Optics Communications

Transcription

1 Optics Communications 283 (2010) Contents lists available at ScienceDirect Optics Communications journal homepage: Analysis of the influence of backscattered optical power over bidirectional PON links J.J. Martínez *, I. Garcés, A. López, A. Villafranca, M.A. Losada Grupo de Tecnologías Fotónicas. I3A, University of Zaragoza, Zaragoza, Spain article info abstract Article history: Received 10 September 2009 Accepted 22 January 2010 Keywords: Passive Optical Networks (PONs) Non-linear backscattering Brillouin Scattering (BS) Rayleigh Scattering (RS) Our aim is to describe the behavior of non-linear scattering effects that arise in standard single mode fiber (SMF), specifically scattering effects that propagate optical power in the reverse direction of the source signal such as Rayleigh Scattering (RS) and Brillouin Scattering (BS). For this purpose, the effects of backscattering phenomena over a bidirectional data transmission in a passive optical network (PON) scheme have been assessed. The impact of these high optical power components over reception at the optical line terminal (OLT) side has been determined when both links use the same wavelength. Bit Error Rate (BER) measurements have been performed with different transmission rates, using several techniques to mitigate the influence of backscattering over the received signal and considering cases with filtered and unfiltered BS. Ó 2010 Elsevier B.V. All rights reserved. 1. Introduction The increasing demand of bandwidth from end users of communication networks has lead to a restricting bottleneck in access networks. The current state-of-the-art solution is the use of wavelength division multiplexing (WDM) applied to Passive Optical Networks (PON) in order to obtain higher capacities and to allow a broader range of subscribers with the same network infrastructure [1]. Amongst the different WDM solutions, a common scheme proposes the use of the same wavelength path in the fiber for both downstream (server or OLT to client or ONU) and upstream (ONU to OLT) transmission [2]. Also, in these architectures ONUs are usually remotely fed (optical loopback) in order to avoid the need for optical sources in the user terminals [3]. The application of this technique and the creation of reflective ONUs involve important advantages by simplifying the design and reducing its cost. Moreover, the absence of a specific light source makes the behavior of the terminal independent from wavelength (colorless). Therefore, a single ONU model could be reused at the same network for different wavelengths, or even in similar networks, without changes, reporting important advantages of scalability and flexibility. On the other hand, reflective colorless ONUs can be the source of important drawbacks in the user terminals that must be assessed. In particular, it is necessary to avoid all the possible interferences derived from reusing the same optical carrier dedicated to the downstream channel. There are several approaches to ensure a * Corresponding author. Tel.: ; fax: addresses: jjosemar@unizar.es (J.J. Martínez), ngarces@unizar.es (I. Garcés), aliclope@unizar.es (A. López), asiervv@unizar.es (A. Villafranca), alosada@unizar.es (M.A. Losada). successful upstream communication, such us using different modulation schemes for downstream and upstream channels [4,5], erasing the previous modulation [6], or feeding with a wavelength not used in the downstream communication. Moreover, an important issue to be taken into account is that signals propagating towards the source (OLT) can be impaired by interference with backscattering effects that take place into the PON link, i.e., a standard single mode fiber (SMF). Both Rayleigh (RS), at the same wavelength of the signal, and Brillouin backscattering (BS), at a lower frequency, may pose a negative influence in the transmission by means of crosstalk, high power reflections etc. Their influence may be critical since they can t be avoided or reduced like the same effects caused by reflections in the plant, easily prevented by angled connectors and splices [7]. It has to be noted that, in the literature, RS is the only backscattering effect that is being taken into account in bidirectional links where it has been shown that it causes a power penalty in reception [8,9]. BS has traditionally not been considered because its frequency is out of the electrical bandwidth of the receiver for commonly used bit rates. However, it inserts a high optical power level at the receiver end and it can be especially deleterious in links working at 10 Gbps because, in this case, the BS frequency is close to the received signal bandwidth. There are different methods that have been proposed and demonstrated both theoretical and experimentally, to reduce the power of backscattering non-linear effects. One of the simplest is the spectral broadening of the optical source signal. Previous literature has shown that the RS signal is characterized by a Lorentzian-shape optical spectrum so that the optical power detected at the receiver depends on the ratio of the receiver bandwidth and the optical bandwidth of the signal. Thus, decreasing this ratio by broadening the spectrum of the optical source has the effect of /$ - see front matter Ó 2010 Elsevier B.V. All rights reserved. doi: /j.optcom

