SOA-BASED NOISE SUPPRESSION IN SPECTRUM-SLICED PONs: IMPACT OF BIT-RATE AND SOA GAIN RECOVERY TIME

Transcription

1 SOA-BASED NOISE SUPPRESSION IN SPECTRUM-SLICED PONs: IMPACT OF BIT-RATE AND SOA GAIN RECOVERY TIME Francesco Vacondio, Walid Mathlouthi, Pascal Lemieux, Leslie Ann Rusch Centre d optique photonique et laser (COPL), Department of Electrical and Computer Engineering Pavillon d optique-photonique, Université Laval Québec, Canada, G1K 7P4 rusch@gel.ulaval.ca ABSTRACT Spectrum-sliced wavelength division multiplexing is becoming an attractive cost effective candidate for future passive optical networks. Since incoherent sources are used, excess intensity noise limits the achievable performances. Here, we propose to use semiconductor optical amplifiers (SOA) to mitigate this noise. We experimentally investigate the advantages and drawbacks of placing the SOA-based noise reduction at the optical line terminal (OLT) as well as at the optical network (ONU) unit (user end side). Two SOAs with different gain recovery times were used at different bit rates. Results show that if the SOA is placed at the ONU, fast SOAs are the most efficient for noise reduction. On the contrary, if the SOA is at the OLT, slow SOAs represent a better solution for noise cleaning, since they introduce less performance penalty. KEY WORDS Semiconductor Optical amplifiers (SOAs), Passive Optical Networks (PONs), Spectrum-Sliced Wavelength Division Multiplexing (SS-WDM). 1. Introduction Incoherent spectrum-sliced wavelength division multiplexed passive optical networks (SS-WDM PONs) are under intense scrutiny for future broadband access. They offer a low cost and robust alternative to conventional wavelength division multiplexing networks by eliminating the need for an array of separately stabilized lasers. However, due to its thermal like nature, incoherent light suffers from important excess noise. This noise drastically limits the achievable spectral efficiency and bit rates. As the bit rate increases, the detected optical noise becomes large, degrading network quality of service and limiting system performance. In order to accommodate higher bit rates, a wider optical bandwidth is required for SS-WDM, reducing the spectral efficiency and increasing the effects of dispersion [1]. Hence, the main challenge for SS-WDM PONs is to find efficient ways to suppress this noise. Some techniques were proposed to achieve this purpose [2-3], however they increase the network complexity by adding a high speed electronic processing at the receiver [2] or by adding several electrical component at both transmitter and receiver [3]. Several recent works proposed semiconductor optical amplifier based noise suppression schemes [4-8]. SOAs are particularly attractive, as a single SOA at the remote node in SS-WDM PONs can serve as both a preamplifier to amplify and suppress noise in the downlink, and as an external modulator of the distributed source in the uplink [9]. The SOA is then the central network component, hence the choice of the right SOA becomes of major importance. We experimentally investigate two configurations for noise suppression: booster at the optical line terminal (OLT), which exploits gain squeezing, and incoherent-tocoherent conversion at the optical network unit that exploits cross-modulation (XGM). The booster configuration at the OLT places the SOA before signal attenuation due to propagation, where it is easier to saturate the SOA, but less effective gain is available from the SOA. The conversion configuration has the SOA at the ONU following propagation and rendering it more difficult to saturate the SOA, but having more available gain. For the conversion configuration, a continuous local laser is required; however an inexpensive CW probe without stabilization is sufficient. In this configuration, the received incoherent slice modulates the SOA gain XGM hence modulating the CW coherent source. In each configuration we use two different SOAs (SOA1 and SOA2) with different gain recovery times in order to assess the impact of both SOA recovery time and transmission speed on noise reduction. SOA1 and SOA2 have respectively fast and slow gain recovery times. We will investigate the impact of gain recovery time and bit-rate on the noise suppression and performance of a single channel SS-WDM link. We will discuss the benefits and the drawbacks of the two schemes, and we will investigate the required SOA design giving maximal noise reduction in the network for each approach. The rest of the paper is organized as follows: in section 2 we detail our experimental setup, along with the

2 OLT ONU Figure 1 SOA-based noise reduction experimental setup with SOA in: (a) booster configuration at OLT, (b) conversion configuration at ONU. BPF: band-pass filter, PC: polarization controller, MZ: Mach-Zender modulator, SOA: semiconductor optical amplifier, CW: continuous-wave emitting laser diode. 1 Normalized Gain Recovery SOA1 SOA Time [ps] Figure 2 (a) Pump-probe experimental setup used to measure SOAs gain recovery dynamics. MLL: mode-lock laser, ODL: optical delay line, VOA: variable optical attenuator, PD: photodiode. (b) Normalized gain as a function of time. description of some important characteristics of the two SOAs we use in the experiment. Section 3 presents and discusses bit error rate () experimental results using the SOA-based noise reduction for the two configurations and for different bit-rates. Finally in section 4 we draw some conclusions. 2. Noise-reduction experiment and SOA characteristics Figure 1(a) and Figure 1(b) depict the experimental setup we used to investigate the performance of the two noise reduction schemes considered in this work. The broadband source is composed of a C-band superfluorescent fiber source followed by a wideband optical filter, an erbium-doped fiber amplifier (EDFA), and a polarization beam splitter (PBS). This gives us a powerful broadband polarized amplified spontaneous emission (ASE) source (around 13 dbm of total output power at the TM port of the PBS). We then spectrally slice the source using optical band-pass filter BPF1 with 0.24 nm (30 GHz) 3-dB optical bandwidth centered at nm. The total power at the output of BPF1 was around 0 dbm, enough to heavily saturate either SOA. The Mach-Zehnder modulator is modulated with a NRZ PRBS sequence at four bit rates (,, 2.5 Gb/s and ) in order to investigate the SOA modulation dynamic on the noise suppression. Several sequence lengths were examined and no significant patterning effect was observed. We thus fixed the sequence length to In each noise reduction scheme, the SOA input power is varied using a variable optical attenuator (VOA) to examine several saturation levels. The input polarization is optimized using a polarization 475

3 Figure 3 Normalized modulation response of ASE as a function of exciting frequency for SOA1 (triangles) and SOA2 (circles). Inset time shows the emitted ASE power as a function of input power for the two SOAs. controller (PC). The detector is a lightwave receiver (Agilent 11982A) followed by the appropriate electrical filter (467 MHz, 933 MHz, 1.87GHz and 3.5 GHz respectively). Since our objective is to understand the relation between noise suppression and SOA recovery time, we used two different commercially available SOAs for each configuration: 1) CIP SOA-NL-OEC-1550 (SOA1), and 2) a linear Covega BOA-2679 (SOA2). The injection currents giving maximal gain are 300 and 500 ma for SOA1 and SOA2 respectively. In all experiments the injection currents were kept fixed at these levels. Both SOAs are InP/InGaAsP multi-quantum well layer structures with a ridge waveguide design. Figure 2 shows the SOAs measured gain recovery times for the two SOAs. Figure 2(a) depicts the pump-probe setup used to characterize SOAs speed [10]. We use a single femtosecond-pulse mode-locked laser (MLL) to generate two counter-propagating pulses that first deplete carriers and thus the gain of the SOA (pump signal) and then probe the gain after a certain time delay (probe signal). The gain is then resolved in time using an optical delay line (ODL) which enables us to vary the delay between the pump and the probe pulses in the SOA. We measured respectively 30 and 650 ps for SOA1 and SOA2 as shown in Figure 2(b). Another important parameter for the noise cleaning performance is the amount and the dynamics of ASE detected noise [11]. Figure 3 shows the ASE normalized response to a modulation signal as a function of the exciting signal frequency. In this case we inject a sinusoidally modulated laser into the SOA and we sweep its frequency while keeping the average input fix to -10 dbm. Then we measure the modulation induced on the detected ASE by XGM. We can see in this figure that Figure 4 results at for SOA1 (triangles) and SOA2 (circles) when SOA is at OTL (booster configuration, solid line curves) and when SOA is at the ONU (conversion configuration, dashed lines curves). The dash-dotted line curve represents the of the link without SOA. SOA1 (the fastest) has the largest ASE modulation bandwidth whilst SOA2 has the smallest bandwidth. Moreover, we can see in the inset of Figure 3 that SOA1 generates more ASE power than SOA2. The signal to noise ratio (SNR) in SS-WDM PONs is proportional to the ratio of the receiver optical bandwidth over the electrical bandwidth B o /B e [12]. Hence, the optical filter BPF1 determines link performance as well as spectral efficiency of the SS-WDM system. BPF2 is matched to BPF1 and selects the WDM channel for reception. We will, however, also experiment with removing the BPF2 for the purpose of bringing out the effect of ASE on the noise cleaning performance of the booster configuration. Noise suppression in the conversion configuration also uses BPF2 select the WDM channel; note that BPF3 ensures that only the newly converted coherent signal is incident on the detector. Conversion efficiency varies with probe power and wavelength. Values of 15 dbm and 1540 nm for SOA1, 9 dbm and 1540 nm for SOA2 were found to be optimal and are used for all conversion noise-suppression experiments. 3. Results and Discussion Note that our experiments were performed back-to-back, and the results are plotted against SOA input power P in. For the booster configuration P in represents power before fiber attenuation, while P in for the conversion configuration represents power received after fiber attenuation. For single mode fiber with 0.25 db/km of attenuation and spans of 20 km, 5 db of loss is incurred; our plots should be interpreted with this in mind. 476

4 10-2 Conversion configuration at ONU SOA2 (slow) 10-2 Booster configuration at OLT SOA2 (slow) P [dbm] in (a) P in [dbm] 10-2 Conversion configuration at ONU SOA1 (fast) 10-2 Booster configuration at OLT SOA1 (fast) filtered filtered filtered P in [dbm] P in [dbm] (b) Figure 5 Bit error rate measurements as a function of the input SOA power for the SOA at the ONU in the conversion configuration at different data bit rates. (a) is for SOA2, (b) is for SOA1; dashed line curves are without BPF2; plain line curves are with BPF2. Figure 4 depicts the results for transmission for BPF1 having 0.24 nm bandwidth (3 db). The dash-dotted curve represents with no SOA present. Results for the booster configuration at the OLT are presented in solid lines, while conversion at the ONU results are presented with broken lines. SOA1 (with faster recovery time) is portrayed with triangular markers and slower SOA2 with circular markers. We see that using a SOA in either configuration leads to improved performance. For the booster configuration we observe a slight power penalty ( 1 db) for the slow SOA2 compared to the faster SOA1. Increasing SOA input power leads to a floor for SOA1 placed at OLT, and performance approaches that when using SOA2. We attribute this behavior to the impact of ASE on the extinction ratio (ER). The ASE is modulated inversely with the data through XGM, hence degrading the extinction ratio. Since SOA1 is both faster and generates Figure 6 Bit error rate measurements as a function of the input SOA power for the SOA at the OLT in the booster configuration at different data bit rates. (a) is for SOA2, (b) is for SOA1; dashed line curves are without BPF2; plain line curves are with BPF2. more ASE power, its ASE power will reduce the ER and cause much greater penalty than SOA2 (slow and negligible ASE power). In the conversion configuration, we can see that using slow SOA2 introduces a 5 db power penalty compared to fast SOA1. This is due to the fact that XGM is more efficient for fast SOAs. We do not observe any floor when using the conversion configuration. Recall the 5 db discrepancy in the definition of P in for these two configurations; both configurations give similar noise suppression at and 0.24 nm optical bandwidth. Figure 5 and Figure 6 show results for bit rates ranging from to. Figure 5(a) and (b) show the results for conversion, while the booster configuration is presented in Figure 6(a) and (b). In both cases figures (a) refer to the results when using the slower SOA, which is SOA2, while figures (b) present the results when using the faster SOA, which is SOA1. 477

5 In the conversion configuration (Figure 5) we observe increasing power penalty as the bit rate increases, showing that XGM is less efficient for high bit rates. Good results are obtained for bit rates lower or equal to without any floors. Comparing Figure 5(a) and (b), the fast SOA1 is more effective than SOA2. Incoherent to coherent conversion is more efficient for fast SOAs, which leads to a better for SOA1. Modulation at with 0.24 nm optical bandwidth is too challenging for any of the suppression methods. In these experiments we wish to demonstrate that the ASE modulation deteriorates the ER and leads to degradation. To this end we also measure in the booster configuration with no channel-selection filter at the receiver (no BPF2); these curves are dashed in Figure 6. When BPF2 is removed all the ASE power is detected. In Figure 6(a) we do not observe any significant penalty for SOA2 in this case. The SOA2 recovery time is so slow that the ASE is not modulated by the data at any of the data rates considered, and no degradation is noted. In Figure 6(b) we see a marked degradation in performance for the dashed curves, i.e., when BPF2 is not used. When the faster SOA1 is used, more ASE power is generated and, more importantly, the ASE is modulated inversely to the data, degrading the ER and closing the eye. The amount of penalty depends on the bit rate, with greatest penalty (7 db) at. The faster SOA1 maybe more effective than slower SOA2 at, but at (an important target for WDM PONs), SOA1 mitigation is crippled by the ASE modulation. 4. Conclusion In this paper, we investigate the impact of SOA characteristics and data modulation speed on the noise reduction capabilities of SOA in SS-WDM PONs. In particular we focus our attention on two possible promising configurations: in the first of them the SOA is used as a booster at the OLT, in the second it is used to perform an incoherent-to-coherent conversion at the ONU. We characterized and used two completely different SOA in order to investigate the impact of the ASE noise dynamics and of the gain recovery time on the noise cleaning process. In the booster configuration at the OLT slow SOAs are more effective in cleaning the noise with respect to fast ones. Even if fast SOAs would introduce good noise cleaning, the ASE noise would degrade the extinction ratio due to its counter-modulation with the signal via XGM, as we show by removing the filter BPF2. The advantage of placing the SOA at the OLT is the possibility of using it as a booster at the edge of the link; the disadvantage of such a scheme is that for bit-rates as high as, floors are visible. On the other hand, in the conversion configuration at the ONU, fast SOAs are more effective for noise cleaning: for bit rates as high as 2.5Gb/s, we were able to get rid of floors. The conversion configuration is also spectrally efficient, since we used a receiver optical filter with the same bandwidth as the transmitter filter (0.24nm). The disadvantage of the conversion-based noise cleaning scheme is the need to saturate the SOA at the receiver so that XGM is efficient. Acknowledgement This work was supported by PADCO. References [1] J. S. Lee, C. H. Chung & D. J. Digiovanni, Spectrum-sliced fiber amplifier light source for multichannel WDM application, IEEE Photon. Technol. Lett., vol. 5, 1993, [2] A. J. Keating, W. T. Holloway & D. D. Sampson, Feedforward noise reduction of incoherent light for spectrum-sliced transmission at, IEEE Photon. Technol. Lett., vol. 7, 1995, [3] J. H. Han, J. W. Ko, J. S. Lee & S. Y. Shin, 0.1-nm narrow bandwidth transmission of a 2.5-Gb/s spectrumsliced incoherent light channel using an all optical bandwidth expansion technique at the receiver, IEE Electron. Lett, 10, 1998, [4] S. Kim, J. Han, J. Lee & C. Park, Intensity noise suppression in spectrum-sliced incoherent light communications systems using a gain-saturated semiconductor optical amplifier, IEEE Photon. Technol. Lett., vol. 11, no. 8, Aug. 1999, [5] A. D. McCoy, P. Horak, B. C. Thomsen, M. Ibsen & D. J. Richardson, Noise suppression of incoherent light using a gain-saturated SOA: implications for spectrumsliced WDM systems, IEEE J. Lightwave Technol., vol. 23, 2005, [6] T. Yamatoya & F. Koyama, Optical preamplifier using optical modulation of amplified spontaneous emission in saturated semiconductor optical amplifier, IEEE J. Lightwave Technol., vol. 22, 2004, [7] M. Menif, W. Mathlouthi, P. Lemieux, L. A. Rusch, & M. Roy, Error-free Transmission for Incoherent Broadband Optical Communications Systems using Incoherent-to-Coherent Wavelength Conversion, J. Lightwave Technol., vol. 23, 2005, [8] W. Mathlouthi, P. Lemieux & L. A. Rusch, Optimal SOA-based Noise Reduction Schemes for Incoherent Spectrum-Sliced PONs., Proc. IEEE 2006 European Conference on Optical Communication (ECOC2006), Cannes, France, September

6 [9] F. Payoux, P. Chanclou, M. Moignard & R. Brenot Gigabit optical access using WDM PON based on spectrum slicing and reflective SOA, IEEE 2006 European Conference on Optical Communication (ECOC2006), vol. 3, Sept. 2005, [10] M. Eiselt, W. Pieper & H. G. Weber, SLALOM: Semiconductor laser amplifier in a loop mirror, J. Lightwave Technol., vol. 13, 1995, [11] T. Yamatoya & F. Koyama, Optical preamplifier using optical modulation of amplified spontaneous emission in saturated semiconductor optical amplifier, J. Lightwave Technol., vol. 22, 2004, [12] G.J Pendock & D.D Sampson, Transmission performance of high bit rate spectrum-sliced WDM systems, J. Lightwave Technol., vol. 14 (10), 1996,