OPTICAL code-division multiple access (CDMA) [1] is

396 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 45, NO. 4, APRIL 2009 Self-Clocked All-Optical Add/Drop Multiplexer for Asynchronous CDMA Ring Networks Konstantin Kravtsov, Yanhua Deng, Student Member, IEEE, and Paul R. Prucnal, Fellow, IEEE Abstract In this paper, we demonstrate a novel self-clocked code-drop unit for incoherent optical code-division multiple-access (CDMA) networks. The unit is based on an all-optical thresholder with internal wavelength conversion that creates a control signal for the drop gate from the incoming data stream. This architecture does not require an external clock signal for drop operation and can be used in asynchronous ring networks. The proposed unit is experimentally demonstrated for two-dimensional time-wavelength optical CDMA codes at a bit rate of 2.5 Gbit/s with two different types of drop gates: terahertz optical asymmetric multiplexer and nonlinear fiber-based loop mirror. Error-free operation is achieved in both configurations. The development of a self-clocked add/drop multiplexer demonstrates a novel concept of asynchronous node operation in multiple access networks. Index Terms Add/drop multiplexer, code-division multiple access (CDMA), optical fiber communications, self-clocked operation. I. INTRODUCTION OPTICAL code-division multiple access (CDMA) [1] is a promising and quickly developing technique for localand metropolitan-area access networks. Broadband CDMA communication schemes were originally developed for wireless communication and proved to be more spectrally efficient, secure, and robust in the presence of multipath interference in metro environments compared with frequency-division multiple access networks. Similarly, optical CDMA systems use spread-spectrum codes in the optical domain, leading to improved network scalability and flexibility, as well as reliability and data security compared with conventional WDM networks [2], [3]. Together with the choice of multiplexing technique, network topology plays an important role at the stage of network design. So far, most efforts in developing optical CDMA networks were focused on star networks [4] [7]. However, it was recently demonstrated [8] that a ring topology is more efficient in terms of survivability and number of codes used. Multifiber rings [9] use a redundant number of links to achieve self-healing operation. The self-healing capability of ring networks provides reliable network operation in case of failure of a node or a link between nodes making the network robust against these failures. Manuscript received June 30, 2008; revised September 02, 2008. Current version published March 25, 2009. This work was supported by the Defense Advanced Research Projects Agency under Contract MDA972-03-1-0006. The authors are with the Department of Electrical Engineering, Princeton University, Princeton NJ 08544 USA (e-mail: kravtsov@princeton.edu; k6@pisem.net; ydeng@princeton.edu; prucnal@princeton.edu) Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JQE.2009.2013105 Another advantage of a ring architecture is in a more flexible code management that allows code reuse. Optical CDMA codes can be re-used in nonoverlapping segments of the ring to reduce the required number of unique codes in the network (see, e.g., [10]), thus increasing effective cardinality and improving network scalability. It is worth mentioning that most contemporary metropolitan and regional optical fiber networks are organized in rings, hence optical CDMA ring networks are more advantageous over star networks because of simpler deployment over the existing infrastructure. The ring architecture, however, requires additional active subsystems in the network. Unlike a passive network with fiber taps for each user, the ring network requires a way to remove codes from the ring once they are received. Code removal is necessary for code reuse and preventing accumulation of noise in the ring. It also provides additional improvement in data security since information becomes unavailable for all subsequent users after the signal has been received and removed from the network. So far, all demonstrated code-removal schemes [8], [11] assume synchronization of the network and are controlled by a global clock source. This approach substantially limits network capabilities allowing only synchronous network operation, thus reducing CDMA effectively to a subset of time-domain multiplexing techniques. Global synchronization is also known to be increasingly more difficult in terms of clock distribution