We demonstrate wavelength conversion based on four-wave mixing in a semiconductor optical amplifier of signals with quadrature amplitude modulation (QAM). We first demonstrate wavelength conversion of 16 Gbaud 16-QAM signals over the entire C-band using two co-polarized pumps with low power penalty at the forward error correction threshold (FEC) for a wide range of input optical-signal-to-noise-ratio (OSNR). We also demonstrate for the first time wavelength conversion of 5 Gbaud 64-QAM signals in a semiconductor optical amplifier with bit-error rate below the FEC threshold over the entire C-band and investigate the dependence of the power penalty on input OSNR with a single pump configuration.
©2013 Optical Society of America
Coherent systems use spectrally-efficient advanced modulation formats, like QPSK, 16-quadrature amplitude modulation (QAM), and 64-QAM, for better exploitation of the channel capacity . Coherent detection furthermore enables filterless transparent networks . The reach and capabilities of these networks can be improved by the use of wavelength conversion to optimize wavelength usage. Wavelength converters are also critical building blocks for optical routers, for instance in data centers. Whether filterless or optically routed, the performance of future all-optical networks will depend on the efficiency of wavelength converters.
Wavelength conversion of 16-QAM was recently reported in periodically-poled lithium niobate waveguides  and in highly nonlinear fibers . However, these methods have limited tuning range and severe pump power requirements (respectively 17 dBm  and 28 dBm ). Wavelength conversion of 64-QAM in highly nonlinear fibers was also recently demonstrated in . Four-wave mixing (FWM) in semiconductor optical amplifiers (SOAs) has been proposed as an efficient and practical wavelength conversion mechanism because of its compactness, low pump powers, high conversion efficiency and transparency to modulation formats . A limitation of FWM in a SOA is the small conversion range, which is of the order of a few nanometers. Fortunately, it is possible to extend the conversion range over the broad gain spectrum of the SOA by using two pumps .
In , we showed that wavelength conversion of 16-QAM signals was feasible in a commercial multi-quantum well (MQW) SOA with low power penalties over the entire C-band with a co-polarized dual-pump configuration and a 40 dB input optical-signal-to-noise-ratio (OSNR). The accumulation of amplified spontaneous emission (ASE) noise as the signal passes through multiple stages of wavelength conversions of a wavelength routed network can limit the maximum number of wavelength conversion stages on a link. In this paper, we extend the work done in done in  and demonstrate wavelength conversion of 16-QAM signals for a wide range of input OSNR.
In addition, we report for the first time, to the best of our knowledge, wavelength conversion of 64-QAM signals in a SOA for which impairments such as nonlinear phase noise and limited conversion efficiency (CE) are even more critical than for 16-QAM. We demonstrate wavelength conversion with measured bit-error rate (BER) below the forward error correction (FEC) threshold over the entire C-band using a dual-pump configuration. Finally, we study the robustness of the SOA-based wavelength conversion of 64-QAM with respect to the input OSNR using a single pump configuration.
2. Principle of wavelength conversion in SOAs
FWM in a SOA is achieved by simultaneously injecting at its input one or two CW signals called “pumps” together with the data modulated signal to be converted. The gain and the refractive index of the amplifier are then modulated at the frequency detuning, defined as the optical frequency separation between the signal and the pump . A new optical field, often called the conjugate, is generated during the propagation within the SOA. Due to the fact that the phase of the converted field reproduces the phase of the signal, FWM is compatible with phase modulation. For a small frequency detuning (<<10 GHz), the FWM term is generated by the carrier density modulation. Due to the relatively slow recovery of the carrier density, determined by the carrier lifetime, which is on the order of several hundred picoseconds, the efficiency of the wavelength conversion rapidly drops for larger frequency detuning. Instead, intraband dynamics, such as spectral hole burning, which involves non-Fermian population pulsations, and carrier heating, which involves carrier temperature pulsations, dominate. The small intraband relaxation times, in the order of 50 fs to 1 ps, imply that the bandwidth of intraband contributions will easily exceed 1 THz.
