Open Access
Add to BibSonomyAbstract
We demonstrate error-free wavelength conversion of 28 GBaud 16-QAM single polarization (112 Gb/s) signals based on four-wave mixing in a dispersion engineered silicon nanowire (SNW). Wavelength conversion covering the entire C-band is achieved using a single pump. We characterize the performance of the wavelength converter subsystem through the electrical signal to noise ratio penalty as well as the bit error rate of the converted signal as a function of input signal power. Moreover, we evaluate the degradation of the optical signal to noise ratio due to wavelength conversion in the SNW.
© 2014 Optical Society of America
1. Introduction
Continuous increase in traffic demand places strain on existing optical systems and networks. Coherent systems, which enable higher-order and spectrally efficient modulation formats such as N-QAM (quadrature amplitude modulation), have eased the strain and increased channel capacity [1]. Additionally, processing functions such as routing and switching are becoming increasingly important, for example in data centers, and can be facilitated by wavelength conversion [2]. To avoid the complexity put upon transmitter and receiver electronics, all-optical wavelength conversion schemes have been researched [3]. One of the key characteristics of a reliable wavelength converter is transparency to bit rates and modulation formats; four-wave mixing (FWM)-based conversion is one of the most promising choices for its ultrafast response and transparency to amplitude and phase of optical signals.
FWM-based wavelength converters for coherent systems have been successfully demonstrated in semiconductor optical amplifiers (SOAs), highly nonlinear fiber (HNLF) and photonic crystal fiber. Full transparency of the wavelength conversion process to both rates and modulation formats was achieved in a quantum dot SOA for quadrature phase shift keying (QPSK), 8-phase shift keying (8-PSK), 16-QAM and orthogonal frequency-division multiplexing with 16 QAM sub-channels (OFDM-16QAM) [4, 5]. Wavelength conversion of QAM signals was also achieved in a multi-quantum well SOA (16 GBaud 16-QAM and 5 GBaud 64-QAM) [6] and HNLF (36-QAM 50 Gb/s and 64-QAM 60 Gb/s) [7]. Conversion of 544 Gb/s signals was demonstrated in HNLF for a dual-carrier polarization-division-multiplexed (PDM-16QAM) signal based on a co-polarized dual-pump scheme [8].
Silicon waveguides are starting to present a promising choice for replacing HNLF and SOAs. Due to a large refractive index contrast, light is tightly confined in the waveguides having a much smaller effective area than single mode fibers and resulting in higher optical power densities. This enhances the nonlinear effects, ultimately translating to shorter interaction lengths which allow compact devices for dense integration of crucial signal processing building blocks [9]. High-density integration was demonstrated in SOAs, however they often suffer from strong dependency on the conversion detuning and the asymmetry in down and up-conversion [5]; a phenomenon not observed with silicon waveguides. Thus far, for coherent systems, FWM-based wavelength conversion of D/QPSK (up to 80 Gb/s) signals [10] and RZ- DPSK signals (640 Gb/s OTDM) [11] in silicon have been demonstrated. In this paper, we demonstrate error-free wavelength conversion of 28 GBaud 16-QAM single polarization (112 Gb/s) signals using FWM in a dispersion-engineered silicon nanowire (SNW). This is the first demonstration of wavelength conversion of higher-order QAM signals (i.e., beyond QPSK) at data rates above 100 Gb/s in silicon.
