throbber
JOURNAL OF LIGHTVVAVE TECHNOLOGY, VOL. 17, NO 5. MAY 1999
`
`Wavelength Add—Drop Switching
`Using Tilting Micromirrors
`
`Joseph E. Ford, Vladimir A. Aksyuk, David J. Bishop, and James A. Walker
`
`Abstract— This paper describes a single-mode optical fiber
`switch which routes individual signals into and out of a wave-
`length multiplexed data stream without interrupting the re-
`maining channels. The switch uses free-space optical wavelength
`multiplexing and a column of micromechanical tilt-mirrors to
`switch 16 channels at 200 GHZ spacing from 1531 to 1556 nm.
`The electrostatically actuated tilt mirrors use an 80 V peak-to-
`peak 300 KHz sinusoidal drive signal to switch between i10°
`with a 20 ps response. The total fiber-to-fiber insertion loss for
`the packaged switch is 5 dB for the passed signals and 8 dB for
`added and dropped signals, with 0.2 dB polarization dependence.
`Switching contrast was 30 dB or more for all 16 channels and all
`input and output states. We demonstrate operation by switching
`622 Mb/s data on eight wavelength channels between the two
`input and output ports with negligible eye closure.
`
`Index Terms—Gratings, microelectromechanical devices, opti-
`cal communications, optical fiber switches, wavelength division
`multiplexing (WDM).
`
`1.
`
`INTRODUCTION
`
`ONVERTING fiber transmission systems from single
`wavelength to wavelength division multiplexing (WDM)
`provides inexpensive bandwidth but can sacrifice routing flex-
`ibility, because diverting part of the traffic in a simple WDM
`line system to an intermediate destination requires that all
`of the remaining wavelength signals must be detected and
`regenerated. As the number of wavelengths increases to 40
`or more, the cost of providing dense WDM repeaters on the
`transrrritted channels becomes prohibitive. These repeaters can
`be eliminated rising wavelength add—drop: a transparent optical
`component to divert selected wavelength signals out of a
`WDM transmission liI1e and also add new signals to reuse the
`dropped wavelengths [1]. Fixed wavelength addazlrop (WAD)
`on a moderate number of channels can be accomplished with
`a set of notch filters [2]. But fully reconfigurable (electrically
`controlled) WAD allows effieient bandwidth allocation and
`fault recovery. Efficient, high-contrast WAD switching has
`beconre a high priority, especially for metropolitan networks.
`Reconfigurable WAD switches can be assembled from dis-
`crete wavelength multiplexers and switches (e.g., connecting
`microoptomechanical
`1 X l switches to arrayed waveguide
`routers [3]) or using an optical circulator with reflective
`fiber grating notch filters tuned by either temperature [4] or
`
`magnetic field [5]. Planar waveguide switches have also been
`integrated with the router onto a single substrate, either as
`separate 2 X 2 switches [6] or using two waveguide grating
`routers comrected by waveguides containing phase-shifters [7].
`Mechanical switches based on macroscopic bulk optics
`and electromechanical actuators have the best insertion loss
`
`aird crosstalk performance of any switch teclmology, but
`are larger, slower, and potentially less reliable than solid-
`state switches. Micromechanical switches may achieve the
`same high performance levels of bulk optics, yet provide
`the compactness and reliability of solid state devices. In this
`paper, we demonstrate WAD using surface-normal operation
`of microoptomechanical switch arrays with free-space optical
`iritercormection to single rriode fiber inputs and outputs [8].
`We describe the design of the wavelength multiplexing optics
`and the rrricromechanical switches, present monochromatic
`aird broad spectrum irreasurerrrents on the WAD switch, then
`conclude with some comments on the ultimate potential of
`this approach.
`
`II. SWITCH DESIGN
`
`The block diagram for our WAD switch (Fig. 1) shows the
`four WDM fiber ports; IN, PASS, ADD, and DROP. The
`main input is connected through an optical circulator to a
`wavelength derriultiplexer and then to a set of individual l
`x 1 switches, each of which can reflect or transmit one wave-
`length channel. Reflected signals retrace their path through the
`wavelength multiplexer and into the circulator, which separates
`the back-reflected light into the PASS output. Transmitted
`signals are collected by a separate wavelength multiplexer and
`directed into the second port of a second optical circulator to
`the DROP output. The ADD port, connected to the first input
`of the second optical circulator, brings in the new data by
`retracing the same optical path created by dropping the input
`channels.
`I11 essence,
`the WAD creates an individual 2 x 2
`switch for each wavelength eharmel where the two allowed
`states are IN to PASS, or IN to DROP and ADD to PASS.
`Data is never routed from ADD to DROP.
`I11 this switch
`
`Manuscript received Deeeniher 29, 1998; revised February 11, I999.
`J. E. Ford and J. A. Walker are with Bell Laboratories Lucent Technologies,
`Holmdel, NJ 07733 USA.
`V. A. Aksyuk and D. J. Bishop arc with Bell Laboratories Luccnt Tech-
`nologies, Murray Hill, NJ 07974 USA.
`Publisher Item Identifier S 0733-8724(99)O3799-8.
`
`design, wavelength multiplexing of the added and dropped
`channels,
`if necessary,
`is done by an external router. Our
`implementation of this design connects two circulators [9]
`to a separate optomechanical package with the free-space
`wavelength multiplexing and rnicromechanical switch array.
`Microelectromechanical systems (MEMS) is a device tech-
`nology using lithographic fabrication techniques developed for
`silicon electronics to create miniature mechanical components.
`Elements are partially released from the substrate using a
`0733-8724/992310.00 (5) I999 IEEE
`
`4
`
`Petitioner Ciena Corp. et al.
`Exhibit 1004-1
`
`

