IJCCCE Vol.13, No.2, 2013
_____________________________________________________________________________________
55
Electronic Signal Processing for Cancelation of OpticalSystems Impairments
Asst. Prof. Dr. Ali Y. Fattah 1, Zainab Faydh Mohammed 1
1 Electrical Eng. Department, University of Technology, Baghdad
email: [email protected], [email protected]
Received: 28/11/2012
Accepted: 22/07/2013
Abstract –In this paper 40 Gb/s DP-QPSK system with coherent reception and DSPunit for optical fiber impairments compensation is proposed . The DSP unit processesthe detected coherent DP-QPSK signal. The Chromatic Dispersion (CD) is compensatedusing a simple transversal digital filter and Polarization Mode Dispersion (PMD) iscompensated using adaptive butterfly equalizer which is realized by applying theconstant-modulus algorithm (CMA). A nonlinear compensator (NLC) is used forcompensating the nonlinear effects based on the technique of multi-span back-propagation. A modified Viterbi-and-Viterbi phase estimation algorithm (workingjointly on both polarizations) is then used to compensate for phase and frequencymismatch between the transmitter and local oscillator (LO). After the digital signalprocessing is complete, the signal is sent to the detector and decoder, and then to theBER test set for direct-error-counting. The presented system is designed and simulatedusing OptiSystem (2011) software interfaced with MATLAB software R2011a forimplementing the DSP unit algorithms. The performance of each part of the system isanalyzed by showing the optical spectrum, RF spectrum, electrical constellationdiagrams, eye diagram and BER performance for different sampling rates and differentbit rates.
Keywords: Coherent Reception, Digital Signal Processing, Optical Fiber Impairments.
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
56
1. IntroductionPhysical impairments in the optical
fiber, in particular, chromatic dispersion,fiber nonlinearities, polarization effects,and amplified spontaneous emissionnoise from the amplifiers, all interact,limiting the data rate and/or thetransmission distances. Solutions formitigating effects of these impairmentsare traditionally based on techniques inthe optical domain, i.e., before thedetection. The primary reason for thistrend has been the background ofresearchers working in the field, who aremostly device physicists. Opticalcompensators, however, rely onadaptive optics and are usually slow inresponding to the system degradation,and are expensive and bulky devices.Electrical domain approaches based onsignal processing, on the other hand,offer great flexibility in design and canbe integrated within the chip sets at thereceiver, reducing bulkiness. Also, theycan potentially operate after the opticalsignal has been partially demultiplexedso that electrical processing is done ata lower rate, hence substantiallylowering the costs. The promise ofsignal processing approaches for opticalcommunications has been noted morethan a decade ago , but their successfuldemonstrations for high-speed opticalcommunications have appeared morerecently [1].
A significant effort has been expendedin industry and academia to identifyelectronic signal processing as a cost-effective technique for upgrading datatransmission to 10Gb/s for variousapplications over installed fibers. Theseapplications include Local Area Networks(LAN), Storage Area Networks (SAN),metro-area networks, and long-haul
systems. The dominant installed fiberinfra-structure for LAN and SAN isMultimode Fiber (MMF) and for metroand long-haul is Single-Mode Fiber(SMF). Such different installed medialead to very different engineeringchallenges due to different dispersionenvironments and introduce very differentperformance bounds. In addition toenabling upgrades to 10Gb/s overinstalled fiber, there is also currentactivity in the industry in using PlasticOptic Fiber (POF) with electronic signalprocessing as the most cost-effective andmost power-efficient technique to enable10-Gb/s transmission within data centersand smaller enterprises, as compared with10GBASE-T over unshielded twistedcopper pairs. Furthermore, there is alsosome effort on the use of electronic signalprocessing in 10Gb/s Ethernet PassiveOptical Networks (10G EPON) forAccess Networks [2].The electronic signal-processingtechniques can be broadly classified asadaptive equalization at the receiver,predistortion at the transmitter, andelectric-field domain signal processing.The Electronic Dispersion Compensation(EDC) at the receiver can be mostconveniently designed to be fullyadaptive and, due to its ease of use andattractive economics, this approach willbe emphasized [2].In this context, Dual-PolarizationQuadrature Phase-Shift Keying (DP-QPSK) transmission emerged as anattractive alternative. Such systemsconvey four bits per symbol(considering both polarizationorientations), consequently reducing thesymbol rate by the same factor incomparison to a binary system at thesame bit rate. In addition to relaxinghardware requirements, the reduced
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
57
symbol rate also accounts for anincreased tolerance to Inter-SymbolInterference (ISI). If coherentlydetected, polarization multiplexedQPSK signals can be separated at thereceiver by signal processingalgorithms, linear and nonlinear effectscan be compensated using digitalprocessing algorithms [3].
2. Channel ImpairmentsCompensation
The symbiotic combination of DigitalSignal Processing (DSP), coherentdetection, and spectrally efficientmodulation formats has resulted in thedigital coherent optical receiver [4].Coherent detection employing multilevelmodulation format has become one of themost promising technologies for nextgeneration high speed transmissionsystem due to the high power and spectralefficiencies. With the powerful DSP,coherent optical receivers allow thesignificant equalization of chromaticdispersion (CD), polarization modedispersion (PMD), phase noise (PN) andnonlinear effects in the electrical domain[5].
Because of the dynamic nature ofsome impairments such as PMD,compensators must be adaptive.Adaptation is not easily achieved in theoptical domain because of the relativelack of flexibility in optical components,and because of the difficulty in extractingan appropriate error signal to control theadaptation. Adaptation that is required totrack changing PMD conditions isrelatively simple to implementelectronically, with established adaptationalgorithms such as the Constant ModulusAlgorithm (CMA) and Least MeanSquare (LMS) algorithm [6].
3. Coherent DetectionThe most advanced detection method
is coherent detection where the receivercomputes decision variables based on therecovery of the full electric field, whichcontains both amplitude and phaseinformation. Coherent detection thusallows the greatest flexibility inmodulation formats, as information canbe encoded in amplitude and phase, oralternatively in both in-phase (I) andquadrature (Q) components of a carrier.Coherent detection requires the receiverto have knowledge of the carrier phase, asthe received signal is demodulated by aLO that serves as an absolute phasereference [7]. In direct detection as shownin Figure (1), in an opt electrical photodetector (a photodiode) the light intensity|E| is converted in an electrical signaland the phase information is totally lost.
Figure 1. Schematic of direct receiver [8]
An alternative way to detect theoptical signal is coherent detection inwhich the received signal is mixed withlocal laser being detected in thephotodiode, and two detectors and properphase delays are used, both amplitude andphase can be preserved as shown inFigure (2) [8].
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
58
Figure 2. Schematic of coherent receiver [9]
While coherent detection wasexperimentally demonstrated as early as1979, its use in commercial systems hasbeen hindered by the additionalcomplexity, due to the need to track thephase and the polarization of theincoming signal. In a digital coherentreceiver these functions are implementedin the electrical domain leading to adramatic reduction in complexity.Furthermore since coherent detectionmaps the entire optical field within thereceiver bandwidth into the electricaldomain it maximizes the efficacy of thesignal processing. This allowsimpairments which have traditionallylimited 40Gbit/s systems to be overcome,since both chromatic dispersion andpolarization mode dispersion (PMD) maybe compensated adaptively using lineardigital filters [9].
4. Digital Signal Processing AidedCoherent Optical Detection
An important goal of a long-hauloptical fiber system is to transmit thehighest data throughput over the longestdistance without signal regeneration.Digital signal processing (DSP) is used atthe receiver to remove the need fordynamic polarization control and also to
compensate for linear (and some extent ofnon-linear) transmission impairments. Anoptical transmission system can berepresented as shown in Figure (3).
where E is the transmitted signal,H( ) is the channel transfer function andE is the received signal. The goal ofDSP is to implement H ( ), that can beinterpreted as the combination of all thelinear effects that affect the signal duringthe propagation, and estimateE thatrepresents the processed signal. In orderto compensate for all these effects, thereceived sampled electrical signal iselaborated with a series of algorithms inorder to minimize the bit error rate (BER)that represents the main evaluationcriterion for digital communicationsystem quality [10].
