In the backbone of today’s high performance networks, optical fibers provide enormous point-to-point communications capacity. With the deployment of DWDM equipment, aggregate throughputs on the order of a few Tbps per fiber are being achieved [1]. However, despite the recent success of fiber optics, it has so far been used primarily as a low loss, high bandwidth replacement to electrical cable in point-to-point transmission links. In these systems, optical signals are usually converted to the electrical domain at intermediate nodes in order to perform switching and signal processing. For example, in the Internet, electronic switches are used to route packets to their destinations. However, in this approach, the maximum serial line rate is limited by the bandwidth of electronics, which is considerably less than the bandwidth available in optical fiber. In effect, an “electronic bottleneck” is created in the system. This article summarizes the research efforts at Princeton University towards the development of network nodes capable of all-optical signal processing and routing.

Fig. 1: The architecture of a generic node in a transmission network with the various signal processing capabilities

Optical time division multiplexing (OTDM) based techniques have demonstrated the capability to provide routing, switching and processing at very high speeds. In OTDM, the electronic baseband signal is modulated on narrow pulses that occupy a small fraction of the bit period. The optical pulses from different baseband channels are multiplexed into different timeslots to create an aggregate line rate equal to the number of channels multiplied by the rate of the baseband electronic channel. Since OTDM assigns different temporal positions within the bit period to different channels, the processing latency of OTDM based techniques is determined by the timeslot access time. Highly scalable architectures of timeslot channel selectors have been demonstrated with timeslot channel access latency of a few bit periods [2,3].

Future optical networks may exploit the benefits of both OTDM and WDM technology. Fig. 1 shows the architecture of an OTDM-WDM hybrid network node capable of signal processing. The node includes several all-optical functions: routing, 3R regeneration (re-amplifying, re-shaping and re-timing), and wavelength and format conversions. Regeneration is done to compensate for the degradation of the signal quality during transmission. Wavelength conversion is useful in avoiding wavelength blocking in transmission systems, thus greatly increasing the flexibility and the capacity of the system. Format conversion is needed as an interface technology between network layers that use different formats.

Fig. 2: The structure of the Terahertz Optical Asymmetric Demultiplexer (TOAD)

All-optical switches in an OTDM-WDM framework present a versatile approach to all-optical processing since a variety of applications can be performed without significantly changing the device architecture. Various approaches have been proposed for the development of such switches. In general, all these approaches involve the utilization of a nonlinearity in various media like optical fiber, passive waveguides and active semiconductor optical amplifiers (SOA). Although passive all-optical switches have demonstrated the fastest switching capabilities to date, they typically require higher switching energies than the active switches based on the nonlinearity of the SOA. Also, ultrafast all-optical SOA-based switches are very compact and offer the possibility for monolithic integration together with other photonic devices. These SOA-based all-optical switches were initially limited by the slow recovery rate associated with the active nonlinearity in the SOA. Research at Princeton University has led to the development of the Terahertz Optical Asymmetric Demultiplexer (TOAD) [4], based on a Sagnac interferometric structure (see Fig.2), that allows signal processing functions at rates faster than the SOA recovery time by utilizing a differential onset of the nonlinearity in the SOA. It can be viewed as a generic 3-port device, with an input, control, and output port.

The TOAD is essentially a fiber loop joined at the base by an optical coupler, which splits an input signal into two equal parts that counter-propagate around the loop and recombine at the coupler. In addition, there is a coupler on the loop for the injection of a control signal, and an SOA in the loop to produce a large phase shift in the data signal.

In the absence of a control signal, the device is off—the two parts of the input pulse see the same medium as they go around the loop, arrive at the coupler in phase, and return to the input fiber. To turn the switch on, a control pulse is inserted that depletes the gain of the SOA and, according to the Kramers–Kronig relations, changes the index of refraction. By carefully timing the control pulse, we can induce a phase difference between the two counter-propagating input pulses, which pass through the SOA at different times. If the phase shift is properly adjusted, the two parts recombine at the input coupler in such a way that the whole signal passes to the output fiber. The control signal is eliminated at the output by a polarization or wavelength filter.

Because the SOA recovers slowly from the blast of the control pulse, input pulses that enter the switch immediately after the control pulse see the SOA in approximately the same recovery state, and are therefore affected similarly. The temporal duration of the switching window t_win can be adjusted by moving the location of the SOA, according to the relation t_win = 2Dx_SOA/c_fiber, where DxSOA is the distance of the SOA from the center position of the loop and cfiber is the speed of light in the fiber. Switching windows with a temporal width approaching 1 ps have been achieved with low control pulse energies (less than 500 fJ).

