(OSSB); Dual-Electroabsorption Modulated Laser. Flexible Solution for Amplified and Dispersion Uncompensated Networks over Standard Fiber

Similar documents
OSSB-OFDM TRANSMISSION PERFORMANCE USING A DUAL ELECTROABSORPTION MODULATED LASER IN NG-PON CONTEXT

Optical fiber beamformer for processing two independent simultaneous RF beams

Relation between C/N Ratio and S/N Ratio

ELEC2202 Communications Engineering Laboratory Frequency Modulation (FM)

A soft decision decoding of product BCH and Reed-Müller codes for error control and peak-factor reduction in OFDM

Fiber Bragg grating based four-bit optical beamformer

Experiment 7: Frequency Modulation and Phase Locked Loops October 11, 2006

Block Diagram of FM Receiver

Notes on Orthogonal Frequency Division Multiplexing (OFDM)

Adaptive Harmonic IIR Notch Filter with Varying Notch Bandwidth and Convergence Factor

COMBINED FREQUENCY AND SPATIAL DOMAINS POWER DISTRIBUTION FOR MIMO-OFDM TRANSMISSION

Iterative Receiver Signal Processing for Joint Mitigation of Transmitter and Receiver Phase Noise in OFDM-Based Cognitive Radio Link

DIGITAL Communications

Optimal Modulation Index of the Mach-Zehnder Modulator in a Coherent Optical OFDM System Employing Digital Predistortion

Keywords Frequency-domain equalization, antenna diversity, multicode DS-CDMA, frequency-selective fading

Power Improvement in 64-Bit Full Adder Using Embedded Technologies Er. Arun Gandhi 1, Dr. Rahul Malhotra 2, Er. Kulbhushan Singla 3

2.5 Gbps Millimeter-Wave Radio over Fiber Transmission based on Dual Octupling of RF Local Oscillator

Transmit Power and Bit Allocations for OFDM Systems in a Fading Channel

Selective Harmonic Elimination for Multilevel Inverters with Unbalanced DC Inputs

TESTING OF ADCS BY FREQUENCY-DOMAIN ANALYSIS IN MULTI-TONE MODE

LOW COST PRODUCTION PHASE NOISE MEASUREMENTS ON MICROWAVE AND MILLIMETRE WAVE FREQUENCY SOURCES

The PAPR and Simple PAPR Reduction of the 2D Spreading Based Communication Systems

HIGH FREQUENCY LASER BASED ACOUSTIC MICROSCOPY USING A CW GENERATION SOURCE

RAKE Receiver. Tommi Heikkilä S Postgraduate Course in Radio Communications, Autumn II.

SECURITY AND BER PERFORMANCE TRADE-OFF IN WIRELESS COMMUNICATION SYSTEMS APPLICATIONS

UNIT - II CONTROLLED RECTIFIERS (Line Commutated AC to DC converters) Line Commutated Converter

Amplifiers and Feedback

Study and Implementation of Complementary Golay Sequences for PAR reduction in OFDM signals

5 Constellation for Digital Modulation Schemes

An orthogonal multi-beam based MIMO scheme. for multi-user wireless systems

ACCURATE DISPLACEMENT MEASUREMENT BASED ON THE FREQUENCY VARIATION MONITORING OF ULTRASONIC SIGNALS

Kalman Filtering for NLOS Mitigation and Target Tracking in Indoor Wireless Environment

A Preprocessing Method to Increase High Frequency Response of A Parametric Loudspeaker

REPORT ITU-R SA Telecommunication characteristics and requirements for space VLBI systems

Dual-modulation of a novel Integrated Laser-modulator for Radio-over-Fiber Systems

Performance Analysis of Atmospheric Field Conjugation Adaptive Arrays

Exploring the Electron Tunneling Behavior of Scanning Tunneling Microscope (STM) tip and n-type Semiconductor

Ignition and monitoring technique for plasma processing of multicell superconducting radio frequency cavities

Overlapped frequency-time division multiplexing

Analysis of mach zehnder modulator response to fiber dispersion in radio over fiber at 60 GHz for multigigabit wireless transmission

A New Simple Model for Land Mobile Satellite Channels

A Novel Frequency Independent Simultaneous Matching Technique for Power Gain and Linearity in BJT amplifiers

] (1) Problem 1. University of California, Berkeley Fall 2010 EE142, Problem Set #9 Solutions Prof. Jan Rabaey

Definition Recall, from 6.7, that PAM signal waveforms are represented

UWB System for Time-Domain Near-Field Antenna Measurement

Boris Krnic Nov 15, ECE 1352F. Phase Noise of VCOs

LETTER Adaptive Multi-Stage Parallel Interference Cancellation Receiver for Multi-Rate DS-CDMA System

A simple charge sensitive preamplifier for experiments with a small number of detector channels

Phase Noise Modelling and Mitigation Techniques in OFDM Communications Systems

Using Adaptive Modulation in a LEO Satellite Communication System

2. Continuous-wave modulation

PREDICTING SOUND LEVELS BEHIND BUILDINGS - HOW MANY REFLECTIONS SHOULD I USE? Apex Acoustics Ltd, Gateshead, UK

Design of an Arrayed Waveguide Grating with flat spectral response

EQUALIZED ALGORITHM FOR A TRUCK CABIN ACTIVE NOISE CONTROL SYSTEM

Energy-Efficient Cellular Communications Powered by Smart Grid Technology

ANALOGUE & DIGITAL COMMUNICATION

Implementation of Wideband Digital Beam Forming in the E-band

Mode spectrum of multi-longitudinal mode pumped near-degenerate OPOs with volume Bragg grating output couplers

EFFECTS OF MASKING ANGLE AND MULTIPATH ON GALILEO PERFORMANCES IN DIFFERENT ENVIRONMENTS

Evaluation of Steady-State and Dynamic Performance of a Synchronized Phasor Measurement Unit

Intermediate-Node Initiated Reservation (IIR): A New Signaling Scheme for Wavelength-Routed Networks with Sparse Conversion

Additive Synthesis, Amplitude Modulation and Frequency Modulation

A Novel Low Power UWB Cascode SiGe BiCMOS LNA with Current Reuse and Zero-Pole Cancellation

70km external cavity DWDM sources based on O-band Self Seeded RSOAs for transmissions at 2.5Gbit/s

Fundamental study for measuring microflow with Michelson interferometer enhanced by external random signal

EXPERIMENTATION FOR ACTIVE VIBRATION CONTROL

Model Development for the Wideband Vehicle-to-vehicle 2.4 GHz Channel

Multicarrier Interleave-Division Multiple Access Communication in Multipath Channels

Modeling Beam forming in Circular Antenna Array with Directional Emitters

POWER QUALITY ASSESSMENT USING TWO STAGE NONLINEAR ESTIMATION NUMERICAL ALGORITHM

Cross-correlation tracking for Maximum Length Sequence based acoustic localisation

Compensated Single-Phase Rectifier

Precise Indoor Localization System For a Mobile Robot Using Auto Calibration Algorithm

AccuBridge TOWARDS THE DEVELOPMENT OF A DC CURRENT COMPARATOR RATIO STANDARD

General Smith Chart Matching

OTC Statistics of High- and Low-Frequency Motions of a Moored Tanker. sensitive to lateral loading such as the SAL5 and

New Adaptive Linear Combination Structure for Tracking/Estimating Phasor and Frequency of Power System

DSI3 Sensor to Master Current Threshold Adaptation for Pattern Recognition

Secondary-side-only Simultaneous Power and Efficiency Control in Dynamic Wireless Power Transfer System

Power Optimal Signaling for Fading Multi-access Channel in Presence of Coding Gap

Part 9: Basic AC Theory

Impact of the Reactive Power Compensation on Harmonic Distortion Level

ORTHOGONAL frequency division multiplexing (OFDM) has become the most popular multicarrier

Alternative Encoding Techniques for Digital Loudspeaker Arrays

Simplified Analysis and Design of MIMO Ad Hoc Networks

Optical Magnetic Response in a Single Metal Nanobrick. Jianwei Tang, Sailing He, et al.

