We demonstrate experimentally and numerically that the SBS based in-band OSNR monitoring technique can be used for dual polarization signals. We also present a novel approach for a drastic enhancement of the sensitivity monitoring range by intentionally adding in-band ASE noise into the signal. Numerical results are provided for 44.6 Gbps DPSK, 44.6 Gbps DQPSK and 112 Gbps Dual Polarization (DP-) QPSK signals, with both 100 GHz and 50 GHz channel spacing scenarios.
© 2010 OSA
The fast growing demands for high speed applications such as triple play, High Definition TV and business services require dramatic changes in optical network concepts and paradigms. The migration from a static point to point network to a fully meshed network forces the system vendors to provide cost effective optical networks with highly manageable capabilities via the deployment of transparent and reconfigurable optical networks. In addition, operators and service providers require new services with high resiliency to hardware failures, flexible resource utilization and offering guaranteed quality of service in term of signal quality or available bandwidth. Deployment of high speed transparent and reconfigurable optical networks requires effective flexible and robust Optical Performance Monitoring (OPM) techniques  in order to ensure high quality of service and high level of resiliency.
The modern high speed networks are susceptible of optical signal degradations, mainly due to the Amplified Spontaneous Noise (ASE) from the optical amplifiers. Real time monitoring of the Optical Signal to Noise Ratio (OSNR) is a mandatory in order to ensure the signal quality and in order to monitor potential failures in the transmission link. The most common method to monitor the OSNR is based on the spectral analysis of the transmitted WDM signals and derives the OSNR by interpolating the out of band noise level into the signal band, namely by estimating the in-band noise level using the out of band noise level . However such a technique suffers from the use of optical filtering and routing in the link path since the out of band noise is filtered out. Therefore the interpolating method leads to severe overestimates of the real OSNR level.
Methods to derive OSNR level by estimating the in-band noise level directly, even in the presence of optical filters in the link, are referred as in-band OSNR methods. Several in-band OSNR methods have been proposed and are based on various approaches such as electrical carrier to noise monitoring, polarization nulling , polarization diversity  optical delay interferometer , nonlinear transfer functions using an optical parametric amplifier , or a nonlinear loop mirror , delay tap asynchronous sampling techniques [7,8] and pairs of Michelson fiber interferometers with different delays . Some of these methods are sensitive to other system impairments such as Chromatic Dispersion (CD) and Polarization Mode Dispersion (PMD) and this makes the OSNR monitoring more challenging.
A method of in-band OSNR monitoring technique based on Stimulated Brillouin Scattering (SBS) effect, as described in , has an advantage in that it is insensible to CD and PMD. It can be also used to monitor WDM signals simultaneously . The SBS effect  is a spectral nonlinear effect which leads to the nonlinear power transfer from the signal spectral component to a Stoke wave (down shifted by ~10 GHz with respect to the signal frequency) propagating in the backward direction with respect to the signal. The OSNR technique based on the SBS effect uses the fact that when a signal has its higher spectral components above the SBS threshold, the efficiency of the power being transferred to the Stoke wave is altered by the noise present within the signal band. In both [10,11], SBS based OSNR technique is demonstrated for 40 Gbps OOK (On-Off Keying) signal. Iredale et al.  demonstrate results for 40 Gbps NRZ (Non return to Zero) OOK signal, which show a dynamic OSNR monitoring range of 15 dB for OSNR levels varying from 15 to 30 dB. Quite high sensitivity (15dB) is demonstrated due to the fact that the 40 Gbps NRZ OOK signal spectrum presents a prominent spectral peak at the carrier wavelength which is sufficiently narrow to stimulate an efficient SBS effect. The efficiency is also enhanced by the fact that a broadband bandpass filter (1nm bandwidth) is used in the experiment. However, such a filter bandwidth is not compliant with operations with 100GHz and 50 GHz channel spacing. Moreover, the NRZ OOK modulation format presents severe system penalties for bit rates of 40 Gbps and beyond; phase modulation formats are preferred and are optionally combined with polarization multiplexing scheme for additional CD and PMD impairment relaxations.
