1. Introduction
All-optical signal processing technology (AOSPT) is expected to improve the efficiency, flexibility, and capacity of optical networks. It is also expected to be useful in reducing the cost, size, and power consumption of optical components used in these networks, primarily because the conversion between optical and electrical signals can be avoided. For these reasons, this technology is being actively pursued not only in the ultrafast regime, but also at speeds of a few tens of Gb/s.

One of the key devices for AOSPT is an all-optical switch. This article reports ultrafast all-optical switches and all-optical signal processing experiments.

2. Symmetric Mach-Zehnder-type all-optical switches

Fig. 1. Schematic illustration of a Symmetric-Mach-Zehnder all-optical switch. The differential phase modulation scheme is shown in the inset.

All-optical switches must satisfy various requirements simultaneously in order to make AOSPT practical and competitive. One is that the control-light-pulse energy must be much less than 1 pJ, otherwise the control light power is too high even at a few tens of Gb/s. To satisfy this requirement, semiconductor optical amplifiers (SOAs) are used for nonlinear elements to enhance the bandfilling nonlinearity in semiconductors by stimulated emission. This enables true femtojoule switching, but the relaxation time of SOAs limits the operating speed to less than a few tens of Gb/s. A promising way to mitigate this problem is to use a differential phase modulation (DPM) scheme. A representative DPM device is the Symmetric Mach-Zehnder (SMZ, sometimes called an SOA-MZI) all-optical switch [1]. In SMZs, the difference in nonlinear phase shifts (fNL) optically induced in the two nonlinear elements defines the state of the switch, so the effect of slow relaxation is circumvented, as illustrated in Fig.1. The DPM scheme also realizes a rectangle-like switching window, which is highly desirable in many applications.

There are two variants of the SMZ switch: a polarization-discriminating SMZ [2,3] (PD-SMZ, often called UNI) and a delayed-interference signal-wavelength converter (DISC) [4]. These devices can be configured in inherently stable forms suitable for bulk optics implementation. The original SMZ switch on the other hand can be unstable unless integrated. Hybrid-integrated (HI-) [5] and monolithic [6,7] SMZ-type switches have been reported.

3. Ultrafast all-optical signal processing experiments
Transmission experiments faster than 100 Gb/s per wavelength have been reported [8,9]. In these experiments, all-optical switches were used as demultiplexers to achieve very high bit rates. For bit-rates faster than 200 Gb/s, however, all-optical switches based on the Kerr effect in optical fibers have been used, but compact and less-power-hungry semiconductor devices are desirable. Recently, we have achieved error-free demultiplexing from 336 to 10.5 Gb/s (multiples of 82 MHz for synchronization with a streak camera etc.) using a HI-SMZ switch [10].

Fig.2. 336 Gb/s demultiplexing with HI-SMZ switch. (a) Cross-correlation trace of input signal pulses. (b) Results of BER measurements.

In the experiment, 336-Gb/s signal pulses (1561 nm) and 10.5-GHz control pulses (1546 nm) were input into the HI-SMZ switch. The durations of the control and signal pulses are important parameters. Our analysis of the crosstalk between the demultiplexed channel and adjacent channels indicates that a 1.5-ps control pulse can open a 3.0-ps switching window for 336-Gb/s demultiplexing. It is not necessary to use an extremely short control pulse, which is also advantageous in reducing the effect of carrier heating in SOAs, which causes degradation of the extinction ratio [11]. On the other hand, the signal pulse-width was set to 0.7 ps, because the crosstalk is more sensitive to it. Fig. 2(a) shows the waveform of the input signal pulses measured by a cross-correlation technique. As shown in Fig. 2(b), measured bit error rates (BERs) of the demultiplexed signal pulses reached an error-free level where BERs were less than 10^-9.

In the above demultiplexing experiment, an all-optical switch was driven by a regular-pulse-train at an electronic bit-rate. However, other applications such as wavelength conversion and 3R regeneration require all-optical switches to operate randomly at a higher transmission bit-rate, which is a tougher situation for optical switches. Here, we describe 3R operation at 84 Gb/s and wavelength conversion at 168 Gb/s.

Fig. 4. 168-Gb/s wavelength conversion with DISC. (a) Eye-diagram of wavelength-converted pulses after demultiplexing. (b) Results of BER measurement.