Figure 3. Transmission and DSP blockscheme [10]
5. Dual-Polarization QuadraturePhase Shift Keying SystemDesign
System setup is established usingOPTISYSTEM(2011) andMATLAB(2011) as shown in Figure (4).
IJCCCE Vol.13, No.2, 2013
_____________________________________________________________________________________
59
Figure 4. DP-QPSK Coherent System with DSP
The system can be divided into fivemain parts: DP-QPSK Transmitter,Transmission Link, Coherent Receiver,Digital Signal Processing, and Detection& Decoding (which is followed by direct-error counting).The signal is generatedby an optical DP-QPSK Transmitterthen propagated through the fiber loopwhere dispersion and polarizationeffects occur. The layout representing theoptical coherent dual-polarization QPSK
transmitter for a single channeltransmission component is shown inFigure (5). In this case, polarizationmultiplexing is used, the laser output issplit into two orthogonal polarizationcomponents by Polarization BeamSplitter (PBS), which are modulatedseparately by QPSK modulators and thencombined using a Polarization BeamCombiner (PBC).
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
60
Figure 5. Optical dual-polarization QPSKtransmitter equivalent layout
For this model, 40Gb/s PseudoRandom Bit Sequence (PRBS)generator is modulated into twoorthogonally polarized QPSK opticalsignals by two QPSK modulators, Bitrate is 40 Gb/s, Sample rate is1.28×10 Hz, Input signal power is 10dBm and Wavelength is 1550 nm. Figure(6) represents a QPSK Modulator whichstarts with the PSK Sequence Generatorto Generate two parallel M-ary symbolsequences from binary signals usingphase shift keying modulation (PSK)(with 2 Bits per symbol).
Figure 6. QPSK Modulator
After that, it passes through M-aryPulse Generator to Generates multilevelpulses according to the M-ary signalinput(with 1 bit Duty cycle),then eachsignal is modulated by Lithium NiobateMach-Zehnder Modulator and combinedtogether to form the QPSK signal.
The transmission link as shown inFigure (7) is composed of 2 fiber spans.Each span contains SSMF with length= 50 km. The optical fiber componentsimulates the propagation of an opticalfield in a single-mode fiber with thedispersive and nonlinear effects takeninto account by a direct numericalintegration of the modified NonlinearSchrödinger (NLS) equation (when thescalar case is considered) and a system oftwo, coupled NLS equations when thepolarization state of the signal isarbitrary.
Figure 7. Transmission link
The optical coherent dual-polarizationQPSK receiver consists of a homodynereceiver design. The component has aLocal Oscillator (LO) laser polarized at45o relative to the polarization beamsplitter, and the received signal isseparately demodulated by each LOcomponent using two single polarizationQPSK receivers. Figure (8) shows thelayout representing the receiver.
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
61
Figure 8. Optical Coherent DP-QPSK Receiver
The optical coherent QPSK receiverconsists of a homodyne receiver design.In a homodyne receiver, the frequency ofthe Local Oscillator (LO) laser is tuned tothat of the TX laser so the photo receiveroutput is at baseband. The component isformed by a set of 3 dB fiber couplers, aLO laser, and balanced detection. Figure(9) shows the layout representing thereceiver.
Figure 9. QPSK Receiver
The four output signals form OpticalCoherent DP-QPSK Receiver are I and Qof the two polarizations(X,Y), whichhave the full information of transmittedsignal can be represented as Output X-I,Output X-Q, Output Y-I and Output Y-Q.These received electrical signals arethen amplified with a set of fourelectrical amplifier having gain =15dB each as shown in Figure (10).After amplification the signals arepassed through Low Pass Gaussianfilters for eliminating the frequenciesabove required band.
