21

1 Introduction

As a result of the widespread use of theInternet, a massive amount of information hasbecome available to nearly everyone at anytime. Individuals who previously receivedinformation in one direction from media suchas television and newspapers now have theopportunity to disseminate information them-selves via the Internet. This change hascaused an explosive growth in network datatraffic due to the huge amount of data beingexchanged between computers. Furthermore,the Internet is now being utilized in a widerrange of application fields, including econom-ic activities, medical services, and public serv-ices, and is penetrating deeper into daily life.

This rapid growth in data exchange hascaused a serious problem with regard to the

capacity of the existing communications net-work. The increase in communications data isexpected to continue, and even accelerate, inthe years ahead, and therefore the informationcommunications network must be improved assoon as possible in terms of speed, capacity,and functionality. In order to realize an ultra-fast, ultra-high capacity network that will ful-fill public needs, a photonic network in whichinformation is exchanged by light wavethrough optical fibers, as opposed to the exist-ing electrical signal system, is essential. Thekey technologies for implementing such aphotonic network are classified into two com-ponents: photonic transmission technology forlinking nodes in networks, and photonic trans-fer technology for routing information to indi-vidual lines. In terms of transmission technol-ogy, it is now possible to send terabit data via

Naoya WADA

3-2 Ultra-fast photonic packet routing tech-nology

Naoya WADA

The great proliferation of the Internet has led to massive data exchanges between termi-nals such as computers and cellular phones. This growth in information exchange hasresulted in a need for higher speed, larger capacity, and enhanced functionality of the infor-mation communications network that serves as the basis of data exchange. A photonic net-work, in which information is transmitted by light, is essential if we hope to the implementa-tion of an ultra-fast, ultra-high capacity network that meets consumer demands. This paperreviews photonic routing technology, describes the concept of photonic packet routingtechnology employing optical orthogonal encoding labels, and then discusses our researchinto this technology. Also outlined are the concept and results of a proof-of-principle exper-iment of photonic packet routing technology using a multi-wavelength label, which is animproved system of MP-Lambda-S (Multi-Protocol Wavelength Switching). This paper con-cludes with an outline of and results from a proof-of-principle experiment of photonic labelprocessing technology using spectral holography.

Photonic packet routing technology is a means of applying optical information technolo-gy to optical communications systems, and is expected to further develop in the future.

Keywords Photonic Network, Photonic Packet Switching, Optical Code Label, All Optical LabelRecognition, All Optical Processing

22

each fiber thanks to the development of WDMtechnology. On the other hand, the switchingcapability at network nodes for each input isstill in the order of one gigabit per second, andit is this that causes the primary bottleneck innetworks. Indeed, there is a gap of two ordersof magnitude between the transmission speedand the transfer speed. In order to eliminatethis bottleneck, photonic technology must beapplied to nodes, because there are limits topresent electrical signal processing capabili-ties (Fig.1).

Major networks are currently shifting con-figurations from the circuit-switching type,represented by the telephone network, to thepacket-switching type, represented by theInternet. As shown in Fig.2, in a circuit-switching type network using point-to-pointprotocol (PPP), a circuit is occupied by com-munication between, for example, terminal Aand terminal D. Another terminal (for exam-ple, terminal B) is not allowed to use this cir-cuit even if there is no traffic on this channel.Thus the communication band channel effi-ciency is low. Meanwhile, in a packet-switch-ing type network using Internet protocol (IP),other terminals are allowed to use this channelfor communication (for example, between ter-minal B and terminal C) by using vacant timeslots, even in the midst of communicationbetween terminal A and terminal B. In thiscase, however, nodes X and Y are required tobe able to send packets after multiplexingthem and demultiplexing/routing them, basedon their packet address information. Thisfunction is known as packet routing. All head-

er processing, including address recognition,is carried out by electrical means in the cur-rent network nodes. Thus even when theinformation has been sent via light signals, itmust be converted into electrical signals forrouting at nodes. The speed of routing, there-fore, is limited by the signal processing capa-bility of the electrical circuit at the node. Inorder to handle packets beyond the capabilityof electrical circuits, photonic packet routingtechnology becomes necessary as it can sendoptical signals as they are, without convertingthem into electrical ones.

Conventional electrical nodes also posesecurity concerns. Tapping or hacking of thetransmission lines is physically possible, butin reality these security breaches rarely occur.More commonly, security problems arise atnodes where light signals are converted intoelectrical ones and the electrical signals aresaved in a buffer until the header processing iscompleted. Photonic routing sends the pack-ets via light signals thus greatly diminishingsuch problems and improving network securi-ty.

This paper first reviews photonic routingtechnology and then describes the outline andresults of research into photonic packet rout-ing technology employing an optical orthogo-nal encoding label. Also described are theoutline and results of research into a proof-of-principle experiment of photonic packet rout-ing technology using the multi-wavelengthlabel, which is an improved form of the MP-Lambda-S (Multi-Protocol WavelengthSwitching). This paper concludes with adescription of the outline and results ofresearch into a proof-of-principle experimentof photonic label processing technology using

Journal of the Communications Research Laboratory Vol.49 No.1 2002

Information transmission speed in net-work

Fig.1

Change in network protocolFig.2

23

time-space conversion processing and spectralholography.

2 Current Status and Issues

Due to the fast-growing popularity of theInternet and the concomitant increase in datatraffic via networks, there has been great inter-est in WDM (wavelength division multiplex-ing) technology, which can accommodate inconventional optical fibers more than 100Gbps of traffic. Networks implementingWDM technology are proliferating in responseto growing demands for increased networkcapacity.

Currently, however, the Internet applica-tions of this technology are limited to linksbetween neighboring routers. When exchang-ing data between transmission and receptionnodes via a number of routers, or transitnodes, the packet handling speed of these tran-sit nodes plays a key role, and the data pro-cessing capability of these transit nodes mustbe upgraded to ease data throughput of thenetwork as a whole. However, if we rely ononly electronic technologies to increase packethandling speed, we can expect relatively limit-ed improvements[1]; there is a higher potentialfor improvement in transmission capacity ofoptical fibers if we utilize WDM technology.In conventional network configurations, basedon a layered structure, the routing capabilitiesof transit nodes causes bottlenecks.

To solve this problem, a number of meth-ods have been proposed regarding the settingup of a photonic network in lower layers thatwill reduce IP routing jobs. For example,MPLS (multi-protocol label switching) tech-nology[2][3] and optical path network technol-ogy (circuit-switching type WDM net-work)[4][5] have been proposed. According tothese methods, IP packets are capsulated in awavelength and then run through the photonicnetwork. The transit node employs opticalcross-connects (OXC) and optical add-dropmultiplexers (OADM) to provide photonicprocessing that has a speed beyond the limitsof electronic methods. These transit methods,

however, only branch light signals throughpredetermined routes wavelength by wave-length. They have no ability to read the infor-mation embedded in the input wavelength androute data of the same wavelength into differ-ent routes. Packets of a pair of transmissionand reception nodes are capsulated in eachwavelength. Furthermore, the transmissionspeed of the optical channel (wavelength) isusually fixed at a certain speed, such as 2.4Gbps or 10 Gbps, for example, and it isimpossible for multiplexing light signals toenter from a number of input ports through asingle route. Thus when an edge node of anetwork uses a bandwidth smaller than thetransmission rate limit, the large bandwidthprovided by optical technologies is not fullyutilized.

In order to efficiently use the bandwidth ofan optical layer it is essential to be able tobranch light signals sent from a transmissionroute (for example, wavelength) into two ormore routes and combine light signals sentfrom a number of routes into a single one. Itis possible to efficiently use the bandwidth asshown in References[7][8] in IP over WDM[6],because the IP router has this capability.However, as previously mentioned, since therouting process still tends to cause bottle-necks, the signals should be multiplexed in theoptical layer. Moreover, when burst traffic isobserved on the Internet, the simple time divi-sion multiplexing system does not allow effi-cient use of bandwidth. Instead, packetexchanging should be adopted and the packetsrouted in the network according to the addressinformation embedded in them.

A number of optical packet exchangershave been proposed[9][10][11]. They read therouting information of packets after convert-ing it into electrical signals and then outputthe packets to corresponding ports. In otherwords, the data itself reaches its destination inthe form of a light signal, but its address infor-mation is not read directly from that signal.The speed of the packet matching process islimited to the same level as that of the IProuters for electrical processing. Optical pro-

Naoya WADA

24

cessing, therefore, is essential for faster signalmatching.

Once these issues are resolved, there willbe heightened expectations for a large-scalepacket exchange network capable of light-based data transmission. Mixed packets withdifferent destination addresses in a transmis-sion route require that the network provide asufficient number of labels. However, if theMPLS network [3] and optical path networktechnologies, which take up a wavelength perlabel, are adopted, it is clear that all wave-lengths will be quickly used. Although around1,000 wavelengths are available[12], they maynot be enough to sufficiently expand the net-work to the scale needed to establish a photon-ic network.

3 Photonic Label Switching UsingOptical Code

Fig.3 demonstrates the configuration ofthe proposed system. The proposed opticalpacket consists of an optical code, in whichthe address information has been mapped, andthe payload data. This optical packet isaccommodated as the optical path payload inan optical frame that has the wavelength rout-ing information in the optical path header.The photonic label switch consists of a wave-length demultiplexer and a photonic labelprocessor. The photonic label processor con-sists of a header processor, a 1×N photonicswitch, and an optical delay. The headerprocessor conducts parallel label matching andconsists of optical amplifiers, optical corre-laters, optical pulse reshapers, and address

encoders. The group of optical correlatersworks as an address bank[13] that recordscodes corresponding to the addresses held inthe routing table.

The input packets are divided wavelengthby wavelength, and then sent to each photoniclabel processor corresponding to individualwavelengths. In the photonic label processor,as shown in Fig.4, the header processor simul-taneously produces replicas of the input codeand conducts label matching, based on the cal-culation of optical correlation, thus outputtingthe photonic switch control signal. A controlsignal of a high peak value can be issued onlywhen the input label matches the intrinsiclabel information that each correlater holds;when this occurs, the other correlaters do notoutput a high peak signal. The photonicswitch conducts the routing of each packetbased on this control signal. This label match-ing process in the optical domain does notrequire logic calculation. Label matching canbe carried out based on optical correlation cal-culation in the time domain between the inputoptical code and the code held in the addressbank. This parallel processing leads to fasterdata handling.

4 Packet Routing Experiment

The experimental setup is shown in Fig.5.This experimental system is comprised of anoptical packet transmitter and a photonic labelprocessor[14]. The optical packet transmitteris made up of a semiconductor mode-lockedlaser diode (MLLD) that generates 2ps-10GHzpulses, LiNbO3 intensity modulators (IM),optical encoders, and an optical delay. Theoptical encoder is composed of optical taps,5ps delay lines, phase shifters, and a combin-er, all of them being integrated into a chip byPLC technology. In the optical encoder, asshown at the top of Fig.6, a single pulse isseparated into eight chip pulses arrayed atintervals of 5 ps. Each pulse is given a phaseshift of either 0 or π, and then an 8-chip opti-cal bipolar code (BPSK-code) is generated[15].The photonic label processor is composed of a

Journal of the Communications Research Laboratory Vol.49 No.1 2002

Optical packet and photonic labelswitching system

Fig.3

25

header processor unit and an optical switchunit. The header processor is composed ofoptical correlaters, photodetector (PD), low-

pass filters, and gain-clamped drivers. The1×2 optical gate switch composes two inten-sity modulators (IM) having a bandwidth of

Naoya WADA

All-optical parallel label processingFig.4

Optical packet routing experiment systemFig.5

All-optical encoder/correlaterFig.6

26

40GHz. The optical correlater and the opticalencoder have the same structure.

The optical packet transmitter generatesburst data at a rate of 64bit@10Gbit/s and 8-chip BPSK optical codes, which are combinedand then output as optical packets. As shownat the bottom of Fig.6, the photonic labelprocessor yields an auto-correlated waveformof a high peak from its correlaters when theBPSK code agrees with the code held in theoptical correlater. Meanwhile, a cross-corre-lated waveform is issued that has a low inten-sity when codes do not match. The output ofan auto-correlated waveform obtained fromthe correlaters in case of code matching isconverted into an electrical signal and thenextended on the time axis by the low-pass fil-ters and the gain-clamped amplifiers to openthe gate switch while the packet passes. Onthe other hand, when there is no match, theoutput cannot open or pass the IM gate switchbecause of its low value. A packet exits onlyfrom the gate the label has matched, and thenthe packet is routed.

Fig.7(a) is a streak camera image of thetop of a packet holding an optical code “0ππππππ0” as a label. Figs.8(b) and (c) showthe outputs from the optical correlaters in theheader processor for the packets having opti-cal codes #1“0ππππππ0” (matched code)and #2 “0π0π0π0π” (unmatched code) asoptical labels. An auto-correlated waveform

of a high peak value is given to the matchedcode, while a cross-correlated waveform of alow value is given to the unmatched code.Therefore, the output from the optical corre-later for the matched code can be the controlsignal strong enough to drive the opticalswitch. The 64bit data following the opticalcode is diffused when passing the optical cor-relater and thus does not affect the control sig-nal. Figs.9(a) and (b) demonstrate the outputsfrom the optical gate switches for the labels ofoptical code #1 or #2. The figures indicatethat the 64bit burst data is successfullybranched in accordance with the attached opti-cal codes. The bit error rates in both ports areas low as 10-10 or less, indicating that the pack-ets have been successfully branched.

In addition, since the photonic labelprocessor employed in this method is a pas-sive and transparent device, it can be used uni-versally for signals of different wavelengthswith no modification. Therefore, the signalsof different wavelengths that have been multi-plexed by WDM can be separately and simul-taneously handled with a single recognitiondevice. As a result, it becomes possible todownsize the system for each packet switch-ing node and provide ultra-high data through-put capability. Fig.10 shows the entire systemconfiguration. The results obtained providestrong support for the feasibility of ultra-high-speed photonic packet routing.

Journal of the Communications Research Laboratory Vol.49 No.1 2002

Generation of an optical packet having an optical code labelFig.7

27

5 Photonic Label Switch UsingMulti-Wavelength Label

Communications Research Laboratory hassuccessfully proposed a method of using themulti-wavelength label and proof-of-principleexperiment in addition to the above-mentionedoptical orthogonal coding method[16][17].Fig.11(a) shows the proposed packet format.The label is made of a series of K-chip pulseschosen from as many as W wavelengths.

Fig.11(a) shows the case of K = W = 4,whereλ1A,λ1B,λ1C,λ1D represent wavelengthsthat are used to generate a multi-wavelength

label. The payload data uses a separate wave-length,λ1E. The wavelengths fromλ1A toλ1E

comprise a wavelength bandλ1A-E. The samenumber of independent multi-wavelengthlabels may be generated in the individualwavelength bands (λ1A-E,λ2A-E,λ3A-E...). Thelabel uses one wavelength for each of the K-pulse series and different wavelengths areassigned to the individual pulses (hereafterrefered to as label configuration A).

Fig.11(b) shows the structure of a networkemploying the multi-wavelength label packet(hereafter refered to as multi-wavelength labelnetwork). The multi-wavelength label net-work is comprised of multi-wavelength edgenodes (MλEN), multi-wavelength labelswitching nodes (MλLSN), and wavelengthrouters (λOXC).

The multi-wavelength edge node is capa-ble of assembling optical packets by addingthe multi-wavelength label to the data that hasarrived from an external system and thensending the packets to the network. It alsoremoves the label from the packet and sendsthe data outside the network when it hasreceived an optical packet. The multi-wave-length label switching node is made up of amulti-wavelength label switch (MλLS) and amultiplexer. The multi-wavelength labelswitch has the function of reading the multi-wavelength label and sending the optical

Naoya WADA

All-optical label matchingFig.8

Packet routing test resultsFig.9

28

packet to the corresponding port. Generationand readout of the multi-wavelength label iscarried out by optical processing using multi-section FBGs. The multiplexer combines theoptical packets that have arrived from a num-ber of multi-wavelength label switches. Thewavelength router sends all the wavelengths ina band to the same output port. Thus thepacket made of wavelengths, which belong tothe same bandwidth shown in Fig.11(a), is not

discomposed by the wavelength router.According to the above configuration, two dif-ferent kinds of nodes are allowed to existwithin the same network, and a flexible net-work can thereby be built in the opticallayer[18].

The necessary number of wavelengthsrequired for label generation can be reduced inthe multi-wavelength label network, comparedwith other networks or an optical path networkwhere one wavelength is needed to make eachlabel. In label configuration A, as many asW!/(W-K)! different labels (W≥K) can be gen-erated when as many as W wavelengths andK-pulse series are used in forming a label.The necessary number of wavelengths may bereduced to the least integer W that meetsW!/(W-K)≥N in the network having as manyas N edge nodes that conduct routing based onthe shortest-path routing method. For exam-ple, assume a case in which the label equiva-lent to the destination address is included inthe packet in a backbone network of approxi-mately N=100. If the network uses one wave-length for each label, as many as 100 wave-lengths will be needed. About 10,000 wave-lengths will be needed in the optical path net-work. On the other hand, only 11 wavelengthsare sufficient in the multi-wavelength labelnetwork if a label consisting of two pulses is

Journal of the Communications Research Laboratory Vol.49 No.1 2002

Ultra-high throughput photonic packet switching systemFig.10

(a).Packet format having a multi-wavelength label

Fig.11

(b).Network configurationFig.11

29

used. If a label consisting of four pulses isused, five wavelengths are sufficient. Evenwhen regarding the hosts in the Internet asedge nodes that route packets of an addresslength of 32 bits, 20 wavelengths (eight puls-es) at most are sufficient in the multi-wave-length label network.

The number of optical labels may increase.Assume that each of the wavelengths forlabeling is used once and assigned to a desiredposition in the label-space of K-pulse (Labelconfiguration B). To recognize the top of apacket, at least one wavelength has to beassigned to the top position. Then the numberof labels becomes KW-(K-1)W[18].

Fig.12 illustrates the relationships betweenthe number of wavelengths, label length, andlabel number for configurations A and B. Ineither case, more than 10,000 labels can beprovided if as few as eight wavelengths areused in labeling. See Reference[19] for a moredetailed analysis.

6 Configuration of Multi-Wave-length Label Switch

Fig.13 shows the block diagram of the sys-tem configuration of the multi-wavelengthlabel switch (MλLS). The multi-wavelengthlabel switch comprises a label/data separator,optical delay generator, a multi-wavelengthlabel processor, a 1×N optical switch, and alabel changer. This configuration of theswitch is the same as that of the aforemen-tioned photonic switch system using the opti-cal coding label but differing in the label/data

separator and label processor unit. Also, theconfiguration of the label changer is differ-ent[20], but is not referred to in this paper.

When a packet having the multi-wave-length label enters this system, the label por-tion is separated by the label/data separatorand sent to the label processor. The labelprocessor conducts label recognition andmatching with the label banking in parallel onan all-optical basis. Based on the result of thislabel matching, the label processor controlsthe optical switch and sends the packet to thecorrect port. If the label needs to be changed,the label changer provides a new label andsends it out.

7 Multi-Wavelength Label Switch-ing Experiment

Fig.14 shows the configuration of theproof-of-principle experiment. The experi-mental setup consists of a multi-wavelengthlabel packet transmitter, 33km dispersionshifted fiber (DSF), and a three-output portmulti-wavelength label switch. The multi-wavelength label packet transmitter consists ofa super continuum (SC) light source[21] (pro-viding multi-wavelength pulses of a band-width of 140 nm @ 10 GHz), LiNbO3 intensi-ty modifiers (IM), a multi-wavelength labelgenerator[22] (comprised of a three-sectionfiber grating (FBG) and an optical circulator,an optical band-pass filter (O-BPF), and adelay. A multi-wavelength light pulse emittedfrom the SC light source is divided into threepulses by the three-section FBG after its repe-tition speed is lowered to the packet rate byIM, and then the multi-wavelength labels (λ1A,λ1C,λ1B in Fig.14) is generated. Meanwhile,

Naoya WADA

Comparison of label numbersFig.12

Multi-wavelength label switchFig.13

30 Journal of the Communications Research Laboratory Vol.49 No.1 2002

the pulse with a center wavelength ofλ1E isextracted from the SC light by BPF, and IMgenerates 64bit burst data. The generatedmulti-wavelength label is laid ahead of theburst data in which the time shift has beencontrolled by the optical delay and then sentout as a packet with a multi-wavelength label.

The three-port output multi-wavelengthlabel switch consists of an O-BPF, opticaldelay, multi-wavelength label correlater (com-prised of three-section FBGs and optical cir-culators), photodetector (PD), low-pass filters(RF-LPF), RF threshold processor, RF wave-form expander, gain-clamped RF amplifier,and 1×3 optical switch. The 1×3 opticalgate switch has 40GHz-bandwidth LiNbO3

intensity modulators. The three-section FBGarray in the multi-wavelength label switchconducts correlation processing for labelrecognition and works as a label bank. Whenthe input label matches the combination ofcorresponding wavelengths of the correlater,the FBG correlater outputs a light pulse of ahigh peak value. On the other hand, when thewavelength array of the input label does notmatch the combination of wavelengths of thecorrelater, a pulse of a low peak value isissued instead of a high peak pulse. The opti-cal output from the correlater is converted into

electrical signal by PD; this signal is electri-cally expanded after a threshold processing toallow the gate switch made of a LiNbO3 inten-sity modulator to be opened for the period oftime needed for the packet to pass. The strongpulse, which has matched the input label andbeen issued by the correlater, is regarded as“1” by the threshold processing and then con-verted into the gate signal by electrical pro-cessing. Meanwhile, the correlater output thathas not matched the input label has no peak inthe pulse, and thus it is regarded as “0” bythreshold processing. The output after theelectrical processing remains at level-zero.Then, only the gate switch connected to thecorrelater whose wavelength set has matchedthe input label will open; the other gateswitches will stay closed. The input packetsare thereby branched to the ports correspon-ding to individual labels.

Fig.15 shows our experimental results.Figs.15(a) and (b) show results obtained witha sampling oscilloscope of the multi-wave-length label “λ1A,λ1C,λ1B” that was generatedby the multi-wavelength label packet transmit-ter and of the intensity of the packet havingsuch a label as header. Figs.15(c) and (d)demonstrate the waveforms of the outputsfrom correlater FGB 1*(λ1B,λ1C,λ1A) and cor-relater FGB 3*(λ1C,λ1B,λ1A). A pulse of ahigh peak value is provided in Fig. 15(c)where the label matches the wavelength set ofthe correlater, while a high peak is not seen inFig.15(d) where the label does not match.These results indicate that all-optic multi-wavelength labeling is viable. Figs.15(e) and(f) show the waveforms of outputs from thethree-port switch. Fig.15(e) shows the outputof the label switch for the case where a packethaving the multi-wavelength label “λ1A,λ1C,λ1B” matching the correlater FBG 1* hasentered the label switch, while Fig.15(f)shows the output for the case where a packethaving the multi-wavelength label “λ1A,λ1B,λ1C” matching the correlater FBG 3* hasentered the label switch. The results indicatethat the multi-wavelength label switches suc-cessfully switch the outputs according to the

Experimental systemFig.14

31

label information of each packet.Fig.16 shows the bit error rate of packets

sent back-to-back on a 33km transmission.Figs.16 (a), (b) and (c) correspond to outputsfrom ports 1, 2, and 3, respectively. It is evi-dent from the port output results that transmis-sion and routing have been successfully per-formed. This investigation shows that it isquite possible to create a photonic networkusing the multi-wavelength label switch.

8 Holographic Label Processor

The above studies describe the opticalcode label switching method (Fig.5) using aPLC-type correlater and the multi-wavelengthlabel switching method using an FGB-type

correlater (Fig.14). Since one correlater corre-sponds to one address in these methods, alarge number of correlaters will be required inthe header processor for exact matching ofaddress information.

We have proposed a holographic addressprocessor[23][24] using spectral holography,which is capable of recording a great deal ofaddress information in a single device, as theaddress bank on behalf of the PLC decoderarray.

Fig.17 is a block diagram illustrating thebasics of holographic label processing. Aftera time-space conversion[25][26], the input opti-cal label is Fourier-transformed and enteredinto an angular-multiplexed spectral hologram(AMSH)[27][28]. The AMSH holds informa-tion corresponding to a multitude of inputlabels. It works as an address bank and out-puts pulses in different angular directionsaccording to the kinds of input label. The out-put pulse runs through a Fourier-transformlens, and it is converted into positional infor-mation and then detected by a positional infor-mation detector. The final output is therebyobtained in accordance with each kind of inputlabel. Matching for a number of labels can beperformed on a real-time basis with a singledevice.

Fig.18(a) shows the system for performingmultiplex writing (holographic label banking)

Naoya WADA

Experimental results (1)Fig.15

Experimental results (2)Fig.16

32

of label information in an AMSH. This sys-tem consists of a time-space conversion grat-ing, Fourier-transform lens, and holographicmedium. The optical label and the referencelabel are time-space converted simultaneouslyand then recorded onto the holographic medi-um by interference after passing through theFourier-transform lens. All optical labelsexperience this processing for multiplexrecording. In this case, the position forinputting optical labels is fixed, while theposition for inputting reference pulses ischanged for each optical label. The AMSHlabel bank is thereby prepared.

Fig.18(b) shows the configuration of theholographic label processor. The processorconsists of a grating, two Fourier-transformlenses, and an AMSH. The optical labelwhose address information has been mappedis time-space converted by the grating andentered into the spectral hologram via theFourier lens. Information about a number ofoptical labels has been written in the spectralhologram by angular-multiplexing using thesystem shown in Fig.18(a). The light emittedfrom the spectral hologram passes through theother Fourier lens and reaches the positioncorresponding to each label in the incident

light. Unlike conventional spectral hologra-phy, there is no need to conduct time-spaceconversion again so as to convert the signalinto a waveform on the time axis.

9 Proof-of-Principle Experimentand Results

Fig.19 shows the experimental system forconfirming the mechanism of an all-opticaladdress matching process at 10 Gbit/s. Thisexperimental system comprises a mode-lockedsemiconductor laser diode (MLLD), opticalfiber amplifier (erbium-doped fiber amplifier:EDFA), filter, polarization controller (PC),isolator, PLC-type 8-chip optical encoder, dis-persion-shifted fiber, collimator, holographicaddress processor, optical spectrum analyzer(OSA), and optical power meter (OPM). Thepulse (central wavelength 1550 nm, 10 Gbit/s,pulse width 2ps) emitted from the MLLD isconverted into an 8-chip optical intensity codewith the PLC-type 8-chip optical encoder.The generated optical code enters the holo-graphic address processor, and a pulse signalcomes out in the position corresponding toeach kind of optical code based on the abovemechanism. The output signal is then

Journal of the Communications Research Laboratory Vol.49 No.1 2002

Principles of holographic label pro-cessing

Fig.17

(a).Writing of label informationFig.18

Experimental system for holographiclabel processing

Fig.19

(b).Label matchingFig.18

33

observed with an optical spectrum analyzer(OSA) and an optical power meter (OPM).

Because a holographic medium for the 1.5μm-band was not available for the experi-ment, we prepared a signal pattern that wasmodified for use in the 1.5μm band and madea holographic address bank using an He-Nelaser with a central wavelength of 632.8nm.We employed Agfa-Gevaert 8E75, originallyprepared for visible light, as the holographicmedium. When employing the He-Ne laser tocreate a quasi hologram for use in the 1.5μmband, an AMSH for optical phase code labelbanking cannot be produced because we can-not project the real input beam (signal) ontothe holographic medium. Thus we used theoptical intensity code as the label in thismechanism confirmation test.

Figs.20 and 21 show the experimentalresults. Figs.20(a) and (b) show the streakcamera traces (waveform along the time axis)of the optical coding labels “10010010” and“10000010”, respectively, which were gener-ated by the PLC-type encoder. In the previousphotonic routing experiment using the PLC-type correlater, we generated a combined opti-cal phase code made of “0 andπ” for creating

the address information. Because of theabove-mentioned reason, however, in thisexperiment we used an intensity label of “1and 0”.

Figs.21(a) and (b) show power spectraobserved at output ports of the address proces-sor for the cases in which the code of incidentlight matches or does not match the code thathas been assigned to each port in advance.When the address information shows a match,a strong peak appears and address recognitioncan be made. Fig.22 is a comparison of inten-sity distributions between the positions of out-put beams corresponding to optical codinglabels “10010010” and “10000101”, in thecase in which an optical coding label of“10010010” is entered into an AMSH inwhich the optical coding labels “10010010”and “10000101” have been recorded inadvance. These experimental results indicateconsiderable potential for practical use of amultiplex recording label recognition devicerelying on an AMSH.

In this experiment we failed to provideoutput light strength that was adequate todrive the optical switch because the light inthe 1.5μm-band showed significant powerloss in a holographic medium that had origi-

Naoya WADA

Generated 8-chip OOK labelFig.20 Recognition of 8-chip OOK label (1)Fig.21

34

nally been prepared for visible light. Thedevelopment of a low-loss hologram for the1.5μm band is one of the keys to its success-ful use in the future.

10 Conclusions

We have reviewed photonic packet routingtechnology and explained the results ofresearch into this technology with regard to anoptical orthogonal coding label. We have alsoexamined the structure of packet switchingnodes capable of ultra-high levels of datathroughput. The outline and results of amechanism confirmation test of photonic

packet routing technology using the multi-wavelength label, which is an improved formof MPλS, have also been discussed. Finally,we have examined the outline and results of aproof-of-principle experiment of photoniclabel processing technology using spectralholography. Photonic packet routing technol-ogy is a means of applying optical informationprocessing technology to optical communica-tions. To further improve this technology it isimportant to develop a material suitable foroptical information process in the 1.5μmband.

Acknowledgments

The author wishes to thank researcherHiroaki Harai, group leader Wataru Chujo,and senior researcher Fumito Kubota of Com-munications Research Laboratory, Prof. Ken-ichi Kitayama at Osaka University, Dr. Joe-wono Widjaja at Suranaree University, Prof.Yukihiro Ishii at Polytechnic University, Prof.Kashiko Kodate, and assistants Kayo Shimizuand Naoko Kawakami at Japan Women’s Uni-versity, for cooperation and useful advice.The author is also grateful to YoshihiroTomiyama and Tomoko Arai for their help inexperiments, and Division Manager ToshikazuItabe and President Takashi Iida of Communi-cations Research Laboratory, for their supportand advice.

Journal of the Communications Research Laboratory Vol.49 No.1 2002

Recognition of 8-chip OOK label (2)Fig.22

References1 S. Nojima, "Technologies for high speed router and problems", IEICE Technical Digest on PNI

(PNI99-5), pp. 29-34, Jun. 1999.

2 E. Rosen, A. Viswanathan, and R. Callon, "Multiprotocol label switching architecture (draft-ietf-mpls-

arch-07.txt)," IETF Internet Draft (Work in Progress), Jul. 2000.

3 D. Awduche, Y. Rekhter, J. Drake, and R. Coltun,"Multi-protocol lambda switching: Combining

MPLS traffic engineering control with optical crossconnects (draft-awduche-mpls-te-optical-02.txt),"

IETF Internet Draft (Work in Progress), Jul. 2000.

4 I. Chlamtac, A. Ganz, and G. Karmi, "Lightpath Communications: An approach to high bandwidth

optical WAN's," IEEE Transactions on Communications, Vol. 40, pp. 1171-1182, Jul. 1992.

5 E. Modiano, "WDM-based packet networks," IEEE Communications Magazine, Vol.37, No.3,

pp.130-135, Mar. 1999.

6 N. Ghani S. Dixit, and T. Wang, "On IP-over-WDM integration", IEEE Communications Magazine,

35Naoya WADA

Vol.38, No.3, pp.72-94, Mar. 2000.

7 H. Harai, F. Kubota, and H. Nakazato, "Design of reconfigurable lightpaths in IP over WDM net-

works," to appear in IEICE Transactions on Communications, Oct. 2000.

8 H. Harai and F. Kubota, "Bandwidth maximization in IP over WDM path networks," to be presented

at OptiComm 2000, Oct. 2000.

9 K. Habara, H. Sanjo, H. Nishizawa, I. Ogawa, and Y. Suzaki, "Large-Capacity Photonic Packet

Switch Prototype Using Wavelength Routing Techniques", IEICE Technical Digest on SSE (SSE99-

148), pp.13-18, Feb. 2000.

10 S. Araki, S. Takahashi, Y. Maeno, Y. Suemura, A. Tajima, H. Takahashi, K. Matsuda, T. Tamanuki,

S. Dohmae, and N. Henmi, "A 2.56 Tb/s throughput packet/cell-based optical switch-fabric demon-

strator", ECOC '98, pp.127-128, 1998.

11 S. Yao, B. Mukherjee, and S. Dixit, "Advances in photonic packet switching: An overview," IEEE

Communications Magazine, Vol.38, No.2, pp.84-94, Feb. 2000.

12 H. Takara, T. Ohara, K. Mori, K. Sato, E. Yamada, K. Jinguji, Y. Inoue, T. Shibata, T. Morioka, and

K-I. Sato, "Over 1000 Channel Optical Frequency Chain Generation from A Single Supercontinuum

Source with 12.5GHz Channel Spacing for DWDM and Frequency Standards", ECOC 2000 Post

Deadline Paper, Sep. 2000.

13 K. Kitayama and N. Wada, "Photonic IP Routing," IEEE Photon. Technol. Lett., Vol.11, p.1689,

1999.

14 N. Wada and K. Kitayama, "Photonic IP routing using optical codes: 10Gbit/s optical packet transfer

experiment," OFC2000, WM51, p.362, 2000.

15 N. Wada, and K.Kitayama, "10Gbit/s optical code division multiplexing using 8-chip optical bipolar

code and coherent detection," J. Lightwave Tech., 17, p1758, 1999.

16 N. Wada, H. Harai, W. Chujo, and F. Kubota, "Photonic packet routing based on multi-wavelength

label switch using fiber Bragg gratings," ECOC2000, Vol.4, No.10.4.6, pp.71-72, Sep. 2000.

17 N. Wada, H. Harai, W. Chujo, and F. Kubota, "Photonic network based on multi-wavelength label

switch", IEICE Technical Digest on PNI (PNI2000-24), pp. 45-50, 2000.

18 H. Harai and N. Wada, “Photonic packet forwarding in a multi-wavelength label switching node,"

ICC2001 Workshop on next generation switching/routing: the optical role, WS2-7, Jun. 2001.

19 H. Harai, N. Wada, F. Kubota, and W. Chujo, "A switching node architecture for multi-rate path

establishment in multi-wavelength label switching networks", IEICE Technical Digest on PS

(PS2000-37), pp. 7-12, 2000.

20 K. Kitayama, N. Wada, and H. Sotobayashi, "Architectural considerations of photonic IP router

based upon optical code correlation (Invited)," IEEE J. Lightwave Technol, Vol.18, No.12, Dec. 2000.

21 H. Sotobayashi and K. Kitayama, "325 nm bandwidth supercontinuum generation at 10 Gbit/s using

dispersion-flattened and non-decreasing normal dispersion fibre with pulse compression technique,"

IEE Electron. Lett., Vol.34, p.1336, 1998.

22 N.Wada, H.Sotobayashi, and K.Kitayama, "2.5 Gbit/s time-spread/wavelength-hop optical code

division multiplexing using fibre Bragg grating with supercontimuum light source," IEE Electron. Lett.,

Vol.36, No.9, pp.815-817, Apr. 2000.

23 N. Wada, J.Widjaja, H. Ishii, and W. Chujo, "Photonic packet routing based on spectral holography,

" Optics Japan 2000, Vol. 1, No.7pA7 pp.61-62, Oct. 2000.

24 J. Widjaja, N. Wada, Y. Ishii, and W. Chujo, "Photonic packet address processor using holographic

correlator," IEE Electron. Lett., Vol.37, No.11, pp.703-704, May 2001.

36 Journal of the Communications Research Laboratory Vol.49 No.1 2002

Naoya WADA, Ph. D.

Senior Researcher, Ultrafast PhotonicNetwork Group, Basic and AdvancedResearch Division

Photonic Network, Photonic PacketSwitching, Photonic Node

25 K. Ema, N. Kagi, and F. Shimizu, "Optical compression using a monochromator and a concave mir-

ror," Opt. Commun., 71, p.103, 1989.

26 T. Konishi and Y. Ichioka, "Ultra-fast image transmission by optical time-to-2-D-space-to-time-to-2-

D-space conversion," J. Opt. Soc. Am. A, Vol.16, p.1076, 1999.

27 F. Mok, "Angle-multiplexed storage of 5000 holograms in lithium niobate," Opt. Lett., 18, p.915,

1993.

28 A.Weiner, D.Leaird, D.Reitze, and E.Paek, "Spectrl holography of shaped femtosecond pulses," Opt.

Lett., 17, p.224, 1992.