2 2244 J.J. Martínez et al. / Optics Communications 283 (2010) reducing the effective RS power [10 12]. For the BS effect, it has been stated that for an optical source signal with a Lorentzian profile spectrally wider than the Brillouin linewidth, the BS gain in the SMF is reduced [13,14], and thus the power threshold at which the BS appears increases. This effect has also been observed for pumping signals with non-uniform modulations, such as amplitude, phase or frequency [6,15,16]. In this paper, we analyze experimentally the influence of RS and BS in bidirectional optical networks. First, measurements of Rayleigh and Brillouin backscattered powers as a function of fiber injected power are presented. In this analysis, different source broadening values were applied to the optical signal in order to assess the reduction of the RS and BS powers. Next, the effects over the received spectra when placing an RSOA are established. Finally, the impact of RS and BS on the performance of an upstream data link is evaluated at different data rates. BER measurements were carried out under different conditions to assess the relative influences of BS and RS on signal reception, first for data rates for which the BS falls well outside the receiver bandwidth and finally, for 10 Gbps. 2. Backscattering behavior The basic experimental set-up depicted in Fig. 1 was used to evaluate the different backscattering effects that take place in a remote-fed PON and affect the reception of the upstream channel. A 1544 nm standard DFB laser of linewidth around 1 MHz, with a linewidth enhancement factor of 3.6 and an adiabatic chirp of 3 GHz/mW was used as the optical source. Discrimination of fiber incoming and outgoing signals was performed by a circulator. Backreflected signal was redirected to an optical High Resolution Optical Spectrum Analyzer (HR-OSA), to obtain a precise measurement of the wavelength and power of the backscattering peaks. Non-linear scattering phenomena take place in a 25-km spool of SMF, with an angled connector in its final extreme to avoid reflections. In the final part of the experiment an RSOA (Reflective Semiconductor Optical Amplifier) was placed at the end of the SMF spool to verify the behavior of the backscattering effects in the presence of amplification. The performance evaluation was carried out by varying the optical power injected to the fiber and the linewidth of the laser source, which was broadened by directly modulating its bias current. Frequency chirping is a well-known phenomenon in directly modulated semiconductor laser sources that is generally considered a detrimental effect. However, in this work we use the effects of adiabatic chirp to our advantage. By means of introducing a weak excursion in the bias current in the range of frequencies where the adiabatic chirp is stronger, it is possible to achieve a broadening in the optical spectrum of the laser source [4,17]. As previously stated, this spectral broadening reduces the power of backscattering effects generated in the fiber. In our experiment we applied over the bias current a 1 GHz signal with the following amplitude values: no modulation, ±1.5 mapp (30 MHz broadening FWHM) and ±2.5 mapp (60 MHz broadening FWHM). Fig. 2 shows the spectra obtained at different laser bias intensities for two injected powers (60 ma for 7 dbm and 110 ma for 10 dbm) and two source broadenings (no modulation and ±2.5 mapp). The Brillouin peak can be clearly seen 10.8 GHz below the Rayleigh peak, which is placed at the central wavelength of the injected source. Higher optical power input to the fiber results in higher BS power. RS also increases, but not as steeply as BS. Measurements obtained when the laser source is slightly directly modulated are also included in both plots. A noticeable reduction of BS power is clearly visible in the presence of spectral broadening. Measured spectra reveal the existence of other minor peak located 10.8 GHz above the central frequency, which is generated as a result of four-wave mixing between the copropagating pump and the stokes peaks that arise in the fiber with the pumping [13,18]. In any case, its level is far beyond RS or BS in any experimental condition. Rayleigh and Brillouin scattering powers as a function of the input optical power are represented in Fig. 3. The total powers for each case are calculated by integration of the peaks in spectra similar to that shown in Fig. 2. Results show that the increase in the DFB output power only causes a small increase in RS power, while BS experiments a steep increase under the same injection conditions. Although the final optical Brillouin power level reached is fairly similar for all the broadenings, the onset point is different. As can be derived from the Figure, broadening of the optical source implies a shift in the injected power