A 1.2V rail-to-rail 100MHz amplifier.

Sound recording with the application of microphone arrays

Overlapping Signal Separation in DPX Spectrum Based on EM Algorithm. Chuandang Liu 1, a, Luxi Lu 1, b

VHDL-AMS Behavioral Modeling and Simulation of M-QAM transceiver system

Investigating Multiple Alternating Cooperative Broadcasts to Enhance Network Longevity

Performance Evaluation of UWB Sensor Network with Aloha Multiple Access Scheme

International Journal of Electronics and Electrical Engineering Vol. 1, No. 3, September, 2013 MC-DS-CDMA

A Novel Control Scheme to Reduce Storage Capacitor of Flyback PFC Converter

Lab 5: Differential Amplifier.

EXPERIMENTAL VERIFICATION OF SINUSOIDAL APPROXIMATION IN ANALYSIS OF THREE-PHASE TWELVE-PULSE OUTPUT VOLTAGE TYPE RECTIFIERS

Performance Analysis of an AMC System with an Iterative V-BLAST Decoding Algorithm

Design of Efficient ZVS Half-Bridge Series Resonant Inverter with Suitable Control Technique

Windowing High-Resolution ADC Data Part 2

Department of Electrical and Computer Systems Engineering

Transcription:

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 The Dual-Electroabsorption Modulated Laser, a Flexible Solution for lified and Dispersion Uncopensated Networks over Standard Fiber Didier Erase, Thoas Anfray, Mohaed Essghair Chaibi, Khalil Kechaou, Juan Petit, Guy Aubin, Kael Merghe, Christophe Kazierski, Jean-Guy Provost, Philippe Chanclou and Christelle Aupetit-Bertheleot Abstract We propose here a onolithically integrated dual RF access electro-absorption odulated laser called D-EML for the generation of optical single-sideband signal for both non-return to zero and orthogonal frequency division ultiplexing forats. The benefits of optical single-sideband over double sideband in ters of resilience against chroatic dispersion effects is presented for high bit rate intensity odulation/direct detection transissions over standard single ode fiber in the C-band. Both siulation and experientation exhibit transission distance enhanceent in the case of dual odulation of the D-EML, i.e. in optical singlesideband configuration, aking this optical source an interesting solution because of its low-cost, low-size and reasonable-power consuption. Index Ters Optical Sources; Optical Counications; Orthogonal Frequency Division Multiplexing (OFDM); Optical Single-sideband (OSSB); Dual-Electroabsorption Modulated Laser (D-EML). N I. INTRODUCTION EW data-hungry counication services whether fixed or obile keep adding stress on the deand for high bit rate fiber optics transission systes. Every level of the network architecture is concerned with this increase related as uch to personal applications such as video-on-deand and on-line gaing as to professional assive data transfer required by the Cloud architectures. Whereas the long haul systes focus on cuulative single fiber data rate transfer and have recently igrated to coherent, ulti-odulation-levels protocols, the etro and the access segents of the network have to deal with cost, energy consuption, footprint constraints that presently prevents the use of expensive coherent receiver and very high Thoas Anfray and Christelle Aupetit-Bertheleot are with XLIM, UMR CNRS 75/University of Lioges, 6 Rue Atlantis, 8768 Lioges, France (e-ail: thoas.anfray@ensil.unili.fr). Didier Erase and Mohaed Essghair Chaibi are with Institute MINES- TELECOM, TELECOM ParisTech, CNRS LTCI, 46 Rue Barrault, 75634 Paris, France (e-ail: didier.erase@teleco-paristech.fr). Khalil Kechaou and Juan Petit were with Institute MINES-TELECOM, TELE- COM ParisTech, CNRS LTCI, 46 Rue Barrault, 75634 Paris, France. Guy Aubin and Kael Merghe are with Laboratory for Photonics and Nanostructures/CNRS, Route de Nozay, 946 Marcoussis, France. Christophe Kazierski and Jean-Guy Provost are with III-V Lab, Coon laboratory of "Alcatel-Lucent Bell Labs France", "Thales Research and Technology" and "CEA Leti", Route de Nozay, 946 Marcoussis, France. Philippe Chanclou is with Orange Labs, Avenue Pierre Marzin, 37 Lannion, France. rate digital signal processing. This situation occurs in any occasions: (i) short ( k) or ediu ( k to 5 k) distances etro sections or interconnections between distant data centers such as those specified by syste anufacturers wishing to transport Gb/s as Gb/s or 4 5 Gb/s (see for exaple [] []) in for exaple so-called CFP (Centu for Factor Pluggable) Multi-source agreeent odules; (ii) long reach (up to 6 k) and/or very high bit rate (4 Gb/s) access PON [3]; (iii) front hauling of coon public radio interface (CPRI) [4]. Hence intensity odulation, direct detection (IM-DD) transission systes reain an attractive solution in any counication situations. When aiing at increasing either the transission range or the bit rate, the designer has to itigate the optical fiber chroatic dispersion, which represents the ain liit to transission in the C or L frequency bands. Low cost deployent would take advantage of solutions that avoid in-line chroatic dispersion copensation while sticking to the C-band wavelength window so as to liit optical attenuation and consequently reduce the use of in-line aplifiers. Unfortunately, having to cope with IM-DD schees, chroatic dispersion is clearly a difficult issue for these systes. Pure IM, which happens to be a double-sideband (DSB) odulation, is progressively converted into optical frequency odulation (and back again) when the signal propagates in a dispersive optical fiber. This leads to the well-known frequency selective IM-DD channel response, which displays fading for specific odulation frequencies [5]. Phase-shaped intensity odulation (eaning the phase of the optical signal is odulated conjointly with the intensity although that odulation is not detected at the receiver side) can be advantageously exploited. As early as 994, J. Binder and U. Kohn [6] have proposed a schee wherein the laser adiabatic chirp would perfor frequency odulation (FM) in order to lead to a quasi single-sideband (SSB) odulation. Unfortunately seiconductor laser adiabatic chirp is too sall for a direct application although it was recently shown that orthogonal frequency division ultiplexing (OFDM) transission could ake use of the idea for access networks [7]. Duo-binary odulation has been extensively studied in long haul systes when a LiNbO 3 Mach-Zehnder odulator is used (this ethod required a signal precoding in order to switch the aplitude sign when required) [8]. In 5, D. Mahgerefteh et al. proposed the chirp anaged laser (CML) [9], which in its latest coercial version cobines in a sall Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 package a distributed feedback (DFB) laser and a atched optical filter providing the proper phase shaping. Record non-return to zero (NRZ) transissions at Gb/s reach up to 5 k [9] and 36 k [] although the latter uses an electronic digital pulse shaping driver generating unipolar negative chirp. A deonstration at higher bit rates has been ade at 5 Gb/s (onolithically ultiplexed to Gb/s) up to 4 k []. However, despite a nice packaging, filter selection and alignent appears not to be totally optial in ters of fabrication cost and teperature stabilization constraints. In addition, the schee used in CML is well suited for NRZ transission but do not fit OFDM odulation schees. In addition to adiabatic chirp, the transient chirp of either the laser in the case of directly odulated laser (DML) or of the external odulator when used, adds up soe