IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 19, NO. 10, MAY 15, 2007
783
All-Optical Asynchronous Serial-to-Parallel Converter Circuit for DPSK Optical Packets Nicola Calabretta, Marco Presi, Giampiero Contestabile, and Ernesto Ciaramella, Member, IEEE
Abstract—We propose a novel asynchronous all-optical circuit for extraction and serial-to-parallel conversion of label bits from differential phase-shift keying (DPSK) packets. The circuit requires only two optical switches regardless of the number of bits to be extracted and parallelized from the packet. Experimental evidence of practical use of the circuit to four bit labels at 10 Gb/s is provided. The circuit is scalable with the number of bits, operates at low input power, and is suitable for photonic integration. The asynchronous nature of the circuit allows us to efficiently extract/read one specific label field of variable length without processing the entire label, leading to a simplified architecture of the label processing circuit. Index Terms—Differential phase-shift keying, header/label processor, optical packet switching (OPS), self-synchronization, serial-to-parallel conversion (SPC), wavelength conversion.
I. INTRODUCTION
I
NTERNET-BASED data traffic generated by advanced telecommunication services and applications boosted the photonic community to develop high-capacity systems. Differential phase-shift-keying (DPSK) wavelength-divisionmultiplexing transmission systems, featuring higher robustness to transmission impairments, recorded capacity of tens of terabits per second [1]. Optical packet switched (OPS) node based on all-optical circuits, with higher speed operation and lower power consumption than electronic circuits, is viewed as a viable solution to route these high-speed data packets. In devising an OPS node, a key function is the all-optical label processing system (AOLS) that provides the forwarding information for routing the packets. Generally, the AOLS includes a label/payload separator (LPS), which extracts the label from the packet, and a serial-to-parallel conversion (SPC) circuit, which parallelizes the label bits. The parallelized bits can be then processed by fast complementary metal–oxide–semiconductor electronics [2]–[5] or by self-routing all-optical circuits [6]–[8]. Several solutions for the all-optically label extractor were presented in [5], [9]. Those solutions require quite involved schemes including nontrivial devices per packet clock-recovery and switches. Moreover, the solution presented in [9] does not simply allow for extracting only a well-defined field within the label bits. All-optical solutions for parallelization of the label
Manuscript received December 28, 2006; revised February 20, 2007. This work was supported in part by Ericsson under a grant. The authors are with Scuola Superiore Sant’Anna, 56124 Pisa, Italy (e-mail:
[email protected]). Color versions of one or more of the figures in this letter are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/LPT.2007.895895
bits were presented in [4], [5], and [10]. In [4] and [10], the schemes require an external synchronous sampling pulse. In [5], the system is based on splitting and delaying the input label copies ( number of bits in a label), which are then in sampled by a synchronous control pulse. However, the splitting losses and the required number of optical switches linearly increase with , which makes the system bulky and difficult to scale. Most of all, the SPC requires a complex preprocessing circuit to retrieve the synchronous sampling pulse. Moreover, some of these approaches do not operate in the -band, are not suitable for photonic integration, depend on the state of polarization, and require high pump power. Furthermore, none has been operated with DPSK packets. Here we demonstrate a novel all-optical circuit that simultaneously extracts and parallelizes the label bits from DPSK packets. The technique is based on the conversion of the serial bits to parallel bits at distinct wavelengths. The main advantages are the scalability with increasing and the asynchronous operation. Indeed, the all-optical circuit is realized by using only two optical switches (regardless of the number of bits), without the need for large splitting losses and additional switches. The intrinsic asynchronous operation of the circuit eases the OPS node architecture eliminating also the need of the LPS or preprocessing circuit providing the sampling pulse. Finally, the circuit is polarization-independent, has low power consumption, and could be made compact by photonic integration. II. PRINCIPLE OF OPERATION The all-optical circuit, shown in Fig. 1, consists of two all-optical blocks: a compact all-optical function (AOF) and an optical AND logic function. The AOF receives at its input continuous-wave (CW) lightwaves. the DPSK packets and It produces two outputs: the DPSK demodulated data packet (label “A” in Fig. 1) and a sequence of colored pulses, with repetition rate equal to the bit time of the label (label “B”). The demodulated label bits and the colored pulses are thus fed into the logic AND. The AND logic function between the two signal is used to transfer the pattern information contained in the label bits to the sequence of colored pulses. By using an arrayed waveguide grating (AWG), the resulting colored pulses are separated at different spatial outputs (label “C”), representing the parallelized copy of the label pattern. These parallel bits are thus ready to be processed by either electrical or all-optical means. The AOF circuit was realized using a linear optical amplifier (LOA) and a one-bit delay interferometer (DI), followed by a wavelength-dependent delay line (WDL) [7]. The incoming local CW DPSK packets are coupled by an AWG with
1041-1135/$25.00 © 2007 IEEE
784
IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 19, NO. 10, MAY 15, 2007
Fig. 1. Experimental setup and packet format employed to demonstrate the all-optical label processing system. PM: phase modulator.
