Experimental demonstrations of all -optical networking functions for WDM optical networks

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content EXPERIMENTAL DEMONSTRATIONS OF ALL-OPTICAL NETWORKING FUNCTIONS FOR WDM OPTICAL NETWORKS by Deniz Gurkan A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment o f the Requirements for the Degree DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING) December 2003 Copyright 2003 Deniz Gurkan Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 3133278 INFORMATION TO USERS The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. ® UMI UMI Microform 3133278 Copyright 2004 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90089-1695 This dissertation, written by ________D eniz GURKAN under the direction o f h e r dissertation committee, and approved by all its members, has been presented to and accepted by the Director o f Graduate and Professional Programs, in partial fulfillment o f the requirements fo r the degree o f DOCTOR OF PHILOSOPHY Director Date Decem ber 1 7 , 2003 Dissertation (^mrfitjtee. Chair Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Acknowledgements I would like to thank all o f the members o f optical communications lab for their friendship, team spirit, and wonderful moments: Early in the mornings, John McGeehan would be the perfect person to start the day with the latest sports critics. Anytime in the day or night, Asaf Sahin would help me figure out anything about any equipment, project, and research in the lab and beyond... Paniz Ebrahimi would give me the motivation to think outside the box and figure out fundamental basis o f things in a philosophical way especially after midnight. Mustafa Cardakli was my first senior member to help me out with everything in the lab, without his help and guidance I would not be able to learn things as easy as I managed to. Lianshan Yan has the most wonderful cheerful character even in the middle o f the night, with 10,000 more data points to be taken. Saurabh Kumar was the first person that I had the opportunity to teach the ways in the lab, it was a wonderful experience on my side to learn even m ore... And I am very lucky that my advisor Professor Alan Willner was in his office on the day that I came to USC to greet me with appreciation and welcome to his wonderful group. I am very lucky that I had the experience to work closely with him in the most dedicated way in every aspect o f research, from presentation to searching. I feel lucky also that Milly Montenegro, Mayumi Thrasher, and Tim Boston were all taking care o f me in iii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. administrative issues as well as sharing my joy and experiences with an infinite support and best attitude. I would like to also thank Dr. Shing-Shiong Chang for the lifelong memorable discussions on life; he will always have a special place in my heart and life philosophy. Special gratitude to my sister, Sevgi Gurkan, for her infinite support and confidence in me at all times, especially the desperate moments... I am already blessed that I have a sister, I am further blessed that she could join me here during the last two years o f my PhD studies. I am very lucky that I had the financial and moral support from my parents during my studies. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table of Contents Acknowledgements..... ...................................... ...................................... iii List ofTables and Figures.................. .............................. .............................. vii Abstract......................................................................................................... . xi 1 Introduction ....................................... ...................................................... 1 1.1 Opticai-to-Electrical-to-Optical (O/E/O) Conversion....................... 1 1.2 Optical Layer .... ..................................................................... 1 1.3 Proposal Outline................................................................................... 2 2 All-Optical Networks........................................................ ................ ............ 4 2.1 Optical Layer............................................................... ......................... 4 2.2 Wavelength Conversion ...... ........................................................ 8 3 Variable-Bit-Rate Header Recognition for Reconfigurable Networks using Tunable Fiber Bragg Gratings as Optical Correlators........................... 16 3.1 Introduction ....... 16 3.2 Experimental Setup of Variable Bit Rate Header Recognition Module...................... 18 3.3 Experimental Results and Summary...................................... 20 4 Tunable All-Optical Time-Slot-Interchange and Wavelength Conversion using Difference-Frequency-Generation and Optical Buffers....................... 24 4.1 Introduction.................................... 24 4.2 Experimental Setup..................... 26 4.3 Results and Discussion ..... 29 5 All-Optical Header Recognition o f WDM Channels Using a PPLN Wavelength Shifter and FBG-Based Correlators .... ................................. 33 5.1 Introduction............................... 34 5.2 Operational Concept ....... 36 5.3 Experimental Setup ....................................................... 39 5.4 System Demonstration................................... 41 6 Simultaneous and independent label swapping o f multiple WDM channels in an all-optical packet-switched network using PPLN waveguides as wavelength converters .... ............ ................. 44 6.1 Introduction........................... 44 6.2 Experimental Setup ....... 47 6.3 Results and Discussion............................................................ 48 v Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 All-Optical Wavelength and Time 2-D Code Converter for Dynamically-Reconfigurabie O-CDMA Networks using a PPLN Waveguide......................................... 53 7.1 2-D OCDMA Networks Utilizing Code-converters....................... 53 7.2 Time/Wavelength O-CDMA Structure and Code Conversion.... 55 7.3 Experimental Setup .............................. 57 7.4 Results and Discussion................................ 60 Bibliography................................................ 62 vi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. List of Tables and Figures Figure 2.1. The functions in a typical node to be performed in an all- optical network ......................... ............................................. .................. 5 Figure 2.2. Reduction in packet loss probability, as a function o f buffer size when using wavelength conversion, and under different load conditions.................................................................................................. ...... 9 Figure 2.3. Factors affecting the performance o f a transparent penalty- free wavelength shifting mode....................... ................ ............................... 10 Figure 2.4. Wavelength conversion by cross-gain modulation in a semiconductor optical amplifier............................................................... ...... 11 Figure 2.5. Wavelength conversion in a Mach-Zehnder configuration. Depending on the phase difference induced by the pump signal, the probe signals in the two arms o f the output coupler combine constructively or destructively ....................................... 12 Figure 2.6. Periodically-poled lithium niobate waveguide mixing operation ........................................................................... 13 Figure 3.1. Conceptual diagram o f optical variable-bit-rate header recognition and switching module................................................................ 17 Figure 3.2. Experimental setup.................................................................. 18 Figure 3.3. (a) Input packet stream to variable-bit-rate header recognition module that recognizes header ‘11’. (b) Grating array #1 output, (c) Grating array #2 output, (d) Decision electronics output, (e) Port 1 output of the optical switch, 2.5-Gb/s packet with header 'O F and 622-Mb/s packet with header ‘10’ pass through, (f) Port 2 output o f the optical switch, 2.5-Gb/s and 155-Mb/s packet with header ‘11’ recognized and dropped............................. 20 v ii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 3.4 (a) Input packet stream to variable bit rate header recognition module that recognizes header ‘10’. (b) Grating array #1 output, (c) Grating array #2 output, (d) Decision electronics output, (e) Port 1 output o f the optical switch, 2.5-Gb/s packet with header '00' and 622-Mb/s packet with header ‘11’ pass through, (f) Port 2 output o f the optical switch, 2.5-Gb/s and 155-Mb/s packet with header ‘10’ recognized and dropped.......................... 21 Figure 3.5. a) 155-Mb/s input signal eye diagram with >3 ns rise and fall times, b) 155-Mb/s decision eye after the optical correlator, c) 2.5-Gb/s input signal eye diagram with <200-ps rise and fall times, d) 2.5-Gb/s decision eye after the optical correlator......................................................... 22 Figure 3.6. BER curves for (a) ‘11’, and (b) ‘10’ header recognition indicating no degradation using our module relative to back-to-back measurements ..................................... 23 Figure 4.1. Conceptual diagram o f our time slot interchange and wavelength conversion experiment..................................................... 27 Figure 4.2. Experimental setup................................................ 28 Figure 4.3. (a) Incoming bit stream at 1555 nm, (b) rectangular pulse train (pump) at 1550 nm, copying odd numbered time slots to 1545 nm, (c) 1545-nm DFG signal after first PPLN device, (d) 1545-nm signal after the FBG array, two bit time delayed with respect to 1555-nm signal...................................................... 30 Figure 4.4. (a) 1555-nm input signal to second PPLN, (b) 1545-nm input signal to second PPLN, (c) rectangular pulse train (pump) at 1550 nm, copying even numbered time slots to 1545 nm, (d) 1545-nm signal output from second PPLN device, time slots interchanged and wavelength converted ....... 31 Figure 4.5. BER curves for back-to-back and after our time-slot- interchange and wavelength-conversion module .................................... 32 Table 5.1. Comparison o f the # of components required for the SOA vs. PPLN WDM-HR implementations ............................................................ 35 Figure 5.1.2x2 optical cross connect implemented with WDM-HR modules................. 36 viii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 5.2. Conceptual diagram o f a WDM-HR module. Each FBG array is uniquely configured to test for a different header pattern on each wavelength. HI = 1s t header bit; X n = channel 1 ,2n d header bit; X p i = pump #1 (header bit selector #1).............................. 37 Figure 5.3. Experimental setup o f a 2-channel WDM-HR module for recognizing a particular combination o f 2-bit packet headers..................... 40 Figure 5.4. PPLN output spectrum showing the two input WDM channels, the two header bit selector pumps, and the four k-shifted header bits. X n = channel 1, 2n d header bit; A , p = pump ........... 41 Figure 5.5. Results o f WDM-HR system demonstration. The FBG array is configured to correlate with header ‘01’ at A ,i and header ‘ 10’ at X i , and the 2x2 OXC is programmed according to the above lookup table, (a) Input WDM channels at X \ & X 2, (b) header bit selector pumps, (c) decision circuit output (used as the switch control signal), (d) port C output at Xi & X 2, (e) port D output at X \ & X 2................... ^ 3 Figure 6.1. Old labels are removed and the payload and new labels are mapped onto new wavelengths. The new label is inserted into the place o f the old one to generate the wavelength- shifted and label-swapped output signal........................................ 45 Figure 6.2. Experimental setup. The payloads and new labels are wavelength converted by individual PPLN waveguides. The outputs are filtered and combined to produce the label swapped WDM channels................................... 47 Figure 6.3. The spectrum of the label swapping subsystem, (al), (a2) are the input spectra to the PPLN waveguides, payload and label-selection spectra, respectively, (bl), (b2) are the DFG spectra after the PPLN waveguides, (cl), (c2) are the label swapped and ^.-shifted outputs 49 Figure 6.4. (a), (b) Four input packets o f the 2 WDM channels, (c) The new-label-selector waveform that is used as the wavelength-converter pump in the first PPLN. (d) The payload-selector waveform to wavelength-convert the payload o f the packets, (e), (f) The wavelength- converted payloads for each channel, (g), (h) The label-swapped output waveforms for each channel................................. 50 ix Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.5. Waveforms o f a single packet from WDM channels 1 and 2 before and after label swapping, (a), (b) The original label on A ,i, 01100111, is swapped with new label, 00101001. (c), (d) The original label onX.2, 1110110, is swapped with new label, 01010001....... 51 Figure 6.6. BER curves for the module. Label swapping introduces a penalty o f 1 dB and 2.5 dB for the WDM channels Xi and X 2, respectively................................................... 52 Figure 7.1. Concept o f the 2-D O-CDMA code converter to resolve code contention for 2 LANs sharing the same particular code.............................. 55 Figure 7.2. (a) Wavelength shifting using a PPLN waveguide, (b) Time shifting o f the 2-D code converter ..... 