Open Access
Add to BibSonomyAbstract
We demonstrate 2.56 Tbit/s/port dual-polarization DWDM/DQPSK variable-length optical packet (20 Gbit/s × 64 wavelengths × 2 polarizations) switching and buffering by using a 2×2 optical packet switch (OPS) system. The optical data plane of the OPS system was constructed of multi-connected electro-optical switches and fiber delay lines. The accumulated polarization dependent loss of each optical path in the data plane was less than 5 dB. This low-polarization-dependence OPS system enabled us to handle DWDM/DQPSK optical packets (1.28 Tbit/s/port) with time-varying polarization after transmission through 100 km fiber in the field.
©2011 Optical Society of America
1. Introduction
To meet the demand of growing internet traffic, improved link capacity and node throughput are required. Advanced multi-level modulation formats with high spectral efficiency, such as quadrature phase-shift keying (QPSK) and quadrature amplitude modulator (QAM), and their combinations with wavelength-division-multiplexing (WDM), orthogonal frequency division multiplexing (OFDM) and polarization-division multiplexing (PDM), are promising techniques to increase the link capacity [1]. Moreover, multi-core fiber transmission at over 100 Tbit/s and high-order mode multiplexing have also been demonstrated [2,3]. In contrast to such progress in improving the link capacity, the low throughput of network nodes may be a bottleneck in future networks. Additionally, increased power consumption in the network nodes is an important issue to be resolved. Optical packet switches (OPSs) can provide higher throughput and higher energy efficiency than conventional electronic packet switching [4–7]. Minimizing the number of optical-to-electronic and electronic-to-optical packet conversions in OPSs is effective in reducing power consumption. Additionally, such transparent systems can handle packets with different data rates and complicated formats without changing the OPS configuration. Using colored (WDM) optical packets with the differential QPSK (DQPSK) format, 1.28 Tbit/s/port optical packet switching and buffering has been demonstrated [8]. However, larger throughput is needed to show the energy efficiency advantages of OPSs compared with electronic packet switching.
Our OPS system has an all-optical data plane that is constructed of optical switches and fiber delay lines (FDLs). Previously, many optical switch technologies have been proposed for use in OPS systems, including arrayed waveguide (AWG) wavelength-selective switches [9,10], semiconductor optical amplifier (SOA) gate switches [11,12], and a phase shift switch using electro-optical (EO) effect or semiconductor carrier injection [13]. However, each has some disadvantages. The AWG based switch with lower power consumption is always used together with a complicated wavelength conversion process. The main issue with the SOA based switch is undesirable non-linear effects, such as four-wave mixing, particularly in the case of large channel numbers typical of dense WDM (DWDM) systems. The phase shift switch is free from such problems, however the strong polarization dependence is undesirable in a network node where input packets may arrive with varying polarization states. For these reasons, in this work we employed EO phase shift switches made of PLZT ((Pb,La)(Zr,La)O3) [14]. The PLZT switches are known as one of EO switches with low polarization dependence. However, the fabrication process is not mature, and it is still difficult to reduce insertion losses, including propagation loss and polarization dependent loss (PDL). A large insertion loss causes degradation of the signal-to-noise ratio in OPSs. In the case of multi-connected PLZT switches, such as those used in OPSs, the accumulated PDL, that is, power leakage to nonselected ports, is nonnegligible. As a result, it is still not easy to provide transparent paths for packets with various polarizations, such as packets in the PDM format or packets exhibiting time-varying polarization after transmission through a long fiber, due to the polarization-dependent characteristics of the switches.
Recently, insertion loss and PDL have been drastically improved by applying a buried waveguide and Mach–Zehnder interferometer [14] to the switch structure. The lower PDL of the improved PLZT switch can overcome the difficulties in handling polarized packets. In the study described here, first we characterized the PDL of multi-connected PLZT switches in OPS, and then we demonstrated dual-polarization optical packet switching and buffering. Also, we investigate handling of optical packets after transmission through a 100 km long fiber in the field, which showed time-varying polarization.
2. Polarization dependent loss of optical data path
Figure 1 shows a schematic diagram of our OPS system, which consists of an optical label processor and an electrical buffer manager in the control plane, and optical switches and optical buffers in the data plane (optical data path). In the case of 2×2 OPS, optical packets pass thorough two PLZT switches. One is used as a label switch in the optical switch and the other is used as an FDL selector in the buffer. In 1×2 and 1×4 switches, packets pass through two and three switching elements, respectively. Therefore, an optical packet passes through five switching elements in the OPS process. Note that each final-stage switching element can work as a gate switch not as a path-selection switch to reduce the crosstalk between the output ports. Therefore, we used 1×4 and 1×8 switches as 1×2 and 1×4 switches, respectively, in the case of a 2×2 OPS with three kinds of FDL buffers.
