A Passive All-Optical Device for 2R Regeneration Based on the Cascade of Two High-Speed Saturable Absorbers

A Passive All-Optical Device for 2R Regeneration Based on the Cascade of Two High-Speed Saturable Absorbers
of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 29, NO. 9, MAY 1, 2011 1319 A Passive All-Optical Device for 2R RegenerationBased on the Cascade of Two High-SpeedSaturable Absorbers Hoang Trung Nguyen, Coraline Fortier, Julien Fatome, Guy Aubin, and Jean-Louis Oudar  Abstract— We discuss the design and realization of a passive all-optical device for 2R regeneration based on a dual-stage of high-speed microcavity saturable absorbers, one for noise reduction of digital zeros (SA-0), and the other for noise reduction of digitalones (SA-1). The numerical and experimental results showed thatby using a simple combination of SA-0 and SA-1 devices, one canobtain an intensity transfer function with a large extinction ratioimprovementoflowpowerlevelsandastronglynonlinearresponsereducing the noise of high power levels. The amplitude and phasecharacterization of a 40-GHz signal transmitted by this device, ob-tained by frequency-resolved optical gating measurements, revealstheintensity-dependantpulse-compressioneffectandthelowchirpintroduced by this device.  Index Terms— All-optical switching gates, asymmetricFabry–Perot devices, high-speed optical techniques, nonlinearoptics, optical regeneration, optical saturable absorption, semi-conductor quantum well (QW). I. I NTRODUCTION H IGH-SPEED networks will need to use ultrafast opticaldevices with extremely low chirp. Optical-communica-tionsystemsaresubjecttoseveralsourcesofsignaldegradation.Some impairments, such as signal attenuation and dispersion,can be cured by amplification and dispersion compensation.However, the random noise from amplifiers requires moregeneral methods of signal restoration. Ultrafast all-opticalreshaping regeneration (2R regeneration) is necessary forhigh-speed networks to improve the SNR through extinctionratio (ER) improvement and noise redistribution [1]. Theseimprovements can be achieved with devices, which have anonlinear transfer function (output power as a function of input power). This nonlinear response enables to increase thediscrimination between logical ones and zeros [1 ]. Severalschemes of2R regeneratorhavebeenproposed [2]–[8].Amongthese, the simple scheme of all-optical 2R regenerator based Manuscript received September 17, 2010; revised December 02, 2010; ac-cepted January 18, 2011. Date of publication February 22, 2011; date of currentversion April 22, 2011. This work was supported in part by the French NationalResearch Agency (ANR—FUTUR).H. T. Nguyen, G. Aubin, and J.-L. Oudar are with the Laboratory forPhotonics and Nanostructures, Centre National de la Recherche ScientifiqueUPR20, 91460 Marcoussis, France (e-mail: Fortier and J. Fatome are with Laboratoire Interdisciplinaire Carnot deBourgogne, UMR 5209 CNRS/Université de Bourgogne, 21078 Dijon, France(e-mail: versions of one or more of the figures in this paper are available onlineat Object Identifier 10.1109/JLT.2011.2117413 on saturable absorbers (SAs) in optical microcavity is partic-ularly attractive, owing to its simplicity and its fully passiveoperation mode (no Peltier cooler, no bias voltage) [9]. Recentexperiments using high-repetition-rate signals showed that withthe benefit of quantum wells (QWs) having an ultrafast carrierrecombination rate [10], [11], SA microcavity devices are very promising devices for 2R regeneration at 160 Gb/s [12 ], [13]. Nevertheless, SAs actually have only been used to achieve anER improvement without impacting the noise in bit-1 level. Inorder to perform a complete regeneration, SAs have typicallybeen used in combination with other devices that can providethe level stabilization at the bit time scale. While the recentadvances in fiber-SA-based gates or SOA-SA-based gates haveallowed to demonstrate attractive characteristics, they typicallysuffer from limited performance concerning the polarizationdependence, the induced chirp, the requirement of strong inputpower [14], [15] and insufficient speed [15], [16 ]. Moreover, they are not adequate for wavelength-division multiplexing(WDM) applications. The recent investigation of a SA-basedmicrocavity passive all-optical device, which can provide anamplitude stabilization, and thus, a noise reduction of bit-1levels, allows a solution for passive all-optical regenerationbased on SAs [17], [18]. The potential