`
`FORT) er al. V»/A'\/'FJ.l-TNGTFT AT)T)—T)ROP SVVTTCHTNG USTNG TU TTNG MT(‘ROMTRRORS
`
`circulator
`
`Single Channel Switches
`(reflectl transmit)
`
`/ Focus Lens
`9». DeMux‘ed
`Device Plane
`
`Collimatlon
`
`ADD
`
`circulalor
`
`Fig. 1. Wavelength add drop switch configuration.
`
`IN PASS
`
`"V
`
`V
`
`7
`
`Fig. 2. Tilting micromirror switch geometry.
`
`(a) Free-space wavelength rnultiplexing optics layout and
`Fig. 3.
`tomechanical package.
`
`op-
`
`selective etch to remove portions of one or more sacrificial
`films. This produces structures which are mechanically active
`yet partially constrained (attached) to the surface [10]. One of
`the earliest commercial MEMS actuators was a display using a
`two-dimensional (2-D) array of tilting rr1icron1irrors a display
`developed by Texas Instruments using [1 1]. These components
`are now driving commercial 800 x 600 pixel projection
`displays, demonstrating that high yield and reliability can be
`achieved with 480 000 element micromechanical device arrays.
`Fig. 2 shows the geometry used in our WAD switch. An
`input signal is imaged onto the tilt mirror so that in one switch
`state (PAS S) the signal is back reflected and in the other state
`(DROP) the signal is tilted to reflect the input toward the
`“add“ signal source, so that the original input and add signals
`are counter-propagating. The switch is never required to route
`light from the ADD to DROP ports. If a switch element set
`to PASS is illuminated from the “add” source, the reflected
`light beam is tilted away from both PASS and DROP ports (a
`path not shown in the diagram). To complete the WAD, each
`element in a linear array of such switches is illuminated by a
`single wavelength picked out of the WDM fiber transmission.
`
`/1. Free—Space Wavelength A/[ultiplexing
`
`The wavelength multiplexing optics used in this switch
`were originally designed for a micromechanical spectral equal-
`izer [12], where a continuous Variable—reflectivity mirror illu-
`minated by a wavelength-dispersed signal enabled dynamic
`
`power equalization. Fig. 3(a) shows the optical system layout.
`Light from an input fiber is collimated by a 25 mm focal
`length doublet lens and illuminates a 600 lines/mm diffraction
`grating blazed at a 34°‘ angle. The diffracted signal is focused
`by a 50 mm focal length triplet lens onto the device plane,
`where the broad spectrum input is distributed over a device
`array. The system uses pupil division to separate the input
`and reflected output signals. The focus lens is shifted down
`relative to the input illumination, so that the collimated input
`beam illuminates the top half of the lens, and the light reflected
`from the device array illuminates the bottom half of the lens.
`The reflected signal diffracts from a second pass off the
`grating, which recombines the spectral components. A small
`fold mirror positioned below the main optical axis picks off the
`reflected signal and directs it into an output collimator which
`focuses the signal into a second optical fiber.
`The grating diffraction efficiency has some polarization
`dependence. The insertion loss at 1543 nm ranges from about
`0.6 to 1.1 dB. If uncorrected, the double-passed grating would
`produce l dB or more of polarization dependent loss (PDL).
`However, PDL can be suppressed using a quarter wave plate
`oriented at 45° to the grating lines so that the reflected signal is
`rotated by 90° before the second pass through the grating. This
`way, any input polarization is attenuated by twice the average
`insertion loss of the grating (1.7 dB for the current grating).
`Fig. 3(b) shows the connectorized opto-mechanical pack-
`age,
`including electrical connections for device control and
`a number of tip/tilt/translation controls for optical alignment.
`
`Petitioner Ciena Corp. et al.
`Exhibit 1004-2
`
`

