We investigate an optical performance monitor based on Stimulated Brillouin scattering (SBS), for enabling the measurement of the in-band optical signal to noise ratio (OSNR) for multiple channels of a wavelength-division multiplexed (WDM) signal simultaneously. The principle relies on propagating the signal in a nonlinear waveguide so that each channel pumps SBS to produce a back-scattered Stokes wave of unique carrier wavelength, and with a power that depends on the in-band OSNR of the channel itself. We experimentally demonstrate a highly sensitive OSNR measurement for a 3 × 40 Gb/s signal, with a small sensitivity to the input state of polarization, and a large dynamic range (25 dB) in the Stokes power. Our results also reveal the insensitivity of SBS to both chromatic and polarization-mode dispersions, and the indirect role these effects can play in mitigating the suppression of SBS from the nonlinear Kerr effect.
© 2010 OSA
Optical performance monitoring plays an essential role in optical communication networks, for diagnosing the signal quality in transmission . A crucial measurement is the optical signal to noise ratio (OSNR), which has been traditionally performed using an optical spectrum analyser (OSA) or filtering devices , . The basis of these methods is to measure the channel power, and then extrapolate the in-band noise of the signal from a power measurement of the out-of-band noise floor. While this is adequate for simple communication links, it becomes prone to errors for transmission of wavelength division multiplexed (WDM) signals in more complex network architectures, where discrepancies in the out-of-band power level can arise (for example from, filtering, or signal spectrum crosstalk or broadening from nonlinear effects). To resolve this, a variety of different approaches have been investigated to enable a direct measurement of the in-band OSNR. A well known method is polarization nulling , which works best when the signal is polarized, and the noise un-polarized (rather than partially polarized), and therefore requires the signal to be free of polarization mode dispersion (PMD), unless special filtering is used . Other methods include electro-optic phase modulation , optical delay interferometers  and a variety of nonlinear optic schemes, , , including optical autocorrelation . These methods are typically limited to single channel operation. Monitoring the in-band OSNR of a WDM signal therefore demands a parallel bank of devices, or a tunable optical filter to select a particular channel to test . A different approach that can simultaneously monitor WDM signals in a single operation is therefore of interest for reducing measurement latency, while providing a more scalable solution. Achieving this without compromising the measurement sensitivity and inter-channel cross talk are also important.
Signal processing based on nonlinear optics can offer unique capabilities for WDM signals  that can be scaled to ultra-high bit-rates. The OSNR monitors demonstrated to date have focused on the use of the optical Kerr effect in nonlinear waveguides, whereby the refractive index increases proportionally to the instantaneous optical intensity. The ultra-fast nonlinear response translates rapid power fluctuations over time, to rapid power fluctuations at new spectral frequencies, which can then be converted to time-averaged power variations by either optical filtering  or interferometry . The corresponding nonlinear power transfer function (PTF) of the instantaneous input to average output powers provides the necessary discrimination of the noise for measuring the OSNR. Demonstrations include using parametric amplification based on four wave mixing (FWM) , and a nonlinear optical fiber loop mirror using self-phase modulation . However, the strict reliance of the optical Kerr effect on instantaneous peak intensity enforces a susceptibility to intensity changes due to dispersion, and crosstalk from cross phase modulation (XPM), (as reported for signal regenerators based on SPM ). Furthermore, these schemes lack sufficient sensitivity to resolve OSNRs beyond 20 dB, where the corresponding changes in the time-averaged intensity become negligible. The overall dynamic range of the power reading for a wide ranging OSNR can also be limited to several dBs .
In this paper, we investigate using the nonlinear effect of Stimulated Brillouin Scattering (SBS) in a nonlinear waveguide for monitoring the in-band OSNR of WDM signals. The basis of the technique is the nonlinear PTF originating from the lower power needed for generating SBS from a signal compared to noise, due to its higher peak spectral power density. Thus, by launching an optical channel of combined signal and noise of fixed total power above the critical level for SBS (of a noise-free signal), then a back-scatted wave is produced with a power that depends on the OSNR of the channel itself. We show that by launching a WDM signal into a nonlinear waveguide, each channel can independently pump SBS to produce a back-scattered Stokes wave at a unique carrier frequency, and with a power that depends on its channel OSNR. We show that by proper choice of the nonlinear waveguide, SBS can be induced at sufficiently low powers where the impact of Kerr nonlinearities, such as SPM, and XPM are minimal. We also highlight the importance of using a waveguide with a large, non-zero dispersion parameter for avoiding inter-channel crosstalk from both FWM and XPM.