lightwaves and fed into the optical circuit. The LOA combined with the DI acts as an optical switch [11], where the packet acts as the pump and the CW lightwaves as probes. In the LOA, the constant-envelope DPSK packet induces a cross-phase modulation on all the CW probes proportional to the packet intensity. Passing through the DI, the above phase modulation of the probes is converted into amplitude modulation. Indeed any significant phase variation at the input results in a pulse at colored pulses synchronous with the DI output. Therefore, the beginning of the packet and with the end of the packet are produced. Simultaneously, the DI demodulates the DPSK label colored packet. The demodulated DPSK packet and the pulses at the DI output are both fed into the WDL. The WDL was implemented by an array of fiber Bragg gratings (FBGs), probe wavelengths and placed each centered at one of the at multiple distances, so that the back-reflected colored pulses pulses with a pulse repetition equal to form a sequence of the bit rate of the label. This emerges at Port 3 of the optical circulator. The demodulated DPSK packet, being at a different wavelength, passes through the FBGs unaffected. As a result, the optical outputs of the AOF consist of the demodulated DPSK label (point “A”) and the sequence of -colored pulses (point “B”). Those outputs are then input to an optical logic AND, realized by means of a terahertz optical asymmetric demultiplexer (TOAD) [12]. The TOAD consists of an optical loop containing offset with respect to the loop center. an LOA placed with a The control pulse, coupled into the loop via a 90 : 10 coupler, . When an optical input pulse opens a switching window of enters the TOAD, the pulse power is split by a 50 : 50 coupler into a clockwise pulse (cw) and counter-clockwise pulse (ccw). If no control pulse is applied, the cw and ccw pulses propagating through the loop recombine in phase at the coupler and the resulting pulse is reflected back at the input port. If a control pulse enters the LOA after the cw pulse but before the ccw pulse, the ccw pulse experiences a different gain and refractive index induced by the control pulse. At the coupler, the output pulse is switched at the other TOAD port [12]. Note that the time required by the TOAD to switch the next pulse (switching repetition) is determined by the LOA recovery time [12]. The optical logic AND between the sequence of pulses (as TOAD input)
and the demodulated label bits (as control signal) is simply obtained at the output of the TOAD by filtering out the control signal. Note that the colored sequence contains only as many pulses as the label bits. Therefore, the AND operation between the sequence and the payload (as part of the control signal of the TOAD) does not produce any pulse. For the same reason, the -pulses sequence produced at the packet end falls within the packet’s guard time; therefore, no signal is ever produced at the output of the TOAD for these pulses. III. EXPERIMENT The experimental setup is shown in Fig. 1. A CW laser at 1538.9 nm was modulated by an intensity modulator acting as an optical gate to generate packets with 3.2-ns guard-time. A LiNbO phase modulator driven by a pattern generator at 10 Gb/s was used to produce sequences of three packets “ ”, with three different labels of four bits each ( “ ”, and “ ”, respectively). Note that the repetition rate of the label bits is of 500 ps to match the delays imposed by the available FBGs based WDL. Four CW nm to nm, spaced lasers (from by 0.8 nm) were coupled with the packet and fed into the first stage of the circuit. The total power of the four probes and the power of the DPSK packets at the LOA input were 4 and 5.8 dBm, respectively. The LOA had 14-dB small-signal gain and saturation output power of 10 dBm at 240 mA, 1.4 dB of polarization-dependent gain, and a gain recovery time of around 100 ps. The DI fiber bit-delay was 100 ps to demodulate the 10-Gb/s DPSK packets. The demodulated DPSK packets after the DI are shown in Fig. 2(a), where the can be seen. three different labels with pattern , , and The WDL consisting of four FBGs wrote along the same fiber ps and centered at the CWs wavelength introduced an ) wavelength-dependent delay, to convert the ( four parallel colored pulses to a sequence of four colored pulses. The four-pulse sequences generated at the beginning and at end of each packet (we recall that only the former sequence is relevant) are shown in Fig. 2(b). The measured extinction ratio (ER) of these pulses was around 14 dB. The demodulated DPSK bits were amplified by an erbium-doped fiber amplifier,
CALABRETTA et al.: ALL-OPTICAL ASYNCHRONOUS SPC CIRCUIT FOR DPSK OPTICAL PACKETS
785
is that it requires only two optical switches. Parallelization of a high number of bits just requires a higher number of CW lightwaves and FBGs with no substantial modification of the setup. Moreover, the use of only two switches hugely reduces the power consumption and the potential complexity for photonic integration. Note also that as the local CW lightwaves can be shared among the AOLS of the OPS node, only one single cost-effective CWs source is required. The operation was demonstrated with 10-Gb/s data packets. However, circuit operation at much higher bit rate is feasible as the technique is sensitive only to the (constant) envelope of the DPSK signal, which makes the function bit-rate transparent, and the TOAD can operate at very high bit rate [12]. Moreover, because the circuit is asynchronous, it allows us to efficiently extract/read one specific label field of variable length (i.e., source or destination address, or time-to-live field, etc.) without processing the entire label, leading to a cost reduction and simplified architecture of the label processing circuit. Finally, the circuit is polarization-independent and it could in principle work also with ON–OFF keying packets by adding an optical packet envelope detector preprocessor.