with increasing bit rates and becomes unfeasible for optical CDMA networks with terabit/s total throughput. The intrinsically asynchronous mechanism of code removal would solve the synchronization problem and reveal full potential of a ring network based on optical CDMA. In this study, we propose and experimentally demonstrate a novel asynchronous code-removal unit. Its self-clocking ability eliminates the need for an external clock source and does not require global network synchronization. The use of all-optical thresholding together with wavelength conversion provides a way of controlling the code drop gate without any external signal. We experimentally demonstrate the use of the proposed unit in a 2.5-Gbit/s optical CDMA network with incoherent 2-D time-wavelength codes. II. INCOHERENT OPTICAL CDMA In this study, we based our analysis and experiments on incoherent optical CDMA with 2-D time-wavelength codes. From a channel capacity perspective, it has been shown theoretically that a properly designed incoherent optical CDMA system can be more robust in the presence of typical nonlinear channel impairments than its coherent counterpart (see [12, p. 109]). This section provides a brief description of incoherent CDMA 0018-9197/$25.00 2009 IEEE

KRAVTSOV et al.: SELF-CLOCKED ALL-OPTICAL ADD/DROP MULTIPLEXER FOR ASYNCHRONOUS CDMA RING NETWORKS 397 Fig. 2. Code-drop module: principle of operation. C1, C2 two optical CDMA codes, each composed of four wavelengths. C1 becomes dropped and C2 passes through. Fig. 1. (a) Carrier hopping prime codes (4, 17) used in the experimental setup; C1, C2 pulses belonging to code1 and code2. (b) FBG array-based optical CDMA encoder. used in our work. More detailed information about 2-D optical CDMA codes can be found in [12] [15]. Two-dimensional time-wavelength optical CDMA codes are combinations of optical pulses with different wavelengths and certain delays between them. In general, they can be represented by a 2-D matrix with wavelengths on one axis and time slots on the other, as shown in Fig. 1(a). In practice, it is convenient to use a small subset of 2-D codes called carrier hopping prime codes (CHPCs) [16]. In these codes, each wavelength is used only once, and delays between each pair of wavelengths are different among all possible CHPCs. To create a CHPC, a broadband picosecond optical pulse containing all wavelengths used in codes (it is typically called a supercontinuum pulse) is split into spectral components, which are combined together after applying proper delays. Data transmission is realized by modulation of the encoded signal. In this demonstration, we used simple on off keying (OOK), although other modulation formats such as M-ary are also compatible with this type of CDMA [17]. Codes created by different users are combined together, hence realizing a multiple access system. To detect the data sent by a particular user, a CDMA decoder is used. In the decoder, the incoming stream is split into different spectral components and complementary delays are applied. At the output of the decoder all spectral components of the proper code coincide in time, resulting in a high-intensity optical pulse called an autocorrelation peak. On the contrary, the codes from all other users are spread in time typically among the whole bit period or even more and can be easily distinguished from the autocorrelation peak. This signal is normally called multiple access interference (MAI). In this particular experimental demonstration CHPCs with four wavelengths and 17 time chips were used. The codes are designed for OC-48 ( 2.5 Gbit/s) bit rate with corresponding chip size of 23.6 ps. The four used wavelengths have 100 GHz ITU grid spacing and are centered at 1552.52, 1551.72, 1550.92, and 1550.12 nm. The two codes used are (0, 7, 14, 4) and (0, 3, 6, 9) where numbers represent a relative time shift of the corresponding wavelength expressed in number of chips; the codes are shown in Fig. 1(a) as C1 and C2, respectively. Note that, for consistency with the experimental results, the first code in the figure is shifted in time, so C1 appears as a cyclic column shift of the code (0, 7, 14, 4). In the experimental demonstration, optical CDMA encoders/decoders were based on the fiber Bragg grating (FBG) array technology. This type of coder is compact and has low power losses compared with thin-film filters- and AWG-based encoders/decoders [17]. Fig. 1(b) schematically shows one of the encoders. It consists of an optical fiber circulator and a FBG array with four FBGs positioned in the fiber to create different delays for different wavelengths. CDMA decoders use the same FBG arrays in reverse direction and thus are complementary with corresponding encoders. III. SELF-CLOCKING CODE-DROP UNIT ARCHITECTURE The code-drop unit that we present in this study consists of two main parts: a code-drop module similar to the one demonstrated in [11] and a clock extraction module that generates the control signal for the code-drop switch. Here, we briefly explain the principle of operation of the code-drop module and then present the novel part of the system the clock extraction system. The purpose of the code-drop module is to remove all optical pulses belonging to a particular CDMA code from the incoming stream. For CHPCs, the number of pulses is equal to the number of wavelengths used in codes, and all of these pulses should be switched out. Although building a fast optical switch for every wavelength cannot be practically affordable, all of the pulses can be aligned in time and switched out by a single all-optical gate. As mentioned before, such an alignment procedure is called decoding of the code and is performed in a CDMA decoder. Therefore, in a code-drop module, the incoming signal first passes through a CDMA decoder and then goes into a fast optical switch, which triggers to switch out the autocorrelation peak, as shown in Fig. 2. The last step is to restore all of the passed codes, which is done in a corresponding CDMA encoder that undoes the changes made in the decoder. After passing through these three elements, one of the CDMA codes gets removed from the incoming signal. Asynchronous network operation, however, implies that the relative position of optical pulses from different codes is not fixed and fluctuates. Sometimes that can lead to undesirable situations when two pulses of the same wavelength overlap in time, so pulses from other codes may be also removed together with the desired code. That potentially limits the number of hops that can be performed within a single ring, hence reducing the

398 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 45, NO. 4, APRIL 2009 Fig. 3. Block diagram of the proposed self-clocked optical CDMA code-drop unit. The all-optical thresholder extracts the control signal for the code drop switch from the incoming data stream. overall efficiency of the network. The solution may be found in a proper code design. CHPCs ensure that no more than one pulse can be removed from any of the codes, thus minimizing negative effects of code dropping. For large code sets with many wavelengths, the loss of pulses is negligible and does not noticeably worsen signal quality. A more detailed analysis of this issue is a rather fundamental question of limitations of optical CDMA and falls beyond the scope of this paper. A clock extraction module is needed for controlling the code drop gate. In an asynchronous network each drop switch should be told independently when to switch because different codes are not synchronized together. The novelty of this work is in the use of the autocorrelation peaks themselves for generation of the control signal for the switch. Effective detection of autocorrelation peaks and reduction of MAI noise was recently demonstrated with the use of an all-optical thresholder [18]. Following the same idea, we applied all-optical thresholding for generation of the control signal, as schematically shown in Fig. 3. Part of the incoming signal after the decoder is passed into an all-optical thresholder where it gets cleaned from MAI noise. The thresholder output can be then used as a control for the drop switch. In practice, however, one more step is to be performed to allow filtering out the control signal from the gate output. The most common and convenient way to perform this signal separation is by spectral filtering, which can be successfully used provided the control and input signals have different wavelengths. To acquire this wavelength separation, the all-optical thresholder was combined with a wavelength converter in a single nonlinear optical loop mirror (NOLM)-based device that is the core element of the clock extraction module. The NOLM-based thresholder/converter is schematically shown in Fig. 4. It is a NOLM-based switch working as a thresholder at small nonlinear phase shifts as in [18]. The input of the switch is connected to a CW light source while its control input is connected to the decoded optical CDMA input stream. The transfer function of such a device can be written as, where is the nonlinear coefficient of the nonlinear fiber in the loop and is the