The same principle of the differential onset of the fast nonlinearity in an SOA has been used to build other all-optical devices in the Mach-Zender [5] and Michelson interferometric configurations. An integrated version of a Mach-Zehnder all-optical switch using an asymmetric twin-waveguide structure was demonstrated at Princeton University [6]. This structure couples the active and passive optical waveguides by growing the active layer on top of the passive guide, thus reducing the optical coupling loss between the active and passive waveguides. This also results in improved yield since device growth can be performed in a few steps and does not require re-growth.

As mentioned before, the temporal length of the switching window is determined by the offset of the SOA in the loop. The position of the SOA in the loop and the configuration of the TOAD can be varied to obtain the desired optical processing functionality. For example, in high-speed demultiplexing applications, a short switching window is used. The aggregate data stream is injected into the input port and the clock pulses into the control port [7]. However, for 3R regeneration, a large switching window is used with the data stream injected into the control port and the local clock pulses into the input port [8]. Short switching windows have also been used in high-speed analog sampling [9], whereas longer switching windows have been used for wavelength and format conversion [10] and all-optical packet routing [11]. The TOAD has been employed in network demonstrations for applications in photonic packet switching [12] and high speed interconnects [13].

In summary, the TOAD has demonstrated the versatility to perform many processing functions all-optically. Due to the low control energy requirements and potential for integration of this device, it is expected to present a viable approach to all-optical signal processing in both transmission systems and local area networks.

References
1. A. K. Srivastava, S. Radic, C. Wolf, J. C. Centanni, J.W. Sulhoff, K. Kantor and Y. Sun, “Ultra-dense terabit capacity WDM transmission in L-band”, IEEE Photonics Technology Letters 12, 1570 (2000).

2. K.-L Deng, K.I. Kang, I. Glesk, P.R. Prucnal, “A 1024-channel fast tunable delay line for ultrafast all-optical TDM networks,” IEEE Photonics Technology Letters 9, 1496 (1997).

3. B. C. Wang, I. Glesk, R. J. Runser, and P. R. Prucnal, “Fast Tunable Parallel Optical Delay Line,” Optics Express 8, 559 (2001).

4. J. P. Sokoloff, P. R. Prucnal, I. Glesk, M. Kane, “A Terahertz optical asymmetric demultiplexer,” IEEE Photonics Technology Letters 5, #7, pp 787-790 (1993)

5. K.I. Kang, I. Glesk, T.G. Chang, P. R. Prucnal, and R. K. Boncek, “Demonstration of all optical Mach-Zehnder demultiplexer,” Electronics Letters 31, 749 (1995).

6. P. V. Studenkov, M. R. Gokhale, J. Wei, W. Lin, I. Glesk, P. R. Prucnal and S. R. Forrest, “Monolithic, all-optical Mach-Zehnder demultiplexer using an asymmetric twin-waveguide structure,” IEEE Photonics Technology Letters 13, 600 (2001).

7. I. Glesk, J. P. Sokoloff, and P. R. Prucnal, “Demonstration of all-optical demultiplexing of TDM data at 250 Gbps,” Electronics Letters 30, 339 (1994).

8. Bing C. Wang, Lei Xu, Varghese Baby, Deyu Zhou, Robert J. Runser, Ivan Glesk, Paul R. Prucnal “Experimental Study on the Regeneration Capability of the Terahertz Optical Asymmetric Demutliplexer” Optics Communications 199, 83 (2001).

9. K. -L. Deng, R. Runser, I. Glesk, and P. R. Prucnal, “Single-shot optical sampling oscilloscope for ultrafast optical waveforms”, IEEE Photonics Technology Letters 10, 397 (1998).

10. R. J. Runser, D. Zhou, C. Coldwell, B. C. Wang, P. Toliver, K. -L Deng, I. Glesk, and P.R. Prucnal, “Interferometric ultrafast SOA-based optical switches: From devices to applications.” Optical and Quantum Electronics 33, 841 (2001).

11. I. Glesk, K. I. Kang, and P. R. Prucnal, “Demonstration of ultrafast all-optical packet routing,” Electronics Letters 33, 794 (1997).

12. P. Toliver, I. Glesk, R. J. Runser, K. -L. Deng, B. Y. Yu, and P. R. Prucnal, “Routing of 100 Gb/s words in a packet-switched optical networking demonstration (POND) Node,” IEEE Journal of Lightwave Technology 16, 2169 (1998).

13. K. -L. Deng, R. J. Runser, P. Toliver, I. Glesk, and P. R. Prucnal, “A highly scalable, rapidly-reconfigurable, multicasting-capable, 100-Gbit/s photonic switched interconnect based upon OTDM technology,” IEEE Journal of Lightwave Technology 18, 1892 (2000).