value at which BS power increase rate rises. The graphs in Fig. 3 clearly demonstrate that spectral broadening has a much stronger effect over BS than over RS total powers. In order to check the behavior of backscattering in an environment that resembles a remote-seeded access network, an RSOA (20-dB gain, +5-dBm output saturation power) has been placed in the final extreme of the fiber on the diagram of Fig. 1. The conditions tested were the same as in the set-up of Fig. 1. Measured spectra are shown in Fig. 4 for input powers of 7 dbm (upper graph) and 10 dbm (lower graph), both obtained using 60 MHz broadening and no broadening. The effects of source broadening are similar to those shown in Fig. 2. The total power of the Brillouin peak is considerably reduced and the mirrored peak practically disappears. The central peak is now the reflected and amplified downstream signal added to the RS and thus, significant information about the variations in RS cannot be extracted. The Brillouin peak is higher now despite the fact that it is not reflected, and in consequence not amplified, by the RSOA. The third peak is also enhanced in the presence of the amplifier because its height is dependent of that of the other two peaks. Thus, its power increases with injected power and decreases by spectral broadening. BS unexpected growth, compared with the measurements shown in Fig. 3, can be explained by considering two different contributions. First, the transference of power from the returned signal down to the Brillouin wavelength into the fiber Fig. 1. Experimental set-up for the measurement of backscattering non-linear effects in SMF.

3 J.J. Martínez et al. / Optics Communications 283 (2010) Fig. 2. Directly measured optical spectra of Brillouin ( 10.8 GHz) and Rayleigh (central) backscattering for two input powers and source spectral broadenings. Fig. 3. Detected total powers at optical spectrum analyzer using the set-up of Fig. 1 for different input power levels and frequency broadenings. which also explains the limiting upper bound in the power of the central peak. Second, the signal amplified in the RSOA and injected into the fiber, generates its own Brillouin peak that goes back into the amplifier, and thus is reflected and amplified to the DFB end of the fiber. The combination of both effects cause an increase of the BS power even when the injected signal is broadened. 3. Influence of backscattering over an upstream data transmission Once the scattering effects have been evaluated under several conditions, we next analyze how backscattered power affects a data transmission when reception is performed at the same point where the source injects power into the fiber. On a first experiment, the set-up proposed in Fig. 5 is used to test the performance at reception. The scheme is similar to the one used in previous experiments, but in this case, an intensity modulation is applied to the RSOA using as source a Bit Error Rate Tester (BERT). In addition, at the reception end the High Resolution OSA is replaced by a variable optical attenuator (VOA) to control the power delivered to the receiver ( 24-dBm 2.5 Gbps, BER = ) that performs direct detection on the upstream optical signal. Detected data is taken back to the BERT to obtain the transmission perfor-

4 2246 J.J. Martínez et al. / Optics Communications 283 (2010) Fig. 4. Directly measured optical spectra of backreflected signals (BS, RS and amplified downstream signal) when a RSOA is placed at the end of the SMF. Fig. 5. Experimental set-up for the measurement of Bit Error Rate (BER) performance of an upstream data transmission. mance of the link. In order to avoid high saturation operation of the RSOA and to maintain an nearly constant output optical power, a 5-dB attenuator is placed in front of it. For this study injected optical powers considered range from 3.8 to 11 dbm. Curves represented in Fig. 6 depict different received spectra when a pseudo random bit sequence (PRBS) at 1.25 Gbps is used to intensity modulate the optical signal at the RSOA. The injection conditions were the same used for the spectra in Figs. 3 and 4. The presence of the modulation can be clearly seen in the central peak and also a small copy of the same modulation can be seen over the Brillouin peak. The later fact confirms our previous assertion that a portion of this peak is generated in the RSOA side and thus, it is modulated and amplified. Other characteristics of RS and BS follow the same behavior observed in prior experiments. However, it is notable that the Brillouin power is lower than in the case where the RSOA does not introduce any modulation to the incoming signal. This fact can be explained by the presence of the fixed 5-dB attenuator in front of the RSOA that lowers the Brillouin peak., In addition, the intensity modulated reflected signal reduces the component of BS that is reflected back into the RSOA, amplified and sent back to the source. Both contribute to lessen the Brillouin peak generated in the RSOA end of the fiber. BER of upstream channel under different broadening conditions