The “3R” regeneration refers to re-timing, re-shaping, and re-amplification of signal pulses that are distorted after transmission. In the experiment [12], a PD-SMZ switch was used as a pulse regenerator. Signal pulses carrying 84-Gb/s data (1560 nm, 2.1 ps) drove the PD-SMZ to encode 84-GHz clock pulses (1547 nm, 2.8 ps) so that the 84-Gb/s data was copied onto the clean clock pulses at 1547 nm. The pattern effect of the switch was suppressed by setting the average power of the clock pulses higher [13]. The average powers of the signal and clock pulses at the input of the SOA module in the PD-SMZ switch were -1 and +5 dBm, respectively. Fig. 3(a) shows the measured BERs after 84-Gb/s pulse regeneration by the PD-SMZ switch and demultiplexing back to 10.5 Gb/s by the HI-SMZ switch, which coincided with the curve for the case without regeneration. This shows that the power penalty due to the regeneration was negligibly small. We also evaluated the retiming capability. Fig. 3(b) shows the variation in BER caused by intentionally displacing the timing between the signal and clock pulses, while other parameters were kept unchanged. Within the range of 2.3 ps, the degradation of the BER was below 1 digit. This characteristic stems from the nearly rectangular switching window of the SMZ-type switch [1].

In the wavelength conversion experiment [14], 168-Gb/s signal pulses (1564 nm, 1.8 ps) drove a DISC to modulate a CW light (1547 nm). The average powers of the signal pulses and the CW light at the input of the SOA module were +10 and +16 dBm, respectively. The DISC generated 168-Gb/s, 2.0-ps pulses at 1547 nm. As shown in Fig. 4(a), a clear eye opening was obtained in the eye-diagram of the signal pulses after demultiplexing to 10.5 Gb/s. Fig. 4(b) shows the results of BER measurement, indicating error-free operation. The results confirm that the SMZ-type switches can be driven by data-modulated 160-Gb/s pulses.

4. Conclusion
We have demonstrated various types of ultra high-speed optical signal processing with Symmetric-Mach-Zehnder-type all-optical switches: error-free 336-Gb/s demultiplexing, penalty- and error-free 84-Gb/s pulse regeneration, and error-free 168-Gb/s wavelength conversion. These are the fastest experiments reported to date in each category. This work was performed partially under the management of the FESTA supported by the NEDO.

References
1. K. Tajima, Jpn. J. Appl. Phys. 32, L1746 (1993).

2. K.Tajima, S. Nakamura, and Y. Sugimoto, Appl. Phys. Lett. 67, 3709 (1995).

3. S. Nakamura, Y. Ueno, and K. Tajima, IEEE Photon. Technol. Lett. 10, 1575 (1998).

4. Y. Ueno, S. Nakamura, K. Tajima, and S. Kitamura, IEEE Photonics Technol. Lett. 10, 346 (1998).

5. K. Tajima, S. Nakamura, Y. Ueno, J. Sasaki, T. Sugimoto, T. Kato, T. Shimoda, M. Itoh, H. Hatakeyama, T. Tamanuki, and T. Sasaki, IEE Electron. Lett. 35, 2030 (1999).

6. R. Hess, M. Caraccia-Gross, W. Vogt, E. Gamper, P. A. Besse, M. Duelk, E. Gini, H. Melchior, B. Mikkelsen, M. Vaa, K. S. Jepsen, K. E. Stubkjaer, and S. Bouchoule, IEEE Photon. Technol. Lett. 10, 165 (1998).

7. J. Leuthold, C. H. Joyner, B. Mikkelsen, G. Raybon, J. L. Pleumeekers, B. I. Miller, K. Dreyer, and C. A. Burrus, IEE Electron. Lett. 36, 1129 (2000).

8. M. Nakazawa, ECOC 2001, Tu.L.2.3, Amsterdam 2001.

9. U. Fieste, R. Ludwig, C. Schubert, J. Berger, S. Diez, C. Schmidt, H. G. Weber, B. Schmauss, A. Munk, B. Buchold, D. Briggmann, F. Kueppers, and F. Rumpf, IEE Electron. Lett. 37, 443 (2001).

10. S. Nakamura, Y. Ueno, and K. Tajima, OFC 2002, FD3(PD), Anaheim 2002.

11. S. Nakamura, Y. Ueno, and K. Tajima, Appl. Phys. Lett. 78, 3929 (2001).

12. Y. Ueno, S. Nakamura, and K. Tajima, IEEE Photonics Technol. Lett. 13, 469 (2001).

13. R. J. Manning, A. D. Ellis, A. J. Poustie, and K. J. Blow, J. Opt. Soc. Am. B14, 3204 (1997).

14. S. Nakamura, Y. Ueno, and K. Tajima, IEEE Photon. Technol. Lett. 13, 1091 (2001).