droite Performance d’un régénérateur optique à base de SOA insensible à la polarisation G. GIRAULT, M. GAY, L. BRAMERIE, V. RONCIN, J.C. SIMON Good morning everyone, I’m going to present you some results from the optronics laboratory at ENSSAT in Lannion, in France. The results concern the study of a Polarization Insensitive 3R Optical Regenerator based on a new SOA-NOLM architecture and its performance in a transmission experiment. First of all we are going to present the regenerator configuration. Then we will quickly have a look at the transmission experiment. And we will finally consider the experimental results. Laboratoire d’optronique CNRS UMR FOTON 6082 ENSSAT / Université de Rennes1 LANNION, FRANCE

Introduction Dégradations du signal  Nécessité de le régénérer : droite Introduction Dégradations du signal  Nécessité de le régénérer : amplification du signal, remise en forme du signal, resynchronisation du signal. Solution pour les transmissions à très hauts débits (>40Gbit/s) : la régénération tout optique qui évite le passage par l’électronique : La régénération 2R (Reshaping Repeater) : Amplification + Remise en forme La régénération 3R (Retiming Reshaping Repeater) : 2R + Resynchronisation. So first of all, we introduce the principle of a 3R optical regenerator: A clock recovery device generates a clean optical clock signal, modulated by the data through a reshaping gate. This leads to an extinction ratio enhancement and an amplitude noise reduction. On the right hand top of the slide, a typical reshaping gate transfer function is plotted. Both non linearities lead to amplitude noise reduction on the ones and the zeros.

Plan Introduction I. Le régénérateur 3R II. Résultats expérimentaux droite Plan Introduction I. Le régénérateur 3R II. Résultats expérimentaux Conclusion I. Le régénérateur 3R So first of all, we introduce the principle of a 3R optical regenerator: A clock recovery device generates a clean optical clock signal, modulated by the data through a reshaping gate. This leads to an extinction ratio enhancement and an amplitude noise reduction. On the right hand top of the slide, a typical reshaping gate transfer function is plotted. Both non linearities lead to amplitude noise reduction on the ones and the zeros.

La régénération 3R en modulation croisée droite La régénération 3R en modulation croisée I- Le régénérateur 3R Puissance en entrée (Ppin) Transmission (Psout/Psin) Fonction de transmission Réduction du bruit d’amplitude Amélioration du taux d’extinction Resynchronisation des données Données dégradées (Ppin) Horloge optique Données régénérées (Psout) Récupération d’horloge Porte optique non-linéaire 1 (Psin) So first of all, we introduce the principle of a 3R optical regenerator: A clock recovery device generates a clean optical clock signal, modulated by the data through a reshaping gate. This leads to an extinction ratio enhancement and an amplitude noise reduction. On the right hand top of the slide, a typical reshaping gate transfer function is plotted. Both non linearities lead to amplitude noise reduction on the ones and the zeros.

L’architecture du régénérateur 3R droite L’architecture du régénérateur 3R I- Le régénérateur 3R Données optique au format RZ (1) 1er convertisseur en longueur d’onde (NOLM-SOA) Sonde (2) 2ème convertisseur en longueur d’onde (DESOA) Données régénérées (1) Horloge optique (1) (2) (1) Récupération d’horloge Our optical regenerator consists of 2 wavelength converters including Semiconductor Optical Amplifiers The first one is a Non Linear Optical Loop Mirror whose non linear element is an SOA. The SOA NOLM function is to reduce the amplitude noise thanks to the suitable transfer function as shown earlier; The 2nd wavelength converter is a Dual Stage of SOA, that improves the output extinction ratio by gain compression in SOA. These stages will be described in detail later. Lastly a clock recovery based on a classical opto-electronical device ensures the 3R regeneration.

Onde contra - propagative droite Le premier convertisseur : NOLM-SOA I- Le régénérateur 3R GSOA temps T Données dégradées (1) ‘1’ ‘0’ ‘1’ ‘1’ 50/50 SOA  T Réflexion Fibré PM XGM   GSOA Onde contra - propagative Onde co - propagative Sonde continue polarisée (2) We describe here the first wavelength converter based on an SOA-NOLM used in a reflecting 2R configuration. In this interferometric device, a continuous probe split into 2 parts which interfere after acquiring a phase shift. The phase shift is created by data injected in the loop that modulates the SOA used in a cross gain modulation configuration. Because the SOA is not at the middle of the loop, the co and counter propagating waves see the SOA modulated gain at 2 different times leading to a gain difference. Thanks to phase amplitude coupling, the gain difference leads to the phase difference necessary for interference. At the reflected output and through an optical circulator device, the probe is modulated by the data and will constitute the data of the dual stage of SOA. Note the output data are inverted as there an extinction of the signal for a phase shift of pi, namely for a one in the data. (The SOA NOLM, usually used in a transmission configuration, was used in a reflecting configuration as which is more appropriate to a large modulation format.) Finally, to enhance the stability of the device, polarization maintaining fibre was used and the probe was linearly polarized. ‘0’ ‘1’ ‘0’ ‘0’ Data(l2) Données à la longueur d’onde de la sonde (2) Inversion de polarité par rapport à l’entrée