Fig. 2. Measured oscilloscope traces. (a) Demodulated input data packets after the first AOF (point “A”). The insets show details of the four labels. (b) The second output of the AOF representing the sequence of colored pulses (point “B”). (c)–(f) Output 1 at 1557.35 nm, Output 2 at 1558.13 nm, Output 3 at 1558.91 nm, and Output 4 at 1559.71 nm, respectively. The time scale is 5 ns/div. The voltage scale is 500 mV/div for (a) and (b), and 200 mV/div for (c)–(f).
filtered by a bandpass filter, and then fed into the TOAD via a 10 : 90 coupler. The four pulses were delayed by a fixed delay line to balance the longer path of the demodulated DPSK signal and then fed into the TOAD. The average optical power of the demodulated signal after the 10 : 90 coupler and the four pulses was 3.3 and 10 dBm, respectively. The LOA in the TOAD had similar characteristic of the other one. The driving current was 180 mA. The TOAD switching window was around 80 ps. The TOAD output was demultiplexed by the AWG, whose representing the parallel bits are reported outputs at in Fig. 2(c)–(f). Indeed, it can be clearly seen that for packets (“ ”) a pulse appears at each output, while for with ”), we get a pulse only at , , and packets with (“ , and finally for packets with (“ ”), a pulse appears only at and . The measured ER of the pulses was around 12 dB. Those results provide evidence that the serial label bits are converted to parallel output ports. IV. CONCLUSION We have demonstrated a novel asynchronous all-optical circuit that simultaneously extracts and parallelizes label bits of DPSK packets. The main advantage of the proposed technique
REFERENCES [1] A. H. Gnauck and P. J. Winzer, “Optical phase-shift-keyed transmission,” J. Lightw. Technol., vol. 23, no. 1, pp. 115–130, Jan. 2005. [2] D. Klonidis, C. T. Politi, R. Nejabati, M. J. O’Mahony, and D. Simeonidou, “OPSnet: Design and demonstration of an asynchronous highspeed optical packet switch,” J. Lightw. Technol., vol. 23, no. 10, pp. 372–381, Oct. 2004. [3] K. G. Vlachos et al., “STOLAS: Switching technologies for optically labeled signals,” Commun. Mag., vol. 41, pp. S9–S15, 2003. [4] K. Chan, F. Tong, C. K. Chan, L. K. Chen, and W. Hung, “An alloptical packet header recognition scheme for self-routing packet networks,” in OFC 2002, 2002, pp. 284–285. [5] R. Takahashi, T. Nakahara, H. Takenouchi, and H. Suzuki, “40-Gbit/s label recognition and 1 4 self-routing using self serial-to-parallel conversion,” IEEE Photon. Technol. Lett., vol. 16, no. 2, pp. 692–694, Feb. 2004. [6] N. Calabretta, A. D’Errico, G. Contestabile, and E. Ciaramella, “Alloptical label processing techniques for pure DPSK optical packets,” IEEE J. Sel. Topics Quantum Electron., vol. 12, no. 4, pp. 686–696, Jul./Aug. 2006. [7] N. Calabretta, M. Presi, G. Contestabile, and E. Ciaramella, “Compact header processing circuit for optical DPSK packets,” Electron. Lett., vol. 42, pp. 871–872, 2006. [8] J. Kurumida, H. Uenohara, and K. Kobayashi, “All-optical label recognition with SOA-MZI multistage switching scheme,” presented at the OFC 2006, Anaheim, CA, JThB51. [9] D. Apostolopoulos, D. Petrantonakis, O. Zouraraki, E. Kehayas, N. Pleros, and H. Avramopoulos, “All-optical label/payload separation at 40 Gb/s,” IEEE Photon. Technol. Lett., vol. 18, no. 19, pp. 2023–2026, Oct. 1, 2006. [10] K. Ema et al., “All-optical serial-to-parallel conversion of Tb/s signals using a four-wave-mixing process,” Opt. Quantum Electron., vol. 33, pp. 1077–1087, 2001. [11] J. L. Pleumeekers et al., “All-optical wavelength conversion and broadcasting to eight separate channels by a single semiconductor optical amplifier delay interferometer,” in OFC 2002, pp. 596–597. [12] J. P. Sokoloff, P. R. Prucnal, I. Glesk, and M. Kane, “A terahertz optical asymmetric demultiplexer (TOAD),” IEEE Photon. Technol. Lett., vol. 5, no. 7, pp. 787–790, Jul. 1993.
2