is presented in Fig. 7. As expected, the overall behavior shows that BER performance degrades with increasing input power. The case without DFB spectral broadening is shown in Fig. 7(a). As it can be observed, the curve for an input power of 3.8 dbm shows a good performance, but when the injected power rises above 5.7 dbm reception is completely lost, i.e., not even BERs of 10 3 can be measured at any received power. The cause for this link failure is not completely clear and it may be produced by either or both RS and BS. RS theory on bidirectional links predicts degradation but not the observed complete link failure, so a possible explanation for the link loss could be the interference caused by high BS powers over the signal, although it is far outside the communication effective bandwidth. On the other hand, Fig. 7(b) and (c) show BER curves when broadenings of 30 and 60 MHz are applied, respectively. As it can be seen, higher spectral broadenings enable link

5 J.J. Martínez et al. / Optics Communications 283 (2010) Fig. 6. Spectra of received 1.25 Gbps modulated signals under different injection conditions (same cases as in Fig. 2 and Fig. 4). operation at higher injected powers. For example, comparison of BER measurements in Fig. 7(b) and (c) shows that spectral broadening enables link operation up to an injected power of 8.1 dbm in the first case and up to 9 dbm in the second case. Performance results of the link are summarized in Fig. 8, where receiver sensitivity to obtain an error rate of 1E-9 versus input power is shown for different frequency broadenings. As the Figure shows, sensitivity degrades slowly with the increase in the injected power up to a point where the communication suddenly fails. The slight loss of quality is fairly similar in the three cases of broadening, while the abrupt fall clearly appears first in the cases with lower spectral width. Thus, we can conclude that sensitivity degradation is due to two different contributions. The first one is responsible for the slow increase observed in the three curves and the other one for the steep one. The first contribution is associated to RS as was stated previously. This conclusion is supported by the fact that, when the input power increases, the output from the RSOA is almost constant due to saturation while RS power is slowly increasing leading to a moderate increase in power penalty. h DP 10log 1 q r i ð1þ T This moderate increase that has been observed in our experiments here is well-known and has been described in the literature and fits the behaviour proposed in Eq. (1) [7 9], where q is a constant dependent of RS polarization, r is the RS to signal ratio and T is the inverse of signal data rate, high RS powers and high data rates increase the penalty. Another fact consistent with our argument is that the slow degradation is independent of the broadening and appears alike in the three sources, in the same way as the RS power increase. We argue that the cause of the abrupt communication loss is related to BS, as it is confirmed by the fact that when its power is reduced by source broadening the point where the link fails is shifted to higher input powers; also RS effect described in equation 1 does not explain this behaviour. To investigate the nature of the link failure we have performed two experiments aimed to separate RS and BS contributions. First, the Brillouin power at the detector has been measured for 1.25 Gbps upstream transmission using the set-up shown in Fig. 5, but changing the receiver by a HR-OSA. The results can be seen in Fig. 9 for different source broadenings. The curves obtained show a similar behavior than those in Fig. 6. The input power level above which communication fails is marked with an arrow for each case. In the Figure, arrows point to approximately the same Brillouin power level in all curves suggesting a common BS power threshold. When this threshold is surpassed BS and data signals may interfere even if the Brillouin peak is far from the effective data signal bandwidth. RS signal, according to our previous measurements does not present important changes. In the second experiment we have filtered out the Brillouin peak so that only the RS and upstream signals are detected at the receiver. For this experiment we have used a narrow upstream modulation of 625 Mbps for the link so a very narrow FP tunable band-pass filter (0.9-GHz FWHM) can be used to limit the influence of BS. Fig. 10 shows measured results of upstream performance at different injected powers, with and without filtering. The BER curves for filtered and unfiltered reception show around a 2 db difference until the unfiltered link fails, as the sensitivity curve of Fig. 11 clearly shows. Sensitivity curves show that the link fails for injected powers higher than 9 dbm when no filtering is performed prior to detection. With the filter, the BS power is attenuated by 18 db (filter insertion loss at 11 GHz), shifting the failure point of the link up to injected powers of 10.5 dbm. Therefore, we can assert that BS, alone or in its interference with RS or other peaks, causes an important degradation on the upstream channel even in 625 Mbps