56 Figure 7.3. (a) O-CDMA encoder with 4 wavelengths and 4 time chips and the 2-D code converter with wavelength and time shifting, (b) 0 - CDMA decoder composed o f the autocorrelation and threshold detection circuit.................................................. 57 Figure 7.4. (a) Decoding o f the input and code-converted codes in the system, (b) Code contention resolution at a routing switch using the code converter module. Whenever a LAN-frame is coded with a code in use in LAN-B, the code-converted signal is routed at the switch output............... 59 Figure 7.5. BER measurements o f our module............................................. 60 x Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abstract The deployment o f optical networks will enable high capacity links between users but will introduce the problems associated with transporting and managing more channels. Many network functions should be implemented in optical domain; main reasons are: to avoid electronic processing bottlenecks, to achieve data-format and data-rate independence, to provide reliable and cost efficient control and management information, to simultaneously process multiple wavelength channel operation for wavelength division multiplexed (WDM) optical networks. The following novel experimental demonstrations o f network functions in the optical domain are presented: Variable-bit-rate recognition o f the header information in a data packet. The technique is reconfigurable for different header sequences and uses optical correlators as look-up tables. The header is processed and a signal is sent to the switch for a series of incoming data packets at 155 Mb/s, 622 Mb/s, and 2.5 Gb/s in a reconfigurable network. Simultaneous optical time-slot-interchange and wavelength conversion of the bits in a 2.5-Gb/s data stream to achieve a reconfigurable time/wavelength switch. The technique uses difference-frequency-generation (DFG) for wavelength conversion and fiber Bragg gratings (FBG) as wavelength-dependent optical time buffers. xi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The WDM header recognition module simultaneously recognizing two header bits on each of two 2.5-Gbit/s WDM packet streams. The module is tunable to enable reconfigurable look-up tables. Simultaneous and independent label swapping and wavelength conversion o f two WDM channels for a multi-protocol label switching (MPLS) network. Demonstration o f label swapping o f distinct 8-bit-long labels for two WDM data channels is presented. Two-dimensional code conversion module for an optical code-division multiple-access (O-CDMA) local area network (LAN) system. Simultaneous wavelength conversion and time shifting is achieved to enable flexible code conversion and increase code re-use among different LANs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 1 Introduction 1.1 Optical-to-Electrical-to-Optical (O/E/O) Conversion As the data rates in optical networks increase to 10 Gbps and further to 40 Gbps, at the intermediate nodes electronic processing will cause a bottleneck. The increase in data rate is not likely to be caught up by electronics. The saturation point for the electronic processing rate is expected to be for data rates higher than 40 Gbps. Then, the all-optical processing comes into the picture. With all-optical processing at switching and/or routing node, we envision to have no O/E/O conversion. Therefore, not only the electronic data processing rate but also protocol and format dependence in the future optical network becomes irrelevant. 1.2 Optical Layer As future networks evolve towards having more bandwidth available to the end users, the backbone optical network will be responsible for more o f the networking functions. These functions include: • Header processing - header recognition, header removal, header replacement (header o f each packet has the destination and the source address as the enablers for routing and switching) • Wavelength routing for lightpaths 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ® Wavelength conversion to consolidate the minimum number of wavelengths with maximum number o f lightpaths • Label swapping The emerging protocols are the Ethernet at the access technology side and MPLS (Multi-Protocol Label Switching) at the backbone technology side of the overall network. These protocols will require more processing at the optical layer. 1.3 Proposal Outline In this proposal, Chapter 2 will present optical networks in general. The main requirements at the optical layer will be described. An important network function in future all-optical transport networks is the recognition of header information all-optically. This information can be used for high-speed switching of optical packets. Another one o f the most required capabilities of optical layer is transparency: being able to handle different formats and bit rates with the same equipment. In Chapter 3, we demonstrate how variable-bit-rate header recognition can be achieved. In a network that includes elements of time-division multiplexing (TDM), the time slot o f each bit defines its output port at each switching node. This makes time-slot-interchange (TSI) the most commonly used method of switching in the time domain [1]. TSI is achieved by shifting a given bit from one time slot into a different one, thereby changing its destination. In Chapter 4, a reconfigurable time/wavelength switch is explained, which optically time-slot-interehanges and wavelength converts a 2.5-Gb/s bit stream using 2 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. difference-frequency-generation (DFG) for wavelength conversion and fiber Bragg gratings (FBG) as wavelength-dependent optical time buffers. With the demand for capacity increase in the optical backbone, wavelength division multiplexing (WDM) enabled more channels to be handled by each single fiber. The all-optical functions, especially header recognition at the switching nodes should be able to handle multiple wavelengths to recognize the header information and route packets on-the-fly. In Chapter 5, our experiment on the demonstration o f multiple channel header recognition is presented. As MPLS emerges as the optical layer protocol to handle differing traffic on the backbone as flow switching, label swapping function becomes a must at each node o f a MPLS architecture. The demonstration of an all-optical multiple wavelength label swapping module is included in the final chapter, Chapter 6. 3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 2 All-Optical Networks Optical carriers should be independent from the intrinsic characteristics of the transported signal in order to perform functions that would not be economically feasible using electronic means. This will be essential prerequisite for creating new generation networks. 2.1 Optical Layer In the optical layer, signals travel transparently and independently o f the transmission format and are subject to the so called “optical transport functions” consisting mainly of the following [2]: ® Optical packet header recognition, ® Time, wavelength, and space switching, • Synchronization, • Power, packet and bit level monitoring, • Compensation of long distance transmission degradations, • Wavelength assignment to the various signals via suitable, switchable and /or tunable sources, • Signal selection via optical filters, which are fixed and/or tunable to desired wavelength, • Stream/packet routing via spatial optical switches, 4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. • Wavelength conversion, achieved by means o f optics to electronics (O/E) and E/O conversions or via integrated converters, • Amplification via doped fiber and/or semiconductor devices. These functions take a part in an optical packet switching router as illustrated in the following diagram. ss?. V a ® ad d ress/lab el recognition contention detection & switching control unit monitoring WDM output packet arrival detection regeneration - routing decision optical switch matrix optical look-up table for switching decision address/label updating (label swapping) O/E conversion electronic processing FULL look-up table Figure 2.1. The functions in a typical node to be performed in an all-optical network. It is advantageous to use WDM, which increases node traffic management capacity and produces greater connection flexibility [3], The other benefits anticipated from “optical transport layer” [4]: ® Enormous aggregate capacity, easily growing to the tens or hundreds o f Tb/sec although no user “sees” an access rate greater than that of an individual channel. • Scalability: the property o f being able to add nodes. By deploying more copies o f enlarging service region while offering higher aggregate capacity. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ® Modularity: the property of being able to add only the desired number o f new nodes, and updating the logical connectivity diagram to include new nodes. • Permanence: Once the optical medium is installed, it assumes the appearance o f “permanence” in those new access nodes. New services, and new network architectures can always be added and/or updated without requiring that the pre-existing medium be retrofitted. • Multiple virtual network integration onto a common physical medium. Some o f the “clear channels” among nodes can be used for an analog network, some for a digital network, some for a voice network, some for a video network. ® Integration of circuit/packet switching capabilities. ® Distributed ATM fabric, rearrangable to optimize performance requirements. • Enhanced reliability: The failure of an access node affects only the users connected via that node; the failed node is then bypassed by changing the connection diagram. Thereby preserving all other virtual and real connections; if the failed node serves a large number of users such as a local area network, then dual homing on two access nodes may be used for further reliability enhancement. 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. • Independence of the physical topology from the logical interconnectivity: The connectivity diagram is easily modified in response to changing traffic patterns, service requests, and equipment failure/restoration, without requiring a physical change to the optical medium. The adaptation o f varying channel densities in WDM and variable-bit-rate processing in each channel will also make it possible to divide the optical layer into a certain number of subnetworks. This means that SDH/SONET, “pure” ATM, TCP/IP, DQDB, dedicated lines, PDH networks and/or any other type o f system could coexist independently in the optical layer. In turn, each sub-network could also support geographically distinct networks provided that an efficient shared management and wavelength reuse protocol is established [5]. The most obvious advantages of separating the complimentary functions o f the optical and electrical layers are as follows: • Transport capacity can be increased gradually through the addition o f wavelengths, and/or increase o f bit-rate without having to use extremely complex and costly ultrahigh speed systems. • Multiplexing, switching, monitoring, and transmission level functions are integrated in the optical interface, thus simplifying and reducing DXC node requirements and associated costs. • Transmission functions from source to destination are simplified. 7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. • Electronic circuits can be simplified by relieving them of the more complex and demanding functions, such as switching large signal blocks at high bit rates. • System reliability is improved by reducing the required electronic circuitry. ® In general, the optical layer will be: ® Transparent, in that signal o f any kind will be able to coexist. • Virtually unlimited in capacity and hence able to accept all future developments, in that it is possible to use extremely close spaced channels (high density WDM). • Flexible, in that it is possible to drop, insert and route signals on the basis o f the associated header and/or wavelength. ® Reliable, in that reconfigurations can be accomplished very quickly. • Easily managed through appropriate control of the electro-optic components. 22 Wavelength Conversion Wavelength conversion is a key enabling technology and function in the experiments demonstrated in this proposal. We used wavelength converters to process information in the optical domain by means o f either correlator gratings or selective filtering. 