control signal intensity. At small nonlinear phase shifts, Taylor expansion of cosine around zero yields, i.e., for small control signals and CW input, the output signal is proportional to the square of the control signal. This nonlinear transfer Fig. 4. Nonlinear fiber-based all-optical thresholder/wavelength converter. PC polarization controller; BPF thin-film bandpass filter; 50/50 and 75/25 splitting ratios of the fiber couplers used; converted wavelength; 0 wavelengths used in CDMA codes. characteristic enables the thresholding operation with inherent wavelength conversion. In asynchronous optical CDMA networks, the intensity of autocorrelation peaks is always higher than intensity of MAI noise. After quadratic transformation in the NOLM, MAI noise becomes suppressed compared with the autocorrelation peaks. Note that the output wavelength of the NOLM is the wavelength of the CW source and can be easily varied. The discussed device performs all necessary steps to generate the control signal for the drop switch. Also note that the output signal of the device is modulated in accordance with the data being sent by the user generating this CDMA code, therefore the control signal should be properly delayed to ensure control and data signal alignment in the drop switch. IV. EXPERIMENTAL DEMONSTRATION To demonstrate compatibility of the proposed clock extraction technique with different switching technologies, the experimental demonstration was performed in two different configurations: one with code-drop switch based on the TOAD [19] and another with a nonlinear fiber-based switch. Although the first configuration is more power-efficient, it suffers from speed limitation of semiconductor optical amplifiers (SOAs). The second all-fiber configuration requires much higher control power level, but provides higher bandwidth of operation. A. TOAD-Based Configuration The experimental setup is schematically shown in Fig. 5. The setup operates at a bit rate of 2.5 Gbit/s and contains two optical CDMA users. Both users share the same data modulator and their signals are interleaved in time to emulate independent transmissions. Signals from the two users are mixed together

KRAVTSOV et al.: SELF-CLOCKED ALL-OPTICAL ADD/DROP MULTIPLEXER FOR ASYNCHRONOUS CDMA RING NETWORKS 399 Fig. 5. Experimental setup with the TOAD-based code-drop switch. MZ mod Mach-Zehnder LiNbO modulator; EDFA erbium-doped fiber amplifier; SC dispersion decreasing fiber-based supercontinuum generator; enc/dec optical CDMA encoder/decoder; PC polarization controller; NL fiber highly GeO -doped nonlinear fiber; CW continuous wave WDM semiconductor laser, T variable time delay. Fig. 6. Measured signal waveforms for the TOAD-based configuration. (a) CDMA code1 (passing). (b) CDMA code2 (dropped). (c) Combination of the two codes before entering code-drop unit. (d) Passed code output. (e) Decoded signal at the input of the drop gate. (f) Control signal for the drop gate generated by the thresholder/wavelength converter. (g) Signal at the output from the code-drop gate. and passed into the code drop unit that decodes one of the codes, extracts the control signal for the drop gate, drops this code, and then restores the remaining code. Both dropped and passed signals are then analyzed with a bit error rate tester (BERT). In the setup, we use an erbium-doped fiber mode-locked laser (MLL) that generates picosecond optical pulses at the central wavelength of 1548 nm. Its signal is then modulated by a Mach Zehnder modulator with a pseudorandom bit sequence. After proper amplification, the signal is passed to a dispersion decreasing fiber pulse compressor to generate broadband supercontinuum pulses. Two copies of its output signal are then passed through FBG array-based CDMA encoders and then combined together with some time delay. At this point, the resulting signal imitates traffic in a typical ring network carrying two independent data streams with two different CDMA codes. In a node, one of the codes is to be dropped from the network and the other is to pass the node. To achieve that, the incoming signal passes through a CDMA decoder matching one of the codes and then enters the control input of the NOLM. The other input of the NOLM is connected to a CW fiber laser operating at nm with 13 dbm of average power that defines the converted wavelength of the thresholder s output. As the active element in the NOLM, we use 15 m of silica-based highly GeO -doped nonlinear fiber [20] with nonlinear coefficient of 35 W km and dispersion of ps/nm km. The thresholder operates at an