Figure 10. Amplification and filtering of thereceived signals
6. Digital Signal Processing (DSP)Unit
After the four signals are amplifiedand filtered, they are passed to the DSPunit for channel impairmentscompensation as shown in Figure (10).The algorithms used for digital signalprocessing are implemented through aMATLAB component . The innerstructure of the DSP modules is shown inFigure (11).
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
62
Figure 11. DSP Structure
The four signals enter the DSP firstthey are converted to digital domain forprocessing then the fiber dispersion iscompensated using a simple transversaldigital filter followed by a butterflyNonlinear Compensator structure (NLC)to compensate the nonlinear effects andthe adaptive PMD compensation isrealized by applying the Constant-Modulus Algorithm (CMA). Amodified Viterbi-and-Viterbi phaseestimation algorithm (working jointlyon both polarizations) is then used tocompensate for phase and frequencymismatch between the transmitter andlocal oscillator (LO).
a) Analog to Digital ConversionThe analog to digital conversion is
basically a down sampling process. A 2-bit sampling is chosen, however samplingrate can be changed.
b) CD CompensationIn the absence of fiber nonlinearity,
the fiber optic can be modeled as a filterwith the transfer function as given inEquation (1).
G(z, ) = e (1)
In order to compensate the ChromaticDispersion simple transversal digitalfilter is used , we multiply the output fieldby the inverse of the channel transferfunction (FIR filter).The magnituderesponse for this filter is shown in Figure(12).The order of the filter increases asthe amount of dispersion (length of thepropagation) increases.
c) Nonlinear Effects CompensationFor the presented system single
channel transmission is used so thenonlinear impairments effect is limited toSPM. SPM affects the phase of signalsand causes spectral broadening, which inturn leads to increases in dispersionpenalties. SPM compensation is done bynonlinear compensator (NLC) as shownin Figure (13) based on the technique ofmulti-span back-propagation.
Figure 12. The magnitude response of CDcompensating filter
Figure 13. (NLC) Compensator
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
63
E = E e ) (2)
) (3)
Where = | | , = is theintra-polarization nonlinearity parameterand is the inter-polarizationnonlinearity parameter and has to beoptimized. The best BER was found for and = 25.
d) PMD CompensationThe Jones matrix of the fiber for
transmission can be written as
= (4)
where and denote the powersplitting ratio and the phase differencebetween the two polarization modes. TheState Of Polarization (SOP) of the outputsignal can be written as:
(5)
By knowing the inverse of matrix T,we can do polarization de-multiplexing.The CMA is a conventional way for this.Figure (14) shows the DSP circuit forchannel expression. The h matrix isbasically an adaptive FIR filter. CMA isused for blind estimation. For theproposed system a 3-tap FIR filter ischosen, however the order can bechanged. The initial values are:
h = (… 010 … ),h = (… 000 … ),h = (… 000 … ) ,h = (… 010 … ) .
e) Carrier Phase Estimation(CPE)Phase locking in the hardware domain
can be replaced by phase estimation indigital domain by DSP. The received
QPSK signal can be presented byEquation (6).
( ) ( ) ( )] (6)
Figure 14. PMD Compensation
During this step frequency and phaseoffset between local oscillator and signalis compensated using "Viterby-and-Viterby" method (working jointly on bothpolarizations) as explained in Figure (15).
Figure 15. Carrier Phase Estimation
After the digital signal processingis completed, the signal is sent to thedetector and decoder, and then to theBER test set for error detection asshown in Figure (16).
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
Figure 16. Detecting and Decoding after theDSP
7. Performance of 40 Gb/s (DP-QPSK) Coherent System withDSP
Figure (17) shows the optical powerspectrum of the transmitted QPSK signalsin polarizations(X,Y) in (a) and (b)respectively that will be transmittedthrough the optical fiber. Figure (18)shows the optical power spectrum of thetransmitted QPSK signals with opticalpower spectrum of the noise (the greensignal) added after it has passed throughthe transmission optical channel. Figures(19) shows RF Spectrum of the (DP-QPSK) receiver's output signals for X,Y-polarizations. Figure (20) displays the In-Phase and Quadrature-Phase of electricalsignals for X,Y polarizations in aconstellation diagram after it passesthrough the Electrical Amplifiers andLow Pass Gaussian Filters. Figure (21)displays the In-Phase and Quadrature-Phase of electrical signals for X,Ypolarizations in a constellation diagramafter it passes through the DSP unit.