6 2248 J.J. Martínez et al. / Optics Communications 283 (2010) Fig. 8. Sensitivity at 10 9 BER as a function of the optical power injected into the fiber, for different spectral broadenings. lies far from the signal bandwidth. In the next section we evaluate the case in which the BS peak enters the receiver bandwidth. 4. Wideband upstream transmissions Fig. 7. BER results for 1.25 Gbps upstream transmission using as a source a DFB signal under different conditions: (a) without spectral broadening, (b) with 30-MHz spectral broadening and (c) with 60 MHz spectral broadening. Nowadays WDM PONs are proposed at rates of 10 Gbps and beyond, so the backreflected BS peak, which has been demonstrated to degrade the quality of a bidirectional link, is very close to the receiver bandwidth. In order to test qualitatively the influence of backscattered optical power in these optical links, an experiment using the set-up shown in Fig. 12 was performed. Its basic design is the same than in previous schemes but it has important differences. In this case, the RSOA in the ONU has been replaced by an EDFA combined with a MZI Modulator and a circulator to maintain the reflective behavior. This change is made to allow the 10 Gbps data rate due to the lack of commercially available RSOAs operating at these high bandwidths. The upstream link modulation was changed into a spectrally wider 10 Gbps PRBS and the detector at the receiver end was substituted in order to work at high bit rates ( 18-dBm 10.5 Gbps, BER ). The spectra shown in Fig. 13 shows how the BS peak is very near the upstream bandwidth, and how source broadening can be used again to reduce Brillouin presence. BER performance curves in Fig. 14 show the behavior already observed at lower bit rates, with the link failure occurring at higher received powers for higher input optical power. An effect that doesn t appear in previous transmissions is a floor that is limiting the lower BER that could be achieved; this floor can be seen more clearly in the curves with low BS i.e., the ones with broadened spectrum at P in = 6.5 dbm and P in = 10.6 dbm. A control experiment was performed to discard the possibility that the floor effect found in our measurements at 10 Gbps was due to experimental conditions. In fact, BER measurements at 2.5 Gbps performed using the set-up depicted in Fig. 12 and using the same input settings did not show a floor effect. In consequence, and observing the behavior of BER curves for this case, the technique of CW wavelength feeding for 10 Gbps upstream channels may have critical consequences in terms of signal quality and has to be analyzed more in detail. links. Before the link loss in the unfiltered case, the reception degrades more slowly than in Fig. 8 due to the lower data rate of the signal, which limits the RS penalty (see Eq. (1)). The irregular behavior in the filtered signal can be explained by small variations in the accurate tuning of the filter. Up to this point we have shown the BS has an influence on data transmission even when its peak 5. Conclusion The powers of backscattering non-linear effects that arise in the fiber have been measured under several conditions in order to assess their behavior. First, backscattered powers have been quantified and their influence over a remote-seeded PON scheme has

7 J.J. Martínez et al. / Optics Communications 283 (2010) Fig. 9. Detected Brillouin powers at upstream receiver end in the set-up of Fig. 5, arrows point at the input optical powers where the link fails. Fig. 10. Upstream transmission (625 Mbps): BER at different injected powers, with or without filter in reception. been assessed. Although RS is well-known as a limiting factor, our experiments have shown that the usually neglected Brillouin Scattering can be an important source of impairments in data transmissions, even when it is placed out of their effective bandwidth, due to the high powers returned to the OLT. The specific effect caused by BS at reception is a complete failure of the link once BS power has surpassed a certain threshold. Some techniques for reducing the BS influence have been tested, including source spectral broadening and filtering at reception. Both of them have proved that it is possible to improve transmission performance at reception by lowering BS power. Finally, upstream transmissions at 10 Gbps have shown the presence of a floor effect that limits the potentially achievable BER. All these effects must be taken into account in the networks with remote-seeded ONUs, where the injected power can be very high. The effects and degradation associated to BS in a remote-seeded PON with a WDM scheme as not yet been analyzed and should definitely be assessed. In this case, BS generated by a particular wavelength may interfere with the rest of the channels and thus operation of this kind of networks can be dramatically impaired. Fig. 11. Sensitivity at 10 9 as a function of the optical power injected into the fiber, for filtered and no-filtered cases.

8 2250 J.J. Martínez et al. / Optics Communications 283 (2010) Fig. 12. Set-up used to perform BS filtered measured and 10 Gbps upstream transmission. Fig. 13. Optical spectra of 10 Gbps upstream transmission in reception with and without source broadening, BS peak can be seen 10.8 GHz below the central peak. References Fig. 14. Upstream transmission (10 Gbps): BER at different injected powers and source spectral broadenings. Acknowledgements Manuscript received December 18, This work was supported by the DGA by the funds provided for the i3a Research Laboratories in Walqa Tech. Park, by the Cátedra Telefónica and by Spanish MEC under project TEC C [1] R.D. Feldman, E.E. Harstead, S. Jiang, T.H. Wood, M. Zirngibl, J. Lightwave Technol. 16 (9) (1998) [2] N.J. Frigo, P.P. Iannone, P.D. Magill, T.E. Darcie, M.M. Downs, B.N. Desai, U. Koren, T.L. Koch, C. Dragone, H.M. Presby, G.E. Bodeep, IEEE Photonics Technol. Lett. 6 (11) (1994) [3] P. Healey, P. Townsend, C. Ford, L. Johnston, P. Townley, I. Lealman, L. Rivers, S. Perrin, R. Moore, Reflective SOAs for spectrally sliced WDM PONs, in: Proceedings of Optical Fiber Communications (OFC), Anaheim, CA, 2202, pp [4] J.J. Martínez, J.I. Garcés, A. López, A. Villafranca, J.C. Aguado, M.A. Losada, J. Lightwave Technol. 26 (3) (2008) 350. [5] J. Prat, V. Polo, C. Bock, C. Arellano, J.J. Vegas-Olmos, IEEE Photonics Technol. Lett. 17 (1) (2005) 702. [6] H. Takesue, T. Sugie, J. Lightwave Technol. 21 (11) (2003) [7] Robert G. Waarts, A.A. Friesem, Eyal Lichtman, Henry Howard Yaffe, Ralf-Peter Braun, Proc. IEEE 78(8) (1990) [8] Roland Karl Staubli, Peter Gysel, J. Lightwave Technol. 9(3) (1991) [9] M. Oskar van Deventer, IEEE Photonics Technol. Lett. 5 (7) (1993) [10] Santanu K. Das, Ed E. Harstead, J. Lightwave Technol. 20(2) (2002) [11] Peter Gysel, Roland K. Staubli, J. Lightwave Technol. 8 (4) (1990) [12] J.A. Lazaro, C. Arellano, V. Polo, J. Prat, IEEE Photonics Technol. Lett. 19 (2) (2007) [13] Govind P. Agrawal, Nonlinear Fiber Optics, Elsevier Science and Technology Books, [14] George C. Valley, IEEE J. Quantum Electron. qe-22 (5) (1986) [15] E. Lichtman, R.G. Waarts, A.A. Friesem, J. Lightwave Technol. 7 (1) (1989) [16] Yasuhiro Aoki, Kazuhito Tajima, Ikuo Mito, J. Lightwave Technol. 6 (5) (1988) [17] L. Bjerkan, A. Røyset, L. Hafskjær, D. Myhre, J. Lightwave Technol. 14 (5) (1996) [18] P. Labudde, P. Anliker, H.P. Weber, Opt. Commun. 32 (1980)