`
`The total fiber-to-fiber insertion loss for this package, mea-
`sured using a simple gold mirror at the device plane, is 4.6
`dB, with 0.2 dB PDL. This package is intended for laboratory
`environments. l-lowever, once the 111ecl1anics are aligned and
`locked down,
`the package can be handled and fibers can
`be attached and removed with rr1ir1i1nal (<0.5 dB) change in
`insertion loss.
`
`For operation as a switch, the device array must be designed
`so that the two mirror states either back reflect the input into
`the first eollimating lens or tilt the beam toward the second
`collimator. Reflections from the fiber to free-space transition
`must also be suppressed, so antirefleetion coated and angle-
`polished FC connectors were used on fibers fusion-spliced
`directly to the circulators.
`
`B. Micromec//Ianical Tilt-/lfirror Switch
`
`The tilt-mirror switch geometry is dictated by the beam path
`through the optical system. The basic requirement to switch
`from a transmissive to a reflective WDM path is to place a
`flat n1irror at the location of each monochromatic signal i11
`the device plane and reflect the incident cone of light back
`toward the focus lens so that it either overlaps the original
`input beam area (in the PASS state), or is shifted to the lower
`half of the focus lens and imaged to the second optical fiber
`output (in the DROP state). Given the numerical aperture of
`the single mode fiber source (about 0.2) and the magnification
`of the imaging between fibers and device plane (two times),
`the full tilt angle required to avoid overlap between the two
`switch states is at least 6°. With a 200 GHZ (1.59 nm) WDM
`signal spacing, the mirrors i11ust n1atcl1 a 57 //L111 pitch of the
`signals at the device plane.
`The devices used in this demonstration were fabricated
`
`through the multiuser MEMS processes (MUMP’s) commer-
`cial MEMS foundry operated by the Microelectronics Center
`of North Carolina (MCNC).1 MUl\/lP’s is a general purpose
`three-layer polysilicon surface micromachining process us-
`ing polysilicon as the structural material, deposited oxide
`(phosphosilicate glass, PSG) as the sacrificial material, silicon
`nitride for electrical
`isolation from the substrate, and a top
`layer of metal. The layer structure, fron1 the bottom up,
`is:
`silicon substrate, 0.6 pm of nitride, 0.5 pm of polysilicon, 2
`pm of PSG, 2 pm of polysilicon, 0.75 pm of PSG, 1.5 pm
`of polysilicon, and 0.5 pm of metal (gold, with a thin chrome
`adhesion layer). In our devices, the moving 111irror switch is
`made from the top 1.5 pm thick polysilicon and the 0.5 pm
`metal layers.
`Fig. 4(a) and (b) shows two SEM mierographs of the
`fabricated switch, with a gold-coated polysilicon tilt-mirror
`suspended 2.75 pm above the silicon substrate. The mirror area
`is 30 X 50 microns, with a 57-pin pitch between the 16 de-
`vices. The photographs reveal a 24° angle between the mirror
`tilt axis and the column of devices. This comes fro111 matching
`the skew-ray beam path through the focus lens to exactly
`overlap the input and output beams. The mirrors are supported
`
`lMCNC’s MEMS Technology Applications Center was recently sep-
`arated as an independent commercial entity called Cronos
`Integrated
`MicroSystems
`lneorporated. For more
`information,
`see
`l-lYl’ERLlNK
`http ://n1en1s.n1cnc. org/n1u111ps .ht1nl', http ://n1en1s.n1cnc. org]n1u1nps.l1tn1l.
`
`JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL.
`
`l7, NO. 5, MAY l999
`
`B825
`
`35KU
`
`xsaa
`
`rail ND14
`
`Fabricated micromeehanical tilt-mirror array: (a) top View and (b)
`Fig. 4.
`perspective view.
`
`by zig-zag torsion bars which allow them to rotate around
`a tilt axis defined by a pair of support points at either end.
`Each device has two electrical contacts leading to electrodes
`under the tilt plates, while the mirror plates are connected to
`a common ground. When the voltage on one of the contacts is
`increased (and the other contact grounded), each Inirror goes
`through a regime of continuously increasing analog deflection.
`However,
`the switch is designed to use digital positions.
`At about 20 V applied the mirror snaps down to contact
`the substrate, producing a repeatable deflection angle of 5°
`relative to the substrate surface. The edges of the mirror plate
`have landing tips to reduce the Contact area and therefore the
`probability of stiction (semipermanent bonding of the mirror
`edge with the substrate). At still higher drive voltages (about
`30 V), the devices apparently snap down into full contact with
`the substrate, becoming parallel to the substrate surface.
`The 1r1irrors were operated with an ac drive voltage to avoid
`electrostatic charging. When operated with a dc voltage for
`several seconds to minutes (depending on humidity), switching
`would lag the drive voltage by longer and longer until the
`devices eventually stopped. This apparently arises from charg-
`
`Petitioner Ciena Corp. et al.
`Exhibit 1004-3
`
`