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Wavelength conversion is the technique of transferring the information modulated on an optical carrier to another optical carrier of a different wavelength. Such a conversion can allow wavelength reuse among several all- optical network sections, by transferring data to any available wavelength. The use of tunable wavelength converters may enable network functions such as routing and switching and may also reduce contention. As shown in Fig. 2.2, with the use of wavelength converters, packet loss probability can be significantly reduced [6]. Load 0.8, lO U t 5 15 25 35 45 55 65 75 85 95 105 Buffer size Figure 2.2. Reduction in packet loss probability, as a function o f buffer size when using wavelength conversion, and under different load conditions. Different network structures may have different requirements on wavelength converters. Figure 2.3 summarizes the main requirements for a transparent, penalty-free wavelength-shifting node. Many wavelength-shifting schemes have been demonstrated but none can simultaneously satisfy all of these requirements. Existing wavelength conversion techniques include optic- 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. electronic-optic (O/E/O) conversion, cross-gain modulation (XGM), cross-phase modulation (XPM), difference frequency generation (DFG), and four-wave mixing (FWM). Extinction Ratio Degradation Format Dependency Polarization Dependence Chirp Bandwidth + Speed Conversion Efficiency Additive Noise Input Power Limited to single V i per S O A y Figure 2.3. Factors affecting the performance o f a transparent penalty-free wavelength shifting node. In the O/E/O converter, the optical signal is converted to an electrical signal and this signal is used to modulate a laser at the converted wavelength, either directly or externally. Using this technique, conversion at 2.5 Gb/s with an extinction ratio o f >40 dB has been demonstrated. However, as bit rates increase, the cost and power consumption of these modules may make the all-optical alternatives preferable. Additionally, O/E/O modules are designed for a single data rate, making any upgrades require a forklift. The simplest all-optical wavelength converters are based on XGM and have shown impressive performance for bit rates up to 40 Gb/s [7, 8]. The gain of a semiconductor optical amplifier (SOA) depends on its input power. Wavelength 10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. conversion makes use of this dependence. As shown in Figure 2.4, when the input power increases, the carrier density; thus, gain decreases. Since the carrier dynamics is very fast, gain modulation can happen in picosecond scale. Gain modulation as high as 40 Gb/s has been observed. If at the same time with the saturating pump signal a lower probe signal at a different wavelength is input to the SO A, it will see a gain modulation in complement of the pump signal. This is the same effect that produces crosstalk in WDM system amplification. Thus, the pump “F’s will be mapped as “0”s and pump “0”s will be mapped as “F’ s into the probe wavelength. n n filter @ X p r o b e Probe A p r o b e ■ j | j j | | Signal J L T L C arrier ~| n r Density tJ L J Gain I f U " Probe n n r Output LI U Time Figure 2.4. Wavelength conversion by cross-gain modulation in a semiconductor optical amplifier. There are three main drawbacks associated with the XGM scheme. First, the extinction ratio of the converted signal is degraded, especially for the wavelength up-sfaift case (shifting to longer wavelengths). The extinction ratio problem can be remedied by using a two-stage configuration as proposed in [9]. The second drawback is the polarization dependence, which can be alleviated with the configuration used in [10], even if the individual semiconductor optical 11 Signal A p u m p Probe X _ _ o b e SOA Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. amplifiers (SOA) depend on polarization. The third drawback is that XGM induces chirp onto the wavelength-converted signal. This chirp, in turn, induces a power penalty when the converted signal is transmitted over standard-single mode fiber, in particular at high bit rates. However, transmission o f a wavelength converted signal 10 Gb/s over 121 km o f dispersion-shifted-fiber was successfully demonstrated [11]. t o Destructive Interference “0” out ^■probe „ Constructive Interference “I” ont Figure 2.5. Wavelength conversion in a Mach-Zehnder configuration. Depending on the phase difference induced by the pump signal, the probe signals in the two arms o f the output coupler combine constructively or destructively. One of the most promising schemes for wavelength conversion is based on cross-phase modulation (XPM) in a semiconductor optical amplifier in interferometric structures such as Michelson (MI) or Mach-Zehnder (MZI) configurations [12]. In the MZI configuration (Figure 2.5), the optical pump signal controls the phase difference Induced on the probe signal. Depending on this phase difference (out of phase or in phase), the probe signal will re-combine at the output y-branch destructively or constructively. Interferometric converters 12 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. allow high operation speeds [13], low chirp [14, 15] and regenerative capabilities [16], Moreover, these converters are wavelength independent [17]. The main drawback of the interferometric wavelength converters is their relatively small input power dynamic range of 3-4 dB at 10 Gb/s. In a recent report [18], an 8 dB dynamic range was demonstrated using a simple method in which the bias current to the converter was controlled. Additionally, the use of an EDFA before the converter enabled the device to achieve more than 40 dB of dynamic range. This method, however, has a slow control scheme due to the gain recovery time o f the EDFA. Replacing the EDFA with an SOA allows for faster control but decreases the dynamic range to 28 dB. Both o f these methods are very straightforward to implement. pump.) i- A L . j into pump mixing DFG : Cascaded y • X process Figure 2.6. Periodically-poled lithium niobate waveguide mixing operation. A powerful wavelength conversion scheme is based on difference frequency generation (DFG). DFG allows data format independence and quantum noise limited wavelength shifting. Traditional schemes for such devices use 13 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. nonlinear fibers, semiconductor optical amplifiers, or nonlinear waveguides for parametric processes [19]. Recently, a new family of very efficient wavelength converter devices based on periodically poled LiNbOa (PPLN) waveguides has been introduced. PPLN devices operate either in difference frequency mode [20] or in cascaded x2;% 2 mode [21], providing much higher conversion efficiencies when compared to fiber based wavelength converters. Conversion efficiencies up to 0 dB from the longest devices (~6 cm) have been theoretically predicted. In 1 0 practice, normalized efficiencies in the range of 0.5 to 1 W ' cm ' for devices m the 1.5-pm window have been demonstrated. For a typical 4-cm long device, efficiencies in the range o f 10 W '1 can be obtained. With a fixed pump power of about 16 dBm, such a device has a conversion loss of 3 dB, with a linear region of more than 60 dB. The primary drawback o f parametric wavelength conversion is polarization dependence. A number of schemes [22, 23] have been demonstrated to solve this problem. Another attractive wavelength conversion scheme is based on four wave mixing (FWM). FWM offers full transparency, to both modulation-format and bit-rate, which can not be achieved by XGM or XPM. In a recent report [24], using FWM in a long semiconductor optical amplifier, 30-nm wavelength down- conversion and 15-nm up-conversion was obtained at 10 Gb/s. This result is a significant improvement over previously reported FWM based wavelength converters, enabling full EDFA band coverage. 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The main drawbacks of FWM wavelength conversion are low conversion efficiency and polarization dependence. Unlike the conversion efficiency issue, polarization dependence can be reduced using two broad categories of solutions. Techniques based on polarization diversity use either two orthogonally polarized pumps and a polarization-insensitive SOA [25, 26, 27], or one pump and two SOAs [28]. The other techniques are based on band-tailored multiquantum-well (MQW) SOA chips [29, 30]. Both types of polarization-insensitive techniques, however, cause an additional 6-dB loss in wavelength conversion efficiency compared to single polarization schemes. This limits their bandwidth and conversion range performance. While wavelength conversion is a well-researched area with potentially very important implications in achieving all-optical networks, no commercial solutions have been introduced, showing that the research field is still ripe for further efforts. For their very good performance, combined with full optical transparency, PPLN devices, although a very recent technology, seem to be a very likely candidate to provide the best solution to wavelength conversion. 1 5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 3 Variable-Bit-Rate Header Recognition for Reconfigurahle Networks using Tunable Fiber- Bragg-Gratings as Optical Correlators 3.1 Introduction One o f the expected features of next-generation optical networks is the ability to have some measure o f transparency to facilitate high-throughput, flexible, and robust operation [31]. A key operational feature in such an optical network could be the ability to transmit and route optical packets through the network over a wide range of data bit rates. Variable-bit-rate (VBR) networks are attractive since they can accommodate a heterogeneity of data users and traffic. One key function in achieving a VBR transparent packet-switched optical network is to quickly and reconfigurably recognize a set of header bits at different bit rates and compare them to a local look-up table for a switching decision. In general, optical header recognition is important for implementing on-the-fly efficient switching decisions for which electronic techniques are too difficult, and such recognition should be able to accommodate changes in the look-up table at each switch in the network. At present, there are no reported results on variable-bit-rate optical header recognition. Related work includes: (a) bit-rate recognition accomplished entirely 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in the electronic domain [32], (b) bit-rate flexible optical clock recovery using a self-pulsating laser [33], and fixed delay lines as optical correlators [34]. Previously, we have reported a technique for fixed-bit-rate optical tunable header recognition [35] which is also presented in Chapter 3. In this chapter, we extend that result to include variability in the data rate of the incoming packets [36], OC-x to OC-16x 2.5 Gb/s 155 Mb/s 2.5 Gb/s Optical Switch VariaMe-BSt-Rate H eader Recognition Module 155 Mb H»: nth bit o f header [H2jHl[ I H2 I HI 2,-Shift: Time-to- Wavelength Mapping te lH ll I H2 1 h i Wavelength dependent delays as optical correlator and detector H I H2 H I No Match \ — <#-Match 1 H2 f HI f — 7 SAMPLES Decision Electronics Figure 3.1. Conceptual diagram o f optical variable-bit-rate header recognition and switching module. We demonstrate variable-bit-rate recognition of the header information in a data packet. Our technique is reconfigurable for different header sequences and uses optical correlators as a header look-up table to generate the correlation signal for the electronics to switch packets. In our technique, cross gain modulation (XGM) in a semiconductor optical amplifier (SOA) maps incoming packets onto two different control wavelengths. This wavelength translation allows the module to operate in a wavelength independent manner. At the same time, the packet itself is passed on without being wavelength shifted, hence it does not incur any penalties due to chirp or limited extinction ratio. Tunable fiber Bragg gratings (FBGs) provide wavelength dependent delays (Figure 3.1) for the optical 17 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. correlator that performs address matching. Since our FBGs are tunable, we can change our optical decoder to match different addresses as in a reconfigurable optical network. When the header bits match a local optical code in the look-up table, a recognition pulse is produced that operates a 2x2 optical switch. We use this optical technique to recognize the header, and switch a series o f incoming data packets at 155 Mb/s, 622 Mb/s, and 2.5 Gb/s to one of two outputs in a reconfigurable network. Penalty-free routing with a 1.6-ns guard time is achieved. This robust technique is tunable for different header sequences and can scale to accommodate higher bit rates (up to 40 Gb/s) and different numbers of header bits. 3.2 Experimental Setup of Variable Bit Rate Header Recognition Module Figure 3.2. Experimental setup. The experimental setup is shown in Fig. 3.2. A packet stream carrying 2.5- Gb/s, 622-Mb/s, and 155-Mb/s packets with a 1.6-ns guard time is tapped to provide an optical input to the variable-bit-rate header recognition module. 1555.7 1551.9 1548.5 1551.9 1555.7 Pattern .Generator K j, fV . \ , i 1 1 MZ M odulator I'Pi'V > j lilicr[- Optical Switch 1 i — rn i . y @*.jS l 1 — > | 1 SOA |--------------------— 111 l f f © b ^ - H 4 © F B G A rray #2 ^ F B G A rray # ! J k -0 1 /1 0 Decision Circuit Decision Circuit o 18 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Synchronization is assumed, in that the transition time between the two header bits is aligned for all bit rates [37]. The correlation between the address bits is accomplished by using a wavelength dependent delay (provided by the FBGs). Only the control wavelengths (from the wavelength-shifted packet) are used in the correlator. Two FBG arrays are used, the first one to identify header ‘ 15 bits and the second one to identify header ‘O ’ bits. FBGs (4 cm apart to provide a 400 ps delay) reflect the mapped information for detection. To identify header ‘1’ bits, the control wavelengths are reflected from the FBGs in the first array and the result is sampled by the first decision circuit. To identify header ‘O ’ bits, the gratings o f the first FBG array are tuned away, so that they no longer reflect the initial two control wavelengths. Instead, the control wavelengths reflect on the FBGs in the second array and are sampled by the second decision circuit. The correlation information is sampled 6.2 ns after the start of the packet. The detection outputs are NAND-gated and the resulting signal is used to control a 2x2 optical switch for packet switching. For example, to match the header ‘10’, the X 2 grating in the first FBG array is tuned away. Thus, the grating array structure formed, i.e., X \ present (‘1’ bit), and X 2 not present (‘O ’ bit), matches the header ‘10’. Since only X% passes through the first grating, we need just a single grating at X 2 in FBG array # 2. 