In the case of phase modulated signals, the quality of the wavelength conversion is limited by the addition of phase noise originating from the SOA amplitude-phase coupling parameterized by the linewidth enhancement factor, often called the α-factor . This nonlinear phase noise (NLPN) typically originates from self-phase modulation and from the internal noise generated by the SOA via cross-phase modulation (XPM) . In the special case of FWM, NLPN arising from the power fluctuations of the pump via XPM was shown to be critical for higher-order phase modulation formats . The phase noise behavior of saturated SOAs with constant envelope modulation schemes (DQPSK, PSK…) has been analyzed in several papers [10, 12, 13], but such studies have not been performed for M-ary QAM formats. Therefore, even though wavelength conversion with SOAs has been extensively studied with OOK formats, the results cannot be extended to phase modulated signals in a straightforward way. Figure 1 shows the received constellations of a 16 Gbaud 16-QAM signal when the pump power is optimized either for maximizing the CE [Fig. 1(a)] or minimizing the error vector magnitude (EVM) [Fig. 1(b)]. As seen in Fig. 1(a), even if the CE is optimal, the received signal suffers more from added NLPN that translates into a lower EVM (BER) compared to Fig. 1(b). This is expected to be even more critical for 64-QAM.
Quantum-Dot (QD) SOAs have also attracted much interest for wavelength conversion based on FWM because of their theoretically delta-function-like density of states providing potentially low or near-zero α-factor [14, 15]. Wavelength conversion based on FWM in a QD-SOA for 10 Gbaud 16-QAM SOA with a single pump configuration  and a dual-pump configuration  yielded similar performances to wavelength conversion of 16 Gbaud 16-QAM in a MQW-SOA with a dual-pump configuration  over the C-band. However, the α-factor of QD-SOAs reported in the literature in the 1.5 µm band still have significant non-zero values [14, 16, 18]. Furthermore, polarization sensitivity is still an issue with QD-SOAs [14, 19] as both the signal and pump need to be precisely aligned along the polarization axis exhibiting the maximum gain [16, 20]. For those reasons, the mature technologies of bulk SOAs and MQW-SOAs may still be the best choice in order to perform 16-QAM and 64-QAM wavelength conversion and further investigation is necessary.
3. Experimental results
3.1 Experimental setup
The experimental setup is presented in Fig. 2. A tunable laser at λs = 1550.3 nm is modulated by an in-phase/quadrature Mach-Zehnder modulator (IQ-MZM). For the 16-QAM experiments, the IQ-MZM is driven by a bit pattern generator (BPG) producing 4-level in-phase and quadrature data streams from a 211 −1 pseudo-random-bit-sequence. For 64-QAM experiments, the IQ-MZM is driven by an arbitrary waveform generator (AWG) with 6 bits resolution. The AWG generates two 8-level electrical signals with a repeated sequence of 393216 bits taken from a pseudo random binary sequence of length 231-1 (limited by AWG memory). The modulator output is fed into a variable optical attenuator (VOA) followed by an erbium-doped fiber amplifier (EDFA), an optical filter (OF) with 1 nm bandwidth tuned at λs and a polarization controller (PC).
The signal OSNR, measured over a bandwidth of 0.1 nm, is adjusted by varying the attenuation and the EDFA gain. The signal and both pumps are combined using a 4x1 coupler (2x1 coupler for single pump experiments). Two other tunable lasers at λP1 and λP2 delivering the pumps are amplified by two EDFAs, followed by two OFs tuned at λP1 and λP2 respectively, and two PCs. The signal wavelength is fixed at 1550.3 nm for all experiments. The wavelength detuning between the signal and the first pump is nm for 16-QAM wavelength conversion and nm for 64-QAM. It is the wavelength of the second pump that is varied in order to tune the wavelength of the converted signal and achieve different wavelength detuning. The modulated signal and the pumps are simultaneously injected into a nonlinear SOA (SOA1117/COVEGA) operating over the C-band (1528 nm to 1562 nm) with 20 dB small-signal gain, 9 dBm output saturation power and < 1 dB polarization dependent gain. Two isolators (not shown in Fig. 2) are used before and after the SOA to suppress back reflections. The FWM term (the conjugate) is filtered with a tunable OF with variable bandwidth (BFV-200-SM-FA/Alnair Labs) and passed through a VOA before going to the pre-amplified receiver. The coherently detected signal was sampled by the real-time oscilloscope at 80 GSa/s with 30 GHz electrical bandwidth. In the digital signal processing stage, we first applied a Gaussian low-pass filter, performed resampling, timing recovery and frequency offset compensation using the estimator suggested in . For 16-QAM, a minimum mean square error filter was applied to mitigate the effect of limited receiver bandwidth, followed by a decision-aided maximum likelihood algorithm to estimate the carrier phase . For 64-QAM, following frequency offset compensation, a Wiener-Hopf-based decision-directed equalizer (WHDD-EQ)  followed by a decision directed phase locked loop was used for phase recovery. We again applied a WH-DD-EQ to further equalize with the more reliable decisions following phase recovery . Finally, hard-threshold decision was performed on I and Q individually, we counted the errors and estimated BER.