`
`FORT) er al. V»/A'\/F,i.FNGTFi AT)T)—T)ROP SVVTTCT-UNG USTNG Tii TTNG MT(‘ROMTRRORS
`
`
`
`
`
`TotalTransmissiontoPassOutput(dB)
`
`-40
`1530
`
`1550
`1545
`1540
`1535
`Wavelength (nm) on 200 GHz grid
`
`1555
`
`r'\
`.:3
`.9-D
`OD.o_
`
`ME.
`
`D2C
`
`.9III
`.9
`EmC:v._
`
`I-
`Eo
`I-
`
`-40
`1530
`
`1550
`1545
`1540
`1535
`Wavelength (nm) on 200 GHz grid
`(b)
`
`1555
`
`(a) PASS and (b) DROP transmission of ASE input dropping eight
`Fig. 6.
`of 16 channels.
`
`two cross sections shown below. With no voltage applied, the
`mirror is very slightly tilted relative to the substrate, but the
`curvature of the cross sections indicate that the stress-induced
`
`sag has been reduced to 0.02 pm (20 nm) across the full 57
`pm aperture. This produces less than /\/20 phase variation in
`there reflected optical wavefront, a flatness comparable to a
`polished glass r11irror.
`
`III. PERFORMANCE
`
`50 nm
`
`
`
`Height(nm)
`
`20
`
`Length (pm)
`
`Fig. 5.
`
`Stlrfaee profile of individual rnicrorneclraniual tilt-mirror.
`
`ing of the exposed dielectric (silicon nitride) surface beneath
`the mirror electrode. After the contacts were grounded to
`dissipate the accumulated shielding charge the devices would
`function normally again. However, these charging effects were
`completely eliminated by replacing the dc drive voltage with
`a high frequency ac drive voltage. The electrostatic switch
`is insensitive to the sign of the drive voltage. When the
`drive oscillates around zero at a frequency much larger than
`the device’s mechanical resonant frequency, the charging is
`proportional to the averaged voltage level (which is zero),
`while the deflection is proportional to the root mean square
`voltage. We verified this experimentally, and found that a
`sinusoidal drive signal oscillating at 300 kHz with a peak-
`to-peak amplitude of 80 V produced reliable switching for
`extended operation times.
`The current MUMP‘s process is not optimized for optical
`l\/[EMS applications. The best mirrors available in the standard
`process use a 0.5-pm thick gold layer principally intended for
`wire bond pads. This gold layer has residual stress and can
`cause curvature of the released polysilicon plates, especially
`over wide thermal excursions. We used an interferometric
`
`surface profilometer to characterize our devices. The original
`devices, fabricated entirely within the MUMP’s process, had
`a 0.17 pm sag across the full 57 pm width of the tilt-mirror
`plate. This can cause the angle of the reflected beam to vary by
`as much as 0.7° depending on where the mirror is illuminated,
`an angle large enough to potentially reduce switching contrast.
`For our mirrors, we deposited 3 nm of chrome, followed
`by 50 nm of gold directly onto the top polysilicon layer of
`the MUMPS die. This composition should produce similar
`reflectivity to the original 500 nm metal layer, but result in
`significantly lower stress. Fig. 5 shows the mirror surface pro-
`file. The top is a false-color map of a single mirror, indicating
`
`Test results for the assembled and aligned WAD switch
`are shown in Figs. 6-11. The total fiber-to-fiber insertion loss
`was approximately 5 dB for the PASS output and 8 dB for
`the DROP output,
`including the circulators. The difference
`between the two states comes from the slightly larger than
`optimal mirror switching angle (we chose to align the optical
`system to minimize PASS losses). The polarization dependent
`loss ranged from 0.1 to 0.2 dB. About 1.8 dB of the loss
`comes from two passes through the circulators; 1.7 dB from
`two passes though the grating, and the remaining 1.5 dB comes
`from residual surface reflections and aberrations. Comparing
`the 5 dB loss to the 4.6 dB loss obtained using a simple gold
`
`Petitioner Ciena Corp. et al.
`Exhibit 1004-4
`
`