19 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 33 Experimental Results and Summary 2.5-Gb/s 2.5-Gb/s 622-Mb/s 155-Mb/s -Tv lt c) d) 01 10 B W I , e) Jftg 11 11 r r f '« t r p f f) I WWIWif " I V i, Figure 3.3. (a) Input packet stream to variable-bit-rate header recognition module that recognizes header 6 11’. (b) Grating array #1 output, (c) Grating array #2 output, (d) Decision electronics output, (e) Port 1 output of the optical switch, 2.5- Gb/s packet with header 'O F and 622-Mb/s packet with header ‘10’ pass through, (f) Port 2 output of the optical switch, 2.5-Gb/s and 155-Mb/s packet with header ‘IF recognized and dropped. Figure 3.3 shows the signals in the module for recognition of header ‘IF . Fig. 3.3 (a) shows the incoming packets and headers. Fig. 3.3(b) shows the output from the FBG array #1. In the ‘IF case, only the first grating array is needed, because both channels are reflected by FBG array #1, so the output of FBG array #2 is irrelevant (Fig. 3.3(c)). Fig. 3.3(d) shows the state o f the switch changing from high to low upon detection of a match. Fig. 3.3 (e) and Fig. 3.3 (f) show the 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. port 1 and port 2 switch outputs, indicating successful switching of the packets with the ‘ 1 r header. 2.5-Gb/s 2.5-Gb/s 622-Mb/s 155-Mb/s a) ' vv' w v i \ V *VV_\V \ v V. V v ;- h) ' ■■ ■' i---------------- t---------------- f d) 00 11 ^ u « \ ! < V ' V - V ' ‘ ? 'U j “ ^ W (vti \l ' » i w v I » •,.' /iV ’ 1 ''! - '? ; s 'f '= 10 / • p * r t f * / » r 1 ^ 1 1 ^ 8 4 % v « V i Figure 3.4 (a) Input packet stream to variable bit rate header recognition module that recognizes header ‘10’. (b) Grating array #1 output, (c) Grating array #2 output, (d) Decision electronics output, (e) Port 1 output o f the optical switch, 2.5- Gb/s packet with header '00' and 622-Mb/s packet with header ‘11’ pass through, (f) Port 2 output of the optical switch, 2.5-Gb/s and 155-Mb/s packet with header ‘10’ recognized and dropped. Fig. 3.4 shows the corresponding process for recognition of the ‘10’ header. In this case, a grating at X 2 in FBG array #1 is tuned to match the array structure with the header. Again, Fig. 3.4(a) shows the incoming packets and headers. Header ‘1’ and ‘O ’ bit correlator outputs are shown in Fig. 3.4(b) and 4(c). Fig. 3.4(d) - (f) show the routing o f the packet with the ‘10’ header. 2 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. s b) 200ps d) 200 ps Figure 3.5. a) 155-Mb/s input signal eye diagram with >3 ns rise and fall times, b) 155-Mb/s decision eye after the optical correlator, c) 2.5-Gb/s input signal eye diagram with <200-ps rise and fall times, d) 2.5-Gb/s decision eye after the optical correlator. In Fig. 3.5, the effect o f rise time is explained. Fig. 3.5(a) shows the incoming packet stream eye diagram for 155 Mb/sec with rise and fall times >3ns. This is achieved using a quasi-Gaussian low pass filter with a bandwidth of 117 MHz. Fig. 3.5 (b) shows the three-level decision eye diagram after the optical ‘11’ correlator for the input signal shown in Fig. 3.5 (a). The correlation decision is made setting the threshold in the lower opening. Longer rise and fall times cause closure of the eye in amplitude, not in time. The eye opening is sufficient for decision even with rise and fall times longer than 3 ns. Fig. 3.5 (c) and Fig. 3.5 (d) show the corresponding eyes for the 2.5 Gb/s case. As expected, shorter rise and fall times provide better decision eye openings. a) 6.4 ns c) 400 ps 22 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. w O Before Switch ■ Port 1 Output ^^ O T t^O ut£U t_ 3 4 5 6 7 8 9 1 0 -27 -26 -29 -28 - 3 1 -30 O Before Switch Port 1 Output X Port 2 Output Optical Power (dBm) -31 -30 -29 -28 -27 -26 Optical Power (dBm) Figure 3.6. BER curves for (a) ‘1 1 % and (b) ‘10’ header recognition indicating no degradation using our module relative to back-to-back measurements. Fig. 3.6 shows BER curves for packet streams before and after the optical switch. The BER curves with our variable-bit-rate header recognition and switching module indicate no significant degradation relative to the back-to-back BER measurements. 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 4 Tunable All-Optical Time-Slot-Interchange and Wavelength Conversion using Differenee- Frequency-Generation and Optical Buffers 4.1 Introduction It may be critical for future networks to provide additional switching functionality in the optical physical layer to ensure high-speed and high- throughput performance. Typically, switching is performed in one o f the following domains: time, wavelength, or space. As traffic grows, the network may require the use o f more than one o f these switching domains concurrently in order to meet the increasing demand. Improved flexibility, throughput, and robustness would thus be achieved in future heterogeneous networks that combine wavelength-, time-, and space-division switching for the data traffic. In a network that includes elements o f time-division multiplexing (TDM), the time slot of each bit defines its output port at each switching node. This makes time-slot-interchange (TSI) the most commonly used method of switching in the time domain [38]. TSI is achieved by shifting a given bit from one time slot into a different one, thereby changing its destination. This can be readily accomplished in the electronic domain by shift registers. However, accomplishing TSI in the optical domain may be highly desirable to avoid optoelectronic data 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. conversion as well as the potentially limited switching speed o f electronic circuits [39]. Similarly, all-optical wavelength conversion would enable a highly-efficient, reconfigurable network switching architecture with wavelength reuse and contention resolution. Two reported techniques for implementing all-optical TSI include: (i) using fixed delay lines as optical time buffers and semiconductor optical amplifiers (SOA) as gates to perform TSI with a 1.6 ns guard time at 125 MHz [40], and (ii) using lithium-niobate space switches followed by fixed delay lines to perform TSI on a 29-Mbit/s signal [41]. The SOA used in the first method has limited switching speed, generates crosstalk due to a limited extinction ratio, and adds noise to the signal. The space switches used in the second method generate signal crosstalk and have scalability problems. Crosstalk is a particularly serious problem in time switching because coherent effects are generated which may induce large power penalties and bit-error-rate floors [42]. On the other hand, all-optical wavelength conversion has been accomplished by several methods, including: cross-gain modulation [43], cross phase modulation [44], and four-wave-mixing [45]. However, these methods generally use an SOA as their wavelength-shifting medium which typically gives rise to at least one o f the following disadvantages: (i) limited conversion speed, (ii) limited wavelength range, (iii) additive noise, (iv) limited output extinction ratio, (v) induced chirp, and (vi) narrow dynamic range of input power. 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This chapter illustrates how the combination of all-optical time-slot- interchange and wavelength conversion o f a 2.5-Gbit/s data signal is used to achieve a reconfigurable time/wavelength switch. A combination of difference frequency generation (DFG) as a wavelength converter and tunable fiber Bragg gratings (FBG) as optical time buffers are employed to generate switch elements. The swapping of adjacent time slots is achieved by using three such switch elements, in which: (i) the first wavelength converter places the odd-numbered bits at a new wavelength, (ii) FBGs introduce a 2-bit delay between wavelengths, and (iii) the second wavelength converter places the non-delayed even-numbered bits to the new wavelength. DFG wavelength-conversion is achieved in periodically poled lithium-niobate, which has the following advantageous properties: (i) adds negligible spontaneous emission noise, (ii) operates with negligible chirp, (iii) has similar up- and down-conversion efficiency, (iv) induces small crosstalk and high extinction ratio at the output, and (v) exhibits >THz bandwidth [46]. Furthermore, tunable FBG optical buffers provide better scalability and are more reconfigurable than conventional buffers based on fiber delay lines. 4.2 Experimental Setup Fig. 4 .1 shows the conceptual diagram o f the swapping o f adjacent time slots. In the first wavelength/time switch, odd numbered time slots are selected and copied from ko to X i by means o f DFG. A wavelength-dependent matrix of discretely tunable time-slot delays, as defined by the tunable set of FBGs, 26 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. introduces a 2-bit delay between X q and Xi, thereby acting as optical time buffers. Then, the second wavelength/time switch selects and copies the even numbered time slots to the already-delayed X i bit stream, resulting in all-optical time-slot- interchange and wavelength conversion. 2-bit delayed slots @ X Time slot L U L 1 L 2 1 1 J . » 1st X - converted slots @ ^ ^ Q i— nr—____ ,rn „m.. . i3ui it2 i * + X .conversion TDM input signal @ Xa ' ^ 1 4 1 3 1 2 l 1 FBGs: 2-bit delay between X {) & A - i 1 4 1 312l 11 1, s i g n a l® ^ 2nd X conversion * E3 -.[I3. > original signal @ X u fTl [2]^ Pulse train @X„ Pulse train @ X p Figure 4.1. Conceptual diagram of our time slot interchange and wavelength conversion experiment. DFG wavelength-conversion is achieved with a periodically poled lithium- niobate device. The device employs a x(2 ) : X ( 2 ) process and creates a new wavelength at A dverted ~ 2 X p u m p - X s i g n a i - DFG is a nearly instantaneous process with a high operational bandwidth in excess o f several THz. In a x(2 ):X (2 ) process, the pump signal generates a second harmonic, which then interacts with the input signal converting it to a new wavelength. The conversion efficiency is proportional to the square of the pump power. It should be emphasized, therefore, that any increase in pump power translates into twice as much improvement in conversion efficiency, i.e. 1 dB into 2 dB. Experimental setup is shown in Fig.4.2. A 2.5-Gb/s bit stream at 1555 nm is input to the time-slot-interchange/wavelength conversion module. Inside the module, the +19-dBm DFG pump signal is generated using an external cavity laser at 1550 nm that is modulated by a 2.5-Gbit/s rectangular pulse train. A 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. WDM coupler is used to combine the pump signal and the incoming bit stream as well as to filter out the ASE noise o f the high power EDFA. The "O N" bits o f the rectangular pulse train (i.e., the high levels) are time-synchronized with the odd numbered data-bit time slots at the input to the first wavelength converter. When the pump power is high, a duplicate of the 1555-nm signal is produced at 1545 nm, effectively selecting and copying the odd numbered time slots to a new wavelength. Wavelength conversion efficiency is -16 dB, but it should be emphasized that -4 dB conversion efficiency is possible by using higher pump powers [47]. Present at the output of the first wavelength converter are the original signal at 1555 nm, the pump signal at 1550 nm, and the newly created odd-numbered time slots at 1545 nm. Signal Input @1555 nm MOD ~ ¥ ~ Pattern Generator r' Wavelength Converter WDM Coupler - ► C p — ► 1st PPLN DFG device Wavelength Dependent Dela’ v s 1 |@ 1555nm @ 1545nm Pump 01550 nm MOD t n n n Pulse Train 2nd PPLN DFG device Coupler BPF @1545 nm 2nd Wavelength Converter Output @1545 nm ^ Time slot interchange and wavelength conversion module Figure 4.2. Experimental setup. The fiber Bragg gratings provide a delay o f two bit times (800 ps) for the 1545 nm signal with respect to the 1555-nm signal and filter out the pump signal at 1550 nm. The wavelength dependent delays provided by the FBGs form the 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. optical time buffers. By tuning the gratings, wavelengths can be reflected from different gratings, resulting discrete time delays from the FBG array structure. In the second wavelength converter, the high level o f the rectangular pulse train is time-synchronized to the even numbered time slots. The second PPLN works under the same operating conditions as the first PPLN, in that it copies only the even numbered time slots to 1545 nm. After the second PPLN, a bandpass filter at 1545 nm is used to filter out the pump (1550 nm) and the original signal (1555 nm), leaving only the wavelength converted TSI signal. The x(2 );X ( 2 ) process of wavelength converters result a very high extinction ratio for time/wavelength switch. Thus, the crosstalk between the non-selected even-numbered time slots from the first DFG and the selected even-numbered time slots from the second DFG is minimal. 