average control power level of 25 dbm, which is enough to create a sufficient nonlinear phase shift in the NOLM. The output of the thresholder passes through a couple of 200-GHz bandwidth spectral filters centered at nm to clean the converted wavelength from the control signal (original CDMA codes). Results of the thresholding clock extraction module operation are shown in Fig. 6. Fig. 6(a) and (b) show waveforms of the original CDMA codes of the two users. These signals are then mixed together as shown in Fig. 6(c) and passed through the CDMA decoder. The decoded signal with the autocorrelation peak and MAI is shown in Fig. 6(e). The output of the decoder passes through the thresholder. Its output at the new wavelength is shown in Fig. 6(f). It clearly duplicates the pattern in Fig. 6(e) with greatly reduced MAI pulses. To make sure that this discrimination is not a result of polarization filtering, we passed the incoming signal through a polarization beam splitter. With the splitter, both CDMA codes share the same polarization state and spectrum, which guarantees that the thresholder works properly, performing amplitude discrimination of the incoming signal. The observed quality of the resulting signal is good enough to control the drop switch. The thresholder output can also be used for reception of data carried by the detected code (drop code). Further signal processing is done in the conventional TOAD switch [19]. Signals coming out from the thresholder swap their roles when entering the TOAD: the thresholder output becomes a control signal for the TOAD, while the thresholder control signal becomes an input signal for the TOAD. Proper delay is added to the latter signal to align the control signal with the incoming data in the TOAD. Output of the TOAD after filtering out the control signal is shown in Fig. 6(g). This signal is then passed through CDMA encoder to restore the remaining code. As expected, its observed waveform shown in Fig. 6(d) is the same as the waveform of the original code. This passing signal is then decoded and received with a photodetector. All observed waveforms have clean eye openings for all signals. Bit error rates were measured explicitly for both dropped

400 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 45, NO. 4, APRIL 2009 Fig. 7. Experimental setup with the NOLM-based code-drop switch. MZ mod Mach-Zehnder LiNbO modulator; EDFA erbium-doped fiber amplifier; SC dispersion decreasing fiber-based supercontinuum generator; enc/dec optical CDMA encoder/decoder; PC polarization controller; NL fiber highly GeO -doped nonlinear fiber; CW continuous wave WDM semiconductor laser; T variable time delay. Fig. 8. Measured signal waveforms for the NOLM-based configuration. (a) CDMA code1 (passing). (b) CDMA code2 (dropped). (c) Combination of the two codes before entering code-drop unit. (d) Passed code output. (e) Decoded signal at the input of the drop gate. (f) Control signal for the drop gate generated by the thresholder/wavelength converter. (g) Signal at the output from the code-drop gate. Fig. 9. BER versus received power for the dropped and passed codes together with back-to-back error rate measurement. and passed codes with a BERT. After proper adjustment of the whole setup we observed error-free transmission for both codes. B. Nonlinear Fiber-Based Configuration Another demonstration was performed for an all-fiber drop code switch. Its schematic is shown in Fig. 7. The setup differs from the TOAD-based demonstration in the code drop gate. Here we use a NOLM-based all-optical switch with similar parameters to the one used in the thresholder. Its nonlinear element is made of 11 m of the same nonlinear fiber with nonlinear coefficient of 35 W km. This gate requires much higher control signal power than the TOAD. In this demonstration, the average control power was 25 dbm. Signal waveforms observed in this system configuration are shown in Fig. 8. As follows from Fig. 8(g), the output of the code drop switch still contains some residual power from the autocorrelation peak that had to be removed. We attribute this insufficient switching ratio to effects of dispersion in the NOLM. It is known that dispersion of a nonlinear element in the NOLM strongly affects operation of the switch [21], although in this demonstration we did not perform any dispersion management. The restoration of the passed code after the drop switch spreads this residual pulse into four smaller pulses and therefore decreases its negative effect. Bit error rates were measured to estimate quality of the output