`
`
`
`
`
`TotalTransmissiontoPassOutput(dB)
`
`1530
`
`1550
`1545
`1540
`1535
`Wavelength (nm) on 200 GHz grid
`
`1555
`
`E5
`E..
`3D...3
`OQ.
`9D
`O..
`C
`.2ID
`.9
`E0'}Cas._
`I-
`EoI-
`
`-40
`1530
`
`1550
`1545
`1540
`1535
`Wavelength (nm) on 200 GHZ grid
`
`1555
`
`(b)
`(a) PASS and (b) DROP transmission of ASE input when channel 8
`Fig. 7.
`is dropped and replaced by light from a tunable laser.
`
`mirror, we see that the surface quality of the micro mirror
`devices is good for this type of rnicrornechanical structure.
`Broad spectrum measurements of switch transmission were
`performed using an Erbium-doped fiber ASE (amplified spon-
`taneous emission) source. Fig. 6(a) shows the PASS and
`Fig. 6(b) shows tl1e DROP output with 9 of the 16 switches set
`to drop. The original source nonuniformity has been subtracted
`to show absolute trar1s111issior1. The 3 dB roll off ir1 both
`
`PASS and DROP passbands occurs at a full width of 0.7 nm,
`as compared to the 1.59 nm pitch between channels. These
`passbands are created by the nonreflective spaces between the
`mirrors. The slightly flattened Gaussian shape of the passbands
`can be calculated by convolving the optical spot size at the
`device plane (full width at half inaximum of about 14 pm) with
`the (roughly) 30 pm wide mirror aperture. Fig. 7 illustrates
`single charmel drop-and-replace: charmel 8 is dropped from
`Fig. 7(a) the pass output to Fig. 7(b) the drop output, and
`replaced by a narrow line width signal from a tunable laser
`source connected to the add port.
`The switch is designed to operate on the ITU frequency
`grid with 200 GHz wavelength charmel spacing. We used a
`
`JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL.
`
`l7, NO. 5, MAY 1999
`
`loss
`wavelength tunable source and polarization dependent
`meter to measure the switching contrast at the center of each
`of the 16 wavelength charmels and each of the four possible
`switch states. Fig. 8 shows a bar graph of the results. The
`switching contrast for IN to PASS, or IN to DROP and ADD
`to PASS states is at
`least 32 dB. and ranges as high as
`47 dB. Signals are never routed from ADD to DROP. The
`extinction between ADD and DROP terminals was at least 36
`
`dB, in either direction. These measurements were taken when
`switching a single channel. The change in crosstalk created
`by switching the adjacent channels, and by switching various
`co111bi11ations of the other 15 channels, was measured to be
`below 1 dB.
`
`Broad spectrum measurernents of the switching contrast
`were made by using the ASE source and spectnrm analyzer
`to store the “ON” transmission,
`then divide by the “OFF”
`transmission. The result
`is plotted in Fig. 9 for switching
`from the IN port to both the PASS and DROP outputs. The
`switching contrast measured using a laser tuned to each of
`the 16 charmels (shown in Fig. 8) is also plotted, to verify this
`result. The switching contrast is sharply peaked at the center of
`the passband, where the optical spot on the tilt-mirror switch is
`entirely reflected. As the wavelength is shifted toward the edge
`of the passband, some of the input spot falls on to the mirror
`edge and is scattered into a wide angular range. A portion
`of the scattered light couples to the opposite switch output,
`resulting in crosstalk. Defining the operating passband as the
`spectral width which maintains a 30 dB or higher switching
`contrast, we see that the PASS output has a 0.57 nm average
`(0.40 run minimum) operating passband, and the DROP output
`has a 0.34 mi average (0.28 1n11 minimum) operating passband.
`The WAD switching response is shown in Fig. 10. The
`vertical lines, separated by 20 us, indicate a switching time
`which is significantly faster the millisecond response typically
`required in SONET ring recovery.
`This switch was intended as a first proof-of-principle
`demonstration, as opposed to a preproduction prototype, so
`reliability was not tested in any systematic manner. However,
`the switch operated as part of a week-long technology
`demonstration in a
`semi-enclosed (covered but not air-
`conditioned)
`temporary pavilion. The
`switch functioned
`normally, without adjustment, despite the wide swings in
`temperature and humidity characteristic of a New Jersey
`summer.
`
`Finally, the ability of the switch to transmit data on mul-
`tiple wavelengths was tested using the arrangement shown in
`Fig. 11. The test used two eight wavelength rnultifrequency
`lasers with 200 GHz pitch between wavelengths [13]. The
`two lasers were temperature tuned so that the wavelengths
`aligned exactly, creating the worst possible case for coherent
`crosstalk noise. Independent data streams from two 622 Mb/s
`(OC-3) word generators was applied to all wavelengths of
`each laser with two external electroabsorption modulators. The
`input signals were connected to the IN and ADD ports of the
`switch, and the transmitted PASS and DROP outputs were
`converted to electrical signals and fed into a GHz oscilloscope
`triggered by either of the word generators. The transmission
`eyes for the data are shown in the lower half of Fig. 11; at
`
`Petitioner Ciena Corp. et al.
`Exhibit 1004-5
`
`