3.2 16-QAM wavelength conversion
We perform wavelength conversion of the 16-QAM signal over the C-band with the conjugate wavelength spanning from nm tonm, with an input OSNR of 40 dB. As discussed in the introduction, the system was not optimized for maximum CE but rather by monitoring the constellation of the conjugate signal and minimizing the EVM. The signal power is −12 dBm and both pumps power is 5 dBm at the SOA input for all the measurements. This optimization was done only once for the whole set of measurements. Table 1 shows the conversion efficiency, defined as the ratio of the converted signal power over the input signal power, for each conjugate wavelength.With our current experimental setup, we could achieve a wavelength conversion bandwidth of 32 nm, limited only by the tuning range of the local oscillator at the coherent receiver. However, it also happens to coincide with the SOA optical bandwidth. The laser linewidths are respectively 22 kHz and 16 kHz for the signal and the LO, and 100 kHz for both pumps.The measured BER as a function of the received power, measured at the input of the pre-amplified receiver, is displayed in Fig. 3. The results in Fig. 3 clearly indicate that the performance in terms of BER is similar over the whole band with very low power penalties measured at each conjugate wavelength, i.e. from 0.5 dB to less than 1.5 dB at the FEC limit (BER = 2.2 × 10−3). This confirms that optimizing EVM also optimizes BER for the regime examined.
In a wavelength routed network, the signal will pass through multiple stages of wavelength conversion and will accumulate noise. In order to investigate the robustness of the wavelength conversion against ASE noise, we measured the BER as a function of received power for degraded OSNR values of 30 dB [Fig. 4(a)] and 20 dB [Fig. 4(b)]. We see that the power penalty at the FEC threshold is smaller than 1.5 dB for every OSNR compared to their respective back-to-back curve. Also, for an OSNR of 30 dB, we observe that the power penalty induced by the wavelength conversion is similar to the 40 dB OSNR case for every conjugate wavelength. However, the results show that the power penalty becomes severe for an OSNR of 20 dB: a BER floor is visible and the power penalty show more variations among the conjugate wavelengths. For instance, 4 dB more power is needed at the receiver in order to achieve a BER of 10-3.5 for nm compared to nm. However, at the FEC threshold, the variations in the power penalty are significantly reduced showing a spread of only 1.5 dB among the conjugate wavelengths.
3.3 64-QAM wavelength conversion
In the second experiment, we demonstrate wavelength conversion of 64-QAM signals with high OSNR (36 dB) over the entire C-band with a dual-pump configuration. The signal power is −12 dBm and both pump powers are 5 dBm at the SOA input for all the measurements. Similarly to the 16-QAM experiments, the signal and pump power optimization was done only once for the whole set of measurements. The CE and the received power penalty at the FEC threshold are displayed in Table 2 for three conjugate wavelengths at the edges and at the center of the band (nm, nm and nm).
For 64-QAM experiments, the laser linewidths are 16 kHz for both the signal and the LO, 22 kHz for the pump closer to the signal and 100 kHz for the other one. The measured BER as function of the received power and constellations diagrams (−26 dBm received power) areshown in Fig. 5. We first notice that BER below the FEC threshold is possible over a 32 nm bandwidth. However, there is a power penalty difference of 5.5 dB at the FEC threshold between the best case (nm) and worst case (nm) as the lower CE leads to a lower OSNR at the input of the coherent receiver for the latter case. Furthermore, the constellation diagram for nm [Fig. 5] shows more nonlinearities compared to nm. This may be explained by the fact that we operate at the very edge of the gain spectrum of our SOA. While dual-pump wavelength conversion of 16-QAM exhibits similar performance over the entire C-band, that is not the case for 64-QAM. As mentioned before, the optimal operating point of the SOA does not maximize the CE, but is rather a compromise between CE and NLPN in order to obtain the best BER. The non-optimal CE becomes a problem for large wavelength detuning as the receiver sensitivity decreases for 64-QAM compared to 16-QAM . Post-compensation for NLPN may be possible to overcome this situation . If so, the CE could be maximized with proper signal and pumps power. Finally, we performed wavelength conversion with a single pump configuration for an input OSNR of 36 dB and a degraded OSNR of 26 dB. The signal and pump power are respectively −12 dBm and 9 dBm with a resulting conversion efficiency of −3.5 dB. The measured BER as function of the received power, measured at the input of the pre-amplified receiver, is displayed in Fig. 6. For high OSNR (36 dB), the BER results show very good performance of 64-QAM wavelength conversion, with a power penalty at the FEC thresholdof 1.7 dB. However, for low OSNR (26 dB) the power penalty is doubled at 3.5 dB. The results show that the number of wavelength conversion stages in a wavelength routed network will be limited for 64-QAM compared to less spectrally dense modulation formats such as QPSK or 16-QAM . Besides the limited CE, nonlinear impairments such as gain compression and nonlinear phase noise also affect the performance, as visible in the constellation diagrams [Fig. 6].