signals. Results of the measurements are plotted in Fig. 9. One can see a 2-dB power penalty for the passed through code compared to back-to-back measurements. This penalty results from imperfection of the code-drop switch we mentioned. Dropped code BER measurements were performed at the converted wavelength output of the thresholder. This signal has even larger power penalty resulting from imperfection of spectral conversion. For a received power of as low as 25 dbm both, passed and dropped codes can be detected error-free. V. CONCLUSION In this paper, we presented a novel approach to implementation of an add/drop multiplexer for asynchronous optical CDMA ring networks. The performed experimental demonstration shows the feasibility of the proposed self-clocked all-optical multiplexer for use in such networks. Although the current demonstration is performed only for two users, there are no immediate limitations on number of users, besides regular constraints of optical CDMA. However, the small size of the network imposes some limitations on the experimental demonstration. Unlike a large-size network where effects of the statistical overlap of autocorrelation peaks and MAI can be studied [22], our demonstration uses a small code space with only four wavelengths, and overlap of one of them has too significant an impact on the codes. Thus, in our case, all measurements were performed for nonoverlapping MAI and autocorrelation peaks. Performance of the novel clock extraction scheme is found to be satisfactory, and it can be potentially

KRAVTSOV et al.: SELF-CLOCKED ALL-OPTICAL ADD/DROP MULTIPLEXER FOR ASYNCHRONOUS CDMA RING NETWORKS 401 used in many applications. Two different realizations of the code-drop unit were presented demonstrating compatibility of the thresholding clock extractor with both SOA- and fiber-based optical gates. Error-free operation of the setup was observed in both configurations. The proposed self-clocked architecture of an add/drop multiplexer for optical CDMA solves the problem of code drop switch synchronization with the incoming data and eliminates the need of clock distribution in ring networks. This approach helps to utilize the full potential of CDMA for asynchronous optical networks. Demonstrations of asynchronous ring networks, which provide increased protection against link failures and improve data privacy, can be accomplished using the proposed all-optical add/drop multiplexers. ACKNOWLEDGMENT The authors would like to thank Oki Electric Industry Company, Ltd. for providing the fiber Bragg gratings. The authors would also like to thank the Fiber Optics Research Center, Moscow, Russia, for their cooperation. The authors also greatly appreciate the help of N. Kostinski in editing the paper and the support from all other members of the Lightwave Communication Research Group, Princeton University, Princeton, NJ. REFERENCES [1] M. A. Santoro and P. R. Prucnal, Asynchronous fiber optic local area network using CDMA and optical correlation, Proc. IEEE, vol. 75, no. 9, pp. 1336 1338, Sep. 1987. [2] A. Stock and E. H. Sargent, The role of optical CDMA in access networks, IEEE Commun. Mag., vol. 40, no. 9, pp. 83 87, Sep. 2002. [3] J. Shah, Optical code division multiple access, Opt. Photon. News, vol. 14, no. 4, pp. 42 47, 2003. [4] C. F. Lam, D. T. K. Tong, M. C. Wu, and E. Yablonovitch, Experimental demonstration of bipolar optical CDMA system using a balanced transmitter and complementary spectral encoding, IEEE Photon. Technol. Lett., vol. 10, no. 10, pp. 1504 1506, Oct. 1998. [5] S. Ayotte, M. Rochette, J. Magne, L. A. Rusch, and S. LaRochelle, Experimental demonstration and simulation results of frequency encoded optical CDMA, in Proc. IEEE Int. Conf. Commun., Jun. 2004, vol. 3, pp. 20 24. [6] E. Park, A. J. Mendez, and E. M. Garmire, Temporal/spatial optical CDMA networks-design, demonstration, and comparison with temporal networks, IEEE Photon. Technol. Lett., vol. 4, no. 10, pp. 1160 1162, Oct. 1992. [7] G. A. Magel, G. D. Laundry, R. J. Baca, D. A. Harper, and C. A. Spillers, Transmission of eight channels 622 mbit/s and 15 channels 155 mbit/s using spectral encoded optical CDMA, Electron. Lett, vol. 37, pp. 1307 1308, 2001. [8] B. Chen, C. Guo, J. Chen, L. Zhang, Q. Jiang, and S. He, Add/drop multiplexing and tdm signal transmission in an optical CDMA ring network, J. Opt. Networking, vol. 6, no. 8, pp. 969 974, 2007. [9] A. Elrefaie, Multiwavelength survivable ring network architecture, in Proc. ICC, 1993, pp. 1245 1251. [10] J. M. Simmons, E. L. Goldstein, and A. A. M. Saleh, Quantifying the benefit of wavelength add-drop in WDM rings with distance-independent and dependent traffic, J. Lightw. Tech., vol. 17, no. 1, pp. 48 57, Jan. 1999. [11] C.