Furthermore, optical networks are pushed towards channel operations at 40 and 100 Gbps using modulation formats such as DPSK (Differential phase shift keying), DQPSK (Differential Quaternary Phase Shift Keying) and DP-QPSK (Dual Polarization Quaternary Phase Shift Keying). With such high speed networks, OSNR requirements become stronger and the network links should be planned to meet OSNR of 15 dB and higher at the link end. This makes the in-band noise level not high enough to cause a significant change in the back reflected power, limiting drastically the sensitivity range of SBS based OSNR monitoring, especially at 50 GHz channel spacing. Additionally, an important challenge is raised for OSNR monitoring of dual polarization modulation format such as 112 Gbps DP-QPSK. Indeed, the common methods of in-band OSNR monitoring, based on the principle of polarization nulling  or polarization diversity  use the fact that the signal is fully polarized while the optical noise is unpolarized. However such an assumption cannot be used for dual polarization signals which are fully unpolarized. In addition, ASE noise can also be partially polarized due Polarization Dependent Loss (PDL) in different network elements. Therefore, it is necessary to develop in-band OSNR monitoring techniques compliant with dual polarization modulation formats and independent of the polarization nature of the ASE noise.
In this paper, we demonstrate experimentally and numerically that the SBS based in-band OSNR monitoring technique can be used for dual polarization signals, using 80 Gbps DP-NRZ OOK signal. We also propose a new concept for SBS based in-band OSNR monitoring with enhanced sensitivity, which is compliant with the OSNR requirements for 44.6 Gbps DPSK, 44.6 Gbps DQPSK and 112 Gbps DP-QPSK signals in optical networks with 100 GHz and 50 GHz channel spacing.
2. In-band OSNR monitoring for dual polarization signals
Figure 1 shows the experimental set up used for the SBS based in band OSNR monitoring. The source signal is a 40 Gbps NRZ OOK signal. The 80 Gbps DP-NRZ OOK signal comprises two 40 Gbps NRZ OOK signal tributaries at the same wavelength but with orthogonal polarization. It is generated by splitting the source signal into two equal power signals whose states of polarization are orthogonalized by using polarization controllers and a polarization beam combiner (PBC). An optical delay of 16 bits is introduced to decorrelate the two signal tributaries. The OSNR is controlled by adding to the signal, amplified spontaneous emission (ASE) noise from a noise source. The ASE power level is adjusted with a variable optical attenuator (VOA). The signal combined with the added noise is then filtered using a 1 nm optical band pass filter (OBPF), amplified in an Erbium Doped Fiber Amplifier (EDFA) and launched with a given fixed optical power into 1 km of highly nonlinear fiber (HNLF) to stimulate the SBS process. The HNLF presents a nonlinearity coefficient of 11.7 W−1km−1 with an effective area of 11.4 μm2. The SBS back reflected power is measured using a photo-detector (PD) at the reflective port of an optical circulator. When propagating into the HNLF, the two polarization tributaries of the 80 Gbps DP-NRZ signal generate two independent and uncoupled SBS induced back reflected waves since the polarization orthogonality is conserved by SBS effect . As explained in , the back reflected power is only function of the signal power and decreases with the increase the ASE noise power (which for a fixed launched composite power leads to a decrease of the signal power). Therefore the polarization nature of the ASE noise (unpolarized, partially or completely polarized) does not affect the in-band OSNR monitoring technique.
Figure 2a and b show the dependence of the back reflected power to the OSNR level for both 40 Gbps NRZ OOK and 80 Gbps DP-NRZ OOK signals with three different launched power levels into the HLNF (15, 17 and 20 dBm). The graphs obtained for both signals are similar and numerical simulations (curves with empty symbols) confirm the experimental results (curves with plain symbols). Numerical simulations were performed using the split step Fourier Transform method including the SBS effect .
The back reflected power efficiency increases with the launched power. For a given launched power, the back reflected power increases with the OSNR level. The back reflected power slope is strong for OSNR levels smaller than 20dB and reduced for higher OSNR levels.
3. SBS based in band OSNR monitoring limitation in real deployed networks
Since the phase modulated modulation formats do not present a peak component at the signal carrier, their SBS threshold level is higher by more than 10dB in comparison to the OOK modulation formats. Therefore, in order to generate the SBS effect, the signal launched power level and/or the HNLF length much be increased.
Figure 3 shows numerical results of the back reflected power as a function of the launched power of 44.6 Gbps DPSK signal with OSNR of 35 dB. The back reflected power increases with the launched power. At a first stage, the increase is linear due to Rayleigh backscattering and Spontaneous Brillouin Scattering (SpBS) effects. For higher launched power the back reflected power increases exponentially due to SBS effect. When using a 1 km long HNLF, launched power levels higher than 23 dBm are required to stimulate SBS effect. Experimental measurements with 1 km of HNLF confirm that up to 22dBm, only Rayleigh back Scattering (RbS) and SpBS occur. When increasing the HNLF length, the SBS threshold decreases. With a 3 km long HNLF and 23 dBm launched power, SBS effect leads to an enhancement of the back reflected power by 16dB in comparison to the linear power increase due to RbS and SpBS.
In optical networks operating at 40 or 100 Gbps with 100 or 50 GHz channel spacing, the amount of in band noise is not enough to significantly change the SBS back reflected power because of the limited filter bandwidth and the higher OSNR requirements (usually higher than 15dB). This leads to a poor OSNR monitoring sensitivity range for the SBS based in band OSNR monitoring technique. Due to physical limitations in the laboratory (1 km long HNLF and an EDFA with output power of 23dBm), we performed to numerical simulations for phase demodulated signals. Figure 4 exhibits numerical results of back reflected power dependence to the OSNR levels for 44.6 Gbps DPSK, 44.6 Gbps DPSK and 112 Gbps DP-QPSK using a 50 GHz gaussian filter (which is compliant with operation for 100 GHz channel spacing) and a 3 km long HNLF. Since the SBS process is dependent on the signal optical spectrum, the SBS induced back reflected power is therefore dependent on the signal modulation format .
In the case of 44.6 Gbps DPSK signal with 20dBm launched power, for an OSNR range varying from 15 to 20 dB, the back reflected power varies by 2dB while for an OSNR range varying from 20 dB to 35 dB the back reflected power changes by 1dB. For 44.6 Gbps DQPSK signal with 20dBm launched power, the back reflected power varies by 1.8dB for OSNR range from 15 to 20 dB and by 1dB for an OSNR range of 20 to 35dB while in the case of 112 Gbps DP-QPSK with 22 dBm launched power, the back reflected power variations are around 1dB and 1.2 dB, respectively.
Such variations of the back reflected power level are not enough for a correct estimation of the OSNR for levels higher than 15 dB. However for OSNR levels lower than 15 dB, the variation of the back reflected power are stronger due to a larger amount of ASE noise which reduces the SBS induced backreflection efficiency more significantly. However, such high bit rate signals cannot be operated with OSNR levels lower than 15dB in the network due to performance constraints.
4. SBS based in band OSNR monitoring technique with enhanced sensitivity
4.1 Principle of operation
The sensitivity monitoring range of the in-band OSNR monitor can be improved by intentionally adding a known amount of Amplified Spontaneous Emission (ASE) noise in the signal band but only for the monitoring purposes. This concept of enhanced in-band OSNR monitoring is illustrated in Fig. 5 , where a given amount of additional in-band ASE noise power (noted Padd) from a noisy source is combined with the tapered WDM signal power whose composite power is Pc=Ps+Pn, where Ps is the signal power and Pn the accumulated ASE noise power resulting from optical amplification in the network.
The Pc and Padd levels are controlled using VOAs to provide a specific Optical signal Power Composite to Noise Ratio (OPcNR), defined as:
By altering OSNRreal with an additional noise power Padd, the SBS back reflected power (measured by PD2) will provide a measurement of the altered OSNR defined as:
The combined signal is then amplified by an Erbium Doped Fiber Amplifier (EDFA) and launched with a fixed known power into a highly nonlinear fiber to enhance the SBS effect. The induced back-reflected power from the HNLF is measured using the optical circulator and a photodiode PD2 and provides an estimation of OSNRalt.
Knowing OPcNR and OSNRalt, OSNRreal can be derived using:
Figure 6 shows the relationship between OSNRreal, and OSNRalt of the degraded signal for different levels of OPcNR with NEB=80GHz. For a reference value OPcNR=50dB, the added external noise is extremely negligible and therefore the measured OSNRalt is identical to OSNRreal. This reference case is therefore identical to the situation d described in , where no additional external noise is combined to the signal extracted from the network. When the amount of combined noise is increased by adding the external ASE noise, the OPcNR level decreases. In this case, OSNRalt<OSNRalt,MAX<OSNRreal where OSNRalt,MAX is the upper limit of OSNRalt, and it is equal to Ps/Padd(Bref/NEB) for high OSNRreal. By reducing the OPcNR level, the operative range of the OSNRalt is shifted toward lower values.
Figure 7 shows the principle of OSNR sensitivity range improvement provided by this novel concept. Figure 7(a) shows the dependence between the back reflected power and the level of OSNRalt. For high levels of OSNRalt, the back reflected power is high but exhibits a low dynamic sensitivity range while at lower OSNRalt levels, the back reflected power exhibits a reduced efficiency but its slope is steeper, providing a large dynamic sensitivity range. Figure 7(b) shows the relationship between OSNRreal and OSNRalt for the case where OPcNR=50 dB (no additional ASE noise is combined to the signal ) and for the case where OPcNR=4dB. For high bit rate optical signals, the OSNR (OSNRreal) range of the signals in the network is usually required to be between 15 and 25dB (see the dashed region). For OPcNR=50 dB, the OSNRalt is equal to OSNRreal since no additional noise is introduced to the signal at the monitoring stage. In the upper graph, for the range of OSNRreal between 15 dB and 30 dB, the change in the back reflected power is only of 1.6 dB. In contrast when adding noise intentionally to the composite signal, the OSNRalt is reduced while enhancing the dynamic range of the back reflected power. For example, when OPcNR is set to 4 dB, the OSNRalt range is shifted to be between 9.8 dB and 11.7 dB and for such OSNRalt range the dynamic range of back reflected power is increased to 3.5 dB.
4.2 Simulation of in-band OSNR monitoring for high bit rate modulation formats
The optical filter used in the simulation is 3rd order flat top Gaussian filter whose 3dB bandwidth is 80 GHz and 45 GHz for 100 GHz and 50 GHz channel spacing respectively.
Figure 8(a) and (b) show the numerical simulation results of the back reflected power variations with the OSNRreal and several OPcNR levels for 44.6 Gbps DPSK in 100 GHz and 50 GHz channel spacing respectively. When no additional ASE noise is combined to the signal (OPcNR=50dB), the back reflected power changes by 1.9 dB only for OSNR varying from 15 dB to 35 dB. The additional curves show the dependence of the back reflected power to OSNRreal for different OPcNR levels. Decreasing the OPcNR leads to a decrease of OSNRalt and to a reduction of the back reflected power efficiency while the slope of the back reflected power increases versus OSNRreal. This enhances the OSNR monitoring sensitivity. At 100 GHz channel spacing, for OPcNR of 6 dB, the back reflected power changes by 6.9 dB for OSNR levels varying from 15 dB to 35 dB. This shows the enhancement of the OSNR monitoring sensitivity by 5 dB. For WDM networks operating with the 50 GHz channel spacing, because of the reduced filter bandwidth, the amount of in band noise power is much lower than in the case of 100 GHz channel grid. Therefore the variations in the efficiency of the SBS induced reflected power is reduced compared to the case of 100 GHz channel spacing. For OPcNR=50dB, when the OSNR varies from 15 dB to 35 dB, the back reflected power chenges by 0.9 dB only. Here again, addition of the local ASE noise leads to an increase of the OSNR monitoring efficiency and for OPcNR of 3dB, the back reflected power varies by 3.6 dB. This shows an enhancement of the OSNR monitoring sensitivity by 2.7 dB.
Figures 8(c) and (d) show how the dynamic sensitivity monitoring range of the OSNR monitor varies as a function of the OPcNR, and depending on the OSNR range of the real optical signal. Sensitivity monitoring ranges are provided for OSNRreal range of 15- 35dB and for the OSNR sub-ranges of 15-20 dB, 20-25 dB and 25-35 dB in the case of 44.6 Gbps DPSK for both 50 and 100 GHz channel spacing cases. In both cases, the decrease of the OPcNR leads to an enhancement of the monitoring sensitivity for all sub ranges of OSNRreal. The sub range of 15-20dB (corresponding to the typical planned OSNR range in the optical network) exhibits the best OSNR sensitivity monitoring range of 3.7 dB and 1.6 dB for 100 and 50 GHz channel spacing respectively.
Similarly, Fig. 9 illustrates compliance and performances of the proposed monitoring technique with 44.6 Gbps DQPSK modulation format for both 100 and 50 GHz channel spacing. More specifically, Figs. 9(a) and (b) show numerical results of the back reflected power as a function of the OSNRreal for several OPcNR levels, in the case of 44.6 Gbps DQPSK signal. For OPcNR of 50dB, the reflected power varies by 5.8 dB and 1 dB for 100 and 50 GHz channel spacing respectively (for the OSNRreal range from 15 to 35 dB). When decreasing the OPcNR level, the variations in the reflected power become stronger. As shown in Fig. 9(c) and (d), the optimum OSNR sensitivity range is 10.2 dB and is obtained for OPcNR=10dB in the case of 100 GHz channel spacing. For networks operating with 50 GHz channel spacing, the OSNR sensitivity range is enhanced up to a level of 3.3 dB, for OPcNR=3dB.
Figure 10 show the compliance and performances of the enhanced approach for in band OSNR monitoring with 112 Gbps DP-QPSK signals for both 100 and 50 GHz channel spacing.
Similarly, Figs. 10(a) and (b) show numerical calculations of the back reflected power as a function of OSNRreal for several OPcNR levels, in the case of 112 Gbps DP-QPSK signal. DP-QPSK modulation format comprises two signals QPSK tributaries at the same wavelength but with orthogonal state of polarization polarization. Each tributary generates a back reflected Stokes wave independently. For OPcNR of 50 dB, the reflected power varies by 4.5 dB and 0.8 dB for 100GHz and 50 GHz channel spacing respectively (for OSNRreal range from 15 to 35 dB). When decreasing the OPcNR level, the variations in the reflected power become stronger. As shown in Fig. 10(c), the optimum OSNR sensitivity range is 7.5 dB and is obtained for OPcNR=10 dB in the case of 100 GHz channel spacing. For networks operating with 50 GHz channel spacing, the OSNR sensitivity range is enhanced to a level of 3 dB for OPcNR=1 dB, as shown in Fig. 10(d).
It is to note that for the three different modulation formats and for 100 GHz channel spacing, the in-band OSNR sensitivity range is reduced after a given OPcNR level (6 dB for 44.6 Gbps DPSK and 10 dB for 44.6 DQPSK and 112 Gbps DP-QPSK). This indicates that the amount of added noise becomes significantly important and deteriorates strongly the SBS back reflected power slope efficiency. For the case of 50 GHz channel spacing, this situation is not observed since the OPcNR level is defined for a NEB of 45 GHz whereas for 100 GHz channel spacing it is defined for NEB of 80 GHz. Therefore for the same OPcNR level the amount of added noise is smaller in the case of the 50 GHz channel spacing where much lower OPcNR levels are required to leads to a reduction of the slope efficiency.
The monitoring accuracy of OSNRreal depends on the monitoring accuracy of the back reflected power, the OPcNR level and the OSNRreal level. If optical power monitoring has an accuracy of +/−ΔP, this will lead to an in band OSNR accuracy of +/−ΔOSNR. Assuming an optical power accuracy of +/− 0.1 dB, Fig. 11 shows the OSNR monitoring accuracy ΔOSNR as function of the OSNRreal for 44.6 Gbps DPDK, 44.6 Gbps DQPSK and 112 Gbps DP-QPSK signals in both 100 and 50 GHz channel spacing scenarios. For each modulation format and channel spacing case, the optimum OPcNR has been used as discussed previously. In all cases, it can be seen that the OSNR accuracy is better in the 100 GHz scenario than in the 50 GHz scenario and it is deteriorated with the increase of OSNRreal. since the slope of the back reflected power decreases. An OSNR accuracy better than +/− 1dB is achieved for all modulation formats with OSNRreal <25dB and OSNRreal<32 dB in the 50 GHz and 100 GHz channel spacing scenarios respectively.