We experimentally studied wavelength conversion of a 16 Gbaud 16-QAM signal using FWM in a SOA. Using a co-polarized dual-pump configuration, we performed successful wavelength conversion over the entire C-band with good performance in terms of BER penalty for all conjugate wavelengths over a wide range of input OSNR. More specifically, we measured a power penalty lower than 1.5 dB at the FEC threshold limit for input OSNR from 20 to 40 dB. These results demonstrate the robustness of SOA-based wavelength conversion to ASE noise accumulation.
We also experimentally studied for the first time wavelength conversion of 5 Gbaud 64-QAM signals based on FWM in a SOA. With a co-polarized dual-pump configuration, we also performed successful wavelength conversion with measured BER below the FEC threshold over the entire C-band. As expected, 64-QAM was found to be more sensitive to input OSNR but nonetheless low power penalty at the FEC threshold was observed for input OSNR degraded by 10 dB. Considering the ongoing developments of SOA technologies coupled to current research efforts in post digital signal processing, it is very likely that these wavelength conversion results will be further improved in the near future. Wavelength converters based on FWM in SOAs is therefore a promising technique for wavelength conversion for next generation wavelength routed optical networks using complex high-order modulation format such as 16-QAM and 64-QAM.
References and links
1. A. H. Gnauck, P. J. Winzer, S. Chandrasekhar, X. Liu, B. Zhu, and D. W. Peckham, “Spectrally efficient long-haul WDM transmission using 224-Gb/s polarization-multiplexed 16-QAM,” J. Lightwave Technol. 29(4), 373–377 (2011). [CrossRef]
2. C. Tremblay, A. Enriquez-Castillo, M. P. Belanger, and F. Gagnon, “Filterless WDM optical core networks based on coherent systems,” in Proceedings of International Conference on Transparent Optical Networks, Stockholm, Sweden, 2011, paper Tu.D1.4. [CrossRef]
3. S. R. Nuccio, Z. Bakhtiari, O. F. Yilmaz, and A. E. Willner, “λ-conversion of 160-Gbit/s PDM 16-QAM using a single periodically-poled lithium niobate waveguide,” in Optical Fiber Communication Conference, OSA Technical Digest (Optical Society of America 2011), paper OWG5. [CrossRef]
4. R. Elschner, T. Richter, M. Nölle, J. Hilt, and C. Schubert, “Parametric amplification of 28-GBd NRZ-16QAM signals,” in Optical Fiber Communication Conference, OSA Technical Digest (Optical Society of America 2011), paper OThC2.
5. A. H. Gnauck, E. Myslivets, M. Dinu, B. P. P. Kuo, P. Winzer, R. Jopson, N. Alic, A. Konczykowska, F. Jorge, J.-Y. Dupuy, and S. Radic, “All-optical tunable wavelength shifting of a 128-Gbit/s 64-QAM signal,” in European Conference and Exhibition on Optical Communication, OSA Technical Digest (Optical Society of America 2012), paper Th.2.F.2. [CrossRef]
6. A. Mecozzi, S. Scotti, A. D’Ottavi, E. Iannone, and P. Spano, “Four-wave mixing in traveling-wave semiconductor amplifiers,” J. Quantum Electron. 31(4), 689–699 (1995). [CrossRef]
7. I. Tomkos, I. Zacharopoulos, D. Syvridis, T. Sphicopoulos, C. Caroubalos, and E. Roditi, “Improved performance of a wavelength converter based on dual pump four-wave mixing in a bulk semiconductor optical amplifier,” Appl. Phys. Lett. 72(20), 2499–2501 (1998). [CrossRef]
8. B. Filion, A. T. Nguyen, S. Amiralizadeh, L. A. Rusch, and S. LaRochelle, “Wideband wavelength conversion of 16 Gbaud 16-QAM signals in a semiconductor optical amplifier,” in Optical Fiber Communication Conference, OSA Technical Digest (Optical Society of America 2013), paper OTh1C.5. [CrossRef]
9. J. Wang, A. Maitra, C. G. Poulton, W. Freude, and J. Leuthold, “Temporal dynamics of the alpha factor in semiconductor optical amplifiers,” J. Lightwave Technol. 25(3), 891–900 (2007). [CrossRef]
10. M. J. Connelly and C. L. Janer, “Modeling of semiconductor optical amplifier RIN and phase noise for optical PSK systems,” in Proceedings of Numerical Simulation of Optoelectronic Devices, Rome, Italy, 2011, pp. 95–96.