`
`FORT) er al. V»/A\/FIFNGTFI AT)T)—T)ROP SVVITCI-IING USING TII TING MT(‘ROMTRRORS
`
`01§
`
`I IN to PASS D IN to DROP
`
`[3 ADD to PASS E ADD to DROP
`
`'5§
`
`‘-3Q
`
`
`
`cantrast/Extinction(:18) Q §
`
`N§
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`1.?
`
`14
`
`16
`
`Channel # (200 GHz pitch on ITU grid)
`
`Fig. 8.
`
`Switching contrast for monochromatic input on a 200 GIIZ grid.
`
`01O
`
`->-O
`
`00O
`
`MC)
`
`._P‘.".‘.’_’.’_.
`
`1535
`
`1540
`
`1545
`
`1550
`
`Wavelength (nm) on 200 GHz grid
`
`DROP output
`
`EE.
`
`-3D
`
`..-:1
`O1'0:1::6
`D.._
`ou.
`
`§:0
`Um
`E.:o.4:
`3U)
`
`
`
`SwitchingContrastforDropOutput(dB)
`
`
`
`
`
`1535
`
`1550
`1545
`1540
`Wavelength (nm) on 200 GHz grid
`(b)
`
`1555
`
`Fig. 9.
`
`Switching contrast for broad spectrum and monochromatic input.
`
`left, the data are switched from IN to PASS; at center, the data
`are switched from IN to DROP; at right, the data are switched
`from ADD to PASS. In each case, the source of crosstalk noise
`
`(:1 100mV§2
`
`soomvo M10.0usTIl‘I1J'
`
`Fig. 10.
`
`Switching dynamics, showing a 20 ps response.
`
`.
`106mV
`
`(the alternate input signals) were turned off and on without any
`perceptible effect on the transmission eye, indicating that there
`was little crosstalk present on any of the eight channels. This
`result is consistent with the 32-45 dB crosstalk suppression
`measured with a tunable laser on the individual channels.
`
`IV. DISCUSSION AND SUMMARY
`
`The prototype described in this paper is a first “proof
`of principle” of the rnicromechanical wavelength add—drop
`switch. There is considerable room for improvement in both
`wavelength resolution and switching crosstalk. The ultimate
`perfonnance limitations come from the grating dispersion
`and the diffraction-limits on optical system resolution It is
`impractical
`to increase the grating spatial frequency much
`beyond the 600 lines/mm used in the prototype (the blaze
`angle would increase from 30° to approach grazing incidence).
`It is possible, however, to double pass the grating to provide
`double the wavelength resolution (with a 1-2 dB excess loss
`
`Petitioner Ciena Corp. et al.
`Exhibit 1004-6
`
`

`
`622 Mhls
`“"7” “A” DATA eauznnon
`
`Ghz Real-Time
`oscilloscope
`SYNC “A”
`
`8 x 200 GHZ
`Mulli-Freq. Laser
`
`Data “ "
`
`M
`
`rm m
`
`,
`
`um
`
`M mu.
`
`.m
`
`.
`
`6“
`
`Fig. 11.
`(summed).
`
`Setup and transmission eyes for eight wavelength data channels
`
`penalty). The optical system was not fully optimized for this
`application. In particular,
`the use of pupil division at
`the
`focus lens to separate the two data paths meant that we used
`the lens off-axis, with broad spectrum light, and also used a
`factor of two magnification between the input fiber and device
`plane to match the lens numerical aperture to the fiber output.
`A full custom design would increase the spatial resolution
`and wavelength range of the switch. Micromecharrical devices
`tailored to a new optical system can then provide a wider and
`flatter operating passband. Our optical system designs indicate
`that 128 charmels at 100 GHZ spacing is certainly possible.
`The long-term reliability of micromechanical switches must
`be well established before they are suitable for telecommunica-
`tions applications. Failure modes may include stress-induced
`fatigue, and stiction or wear on contacting surfaces. One of
`our tilt-mirror switches has been in continuous operation for
`over a year without failure, a total of 55 million switching
`cycles. However, some devices have shown intermittent delays
`in switching, apparently due to temporary stiction of the mirror
`edge. These types of problem have been extensively studied
`by Texas Instruments [11] and Sandia National Laboratories
`[14]. Their results indicate that with proper design of device
`structures and electrical drive signals it is possible to achieve
`highly reliable operation,
`in terms of both the number of
`operating cycles and in switching after extended pauses. Most
`micromechanical device failure modes come from surface
`
`contact. Flexural actuation, where a beam or membrane is
`elastically deflected by an applied force (without friction
`or contact) is probably the most reliable approach. This is
`the basis of ultra-reliable micromechanical accelerometers
`
`used as air bag sensors since about 1993. This indicates
`that suitably designed micromechanical tilt mirrors can meet
`telecom reliability standards.
`The free-space optornechanical packaging of the switch is
`another concern. The long path lengths raise the possibility of
`failures from thermal expansion or shock. However, a number
`of fiber optic multiplexers based on planar gratings with simi-
`lar free-space optical systems have been developed as products
`for telecom [15]. These systems use transmissive imaging
`of the diffracted output into an array of output waveguides.
`
`JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 17, NO. 5, MAY 1999
`
`The fundamental design of our switch uses imaging of the
`input fiber onto reflective micromechanical modulators, then
`reimages the reflected signal toward the input fibers. If the
`device aperture is large compared to the fiber 111ode, the lateral
`alignment tolerance for the long optical path can be determined
`by the modulator aperture rather than the fiber mode. This can
`greatly simplify maintaining wavelength and loss tolerances,
`even in relatively hostile operating conditions.
`Using free-space optical interconnections between optical
`fibers and surface-nonnal device arrays presents one additional
`opportunity not present in planar optic devices. The single in-
`put a11d output fibers used in the prototype could be extended to
`one-dimensional (1-D) fiber arrays, and the micromechanical
`device extended to a 2-D array of tilt-111irror switches. This
`would make full use of the optical imaging and lithographic
`device fabrication capabilities, and distribute the packaging
`costs among mrrltiple fiber inputs. This significantly extends
`the number of channels which can be switched.
`I11 fact, the
`practical limit on the number of switched channels will be
`determined by the electrical pin-out of the micromechanical
`chip, which can number several hundred.
`In summary, we have demonstrated a reconfigurable wave-
`length add—drop switch with 20 as switching of 16 input and
`output channels on a 200 GHz pitch. The switch is based on
`free-space optical demultiplexing and a microoptomechanical
`tilt-mirror switch array. The total insertion loss of the passed
`and dropped channels is 5 and 8 dB, respectively, with 0.2
`dB polarization dependence. The switching contrast for 16
`channels on the ITU grid was greater than 30 dB for both
`passed and dropped outputs. This approach has the potential to
`scale to larger charmel counts, and to operate on multiple input
`fibers, creating a cost-effective switch for telecom applications.
`
`ACKNOWLEDGMENT
`
`The authors wish to thank R. Ellard and F. Beisser of Lucent
`
`Technologies Bell Laboratories, for custom machining and
`electronics, C. Doerr for the use of prototype rnultifrequency
`lasers for parallel switch operation, and C.-C. Chang for help
`with the data trar1srr1issior1 tests.
`
`REFERENCES
`
`[1] C. Brackett, A. Acampora, J. Sweitzer, G. Tangonan, M. Smith, W.
`Lennon, K. Wang, and R. Hobs, “A scalable multiwavclcngth multihop
`optical network: A proposal for research on all-optical networks,” J.
`Lightwave Technol., vol. 11, pp. 736—753, 1993.
`J. Lemaire, W. M.
`[2] V. Mizrahi, T. Erdogan, D.
`J. DiGiovanni. P.
`MacDonald, S. G. Kosinski, S. Cabot, ar1d J. E. Sipe, “Four channel fiber
`grating demultiplexer,” Electron. Lett, vol. 30, no. 10, pp. 780—781,
`May 1994.
`[3] C. R. Giles, T3. Barber, V. Askyuk, R. Rue], J,. Stulz, and T). J. Bishop,
`“Reconfigurable 16-channel WDM DROP module using silicon MEMS
`optical switches,” IEEE Photon. Technol. Lett.. 1998.
`[4] T. Eftimov, M. C. Earries, S. Huang, N. Duricic, T). Grohnic, T3.
`Keyworth, and J. S. Ohbi, “S-channel tunable drop device with thermal
`tuning for 100 Ghz channel spacing.” in Proc. 1998 European Conf
`Optic. Commun., Madrid, Spain, pp.
`l27—l28.
`[5] S. Jin, R. P. Espindola, H. Mavoori, T. A. Strasser, and P. J. Lemaire,
`“Magnetically programmable fiber Bragg grating,” Electron. Lani, Vol.
`34, p. 2158, 1998.
`S. Suzuki, A. Hirneno, and M. Ishii, “Integrated nrultichannel optical
`wavelength selective switches incorporating an arrayed-waveguide grat-
`ing multiplexer and thermooptic switches,” J. Lig//ztwave Tec/7nol., vol.
`16, pp. 650—655, Apr. 1998.
`
`Petitioner Ciena Corp. et al.
`Exhibit 1004-7
`
`

`
`FORD er al. Vi/A'\/ELENGTH AT)T)—T)ROP SVVITCI-IING USING Tll TING MT(‘ROMTRRORS
`
`[10]
`
`[8]
`
`[7] C. R. Doerr, “Proposed WDM cross connect using a planar arrangement
`of waveguide routers and phase shifters,” IEEE Photon. Technol. Lel1.,
`vol. 10. pp. 117 119, 1998.
`J. E. Ford, V. Aksyuk, J. A. Walker, and D. J. Bishop, “Wavelength-
`selectable add/drop with tilting micro mirrors,” in Proc. IEEE LEOS
`Armu. illeeting, Nov. 1997, postdeadline paper PD2.3.
`[9] JDS/Fitel CRZSOO double isolation circulators, with < 1 dB loss/pass
`and > 50 dB isolation and return loss.
`Sensors,
`Issue:
`Integrated
`K.D. Wise, Guest Editor,
`Special
`IEEE,
`Proc.
`(MEMS),
`Microactuators,
`and Microsystems
`and
`ht'tp://vwvw-
`Aug.
`1998. Also
`available:
`http://mems.isi.edu
`bsac.eecs.berke|ey.edu/visitors.htm|.
`P. F. Van Kessel, L. J. Hombeck, R. E. Meier, and M. R. Douglass, “A
`.\4EMS-based projection display,” 1’roc. UEEE, vol. 86, pp. 1687—1704,
`Aug. 1998.
`J. E. Ford and J. A. Walker, “Dynamic spectral power equalization
`using micro-opto-mechanics,” IEEEP/101071. Techno]. Letzh, vol. 10, pp.
`1440—1-142, Oct. 1998.
`C. R. Doerr, C. H. Joyner, L. W. Stulz, and J. C. Centanni, “Wavelength
`selectable laser with inherent wavelength and sing1e—mode stability,”
`IEEE Photon. Technol. Lett., vol. 9, pp. 1430—l432, 1997.
`D. M. Tanner, N. F. Smith, D. J. Bowman, W. P. Eaton, and K. A.
`Peterson,

This document is available on Docket Alarm but you must sign up to view it.


Or .

Accessing this document will incur an additional charge of $.

After purchase, you can access this document again without charge.

Accept $ Charge
throbber

Still Working On It

This document is taking longer than usual to download. This can happen if we need to contact the court directly to obtain the document and their servers are running slowly.

Give it another minute or two to complete, and then try the refresh button.

throbber

A few More Minutes ... Still Working

It can take up to 5 minutes for us to download a document if the court servers are running slowly.

Thank you for your continued patience.

This document could not be displayed.

We could not find this document within its docket. Please go back to the docket page and check the link. If that does not work, go back to the docket and refresh it to pull the newest information.

Your account does not support viewing this document.

You need a Paid Account to view this document. Click here to change your account type.

Your account does not support viewing this document.

Set your membership status to view this document.

With a Docket Alarm membership, you'll get a whole lot more, including:

  • Up-to-date information for this case.
  • Email alerts whenever there is an update.
  • Full text search for other cases.
  • Get email alerts whenever a new case matches your search.

Become a Member

One Moment Please

The filing “” is large (MB) and is being downloaded.

Please refresh this page in a few minutes to see if the filing has been downloaded. The filing will also be emailed to you when the download completes.

Your document is on its way!

If you do not receive the document in five minutes, contact support at support@docketalarm.com.

Sealed Document

We are unable to display this document, it may be under a court ordered seal.

If you have proper credentials to access the file, you may proceed directly to the court's system using your government issued username and password.


Access Government Site

We are redirecting you
to a mobile optimized page.





Document Unreadable or Corrupt

Refresh this Document
Go to the Docket

We are unable to display this document.

Refresh this Document
Go to the Docket