OPTICS IN - Sunny Bainsmission was obtained with a 12.5% duty ratio. On the other hand, because the download data rate is easily over 100Mbps, the data trans-fer ratio of download

1SPIE’s International Technical Group Newsletter

OPTICS IN INFORMATION SYSTEMS MARCH 2003

OPTICS ININFORMATION

SYSTEMS

MARCH 2003VOL. 14, NO. 1

SPIE’sInternational

TechnicalGroup

Newsletter

NEWSLETTER NOW

AVAILABLE ONLINETechnical Group members are beingoffered the option of receiving theOptics in Information SystemsNewsletter electronically. An e-mail isbeing sent to all group members withadvice of the web location for thisissue, and asking members to choosebetween the electronic and printedversion for future issues. If you are amember and have not yet receivedthis message, then SPIE does nothave your correct e-mail address.

To receive future issues electroni-cally, please send your e-mail addressto:[email protected] the word OIS in the subject lineof the message and the words elec-tronic version in the body of themessage.

If you prefer to receive thenewsletter in the printed format, butwant to send your correct e-mailaddress for our database, include thewords print version preferredin the body of your message.

Calendar —See page 10

Technical Group RegistrationForm —See page 11

Scalable optical interconnectionnetworks for symmetricmultiprocessors (SMPs)

Figure 1. An overview of a single board of SYMNET, which shows the interconnections for fourprocessors. These are connected to dual Y-couplers by optical waveguides/fibers.

Continued on page 7.

nected modules. As each processor accesses thebus it inserts its address request, which is thensnooped by all processors and memory modulesconnected .

Contention problems, combined with the fasterprocessing capabilities of current processors (1-2GHz), degrade the bus’s performance. Further-more, coherence overhead and bus speed(100MHz) limit the number of processors that canbe connected, thereby affecting the scalability ofsnoopy-bus based SMPs. The snoop or address

Market demands and application trends, coupledwith the ever-increasing need for higher perfor-mance and sustained productivity in many scien-tific applications, are accelerating the need forhigh-performance, scalable, parallel computingsystems. Symmetric multiprocessors (SMPs) areattractive parallel computers and are used widelyas they provide a global physical address spaceand symmetrical access to the entire memory, of-fering increased flexibility and programmability.SMPs use fast snoopy protocols to maintain cachecoherence: facilitated by the time-multiplexed,shared bus. Every request is broadcast to all con-

SPIE’s International Technical Group Newsletter2


Athermal design of WDM holographic filtersHolographic filters can be used for WDMapplications, but temperature dependenceis a critical concern in telecommunica-tion. Though we can combat thermal driftusing thin-film filter technology andclever selection of materials—alternatelayers of materials with opposite thermalexpansion coefficients can be depositedto fabricate athermal thin-film filters—such flexibility isn’t available in bulk ho-lographic filters. In this article we presentan athermal design that makes the char-acteristics of WDM filters (central wave-length, bandwidth, etc.) as invariant aspossible with respect to temperature fluc-tuations.

In our experiments, holographic fil-ters are recorded in lithium niobate(LiNbO

3) by interfering two continuous

wave laser beams inside the crystal (seeFigure 1).1 A stabilization system hasbeen incorporated into the recording setupin order to prevent the interference pat-tern from drifting. By properly choosingthe angle 2θ between the recording beams,we’re able to control the Bragg wavelength ofthe grating operated in the reflection geometryas a WDM filter (see Figure 2). The reflectanceof the filter at normal incidence (θ

out=0) is

shown in Figure 3 for three different tempera-tures.

The Bragg wavelength λB (center wave-

length) of the filter at 24.3°C is estimated to be1569.45nm with a bandwidth of 0.134nm(~16.75GHz). The criterion used to calculatethem is as follows: we first find the two pointsof the filter whose through-channel transmit-tances are 0.5dB higher than the minimumtransmittance: these are defined as the bandedges of the filter. The Bragg wavelength isthen calculated as the average of the two edgewavelengths and the bandwidth as the differ-ence between them. We will refer to this as the0.5dB criterion from here on.

By inspecting the grating equation:

2nΛcosθin=λ

B

Λ is the period of the index grating. We con-clude that by tilting the incident beam awayfrom the normal, we’re able to Bragg-matchthe grating to a shorter wavelength.

Temperature changes affect holographic fil-ters mainly through two mechanisms: thermalexpansion or contraction of the bulk material(in our case, LiNbO

3), i.e. the grating period is

a function of temperature; and thermal depen-dence of the dielectric constant of the bulk ma-terial, i.e. the refractive index n is a function oftemperature. There are other possible effects—e.g. the thermal dependence of the piezoelec-tric tensor will manifest itself when stress is

being applied—but they will be neglected here.Within the temperature range of interest, we

may assume that both thermal expansion andthermal-dielectric-constant changes are linear.In addition, we know that both coefficients arepositive.2, 3 The athermal design of the WDMfilters can therefore be implemented as follows:as temperature rises, the Bragg wavelength ofa given filter will shift upward, i.e. towardslonger wavelengths. To compensate for such ashift, we tilt the beam away from the normal.

On the other hand, to undo the effectcaused by a temperature drop, we’ll needto adjust the beam towards the normal.

To verify the statement above, we firstfigure out the 0.5dB criterion center wave-lengths for a series of incident angles atfour different temperatures (22.9°C,41.9°C, 49.1°C, 62.9°C). Temperaturemonitoring is made possible by readingthe resistance off a thermistor in closecontact with the LiNbO

3 crystal when the

whole system is in thermal equilibrium.A thermal-electric cooler is used to con-trol the temperature of the system. Thecenter wavelength corresponding to theincident angle θ

out = 10° (θ

in ≅ 4.5°) at the

lowest temperature is chosen as our tar-get wavelength for angular compensation.For each of the other temperatures, we’reable to pick a center wavelength that’sclosest to the target wavelength along withthe corresponding incident angle. Wetherefore end up with the compensationangle as a function of temperature change.

This is plotted in Figure 4.The fit is done according to the following

formula:

cos(θB + ∆θ) 1

cosθB

(1+ a∆T)(1 + b∆T)

Here a is the thermal expansion coefficient, bis the thermal coefficient of dielectric constant,and θ

B is the Bragg angle corresponding to the

target wavelength when ∆T = 0. We conclude

Figure 1. Recording of a holographic grating.

Figure 2. Holographic grating operated as a WDM filter inreflection geometry.

Figure 3. Filter response measured in the through channel at normal incidence for threedifferent temperatures.


=



Low-power spatial optical interconnection technique forlocation-based information service environmentIn the future, almost all elec-tronic devices will be connectedto the broadband informationenvironment. Location-basedinformation services for deviceslike navigation systems—usingGPS (Global Positioning Sys-tem) or VICS (Vehicle Informa-tion and Communication Sys-tem),1 for example—will be-come increasingly important.To realize the appropriate infor-mation service environment, ahandheld information device—called My Button2—and a basestation—i-lidar 3—have beenproposed. My Button has a spatial optical com-municating module—called HV (hyper versa-tile) target—to communicate with the i-lidar.

Here we report a novel technique for spa-tial optical data interconnection using the HVtarget that uses a liquid crystal light modula-tor, and we describe an approach to getting ahigher data-transfer rate with lower consump-tion.

Hyper versatile (HV) targetHV target is a spatial optical communicationmodule that incudes a corner-cube reflector, thereflectivity of which can be varied with lowpower using a liquid crystal (LC) light modu-lator. HV target has several implementation lev-els as shown in Table 1, from simple (level 0)to complicated (level 5).

UMU film 4 is an attractive LC modulatorbecause of its flexibility and polarization in-sensitivity. Figure 1 shows the principle of

Figure 1. Operation principle of UMU film: (a) OFF-state (translucent); (b) ON-state (transparent).

(a) (b)

UMU film operation. A liquid crystallayer is laminated between two piecesof plastic sheet with a transparent In-dium Tin Oxide (ITO) electrode. In theliquid crystal layer, small droplets ofnematic LC are dispersed in a porouspolymer sheet. When no electric fieldis applied on the device, the LC mol-ecules are randomly aligned. Conse-quently, incident light is scattered andcannot pass straight through: it there-fore gives the device the appearance ofa frosted glass window (translucent). Onthe other hand, when enough of an elec-tric field is applied to the device, themolecules align in the direction of thefield and the light can propagate throughthe device with relatively little scatter-ing (making it look clear). Thus, trans-parent optical intensity can be switchedwith applied voltage without using ex-

ternal polarizing optical deviceslike conventional LC light modu-lators.

Experimental results anddiscussionThe response time for the LC toswitch from translucent to clear(rise time) at AC24V is 0.4ms. Togo from and clear to translucent(fall time) takes 20ms. Figure 2shows modulation bandwidthcharacteristics as a function of on/off duty ratio at 1kHz and 5kHzcarrier frequencies. Because thefall time is much longer than therise time, wider bandwidth wasmeasured at the small duty ratiocondition. Figure 3 shows eye dia-

grams with 50%, 25%, and 12.5% duty ratio,at 30Hz, 100Hz, and 300Hz modulation fre-quency. No eye was observed with 50% dutyratio at 300Hz.

Because the rise time of the UMU film ismuch faster than fall time, a reduction of theduty ratio is effective for higher modulatingspeeds. At the maximum, over 100 bps trans-mission was obtained with a 12.5% duty ratio.

On the other hand, because the downloaddata rate is easily over 100Mbps, the data trans-fer ratio of download to upload can be morethan 106. We call this spatial optical intercon-nection Asynchronous speed Spatial OpticalBroadband Interconnection, ASOBI. Despite itslarge asynchronous data rate, the system canperform useful tasks. For instance, many ap-plications require broadband data download—

Table 1. Implementation levels of HV targets.

Figure 2. Modulation frequency characteristics as a functionof signal duty ratio.




Optical interconnection system basedon fiber image guidesOptoelectronic interconnection systems canprovide high-speed, high-bandwidth data trans-fer with low power and low crosstalk.1 In thesesystems the optical interconnection betweentransmitter and receiver arrays must be donewith low cost, low loss, and tolerance to mis-alignment. Recently, fiber image guides (FIGs)have been used to transfer two-dimensional(2D) optical data packets as an alternative tofree-space interconnections.2,3 Free-space op-tical interconnects are very useful for short-dis-tance 2D parallel interconnects but they are notsuitable for multi-channel interconnects overlong distances (over 1m) because of beamspreading and the difficulty of alignment. FIGsare tightly-packed arrays of thousands of opti-cal fibers, and are capable of 2D parallel im-age transmission with more flexible alignmentand packaging than free-space alternatives.

Figure 1 shows the construction of a typi-cal FIG. Those we used in the experiment areproduced at Schott Optovance, Inc. Their den-sity is approximately 104mm-2, and each com-ponent strand is a step-index multi-mode fiber.The FIGs consist of more than 15,000 of thesecore fibers, each with a 9.1µm core diameterand 13.6µm pitch. Available numerical aper-tures (NA) are 1.0, 0.55, and 0.25, and the at-tenuation level is less than 0.4dBm-1 in thewavelength range from 700nm ~ 1100nm. Theoptical attenuation of 0.25 FIGs is even lowerat 0.1dB/m. The individual fibers are rigidlybounded to each end by a background acid-soluble glass holder 2.5cm thick. Between thesepoints the individual fibers are loose so thatthe FIG can be bent and positioned easily.

We interconnect two or more transceiverarrays by mounting each FIG end close to atransmitter and detector array. Great freedomin alignment and packaging is possible thanksto the FIG flexibility: optical alignment is eas-ily achieved by moving the bundle ends. Wemounted the image guides between two opto-electronic transceiver array boards calledTranspar:4 a modular optoelectronic intercon-nection test system containing a field-program-mable gate array (FPGA) chip forreconfigurable logic, vertical-cavity surface-emitting laser/metal-semiconductor-metal(VCSEL/MSM) smart-pixel devices for opti-cal input/output, and transimpedance receiverarrays.

Transpar can implement both a photonicmulti-token ring network and signal process-ing functions on three-dimensional data pack-ets being transferred among nodes. The distancebetween each VCSEL/detector and FIG is criti-cal in determining optical power and beam col-

limation. There are three different distance vari-ables in our setup. The distance between thetop of the VCSELs and the input end of theFIG is given by d

in, the distance between the

output end and the detector plane is dout

and thedistance along the length of the FIG from theinput end is given by d

FIG.

To optimize the interconnection efficiency,it is important to match the optical spot size tothe detector size. The MSM detector size isabout 75×75µm, so we need the optical spotsize to be less than 75µm in diameter. We ob-served the spot using a CCD camera directlyfrom the image guide output as shown in Fig-ure 2. The output spot size is well defined andless than 75µm in diameter when the couplingdistance d

in is less than 100µm.

A detailed computer simulation ofwaveguide propagation in a FIG was investi-gated using RSoft’s BeamPROP. The simula-tions were performed on different x-y-z direc-tional alignments to analyze the optical poweroutput. The maximum power output is obtainedwhen the center of the laser aperture is wellaligned to the core fiber. Figure 3 shows onesimulation result with d

in = 100µm, d

FIG =

200µm, dout

= 0µm. The core fiber boundariesare shown by white circles. For the simulationwe launched a 15µm-wide laser in variouslaunch positions. These simulations are wellsupported by the experimental results.

Sunkwang Hong and Alexander A.SawchukSignal and Image Processing InstituteIntegrated Media Systems CenterUniversity of Southern California, LosAngeles, CA 90089-2564Tel: +1 213/740-4622Fax: +1 213/740-4651E-mail: [email protected]

References1. Special Issue on Optical Interconnections for

Digital Systems, Proc. IEEE 88 (6), June 2000.2. D. M. Chiarulli, S. P. Levitan, P. Derr, R.

Hofmann, B. Greiner, and M. Robinson,Demonstration of a Multichannel OpticalInterconnection by Use of Imaging Fiber BundlesButt Coupled to a Optoelectronic Circuits, Appl.Opt . 39, pp. 698-703, 2000.

3. S. Hong and A. A. Sawchuk, Testing andSimulation of an Optical Interconnection SystemUsing Fiber Image Guides, Proc. SPIE 4987, 2003

4. S. Hong, Z. Y. Alpaslan, L. Zhang, C. Min, and A.A. Sawchuk, Testing of Reconfigurable TranslucentSmart Pixel Array (R-Transpar) for Optical Multi-Token-Ring Network, International TopicalMeeting on Optical Computing, (OC 2002), pp.141-143, Taipei, Taiwan, April 2002.

Figure 1. Fiber image guide structure.

Figure 2. FIG output image of Honeywell VCSEL-MSM chip.

Figure 3. Transverse field profile outputs of aVCSEL through fiber image guides.



Holding light assists four-wave-mixing-basedsemiconductor wavelength convertersAll-optical wavelength conversion usingthe four-wave mixing (FWM) effect in asemiconductor optical amplifier (SOA) canprovide functions necessary for optical net-works. However, the technique suffers frompoor conversion efficiency. Significant im-provement both here and in the noise fig-ure are required to make the technologypractical for application. FWM conversionefficiency increases approximately with thesquare of the unsaturated gain and satura-tion intensity.1 Consequently, increasing thegain and/or saturation power of a SOA isthe key to improve conversion efficiency.A holding light, also called an assistingbeam, was suggested as a way of increas-ing the saturation intensity of the SOA.2,3

We will demonstrate that using such a beamcan also improve the efficiency and signal-to-noise ratio of a FWM-based wavelengthconverter.

The effect of an assisting beam on wave-length conversion depends on what happensas the beam passes through the SOA. Whenthe device is operated in the gain regimefor the assisting light, the saturation inten-sity is raised but the amplifier gain lowered.If the SOA is biased in the absorptive re-gime, both its gain and saturation power canbe increased. However the current required forthis condition is usually too small to provideenough gain for wavelength conversion. Fi-nally, if the SOA is operated at or close to thetransparent condition for the assisting beam,the saturation intensity can be enhanced with-out sacrificing the gain.

Figure 1 shows the schematic of a light-as-sisted wavelength converter. A high-power as-sisting beam is propagates through the SOA inthe opposite direction as the pump and probebeams. The latter two beams interact inside theSOA to generate a conjugated signal throughnonlinear effects. To prove the improvementin wavelength-conversion efficiency with anassisting beam, the measurement was carriedout using two different wavelengths: 1480nmor 1430nm. The output signal was measuredwith an optical spectrum analyzer (OSA) thatcan clearly distinguish the conjugated signalfrom the amplified spontaneous emission noise(ASE). The conversion efficiency can be en-hanced by 5.8dB at the corresponding trans-parent current for the 1480nm assisting beamwhen the detuning (frequency spacing betweenthe pump and probe wavelengths) is 100Gbps.The enhancement increases slightly as the

detuning becomes larger. The enhancementdrops to 3.3dB when the SOA bias is raised to200mA. This is due to the fact that the assist-ing beam increases the saturation power butdecreases the available gain.

In principle, it is possible to use a shorter-wavelength assisting beam—one whose trans-parent current is higher—to obtain larger en-hancement at high SOA bias. Figure 2a showsthat the assisting beam provides for larger im-provement at 1430nm than 1480nm for thesame assisting power (18dBm). In fact, the dif-ference between the results of using 1480nmor 1430nm is not as significant as expected dueto the fact that the gain for the assisting beamdepends on its wavelength. With a 200mA bias,which is closer to the transparent current forthe 1430nm beam, the net gain is larger for the1480nm assisting beam. Therefore, the largerenhancement to wavelength conversion effi-ciency by using a shorter-wavelength assistingbeam is compensated by the smaller net gain.The advantage gained by using a shorter-wave-length assisting beam is more pronounced athigher SOA bias currents, where the 1480nmbeam depletes more carriers than the 1430nmbeam does. Also, the gain difference provided

by the SOA to the assisting beam of differentwavelengths tends to be smaller at higher bias.

In addition to the improvement to conver-sion efficiency, the assisting beam can increasethe signal-to-background-noise ratio (SBR) byabout 3dB and 1.5dB at the bias levels of105mA and 200mA, respectively. The mea-sured FWM spectra indicate that the presenceof an assisting beam also increases ASE, but itis less than the increase in the conjugated sig-nal. The improvement on the conversion effi-ciency and SBR by using an assisting beam canbe seen clearly by measuring the eye-diagramfor the received signals when the probe beamis modulated at 10Gbps data rate. The resultsare shown in the inset to Figure 2b. The 1480nmassisting beam increases the signal amplitudeby >5dB at 105mA and by >1.5dB at 200mA.Without using an assisting beam, the conju-gated signal after wavelength conversion suf-fers from a 4.4dB power penalty due to thecontribution from the signal-spontaneous-emis-sion noise, as shown in Figure 2b. The powerpenalty is reduced by 2.6dB when an assistingbeam is applied. This proves that the assistingbeam improves not only the conversion effi-ciency but also the signal quality of a FWM-

Figure 1. Schematic of wavelength conversion using the FWM effect of a SOA and an assisting beam.

Figure 2. Performance enhancement of wavelength conversion by using an assisting beam: (a) improvementof conversion efficiency at different bias currents for 1480nm and 1430nm assisting beams; (b) bit error rateand eye diagrams.

Continues on page 9.



Integrated optical devices basedon a lithium niobate substrateWe have developed integrated optical devicesbased on lithium niobate (LiNbO

3). Integrated

optical bistable circuits consisting of two par-allel reversed-∆-directional couplers intercon-nected by single mode waveguides were fabri-cated on a single LiNbO

3 crystal, as shown in

Figure 1.1 Hybrid optical bistable performancewas demonstrated. By controlling parametersof the applied dc voltage, both the complemen-tary and consistent types of hysteresis loops anddifferential curves are obtained. Optical logicgates and optical amplifiers can therefore beconstructed.

In information and optical communicationsystems, external modulation of laser sourceshas been employed for high speed applications.The Mach-Zehnder modulator based onLiNbO

3 has been widely used for this purpose.

However, the linearity of the output optical sig-nal is a critical issue because of the inducedBessel-function distortions of the Mach-Zehnder modulator. The cascaded modulatorstructure was therefore investigated to reducethe harmonic and intermodulation distortions.2

Distortions of -47dB and -39dB can be ob-tained, respectively, using this structure.

To evaluate the performance of integratedoptical devices used in information and opti-cal communication systems, the propagationloss for an optical channel waveguide is a veryimportant issue. We developed a nondestruc-tive method for measuring the propagation andbending losses of separated individual modeswithin single- and multimode waveguides. Thisnovel technique is a combination of the butt-couple, prism-couple, and phase-modulationmethods.3,4

To reduce the bending propagation loss forintegrated optical devices, the microprismstructure was investigated. To construct a low-loss and wide-branching-angle symmetric Y-junction waveguide, a novel structure with fullphase compensation was used.5,6 The structure,as shown in Figure 2, consists of a pair of mi-croprisms for pre-compensation and a micro-prism for phase-front acceleration. A normal-ized transmitted power of more than 95% isobtained, even though the branching angle ofthe Y-junction waveguide is up to 20°.

Ching-Ting LeeInstitute of Optical SciencesNational Central UniversityChung-Li, TaiwanRepublic of China

Figure 1. Experimental setup of the complementary optical bistable operator.

Figure 2. Top view of the full-phase-compensation microprism-type Y-junction waveguide.

References1. C.T. Lee, H.C. Lee, H.H. Lai, and L.G. Sheu,

Complementary optical bistable operation withintegration of two directional couplers on LiNbO

3

crystal, Jpn. J. Appl. Phys. 35 (5A), p. 2686, May1996.

2. C.T. Lee, H.C. Lee, and L.G. Sheu, The reductionof harmonic and intermodulation distortions with acascaded Mach-Zehnder modulator, Opt. Rev. 3(5), p. 341, 1996.

3. C.T. Lee, Nondestructive measurement ofseparated propagation loss for multimodewaveguides, Appl. Phys. Lett. 73 (2), p. 133, July,1998.

4. C.T. Lee, Nondestructive measurement ofindividual modes loss for waveguides, U.S. PatentUS6219475B1.

5. J.M. Hsu and C.T. Lee, Design of microprism-typesymmetric Y-junction waveguides with the fullphase compensation method, Appl. Opt. 38 (15),p. 3234, May 1999.

6. J.M. Hsu and C.T. Lee, Systematic design of novelwide-angle low-loss symmetric Y-junctionwaveguides, IEEE J. Quantum Electron. 34 (4),p. 673, April 1998.



bandwidth required to broadcast the address re-quests is, therefore, the major limitation ofSMPs. Electrical solutions to this address band-width have ranged from using split-transactionor multiple-address buses, or a data crossbar(such as UE100001 from Sun) that can scaleup to 64 processors. Beyond this, the addressbuses are saturated.

Optical technology can provide the highcommunications bandwidth needed to solve theaddress bandwidth limitation. The key technol-ogy is the low-cost, parallel, optical intercon-nects consisting of an array of vertical-cavitysurface-emitting lasers (VCSELs) and photo-detector (PD) transceivers with data rates in ex-cess of 3Gb/s per channel.2 An array of suchtransceivers enable address transmission at arate of 200-300Gbps, which will be the addressbandwidth required by future SMPs. The pro-posed optical symmetric multiprocessor, calledSYMNET, consists of four processors intercon-nected using a binary tree is shown in Figure1. The tree is constructed using a dual array ofY-couplers that allow two-way address trans-mission (see inset to Figure 1). The up-streamarray of Y-couplers is used to combine the ad-dress requests from the processors. After reach-ing higher levels, these address requests are re-routed through the downstream array of Y-split-ters, enabling the broadcast of the address re-quests to all the processors and memories.

Time-division multiple-access (TDMA)protocol is used as a control mechanism toachieve mutually-exclusive access to the sharedchannel. Optical token propagation ensures thataddress requests are transmitted from succes-sive processors without collision. This way,address requests from different processors canbe pipelined and more than one address requestcan be inserted into the tree interconnect.3 Theoptical token is a single pulse that provides atime reference for insertion of address requestsinto the tree interconnect by the individual pro-cessor. As shown in Figure 1, in cycle 1(square) the optical token is received by pro-cessor 1, which in turn transmits an addressrequest. In cycle 2, when this address from pro-cessor 1 is in propagation at the next level ofthe tree, the token is received by processor 2,which can transmit its own address request(shaded diamonds). In cycle 3, the address re-quest from processor 1 is being re-routed us-ing the downward Y-splitter, and at the sametime the address request from processor 2 hasmoved up the binary tree. The optical token isnow received by processor 3, which can trans-mit an address. In cycle 4, the address requestfrom processor 1 has reached all the proces-sors, and all snoop for it simultaneously. Broad-casting the address request results in simulta-neous reception and snooping of the request

by all the processors/memory modules. Theaddress requests are thereby serialized.

A possible optical implementation ofSYMNET using parallel optical transceivers,polymer waveguides, arrays of Y-couplers/splitters, and semiconductor optical amplifiers,is shown in Figure 2. Parallel optical links—1D and 2D (1×12, 4×12, and 6×12) VCSELarrays operating in the 850-980nm wavelengthrange at 3Gbps/channel data rate—are commer-cially available. There are several methods forintegrating VCSELs with optical receivers suchas standard complementary metal-oxide silicon(CMOS) circuitry. These include substrate re-moval, coplanar flip-chip bonding (for bottom-emitting VCSELs), and various top-contactbonding methods.4 We use polymerwaveguides to route the optical address requestsfrom the discrete processors as they exhibitexcellent thermal stability, low optical loss andmechanical robustness.5 Y-couplers/splittersare efficient at splitting light, but the associ-ated losses are tremendous for a given BER.Hence, we use an array of semiconductor opti-cal amplifiers (SOAs) at the root of the tree toboost the power of the optical address request.SOAs operating in the 980nm range provide afiber-to-fiber gain of 20dB with a response timeof less than 60psec.6 The address request is fedto the CMOS transmitter IC through the ad-dress-port controller, which in turn drives theVCSELs. The direction of the light launchedinto the waveguides is changed by 90° using a45° reflector. The address pulses propagate in

the waveguides through different levels of cou-plers and splitters. At the receiving end of thewaveguides, the direction of light pulses is onceagain changed by 90°. The light pulses areeventually received by the photodetector afteramplification, and are returned to the address-port controller from the CMOS receiver IC forprocessing the received address request.

Figure 3 shows the layout of the tree-basedaddress sub-network interconnecting the vari-ous processors through different levels of thesystem boards. At the intra-board level (k=0),2-4 processor/memory modules are connectedusing polymer-waveguide-based arrays of up-stream Y-couplers and down-stream Y-split-ters. At the next level (k=1), similar couplersand splitters are connected from different in-tra-boards. At the highest level (k=2), these up-stream and down-stream couplers are con-nected to an array of SOAs. The optical pulsesare routed using waveguides at the intra-board/inter-board levels, and between boards usingpolymer fibers. MT-ferrule-type push-pulls areused to connect the waveguides and fibers.7 Thesystem can be scaled in powers of two by sim-ply replicating it and then adding to it. The ad-dition of newer boards is facilitated by discon-necting the root board (level k=n), insertinganother board between it and level k=n-1, andthen re-connecting the root board to the newone. Hence, SYMNET provides easy additionof boards without significantly altering the ex-

Scalable optical interconnection networks for symmetric multiprocessors (SMPs)Continued from cover.

Figure 2. The proposed optical implementation of SYMNET using parallel optical interconnects, polymerwaveguides, SOAs, and an array of Y-couplers/splitters, is shown.




isting architecture. This facilitates thescalability of the address sub-network.

Cache coherence cannot be maintained byusing standard snoopy protocols: they must bemodified to take into account the simultaneousinsertion of address requests into the tree in-terconnect. By adding transient states to theexisting stable states, it is possible to satisfycache coherence and sequential memory con-sistency requirements. The theoretical power-budget analysis reveals a high scalability: asmany as 128 processors could be supported bythis architecture for a BER of 10-15 at an oper-ating wavelength of 980nm.

Ahmed Louri and Avinash Karanth KodiDepartment of Electrical and ComputerEngineeringUniversity of ArizonaTucson, AZ 85721E-mail: [email protected]

References1. A. Charlesworth, STARFIRE: Extending the SMP

Envelope, IEEE Micro 18 (1), pp. 39-49,1998.

Scalable optical interconnection networksfor symmetric multiprocessors (SMPs)Continued from page 7.

2. Parallel Links transform network equipment,FiberSystems International, pp. 29-32, Feb 2002.

3. Donald M. Chiarulli, Stephen P. Levitan, Rami G.Melhem, Manoj Bidnurkar, Robert Ditmore,Gregory Gravenstreter, Zicheng Guo, ChungmingQiao, Majd F. Sakr, and James P. Teza, Optoelec-tronic Buses for High-Performance Computing, inProc. IEEE 82 (11), pp. 1701-1710, 1994.

4. R. Pu, E. M. Hayes, C. W. Wilmsen, K. D.Choquette, H. Q. Hou, and K. M. Geib, Compari-son of techniques for bonding VCSELs directly toICs, J. Opt. A: Pure and App. Opt. 1 (2), pp. 324-329, 1999.

5. Louay Eldada and Lawrence W. Shacklette,Advances in Polymer Integrated Optics, IEEE J.Selected Topics in Quantum Electronics 6 (1),pp. 54-68, Jan/Feb 2000.

6. D. Wiedenmann, C. Jung, M. Grabherr, R. Jager,U. Martin, R. Michalizk, and K. J. Ebeling, Oxide-confined vertical-cavity semiconductor opticalamplifier for 980nm wavelength, Technical DigestLasers and Electro-Optics, pp. 378-379, 1998.

7. Y. S. Liu, R. J. Wojnarowski, W. A. Hennessy, J. P.Bristow, Yue Liu, A. Peczalski, J. Rowlette, A.Plotts, J. Stack, M. Kader-Kallen, J. Yardley, L.Eldada, R. M. Osgood, R. Scarmozzino, S. H. Lee,V. Ozgus, and S. Patra, Polymer OpticalInterconnect Technology (POINT)-OptoelectronicPackaging and Interconnect for Board andBackplane Applications, Proc. ElectronicComponents and Tech. Conf., pp.308-315, 1996.

Figure 3. Shown is a complete layout illustrating the interconnections between discrete proces-sors/memories using polymer waveguides, with fiber ribbons at the intra-/inter-board levels.

routed onto a 2D optoelectronic chip in the cen-ter that is set up as a smart pixel array. It con-tains detectors, via which the switch matrix canbe programmed optically, and modulators forthe actual switching. The linear detector arraythat acts as the final stage of the crossbar switchhas not yet been bonded to the substrate in thesetup of Figure 2. It could be shown that opti-cal signals can be reliably routed onto their re-spective targets with this demonstrator.

For bridging larger interconnection dis-tances, optical fiber links can be attached toplanar-integrated MOEMS. Experimentally, anapproach using fiber ribbons with standard MT-type connectors that are attached via a micro-machined socket plate was realized.5 This al-lows one to connect cables with 12 optical fi-bers. Here, reliable fiber-free-space opticallinks could also be established. Future researchnow has to tackle the problem of improvingthe coupling efficiency: it’s been ignored sofar in the above proof-of-principle experiments.

Matthias Gruber and Jürgen JahnsFernUniversität HagenUniversitätsstr. 2758084 Hagen, GermanyTel: +49 2331 987 1131Fax: +49 2331 987 352E-mail: [email protected]

References1. J. Jahns and A. Huang, Planar integration of free-

space optical components, Appl. Opt. 28, p. 602,1989.

2. S. Sinzinger and J. Jahns, Integrated micro-opticalimaging system with a high interconnectioncapacity fabricated in planar optics, Appl. Opt.36, p. 4729, 1997.

3. D. Fey, W. Erhardt, M. Gruber, J. Jahns, H. Bartelt,G. Grimm, L. Hoppe, and S. Sinzinger, OpticalInterconnects for Neural and Reconfigurable VLSIArchitecture, Proc. IEEE 88, p. 838, 2000.

4. W. Eckert, S. Sinzinger, and J.Jahns, Compactplanar-integrated optical correlator for spatiallyincoherent signal, Appl. Opt. 39, p. 759, 2000.

5. M. Gruber, J. Jahns, E. El Joudi, and S. Sinzinger,Practical Realization of Massively Parallel Fiber-Free-Space Optical Interconnects, Appl. Opt. 40,p. 2902, 2001.

Planar integration ofcomplex systemsfor opticalcommunication andcomputingContinued from page 12.



that, by slightly tilting the incident beam, we’reable to compensate for Bragg wavelength driftsdue to changes in the ambient temperature.

Our data suggest that, for operation aroundan incident angle θ

out of 10°, an angular cor-

rection of 0.88° will be required for a tempera-ture change of 100°C. Such an angular fine-tuning can be achieved by a bimetallic com-posite beam, which makes use of the TEC (ther-mal expansion coefficient) discrepancy be-tween two properly chosen materials.4 The prin-ciple of operation is illustrated in Figure 5. Thealuminum-silicon composite beam will be de-signed in such a way that it deflects 0.44° for atemperature change of 100°C.

Athermal design of WDM holographic filters

Figure 4. Compensation angle for thermal drift ofBragg wavelength as a function of temperaturechange.

Figure 5. Athermal design utilizing an Al-Sicomposite beam microactuator.

Continued from page 2.

Hung-Te Hsieh, George Panotopoulos, andDemetri PsaltisCalifornia Institute of TechnologyMS 136-93Pasadena, CA 91125E-mail: [email protected]

References1. G. A. Rakuljic and V. Leyva, Volume holographic

narrow-band optical filter, Optics Letters 18 (6),p. 459, 15 March 1993.

2. R. T. Smith and F. S. Welsh, Temperaturedependence of elastic, piezoelectric, and dielectricconstants of lithium tantalite and lithium niobate,Journal of Applied Optics 42 (6), p. 2219, May1971.

3. U. Schlarb and K. Betzler, Refractive indices oflithium niobate as a function of temperature,wavelength, and composition – a generalized fit,Physical Review B 48 (21), p. 15613, 1 December1993.

4. W. H. Chu, M. Mehregany, and R. L. Mullen,Analysis of tip deflection and force of a bimetalliccantilever microactuator, Journal ofMicromechanics and Microengineering 3, p. 4, 5February 1993.

as high-resolution movie, for instance—butonly need require small bits in response: suchas YES, 1, or A, etc.. Furthermore, since the lo-cation or direction of the My Button can be de-tected by the i-lidar indoor laser radar commu-nication system without location informationbeing explicitly sent, Button holders can auto-matically /receive various information servicesbased on its location.

Hideo Itoh, *Takeshi Akiyama, TakuichiNishimura, Yoshinov Yamamoto,*Takehiko Hidaka, and HideyukiNakashimaCyber Assist Research Center,AIST CREST, JST and*Shonan Institute of Technology

Low-power spatial optical interconnection techniqueContinued from page 3.

Figure 3. Eye diagram of data transmission.

References1. http://www.vics.or.jp/eng/

index.html2. H. Nakashima and K. Hashida,

Location-based communicationinfrastructure for SituatedHuman Support, Proc. ofWorld Multiconference onSystemics, Cybernetics andInformatics 2001 (SCI2001)IV, p47 2001.

3. H. Itoh, S. Yamamoto, M.Iwata, and Y. Yamamoto, Guestguiding system based on theindoor laser radar system usingHV targets and a frequencyshifted feedback laser, Proc. ofInt’l Topical Meeting onContemporary PhotonicsTechnologies 2000 (CPT2000),Tc-23, pp.117-118 (2000).

4. http://www1.nsg.co.jp/umu/

based wavelength converter.

San-Liang LeeProfessor, Dept. of Electronic EngineeringDirector, Center for Optoelectronic Scienceand TechnologyNational Taiwan University of Science andTechnologyTel: +11 886 2737 6401Fax: +11 886 2737 6424E-mail: [email protected]

References1. A. D’Ottavi, F. Martelli, P. Spano, A. Mecozzi, and

S. Scotti, Very high efficiency four-wave mixing ina single semiconductor traveling-wave amplifier.Appl. Phy. Lett. 67, pp. 2753-2755, 1995.

2. M. A. Dupertuis, J. L. Pleumeekers, T. P. Hessler,P. E. Selbmann, B. Deveaud, B. Dagens, and J. Y.Emery, Extremely fast high-gain and low-currentSOA by optical speed-up at transparency, IEEEPhoton. Technol. Lett., 12 (11), pp. 1453-1455,Nov 2000.

3. M. Yoshino and K. Inoue, Improvement ofsaturation output power in a semiconductor laseramplifier through pumping light injection, IEEEPhoton. Technol. Lett. 8 (1), pp. 58-59, Jan 1996.

Holding light assists four-wave-mixing-basedsemiconductor wave-length convertersContinued from page 5.



C A L E N D A R

Featured Publication

Optical Information Processing:A Tribute to Adolf Lohmann

Editor: H. John CaulfieldH. John CaulfieldH. John CaulfieldH. John CaulfieldH. John Caulfield␣

Adolf Lohmann's significant contributions to the field of optics arerecognized by key people whose own works have been based on orinfluenced by his ground-breaking discoveries and inventions.Chapters describe the history, research, and latest trends in opticalinformation processing.

SPIE PRESS Vol. PM117 PM117 PM117 PM117 PM117 • July 2002 • 370 pagesHardcover • ISBN 0-8194-4498-7SPIE Member $54; List Price $66SPIE Member $54; List Price $66SPIE Member $54; List Price $66SPIE Member $54; List Price $66SPIE Member $54; List Price $66

To order visit:

www.spie.org/bookstore

[email protected] • Tel: +1 360 676 3290 • Fax: +1 360 647 1445

2003SPIE's AeroSenseAerospace/Defense Sensing,Simulation, and Controls21–25 AprilOrlando, Florida USAProgram • Advance Registration Ends 10 April 2003ExhibitionURL: spie.org/conferences/Programs/03/or/

Optical Data Storage11–14 MayVancouver, BC CanadaSPIE is a co-sponsorProgramURL: www.osa.org/meetings/topicals/ODS/

Microtechnologies for theNew Millennium 200319–21 MayMaspalomas, Gran Canaria, Canary Islands, SpainProgram • Advance Registration Ends 9 May 2003URL: spie.org/conferences/Programs/03/emt/

Optics in Computing18–20 JuneWashington, DC USASPIE is a cooperating organizationProgramURL: www.osa.org/meetings/topicals/OC/

The International Symposium on OpticalScience and TechnologySPIE's 48th Annual Meeting3–8 AugustSan Diego, California USACall for Papers • Abstracts Due 6 January 2003ExhibitionURL: spie.org/conferences/calls/03/am/

ITCom 2003Information Technologies and Communications7–11 SeptemberOrlando, Florida USACo-located with NFOECCall for Papers • Abstracts Due 10 February 2003URL: spie.org/conferences/calls/03/it/

International Conference onMicro- and Nanoelectronics6–10 OctoberZvenigorod, RussiaSponsored by SPIE Russia Chapter.Program

2004Photonics West24–29 JanuarySan Jose, California USAProgram

For More Information ContactSPIE • PO Box 10, Bellingham, WA 98227-0010

Tel: +1 360 676 3290 • Fax: +1 360 647 1445 • E-mail:[email protected] • Web: www.spie.org



Optics in Information SystemsThis newsletter is published semi-annually by SPIE—The International Society for

Optical Engineering, for its International Technical Group on Optics in Information Systems.

Editor and Technical Group ChairsBahram Javidi Technical Editor Sunny BainsDemetri Psaltis Managing Editor Linda DeLano

Articles in this newsletter do not necessarily constitute endorsement or the opinionsof the editors or SPIE. Advertising and copy are subject to acceptance by the editors.

SPIE is an international technical society dedicated to advancing engineering, scientific, andcommercial applications of optical, photonic, imaging, electronic, and optoelectronic technologies.

Its members are engineers, scientists, and users interested in the development and reduction to practiceof these technologies. SPIE provides the means for communicating new developments and applicationsinformation to the engineering, scientific, and user communities through its publications, symposia,education programs, and online electronic information services.

Copyright ©2003 Society of Photo-Optical Instrumentation Engineers. All rights reserved.

SPIE—The International Society for Optical Engineering, P.O. Box 10, Bellingham, WA98227-0010 USA. Tel: +1 360 676 3290. Fax: +1 360 647 1445.

European Office: Karin Burger, Manager, [email protected], Tel: +44 7974 214542.Fax: +44 29 2040 4873.

In Japan: c/o O.T.O. Research Corp., Takeuchi Bldg. 1-34-12 Takatanobaba, Shinjuku-ku,Tokyo 160, Japan. Tel: +81 3 3208 7821. Fax: +81 3 3200 2889. E-mail: [email protected]

In Russia/FSU: 12, Mokhovaja str., 121019, Moscow, Russia. Tel/Fax: +7 95 202 1079.E-mail: [email protected]

Join the Technical Group...and receive this newsletter

Membership Application

Please Print ■■ Prof. ■ ■ Dr. ■ ■ Mr. ■ ■ Miss ■ ■ Mrs. ■ ■ Ms.

First Name, Middle Initial, Last Name ______________________________________________________________________________

Position _____________________________________________ SPIE Member Number ____________________________________

Business Affiliation ____________________________________________________________________________________________

Dept./Bldg./Mail Stop/etc. _______________________________________________________________________________________

Street Address or P.O. Box _____________________________________________________________________________________

City/State ___________________________________________ Zip/Postal Code ________________ Country __________________

Telephone ___________________________________________ Telefax _________________________________________________

E-mail Address/Network ________________________________________________________________________________________

Technical Group Membership fee is $30/year, or $15/year for full SPIE members.

❏ Optics in Information SystemsTotal amount enclosed for Technical Group membership $ __________________

❏ Check enclosed. Payment in U.S. dollars (by draft on a U.S. bank, or international money order) is required. Do not send currency.Transfers from banks must include a copy of the transfer order.

❏ Charge to my: ❏ VISA ❏ MasterCard ❏ American Express ❏ Diners Club ❏ Discover

Account # ___________________________________________ Expiration date ___________________________________________

Signature ___________________________________________________________________________________________________(required for credit card orders)

OPTICSONLINE

Optics WebDiscussion Forum

You are invited to participate inSPIE’s online discussion forum onOptics in Information Systems. Topost a message, log in to create auser account. For options see“subscribe to this forum.”

You’ll find our forums well-designed and easy to use, withmany helpful features such asautomated email notifications,easy-to-follow ‘threads,’ andsearchability. There is a full FAQfor more details on how to use theforums.

Main link to the new Optics inInformation Systems forum:h t t p : / / s p i e . o r g / a p p /

forums/tech/Related questions or

suggestions can be sent [email protected].

T his newsletter is printed as a benefit ofthe Optics in Information Systems Tech-

nical Group. Membership allows you to com-municate and network with colleagues world-wide.

As well as a semi-annual copy of the Opticsin Information Systems newsletter, benefits in-clude SPIE’s monthly publication, oemagazine,and a membership directory.

SPIE members are invited to join for thereduced fee of $15. If you are not a member ofSPIE, the annual membership fee of $30 willcover all technical group membership ser-vices. For complete information about SPIEmembership and an application form, pleasecontact us.

Send this form (or photocopy) to:SPIE • P.O. Box 10Bellingham, WA 98227-0010 USATel: +1 360 676 3290Fax: +1 360 647 1445E-mail: [email protected]

http://www.spie.org/info/ois

Please send me■■ Information about full SPIE membership■■ Information about other SPIE technical groups■■ FREE technical publications catalog

Reference Code: 3537



P.O. Box 10 • Bellingham, WA 98227-0010 USA

Change Service Requested

DATED MATERIAL

Non-Profit Org.U.S. Postage Paid

Society ofPhoto-Optical

InstrumentationEngineers


Planar integration of complex systemsfor optical communication and computingThe tremendous increase in per-formance of communication andcomputing hardware—whichwe have been observing formore than three decades—isbased on a continued miniatur-ization and integration intomonolithic blocks of ever morecomponents and functionalities.In recent years, considerable ef-fort has been made to transferthis concept to neighboring dis-ciplines, in particular to mechan-ics and optics, to achieve a simi-lar performance boost there.This will eventually lead to pow-erful complex micro-opto-electro-mechanical systems(MOEMS). However, integra-tion approaches in these disci-plines have to be sufficiently compatible withthe (de-facto) standards set by microelectronicsin order to be adopted and applied in commer-cial systems. For that reason, the concept of pla-nar-integrated free-space optics1 was developed.

Planar integration of free-space opticsThis concept provides compatibility with mi-croelectronics both in terms of dimensions andfabrication technologies. It thus makes the pow-erful parallel information processing tools offree-space optics available for VLSI. The ideais to miniaturize and “fold” a free-space opti-cal setup into a transparent substrate of, typi-cally, a few millimeter thickness in such a waythat all optical components fall onto the plane-parallel surfaces. Passive components—likelenses or diffraction gratings—can then be re-alized through surface relief structuring, whileactive components (like optoelectronic chipswith arrays of emitters, modulators, or detec-tors) can be bonded onto the substrate. A re-flective coating ensures that optical signalspropagate along zigzag paths inside the sub-strate (see Figure 1).

Since all passive components are located atthe substrate surfaces, the optical system can

sophisticated etching techniques ofthe semiconductor industry can beapplied. Due to the integration intoa rigid substrate, the passive opti-cal system does not require furtheradjustment and it is well-protectedagainst deleterious environmentalinfluences.

From a topological point ofview, the planar-integration ap-proach yields a fully three-dimen-sional (3D) system architecture butrequires only two-dimensional(2D) fabrication complexity. Thisallows one to implement highly-parallel and complex (regular andirregular) optical interconnectschemes involving large fan-out,fan-in, and filtering operations thatwould be difficult to implement

otherwise. Experimental demonstrations in-clude point-to-point interconnection of 2500channels in a multi-chip module architecture,2

optical 1-to-10 fan-out and 10-to-1 fan-in inan integrated 10×10 crossbar architecture,3 andoptical correlation.4 In all three cases, the chan-nel density is typically 400mm-2. Thus, crucialrequirements for interconnects—such as thoseenvisaged by the ITRS roadmaps for the nearfuture—can be met.

Planar-integrated MOEMSIn addition to housing the free-space opticalsystem, the transparent substrate can also serveas a board for electrical circuitry as indicatedin Figure 1. In an experimental demonstration,electrical wires and bond pads were realizedthrough lithographic structuring and evapora-tion of metals (aluminum, gold). Optoelectronicchips were then flip-chip bonded onto the trans-parent substrate. Figure 2 shows an optoelec-tronic MOEMS fabricated in this way. It com-prises three chips and implements a 10×10crossbar switch.3 Optical signals are generatedby two linear arrays of VCSEL diodes and

Figure 1. Schematic of planar-integrated MOEMS with free-space opticalinterconnects.

Figure 2. Experimental demonstratorimplementing an optical crossbar switch.

be fabricated as a whole using mask-based tech-niques. This approach provides lithographicprecision concerning the positioning of com-ponents, and high accuracy for the required re-lief structures because the well-established and

Documents

OPTICS IN - Sunny Bainsmission was obtained with a 12.5% duty ratio. On the other hand, because the download data rate is easily over 100Mbps, the data trans-fer ratio of download