We have proposed a novel and improved approach for in-band OSNR monitoring based on the SBS effect in fiber. We have demonstrated experimentally and numerically that SBS based in-band OSNR monitoring technique is feasible for 80 Gbps DP-NRZ OOK signal, showing compliance of in-band OSNR monitoring for dual polarization modulation formats. By intentionally introducing in-band ASE noise to the signal to be monitored, we have shown that the OSNR dynamic monitoring range is drastically enhanced for 44.6 Gbps DPSK, 44.6 Gbps DQPSK and 112 Gbps DP-QPSK signals in both 100 and 50 GHz channel spacing. When using optical power accuracy of +/−0.1 dB, in-band OSNR accuracy better than +/−1 dB can be achieved for OSNRreal<25dB and OSNRreal<32 dB in the 50 GHz and 100 GHz channel spacing scenarios respectively. The proposed technique provides a practical approach for the in band OSNR monitoring of high bit rate modulation formats compliant with the OSNR requirements in optical networks with 100 GHz and 50 GHz channel spacing.
References and links
1. D. C. Kilper, R. Bach, D. J. Blumenthal, D. Einstein, T. Landolsi, L. Ostar, M. Preiss, and A. E. Willner, “Optical performance monitoring,” J. Lightwave Technol. 22(1), 294–304 (2004). [CrossRef]
2. J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photon. Technol. Lett. 13(1), 88–90 (2001). [CrossRef]
3. T. B. Anderson, K. Clarke, S. D. Dods and M. Bakaul, “Robust, low cost, in-band optical signal to noise monitoring using polarization diversity”, presented at OFC/NFOEC 2007, paper OMM3 (2007).
4. X. Liu, Y.-H. Kao, S. Chandrasekhar, I. Kang, S. Cabot, and L. L. Buhl, “OSNR Monitoring Method for OOK and DPSK Based on Optical Delay Interferometer,” IEEE Photon. Technol. Lett. 19(15), 1172–1174 (2007). [CrossRef]
5. T. T. Ng, J. L. Blows, M. Rochette, J. A. Bolger, I. Littler, and B. J. Eggleton, “In band OSNR monitoring and chromatic dispersion monitoring using a fiber optical parametric amplifier,” Opt. Express 13(14), 5542–5552 (2005). [CrossRef] [PubMed]
6. R. Adams, M. Rochette, T. T. Ng, and B. J. Eggleton, “All optical in band OSNR monitoring at 40 Gb/s using nonlinear optical loop mirror,” IEEE Photon. Technol. Lett. 18(3), 469–471 (2006). [CrossRef]
7. K. Clarke, T. Anderson and S. D. Dods, “Monitoring of multiple modulation formats using asynchronous delay-tap sampling”, presented at COIN-ACOFT 2007, (2007).
8. D. Dahan, D. Levy and U. Mahlab, “Low cost Multi-impairment monitoring technique for 43 Gbps DPSK and 86 Gbps DP-DPSK using delay tap asynchronous sampling method”, presented at ECOC 2009, P3.01 (2009).
9. E. Flood, W. H. Guo, D. Reid, M. Lynch, A. L. Bradley, L. P. Barry, and J. F. Donegan, “In-band OSNR monitoring using a pair of Michelson fiber interferometers,” Opt. Express 18(4), 3618–3625 (2010). [CrossRef] [PubMed]
10. T. B. Iredale, M. D. Pelusi, and B. J. Eggleton, “Highly-sensitive all optical in band OSNR monitoring using stimulated Brillouin scattering”, presented at COIN/ACOFT.2007, (2007).
11. A. Fu, M. D. Pelusi, B. J. Eggleton, “Simultaneous Monitoring of In-Band Optical Noise for WDM Signals using Stimulated Brillouin Scattering” presented at OFC/NFOEC 2010, paper OWB2 (2010).
12. G. P. Agrawal, “Nonlinear Fiber Optics”, 3rd Ed., (Academic Press, 2001).
13. L. Thévenaz et al., “All-optical polarization control through Brillouin amplification”, presented at OFC/NFOEC 2008, paper OML73 (2008).