Experimental results demonstrate the measurement of the in-band OSNR for 3 × 40 Gb/s WDM signals, encoded as non-return to zero (NRZ) on-off keying (OOK), and with a 200 GHz channel spacing. A highly sensitive measurement of arbitrary OSNR in the 5-30 dB range is achieved for all channels, along with a large dynamic range (~25 dB) in the Stokes power, and a small sensitivity to the state of polarization of the input signal. Furthermore, our results reveal the insensitivity of SBS to both chromatic and polarization mode dispersions (PMD), and the indirect role these effects can play in mitigating SBS suppression from XPM-induced frequency modulation of WDM signals.
A. Stimulated Brillouin Scattering critical power
The in-band OSNR monitoring of WDM signals relies on tapping a portion of the signal from the network, and launching it into a nonlinear waveguide with sufficient power to pump SBS, as shown schematically in Fig. 1 . With no other fields present, the SBS process in classical terms arises from the spontaneous inelastic scattering off acoustic waves that are excited thermally within the waveguide . The moving interference beat pattern between the incident and generated scattered fields (of different carrier frequency), creates a pressure wave by electrostriction, which creates a moving, periodic compression of the fiber’s density. This in turn modulates the optical susceptibility of the waveguide by the elasto-optic effect, to create a moving refractive index grating, which stimulates further scattering of the pump light by diffraction, leading to enhancement of both the acoustic and back-scattered Stokes waves, and the overall efficiency of SBS. The Stokes wave produced is downshifted from the carrier frequency of the pump, typically by ≈11 GHz at the 1550 nm wavelength for silica at room temperature, due to the Doppler effect associated with the grating moving at the velocity of the acoustic wave.
An equation for roughly estimating the critical launch power needed for significant SBS is gB⋅ P0sg⋅ Leff /Aeff ≈21 where P0sg is the critical pump power for the signal, gB is the peak value of the Brillouin gain, and Leff and Aeff are the fiber’s effective length (accounting for propagation loss) and effective mode area respectively . The accuracy of this equation and its length dependence has been investigated . In practice, the actual P0sg depends on the peak spectral power density of the signal relative to the Brillouin gain spectrum , which has a typically narrow bandwidth, Γ, on the order of 27 MHz in silica fibers, (corresponding to the acoustic phonon lifetime of ≈12 ns). Consequently, P0sg is lowest for a narrow linewidth CW laser. Although typical high-speed optical communication signals have spectral widths much broader than that of the Brillouin-gain spectrum, signals encoded by on-off keying for example, have prominent, and narrow spectral peaks (in the case of NRZ-OOK, an intense spike at its carrier wavelength), that can promote SBS at just a moderately higher P0sg .
The P0sg for a NRZ signal encoded with a pseudo-random bit sequence (PRBS) is predicted to be 3 dB higher than for a CW laser of the same launch power , . In practice, however, P0sg has been found to be only 2 dB more at higher bit rates of 10 Gb/s, due to the contribution of Raleigh backscattering, that arises when its spectral width becomes broad enough (at higher signal bit-rate) to overlap with the narrow Brillouin gain zone (at ≈11 GHz offset from the forward propagating signal acting as the pump) . The measured P0sg for data signals is further complicated by the impact of Kerr nonlinearities, such as SPM ,  and XPM [21–23], as discussed in Section 4.
B. OSNR monitoring principle
The basis for OSNR monitoring is the dependence of the back-scattered power from the nonlinear waveguide, Pbs, on the OSNR of the input channel launched with a total power (Pch), constituted by signal and noise, of powers Psig and Pnse respectively i.e. Pch = Psig + Pnse. The principle is based on the signal bandwidth being much broader than Γ so that a change in its peak spectral power density translates to significant change in Pbs . For low launch powers (where Pch « P0sg), the PTF of Pbs increases linearly with both Psig and Pnse, due to both Fresnel reflections (primarily from fiber connections) and spontaneous scattering. On the other hand, the PTF of the signal given by Pbs = T(Psig ), exhibits a critical launch power (i.e. P0sg), whereupon Pbs increases sharply due to SBS, as drawn schematically in Fig. 1. The equivalent PTF for noise, however, remains linear over the same range, since its critical power (i.e. P0ns), is much higher, given its lower spectral power density relative to the narrow SBS gain bandwidth. Thus, by keeping Pch fixed, while satisfying Pch > P0sg, gives an OSNR dependent Pbs, which is maximum for a noiseless signal (i.e. Pch = Psig), and decreases with increasing Pnse, according to Pbs ≈ T(Pch − Pnse ). The same applies to a multi-channel signal, in which case the monitor must satisfy Pch > P0sg for all channels. Simultaneous OSNR monitoring can then be performed by implementing the scheme as a nonlinear signal processing front end on either an OSA, or WDM demultiplexer with an array of photodetectors to simultaneously measure Pbs for each channel.
Monitoring a WDM signal also relies on each channel of different carrier frequency pumping SBS at a different frequency, without crosstalk from neighboring channels. It has been shown for the simple case of two CW pumps, this is satisfied when the channel spacing, Ω, satisfies Ω » Γ. i.e. for Ω on the order of at least 100 MHz .
A. Preliminary SBS measurements
The OSNR monitor was demonstrated using the experimental set-up shown in Fig. 2 . A 3 × 40 Gb/s NRZ signal was generated using a WDM multiplexer (MUX) to combine three CW lasers, of specified spectral width < 30 MHz. These were separated by 200 GHz on the ITU grid at wavelengths 1549.32 nm, 1550.92 nm, and 1552.52 nm, denoted channels 1, 2 and 3 respectively, and co-polarized by use of polarization controllers (PCs). The combined output was amplitude modulated by a LiNbO3 Mach-Zehnder (MZ) modulator to encode a 40 Gb/s NRZ-OOK, PRBS of 231−1 pattern length. The channels were then temporally decorrelated by the group delay from propagation through 4 km of standard single mode fiber (SSMF). Its impact on SBS is discussed in Section 4. The OSNR was then controlled by adding noise from a cascade of two amplified spontaneous emission (ASE) sources, with a tunable bandpass optical filter (BPF) placed in between. The noise power was adjusted by a variable optical attenuator (VOA), before being combined with the signal. Both were then amplified in an Erbium-doped fiber amplifier (EDFA) and launched into the nonlinear fiber via a 99:1 coupler and optical circulator, which enabled Pch and Pbs for each channel to be measured by OSAs, respectively, using a 1.0 nm resolution bandwidth (RBW) setting.
The SBS efficiency for a given pump power is determined by the nonlinear waveguide characteristics. A large Leff /Aeff, reduces the launch power needed to satisfy Pch > P0sg. Of importance also, was choosing a waveguide with a sufficient non-zero dispersion, in order to mismatch the phase and group velocities of the co-propagating waves, and thereby suppress FWM and XPM, respectively. The impact of XPM is discussed further in Section 4. With these factors in mind, our experiment used a 1 km length of highly nonlinear fiber (HNLF) with a nonlinearity coefficient of 11.5 W−1km−1 (Aeff ≈12 μm2) and a dispersion equal to −2.47 ps/nm.km (of normal sign) at the 1550 nm wavelength, and a dispersion-slope of 0.02 ps/nm2.km. Importantly, the critical powers needed for SBS in this fiber were at a sufficiently low level, where the undesired Kerr nonlinearities play a minor role, as discussed in the following sections.
The SBS efficiency for the HNLF was characterized by measuring the PTFs for various sources, including a CW laser, and both single channel (Ch. 2) and WDM 40 Gb/s intensity modulated signals, encoded as either NRZ, or carrier-suppressed, return to zero (CS-RZ). The curves in Fig. 3(a) show that P0sg was lowest for the CW laser, and increased for the NRZ and CS-RZ signals, as expected from their lower peak spectral power density relative to the Brillouin gain bandwidth [15–17]. Both Figs. 3(a) and 3(b) show a ≈2 dB increase in P0sg for the single channel 40 Gb/s NRZ signal versus single channel CW laser; consistent with the observations reported in  for a 10 Gb/s NRZ signal. The highest P0sg was for CS-RZ, as expected from the lower peak power density of its spectrum containing two prominent, narrow lines, instead of one as for both NRZ and the CW laser. The consequences for scaling the OSNR monitoring to alternative modulation formats and different bit rates are discussed in Section 4. A further increase in P0sg was also observed for WDM intensity modulated signals, because of the Kerr nonlinearities in the HNLF, as also discussed in Section 4.
B. OSNR monitoring results
The fiber was then applied to the OSNR monitoring of the WDM 3 × 40 Gb/s NRZ signal. The initial experiment considered noise applied to only the central channel by tuning the center wavelength of the 0.5 nm bandwidth, BPF in the noise circuit to Ch. 2. To test the wavelength independency of the monitor, all three WDM channels were set at equal powers of Pch = 59 mW, which satisfied Pch > P0sg. The OSNR for Ch. 2 was determined by measuring the noise and signal powers separately at the input to the EDFA using an OSA. The input power to the EDFA was 0.2 mW, which ensured the EDFA added minimal noise. The OSNR was varied by adjusting the VOA in the noise circuit of Fig. 2. For each setting, the power of the Ch. 2 CW laser was adjusted to keep Pch constant.