advantage of this ap- proach is to make a complete passive all-optical high-bit-rate2R regeneration relying upon a single technology. SA-basedmicrocavities, on the other hand, offer a number of uniqueadvantages, including polarization-independent operation, lowinsertion loss, high contrast ratio, and high speed and compati-bility with WDM applications [19].In this paper, we propose and study experimentally passiveall-optical-device-based SAs that could be used for 2R regen-eration. We demonstrate that some important properties of thepower transfer function, such as a large attainable ER and astrong nonlinearity, can be achieved by using a simple combi-nation of two SA devices named SA-0 and SA-1. The SA-0 isused to improve the ER, while the SA-1 provides a significantstabilization at high amplitude levels, representing the logicalones.Furthermore,a40-GHzsignalprocessedbythedual-stagedevice has been characterized in intensity andphase by frequency-resolved optical gating (FROG) measure-ments [20].II. R EVIEW OF  D EVICE  D ESIGN The nonlinearity of SA-based devices is achieved by themodulation of carrier density in active layer. In order to ob-tain good saturation characteristics, InP-based multi-QWs, 0733-8724/$26.00 © 2011 IEEE  1320 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 29, NO. 9, MAY 1, 2011 with a strong excitonic absorption in the telecommunicationwavelengths window, were used as active layers. To amplifythe nonlinear response and to reduce the switching energy, theactive layer is contained in an asymmetric Fabry–Perot micro-cavity (AFPM). The QWs are suitably located at one antinodeof the intracavity intensity. The AFPM enhances the intracavityintensity of electromagnetic field and, thus, combined with theexcitonic absorption, improves the nonlinearity and decreasesthe switching energy. Furthermore, the SA microcavity deviceis used at normal incidence, thus yielding intrinsic polariza-tion-insensitive operation.The total field reflectivity of AFPM results from inter-ferences between the front mirror with reflectivity and theback mirror with reflectivity , modulated by the absorption inthe active region(1.1)Equation(1.1)expressesthe intermsof ,theeffectivebottom reflectivity , and the single-pass dephasing . Here, where is the QW number, is the (carrierdensity dependent) single-pass absorption per QW, and is thelongitudinal confinement factor of the microcavity.As shown in (1.1), the total reflectance atthe resonant wavelength is strongly dependent on the values of and on the absorption in the cavity. Mathematically, inthe case of SA-1, where decreases when ab-sorption decreases due to a partial cancellation of the reflectionfrom the top and the bottom reflectors. is null when theinternal absorption exactly balances the two reflections. A de-crease of absorption beyond this point will cause an increase of . This design provides a stabilization of the reflected outputpower, which can be used for noise reduction of logical ones. Inthe case of SA-0, where increases when the ab-sorption decreases, and reaches its maximum when the absorp-tion is fully saturated. The SA-0 can be used for ER improve-ment and noise reduction of logical zeros. Thus, with an initialstructure, by modifying the cavity parameters of the AFPM, wecan perform a bit-1 regeneration device or an ER improvementdevice.MQW-SAstypicallyhavearatherlongrecoverytime,arounda few nanoseconds. In order to reduce significantly the responsetime of SAs and make them compatible with high-bit-rate oper-ation, recombination centers must be introduced during or aftercrystalgrowthbymeanoflow-temperaturemolecularbeamepi-taxy [21], ion implantation [22], heavy ion irradiation [10], or Be [23] or Fe doping [24]. Recoverytimes in the subpicosecond range can be achieved by these methods of damage creation.In order to study the transfer function of the device for 2Rregeneration, several microcavities SA-0 and SA-1 have beenfabricated. SAs used in this paper are based on the same initialstructure.Theabsorptionregionincludes7 (AlGaAs/InGaAs)QWs, grown by metal-organic vapor-phase epitaxy (MOVPE).The QWs are contained in an asymmetric Fabry–Perot andlocated at one antinode of the intracavity intensity. The back mirror was made by deposition of a silver layer, with a cal-culated reflectivity of 0.945. The sample was mounted on a Fig. 1. Transfer functions of SA-0 (a) and SA-1 (b) with different reflectivitiesof front mirror. The straight lines are the linear functions. Si substrate by