4.3 Results and Discussion Fig. 4.3 shows the waveforms at the various stages o f time switching and buffering in the first wavelength converter and FBG array. Odd numbered time slots are selected and copied at the first wavelength converter/time switch, and are delayed by the optical-buffer FBG array. Fig. 4.3(a) shows the incoming bit stream (1555 nm) with odd and even numbered time slots. The rectangular pulse train at 1550 nm (Fig. 4.3(b)) is the pump input to the first wavelength converter and high level is time-synchronized to the odd numbered time slots, selecting and copying them to 1545 nm. Fig. 4.3(c) shows the newly created 1545-nm bit stream with odd numbered time slots only. The even numbered time slots are not 29 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. selected for wavelength conversion because they are aligned with the low "O FF" level o f the 1550-nm pump. Fig. 4.3(d) shows the 1545-nm signal after the FBG grating array, delayed by two bit times with respect to the 1555-nm original signal. 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 45 67 890 Slot N um bers 00 0 10 1 10 1 1 1 1 1 0 1 0 0 0 1 1 101 1 0 0 0 0 1 0 B it Stream (a) (b ) v*. . v*. - V5 . V 1 . . V i '•v ' * w < * .• ’ Pulse Train o - o - o - 1- 1- 1- 1- 1- 0- 1- 1- 1- 0- 0- 1- Odd Slots (c) 1 - 0 - 0 - 0 - 1- 1- 1- 1- 1 - 0 - 1- 1- 1 - 0 - 0 - Odd Slots-delayed (d) Figure 4.3. (a) Incoming bit stream at 1555 nm, (b) rectangular pulse train (pump) at 1550 nm, copying odd numbered time slots to 1545 nm, (c) 1545-nm DFG signal after first PPLN device, (d) 1545-nm signal after the FBG array, two bit time delayed with respect to 1555-nm signal. The input and output waveforms o f the second wavelength converter are shown in Fig. 4.4. In the second wavelength converter, only the even numbered time slots are selected and copied. Fig. 4.4(a) shows the original 1555-nm bit stream as the input to the second wavelength converter. Fig. 4.4(b) shows the 1545-nm signal input to the second wavelength converter with no even numbered time slots present. The rectangular pulse train at 1550 nm (Fig. 4.4(c)) is now time aligned to the even numbered slots so that only these slots are selected and 30 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. copied to 1545 nm. Fig. 4.4(d) shows the 1545-nm output signal after the converter and bandpass filter. The empty, even-numbered slots of the 1545 nm signal are now filled in with the even-numbered slots from the original 1555-nm signal. Since the odd numbered time slots were buffered by two bit times before entering the second converter, the time slots o f the original signal are interchanged and wavelength converted at the output o f our module. 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 Slot N um bers 0 0 0 1 0 1 101 1 1 1 1 0 1 0 00 1 1 101 1 0 0 0 0 1 0 B it Stream (a) 1 - 0 - 0 - 0 - 1 - 1 - 1 - 1 - 1 - 0 - 1- 1 - 1 - 0 - 0 - Odd slots-d elayed ' P ulse Train 1 0 0 1 0 1 0 0 1 1 1 1 1 0 1 0 1 0 0 1 1 0 1 1 1 0 0 0 0 0 T im e slot interchanged and ( j ) | w avelen gth converted Figure 4.4. (a) 1555-nm input signal to second PPLN, (b) 1545-nm input signal to second PPLN, (c) rectangular pulse train (pump) at 1550 nm, copying even numbered time slots to 1545 nm, (d) 1545-nm signal output from second PPLN device, time slots interchanged and wavelength converted. 3 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Fig. 4.5 shows the BER curves for the signals before and after the time/wavelength switch with <4-dB power penalty. This penalty is primarily due to the low wavelength-conversion efficiency and various insertion losses, which can be reduced by higher pump powers and better pigtailing. Moreover, it is possible to increase the selecting pulse’s extinction ratio by >30 dB. This will result in time-switching extinction ratios in excess of 50 dB, effectively eliminating crosstalk interference. — After switch -® ~Back-to-back w > -40 -38 -36 -34 -32 -42 Optical Power (dBm) Figure 4.5. BER curves for back-to-back and after our time-slot-interchange and wavelength-conversion module. 32 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 5 All-Optical Header Recognition of WDM Channels Using a PPLN Wavelength Shifter and FBG-Rased Correlators All-optical packet-header recognition of multiple WDM channels is accomplished simultaneously in a single module. The technique uses: (i) a periodically-poled lithium niobate (PPLN) waveguide to wavelength-shift several header bits from multiple incoming WDM packets, and (ii) fiber Bragg grating arrays that perform optical correlation between the wavelength-shifted incoming header bits and the desired look-up-table pattern. A key feature o f the PPLN that significantly reduces the necessary number of components is that it can accommodate several independent WDM data-packet streams and several independent pumps, with the number of header bits determined by the number of pumps. We demonstrate the WDM header recognition module by simultaneously recognizing two header bits on each of two 2.5-Gbit/s WDM packet streams. These packets are routed independently to different output ports of a 2x2 optical cross-connect, with a packet guard time of 1.6 ns. The FBG arrays can be tuned to enable reconfigurable look-up tables. 33 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.1 Introduction The next generation o f high-performance optical networks may require data packets to be rapidly routed by all-optical switches. In such a scenario, it may be desirable to recognize the header bits o f each data packet to enable efficient and high-throughput switching [48], One brute-force method is to tap off a small portion of the signal and electronically detect the bits. This approach, however, may not be suitable for high-bit-rate packets for which electronic techniques do not exist. A potentially faster approach is to decode and recognize the header bits all-optically so that a given routing decision can be made on-the- fly [49], Previously, we demonstrated a technique for all-optical header-recognition that used X-shifting in a semiconductor optical amplifier (SOA) and tunable fiber Bragg grating (FBG) arrays as optical header-bit correlators [50]. This technique, however, is not easily scalable to multiple WDM packet streams since an SOA can only accommodate a single input data stream. The header processing of N WDM channels would require wavelength demultiplexing and N distinct SOA modules, which is perhaps too many components for practical WDM network implementation. 34 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. # of Header Bits Detected = M # of WDM Channels = N SOA PPLN # of X shifters N 1 # of FBG Array Correlators N*2m 2m # of FBGs per array 2M 2M Total # of FBGs N*2m*2M 2m*2M Total # of Components N+(N*2m*2M) 1+(2m*2M) Table 5.1. Comparison of the # o f components required for the SOA vs. PPLN WDM-HR implementations. We demonstrate simultaneous all-optical packet-header recognition o f two WDM channels at 2.5 Gbps in a single module. The technique uses: (i) a periodically-poled lithium niobate (PPLN) waveguide [4] to wavelength-shift several header bits from multiple incoming WDM packets, and (ii) fiber Bragg grating arrays that perform optical correlation between the wavelength-shifted incoming header bits and the desired look-up-table pattern. The PPLN accomplishes wavelength shifting by means of difference frequency generation (DFG) in which the X-shifted signals are the mirror images o f the input signals, with respect to the pump. A key feature of the PPLN that significantly reduces the necessary number of components is that it can accommodate several independent input WDM data-packet streams and several independent pumps. The number of pumps determines the number of header bits of each packet. These packets are routed independently to different output ports of a 2x2 optical cross-connect, with a packet guard time o f 1.6 ns. This technique can be 35 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. efficiently scaled to accommodate more WDM channels at higher bit rates, and the FBG arrays can be tuned to enable reconfigurable look-up tables. 5.2 Operational Concept A conceptual diagram for the placement o f the WDM-HR modules in an optical cross-connect is shown in Fig. 5.1. The optical power from the WDM channels on each incoming fiber is tapped off and directed to a WDM-HR module. The module simultaneously tests all possible header bit combinations for each WDM channel using multiple arrays of FBG-based optical correlators. v . v Optical Port A v , v WDM-HR control electronics WDM-HR PortB 5 * 5 A s ' Port C Optica Switch for ^ PortD 2x2 WDM Optical Cross Connect Figure 5.1. 2x2 optical cross connect implemented with WDM-HR modules. Fig. 5.2 traees through the basic architecture of the WDM-HR module for the case of two WDM channels with two header bits. The incoming WDM signal is tapped off and mixed in the PPLN with two pump signals that are gated to X - shift only the header bits on each channel. With each header bit now on a unique wavelength, the signal is passively split and directed to multiple FBG arrays to test each possible combined header-bit combination o f wavelength channels. The 36 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. FBG correlators provide wavelength dependent delays, stacking the header bits on top of each other to produce the output correlation peak. Each FBG array is uniquely configured to produce an autocorrelation peak for a particular header pattern on each incoming wavelength. Input header bit ; selectors : I ! ! r ^-conversion FBG Array Correlators 4 A-shifted header bits 2-Channei PPLN / data H2 data H2 Decision Circuit WDM Signals; M = # header bits X12 = Xt, header bit 2 Decision Circuit Match/ No Match? Switch Controller j Match/ INo Match? Figure 5.2. Conceptual diagram o f a WDM-HR module. Each FBG array is uniquely configured to test for a different header pattern on each wavelength. HI = 1s t header bit; % n ~ channel 1 ,2n d header bit; A . p i = pump #1 (header bit selector #1). To test all possible header patterns, there must be two complementary correlators in each FBG array. The first, called the “ones correlator,” tests for the presence o f the desired ‘ 1 ’ bits, and the second, called the “zeros correlator,” tests for the presence of the desired ‘0’ bits. With only one of these correlators, a ‘101’ pattern would produce the same correlation peak as a ‘110,’ causing ambiguity. Simple threshold detectors are then used to sample the outputs o f the two correlators at the packet rate (MHz). A NAND gate tests the two threshold- detector outputs and produces a high signal to indicate a ‘match,’ and a low signal 37 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. to indicate a ‘no match’ to the pattern that the FBG array was configured to recognize. The switch controller determines the header information of each WDM channel by detecting which FBG correlator array produces a ‘match’ signal. The controller then configures the switches to appropriately route the channels according to its local look-up table. (The PPLN waveguide has the following advantages over an SOA-based X-shifter: (i) negligible spontaneous emission noise, (ii) negligible chirp, (iii) similar up- and down-conversion efficiency, (iv) low crosstalk, (v) high extinction ratio, and (vi) >THz, bandwidth [51].) There are two ways to implement the FBG arrays, both of which require the use o f sampled FBGs to reflect multiple wavelengths at the same point in the fiber. The first, which is demonstrated in this letter, requires a separate FBG array to test each combination of header bits and wavelengths. If there are m header bits and n WDM channels, then there are m * n total header bits, requiring 2 m *n unique FBG arrays. The number o f FBG arrays can be reduced to 2 m if the correlators are instead configured to serially output the autocorrelation peaks for each WDM channel. In this case, a wavelength-dependent, one-bit-time delay is placed before the correlators so that the ki header is tested one bit time before X 2, and so forth. Each FBG array therefore produces a separate match/no match output for each wavelength channel. A comparison o f the total number of 38 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. components required (1-shifters + FBGs) for the PPLN versus the SOA implementations o f WDM-HR is shown in Table 1. 5.3 Experimental Setup Our experimental setup is shown in Fig. 5.3, using two WDM input signals with 2-bit headers. Two pump signals are modulated with a single ‘1’ bit followed by ‘O ’ bits for the duration of a packet. They are then amplified and delayed by one bit time through a dispersive fiber. The first pump is aligned with the first header bit and the second pump with the second header bit at the input to the PPLN waveguide. Via the cascaded x(2 );X (2 ) nonlinear process, the two header bits from each signal mix with the two pumps in the waveguide to generate four DFG wavelengths, each containing one header bit. Figure 4 shows the spectrum of the incoming WDM channels, the two pump signals, and the four 1-shifted header bits. The DFG signals then pass through two FBG arrays composed o f a zeros and a ones correlator that are configured to test for a particular header pattern on the two WDM channels. In our case, the correlators are configured to match a ‘10’ on X \ (1st incoming header bit is “0” and 2n d header bit is “1”), and a ‘ 11’ on li- Then, in the ones correlator, the first grating must be a sampled FBG that reflects both X u (signal 1, 2n d header bit) and X 22, since we are looking for the second header bit on both wavelengths to he a 6 F. The second FBG will reflect only X 21 since we are looking for a ‘ 1’ in the first header bit of X 2 and a ‘O ’ on A 4 . 