Figure 4(a) shows the optical spectra traces of the input, throughput and back reflected emission from the HNLF for a 15 dB OSNR (Ch. 2). Note, all OSNR values stated in this paper are with respect to a 0.1 nm noise bandwidth. The throughput spectrum in Fig. 4(a)(iii) showed that FWM in the HNLF remained small, thanks to the non-zero dispersion of the fiber. Comparison of the input and reflected emission spectra showed that although Pch was equal for all 3 channels, a drop in Pbs was observed only for Ch. 2. This is explained by the corresponding change in T(Pch−Pnse) for the drop in Psig at lower OSNR. Fig. 4(b) plots the measured Pbs versus OSNR for Ch. 2, highlighting the large dynamic range (exceeding 25 dB) in Pbs, as the OSNR varied between 5-25 dB. In contrast, Pbs for the noise-free channels 1 and 3 remained unchanged. The dynamic range is notably larger than reported for other nonlinear methods based on the optical Kerr effect , . Similar results were also obtained with the 0.5 nm bandwidth noise source tuned to Ch. 3, (leaving channels 1 and 2 noise-free), as plotted in Fig. 4(b). The wavelength dependence of the SBS process was also noted to be minor between channels as shown in Fig. 5(a) . The change in Stokes power over wider wavelength spans is also expected to be small, from the corresponding change in gB /Aeff .
We also investigated the capability of measuring the OSNR for a broadband noisy WDM signal. This was performed by swapping an 8 nm bandwidth BPF into the noise circuit. Again, Pch = 59 mW was used for each channel. As the proportion of noise power was increased by varying the VOA, Pch was equalized by adjusting each CW laser power, and tweaking the EDFA gain. The OSNR of each channel was determined by comparing the signal and noise powers at the input to the EDFA. Fig. 4(c) shows the curves of Pbs versus OSNR closely overlapped for all channels, with a dynamic range and OSNR sensitivity similar to the 0.5 nm bandwidth noise result in Fig. 4(b).
A. Power transfer function (PTF) analysis
The OSNR monitoring principle, as described in Section 2, was validated by calculating the expected back-scattered power variation as a function of OSNR using the measured Ch. 2 SBS PTF for the 3 × 40 Gb/s WDM signal. The procedure was as follows:
- • Psig was first mapped to the theoretical OSNR for a fixed Pch = 59 mW, according to Psig = Pch × OSNR/(OSNR + F), where F is a calibration factor given by the ratio of the total noise power, to that within a 0.1 nm RBW at its center wavelength. The optical filter used in the noise circuit gave F = 5.6. The signal contribution to Pbs was then calculated by designating Psig as the x-axis channel power of the SBS-PTF.
- • The noise contribution was then added by designating the corresponding noise power (Pch − Psig) as the x-axis channel power, and calculating Pbs by interpolation from the linear fit to the PTF for the signal at low channel powers, according to the expectation that P0ns » P0sg. It was verified experimentally, that the PTF for noise alone remained linear over the same channel power range.
Figure 4(b) shows that the calculated back-scattered power versus OSNR followed the experimentally measured curve.
The OSNR monitoring of WDM signals relied on the independent SBS for each channel, as discussed in Section 2. This is evident in Figs. 3(b) and 5(a), from the closely matching PTFs generated simultaneously for each channel of the 3 × 40 Gb/s NRZ signal, at different center wavelengths of 1558.98 and 1550.92 nm, respectively. This ensures that an arbitrarily measured value of Pbs for all channels with the same Pch corresponds to a unique OSNR. The value of Pch itself, can tailor the overall sensitivity and dynamic range of the technique, as highlighted in Fig. 5(b), which plots the expected Pbs versus OSNR for different Pch as calculated from the measured Ch. 2 SBS-PTF for the 3 × 40 Gb/s WDM input signal, (using the procedure described above). It shows that using a lower Pch approaching P0sg, can take advantage of the greater nonlinearity of the PTF to resolve OSNRs beyond 30 dB. The trade-off, however, is the reduction in the dynamic range for Pbs.
B. SBS suppression factors
As discussed in Section 2, the power efficiency of SBS is determined by the peak spectral power density of the pump, relative to the narrow Brillouin gain bandwidth. In applications where SBS is an undesired effect, it is often suppressed by purposely dithering the carrier frequency of the pump laser (i.e. signal) beyond Γ (i.e. at around 100 MHz). Doing so broadens the pump linewidth and reduces its spectral power density, thereby increasing P0sg. The conventional method is by electrically dithering the laser current , , , or by frequency modulating the laser output with an external electro-optic phase modulator , . Alternative approaches that avoid the signal distortion induced by frequency modulation, include designing fiber transmission links comprised of different segments with varying parameters , ,  or non-uniform temperature distribution . While SBS suppression is contrary to the needs of our experiment, frequency modulation of the pump can arise from the optical Kerr effect in the waveguide, by SPM ,  and XPM [21–23]. Notably, the corresponding Kerr-induced SBS suppression, which we refer to as “KISS”, only arises for an intensity modulated pump source, and not the CW case, as highlighted below.
The impact of KISS on our experiment is evidently shown in Fig. 3(b) by comparing the PTFs for the 40 Gb/s NRZ signals, versus the single and multi-channel CW laser sources, of 100 and 200 GHz spacing. It shows the PTFs of both single and multi-channel CW sources, exhibit negligible difference, as expected from the absence of KISS by both SPM and XPM.
In contrast, for the 40 Gb/s NRZ signals, the PTFs indicate a higher P0sg for the WDM versus single channel case (where XPM is zero). The nonlinearity of the PTF for SBS is so steep, that even just a slight dip of the main spectral peak in the NRZ spectrum from weak spectral broadening from either SPM or XPM, can reduce the corresponding Stokes power by several dBs. Fortunately for our experiment, KISS played a minor overall, since the nonlinear phase shifts associated with the optical Kerr-effect for the channel powers used, were much less than π.
Figure 3(b) shows the effect of FWM for different WDM channel spacing, by comparing the SBS power transfer function for a single CW laser to both three channel CW sources of 100 and 200 GHz channel spacing. The close match for all curves highlights the small impact of FWM, even for a narrow 100 GHz channel spacing, as intended by choosing a HNLF with non-zero dispersion parameter to mitigate phase-matched FWM. Notably, our experiment considered the worst case scenario for FWM, with all wavelengths co-polarized. If FWM became an issue (such as for even narrower channel spacing), then the choice of both the length and dispersion parameters of the HNLF could be further optimized with respect to the coherence length for phase-matched FWM  to mitigate its effect. On the other hand, Sub-section 4(d) explains that although the HNLF was effective in reducing FWM, the impact of XPM was non-negligible. In this case, the suppression of XPM required exploiting the dispersion properties of the HNLF or pre-distorting the WDM signal itself.
C. Input signal effects on KISS
The undesired impact of KISS can be mitigated by the fact that SBS itself is insensitive to many characteristics of the input pump, to which the Kerr nonlinearity is not. These include input state of polarization, and distortion from chromatic dispersion and PMD, including differential group delay (DGD). The intrinsic insensitivity of SBS to the input state of polarization of the WDM signal was observed experimentally, as predicted from theory . The difference in Pbs was within 0.5 dB, for launching the 3 × 40 Gb/s NRZ signal of arbitrary state of polarization into the HNLF. The PTF for the single channel 40 Gb/s signal is also insensitive to distortions both chromatic and DGD dispersion effects, (given that the typical dispersion time-scales are much shorter than the acoustic phonon lifetime).
The effect on KISS from pre-distorting the signal by chromatic dispersion was investigated by varying the length of SSMF in the Fig. 2 setup. While signal dispersion itself, doesn’t directly impact SBS in the waveguide, it does suppress the KISS from XPM because of the reduced peak power associated with pulse broadening (and for our experiment, by the simultaneous decorrelation of the data patterns for the WDM channels). The PTF of the single channel 40 Gb/s NRZ signal shifted only slightly when the 4 km SSMF was removed, as shown in Fig. 6(a) . Similarly, the PTF of the WDM-NRZ signal exhibited a slight shift from the single channel case, due to the onset of KISS from XPM. Furthermore, changing the SSMF length from 4 km to 2 km had negligible effect on both the PTF and OSNR curves as shown in Fig. 6(a) and 6 (b) respectively, highlighting the insensitivity of the scheme to dispersion (in the limit where XPM is suppressed).
The accuracy of the monitoring scheme was investigated by using the data in Fig. 6(b) (comparing the OSNR response for a NRZ WDM signal pre-distorted by either a 2 or 4 km length of SSMF). The calculation defined the “4 km” curve as the reference for the purpose of mapping an arbitrarily measured Pbs (in this example, the back scattered power points from the “2 km” curve) to the inferred OSNR by linear interpolation. The error between the inferred OSNR and its true value was found to be less than ± 0.5 dB for 9 < OSNR < 23 dB, and less than ± 1 dB for 3 < OSNR < 36 dB, as shown in Fig. 6(c). It should be noted performing linear interpolation with such a limited data set as the reference (i.e. only 7 data points across the entire OSNR range in this case) induced some error. Nevertheless, the result highlights both the accuracy of the scheme and insensitivity to dispersion achieved.
Notably, our experiment had considered the worst case scenario for signal polarization, with all WDM channels co-polarized. In the case of launching a WDM signal with orthogonal state of polarization between two channels, the XPM effect is expected to reduce by a factor of three .