Le second convertisseur : DESOA droite Le second convertisseur : DESOA I- Le régénérateur 3R Données issues du NOLM-SOA (2) ‘0’ ‘1’ ‘0’ ‘0’ SOA 1 1er étage 2ème étage SOA 2 SOA 1 1er étage 2ème étage XGM SOA 2 Sonde modulée (1) SOA 1 1er étage 2ème étage XGM SOA 2 Données régénérées (1) XGM Horloge optique à la longueur d’onde des données initiales (1) Données régénérées à la longueur d’onde que celle des données en entrée du NOLM-SOA et de même polarité. Intérêt du DESOA par rapport à un SOA seul : le taux d’extinction en sortie est deux fois plus important en dB (en considérant les SOA identiques). Let’s now describe the dual stage of SOA used in a 3R configuration. In each stage, the data and the probe are coupled just before an SOA; The data modulates the SOA gain so that at the SOA output, the probe is modulated by the data. So, whereas the extinction ratio of the modulated probe at the output of a single SOA is limited by the gain compression, with a dual stage of SOA, the output extinction ratio is enhanced by 2 successive gain compressions. And finally, the transfer function of the dual stage of SOA is the square of that of a single SOA. So, the extinction ratio is twice in dB the extinction ratio of a single stage. The typical extinction ratio we can obtain with this kind of device is about 14dB. The device is then really appropriate for optical regeneration and constitutes the 2nd wavelength converter of our regenerator.

Plan Introduction I. Le régénérateur 3R II. Résultats expérimentaux droite Plan Introduction I. Le régénérateur 3R II. Résultats expérimentaux Conclusion So first of all, we introduce the principle of a 3R optical regenerator: A clock recovery device generates a clean optical clock signal, modulated by the data through a reshaping gate. This leads to an extinction ratio enhancement and an amplitude noise reduction. On the right hand top of the slide, a typical reshaping gate transfer function is plotted. Both non linearities lead to amplitude noise reduction on the ones and the zeros.

Fonction de transmission du régénérateur droite Fonction de transmission du régénérateur II- Résultats expérimentaux Mesure du taux d’extinction en sortie du régénérateur en fonction de la puissance crête de pompe en entrée: TEout TEout= 0 TEin Régénérateur 3R TEout We study the polarization dependence of the regenerator by measuring the SOA cross gain polarization dependence. For this experiment, pump and probe are continuous signals. The pump is successively linearly polarized on the two normal modes of polarization, whereas the probe is linearly polarized. The SOA polarization dependent gain is small in the small signal regime but can reach 2dB in the output power saturation regime. This SOA was then introduced in the NOLM and we have plotted here the transfer function of the device with the same conditions as before for pump and probe polarizations. This time the SOA NOLM shows a low PDL. So the experimental results show that polarization dependent losses of the SOA NOLM seem to be smaller than the polarization dependent gain of a single SOA. Numerical analysis are being carried out in the laboratory in order to study in which cases this can occur. (Hence if the SOA NOLM is installed just before the dual stage SOA, the overall regenerator appears to be polarization independent. This arises as the low polarization dependence of the SOA NOLM stabilises the dual stage SOA input power.) Nb: continuous pump and probe are equivalent to a modulated signal in a symmetric configuration (co and counter propagating waves reach the SOA at the same time). Hypothèse : est une constante.

Fonction de transmission du régénérateur (2) droite Fonction de transmission du régénérateur (2) II- Résultats expérimentaux TEin Régénérateur 3R TEout 2 non-linéarités, Possibilité d’obtenir un taux d’extinction en sortie de 13 dB pour un taux d’extinction en entrée de 8 dB. 6 -14 -12 -10 -8 -6 -4 -2 Transmission (dB) Puissance crête des données en entrée(dBm) 2 4 8 dB 13 dB We study the polarization dependence of the regenerator by measuring the SOA cross gain polarization dependence. For this experiment, pump and probe are continuous signals. The pump is successively linearly polarized on the two normal modes of polarization, whereas the probe is linearly polarized. The SOA polarization dependent gain is small in the small signal regime but can reach 2dB in the output power saturation regime. This SOA was then introduced in the NOLM and we have plotted here the transfer function of the device with the same conditions as before for pump and probe polarizations. This time the SOA NOLM shows a low PDL. So the experimental results show that polarization dependent losses of the SOA NOLM seem to be smaller than the polarization dependent gain of a single SOA. Numerical analysis are being carried out in the laboratory in order to study in which cases this can occur. (Hence if the SOA NOLM is installed just before the dual stage SOA, the overall regenerator appears to be polarization independent. This arises as the low polarization dependence of the SOA NOLM stabilises the dual stage SOA input power.) Nb: continuous pump and probe are equivalent to a modulated signal in a symmetric configuration (co and counter propagating waves reach the SOA at the same time). Remarque : les 2 non-linéarités atteintes avec un TEin de 11 dB. Or dans l’expérience TEin>11 dB  l’hypothèse, = constante, est validée.

II- Résultats expérimentaux droite Une faible pénalité II- Résultats expérimentaux Taux d’erreurs binaires : btb = back-to-back Pénalité de 0,5 dB pour un TEB de 10-9. We study the polarization dependence of the regenerator by measuring the SOA cross gain polarization dependence. For this experiment, pump and probe are continuous signals. The pump is successively linearly polarized on the two normal modes of polarization, whereas the probe is linearly polarized. The SOA polarization dependent gain is small in the small signal regime but can reach 2dB in the output power saturation regime. This SOA was then introduced in the NOLM and we have plotted here the transfer function of the device with the same conditions as before for pump and probe polarizations. This time the SOA NOLM shows a low PDL. So the experimental results show that polarization dependent losses of the SOA NOLM seem to be smaller than the polarization dependent gain of a single SOA. Numerical analysis are being carried out in the laboratory in order to study in which cases this can occur. (Hence if the SOA NOLM is installed just before the dual stage SOA, the overall regenerator appears to be polarization independent. This arises as the low polarization dependence of the SOA NOLM stabilises the dual stage SOA input power.) Nb: continuous pump and probe are equivalent to a modulated signal in a symmetric configuration (co and counter propagating waves reach the SOA at the same time).

Une faible dépendance à la polarisation droite Une faible dépendance à la polarisation II- Résultats expérimentaux Dépendance en polarisation du SOA en pompe/sonde Dépendance en polarisation du NOLM-SOA Dispositif PM, sonde polarisée TE, signaux continus -35 -30 -25 -20 -15 -10 -5 5 8 9 10 11 12 13 Gain sur la sonde (dB) Puissance de pompe en entrée (dBm) pompe TE pompe TM -8 -6 -4 -2 Fonction de transmission de la sonde avec normalisation (dB) We study the polarization dependence of the regenerator by measuring the SOA cross gain polarization dependence. For this experiment, pump and probe are continuous signals. The pump is successively linearly polarized on the two normal modes of polarization, whereas the probe is linearly polarized. The SOA polarization dependent gain is small in the small signal regime but can reach 2dB in the output power saturation regime. This SOA was then introduced in the NOLM and we have plotted here the transfer function of the device with the same conditions as before for pump and probe polarizations. This time the SOA NOLM shows a low PDL. So the experimental results show that polarization dependent losses of the SOA NOLM seem to be smaller than the polarization dependent gain of a single SOA. Numerical analysis are being carried out in the laboratory in order to study in which cases this can occur. (Hence if the SOA NOLM is installed just before the dual stage SOA, the overall regenerator appears to be polarization independent. This arises as the low polarization dependence of the SOA NOLM stabilises the dual stage SOA input power.) Nb: continuous pump and probe are equivalent to a modulated signal in a symmetric configuration (co and counter propagating waves reach the SOA at the same time). NOLM-SOA est donc beaucoup moins sensible à la polarisation que le SOA seul

La boucle à recirculation droite La boucle à recirculation II- Résultats expérimentaux Données émises : 10 Gbit/s - PRBS 215 -1 Impulsions de 50 ps @ 1552 nm au format RZ Emission ASE EDFA NZDSF Pompe Raman DCF Emission Régénérateur Optique 3R Brouilleur de polarisation Réception EDFA et pompages Raman contra-propagatifs  compensation des pertes dans la fibre avec minimisation de l’accumulation de bruit dans la ligne ainsi que des effets non-linéaires. Pompe Raman EDFA Le Régénérateur 3R placé en fin de boucle Régénérateur Optique 3R DCF NZDSF NZDSF : 2  50 km - DCF : dispersion quasi-compensée Brouilleur de polarisation : test de la sensibilité à la polarisation du régénérateur Fréquence de modulation ~ 1MHz Brouilleur de polarisation OSNR dégradé par l’ajout d’une source d’émission spontanée amplifiée : Possibilité d’étude du TEB en fonction du nombre de tours. ASE Let’s now have a look at the transmission experiment which was carried out using a 10 Gbit/s recirculating loop. The signal transmitted was a single wavelength RZ signal. The transmission fibre was Non Zero Dispersion Shifted Fibre so slightly dispersive at our working wavelength. Spans of 100 km were realised. Fiber losses were compensated with Erbium doped fibre amplifiers and counter-propagating Raman pumping ensuring a low noise accumulation line. The 3R regenerator was installed at the end of the loop and a polarization scrambler was installed in front of the regenerator in order to study its polarization dependence. Under these conditions, with an Optical Signal to noise ratio of 33dB in front of the first regenerator, no error was measured after 100 000 km propagation, during more than ha(l)f an hour, ensuring a Bit error rate better than 10 to the power of -10. This ensures a really good stability of the device. Then we artificially degraded the OSNR in order to measure the BER evolution in the regenerated line. This was done by introducing an Amplified Spontaneous Emission Source. Pour un OSNR (rapport signal à bruit) de 33 dB (0,1 nm) au premier tour : aucune erreur n’a été mesurée après une transmission de 100 000 km (1000 passages dans le régénérateur) pendant 30 minutes (TEB<10-10).

Evolution du taux d’erreurs binaires droite Evolution du taux d’erreurs binaires II- Résultats expérimentaux 1 10 100 1000 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 log(TEB) Nombre de tours (N) OSNR (0.1nm) 21dB 24dB [1] : J.Mork et al., ‘Analytical expression for Bit Error Rate of cascaded All-optical Regenerators’, IEEE Phot. Tech. Lett., vol. 15, no.10, oct. 2003 [1] Let’s now consider the results: We show here the BER evolution with the number of laps, with (in red) and without (in blue) a regenerator, with an OSNR of 21dB in front of the 1st repeater for the triangles and an OSNR of 24dB for the squares. Firstly, without regenerator (in blue), less than 10 laps were achieved with an acceptable BER. This is not surprising since the BER accumulation in a linearly amplified transmission link varies exponentially with the number of laps. The results then obtained including the regenerator are shown in red. In the theoretical literature, for a perfect step gate, the BER should increase linearly with the number of laps. This is not exactly the case in our results as the non linear gate is smoother than an ideal step gate. However, extrapolation of the BER evolution shows that the BER tends to become linear after a few cascades of regenerators. Unfortunately, this is less obvious in the case of the 24dB OSNR. This arises because the measurement time needs to be really long for a good estimation of BER of 10-8 after 100 or more laps in a recirculating loop experiment. However the BER evolution demonstrates once again the very good stability of the device.

droite Conclusion Expériences avec un régénérateur optique 3R insensible à la polarisation présentant une architecture originale à base de SOA. Grande stabilité du dispositif expérimental qui a permis de réaliser une transmission comprenant 1000 passages dans le régénérateur. Possibilité de propager le signal sur 100 000 km avec un TEB d’environ 10-8 pour un OSNR de 24 dB (0.1nm) mesuré devant le régénérateur au premier tour (sans code correcteur d’erreur). Possibilité d’étendre les expériences au débit de 40 Gbit/s en utilisant des SOA adaptés. Expérience sur l ’évolution du TEB en fonction du nombre de tours et donc de passages dans le régénérateur en adéquation avec la théorie [1]. As a conclusion, we have described in our presentation a polarization insensitive, optical 3R regenerator based on SOA gates. A transmission experiment showed good device stability after 1000 cascades of regenerators. Moreover by deliberately degrading the signal, we show that it was possible to propagate the signal over 100 000km with a Ber of about 10-8 with an OSNR as low as 24dB. Moreover this experiment could be extended to 40Gbit/s using more recent SOA chips. This device offered us the possibility of studying BER evolution as a function of the number of laps in a regenerated line; experimental results were in accordance with the theory. Finally the question of the assessment of low BER characterisation of optical regenerators was raised. [1] : J.Mork et al., ‘Analytical expression for Bit Error Rate of cascaded All-optical Regenerators’, IEEE Phot. Tech. Lett., vol. 15, no.10, oct. 2003

Des questions ? Remerciements Conseil régional de Bretagne droite Des questions ? Remerciements Conseil régional de Bretagne Commission européenne (F.E.D.E.R.) Ministère de la Recherche et Nouvelles Technologies Thank you for your attention, and feel free to ask questions. Thank you.

Une faible dépendance à la polarisation (2) droite Une faible dépendance à la polarisation (2) II- Résultats expérimentaux Fonction de transmission du NOLM-SOA en réflexion : avec K : une constante, Gco : le gain du SOA vu par l’onde co-propagative, Gcontra : le gain du SOA vu par l’onde contra-propagative,  : le coefficient de couplage phase/amplitude dans le SOA.  où   Let’s now have a look at the transfer function of the SOA NOLM to understand its low polarization dependence. The reflected output power depends on the SOA co-propagating gain, the co and counter gain difference, the phase shift difference and the probe input power. Then the SOA NOLM PDL in dB is the sum of the SOA PDG and a term that is a correlation of many parameters. Considering this equation, it arises when the correlated term is negative; Simulations are being explored to find in what conditions is this realised.  Ainsi, PDLNOLM(dB) < PDGSOA si A(dB)TE - A(dB)TM < 0

Une faible dépendance à la polarisation (3) droite Une faible dépendance à la polarisation (3) II- Résultats expérimentaux où  et Ainsi, PDLNOLM(dB) < PDGSOA si A(dB)TE - A(dB)TM < 0 Si tout est parfaitement symétrique dans le NOLM-SOA, avec des signaux continus, on a GSOA = 0 et donc AdB = 0, d’où PDLNOLMdB = PDGSOAdB. Or, s’il existe une dissymétrie (place du SOA dans la boucle en dynamique ou différence de couplage fibre/puce dans le SOA en statique). Ainsi, GSOA  0 et comme GTE  GTM alors, GSOATE  GSOATM. Let’s now have a look at the transfer function of the SOA NOLM to understand its low polarization dependence. The reflected output power depends on the SOA co-propagating gain, the co and counter gain difference, the phase shift difference and the probe input power. Then the SOA NOLM PDL in dB is the sum of the SOA PDG and a term that is a correlation of many parameters. Considering this equation, it arises when the correlated term is negative; Simulations are being explored to find in what conditions is this realised. Ainsi, Dissymétrie dans le SOA  possibilité d’avoir A(dB)TE - A(dB)TM < 0

Dispositif opto-électronique de récupération de l’horloge optique droite Dispositif opto-électronique de récupération de l’horloge optique Récupération d’horloge BA BA PD LiNbO3 modulateur laser DFB (1552 nm) Une photodiode à 10 GHz suivie d’un amplificateur large bande (BA) Un amplificateur limiteur Une récupération d’horloge Le signal remis en forme et resynchronisé module un signal optique issu d’un laser DFB via un modulateur à LiNbO3.

droite

droite We describe here the first wavelength converter based on a SOA-NOLM used in a reflected configuration. In this interferometric device the optical clock split into 2 parts by a coupler that interfere after acquiring a phase shift. They have the function of probe for the SOA used in a cross gain modulation configuration. Data, that have the function of pump, are injected in the loop and modulate the SOA gain. They reach the SOA at the same time than the co-propagating wave. Whereas the counter-propagating wave is delayed and doesn’t see any gain modulation leading to a gain difference between the 2 waves. Thanks to phase amplitude coupling, this gain difference leads to phase difference necessary for interference. At the reflected output, probe is modulated by the data and polarity is inverted as there is an output extinction for a phase difference of pi namely with a one in data . The SOA NOLM is usually used for optical regeneration in a transmission configuration as it leads to a better extinction ratio. Nevertheless it was here used in a reflective configuration as it leads to a low phase dependence of the device and then a more stable configuration. Moreover, to enhance that stability, polarization maintaining fibre was used. Finally the SOA NOLM transfer function presents 2 non linearities allowing ones and zeros regeneration has shown before.

Dispersion Dispersion partially compensated : 22 ps/nm/laps . droite Dispersion Dispersion partially compensated : 22 ps/nm/laps . But dispersion not critical under 10 laps of propagation : BER deduced from OSNR in a linear assumption = BER measurement .