39 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. data #1 WDM ^ channels ^ 2 MOD MOD T ~ data #2 header I C A l A T f f t r pumps L EDF^>—[ m o d - czzgat' PPLN U & > . S- _« « 9 » > T S £ -ts t A O.^ .3 © FBG array correlator zeros ones correlator correlator 7?L H qHHJ-©- 4 f* gated pump bits header header bit #2 bit #1 ,rFBG FBG decision circuit control signal toOXC Figure 5.3. Experimental setup o f a 2-channel WDM-HR module for recognizing a particular combination of 2-bit packet headers. The threshold detector is set to detect a ‘level 3’ signal which occurs when all three ‘ 1 ’ bits are stacked up by the FBG delays. Conversely, the gratings in the zeros correlator are configured to reflect when looking for a ‘O ’ bit and to transmit otherwise (and the threshold detector is always set to detect a ‘0’ level). Thus, the first grating can be absent since a ‘ 1’ bit is desired for the second header bit on both channels, and the second grating would be designed to reflect A ,n only. The choice o f gratings is therefore different for each header pattern and multiple grating arrays are required to test all possibilities. 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. channels 1s t WDM X pl V J_5 dB K head,S header bit selector selector ' ■ v d X-shifted header bits 1542 1551 1560 W a v e l e n g t h (nm) Figure 5.4. PPLN output spectrum showing the two input WDM channels, the two header bit selector pumps, and the four X-shifted header bits. X n = channel 1, 2n d header bit; X p = pump. The FBGs are separated by 1/2 of a bit-time so that the round trip time between gratings is a full bit. The first incoming header bit is therefore delayed by one bit-time, causing it to stack on top of the second header bit at the circulator output. Using a timing signal that would originate from a packet arrival detection circuit in an actual cross-connect, the two correlation outputs from the ones and zeros correlators are sampled at the packet rate and combined with a NAND gate to provide a simple ‘match’ or ‘no match’ signal to the switch controller. 5.4 System Demonstration We utilize the WDM-HR module to demonstrate independent routing of two WDM data-packet streams to different output ports of a 2X2 switch. The two WDM packet streams at 2.5 Gbits/s have 53 bytes per packet with 2-bit headers 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and a 1.6 ns guard time between packets. LiNb03 switches are used to implement the 2x2 optical cross-connect (OXC). The signals are input to port A and can be output to either port C or D. In this experiment, a single FBG array is used to demonstrate header recognition for the header pattern “01” on Ai and “10” on A 2. The following lookup table is used to determine which headers correspond to the two cross-connect output ports: 1 Header bit pattern Ai Output Port A 2 Output Port 01 10 D C else Else C D The results of the system demonstration are shown in Fig. 5.5. The packets on Ai and A 2 enter the 2x2 OXC on port A (Fig. 5.5a). After mixing with the pump header selectors (Fig. 5.5b) and traversing the correlators, the decision circuit outputs a high level when there is a match (Fig. 5.5c). Figures 5d and 5e show the successful routing of the “matched” packets on A i to port D and A 2 to port C. Since the WDM-HR module operates on optical power tapped from the main data stream, it does not degrade the bit-error-rate (BER) performance o f the transmission link. BER measurements o f the main signals transmitted through the optical switches show no power penalty due to the WDM-HR module. 42 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (b) (c) (d) (e) L L r i 1s t header selector 2n d header selector control signal OXC > - port C output OXC >-portD output Figure 5.5. Results of WDM-HR system demonstration. The FBG array is configured to correlate with header ‘01’ at A ,i and header ‘ 1 O ’ at X 2, and the 2x2 OXC is programmed according to the above lookup table, (a) Input WDM channels at X i & (b) header bit selector pumps, (c) decision circuit output (used as the switch control signal), (d) port C output at X \ & X 2, (e) port D output at X i & X 2. 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 6 Simultaneous and independent label swapping of multiple WDM channels in an all-optical packet- switched network using PPLN waveguides as wavelength converters 6.1 Introduction We demonstrate simultaneous and independent label swapping and wavelength conversion of many WDM channels using PPLN waveguides as 1 - shifters. Our method is bit-rate, label length, and format independent. We demonstrate label swapping of distinct 8-bit-long labels for 2 WDM data channels at 10 Gbit/s with a guard time o f400 ps and a power penalty <3 dB. This method can potentially accommodate 10 WDM channels simultaneously over the PPLN’s -4 0 nm l-shifting bandwidth. Next-generation packet-switched all-optical networks may require the use of labels to efficiently route packets to an appropriate destination, such as in multi-protocol label switching (MPLS) [52], Such labels can be located in the digital baseband data stream or on an ancillary subcarrier [53,54]. One o f the requirements of using a label for routing is that the label should be able to be read rapidly and on-the-fly. Various all-optical techniques for rapid label recognition have been presented [55,50], However, a truly flexible and reconfigurable optical 44 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. network also necessitates the ability for the label be rapidly replaced, or "swapped." Previous reports o f all-optical techniques for label swapping include: (i) employment o f XOR logic in an integrated interferometric wavelength converter [53], (ii) incorporation o f fiber Bragg grating filters to separate and replace the subcarrier labels using dispersion-induced fading [56], and (iii) use o f a Fabry- Perot filter as a notch filter for erasure and replacement of subcarrier labels [57, 58]. However, no reported method to date has shown the functionality of simultaneous and independent label swapping of many WDM channels in a single subsystem. p a y lo a d old label “ ■ . in ---------------- ii ; — 1 iii f I I IN P U T s y n c h r o n o u s W D M c h a n n e ls i[ p a c k e t p a y lo a d to A , m ap p in g u sin g a P P L N as a X -shifter label recognition and synchronous new la b els Iprocessina J .... p a c k e t la b e l to X m ap p in g usin g a P P L N as a X -shifter t new i p ayload label out payload new O U T P U T X -sh ifte d a n d la b e l-s w a p p e d W D M c h a n n e ls i s % _M u 11 j - w a v e le n g th J r te ljw a p jg iH g _ m ® d n ie , Figure 6.1. Old labels are removed and the payload and new labels are mapped onto new wavelengths. The new label is inserted into the place o f the old one to generate the wavelength- shifted and label-swapped output signal. We demonstrate the simultaneous and independent label swapping of many WDM channels using periodically-poled lithium niobate (PPLN) 1-shifters. The unique property of the PPLN for this application is that it can accommodate multiple independent input WDM data packets. The PPLN performs difference 45 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. frequency generation (DFG) in which all input wavelengths are mapped to their mirror images with respect to a pump signal. The PPLN waveguide: (i) adds negligible spontaneous emission noise, (ii) operates with negligible chirp, (iii) has similar up- and down-conversion efficiency, (iv) induces negligible crosstalk and high extinction ratio at the output, and (v) has >THz bandwidth. The pump signal is time-gated so that only the payloads o f the WDM packets are 1-shifted, thus ‘removing’ the old labels. A second PPLN 1-shifts the new labels to the same wavelength as the X -shifted payloads. These two signals are optically combined, forming the label swapped packet as shown in figure 6.1. We demonstrate successful label swapping and X-shifting of 2 WDM channels by replacing the labels of individual channels independently. Our method is bit-rate, label length, and format independent, and can potentially accommodate up to 10 WDM channels over the PPLN’s ~40 nm X-shifting bandwidth in the same module. 46 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.2 Experimental Setup Our experimental setup is shown in figure 6.2. There are 2 WDM input channels at 10 Gbit/s with 8-bit labels in 60-bit-long packets with a 4-bit guard time (400 ps) between each label and the payload. The guard time is needed to prevent any potential distortion from the rise (or fall) times o f the low frequency payload and label selector signals. It is assumed that the first bits of each packet o f the WDM channels arrive at the label swapping module at the same time and that their labels are aligned with each other. In this subsystem, we use 2 identical PPLN waveguides (i.e., designed to have the same pump wavelengths and phase matching temperature) manufactured and pigtailed on the same substrate. The PPLN waveguides perform a cascaded x(2): % (2 ) process that produces difference frequency generation (DFG), which is much more efficient than four-wave d=1546.1 pc decorrelation fiber ► jpP L N -1 » BPF Xl-1547. 10 Gbps EDFA pattern generator label removal p c i p C ^H »[ m o d | — P > -^ JEDFA P C * EDFA um »gq5L -f> EDFA pc " iK L new label ^-shifting * |PPLN-2|— > 1 BPF out =1554.5 out ^ 2 = 1 5 5 6 .1 label swapping module Figure 6.2. Experimental setup. The payloads and new labels are wavelength converted by individual PPLN waveguides. The outputs are filtered and combined to produce the label swapped WDM channels. mixing and can have minimum required pump powers as low as 50 mW [59]. 47 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The data channels are generated using the same modulator and are subsequently decorrelated by 1-bit time (100 ps), using a spool o f DCF fiber. The payload-selector pump is a square wave at the payload rate. It is modulated with “1” bits for the duration of the payload and “0” bits for the duration o f the label. In order to remove the old labels, only the payloads of the packets o f both channels are wavelength converted using PPLN-1, as shown in figure 6.2 (label removal module). The new-label-selector pump is modulated by the inverse pattern o f the payload-selector pump. Since we do not have an additional pattern generator to generate the new labels, we place the new label bits within the payload of the input packets so they could be extracted for use as new labels. Fiber delays are used to align the new labels with the label-selector pump at the input to PPLN-2 (new label ^-shifting module). This PPLN wavelength-shifts the new labels to the same wavelengths as the payloads shifted in PPLN-1. The DFG outputs of both waveguides are then filtered and amplified. The A,-shifted payloads and the new labels are then combined such that the new labels are located in the same position as the old labels. 6.3 Results and Discussion The input and output PPLN spectra are shown in figure 6.3. The payload- selection spectra are shown in 3(al) and (bl). Similarly, 3(a2) and (b2) show the label-selection spectra. The input spectra to the PPLN waveguides, which is comprised of the 2 WDM input channels and the pump, are shown in 3(al) and (a2). At the output of the PPLNs, the spectra include the wavelength converted 48 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. payloads, 3(bl), and labels, 3(b2). The converted signals are mirror images of the input wavelengths with respect to the pump. The spectra o f the two X-shifted WDM channels containing their new labels axe shown in 3(cl) and (c2). These output spectra are obtained after combining the wavelength converted payloads with the new labels. The new labels are aligned to the position o f the old labels. (cl) (c2) X-shifted with new labels new label X-shifted pump Wavelength (nm) Figure 6.3. The spectrum of the label swapping subsystem, (al), (a2) are the input spectra to the PPLN waveguides, payload and label-selection spectra, respectively, (bl), (b2) are the DFG spectra after the PPLN waveguides, (cl), (c2) are the label swapped and X-shifted outputs. Figure 6.4 shows time domain waveforms at each stage o f the experiment. The waveforms in figure 6.4(a) and (b) are the input WDM channels. The new- label-selector waveform, 4(c), is aligned with the new labels that are coded in the packet payloads. The payload-selector, 4(d), is gated “on” only during the payload portion o f the packet. The wavelength-shifted payloads o f the packets are shown in 4(e) and (I). Wavelength converting only the payloads removes the original labels. Figures 6.4(g) and (h) show the label swapped packets. A magnified view of the input and label-swapped packets o f the 2 WDM channels is shown in figure 6.5. The packet patterns for Xj with the original label and the new label are shown in 5(a) and (b). Figures 6.5(c) and (d) show similar waveforms 49 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. for % 2- As can be seen from the figures, the wavelength converted and label swapped packets do not have any significantly visible added noise or distortion. ; • ' , ' ’ 1 > ^ ^ i ( a ) 1 '- T , ’:.v ' I . . £ f „ \ ' ■ pi' ^ i I ' ; ■ ; ; f : ! : ! S label I j i. ; ! > i i j ^_^jelectors ’ ‘ ' i >*'*'$! •p'i' Yvr v i -ijy i'f ■ ■ V - t<* ) | v J [ ■ ;• ? '* j 3 ? “ ■ y \ ' ; * '■ f f payload ! f ]:'!;■ i r ; i selectors ! ■ « t .yv<» 3 * . f **; < < • 1 1 fin' ■ ' ilf*, ; UviM*** * ' ^ i !W?' »/' * I fl* l 0U^f , 1 : iT-LliT j." : - A :';:= ! -" I T ■ ; ^1 label ^ i - 4 i ' PI A * rei”oved ; ! : I i 1 ; „ ! i . # s O U l w v v \: X 2 label -T: removed " ........? , f! M /-f ' i f j ! 