KISS is also mitigated by pre-distorting the signal with PMD. The impact of PMD in general on suppressing XPM of WDM signals is well known . For our experiment, a PMD emulator with programmable DGD was inserted immediately after the data modulator in the set-up of Fig. 2. The instrument converted each incoming linearly polarized data channel into two linear orthogonal polarization states, with a programmable relative time delay shift. The XPM suppression arises from both the temporal delay between the two linearly orthogonal polarization states of the channel, (providing an effective data pattern decorrelation), and the polarization dependence of XPM itself, as described above. While DGD had little direct impact on SBS generation for a single channel signal, a shift of the PTF to lower P0sg was observed for the pre-distorted WDM signal, as shown in Fig. 6(a), indicating a counter-effect to KISS. In particular, for a single channel 40 Gb/s NRZ, with Pch = 18.8 dBm, Pbs remained at 9.7 dBm for arbitrary DGD between 0.18 to 22 ps (while keeping the 4 km SSMF in the set-up), highlighting the insensitivity of SBS to DGD. On the other hand, for the 3 × 40 Gb/s NRZ, with the same Pch, the Pbs varied between 7~9 dBm.
The combined effect of signal pre-distortion by DGD and chromatic dispersion (from propagation in 4 km SSMF) is shown in Fig. 6(a), and highlights the convergence of the SBS PTF toward the single channel 40 Gb/s case as XPM is suppressed. The results overall validate the benefit of pre-distorting the WDM signal by either DGD, or chromatic dispersion.
D. Waveguide effects on KISS
Another method of countering KISS is by taking advantage of the chromatic dispersion within the nonlinear waveguide itself. In this case, XPM is countered by an induced time-delay between channels (i.e. “walk-off”), rather than dispersion of the signal itself (as discussed above). The walk-off arising from the refractive index variation with wavelength causes
WDM channels of different carrier wavelength to propagate with different group velocities, and therefore experience a carrier frequency dependent decrease in their nonlinear interaction (i.e. XPM). The worst case XPM corresponding to the least walk-off would therefore be expected for channels propagating at wavelengths where the fiber refractive indices are more closely matched (i.e. for adjacent channels with our HNLF) . Considering a wavelength channel spacing of Δλ, the walk-off length, Lw, corresponding to the delay equaling half the bit period for signal of bit rate, B, is given by, Lw = 1/|B·Δλ·(2D − S·Δλ)|, where D and S are the fiber dispersion and slope, respectively, specified at the pump wavelength. A reduced inter-channel nonlinear interaction is therefore expected when L /Lw » 1. However, the HNLF used in the experiment, had L /Lw = 0.32 and 0.16 for a channel spacing of 200 and 100 GHz respectively (i.e. Δ λ = 1.6 nm and 0.8 nm respectively), which suggests that walk-off played a minor role in reducing XPM in the 1 km HNLF (for adjacent channels at least).
Alternatively, XPM (and KISS) could also be mitigated by using a nonlinear waveguide with an intrinsically high DGD (i.e. frequency dependent birefringence), such as from a polarization maintaining HNLF. Such a fiber has been shown effective for suppressing XPM of WDM signals in the application of simultaneous multi-wavelength signal regeneration . It would therefore be equally beneficial for countering the KISS from XPM.
Scaling the technique to monitor more WDM channels is not expected to significantly impact XPM, since the worst case is expected for adjacent channels . Operation at a higher bit-rate, or with alternate data modulation formats such as return-to-zero (RZ) OOK, and CS-RZ-OOK, and phase shift keying, [15–17], will change P0sg. It has been observed experimentally, that P0sg for RZ-OOK is 3 dB higher than NRZ-OOK . Thus, applying the technique to monitor a WDM signal of different modulation formats or bit-rate, would require adjusting the power of each channel accordingly to satisfy Pch > P0sg. The total average launch power requirement scales with Pch and ultimately constrains the maximum number of channels that can be monitored simultaneously by the technique. This was limited in our experiment by the available optical amplifiers. The practical application of the technique would therefore benefit from further progress in the development of highly nonlinear glasses such as chalcogenides , and waveguides with larger Leff /Aeff to further reduce P0sg. This would be particularly important for advanced phase shift keying data formats, which have predicted P0sg to be on the order of 10 times higher than the CW case, and more sensitive to the bit-rate , .
Applying the OSNR monitoring scheme to signals that have been pre-distorted by frequency dithering, will also induce an undesirable increase in P0sg. In such cases, the application of the monitor would be more compatible with signaling systems using alternative methods of SBS suppression (if required) , , .
We investigated the use of SBS in a nonlinear fiber for enabling the simultaneous multi-channel OSNR monitoring of a WDM signal. Experimental results showed the capability to measure the OSNR of a 3 × 40 Gb/s NRZ WDM signal with a 200 GHz channel spacing. A high dynamic range in power of >25 dB for the back-reflected Stokes wave for varying OSNR between 5-25 dB, was demonstrated. The measurement was also insensitive to the launched state of polarization of the WDM signal into the nonlinear waveguide. We also investigated the impact of XPM from the optical Kerr effect in the waveguide on suppressing SBS for a WDM signal pump. We showed that while XPM played a role in suppressing SBS, its effect could be mitigated by chromatic dispersion and differential group delay, which have no direct impact on SBS itself. The SBS suppression could therefore be mitigated by applying dispersion to the signal, either as pre-distortion, or within the waveguide itself. By doing so, it was shown that the SBS-PTF for the WDM signal converged closely toward the single channel case. Applying the technique to more channels, or alternate data modulation formats, will require scaling the launch power.
The authors thank Prof. Luc Thévenaz from the EPFL Swiss Federal Institute of Technology for his constructive feedback. This work was supported in part by the Australian Research Council (ARC) through its ARC Centres of Excellence and Federation Fellowship programs.
References and links
1. Z. Pan, C. Yu, and A. E. Willner, “Optical performance monitoring for the next generation optical communication networks,” Opt. Fiber Technol. 16(1), 20–45 (2010). [CrossRef]
2. W. Chen, S. Zhong, Z. Zhu, W. Chen, and Y.-J. Chen, “Adding OSNR and Wavelength Monitoring Functionalities on a Double-Resolution-AWG-Based Power Monitoring Circuit,” IEEE Photon. Technol. Lett. 15(6), 858–860 (2003). [CrossRef]
3. C.-L. Yang and S.-L. Lee, “OSNR monitoring using double-pass filtering and dithered tunable reflector,” IEEE Photon. Technol. Lett. 16(6), 1570–1572 (2004). [CrossRef]
4. 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]
5. M.-H. Cheung, L.-K. Chen, and C.-K. Chan, “PMD-insensitive OSNR monitoring based on polarization-nulling with off-center narrow-band filtering,” IEEE Photon. Technol. Lett. 16(11), 2562–2564 (2004). [CrossRef]
6. Y.-C. Ku, C.-K. Chan, and L.-K. Chen, “Robust optical signal-to-noise ratio monitoring scheme using a phase-modulator-embedded fiber loop mirror,” Opt. Lett. 32(12), 1752–1754 (2007). [CrossRef] [PubMed]
7. 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]
8. T. T. Ng, J. L. Blows, M. Rochette, J. A. Bolger, I. Littler, and B. J. Eggleton, “In-band OSNR and chromatic dispersion monitoring using a fibre optical parametric amplifier,” Opt. Express 13(14), 5542–5552 (2005). [CrossRef] [PubMed]
9. R. Adams, M. Rochette, T. T. Ng, and B. J. Eggleton, “All-Optical In-Band OSNR monitoring at 40 Gb/s using a nonlinear optical loop mirror,” IEEE Photon. Technol. Lett. 18(3), 469–471 (2006). [CrossRef]
10. M. Dinu, D. C. Kilper, and H. R. Stuart, “Optical performance monitoring using data stream intensity autocorrelation,” J. Lightwave Technol. 24(3), 1194–1202 (2006). [CrossRef]
11. L. Provost, F. Parmigiani, P. Petropoulos, D. J. Richardson, K. Mukasa, M. Takahashi, J. Hiroishi, and M. Tadakuma, “Investigation of four-wavelength regenerator using polarization- and direction-multiplexing,” IEEE Photon. Technol. Lett. 20(20), 1676–1678 (2008). [CrossRef]