11. R. Elschner and K. Petermann, “Pump-induced nonlinear phase noise in wavelength converters based on four-wave mixing in SOAs,” in Proceedings of IEEE Lasers and Electro-Optics Society (Institute of Electrical and Electronics Engineers, Belek-Antalya, 2009), pp. 779–780. [CrossRef]
12. C. Porzi, A. Bogoni, and G. Contestabile, “Regeneration of DPSK signals in a saturated SOA,” IEEE Photon. Technol. Lett. 24(18), 1597–1599 (2012). [CrossRef]
13. X. Wei and L. Zhang, “Analysis of the phase noise in saturated SOAs for DPSK applications,” J. Quantum Electron. 41(4), 554–561 (2005). [CrossRef]
14. T. Akiyama, M. Sugawara, and Y. Arakawa, “Quantum-dot semiconductor optical amplifiers,” Proc. IEEE 95(9), 1757–1766 (2007). [CrossRef]
15. T. Akiyama, H. Kuwatsuka, N. Hatori, Y. Nakata, H. Ebe, and M. Sugawara, “Symmetric highly efficient (~0 dB) wavelength conversion based on four-wave mixing in quantum dot optical amplifiers,” IEEE Photon. Technol. Lett. 14(8), 1139–1141 (2002). [CrossRef]
16. G. Contestabile, Y. Yoshida, A. Maruta, and K. Kitayama, “Coherent wavelength conversion in a quantum dot SOA,” IEEE Photon. Technol. Lett. 25(9), 791–794 (2013). [CrossRef]
17. G. Contestabile, Y. Yoshida, A. Maruta, and K. Kitayama, “Ultra-broad band, low power, highly efficient coherent wavelength conversion in quantum dot SOA,” Opt. Express 20(25), 27902–27907 (2012). [CrossRef] [PubMed]
18. A. J. Zilkie, J. Meier, M. Mojahedi, A. S. Helmy, P. J. Poole, P. Barrios, D. Poitras, T. J. Rotter, C. Yang, A. Stintz, K. J. Malloy, P. W. E. Smith, and J. S. Aitchison, “Time-resolved linewidth enhancement factors in quantum dot and higher-dimensional semiconductor amplifiers operating at 1.55 µm,” J. Lightwave Technol. 26(11), 1498–1509 (2008). [CrossRef]
19. N. Yasuoka, K. Kawaguchi, H. Ebe, T. Akiyama, M. Ekawa, K. Morito, M. Sugawara, and Y. Arakawa, “Quantum-dot semiconductor optical amplifiers with polarization-independent gains in 1.5-µm wavelength bands,” IEEE Photon. Technol. Lett. 20(23), 1908–1910 (2008). [CrossRef]
20. M. Matsuura and N. Kishi, “High-speed wavelength conversion of RZ-DPSK signal using FWM in a quantum-dot SOA,” IEEE Photon. Technol. Lett. 23(10), 615–617 (2011). [CrossRef]
21. M. Selmi, Y. Jaouen, and P. Ciblat, “Accurate digital frequency offset estimator for coherent PolMux QAM transmission systems,” in European Conference and Exhibition on Optical Communication, Vienna, Austria, paper P3.08 (2009).
22. S. Zhang, C. Yu, P. Y. Kam, and J. Chen, “Parallel implementation of decision-aided maximum-likelihood phase estimation in coherent M-ary phase-shift keying systems,” IEEE Photon. Technol. Lett. 21(19), 1471–1473 (2009). [CrossRef]
23. J. G. Proakis and D. K. Manolakis, in Digital Signal Processing: Principles, Algorithms and Applications, (Prentice Hall, 1995).
24. W.-R. Peng, T. Tsuritani, and I. Morita, “Transmission of high-baud PDM-64QAM signals,” J. Lightwave Technol. 31(13), 2146–2162 (2013). [CrossRef]
25. R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photon. Technol. Lett. 24(1), 61–63 (2012). [CrossRef]
26. A. Ghazisaeidi and L. Rusch, “On the efficiency of digital back-propagation for mitigating SOA-induced nonlinear impairments,” J. Lightwave Technol. 29(21), 3331–3339 (2011). [CrossRef]