1 label <g>' ' ■ - I * . V ,1 iV ‘" | t! ; 1 ’ - - P ' 1 1 ^ ; aPPe d out , ' . i t , ' ' ,■ 0 l ^ l t h ' r' " .'C ’l '\ ' A'T ; > " li, ' label (h)f,> 1 ' '‘j.1 '* I , '« l. '• I 1 " j “PI I swapped q u j 1 \ U v.'*J V --t' -f > .• ’• > ' • U 1 ' , - ^ r .^ ^ output % 2 800 ps time Figure 6.4. (a), (b) Four input packets o f the 2 WDM channels, (c) The new-label-selector waveform that is used as the wavelength- converter pump in the first PPLN. (d) The payload-selector waveform to wavelength-convert the payload of the packets, (e), (f) The wavelength-converted payloads for each channel, (g), (h) The label-swapped output waveforms for each channel. 50 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. input packet p a y lo a d old label 01100111 f\ n \ label-swapped packet ttew 1 ^ ' payload old label m o n o old label m o n o Xj label-swapped packet input packet 00101001 new label new label 01010001 Figure 6.5. Waveforms o f a single packet from WDM channels 1 and 2 before and after label swapping, (a), (b) The original label on X u 01100111, is swapped with new label, 00101001. (c), (d) The original label on X 2 , 1110110, is swapped with new label, 01010001. Figure 6.6 on the next page shows the power penalty introduced by this module. The wavelength conversion using the PPLN waveguide introduces a penalty of about 2 dB for this system. The wavelength shifting of the payloads introduced a power penalty o f about 1.5 dB for A ,i and about 2 dB for X%. The insertion of the wavelength-converted new labels contributed about 1 dB for X 2 and about 2.5 dB for X u The additional penalty from insertion of the labels is due to a misalignment of the bandpass filters with the output signals and is not inherent to this method. 5 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. -log(BER) 2 3 4 5 6 7 8 9 1-10 -8 -6 -4 -2 received power (dBm) Figure 6.6. BER curves for the module. Label swapping introduces a penalty of 1 dB and 2.5 dB for the WDM channels A ,i and X 2 , respectively. AA-sbifted payload of A ., O/.-shifted payload of A ^ A A-shifted & label swapped . .oul a i A, • A-shifted & labeLswapped at Aj Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 7 All-Optical Wavelength and Time 2-D Code Converter for DynamicaUy-Reconfigiirable O- CDMA Networks using a PPLN Waveguide 7.1 2-D OCDMA Networks Utilizing Code-Converters There has recently been much renewed interest in optical code-division- multiple-access (O-CDMA) due to its potential for enhanced data security and spectral efficiency, especially when considering the fine granularity o f traffic in local-area-networks (LANs). However, a key drawback for O-CDMA has been the necessity of generating, propagating, and detecting extremely short chip times (i.e the time- domain subdivisions of a bit) such that there are sufficient orthogonal codes. One approach for alleviating the small chip time has been the introduction of a two-dimensional O-CDMA architecture, in which each bit is sub-divided into a combination o f chip times and a discrete set of wavelengths. Even with a time/wavelength approach, a reasonable number o f wavelengths and chip times cannot accommodate many simultaneous users. Therefore, it may be of great value for an O-CDMA network to re-use a finite set o f 2-D codes across different parts of the network. Moreover, such code re-use, which is analogous to wavelength re-use in a WDM network, should be reconfigurable in order to account for changing traffic patterns and to alleviate congestion. 53 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In general, a 2-D code converter would need to redistribute the optical energy in both dimensions, namely, the chip times and the wavelengths. A brate- force electronic approach for code conversion would be to decode the O-CDMA signal using autocorrelation and threshold detection, change the code in the electronic domain, and then re-encode the data on an optical signal. A potentially more rapid, efficient and transparent approach for high-data-rate signals is to perform the code conversion in the optical domain. Although there were generic demonstrations o f all-optical wavelength conversion for wavelength routing in WDM networks and separate demonstrations of all-optical time shifting for time slot routing in TDM networks, there has been no reported demonstration of an all-optical O-CDMA 2-D code converter. We demonstrate all-optical, wavelength and time, code conversion for OCDMA networks at a user data rate of 2.5 Gbit/s with 4 chips/bit and 2 wavelengths/code. Difference frequency generation (DFG) in a periodically- poled lithium-niobate (PPLN) waveguide enables wavelength conversion and fiber Bragg gratings (FBGs) are used to provide cyclic time shifts to the incoming code to generate a new time/wavelength code. We also demonstrate switching of input frames to code-converted frames to resolve code contention between 2 LANs sharing the same particular code. Our technique for code conversion introduces less than 0.7 dB power penalty. 54 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. LAN-A IN: code-1 If input code is in use in LAN-B -> Code contention input code: A ^ O O A ^ (Wavelength)C ode Converter X f i O t X! f .2 1 3 > c S £ V x , 2 h M > # - # Conversion (DFG in sPPL N L LAN-B OUT: code-2 output Cyclic Time Shift, (FBG s)J code: Q P C ^ O i A A x3 \ Q time Figure 7.1. Concept of the 2-D O-CDMA code converter to resolve code contention for 2 LANs sharing the same particular code. 7.2 Time/Wavelength O-CDMA Structure and Code Conversion Figure 1 explains the structure of a time/wavelength 2-D O-CDMA code conversion. Interconnectivity between multiple O-CDMA LANs can be made efficient by incorporating code re-use. To provide this functionality, a code converter acts as a bridge between two LANs. If a node in LAN-A needs to communicate with another node in LAN-B using code-1, the network must ensure that code-1 is not already being used in LAN B. If code-1 is being used in LAN- B, the network can select another code common to both the LANs. This, however, requires a large number of orthogonal codes to provide adequate networking. A potentially better alternative is to employ a code converter that takes frames containing code-1 from LAN-A and changes the data to a new time/wavelength code that is currently not being used in LAN-B. 5 5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The code conversion process is carried out in two parts. First, the incoming set o f wavelengths is mapped onto a new set of wavelengths by DFG in a PPLN waveguide. Figure 2a shows the spectrum of the signal at the PPLN output with all input wavelengths mapped onto the mirror images o f themselves with respect to the pump. The device employs a x(2) : % (2) nonlinear process to shift data to new wavelengths at X c ~ 2 X p u m p - ^ - s i g n a l - DFG is an instantaneous process with a wide operational bandwidth of ~ 50 nm. Moreover, the PPLN has negligible spontaneous emission noise and no intrinsic chirp. O-CDMA encoded input wavelengths - > 1 5 dB ,,--------------- A --------------------- j pump A I code-converted I 1 I I outP ut wavelengths IS I A i*4 m ? ongm shifted i chip time bit time Figure 7.2. (a) Wavelength shifting using a PPLN waveguide, (b) Time shifting of the 2-D code converter. The second step provides code conversion in the time dimension through wavelength dependent delays (FBGs). The FBGs are tuned to the output 56 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. wavelengths of the PPLN waveguide for the particular code to be converted, lei and lc4. They are separated by a distance of 4 cm which produces a cyclic shift of 4 chip times for the reflected wavelengths. This causes the pulses of the output wavelengths, lcl and lc4, to change position with respect to one another, as shown in figure 2b. Thus, the data leaving the code converter is coded with a new set of wavelengths and cyclically shifted time chips (1001 a 0110) compared to the original code. 7.3 Experimental Setup /„v OCDMA encoder '110 Gchip/s Aj,=1550.12nm P P L N 1539 nm 1542 rnn 1545 nm 1547 nm 18G patters tim e dim ension A,-shifter time-shifter delay wavelength dimension in p u t; f o de j 1 data bit j 4 chips/bit £ A code converted signal ~ S ( v \ & u t p i l t \ i l l converter FBGs as wavelength dependent delays O-CDMA 10 Gchip/s' I V 2 c h t p | I 1 delay lh |r c v r |- » decision circuit (b ) ;0-CDMA decoder I decoder ► ou tp u t 2.5 Gbps Figure 7.3. (a) O-CDMA encoder with 4 wavelengths and 4 time chips and the 2- D code converter with wavelength and time shifting, (b) O-CDMA decoder composed o f the autocorrelation and threshold detection circuit. Our experimental setup is shown in figure 3. To encode the data, four lasers are modulated with the same pulse pattern and the introduction of 1-bit delays between the individual wavelengths produces the chips for each code. The original code has of the chip pattern C l=‘1001’ for “1” data bits with the first 57 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. time chip at 1 1 = 1539.04 ran and the fourth at 14 = 1547.68 nm. To represent other codes (C2=, 1100’ and C3=’1010i) propagating alongside our target code in the network, pulses at 1 3 = 1545.29 ran and 1 2 = 1543.56 nm occupying the third and second time chips o f each bit, respectively. Since the module operates at a user data rate o f 2.5 Gbit/s, each bit is split into four chip times, and the modulator operates at 10 Gbit/s. The pulse waveforms of code Cl are shown in figure 2b. The amplified and ASE filtered O-CDMA signal and the pump are coupled into the PPLN waveguide. All four incoming wavelengths are mapped onto their DFG counterparts. The time-shifter (FBGs) is tuned to two o f the new wavelengths, namely lcl = 1561.60 nm and lc4 = 1552.72. These wavelengths correspond to the first and last pulses in the original code, C l. The first FBG encountered is tuned to reflect the wavelength (1552.72 nm) of the last chip in the bit and the second FBG to the wavelength (1561.60 nm) o f the first chip in the bit. Since the FBGs are separated by two bit times, the total delay will be of 4-bit duration, and the output from the circulator will contain adjacent pulses at 1552.72 nm and 1561.60 ran, but with their order reversed. The correspondence between the old and new code is depicted in figure 2b. 5 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ‘1 J U V U J L ; *"f w gsaye" * e p i p iw P i' *W8 fW"in "* p[B!|™ i i " r" autocorrelation of i original code \ f \ f \ / \ decoded 1 1 0 / 1 \ 0 / 1 1 \ 0 0 /1 \ 0 j data pattern autocorrelation of output from the .code-converter \ / \ / \ / \ decoded 1 10 / 1 \ 0 / 1 1 \0 0 /1 \ 0 /code-converted V y / \ J data pattern switch c °^ro l "^code-convs converter scoder - decision circuit! (b) /■ ► n o code cont Figure 7.4. (a) Decoding o f the input and code-converted codes in the system, (b) Code contention resolution at a routing switch using the code converter module. Whenever a LAN-frame is coded with a code in use in LAN-B, the code-converted signal is routed at the switch output. To decode the data we employ an optical threshold technique using an optical power adder. Two FBGs are separated by a distance equivalent to half of the time between the two pulses forming the code. Reflection from these gratings causes a time shift between the chips, which results in the pulses adding up and generating a correlation peak whenever there is a “1” bit in the input data stream. An electronic threshold circuit converts this output signal to the decoded data stream. The correlation peaks and the matching decoded signals are shown in figure 4a for the input O-CDMA signal and the code-converted output signal. The noise in the decoder input for the code-converted signal resulted from EDFA ASE noise. 59 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. We also demonstrate switching o f optical packets based on whether a particular code is present in the data frame. LAN-ffames o f 16 bytes each are encoded into O-CDMA codes with some frames containing the original code. We assume that our original code is in use in LAN-B and therefore will cause contention. A set of decoder gratings is used to detect the presence o f the original code and to generate a switching signal, which is used to control a lithium-niobate electro-optic switch to route the code-converted signal in the case o f code contention. Figure 4b shows the pulse waveforms and the switch setup for the routing portion of our demonstration. 7.4 Results and Discussion 2-D O-CDMA code conversion was performed successfully using wavelength conversion and cyclic time shifting. A back-to-back • code-converted W 5 -27 -33 -31 -29 power (dBm) Figure 7.5. BER measurements o f our module. 