2. Silicon nanowire design and FWM characterization
FWM is prevalent for wavelength conversion of phase modulated signals as it enables generation of an idler carrying the same information as the original signal while preserving the phase (although conjugated). FWM is most effective when the phase-matching condition is ensured [12]. As phase matching is affected by chromatic dispersion, it is desired to have the pump close to the zero dispersion wavelength in the anomalous group velocity dispersion (GVD) regime (i.e., around 1550 nm for fiber optic communications). Since material dispersion of silicon is large and normal, the waveguide structure and dimensions should be carefully designed to tailor the waveguide dispersion such that anomalous GVD can be achieved. Due to fabrication constraints with using the IMEC SOI fabrication technology [13], we consider SNWs based on strip waveguide structures with a height of 220 nm and a top oxide cladding see Fig. 1(a). We then vary the waveguide width (w) and calculate the corresponding effective index, group index and GVD using Lumerical MODE and MATLAB. Only the first TE mode is considered in our simulations. The results are shown in Fig. 1. Anomalous dispersion can be realized if the waveguide width is smaller than 540 nm and greater than 440 nm. It is worth noting that at a wavelength of 1550 nm, the waveguide exhibits a negative dispersion slope when w < 490 nm and a positive slope when w > 490 nm; moreover, the dispersion slope is almost 0 when w ~490 nm, which is extremely good for wideband FWM applications. So the optimal waveguide width is ~500 nm since it exhibits anomalous dispersion as well as a flat dispersion slope around 1550 nm.
Fig. 1 (a) Cross section of and (b) simulated dispersion curve for strip silicon waveguide structure.
The SNW used in our experiment is 20 mm in length and has a width of 500 nm. It exhibits a 22 dB fiber-to-fiber loss, primarily due to the vertical grating couplers used to couple light in and out of the waveguide, ~10 dB for both input and output (these couplers also ensure operation with TE-polarized light). The losses can be decreased by redesigning the grating couplers [14] or using an overlaying polymer waveguide and a tapered SNW waveguide design [15, 16]. The two-photon absorption (TPA) and free carrier effect induced propagation loss was measured to be negligible (i.e., linear absorption dominated) when the input optical power is less than 24.5 dBm. The SNW was characterized for FWM conversion efficiency, defined as the power ratio between generated idler and input signal (Fig. 2). The measured conversion efficiency varies between −24.7 dB to −17.8 dB across the 10 dB conversion bandwidth of up to 20 nm above the pump wavelength. These values are comparable to those reported in [10] for a 2.5 cm long silicon germanium waveguide. It is expected that the total conversion bandwidth is 40 nm (20 nm on each side of the pump), thus covering the entire C-band using a single pump.
Fig. 2 (a) Optical spectra of the wavelength conversion; (b) FWM conversion efficiency vs. wavelength offset (of idler wavelength with respect to pump wavelength).
3. Experimental setup
Fig. 3 depicts the experimental setup. A tunable external cavity laser (ECL) is modulated by a single polarization in phase quadrature dual-parallel Mach-Zehnder modulator (IQ MZM) driven by an arbitrary waveform generator (AWG) of bandwidth 18 GHz to obtain a 16-QAM repeated sequence of length 217 symbols at 28 GBaud. The signal is amplified with an erbium-doped fiber amplifier (EDFA1) and is coupled with an amplified spontaneous emission (ASE) source, followed by a polarization controller (PC). Variable optical attenuators (VOA) are used to control the power levels. A second continuous-wave ECL is used as the pump; it is coupled with the 16-QAM signal using a WDM coupler with the pass band from 1547.47 nm to 1552.77 nm, and the reflection bands from 1520 nm to 1546.37 nm and 1553.88 nm to 1570 nm. The signal and pump are then amplified (EDFA2) and coupled into the SNW. The converted signal is filtered using a tunable bandpass filter (BPF) with a bandwidth of 0.68 nm, amplified (EDFA3), and filtered again (bandwidth of 0.65 nm) before being coherently detected by a fully integrated silicon photonics coherent receiver (SiP CRx) [17] and real-time sampled. Offline digital signal processing (DSP) [18, 19] is applied to recover the signal constellation, electrical signal to noise ratio (SNR), and bit error rate (BER) calculations. The local oscillator of the CRx is an ECL of linewidth smaller than 100 kHz, tuned within ± 1 GHz of the frequency of the signal of interest.
Given the relatively high input and output coupling losses of the SNW, it is necessary to amplify the signal (using EDFA2) at its input in order to have sufficient power for FWM-based wavelength conversion. Moreover, it is necessary to amplify (with EDFA3) the output of the SNW to obtain the power level for proper detection by the SiP CRx. Thus, in our experiments, we consider the wavelength converter subsystem, to consist of EDFA2, the SNW, the 0.68 nm BPF, EDFA3, and the 0.65 nm BPF (shown in the dotted box in Fig. 3), and not just the SNW.
Unless otherwise stated, the BER, SNR, and OSNR measurements compare the performance of the wavelength converter subsystem to that of the back-to-back (i.e., from the point after the WDM coupler to point D shown in Fig. 3). Accordingly, the performance characterization not only reflects the FWM-based wavelength conversion process in the SNW, but also the performance/characteristics of both EDFA2 and EDFA3.
4. Results and discussion
The pump wavelength is 1545.5 nm and the corresponding power is 8.8 dBm measured at point A. Figure 4 shows the optical spectra of the converted signals, measured at point C, for input signal wavelengths of 1547.5 nm, 1548.5 nm, 1550.5 nm, and 1552.5 nm, all with a power of 4.0 dBm measured at point A. The total power into the SNW is kept below 24.5 dBm to ensure that linear absorption dominates. The conversion efficiency varies between −21 dB to −17.5 dB, in agreement with the values shown in Fig. 2. The inset on the right in Fig. 4 shows the constellation of the input data signal at 1548.5 nm measured at point A; the SNR is 21.91 dB after detection, providing error-free operation. The inset on the left shows the constellation of the converted signal at 1542.5 nm measured at point D. The SNR is 18.08 dB and the corresponding BER is 1.8 × 10−4. For the other converted wavelengths, their SNRs are between 17.64 dB and 18.08 dB and their BERs are between 1.8 × 10−4 and 3.2 × 10−4, all below the forward error correction (FEC) limit of 3.8 × 10−3 to obtain error-free operation. The typical measured SNR penalty of the wavelength converter subsystem is ~4 dB. By replacing the WDM coupler with a 50/50 power coupler, conversion over 40 nm can be achieved, as shown in Fig. 2.
Fig. 4 Optical spectra of pump, 16-QAM signals (at 1547.5 nm, 1548.5 nm, 1550.5 nm, and 1552.5 nm) and converted signals; inset shows the constellation of original and converted signals at 3 nm offset.
To further characterize the overall performance of the wavelength converter subsystem, we examine the minimum signal power (at point A) required for error-free conversion. Figure 5 shows the BER of the converted signal as a function of input signal power. The input signal wavelength is 1548.5 nm with an OSNR of 39.07 dB and SNR of 21.91 dB (all measured at point A). Moreover, EDFA2 is set to provide an output power of 24.3 dBm and the power received at the SiP CRx is fixed at −5 dBm. Error-free conversion (within the FEC limit) is possible until the signal input power drops below −4.5 dBm. As mentioned in Section 3, both EDFA2 and EDFA3 are required for our wavelength converter subsystem and as such, the results shown in Fig. 5 not only reflect the FWM-based wavelength conversion process but the performance of both EDFAs.
The previous results characterized the performance of the wavelength converter subsystem by comparing the SNR and/or BER with respect to the back-to-back configuration (i.e., between point A or after the WDM coupler and point D). We now complement these results and perform a separate set of experiments to characterize the FWM-based wavelength conversion process directly in the SNW by comparing the OSNR of the input data signal (at point B) with that of the output idler signal (at point C). For these measurements, we use an input signal wavelength of 1548.5 nm. For an input OSNR (at point B) of 38.14 dB, 35.76 dB, 32.93 dB, and 30.61 dB (all with signal powers of 4.0 dBm at point A), the OSNR of the converted signal (at point C) is 27.01 dB, 26.75 dB, 25.95 dB, and 25.29 dB, respectively. Figure 6 shows the corresponding optical spectra measured at point C.
Fig. 6 Optical spectra of pump, 16-QAM signals at 1548.5 nm (with OSNR of 38.14 dB, 35.76 dB, 32.93 dB, and 30.61 dB) and converted signals measured at point C.
Figure 7 shows the OSNR of the converted signal out of the SNW as a function of the OSNR of the input signal into the SNW. After linear regression, we can observe that an improvement of x dB in the input OSNR is not mapped to an equivalent improvement in the output OSNR, but rather an amount of only 0.24x dB. This means that an increased noise floor around the signal is not fully transferred by the FWM process. A detailed analysis of the noise transfer process (see e.g., [20]) is beyond the scope of this work and is the subject of future investigations. Ultimately, EDFA3, which is necessary to increase the power of the output idlers from the SNW for proper detection by the SiP CRx, further degrades the OSNR to ~20 dB at point D for all cases. The resulting BER varies between 1.4 × 10−3 and 2.4 × 10−3, still below the FEC limit. The resulting SNR penalty varies between 5.32 dB and 5.46 dB.
Fig. 7 OSNR of the wavelength converted signal measured at point C as a function of the OSNR of the input data signal measured at point B. The input signal wavelength is 1548.5 nm.
The results summarizing the performance of the wavelength converter subsystem as well as the wavelength conversion process in the SNW are summarized in Table 1.
Table 1. Summary of the Performance of the Wavelength Converter Subsystem as well as Wavelength Conversion. The Input Signal Wavelength is 1548.5 nm.
5. Summary and conclusion
In this paper, we have demonstrated error-free wavelength conversion of 28 GBaud 16-QAM single polarization (112 Gb/s) signals based on FWM in a dispersion-engineered SNW. This is the first demonstration of wavelength conversion of higher-order QAM signals (i.e., beyond QPSK) at data rates above 100 Gb/s in silicon. Wavelength conversion covering the entire C-band is achieved using a single pump. We characterize the performance of the wavelength converter subsystem through the SNR penalty and BER of the converted signal as a function of input signal power. We also evaluate the degradation of the OSNR due to the FWM-based wavelength conversion process in the SNW. The silicon-based approach provides a setup size reduction compared to HNLF-based approaches. It also does not suffer from the conversion detuning dependency and asymmetric down and up conversion that can be observed when using SOAs. The demonstrated properties and results display its potential for on-chip integration for future optical transmission system building blocks.
Acknowledgments
This work was supported by the NSERC NGON CREATE program, the NSERC SiEPIC CREATE program, the Passive Silicon Photonics Workshop, and the FRQNT Programme de recherche pour les enseignants de collège. We thank TeraXion for providing the silicon photonics coherent receiver, CMC Microsystems for the SNW fabrication assistance, Lumerical Solutions and Mentor Graphics for design software We also thank L. Chrostowski, and N. A. F. Jaeger (University of British Columbia) and D. Deptuck (CMC) for discussions and assistance in SNW characterization, as well as Lucas A. Crea and Philip Roberge for their contributions.
References and links
1. A. Gnauck, P. Winzer, S. Chandrasekhar, X. Liu, B. Zhu, and D. Peckham, “Spectrally efficient long-haul WDM transmission using 224-Gb/s polarization-multiplexed 16-QAM,” J. Lightwave Technol. 29(4), 373–377 (2011).
2. X. Ye, Y. Yin, S. B. Yoo, P. Mejia, R. Proietti, and V. Akella, “DOS: A scalable optical switch for datacenters,” in Proc. of the 6th ACM/IEEE Symposium on Architectures for Networking and Communications Systems, 2010, p. 24. [CrossRef]
3. S. B. Yoo, “Wavelength conversion technologies for WDM network applications,” J. Lightwave Technol. 14, 955–966 (1996).
4. 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]
5. 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]
6. B. Filion, W. C. Ng, A. T. Nguyen, L. A. Rusch, and S. Larochelle, “Wideband wavelength conversion of 16 Gbaud 16-QAM and 5 Gbaud 64-QAM signals in a semiconductor optical amplifier,” Opt. Express 21(17), 19825–19833 (2013). [CrossRef] [PubMed]
7. S. T. Lu, G. Wei, and T. Kawanishi, “Experimental demonstration of wavelength conversions of optical 36 QAM and 64 QAM through FWM in HNLF,” Proc. IEEE Photonics Conference, Bellevue, WA, USA, September 2013, p.ThG1.4.
8. X. Li, J. Yu, Z. Dong, and N. Chi, “Wavelength conversion of 544-Gbit/s dual-carrier PDM-16QAM signal based on the co-polarized dual-pump scheme,” Opt. Express 20(19), 21324–21330 (2012). [CrossRef] [PubMed]
9. J. Leuthold, C. Koos, and W. Freude, “Nonlinear silicon photonics,” Nat. Photonics 4(8), 535–544 (2010). [CrossRef]
10. M. A. Ettabib, K. Hammani, F. Parmigiani, L. Jones, A. Kapsalis, A. Bogris, D. Syvridis, M. Brun, P. Labeye, S. Nicoletti, and P. Petropoulos, “FWM-based wavelength conversion of 40 Gbaud PSK signals in a silicon germanium waveguide,” Opt. Express 21(14), 16683–16689 (2013). [CrossRef] [PubMed]
11. H. Hu, H. Ji, M. Galili, M. Pu, C. Peucheret, H. C. H. Mulvad, K. Yvind, J. M. Hvam, P. Jeppesen, and L. K. Oxenløwe, “Ultra-high-speed wavelength conversion in a silicon photonic chip,” Opt. Express 19(21), 19886–19894 (2011). [CrossRef] [PubMed]
12. H.-S. Jeong, D. W. Kim, K. H. Kim, and J.-M. Lee, “All-optical signal-conversion efficiency with a parameter-dependent four-wave-mixing process in a silicon nanowaveguide,” J. Korean Phys. Soc. 62(3), 428–434 (2013). [CrossRef]
13. Silicon Photonics Platform - ePIXfab: http://www.epixfab.eu (2013).
14. IMEC News Release, http://www2.imec.be/be_en/press/imec-news/photonics2013.html (2013).
15. L. K. Oxenløwe, H. Ji, M. Galili, M. Pu, H. Hu, H. C. H. Mulvad, K. Yvind, J. M. Hvam, A. Clausen, and P. Jeppesen, “Silicon photonics for signal processing of Tbit/s serial data signals,” IEEE J. Sel. Top. Quantum Electron. 18(2), 996–1005 (2012). [CrossRef]
16. M. Pu, L. Liu, H. Ou, K. Yvind, and J. M. Hvam, “Ultra-low-loss inverted taper coupler for silicon-on-insulator ridge waveguide,” Opt. Commun. 283(19), 3678–3682 (2010). [CrossRef]
17. M. Morsy-Osman, M. Chagnon, X. Xu, Q. Zhuge, M. Poulin, Y. Painchaud, M. Pelletier, C. Paquet, and D. V. Plant, “Colorless and preamplifierless reception using an integrated Si-photonic coherent receiver,” IEEE Photon. Technol. Lett. 25(11), 1027–1030 (2013). [CrossRef]
18. S. J. Savory, “Digital filters for coherent optical receivers,” Opt. Express 16(2), 804–817 (2008). [CrossRef] [PubMed]
19. M. Kuschnerov, F. N. Hauske, K. Piyawanno, B. Spinnler, M. S. Alfiad, A. Napoli, and B. Lankl, “DSP for coherent single-carrier receivers,” J. Lightwave Technol. 27, 3614–3622 (2009).
20. N. Cheng and J. C. Cartledge, “Noise transfer characteristics of cross-absorption modulation in an electroabsorption modulator,” IEEE Photon. Technol. Lett. 17(4), 780–782 (2005). [CrossRef]