Au–In bonding to make easier the heat dissi-pation and limit the thermo-optic effects [25]. As previouslyindicated, the SA characteristics depend strongly on the cavitymirror reflectivities and on the absorption of the active layerincluded in cavity. In order to investigate the influence of eachcomponent on the operation of the 2R regeneration device, dif-ferent front mirror reflectivities were used. In the case of SA-0structures, two devices with different ER improvements havebeen fabricated. The first was designed to have a reflectivityclose to zero at low intensity (impedance matching) giving ahigh ER improvement. The top mirror in this case consistedof two pairs of [SiO /TiO : ] with 0.78 powerreflectivity. The second device had a front mirror made of 6.5[InP/InGaAsP] with 0.70 reflectivity, resulting in a lower ERimprovement but a lower insertion loss. For SA-1 structures,the top mirrors were, respectively, made by deposition of 3 [YF /ZnS: ], with a calculated reflectivity of 0.92, and by deposition of 5 [InP/InGaAsP: ] anda one-period dielectric coating [YF /ZnS: ] on thetop, to obtain a front mirror reflectivity of 0.88. The changeof front mirror reflectivities allows to modify the stabilizationpower level of the device. Fig. 1 displays the reflected outputfluence (energy density) of SA-0 [see Fig. 1(a)] and of SA-1[see Fig. 1(b)], as a function of input fluence. These transferfunctions were measured at 1550 nm with a pulse source at10-MHz repetition and 0.8-ps pulse duration. The influence of the front mirror reflectivity on the behavior of each SA deviceis clearly apparent on the figures. A larger ER improvementcan be achieved with the structure SA-0 working at impedancematching.The transferfunctionsof SA-1sshowthat therelativeamplitude fluctuation at the output would be reduced in theinput energy fluence of [1–6 J/cm ] and of [0.6–2 J/cm ]in the case of and respectively. More-over, the threshold of high power stabilization of SA-1  1322 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 29, NO. 9, MAY 1, 2011 Fig.2. Numericaltransferfunctionsforthe(a)(SA-1    SA-0)and(b)(SA-0   SA-1) configurations, where SA-1 is fixed and SA-0 varies and (c) SA-1 varies. IV. N ONLINEAR  R ESPONSE OF THE  D EVICES ToestimatethepotentialfornoisereductionandERimprove-ment of this device, we measured its transfer functions, i.e., thetime-averaged output power versus input power. The experi-mental setup consists of a mode-locked fiber laser producing0.8-ps pulses at 10 MHz, with wavelength adjustable in therange of 1540–1565 nm. The dual-stage (SA-0 SA-1) actsas a complete 2R regenerator. The input signal, after passingthrough a variable attenuator, is focused onto the sample SA-0.The signal transmitted by SA-0 is injected directly into SA-1duetotheopticalcirculator.Microlensedfibersproducingaspotsize of 3.1 m diameter at 1/  e  intensity were used to focus op-ticalsignalsontheSAdevices.Thereflectedsignalwasdetectedand analyzedby an optical spectrum analyzer (OSA).In this ex-periment, there was no amplifier between SA-0 and SA-1, andthe coupling losses were estimate to dB.The experimental and numerical transfer functions of (SA-0 SA-1) devices are displayed in Fig. 3. The parametervalues of SA-0 and SA-1 used for simulation are indicated inTable I. As can be seen, a good agreement is obtained betweenthe calculated and experimental results for incident fluences upto 1.5 J/cm . The disagreement at low input fluence in thecase, where SA-1 is fixed and SA-0 with is related TABLE IP ARAMETERS OF  SA-0  AND  SA-1  FOR  S IMULATION Fig.3. Reflectedoutputfluenceasafunctionofinputfluence,forseveraldevicecombinations. (a) (SA-0+SA-1) SA-1 fixed with         , SA-0 varies. (b)(SA-0    SA-1) SA-0 fixed with         , SA-1 varies. to the limited sensitivity of the OSA. As shown on Fig. 3,the effect of ER improvement and stabilization at high levelcan be obtained in this device combination. The possibility toadjust the threshold and ER improvement of the module usingthe SA-0 and SA-1 individual characteristics is demonstrated.Fig. 3(a) shows the transfer functions of the (SA-0 SA-1)module when the front mirror of SA-1 is fixed to .Our measurement showed that the ER is significantly improvedwith a SA-0 of 0.78 (impedance matching). The change inSA-0 does not change the nonlinearity of the full device athigh power level. The increase of stabilization threshold from22 J/cm to 37 J/cm when changing the front mirror of SA-0 from to is related to the higherinsertion loss of the SA-0 at impedance matching. In order todetermine the influence of SA-1 on the operation of the (SA-0SA-1) module, we measured several transfer functionswith different SA-1 parameters. The SA-0 was then fixed within this case. Fig. 3(b) shows that the threshold forpower stabilization increases from 10 J/cm to 22 J/cmwhen the front mirror of SA-1 changes from 0.88 to 0.92.The ER improvement of the module did not change.  