(a)
(b)
Figure 17. Optical power spectrum of thetransmitted QPSK signals.(a) X-Polarization, (b)
Y-Polarization.
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
(a)
(b)Figure 18. Optical power spectrum of the
QPSK signals after the transmission channel.(a) X-Polarization, (b) Y-Polarization.
(a)
(b)Figure 19. RF Spectrum of the (DP-QPSK)
receiver's output signals.(a)X-Polarization, (b) Y-Polarization.
(a)
(b)Figure 20. I-Phase and Q-Phase of electricalsignals.(a)X-polarization,(b) Y-polarization.
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
(a)
(b)Figure 21. The In-Phase and Quadrature-Phase of
electrical signals.(a) X-polarization,(b) Y-polarization
8. Performance of the DSP UnitThe algorithms used for digital
signal processing are implementedthrough a Matlab component. Bysetting the Matlab component to debugmode, the generated electricalconstellation diagrams before DSPshown in figure (22).Figure (23) showselectrical constellation diagrams for bothsignals (X,Y) polarizations after CDcompensation .Figure (24) showselectrical constellation diagrams for bothsignals (X,Y) polarizations afterNonlinear Effects Compensation .Figure(25) shows electrical constellationdiagrams for both signals (X,Y)polarizations after PMD compensation.Figure (26) shows electricalconstellation diagrams for both signals
(X,Y) polarizations after Carrier PhaseEstimation .
(a)
(b)
Figure22. Electrical constellation diagramsbefore the DSP.(a) X-polarization,(b) Y-
polarization.
(a)
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
(b)Figure 23. Electrical constellation diagrams
after CD compensation.(a) X-polarization,(b) Y-polarization.
(a)
(b)Figure 24. Electrical constellation diagrams
after Nonlinear Effects Compensation.(a) X-polarization,(b) Y-polarization.
(a)
(b)Figure 25. Electrical constellation diagrams
after PMD compensation.(a) X-polarization,(b) Y-polarization.
(a)
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
(b)Figure 26. Electrical constellation diagrams
after Carrier Phase Estimation.(a) X-polarization,(b) Y-polarization.
9. Eye DiagramsThe generated eye diagrams after each
step of the DSP obtained using(OptiSystem2011) are shown in thefigures (27-30). Figure (27) in (a) and (b)show the eye diagram and Q factorrespectively for the X-polarization OPSKsignal before CD, PMD and Nonlineareffect compensations. Figure (28) in (a)and (b) show the eye diagram and Qfactor respectively for the X-polarizationOPSK signal after CD compensation.Figures (29) in (a) and (b) show the eyediagram and Q factor respectively for theX-polarization OPSK signal after PMDcompensation .Figure (30) in (a) and (b)show the eye diagram and Q factorrespectively for the X-polarization OPSKsignal after Nonlinear effectscompensation.
(a)
(b)Figure 27. Eye diagram and Q factor before
CD,PMD, Nonlinear effect compensations.
(a)
(b)Figure 28. Eye diagram and Q factor for X-
polarization after CD Compensation.
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
(a)
(b)Figure 29. Eye diagram and Q factor for X-polarization after PMD Compensation.
(a)
(b)Figure 30. Eye diagram and Q Factor for X-
polarization after Nonlinear Effects compensation.
10. Analysis of BER PerformanceFigure (31) represents the BER
performance with Optical Signal to NoiseRatio (OSNR) (DP-QPSK) Coherentoptical system for two sampling rates 2and 4 samples per symbol. According tothe results the system performance showsan obvious improvement with theincrement of the sampling rate. Figure(32) represents the BER performancewith OSNR for two bit rate 40 Gb/s and100 Gb/s. The simulation result showsthat 40 Gb/s system has betterperformance than 100 Gb/s.