-S. Brès, I. Glesk, R. J. Runser, and P. R. Prucnal, All-optical OCDMA code-drop unit for transparent ring networks, IEEE Photon. Technol. Lett., vol. 17, no. 5, pp. 1088 1090, Mar. 2005. [12], P. R. Prucnal, Ed., Optical Code Division Multiple Access: Fundamentals and Applications. New York: Taylor & Francis Ltd., 2006. [13] L. Tančevski and I. Andonovic, Wavelength hopping/time spreading code division multiple access systems, Electron. Lett, vol. 30, pp. 721 723, 1994. [14] G. C. Yang and W. C. Kwong, Two-dimensional spatial signature patterns, IEEE Trans. Commun., vol. 44, no. 2, pp. 184 191, Feb. 1996. [15] R. M. H. Yim, L. R. Chen, and J. Bajcsy, Design and performance of 2-d codes for wavelength-time optical CDMA, IEEE Photon. Technol. Lett., vol. 14, no. 5, pp. 714 716, May 2002. [16] G.-C. Yang and W. C. Kwong, Prime Codes With Applications to CDMA Optical and Wireless Networks. Boston, MA: Artech House, 2002. [17] C.-S. Brès, Y.-K. Huang, I. Glesk, and P. R. Prucnal, Scalable and asynchronous incoherent optical CDMA, J. Opt. Networking, vol. 6, no. 6, pp. 599 615, 2007. [18] K. Kravtsov, P. R. Prucnal, and M. M. Bubnov, Simple nonlinear interferometer-based all-optical thresholder and its applications for optical CDMA, Opt. Exp., vol. 15, no. 20, pp. 13 114 13 122, 2007. [19] 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. [20] E. M. Dianov and V. M. Mashinsky, Germania-based core optical fibers, J. Lightw. Technol., vol. 23, no. 11, pp. 3500 3508, Nov. 2005. [21] D. J. Richardson, R. I. Laming, and D. N. Payne, Switching and passive mode-locking of fibre lasers using nonlinear loop mirrors, Proc. SPIE, vol. 1581, pp. 26 39, 1991. [22] V. J. Hernandez, A. J. Mendez, C. V. Bennett, R. M. Gagliardi, and W. J. Lennon, Bit-error-rate analysis of a 16-user gigabit ethernet optical- CDMA (O-CDMA) technology demonstrator using wavelength/time codes, IEEE Photon. Technol. Lett., vol. 17, no. 12, pp. 2784 2786, Dec. 2005. Konstantin Kravtsov received the M.S. degree from the Moscow Institute of Physics and Technology, Moscow, Russia, in 2005. He is currently working toward the Ph.D. degree in electrical engineering at Princeton University, Princeton, NJ. Yanhua Deng (S 07) received the B.Eng. degree in electrical engineering from Cooper Union for the Advancement of Science and Art, New York, in 2006. She is currently working toward the Ph.D. degree in electrical engineering at Princeton University, Princeton, NJ. Her graduate research involves all-optical networks and network system design based upon optical CDMA technologies. Paul R. Prucnal (F 92) received the A.B. degree from Bowdoin College, and the M.S., M.Phil. and Ph.D. degrees from Columbia University, New York. He was a faculty member with Columbia University until 1988, when he joined Princeton University as a Professor of electrical engineering. From 1990 to 1992, he served as Founding Director of Princeton s Center for Photonics and Optoelectronic Materials. He has also held positions as Visiting Professor with the University of Tokyo and the University of Parma. He is the inventor of the Terahertz Optical Asymmetric Demultiplexer, an ultrafast all-optical switch, and is credited with doing seminal research in the areas of all-optical networks and photonic switching, including the first demonstrations of optical code-division and optical time-division multi-access networks in the mid 1980s. With DARPA support in the 1990s, his group was the first to demonstrate a 100-gigabit/s photonic packet switching node and optical multiprocessor interconnect, which was nearly 100 times faster than any system with comparable functionality at that time. For the past several years, he has been doing research on optical CDMA networks. He has published over 200 journal papers and holds 17 patents. He is editor of the book, Optical Code Division Multiple Access: Fundamentals and Applications (Taylor and Francis, 2006). Prof. Prucnal is a Fellow of the Optical Society of America (OSA). He is currently an Area Editor of the IEEE TRANSACTIONS ON COMMUNICATIONS for optical networks. He was general chair of the OSA Topic Meeting on Photonics in Switching in 1999 and was a recipient of the Rudolf Kingslake Medal from SPIE. In 2005, he was the recipient of a Princeton University Engineering Council Award for Excellence in Teaching and, in 2006, the recipient of the Graduate Mentoring Award in Engineering at Princeton.