signal distortion. G.P. Agrawal et al. [] have derived that optial chirp for abnoral dispersive fiber (such as standard single ode fiber at 55 n) was given generally by a chirp coefficient α s equal to - (see part II for definition of α s although the distortion is strongly signal dependent and can be tuned to iprove perforances in specific situation such as that described in []. Finally, odulating conjointly the laser and an electro-absorption odulator (EAM) was proposed by H. Ki et al. [3] based on the idea that a sall current odulation of the laser would perfor frequency odulation whereas the EAM acts as an intensity odulator. First deonstration was done with bulk devices [3] and analyzed in [4]. We perfored the first experients with an integrated device, which we naed a D-EML for dual odulation electro-absorption odulated laser and deonstrated the whole interest of the concept reaching 3 k at Gb/s [5]. This concept relies ainly on atching the condition derived by J. Binder and U. Kohn for the adiabatic chirp and, to a lesser extent, reversing the data signal driving the laser and delaying it in order to better cope with the adiabatic chirp. Record NRZ chroatic uncopensated experients were done at Gb/s by K. Hasebe [6] up to 8 k and at Gb/s, 5 Gb/s and 4 Gb/s by ourselves up to 39.6 k, 5 k and k respectively [7] [8]. Finally, we deonstrated that the SSB schee obtained with D-EML was particularly well suited to IM-DD OFDM transissions, which can represent a powerful solution for the applications entioned above (even though ODFM deodulation requires soe post detection signal processing). We recently deonstrated a chroatic dispersion uncopensated k OFDM transission at. Gb/s [9]. The reaining of the paper is organized as follows. The following section presents a short suary of the theoretical analysis of the transission of DSB and SSB signals in a dispersive optical fiber. The principle of the SSB signal generation based on IM/FM ixing is also explained. The third part is a description of the D-EML device that was designed and tested for this application. The fourth and fifth parts are the description of the siulation and experiental results that we obtained using a D- EML in the case of NRZ and OFDM odulation schees. Finally, the last part draws conclusions. II. PRINCIPLE OF SSB SIGNAL GENERATION BASED ON IM/FM MIXING AND COMPARISON WITH DSB The principle of signal transission in an optical fiber is well known especially when non-linear effects are neglected due to liited optical power levels and polarization ode dispersion are non-relevant thanks to liited transission ranges. This part of the paper suarizes well-known results for transission characteristics of phase plus aplitude odulated signals in dispersive optical fiber. Let us first consider pure sall-signal IM: P ( t) P cos f t () s o where P s (t) is the odulated eitted optical power, P the average power, f the odulation frequency and IM the intensity odulation index. Since haronic analysis will be required, it will be ore convenient to write this signal in ters of coplex aplitude leading to: A ( t) A s o IM AM cos f t () given P = A A, IM = AM and AM is neglected with respect to. Sall signal phase (PM) or frequency odulation (FM) can be added to this expression by writing now: A s t) A o j f t cos f te cos ( (3) AM where β is the phase odulation index (often equally referred to as the frequency odulation index). The instantaneous angular optical frequency varies as: t f sin f t (4) o where ω is the optical frequency without odulation applied. θ is the phase shift between the optical phase odulation and the optical aplitude odulation. It ust be noted that θ ay vary with f. This happens in seiconductor lasers where the adiabatic chirp, which doinates at low odulation frequencies, and the transient chirp, which doinates at high odulation frequencies, correspond to θ = π/ and θ = respectively. The haronic analysis of this signal has been done elsewhere (see for exaple [5]). When transitted in a dispersive fiber the initial haronics are phase shifted relative to each other and when recobined in the square-law detector the output electrical signal leads to the well-known frequency dependent channel response first derived by F. Devaux et al. (for the case θ = only) [], which exhibits typical dips for specific odulation frequency depending on the accuulated dispersion and corresponding to the situation where the ixing on the detector of the central carrier with the left haronic is π out-of-phase with that of the central carrier with the right haronic. A given set of θ, IM and β given by θ = ±π/ and IM = 4 J (β)/j (β) β (where J n denotes the Bessel function of order n) is a noticeable feature in so far as it suppresses the left or right odulation haronic and leads to flat channel response characteristic of SSB odulation. This is the ain feature that will be exploited in a series of odulation schees such as duobinary, CML and D-EML. This cobination of IM and PM can be related to the odulator device chirp. We will define the coplex sall signal chirp coefficient as: Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 3 s j e (5) corresponding to the F. Koyaa definition for external odulator when θ = []. When considering directly odulated laser, this coefficient is directly related to the Henry linewidth enhanceent factor α H such as: IM H j f c s (6) f Where f c is the so-called corner, chirp or characteristic frequency depending on the authors []. Siultaneous IM and PM can also be generated both independently by a D-EML using two different effects and two different parts of this single device. In CML, FM to IM conversion is obtained by eans of optical filter. The liits to sall signal analysis when trying to reach optial SSB conditions are nuerous and are to be analyzed successively and differently depending on applications. A. Second haronics Sall signal SSB is obtained when the first haronic generated by IM is equal in agnitude and opposite in sign to one of the FM haronics. Even if the IM is supposed to be linear thanks to a proper tuning of the odulator bias, the second FM haronics exist and their relative aplitude with respect to the reaining first haronic (doubled due to addition of the IM and FM haronic) is equal to J (β)/j (β) in aplitude (approxiately β/8 for sall β and equal to -4 db for β = /, which correspond to specific conditions described below). These second haronics although affected by the fiber dispersion leading to signal distortion at the receiver reain quite sall. litude (a.u.) litude (a.u.) (a) T b Tie (a.u.) Tie (a.u.) Fig.. Siple -- bit pattern (Gaussian shape). The bit envelops which were well separated at the fiber input, are enlarged through fiber dispersion. The top figures show the bits taken individually whereas the botto figures depict the su of the blue and green curves. The case (a) presents the results for no fiber. The case considers that the phase is unaltered all across the tie span whereas the case (c) corresponds to the situation where a π-phase shift occurs during sybol. It can be seen easily that the bit is preserved nicely in the case (c). Note that the chirp induced frequency evolution within a disperse pulse is not taken in to account. A siple calculation shows that this would not strongly affect the results. Soe ore sophisticated siulation will be displayed latter in the paper. (c) Tie (a.u.) Fig.. Optical spectra for Gb/s at the output of an ideal intensity odulator coupled with an ideal frequency odulator. (a) is the case of pure IM with IM =., (c) and (d) are the cases of IM/FM ixed with respectively IM = and Δf pp = 5 GHz, IM =.8 and Δf pp = 4 GHz, IM =.6 and Δf pp = 3 GHz. B. Non-linear odulator response Every optical intensity odulation (whether external or via direct odulation) has a non-linear response. The linear range is often rather liited to a sall zone of the total odulator intensity span ranging fro zero power to the axiu eitted light output. This drawback is particularly sensitive for EAMs. Large signal odulation will generate extra haronics that, as above, will be dispersed in the fiber. Electronic pre-linearization is coonly used in EAM and directly odulated laser based systes whether for NRZ or OFDM odulations. C. Broadband large signal odulation Optical counication signals are rarely narrow band. When IM-DD signal are concerned, odulation will generally be on-off keying NRZ. For this case J. Binder and U. Kohn have proposed to apply the SSB condition to the frequency corresponding to half the bit rate B, i.e. having an optical frequency difference between arks and spaces of Δf pp = B/ (in the case of a sin-shape frequency odulation Δf pp = βf and f = B/ giving β = /). The arguent given in that paper is not totally clear and optiality of this choice has not been derived forally. We actually show that the choice happens to be only close to optiu in soe particular cases. One notices that when the extinction ratio is high, the resulting situation is that the phase shift of the optical wave in two arks is equal to π ultiply by the nuber of spaces separating the two arks. Fig. displays a siple scheatic diagra explaining the advantage of this configuration in the tie doain as entioned in [4]. It corresponds exactly to the situation in duo-binary odulation. When the extinction ratio ER given as the ratio between the optical power of arks and spaces is not equal to infinity, J. Binder and U. Kohn propose using the general forula: with - - -3-4 (a) (c) (d) -5 536.5 536.3 536.35 536.4 536.45 536.5 536.55 Wavelength (n) B f pp IM (7) Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 4 Chroatic dispersion penalty (db) 8 7 6 5 4 3 (c) (d) 3 4 5 6 7 8 9 Frequency deviation (GHz) Fig. 3. Siulated chroatic dispersion penalty P D versus peak-to-peak frequency deviation Δf pp at 5 k for Gb/s. The cases, (c) and (d) are differentiated by their intensity odulation index which is respectively IM =, IM =.8 and IM =.6. Chroatic dispersion penalty (db) 6 5 4 3 (a) (c) (d) (b') (c') (d') 5 5 5 3 35 Distance (k) Fig. 4. Siulated chroatic dispersion penalty P D versus distance for the cases (a),, (c) and (d) depicted in Fig. at Gb/s. The additional cases (b'), (c') and (d') use the sae intensity odulation indexes as, (c) and (d) with the optiu frequency deviations naely respectively 5.5 GHz, 4.75 GHz and 4.5 GHz. The overall penalty curves are obtained in cases c/c' and d/d' by an up-shift corresponding to P E, which is equal to.97 db and. db respectively. ER ER IM (8) Fig. presents nuerically coputed optical spectru at the output of cobined intensity and frequency odulation by pseudo-rando binary Gb/s sequence of rectangular sybols for several ER and Δf pp showing the evolution towards the SSB regie. With this ideal dual odulator (IM+FM) we have investigated basic cases including the chroatic dispersive fiber (7 ps/n/k at 55 n), an ideal photodiode and an electrical reception filter (3 rd order low-pass Bessel filter with a - 3 db bandwidth of 6.5 GHz at Gb/s). Hence, we have calculated the chroatic dispersion penalty which is coputed fro the eye opening (EO) of the eye diagra. The EO is the difference between the ean value of arks and spaces inus the su of the standard deviation of arks and spaces. The equation (9) gives the chroatic dispersion penalty (P D ), expressed in db, coparing the eye opening in back-to-back condition (EO i ) and at a given transission distance (EO o ). P D EOi log EOo (9) The chroatic dispersion penalty is plotted versus the peakto-peak frequency deviation operated by the frequency odulator in Fig. 3 for a distance equal to 5 k in order to exhibit noticeable chroatic dispersion effects at Gb/s. These siulations display iniu P D for frequency deviation equal to 5.5 GHz, 4.75 GHz and 4.5 GHz that can be copared to J. Binder and U. Kohn epirical proposal of 5 GHz, 4 GHz and 3 GHz for intensity odulation indices of.,.8 and.6. The chroatic dispersion penalty has also been plotted versus distances (Fig. 4). A typical P D = db criterion allows coparing the different cases. It appears that the pure intensity odulation case is obviously very sensitive to the chroatic dispersion and the eye diagra closes rapidly. A distance of about 95 k can be reached for a db penalty in case (a) whereas 5 k, k and 5 k are respectively reachable for cases, (c) and (d) for which IM and FM are ixed according to condition (7). The cases (c) and (d) present better perforances than because of the requireent of low IM and FM indices to better satisfy the SSB condition. However, it should be noticed that decreasing values of extinction ratio leads to an increasing power penalty P E which is calculated by the forula: P log E IM () The power penalty resulting fro zero-level optical power has to be added to P D in order to get the overall penalty P T. The additional cases (b'), (c') and (d'), obtained using the optial frequency deviation values derived fro Fig. 3, show soe sall iproveents of perforances copared to, (c) and (d) eaning that J. Binder and U. Kohn epirical proposal is a reasonable approxiation. D. Modulator's dynaics The previous analysis is funded on linear algebraic-type reasoning. When the physics of the coponent are taken into account, the actual signal response of the eitter is often ore coplicated. This is particularly true in the case of directly odulated lasers in which the dynaics of carriers and photon are different and ay be spatially inhoogeneous (spatial hole burning). III. DESIGN, CHARACTERISTICS AND MODELS OF THE D-EML The D-EML is an integrated electro-absorption odulated laser with two independent odulation accesses, one for driving the laser and the other one for the EAM. The principle of operation relies in balancing both odulations in order to obtain an output signal verifying SSB conditions as exposed in part II. Practically, we could send two different electrical signals on the two icrowave inputs. However this does not correspond to the objective of siplifying the eitter and liiting operation coplexity. We chose to send either identical signals (or possibly in the case of NRZ inverted signals) on both inputs. The only authorized discrepancy will be related to tuning the power ratio and the relative delay of both inputs. An extra reasonably low cost solution consists in adding an ad-hoc electrical filter allow- Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 5 Zn-doped cladding p-inp Spacer InP Active layer MQW AlGaInAs AR DFB Laser access Ti/Au EAM access Ti/Au EAM section DFB Section Substate n-inp Grating GaInAsP SCH GaInAsP Fig. 5. Sectional view of the D-EML with all iportant layers. ing soe sort of equalization particularly useful in OFDM applications and possibly NRZ. A. Design and fabrication of the D-EML The design technology of D-EML is based on AlGaInAs-QW (quantu well) aterial for its large electronic confineent providing enhanced electro-absorption properties and reduced theral carrier leakage. The sae active layer is used for both laser and odulator section. Therefore, a positive detuning of 5 n was realized between DFB and laser gain peak wavelengths. This led to a 45 n optiu detuning between laser and unbiased EAM absorption edge wavelengths ainly due to the laser gap noralization effect. The positive detuning design favors FM efficiency of the laser section. The coponent waveguide is selectively buried with a tande layer of sei-insulating InP separately doped with Fe and Ru. The sei-insulating buried structure assures low EAM capacitance and low theral resistance of the laser. This is obtained by Fe doped InP due to its high resistivity. Ru doping was introduced to prevent Zn-Fe inter-diffusion and to provide InP single optical index confineent for all waveguide [3]. The diodes are copleted by a standard p-layer doped confineent and contact structure. Large p-type regrowth surface favors a low series resistance of both coponent sections in order to enhance their RC bandwidth liits. EAM section is 75 µ long and ay be odulated up to 4 GHz. DFB laser section is 47 µ long and is designed to iprove the FM odulation properties versus standard EML version (device sectional view is shown in Fig. 5). The coupled optical power in the single ode fiber is around 4 db. The D- HR Si-doped cladding n-inp Ground Ti/Au Zn-doped GaInAs p-inp Ru-doped InP Fe-doped InP EML is design to operate in C-band. The laser used in our experientations had a wavelength of 537 n. With such a design, it is possible to separate copletely optical FM and IM by applying digital or analog odulation respectively on the high FM-index laser section and on the high IM-index odulator section. Residual IM produced by the laser section is sall copared to IM produced by the odulator. B. Characterization and odels of the D-EML The original principle of the D-EML is to consider the laser as a pure frequency odulator and the EAM as an aplitude odulator. However, this is not practically the case and a good evaluation of the devices chirp is required. We have perfored both sall and large signal evaluation of the devices chirp characteristics. Ideally, we would like to consider both devices independently and easure the laser characteristics while keeping the EAM at a static bias and vice versa. Unfortunately, they appear not to be totally uncorrelated. A change in the odulator static bias voltage will affect slightly the teperature due to theral heating of the atch resistance and we know that the laser wavelength changes with teperature and the odulator absorption responds in wavelength dependent. A second effect is related to the feedback fro the front facet of the device (odulator side). This feedback tends to odify the laser wavelength but also iprove its odulation bandwidth [4]. This feedback is clearly dependent on the odulator bias since the light goes through it before and after reflection. On the other hand, it was shown on earlier device easureents that a good electrical isolation could be obtained. The odulator effective insertion loss results fro the choice of the bias point and the voltage swing. The forer typically corresponds to a 4 db attenuation leading to a easured db to.5 db attenuation for the -sybols. This copares favorably to the loss induced by the filter in a CML. The output power of D-EML, just as EML is liited by the EAM saturation, which prevents using laser eission as high as could be with DML or CML. Expected output power can reach 7 db, when our own devices would eit around db. A phenoenological odel of the D-EML has been developed in order to perfor syste siulations for extending experiental results. This odel is based on experiental easureents. Fig. 6 depicts the easured electro-optic frequency Transission (db) 3-3 -6-9 - (a) -5-8 Measures Approxiation - 4 8 6 4 8 Frequency (GHz) 3 4 5 6 7 Frequency (GHz) Fig. 6. Measured and siulated electro-optic bandwidth of the laser (a) and odulator of the D-EML. Measures Approxiation Attenuation (db) Henry's factor - -6 - -4-8 - 3 (a) Measures Approxiation - Measures Approxiation -.5.5.5 3 3.5 4 Reverse voltage (V) Power (W) 3 8 4 6 4 6 8 4 6 Current (A) Fig. 7. Measured and siulated static characteristics of the odulator (a) and laser of the D-EML. 8 4 Non saturated nd order saturated Measures Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 6 TABLE I SUMMARY OF EXPERIMENTAL TRANSMISSION DISTANCES OBTAINED IN DUAL MODULATION FOR DIFFERENT BIT RATES AS WELL AS SIMULATED CASES DEPICTED IN FIG. AND REALIZED WITH THE PHENOMENOLOGICAL MODEL OF D-EML. Conditions Data rate Pattern Experiental results with D-EML Gb/s 7-3 k for BER < - [4] 4 k for BER < - 65 k for BER < -8 Siulation results with phenoenological odel of D-EML (a) (c) (d) Distance reach for a given BER [Ref.] 5 k at -3 9 k at -9 6 k at -3 k at -9 3 k at -3 5 k at -9 Gb/s 7-4 k for BER < -8 [6] - - - - 5 Gb/s 7-4 k for BER < -9 8 k at -3 4 Gb/s 3 - k for BER < -8 [7] k for BER < -4 [7] k at -9 6.5 k at -3 4 k at -9 5 k at -3 6 k at -9 8 k at -3 4 k at -9 54 k at -3 8 k at -9 k at -3 4 k at -9 3 k at -3 k at -9 56 k at -3 4 k at -9.5 k at -3 5 k at -9 response of the laser and of the odulator as well as the nuerical fit based on a Butterworth filter odel that we used in siulations. Fig. 7 shows the easured static characteristics of the laser and the odulator. For the odulator a polynoial approach has been used to fit the attenuation and the chirp coefficient characteristics ( th order polynoials). For the laser a basic transfer function without saturation has been coupled with nd order saturation function for reaching the experiental characteristic. Sall-signal chirp easureents have been perfored using the vectorial network analyzer ethod [] [5]. The EAM chirp characteristic is obtained at a fixed laser bias and various odulator biases. The evolution of the EAM chirp going fro a positive value for low reverse bias to a negative value for high bias is observed on Fig. 7(a). This negative chirp can be used in order to iprove the transission range on standard dispersive fibers at 55 n. However the obvious drawback coes fro the high absorption at high reverse bias. We tend to use the EAM around the zero-chirp point. The sall-signal laser chirp is obtained with a constant odulator bias. Its value is not strongly laser bias dependent. Using a fit siilar to that described in [5] we get α H =.5 and f c =.7 GHz. Large signal easureent is perfored using a frequency discriinator based on an interferoetric structure siilar to that studied in [5]. The sall signal results could be thus coforted and high bit rate NRZ-type signals cobining FM and Optical intensity (W).4..8.6.4...4.6.8..4.6.8-5 Tie (ns) Fig. 8. Optical intensity and frequency deviation responses of the laser to a ps square signal. - - -3-4 Optical frequency deviation (GHz) IM are analyzed. The single-sideband condition was verified by first tuning the device so as to withdraw one lateral sideband on a spectru analyzer trace and then easuring the IM and PM odulation index in odulus and phase. Fig. 8 shows the response to a ps square signal applied on the laser biased around 6 A. One notices that the transient chirp is doinating. This is to be taken into account when analyzing NRZ-SSB experient. IV. SIMULATION AND EXPERIMENTAL RESULTS FOR NRZ As seen in part II and III, perforing NRZ transission with a D-EML leads to adjusting experiental conditions to stick to J. Binder and U. Kohn conditions, i.e. atching the adiabatic laser chirp which provides the FM so as to have an optical frequency difference between ones and zeros equal to half the bit rate. This eans 5 GHz at Gb/s and GHz at 4 Gb/s. Such a frequency deviation appears too high to reach for the standard DFB laser. However a nuber of arguents leads to iproving the situation: (i) lowering the extinction ratio brings a siilar lowering of the required chirp. In high-rate tests, this effect has been verified. We notice experientally and by siulation that reducing the extinction is beneficial in dual odulation; (ii) transient chirp ust not be oitted as seen in part III; (iii) the FM frequency response of laser above the relaxation oscillation frequency decreases as f whereas the aplitude odulation decreases as f. Therefore despite the liited odulation bandwidth of our laser (about GHz) 4 Gb/s experients have been perfored. Modulating the laser current creates also an IM, which adds up to or is deducted fro the odulator IM. The first case corresponds to sending the sae signal on both inputs (we call it DATA/DATA). It tends to increase the extinction ratio, which ay or ay not be beneficial as observed in siulation. The second one (naed /DATA) uses the inverted output of the bit-sequence generator. It leads to an overall negative transient chirp [7] which can be tuned by further delaying signals one with respect to the other. All our record experients have been perfored in such a way. A. Experiental results The Table I reports the bit error rates easured in diverse experiental conditions at Gb/s, Gb/s, 5 Gb/s and 4 Gb/s. The experiental settings of the D-EML, in ters of Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 7 TxOFDM Variable delay lines Variables attenuators Custo filters ATT ATT Dual driver Either filters are active or attenuators and variable delays are active. τ τ Bias - V D-EML Laser Modulator Bias 9 A OSA 5 k k 5 k k Booster aplifier P out = W G ax = 3 db NF = 4 db Standard single ode fiber A lin =. db/k at 55 n D λ = 6 ps/n/k at 55 n x5 k x5 k 3x5 k 4x5 k In-line aplifier P out = W G ax = 6 db NF = 4.5 db lification schees x5 k 3x5 k 4x5 k Preaplifier P out = W G ax = 6 db NF = 4.5 db Band-pass filter st order Gaussian type Band = 8 GHz PIN G = db TIA R =,95 A/W Low-pass filter I 3 rd order d = 5 na N Butterworth type th =.5 pa/hz / Band = GHz RFSA RxOFDM Fig. 9. Synoptic for the OFDM siulations at. Gb/s. Paraeters has been chosen to correspond to experiental ones. The custo filter is a proposal that was not designed practically. It was ipleented in siulations. For experient, synchronized outputs of the AWG are used, one of which being digitally filtered. odulator bias voltage, laser bias current and odulating aplitudes on both inputs are optiized in each situation. The transissions are perfored without any dispersion copensation whatsoever. The Gb/s results are reported in [5] at 3 k. In order to reach 65 k a slight delay (about 6 ps) between both odulations was added. This allows a fine tuning of the transient chirp so as to have soe copensation between odulator and laser chirp. The Gb/s and 4 Gb/s results are respectively reported in [7] and [8]. In all these cases, a transission with bit error rate (BER) below the forward error correcting (FEC) liit is obtained with optical aplification soeties but no chroatic dispersion copensation. In Table I, the shorter pattern length at bit rates lower than 4 Gb/s is liited by our transitter equipent and it is not otivated by any fundaental issue. The evolution of the reachable transission distance with respect to bit rate follows a trend close to the expected square-law for chroatic dispersion liited systes. B. Siulation results Siulation work perfored using VPItransissionMaker TM and our odel of the D-EML confirs the potentiality of the dual odulation technique by giving the liits of the expectation fro the D-EML used with NRZ signals. Table I includes siulated ranges corresponding to different odulation cases of the D-EML odel. Note that (a) is the case of pure IM with IM =., (c) and (d) are the cases of IM/FM ixed with respectively IM = and Δf pp = 5 GHz, IM =.8 and Δf pp = 4 GHz, IM =.6 and Δf pp = 3 GHz. As observed in experientation, decreasing the IM index (fro case to (d)) in dual odulation allows reaching higher distances when total optical attenuation is not detriental. The 4 Gb/s results have been obtained by iproving the FM bandwidth of the laser. They exhibit thereby an increase factor fro case of single odulation (a) to case of dual odulation (d) greater than.7 eaning that the distance could be drastically iproved copare to standard EML. V. SIMULATION AND EXPERIMENTAL RESULTS FOR OFDM A. Experiental results In analog optic counication systes, H. Ki technique for the generation of optical SSB signals has only been experienced with narrowband radio over fiber (RoF) signals [3]. Previously OSSB-OFDM has been deonstrated by several authors using an odulator and an optical filter [6]. As for CMLs, this requires adding a coponent that ust be tuned to the laser. We experientally deonstrate here that direct SSB generation can be extended for wideband OFDM signals using the integrated D-EML. To do so, H. Ki's conditions [3] have to be strictly satisfied for each odulation frequency, i.e. FM and IM have to be either or π phase shifted, and the ratio between IM and FM indices ( IM /β) has to be equal to. Keeping the required phase shift between IM and FM across a wide frequency band is not straightforward. This is due to the - - -3-4 -5 OSSB - - -3-4 -5 Unodulated OVSB OSSB -6-6 -7-7 -8 536.5 536.3 536.35 536.4 536.45 536.5 536.55 Wavelength (n) Fig.. Measured optical spectra at the output of the D-EML (resolution.44 p). -8 536.5 536.3 536.35 536.4 536.45 536.5 536.55 Wavelength (n) Fig.. Siulated optical spectra at the D-EML output for the three odulation configurations (resolution p). Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 8 Transission (db) Transission (db) -8 - - -4-6 -8-8 - - -4 5 k k -6 OSSB k -8 3 4 5 6 7 8 9 3 Subcarrier index (-) Fig.. Measured channel response versus OFDM subcarriers (i.e. frequency) in configuration for different distances (top) as well as in and OSSB configurations at k (botto). non-linear phase shift (ΔΨ), between the FM and the laser odulating signal which goes fro to π/ as we ove up in frequency and the linear phase shift (Δφ), introduced by the difference in electrical length between the two access circuits. It is also coplicated to aintain the required ratio of between IM and FM indices across a large frequency band as the FM index is a frequency-dependent value. One key idea is the insertion of a passive filter in the laser access for odulation. While its phase response takes into account ΔΨ and Δφ to satisfy the phase condition for all odulating frequencies, the aplitude response of the filter is designed to keep β constant over the whole frequency band. Meanwhile constructing a real RF filter with the required characteristics, we opted for ipleenting it nuerically. To deonstrate the feasibility of the proposed technique, 5.3 GHz real-valued baseband OFDM signal is used to odulate the D-EML according to setups close to those depicted in Fig.. The nuerical OFDM signal consists of 3 sybols. Each of the is generated as follows: 56 sybols at the input of the inverse fast Fourier transfor (IFFT) are set to have a Heritian syetry. Only 4 subcarriers, QPSK apped, carry inforation data and 6 saples are used as a cyclic prefix. Two copies of this signal are used: while the first is directly -35-4 -45-5 -55-6 -65-7 -35-4 -45-5 5 k k -55-6 OVSB -65 OSSB k -7.5.5.5 3 3.5 4 4.5 5 5.5 6 Frequency (GHz) Fig. 4. Siulated electrical spectra after receiver in configuration for different distances (top) as well as in the three configurations, OVSB and OSSB at k (botto). sent to the digital-to-analog converter (DAC) to odulate the EAM, the second is nuerically filtered before feeding the DAC and odulating the DFB laser. Analog OFDM signals are generated with an arbitrary wavefor generator (AWG) at GS/s. The EAM aplifier gain is adjusted so that the electrooptical conversion is perfored in the linear region of the EAM. For the fixed IM value, the DFB laser aplifier gain is optiized to get the adequate β. Transissions are perfored over 5 k and k IM-DD dispersive channel which are long enough to show the discrepancy between SSB and DSB signals. At the receiver side, a 5 GHz photo-detector operates the optical to electrical conversion. A digital sapling oscilloscope (DSO) digitizes the electrical signal at 5 GS/s. OFDM signals are proceeded offline with Matlab. The optical spectru of the generated SSB signal, superiposed with the corresponding DSB spectru obtained by only odulating the EAM, is depicted in Fig. 9. The sideband power ratio exceeds db for all odulating coponents. Fig. 3 shows the channel response per OFDM subcarriers after propagation through 5 k and k IM-DD dispersive channel in configuration as well as after k in and OSSB configuration for coparison. While the SSB signal perits a flat response whatever the span length, the characteristic EVM RMS (%) 9 8 7 6 5 4 3 DSB 5 k DSB k SSB 5 k SSB k 3 4 5 6 7 8 9 Subcarrier index (-) Fig. 3. Measured EVM versus OFDM subcarriers (i.e. frequency) in and OSSB configurations. BER - - -3-4 -5-6 -7-8 OVSB OSSB -9 5 5 75 5 5 75 5 5 Distance (k) Fig. 5. Siulation results for the transission of OFDM signal in different configurations. Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 9 - - -3-4 -5-6 D-EML D-EML OVSB D-EML OSSB DPMZM OSSB -7 536.5 536.3 536.35 536.4 536.45 536.5 536.55 Wavelength (n) Fig. 6. Siulated optical spectra at the output of D-EML and DPMZM for coparison (resolution p). dips appear clearly with the DSB signal. As depicts in Fig. 5, OFDM subcarriers affected by frequency fading exhibit a high error vector agnitude (EVM) exceeding 8 %. In the SSB case, the EVM is kept below 33 % and its ean value over all the OFDM subcarriers is.78 % and 5.5 % respectively after 5 k and k IM-DD dispersive channel. This constitutes an iproveent with respect to our previous results [9] where no filter was used. B. Siulation results Soe retro-siulations have been carried out using VPItransissionMaker TM. Fig. depicts the synoptic of the siulated link. The OFDM generator delivers a purely real OFDM signal to the dual driver which uses either custo electrical filters or variable attenuators and delay lines to satisfy the SSB condition. In the case of filtering, this condition is respected for a large frequency bandwidth since we developed a ethodology to custo (phase and agnitude), in siulation, electrical filters (in experientations it would be digital filters) to produce the better sideband extinction at large scale. In the case of variable attenuators and delay lines, the SSB condition is only achieved for one precise frequency which is chosen close to the center of the OFDM signal bandwidth. The OFDM signal uses 56 subcarriers (5 pilot tones, 3 data subcarriers and 8 subcarriers for Heritian syetry) with a QPSK coding for a total bit rate of 4 Gb/s. The cyclic prefix is fixed to.5%. At the output of D-EML a fixed optical attenuator account for fiber coupling losses and siulated also a 9/ coupler. The % of the optical signal are used to feed the optical spectru analyzer (OSA). Several aplification schees have been used to reach k (Fig. ). Fig. shows the siulated optical spectra at the output of the D-EML in four configurations. The first one corresponds to the unodulated D-EML. The second one represents the odulation of the odulator alone (: optical double sideband). The third configuration uses the variable attenuators and delay lines of the dual driver to odulate both laser and odulator (OVSB: optical vestigial sideband). Finally, the last configuration represents the case using the custo filters to drive both laser and odulator in order to produce large bandwidth OSSB-OFDM. These spectra copare well with experiental ones. Fig. depicts the siulated electrical spectra after receiver for different distances in as well as for, OVSB BER - - -3-4 -5-6 -7-8 -9 D-EML D-EML OVSB D-EML OSSB DPMZM OSSB 3 4 5 6 7 Distance (k) Fig. 7. BER versus distance at 5 Gb/s for D-EML in the three configurations and for the DPMZM for coparison. and OSSB at k. The sae frequency fading as in experientation can be observed. They degrade significantly the signal perforances whereas the OSSB signal presents the flattest response. The BER has been evaluated at Gb/s versus the transission distance as it is shown in Fig. 4. In siulation the BER for a M-QAM is inferred fro the sybol error rate (SER) with the equation: BER = SER log (M). The SER is estiated by a statistical Gaussian ethod. A typical value of -3 for BER will be used as criterion considering the possible use of forward error corrector (FEC). The configuration OSSB allows the lowest BER with approxiately -6, -5 and 4-4 respectively for k, 5 k and k. The syste siulations including our odel of D-EML shows results in good agreeent with experiental ones thus we push our study to higher bit rate with conventional OFDM. The targeted useful bit rate is 5 Gb/s for NG-PON applications. The OFDM signal is coposed of 56 subcarriers (7 useful and 9 pilot) foring a 7.75 GHz signal transposed at 5 GHz. A 6-QAM odulation is used to increase further spectral efficiency. The cyclic prefix used is.5 %. The total bit rate is about 3.7 Gb/s allowing a 5 Gb/s useful bit rate. The siulation setup is odified in order to odulate the D- EML by the OFDM signal in a passive point-to-ultipoint architecture. No optical aplification is inserted and there is no specific chroatic dispersion copensation (except OFDM equalization as in the previous part). The output power of the D-EML is set to 5 db assuing that the coponent is packaged and fibred. The optical budget, i.e. the total attenuation between optical eitter and receiver, is controlled by the eans of an optical attenuator which accounts for luped losses due to splitting node. It is fixed to 5 db in order to be copliant with the class B of G-PON according to the ITU-T G.984 standard. The dual driver still allows the three configurations, OVSB and OSSB for the D-EML. The sae link has been studied with the use of a dual parallel Mach-Zehnder odulator (DPMZM) [7] in OSSB configuration for coparison. Fig. 6 presents the optical spectra at the output of the D- EML in the three configurations and for the DPMZM in OSSB configuration. The OVSB configuration exhibits a deep frequency fading inside the OFDM signal band which corresponds to the frequency used to satisfy the SSB condition in the dual driver. Contrary to OVSB, the OSSB configuration of the D- EML presents a good sideband power ratio (SBPR) of about db over the whole OFDM signal band. The SBPR is defined Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.

The final version of record is available at http://dx.doi.org/.9/jlt.4.34647 as the absolute difference between the power of the lower and the upper sideband. The OSSB configuration of the DPMZM perfors the best SBPR close to 6 db. Fig. 7 depicts the perforances versus distance for all the odulation schees at 5 Gb/s. In and OVSB configurations the liitation is around 4 k for a BER at -3 because of the transission ipairent due to the first frequency fading of the IM-DD channel. This fading is significantly reduced in the OSSB configuration leading to an increase of axial transission distance around 55 k. The dual odulation of the D-EML leads to perforances very close to those obtained with DPMZM both in OSSB condition. Copared to the DPMZM using LiNbO 3, the D-EML is low-cost, ore copact, and requires lower control voltage aking it a very interesting candidate as SSB onolithically integrated transitter. In pure SSB odulation, the OFDM results obtained are liited by the SNR. The linearity of the odulator response liits also the sub-carrier odulation span. One can expect that using advanced electronic pre-copensation technics ay lead to soe ajor iproveent. When coparing OSSB-NRZ, as described in Table I (also coparable to CML results), to OSSB- OFDM, one can notice that the perforances are coparable. The range tends to be higher with OFDM thanks to the ability to obtain real OSSB over the whole band of the signal, although one ay expect siilar iproveent if a custo ade filter were design for NRZ transissions. The choice between both odulation forats in a syste will depend on other criteria. The ulti-band aspect of OFDM, gives soe very interesting flexibility in a network architecture using ultiple access. However, OFDM requires soe digital processing that needs an electronic processor leading to higher power consuption. VI. CONCLUSION This paper sus up soe siulation and experiental results obtained with the dual electro-absorption odulated laser. On the one hand, we deonstrate here the possibility to generate an optical single-sideband signal with this onolithically integrated dual RF access EML by ixing FM and IM for both NRZ and OFDM odulation forats. On the other hand, we show that OSSB signal presents optiu iunity against chroatic dispersion effects in opposition to for high bit rate IM-DD transissions. We perfored transissions at the state of the art for high bit rate ainly for access networks. Experiental NRZ transissions without any chroatic dispersion copensation at rates of Gb/s, Gb/s, 5 Gb/s and 4 Gb/s were perfored with range of 6 k, 4 k, 4 k and k respectively atching siulation results. Very good quality SSB odulation for OFDM signals could be obtained with digital filtering leading to soe very flat dispersive fiber channel response. Over k transission could be obtained at. Gb/s and siulations deonstrate higher rate possibilities. It has also been noticed through experients and siulations that FM speed and FM efficiency of the laser are key eleents. The interest of D-EML for high speed SSB transissions over IM-DD channel relies on its siplicity, copactness, low-cost and reasonable power consuption. ACKNOWLEDGMENT The authors are grateful to the MODULE project fro the French national initiative ANR-VERSO progra, Systeatic Paris-region and Elopsys Liousin-region copetitiveness clusters for financial support. REFERENCES [] "ADVA Optical Networking," [Online]. Available: http://www.advaoptical.co/~/edia/resources/application Notes/G Product Faily.ashx. [] "Finisar," [Online]. Available: http://www.finisar.co/products/opticalcoponents/cml-transitters. [3] P. Chanclou, A. Cui, F. Geilhardt, H. Nakaura and D. Nesset, "Network Operator Requireents for the Next Generation of Optical Access Networks," IEEE Network, vol. 6, no., pp. 8-4,. [4] Q. Deniel, F. Saliou, S. D. Le, P. Chanclou, D. Erase and R. Brenot, "lified RSOA Self-Tuning Laser for WDM PON Using Saturated SOA for Noise Reduction and Data Cancellation," in European Conference on Optical Counication, London, 3. [5] L. Anet Neto, D. Erase, N. Genay, P. Chanclou, Q. Deniel, F. Traore, T. Anfray, R. Hadou and C. Aupetit-Bertheleot, "Siple Estiation of Fiber Dispersion and Laser Chirp Paraeters Using the Downhill Siplex Fitting Algorith," IEEE Journal of Lightwave Technology, vol. 3, no., pp. 334-34, 3. [6] J. Binder and K. U., " Gb/s Dispersion Optiized Transission at.55 μ Wavelength on Standard Single Mode Fiber," IEEE Photonics Technology Letters, vol. 6, no. 4, pp. 558-56, 994. [7] L. Anet Neto, P. Chanclou, B. Charbonnier, A. Gharba, N. Genay, R. Xia, M. Ouzzif, C. Aupetit-Bertheleot, J. Le Masson, D. Erase, E. Grarg and v. Rodrigues, "On the Interest of Chirped Lasers for AMOOFDM Transissions Through Long Distance PON Networks," in Optical Fiber Counication Conference, Los Angeles,. [8] D. Penninckx, M. Chbat, L. Pierre and J. P. Thiery, "The Phase-Shaped Binary Transission (PSBT): A New Technique to Transit Far Beyond the Chroatic Dispersion Liit," IEEE Phtononics Technology Letters, vol. 9, no., pp. 59-6, 997. [9] D. Mahgerefteh, Y. Matsui, L. C., B. Johnson, D. Walker, X. Zheng, Z. F. Fan, K. McCallion and P. Tayebati, "Error-free 5 k Transission in Standard Fiber Using Copact Gb/s Chirp-anaged Directly Modulated Lasers (CML) at 55 n," IET Electronics Letters, vol. 4, no. 9, pp. 43-44, 5. [] X. Zheng, S. Priyadarshi, D. Mahgerefteh, Y. Matsui, T. Nguyen, J. Zhou, M. Deutsch, V. Bu, K. McCallion, J. Zhang and P. Kiely, "Transission fro -36 k (6 ps/n) at Gb/s without Optical or Electrical Dispersion Copensation Using Digital Pulse Shaping of a Chirp Managed Laser," in Optical Fiber Counication Conference, San Diego, 9. [] S. Matsuo, T. Kakitsuka, T. Segawa, R. Sato, Y. Shibata, R. Takahashi, H. Oohashi and H. Yasaka, "4x5 Gb/s Frequency-Modulated DBR Laser Array for -GbE 4-k Reach Application," IEEE Photonics Technology Letters, vol., no. 7, pp. 494-496, 8. [] G. Agrawal and P. M., "Effect of Frequency Chirping on the Perforance of Optical Counication Systes," IEEE Optics Letters, vol., no. 5, pp. 38-3, 986. [3] H. Ki, S. K. Ki, H. Lee, S. Hwang and Y. Oh, "A Novel Way to Iprove the Dispersion-liited Transission Distance of Electroabsorption Modulated Lasers," IEEE Photonics Technology Letters, vol. 8, no. 8, pp. 947-949, 6. [4] H. Ki, "EML-Based Optical Single Sideband Transitter," IEEE Photonics Technology Letters, vol., no. 4, pp. 43-45, 8. [5] J. Petit, D. Erase, C. Kazierski, C. Jany, J. Decobert, F. Alexandre, N. Dupuis and R. Gabet, "Enhanced Gb/s NRZ Transission Distance using Dual Modulation of an Integrated Electro-absorption Modulated Laser Transitter," in Optical Fiber Counication Conference, San Diego, 9. Copyright (c) 4 IEEE. Personal use is peritted. For any other purposes, perission ust be obtained fro the IEEE by eailing pubs-perissions@ieee.org.