Figure 31. BER performance with OSNR fordifferent sampling rates
IJCCCE Vol.13, No.2, 2013
Dr. Ali Y. Fattah and Zainab FaydhMohammed
Electronic Signal Processing forCancelation of Optical SystemsImpairments
____________________________________________________________________________________
70
Figure 32. BER performance with OSNR fordifferent bit rates.
11. Conclusions1.40 Gb/s (DP-QPSK) Coherent
system with DSP unit is designedusing OptiSystem(2011) interfacedwith MATLAB R2011a forimplementing the DSP unitalgorithms. The performance ofeach part of the system is analyzedby showing the opticalspectrum,RF spectrum andelectrical constellation diagrams forthe transmitted signals with noisesignals in both polarizations (X,Y)
2.The performance of the DSP unitfor optical impairmentscompensation is analyzed byelectrical constellation diagrams,Eye diagrams with Q factor aftereach step (CD compensation,Nonlinear effect compensation,PMD compensation and CarrierPhase Estimation) are presented. Itwas found from constellationdiagram that adding DSP unitimproves the system performancedrastically by clearly distinguishingthe constellation points at thedesired bit positions.
3. The BER performance with OpticalSignal to Noise Ratio (OSNR) of(DP-QPSK) Coherent optical systemfor two sampling rates 2 and 4samples per symbol is analyzed. The
system performance shows anobvious improvement with theincrement of the sampling rate. BERperformance with OSNR for two bitrate 40 Gb/s and 100 Gb/s ispresented, the simulation resultshows that 40 G b/s system has betterperformance than 100 Gb/s.
References[1] T. Adalr, W. Wang, and A. O. Lima, “Electronic
Equalization in Optical Fiber Communications”,IEEE International Conference on Acoustics,Speech, and Signal Processing (ICASSP), Vol. IV,pp. 497-500,2003.
[2] I. P. Kaminow, T. Li, A. E. Willner, “Optical FiberTelecommunications V A Components andSubsystems”, Elservier Science, (USA), 2008.
[3] T. F. Portela, D. V. Souto, V. N. Rozental, H. B.Ferreira, D. A. A. Mello, “Analysis of SignalProcessing Techniques for Optical 112 Gb/s DP-QPSK Receivers with Experimental Data”, Journalof Microwaves, Optoelectronics andElectromagnetic Applications, Vol. 10, No. 1, June2011.
[4] S. J. Savory, “Digital Coherent Optical Receivers:Algorithms and Subsystems”, IEEE Journal ofSelected Topics in Quantum Electronics, Vol. 16,No. 5, pp. 1164-1179,October, 2010.
[5] T. Xu, “Digital Dispersion Equalization and CarrierPhase Estimation in 112-Gbit/s Coherent OpticalFiber Transmission System”, Licentiate Thesis,Royal Institute of Technology, Stockholm, Sweden,2011.
[6] J. Sewter , “Electronic Equalization of PolarizationMode Dispersion in 40 G b/s Optical Systems”,MSc. Thesis, Electrical and Computer Engineering ,University of Toronto, 2005.
[7] E. Ip, “Coherent Detection and Digital SignalProcessing for Fiber Optic Communications”, PhD.Thesis, Department of Electrical Engineering,Stanford University, December, 2008.
[8] M. Mussolin, “Digital Signal ProcessingAlgorithms for High-Speed Coherent Transmissionin Optical Fibers”, M Sc. Thesis,Telecommunications Engineering, PadovaUniversity, Italy, 2010.
[9] S. J. Savory, “Digital Filters for Coherent OpticalReceivers”, Optical Society of Americas, OpticalExpress, Vol. 16, No.2, pp. 804-816,2008.
[10] S. J. Savory, G. Gavioli, R. I. Killey, P. Bayvel,“Electronic compensation of chromatic dispersionusing a digital coherent receiver”, Optical Societyof America, Optical Express, Vol. 15, No. 5, pp.2120-2126 ,2007.