Fig. 1 Schematic diagram of 2×2 OPS system (2×1 buffer with three fiber delay lines). Each final-stage switching element in 1×2 and 1×4 switches can work as a gate switch to reduce the crosstalk between the output ports.
Figures 2(i) and 2(ii) show the wavelength (C-band) dependent PDLs of the 1×2 and 1×4 PLZT switches, respectively. In this PDL measurement, we used Jones-Matrix eigenanalysis and the Muller-Matrix Method, and the two results were almost the same. A PDL of less than 1 dB was obtained with the 1×2 switch. However, the PDL of the 1×4 switch was about 2 dB, because the switch had more switching elements than the 1×2 switch. Figure 2(iii) shows the PDL measured at each output port of the 1×4 switch when the 1×4 switch was connected behind the 1×2 switch. In this case, the optical packets pass through five switching elements. We found that the typical accumulated PDL was about 3 dB, and it was almost independent of the input wavelength. In comparison with a LiNbO3 (LN) switch with the same waveguide structure, the total PDL of the PLZT switches is much smaller, because the roughly estimated PDL (from the difference of the electrooptic coefficients r33/r13 (for TM/TE)) of a switching element in an LN switch is over 5 dB. A lower PDL of the PLZT switches can be obtained by optimizing the fabrication process; however, the total PDL increases with the number of switching elements that packets pass through. The scalability of OPSs in terms of the total loss and the PDL of switches will be an issue in the future.
Fig. 2 Polarization dependent losses (PDLs) of (i) 1×2, (ii) 1×4 (output port 1) and (iii) 1×2 + 1×4 PLZT switches (each output port), measured by a polarization analyzer using Jones and Muller matrix methods.
3. Dual-polarization optical packet switching and buffering experiment
We attempted dual-polarization optical packet switching and buffering using the 2×2 OPS system with low PDL described above. Figure 3 shows the experimental setup of a transmitter for dual-polarization, wide-colored, DQPSK packets. We used 64 arrayed distributed feedback lasers (1537.0–1562.2 nm, 50 GHz spacing) as payload light sources; their spectrum is shown in Fig. 4(a) . The continuous waves of the 32 odd and 32 even channels were collectively modulated by LN intensity modulators (LN-IM-1, −2), respectively. Then, LN-DQPSK modulators (DQPSK-1, −2) generated 20 Gbit/s NRZ-DQPSK payloads with 10 Gbit/s symbol-rate data (I1, Q1 for odd channel, I2, Q2 for even channel). As a result, 64-channel, 1.28 Tbit/s DWDM/NRZ-DQPSK payloads were generated. A set of eight payloads of differing lengths of 700–1500 bits (payload train) were divided, and one payload train was given a fiber delay of about 3080 ns, corresponding to the repetition time to generate a set of payloads, to suppress interference between payload trains after coupling at a polarization beam coupler (PBC). One payload train had a delay of several bits from the other; therefore, we regarded two payload trains with different polarizations as different data in this demonstration. As a result, the payload bit rate increased to 2.56 Tbit/s. For label generation, 10 GHz optical pulses generated by a mode-locked laser diode (MLLD, 1530 nm) were modulated according to the payload pattern by LN-IM-3. The pulses were input to a 200 Gchip/s multiple optical encoder (MOE) with an AWG configuration [15], and different phase-shift-keying (PSK) optical codes were generated as label-A and label-B. After the PBC, the payload was sandwiched by the PSK labels, as shown in Fig. 4(b), generating 2.56 Tbit/s dual-polarization optical packets. The packet train was divided to assume that each train was generated from different transmitters TX-1 and TX-2, and one packet train was given a delay of about 77 ns to bring about a situation where packets collide in the OPS, as shown in Figs. 4(c) and 4(d). The output polarization of each packet from TX-1 and TX-2 was adjusted by a polarizer set on each transmission line, and the angles of the polarizers were fixed during the demonstration.
Fig. 3 Experimental setup of transmitter for dual-polarization, wide-colored, DQPSK packet generation (20 Gbit/s DQPSK × 64 wavelengths (50 GHz spacing, C-band) × 2 polarizations).