NGUYEN  et al. : PASSIVE ALL-OPTICAL DEVICE FOR 2R REGENERATION 1323 Fig. 4. Experiment setup. Att: variable optical attenuator; EDFA: erbium-doped fiber amplifier; SHG-FROG: second-harmonic regeneration-frequency-resolved optical gating. The experimental results showed that a complete 2R regener-ation can be obtained with the simple high-speed SAs (SA-0SA-1) combination device. The optimum configuration couldbe obtained with the SA-0 working at impedance matching. Areduction of the threshold of bit-1-level stabilization would beexpected by reducing the coupling loss between two compo-nents. The experimental study of noise reduction involving thetwo devices operating at the bit rate and bit pattern differencespresented in [30] demonstrate the promising character of (SA-0SA-1) devices for 2R regeneration at 40 Gb/s.V. A MPLITUDE AND  P HASE  C HARACTERIZATION The chirp parameter is one of the most important character-istics for modulators. It is proportional to the ratio between thephasevariationandtherelativeamplitudevariation.Anaccurateestimation of the system performance requires the determina-tion of amplitude and phase of the processed signals. In order tocharacterize the signal in amplitude and phase at high repetitionrates we measured the characteristics of a 40 GHz pulse traintransmitted by the device, by using the SHG-FROG technique.  A. SA-1 Characterization The experimental setup is shown in Fig. 4. A high-quality40 GHz, 7 ps transform-limited pulse train was first generatedat 1551 nm (the resonance wavelength of the SA-1 device) byusing the multiple four-wave-mixing technique in a 1420 mlong Teralight fiber. The signal was focused on the SA-1 de-vice thanks to a fiber-pigtailed high aperture lens producing afocused spot 5 m diameter at of maximum intensity.The transmitted signal was analyzed by a power meter and asecond-harmonic-generation FROG device [13].The characterization in amplitude and phase of the SA-0 de-vice has already been presented in [12]. The FROG measure-ments showed the pulse compression effect and no nonlinearphase distortion of SA-0. In this paper, we focus the work onthe SA-1 device.In order to determine the operating regimes of the SA-1 de-vice, we measured it transfer function at 40 GHz. The results,presented in Fig. 5(a), showed that at resonance, SA-1 providesa significant amplitude stabilization of the reflected signal inthe average input power range from 40 to 90 mW. Fig. 5(b)showstheintensityandphaseprofileswithdifferentinputpowerlevels. At 15 dBm, the SA-1 is in the linear area. The pulse andphaseprofiles areidenticaltothemeasured reference.Thepulsecompression effect occurs when the incident power increases,inducing the nonlinear behavior of SA-1. The phase profile isslightly changed. The experimentalresults are demonstrated the Fig. 5. (a) Transfer function of SA-1 device. (b) FROG result. Intensity profileand phase of the 40 GHz pulse train at the SA-1 input: Au mirror reference andafter SA-1 output for linear and nonlinear regimes.Fig. 6. Experiment setup. Att: variable optical attenuator. 1     delay line.EDFA: erbium-doped fiber amplifier. OC: optical circulator. SHG-FROG:second-harmonic regeneration—FROG. dependence of the pulse compression effect on the input power,and showed the low chirp property of SA-1.  B. (SA-0 SA-1) Module Characterization In this section, we describe the characterization of a signaltransmitted by a module composed of successive devices SA-0and SA-1. The experimental setup is presented in Fig. 6. Thesignal is composed of a pulse train with duration of 5 ps at 40GHz repetition. Then, a 50:50 coupler combined with a 12.5 psdelay line was use to generate through multiplexing in the timedomain, the following simplified bit pattern “ ”at 80 GHz. A variable attenuator has also been inserted intothe delay line in order to adjust the energy of the ghost-pulsesinjected into the “0” bit slots.The transfer function measurements have been realized ondual-stage SA-0 and SA-1. As can be seen in Fig. 7(a), the
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks