

51 4.4.2 OOK performance OOK should be the less resilient format and simulation confirmed it. Bit-patterndependence introduced the highest amount of amplitude noise to eye diagram. Narrower linewidth has no effect on the performance; see Figure 35, Figure 36. a) b) Figure 35: Eye diagram of OOK transmitter, a) 2 MHz linewidth, b) 500 khz linewidth 42

52 Figure 36: Eye diagram after first loop. Figure 37: Spectrum after first loop. 43

53 4.4.3 DPSK performance This format is more sensitive on the laser linewidth, Figure 38. On the other Figures, there are dependences with 500 khz linewidth. The eye diagrams (Figure 39) are smoother than OOK ones. Signal s bands in the spectrum have got approximately the same amplitude for OOK and DPSK as well. The nearest noise bands are higher for DPSK (Figure 40) compared to OOK spectrum (Figure 37) but on the other hand, other FWM products have to be higher in the OOK spectrum, due to energy conservation. The eye diagrams opening are approximately identical, but in the back-to-back condition OOK has got advantage of ~2.5 mw higher contrast (difference between low and high level). It means that DPSK is more resilient than OOK. a) b) Figure 38: DPSK back-to-back eye diagram of a) 2 MHz and b) 500 khz linewidth 44

54 Figure 39: DPSK eye diagram after the first loop. 45

55 Figure 40: DPSK spectrum after the first loop DQPSK performance Investigation of modulated DQPSK format suffers bit interference at symbol. The distortion assessment on the DQPSK may bring the focusing on the bottom curve in the eye diagram (Figure 42). It shows similar distortion as DPSK (Figure 39). Certainly the final impairment is more serious than DPSK case and it has to consider also fidelity of demodulation. The FWM products generation (Figure 43) has got similar behavior compared to OOK format (Figure 37). 46

56 Figure 41: DQPSK eye diagrams in the back-to-back condition a) 2 MHz, b) 500 khz. Figure 42: DQPSK eye diagram after the first loop. 47

57 Figure 43: DQPSK spectrum after the first loop Potential of best setup DPSK format deployed 500 khz laser linewidth reached the best performance, which can be enhanced by using new type of optical fibre with large effective area and sources with even narrower linewidth. Following Figures (44-46) show the best results of my simulation. 48

58 Figure 44: DPSK eye diagram after first loop Figure 45: DPSK spectrum after firrst loop 49