This demonstration was designed to forward optical packets with only label-A input from TX-1 and TX-2 without packet collision. Figure 5 shows the setup of the OPS system. When multiple optical decoders (MODs) [15] recognized an input label as label-A, the switch controllers output gate signals to open and close 1×2 PLZT switches, so that packets with label-A were sent to the buffers (Figs. 5(i) and 5(ii)). An asynchronous buffer manager [16] received packet arrival information from the label switches. When packet collision was expected, the buffer manager controlled switches in the buffer so that each packet was sent to an appropriate FDL. We set automatic-level-control, transient-suppressed EDFAs (ALC-TS-EDFA) [17,18] in each FDL to improve signal degradation. A conventional EDFA is not suitable as an amplifier in the FDL, because amplification gain given to packets by the EDFA should be dependent on the packet duty cycle, which is generally variable in an FDL. In contrast, the ALC-TS-EDFA achieved packet-duty-cycle–independent amplification using electrical feedback circuits [19]. By using the ALC-TS-EDFAs in the FDLs, output power fluctuation of the packets, caused by differences in propagation loss and PDL between paths in the PLZT switches, became smaller compared with the case where an EDFA is placed after a 3×1 coupler, as in the previous setup [20]. Furthermore, we were able to achieve almost the same signal-to-noise ratio of packets after the OPS process as that reported in [20], even though the number of channels increased from 30 to 64. Figures 5(iii) and 5(iv) show the buffered packets in FDLs. Some packets were shifted by the FDLs, and all packets with label-A were output with no collisions, as shown in Fig. 5(v). Power fluctuation of 5 dB shown in the output spectrum of Fig. 5(v) was due to the PDL (i.e., the different polarization states between channels in the OPS) and the gain profile of the EDFAs in the C-band.
Fig. 5 Experimental setup of OPS system and packet receiver for dual-polarization DWDM/DQPSK optical packets. (i-v) Waveforms in the OPS. Right bottom graph shows an output spectrum after buffering.
After buffering, optical payloads with different polarizations were separated by a polarization beam splitter (PBS). Then, I and Q data of the DQPSK payload were demodulated by 1-bit delay interferometers and received by a balanced detector. The clock and data were recovered by an optical packet receiver, and the bit-error-rate was measured by a packet error detector, as shown in Fig. 6(a) . Clearly opened eye diagrams were obtained in all channels. Note that center lines in the diagrams, as shown in Fig. 6(b), are derived from the zero-power-level span from packet to packet. Also, our packet error detector can count errors except for such a span (only while a packet is being received). Figure 6(c) shows BERs of the demodulated I and Q data on two different polarizations for the back-to-back configuration (without the OPS) and after the OPS. Error-free (BER<10−9) operation for all payload channels was confirmed. The average power penalty was 4 dB. We considered that the distribution of BERs after the OPS was due to differences in the PDLs of the switches and/or the gains of the EDFAs in the OPS between channels.
Fig. 6 (a) Experimental setup of packet receiver for dual-polarization DWDM/DQPSK optical packets. (b) eye diagrams of 32 ch. (c) bit-error-rate of all payloads (back to back data were measured at channels 1, 16, 32, 48 and 64).
4. Field demonstration using JGN2plus optical test bed
Our success in handling dual-polarization optical packets with the low-polarization-dependence OPS system opens up the possibility of handling optical packets with time-varying polarization after transmission through a fiber in the field. Due to the larger accumulated PDL of the older installed PLZT switches, we could demonstrate only label switching by using a 1×2 PLZT switch [21]. We attempted both switching and buffering for DWDM/DQPSK optical packets after field fiber transmission for the first time.
In this demonstration, we used the fiber installed in the JGN2plus optical testbed [22] located between Koganei and Otemachi in Tokyo, as shown in Fig. 7(a) . One round-trip fiber line consists of 85 km of installed single-mode fiber and 15 km of dispersion compensating fiber, with a total loss of typically 38 dB. Chromatic dispersion was compensated at –20 to + 25 ps in the C-band. Differential group delay (DGD) of less than 2.5 ps (Fig. 7(b)) was sufficiently tolerable for our 10 Gb/s NRZ colored packets. Optical packet transmitters (TX-1, TX-2), a 2 × 2 OPS and a packet receiver (RX) were located in Koganei. Their setups were the same as those described in Section 3, except for the polarization MUX/DEMUX part (dashed-line square in Figs. 3 and 6(a)). The optical packet format was 64-channel DWDM and 20 Gbit/s NRZ-DQPSK (1.28 Tbit/s, without polarization multiplexing). Optical packets generated by transmitters were sent to Otemachi and looped back to Koganei via different fibers. In the OPS, the directions of optical packets with label-A or label-B were switched to port-A or port-B, respectively, without collisions between packets from TX-1 and TX-2, as shown in Fig. 7(c). In this demonstration, output packets from port-A were received by the RX, and their BERs were measured.
Fig. 7 (a) Schematic diagram of demonstration system. (b) DGD/PMD of 100-km field-installed fiber for 1 hour, in comparison with a bobbin fiber. (c) Image of packet flow in this demonstration.
5. Experimental results
Figure 8(a) shows a spectrum and waveforms of 64ch (50 GHz spacing in the C-band) colored optical packets of TX-1 and TX-2 after field transmission. A mode-locked laser diode with a center wavelength of 1530 nm was used as a light source for PSK code labeling. A peak power fluctuation of less than 3 dB was measured in the spectrum after the field transmission. The polarization of packets output from each transmitter was defined by a fixed polarizer; however, the packets arriving at the OPS had time-varying polarization. At the input of the OPS, we set a 77 ns delay time between the packets from TX-1 and TX-2 by using a FDL for bringing about a situation where packets collide in the OPS.
Fig. 8 (a) Spectrum of DWDM/DQPSK packets from TX-1 and TX-2 after field transmission (measured at inputs of OPS) and (b) after OPS process. (c) Bit error rate of optical packet switching and buffering for DWDM/DQPSK optical packets after transmission through 100 km-long field fiber. (Back-to-back data were measured at channels 16, 32 and 48.)
By the same OPS process as described in Section 3, we demonstrated label switching and buffering of optical packets with label A. Figure 8(b) shows the output spectrum after the OPS process. We observed a time-varying power fluctuation of about 7 dB, which was also caused by the time-varying polarization of packets in the field transmission. In the RX, DQPSK packets of each channel were demodulated, and the clock and data were recovered. BERs were measured by the error detector. Figure 8(c) shows BERs of the demodulated data after the OPS process and in the back-to-back configuration. Error free (BER<10−9) operation was obtained for all channels.
6. Conclusion
We demonstrated 2.56 Tbit/s/port dual-polarization, wide-colored DQPSK packet switching and buffering with an OPS system having an accumulated PDL of less than 5 dB for the first time. We were able to maintain a high signal-to-noise ratio of packets after the OPS process by using packet-duty-cycle–independent EDFAs in each FDL. Furthermore, we achieved both optical packet switching and buffering for 1.28 Tbit/s (64 ch×20 Gbit/s DQPSK) optical packets after transmission through a 100 km long field fiber for the first time.
Figure 9 shows the progress of our OPS system. Although the configuration of the OPS system was not changed, the throughput was increased merely by changing the packet format. We could keep the power consumption of the 2×2 OPS system at about 1 kW; therefore, the energy efficiency was less than 0.2 nJ/bit. Using other multiplexing formats and digital coherent processing for the optical packets will enable more energy-efficient optical packet switching.
References and links
1. D. Qian, M. F. Huang, E. Ip, Y. K. Huang, Y. Shao, J. Hu, and T. Wang, “101.7-Tb/s (370×294-Gb/s) PDM-128QAM-OFDM transmission over 3×55-km SSMF using pilot-based phase noise mitigation,” in Proceedings of Optical Fiber Communication Conference 2011 (2011), PDPB5.
2. J. Sakaguchi, Y. Awaji, N. Wada, A. Kanno, T. Kawanishi, T. Hayashi, T. Taru, T. Kobayashi, and M. Watanabe, “109-Tb/s (7×97×172-Gb/s SDM/WDM/PDM) QPSK transmission through 16.8-km homogeneous multi-core fiber,” in Proceedings of Optical Fiber Communication Conference 2011 (2011), PDPB6.
3. M. Salsi, C. Koebele, D. Sperti, P. Tran, P. Brindel, H. Mardoyan, S. Bigo, A. Boutin, F. Verluise, P. Sillard, M. Astruc, L. Provost, F. Cerou, and G. Charlet, “Transmission at 2×100Gb/s, over two modes of 40km-long prototype few-mode fiber, using LCOS based mode multiplexer and demultiplexer,” in Proceedings of Optical Fiber Communication Conference 2011 (2011), PDPB9.
4. D. J. Blumenthal, A. Carena, L. Rau, V. Curri, and S. Humphries, “All-optical label swapping with wavelength conversion for WDM-IP networks with subcarrier multiplexed addressing,” IEEE Photon. Technol. Lett. 11(11), 1497–1499 (1999). [CrossRef]
5. K. Kitayama and N. Wada, “Photonic IP routing,” IEEE Photon. Technol. Lett. 11(12), 1689–1691 (1999). [CrossRef]
6. H. J. S. Dorren, M. T. Hill, Y. Liu, N. Calabretta, A. Srivatsa, F. M. Huijskens, H. de Waardt, and G. D. Khoe, “Optical packet switching and buffering by using all-optical signal processing methods,” J. Lightwave Technol. 21(1), 2–12 (2003). [CrossRef]
7. S. J. B. Yoo, F. Xue, Y. Bansal, J. Taylor, Z. Pan, J. Cao, M. Jeon, T. Nady, G. Goncher, K. Boyer, K. Okamoto, S. Kamei, and V. Akella, “High-performance optical-label switching packet routers and smart edge routers for the next-generation internet,” IEEE J. Sel. Areas Commun. 21(7), 1041–1051 (2003). [CrossRef]
8. S. Shinada, H. Furukawa, M. Nakamura, and N. Wada, “Record switching throughput of 1.28-Tbit/s/port (64-wavelength×20-Gbit/s) by DWDM/NRZ-DQPSK optical packet switch system,” in Proceedings of 35th European Conference and Exhibition on Optical Communication (2009), PD3.1.
9. M. Funabashi, Z. Zhu, Z. Pan, and S. J. B. Yoo, “Packet-by-packet all-optical burst-mode 3R regeneration in an optical-label switching router,” in Proceedings of Optical Fiber Communication Conference 2006 (2006), OFJ1.
10. Y. Suzaki, R. Urata, T. Nakahara, H. Takenouchi, and R. Takahashi, “Hybrid optoelectronic router for optical packet switching,” in Proceedings of 15th OptoElectronics and Communications Conference (2010), 8D4–2.
11. N. Wada, H. Harai, and F. Kubota, “40 Gbit/s interface, optical code based photonic packet switch prototype,” in Proceedings of Optical Fiber Communication Conference 2003 (2003), FS7.
12. N. Calabretta and H. Dorren, “All-optical label processing in optical packet switched networks,” in Proceedings of Optical Fiber Communication Conference 2010 (2010), OThN6.
13. T. Tanemura, I. M. Soganci, T. Oyama, T. Ohyama, S. Mino, K. A. Williams, N. Calabretta, H. J. S. Dorren, and Y. Nakano, “Optical buffer based on monolithic InP phased-array 1×16 switch with silica-PLC pitch converter and ultra-compact coiled fiber delay lines,” in Proceedings of Optical Fiber Communication Conference 2010 (2010), PDPA5.
14. K. Nashimoto, D. Kudzuma, and H. Han, “High-speed switching and filtering using PLZT waveguide devices,” in Proceedings of 15th OptoElectronics and Communications Conference (2010), 8E1–1.
15. G. Cincotti, N. Wad, and K. Kitayama, “Characterization of a full encoder/decoder in the AWG configuration for code-based photonic routers - part I: modeling and design,” J. Lightwave Technol. 24(1), 103–112 (2006). [CrossRef]
16. S. Shinada, H. Furukawa, H. Harai, and N. Wada, “Variable-length DQPSK optical packet switching and buffering using asynchronous buffer manager,” in Proceedings of 36th European Conference and Exhibition on Optical Communication (2010), We.8.A.2.
17. Y. Oikawa, Y. Horiuchi, Y. Tanaka, Y. Shiga, M. Shiga, and N. Nagaeda, “Super-fast AGC-EDFA for the burst-mode systems without gain excursion in 20-ns and 21-dB ramped input,” in Proceedings of Optical Fiber Communication Conference 2008 (2008), JThA15.
18. Y. Awaji, H. Furukawa, N. Wada, P. Chan, and R. Man, “Mitigation of transient response of erbium-doped fiber amplifier for burst traffic of high speed optical packets,” in Proceedings of Conference on Lasers and Electro-optics 2007 (2007), JTuA133.
19. H. Furukawa, N. Wada, Y. Awaji, T. Miyazaki, H. Iiduka, N. Shiga, N. Sato, and H. Harai, “Optical packet and circuit simultaneous transmission technologies for dynamic lightpath setup/release and packet traffic change,” in Proceedings of Optical Fiber Communication Conference 2011 (2011), OMK2.
20. S. Shinada, H. Furukawa, and N. Wada, “Over 1-Tbit/s/port polarization-multiplexed DWDM/DQPSK optical packet switching and buffering,” in Proceedings of Optical Fiber Communication Conference 2011 (2011), OMW6.
21. N. Wada, N. Kataoka, T. Makino, N. Takezawa, K. Nashimoto, and T. Miyazaki, “Field demonstration of 1.28T bit/s/port, ultra-wide bandwidth colored optical packet switching with polarization independent high-speed switch and all-optical hierarchical label processing,” in Proceedings of 33th European Conference and Exhibition on Optical Communication (2007), PDS 3.1.
22. JGN2plus, http://www.jgn.nict.go.jp/.