12. R. W. Boyd, Nonlinear Optics (Academic Press, 3rd edition, 2008).
13. G. P. Agrawal, Nonlinear Fiber Optics (Academic Press, San Diego, California, 3rd edition, 2001).
14. S. Le Floch and P. Cambon, “Theoretical evaluation of the Brillouin threshold and the steady-state Brillouin equations in standard single-mode optical fibers,” J. Opt. Soc. Am. A 20(6), 1132–1137 (2003). [CrossRef]
15. T. Sugie, “Maximum repeaterless transmission of lightwave systems imposed by stimulated Brillouin scattering in fibres,” Opt. Quantum Electron. 27(7), 643–661 (1995). [CrossRef]
16. E. Lichtman, R. G. Waarts, and A. A. Friesem, “Stimulated Brillouin Scattering excited by a modulated pump wave in single-mode fibers,” J. Lightwave Technol. 7(1), 171–174 (1989). [CrossRef]
17. D. A. Fishman and J. A. Nagel, “Degradations due to Stimulated Brillouin Scattering in multigigabit intensity-modulated fiber-optic systems,” J. Lightwave Technol. 11(11), 1721–1728 (1993). [CrossRef]
18. M. Ferrario, L. Marazzi, P. Boffi, A. Righetti, and M. Martinelli, “Impact of Rayleigh backscattering on Stimulated Brillouin Scattering threshold evaluation for 10 Gb/s NRZ-OOK signals,” Opt. Express 17(20), 18110–18115 (2009). [CrossRef] [PubMed]
19. V. Yu. Golyshev, E. A. Zhukov, I. É. Samartsev, and D. G. Slepov, “The effect of Self-Phase-Modulation on the Stimulated Mandelstam–Brillouin Scattering in fiber-optic communication lines,” Tech. Phys. 49(7), 872–875 (2004). [CrossRef]
20. A. Sano, Y. Miyamoto, T. Kataoka, H. Kawakami, and K. Hagimoto, “10Gbit/s, 300km repeaterless transmission with SBS suppression by the use of the RZ format,” Electron. Lett. 30(20), 1694–1695 (1994). [CrossRef]
21. Y. Horiuchi, S. Yamamoto, and S. Akiba, “Stimulated Brillouin scattering suppression effects induced by cross-phase modulation in high power WDM repeaterless transmission,” Electron. Lett. 34(4), 390–391 (1998). [CrossRef]
22. S. S. Lee, H. J. Lee, W. Seo, and S. G. Lee, “Stimulated Brillouin scattering suppression using cross-phase modulation induced by an optical supervisory channel in WDM links,” IEEE Photon. Technol. Lett. 13(7), 741–743 (2001). [CrossRef]
23. J. D. Downie and J. Hurley, “Experimental study of SBS mitigation and transmission improvement from cross-phase modulation in 10.7 Gb/s unrepeatered systems,” Opt. Express 15(15), 9527–9534 (2007). [CrossRef] [PubMed]
24. E. Lichtman, A. A. Friesem, R. G. Waarts, and H. H. Yaffe, “Stimulated Brillouin scattering excited by two pump waves in single-mode fibers,” J. Opt. Soc. Am. B 4(9), 1397–1403 (1987). [CrossRef]
25. Y. K. Park, O. Mizuhara, L. D. Tzeng, J.-M. P. Delavaux, T. V. Nguyen, M.-L. Kao, P. D. Yeates, and J. Stone, “A 5 Gb/s repeaterless transmission system using Erbium-doped fiber amplifiers,” IEEE Photon. Technol. Lett. 5(1), 79–82 (1993). [CrossRef]
26. R. Elschner, C.-A. Bunge, B. Huttl, A. G. i Coca, C. Schmidt-Langhorst, R. Ludwig, C. Schubert, and K. Petermann, “A.G. i Coca, C.L. Schmidt, R. Ludwig, C. Schubert, K. Petermann, “Impact of pump-phase modulation on FWM-based wavelength conversion of D(Q)PSK signals,” IEEE J. Sel. Top. Quantum Electron. 14(3), 666–673 (2008). [CrossRef]
27. M. R. Lorenzen, D. Noordegraaf, C. V. Nielsen, O. Odgaard, L. Grüner-Nielsen, and K. Rottwitt, “Suppression of Brillouin scattering in fibre-optical parametric amplifier by applying temperature control and phase modulation,” Electron. Lett. 45(2), 125–126 (2009). [CrossRef]
28. K. Shiraki, M. Ohashi, and M. Tateda, “SBS threshold of a fiber with a Brillouin frequency shift distribution,” J. Lightwave Technol. 14(1), 50–57 (1996). [CrossRef]
29. A. Zadok, E. Zilka, A. Eyal, L. Thévenaz, and M. Tur, “Vector analysis of stimulated Brillouin scattering amplification in standard single-mode fibers,” Opt. Express 16(26), 21692–21707 (2008). [CrossRef] [PubMed]
30. Q. Lin and G. P. Agrawal, “Effects of Polarization-Mode Dispersion on Cross-Phase Modulation in Dispersion-Managed Wavelength-Division-Multiplexed Systems,” J. Lightwave Technol. 22(4), 977–987 (2004). [CrossRef]
31. R. Lapp, “Experimental investigation of signal distortions induced by cross-phase modulation combined with dispersion,” IEEE Photon. Technol. Lett. 9(12), 1592–1594 (1997). [CrossRef]
32. A. Prasad, C.-J. Zha, R.-P. Wang, A. Smith, S. Madden, and B. Luther-Davies, “Properties of GexAsySe1-x-y glasses for all-optical signal processing,” Opt. Express 16(4), 2804–2815 (2008). [CrossRef] [PubMed]