60 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. BER measurements (Figure 5) taken on the code-converted 10 Gchip/s bit streams show that our technique o f code conversion introduces less than 0.7 dB power penalty. Our technique for code conversion is capable o f accommodating several codes. The PPLN has an operational bandwidth -5 0 nm, which may accommodate up to 20 wavelengths with -0.8 nm separation. O-CDMA encoding based on multiple wavelengths per chip time increases the number of available codes and improves their cross-correlation characteristics resulting in lower crosstalk. Sampled FBGs can be used to reflect more than one wavelength simultaneously, thereby allowing detection of multiple wavelengths in a single chip time. However, the grating length grows with the number o f wavelengths to be detected in a single chip time. This places a constraint on the spacing between two gratings if the chip rate is higher than 10 Gchip/s. 61 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Bibliography [37] O. H. Adamczyk, M. C. Cardakli, J. X. Cai, M. I. Hayee, C. Kim, A. E. Willner, “Coarse and Fine Bit Syncronization for WDM Interconnections using Two Subcarrier Multiplexed Control Pilot Tones,” IEEE Photonics Tech. Lett., vol. 11, no. 8, pp. 1057-1059, 1999. [45] P. Agrawal, “Population Pulsations and Nondegenerate Four-Wave-Mixing in Semiconductor Lasers and Amplifiers,” Journal o f Optical Soc. Amer. B, no. 1, pp. 147-159,1988. [32] T. C. Banwell and N. K. Cheung, “ A Programmable Rate Detector for Rapid-Reconfigurable Rate-Transparent Optical Networks,” Proc. of Optical Fiber Communication Conference, paper PD -10,1999. [3] A. Brackett, “Dense Wavelength Division Multiplexing Networks: Principles and Applications”, IEEE J. Selected Areas on Communications, no. 6, pp. 948- 964,1990. [4] A. Brackett, A.S. Acampora, J. Sweitzer, G. Tangonan, M. T. Smith, W. Lennon, K.-C. Wang, R. H. Hobbs, “A Scalable Multiwavelength Multihop Optical Network: A Proposal for Research on All Optical Networks,” IEEE J. Lightwave Technol., no. 5, pp. 736-753,1993. [46,51] I. Brener, M. H. Chou, and M. M. Fejer, “Efficient wideband wavelength conversion using cascaded second-order nonlinearities in LiNb03 waveguides,” Optical Fiber Communications Conference, paper FB6, vol. 4, pp. 39-41,1999. [36] M. C. Cardakli, D. Gurkan, S. A. Havstad, and A. E. Willner, “Variable-Bit- Rate Header Recognition for Reconfigurable Networks using Tunable Fiber- Bragg-Gratings as Optical Correlators,” Proc. o f Optical Fiber Communication Conference, paper TuN2, 2000. [39] M. C. Cardakli, D. Gurkan, S. A. Havstad, A. E. Willner, K. R. Parameswaran, M. M. Fejer, and I. Brenner, “All-Optical Time-Slot-Interchange and Wavelength Conversion using Difference-Frequency-Generation and FBGs,” post-deadline paper, Proc. of Optical Fiber Communication Conf., paper PD-10, 2000. 62 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. [35, 50] M. C. Cardakli, S. Lee, A. E. Willner, V. Grubsky, D. Starodubov, and J. Feinberg, “All-Optical Packet Header Recognition and Switching in a Reconfigurable Network using Fiber Bragg Gratings for Time-to-Wavelength Mapping and Decoding,” Proc. of Optical Fiber Communication Conference, paper TH-M 4,1999. [31,48] V. W. S. Chan, K. L. Hall, E. Modiano, and K. A. Rauschenbach, “Architectures and Technologies for High-Speed Optical Data Networks,” IEEE J. Lightwave Tech., vol. 16, pp.2146-2168,1998. [16] D. Chiaroni, N. Vodjani, P. Dousierre, C. Chauzat, D. De.Bouard, M. Sotom, F. Masetti, A. Jourdan, G. Soulage, and F. Ratovelomanana, “New Regeneration Functionalities o f All-Optical Wavelength Converters for Packet Switching Applications,” Proc. of Photonics in Switching, Sendai, Japan, 1996. [20] M. H. Chou, T. Hauden, M. A. Arbore, M. M. Fejer, “Efficient Wavelength Conversion based on Difference Frequency Mixing in LiNb03 Waveguides with Integrated Coupling Structures”, Proc. o f Optical Fiber Communication Conf, paper W B 4,1998. [21] M. H. Chou, I. Brener, M. M. Fejer, E. E. Chaban, S. B. Christman, “1.5 □- band Wavelength Conversion based on Cascaded Second-Order Nonlinearity in LiNb03”, IEEE Photonics Technol. Lett., no. 6, pp. 653-655,1999. [47] M. H. Chou, I. Brener, M. M. Fejer, E. E. Chaban, and S. B. Christman “1.5- pm-Band Wavelength Conversion Based on Cascaded Second-Order Nonlinearity in LiNb03 Waveguides,” IEEE Photonic Technol. Lett., no. 6, pp. 653-655,1999. [26] P. Y. Cortes, M. Chbat, S. Artigaud, J. L. Beylat, and J. Chesnoy, “Below 0.3 dB Polarization Penalty in 10 Gbit/s Directly Modulated DFB Signal over 160 km using mid-span Spectral Inversion in a Semiconductor Optical Amplifier,” Proc. o f European Conf. on Fiber Optic Communications, 1995. [18] S. L. Danielsen, P. B. Hansen, K. E. Stubkjaer, M. Schilling, K. Wunstel, W. Idler, P. Doussiere, and F. Pommerau, “All Optical Wavelength Conversion Schemes for Increased Input Power Dynamic Range,” IEEE Photonics Technol. Lett., no. 1, pp. 60-62,1998. 6 3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. [6] S. L. Danielsen, C. Joergensen, B. Mikkelsen, and K. E. Stubkjaer, “Analysis O f a WDM Packet Switch with Improved Performance under Bursty Traffic Conditions due to Tunable Wavelength Converters,” J. Lightwave Technol., no. 5, pp. 729-735, May 1998. [8] S. L. Danielsen, C. Joergensen, M. Vaa, K. E. Stubkjaer, P. Doussiere, F. Pommerau, L. Goldstein, R. Ngo, and M. Goix, “Bit error rate assessment o f a 40 Gbit/s all-optical polarization independent wavelength converter,” Proc. of Optical Fiber Communication Conf., paper P D 12,1996. [24] D’Ottavi, P. Spano, G. Hunziker, and R. Paiella, “Four-Wave Mixing Wavelength Conversion at 10 Gb/s over a 30-nm Interval,” IEEE Photonics Technol. Lett., no. 7, pp. 952-955,1998. [44] T. Durhuus, C. Joergensen, B. Mikkelsen, R. J. S. Pedersen, and K. E. Stubkjaer, “All Optical Wavelength Conversion by SO As in a Mach-Zehnder Configuration,” IEEE Photonic Technol. Lett., no. 1, pp. 53-55, 1994. [12] T. Durhuus, B. Mikkelsen, C. Joergensen, S. L. Danielsen, and K. E. Stubkjaer, “All-Optical Wavelength Conversion by SOA’s,” J. Lightwave Technol., no. 6, pp. 943-954,1996. [7] D. Ellis, D. A. O. Davies, A. Kelly, and W. A. Pender, “Data driven operation of semiconductor amplifier loop mirror at 40 Gbit/s,” Electronics. Lett., no. 15, pp. 1245-1247,1995. [53] T. Fjelde, A. Kloch, D. Wolfson, B. Dagens, A. Coquelin, I. Guillemot, F. Gaborit, F. Poingt, M. Renaud, “Novel scheme for simple label-swapping employing XOR logic in an integrated interferometric wavelength converter,” IEEE Photonics Tech. Letters, 13, 750-752, 2001. [1, 38] H. S. Hinton, “Photonic Switching Fabrics,” IEEE Communications Mag., no. 4, pp. 71-89,1990. [27] Hunziker, R. Paiella, D. F. Geraghty, K. J. Vahala, and U. Koren, “Polarization-Independent Wavelength Conversion at 2.5 Gb/s by Dual-Pump Four-Wave Mixing in a Strained Semiconductor Optical Amplifier,” IEEE Photonics Technol. Lett., no. 12, pp. 1663-1635,1996. 64 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. [15] W. Idler, M. Schilling, K. Daub, D. Baums, U. Koemer, E. Each, G. Laube, and K. Wunstel, “Signal Quality and BER Performance Improvement by Wavelength Conversion with an Integrated Three-Port Mach-Zehnder Interferometer,” Electronics Lett., no. 6, pp. 454-455, 1995. [5] S. Johansson, M. Lindblom, P. Granestrand, B. Lagerstrom, L. Thylen “Optical Cross Connect System in Broadband Networks: System Concept and Demonstrator Description,” IEEE J. Lightwave Tech., no 5, pp. 688-694,1993. [17] C. Joergensen, S. L. Danielsen, T. Durhuus, B. Mikkelsen, K. E. Stubkjaer, N. Vodjani, F. Ratovelomanana, A. Enard, G. Glastre, D. Rondi, and R. Blondeau, “Wavelength Conversion by Optimized Monolithic Integrated Mach- Zehnder Interferometer,” IEEE Photonics Technol. Lett., no. 4, pp. 521-524, 1996. [13] Joergensen, S. L. Danielsen, P. B. Hansen, K. E. Stubkjaer, M. Schilling, K. Daub, E. Lach, G. Laube, W. Idler, and K. Wunstel, “All-Optical 40 Gbit/s Compact Integrated Interferometric Wavelength Converter,” Proc. of Optical Fiber Communication Conf., paper TuOl, 1997. [28] J. P. R. Lacey, S. J. Madden, and M. A. Summerfield, “Four-Channel Polarization Insensitive Optically Transparent Wavelength Converter,” IEEE Photonics Technol. Lett., no. 10, pp. 1355-1357,1997. [42] P. J. Legg, D. K. Hunter, P. E. Barnsley, and I. Andonovic, “Crosstalk Effects in Optical Time Division Multiplexed Switching Networks,” IEE Colloquium on Optical Switching, pp. 11/1-11/4, 1993. [25] R. M. Jopson and R. E. Tench, “Polarization-Independent Phase Conjugation of Lightwave Signals,” Electronics Lett., no. 25, pp. 2216-2217,1993. [56] Hyuek Jae Lee, S. J. B. Yoo, V. K. Tsui, S. K. H. Fong, “A simple all-optical label detection and swapping technique incorporating a fiber Bragg grating filter,” IEEE Photonics Tech. Letters, 13,635-637,2001. [57] Y. M. Lin, W. I. Way, G. K. Chang, “A novel optical label swapping technique using erasable optical single-sideband subcarrier label,” IEEE Photonics Tech. Letters, 12,1088-1090,2000. 65 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. [58] B. Meagher, G. K. Chang, G. Ellinas, Y. M. Lin, W. Xin, T. F. Chen, X. Yang, A. Chowdhury, J. Young, S. J. Yoo, C. Lee, M. Z. Iqbal, T. Robe, H. Dai, Y. J. Chen, W. I. Way, “Design and implementation o f ultra-low latency optical label switching for packet-switched WDM networks,” Journal o f Lightwave Tech., 18,1978-1987, 2000. [14] B. Mikkelsen, T. Durhuus, C. Joergensen, R. J. S. Pedersen, S. L. Danielsen, and K. E. Stubkjaer, “10 Gbit/s Wavelength Converter realized by Monolithic Integration o f Semiconductor Optical Amplifiers and Michelson Interferometer,” Proc. o f European Conf. on Optical Fiber Communication, Heidelberg, Germany, 1994. [23] T. Morioka, S. Kawanishi, K. Uchiyama, H Takara, M. Saruwatari, “Polarization-Independent 100 Gbit/s All-Optical Demultiplexer using Four- Wave Mixing in a Polarization-Maintaining Fiber Loop,” Electronics Lett., no. 7, pp. 591-592, 1994. [29] K. Ovsthus and V. Khalfin, “A Novel Method for Polarization Insensitive Four-Wave Mixing in Semiconductor Optical Amplifiers,” IEEE Photonics Technol. Lett., no. 4, pp. 527-529,1996. [30] R. Paiella, G. Hunziker, J. Zhou, Kerry J. Vahala, U. Koren, and B. I. Miller, “Polarization Properties o f Four-Wave Mixing in Strained Semiconductor Optical Amplifiers,” IEEE Photonics Technol. Lett., no. 6, pp. 773-775,1996. [59] R. Parameswaran, R. K. Route, J. R. Kurz, R. V. Roussev, M. M. Fejer, “Highly efficient SHG in buried waveguides formed using annealed and reverse proton exchange in PPLN,” 752-753, LEOS 2001, paper ThL3. [11] J. S. Perino, J. M. Wiesenfeld, B. Glance, “Fiber Transmission o f 10 Gbit/s Signals following Wavelength Conversion using a Travelling-Wave Semiconductor Optical Amplifier,” Electronics Lett., no. 3, pp. 256-257, 1994. [55] P. Petruzi, Proc. o f ECOC’Ol, paper We.B.2.1, Oct. 2001. 66 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. [52] E. Rosen, A. Viswanathan, R. Gallon, “Multiprotocol Label Switching Architecture,” RFC 3031, Jan. 2001. [49] D. Sadot and I. Elhanany, “Optical switching requirements for terabit/second packet over WDM networks,” IEEE Photonics Tech. Letters, 12, 440-442, 2000. [33] B. Sartorius, C. Bomholdt, O. Brox, H. J. Ehrke, D. Hoffmann, R. Ludwig, and M. Mohrle, “Bit Rate Flexible All-Optical Clock Recovery,” Proc. o f Optical Fiber Communication Conference, paper F B -1,1999. [10] W. Shieh, S. H. Huang, A. E. Willner, “A Polarization-Independent and Contrast-Ratio-Enhancing Module for All-Optical Wavelength Shifting using SOAs,” IEEE Photonics Technol. Lett., no. 4, pp. 533-535,1996. [34] J. Shin, M. Jeon, and C. Kang, “Fiber-Optic Matched Filters with Metal Films Deposited On Fiber Delay-Line Ends for Optical Packet Address Detection,” IEEE Photonics Tech. Lett., vol. 8, pp. 941-943,1996. [9]J. C. Simon, L. Lablonde, I. Valiente, L. Billes, P. Lamouler, “Two-stages wavelength converter with improved extinction ratio,” Proc. o f Optical Fiber Communication Conf, paper PD15, 1996. [41] R. A. Thompson and P.P. Giordano, “An Experimental Photonic Time-Slot Interchanger Using Optical Fibers as Reentrant Delay-Line Memories,” IEEE J. Lightwave Technol., no. 1, pp. 154-162, 1987. [43] J. M. Wiesenfeld, J. S. Perino, A. H. Gnauck, and B. Glance, “Bit Error Rate Performance for Wavelength Conversion at 20 Gbit/s,” Electronics Lett., no. 9, pp. 720-721, 1994. [54] M. D. Vaughn, D. J. Blumenthal, “All-optical updating o f subcarrier encoded packet headers with simultaneous wavelength conversion of baseband payload in semiconductor optical amplifiers,” IEEE Photonics Tech. Letters, 9, 827-829, 1997. [2] A. E. Willner, M. C. Cardakli, “Network Functions Enabled by Optics,” invited paper, 37th Annual Allerton Conference on Communications, Control, and Computing, Sept. 1999. 67 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. [40] J. Yao, P. Barnsley, N. Walker, and M. O’Mahony, “Time Slot Interchanging Using Semiconductor Laser Amplifiers,” Electronics Lett., no. 12, pp. 1053-1054, 1993. [19] S. J. B. Yoo, “Wavelength Conversion Technologies for WDM Network Applications,” J. Lightwave Technol, no. 6, pp. 955-966, June 1996. [22] S. J. B. Yoo, O. Rajhel, C. Caneau, R. Bhat, M. A. Koza, “Multichannel Polarization-Independent Wavelength Conversion by Difference-Frequency Generation in AlGaAs Waveguides”, Proc. o f Optical Fiber Communication Conf., paper Tu06, 1997. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Asset Metadata

Creator Gurkan, Deniz (author)

Core Title Experimental demonstrations of all -optical networking functions for WDM optical networks

Contributor Digitized by ProQuest (provenance)

School Graduate School

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag engineering, electronics and electrical,OAI-PMH Harvest,physics, optics

Language English

Advisor Willner, Alan E. (committee chair), Lee, Daniel C. (committee member), Rich, Daniel (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-483599

Unique identifier UC11335040

Identifier 3133278.pdf (filename),usctheses-c16-483599 (legacy record id)

Legacy Identifier 3133278.pdf

Dmrecord 483599

Document Type Dissertation

Rights Gurkan, Deniz

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA