(12) United States Patent
`Garverick et al.
`
`111111111111111111111111111111111111111111111111111111111111111111111111111
`US006543286B2
`
`(LO) Patent No.:
`(45) Date of Patent:
`
`US 6,543,286 B2
`A1>r. 8, 2003
`
`(54) HlGH FREQUENCY PULSE WTOTH
`MODULATION DRIVER, PARTICUlARLY
`USEFUL FOR ELECTROSTATICALLY
`ACTUATED MEMS ARRAY
`
`(75)
`
`Inventors: Steven L. Can-erick, Soloo, Oil (US);
`Michael L Nagy, Lawrenceville, GA
`(US)
`
`(73) As.<>ignee: Movaz Netwm·ks, lnc., Norcross, GA
`(US)
`
`( • ) Notice:
`
`Subject to any disclaimer, lhe term of lbis
`patent is extended or adjusted under 35
`U.S.C. 154(b) by 0 days.
`
`(21) AppL No.: 09/884,676
`
`(22) Filed:
`
`Jun. 19, 2001
`
`(65)
`
`Prior Publication Data
`
`US 2002/0101769 Al Aug. l, 2002
`
`(60)
`
`(51)
`(52)
`(58)
`
`Related U.S. Application Data
`Provisional application No. 601264.267, filed on Jan. 26,
`2001, and provisional nppHcation No. 60(267,285, filed on
`Feb. 7, 2001.
`lnt. C l.7
`. . ... ..•.... •..•........• ~lJ> 15/00; GOlP 15/125
`U.S. Cl ................................... 73/514.18; 73/514.32
`F ield of Search ......................... 73/514.16. 514.18,
`73/514.26, 514.32, 514.36; 361/271, 277;
`310/309
`
`(56)
`
`Retcrenct.>S C ited
`
`U.S. PATENT DOCUMENTS
`
`• 7/2000 Mizuoo c1 al ........... 73/514.32
`6.082,197 A
`6,242,989 131 • 6!200 I 13arber ct .al. ............... 361(271
`6,333.584 BJ • 12/2001 Jerman et al ............... 310/309
`
`* cited by examiner
`
`Primary Examiner-Trong Phan
`(74) Allorfle)'> Agent, or Firm--Charles Guenzer
`
`(57)
`
`ABSTRACT
`
`Pulse-width modulation (PWM) drive circuitry particularly
`applicable to an array of e lectroslatic actuators formed in a
`micro electromechanical system (MEMS), such as used for
`optical switching. A control cell associated with each actua(cid:173)
`tor includes a regi'>ter selectively stored with a desired pulse
`width. A clocked counter distributes its outputs to ail control
`cells. When tbe counter matches lhe regisler, a polarity
`signal corresponding to a drive clock is latched aod controls
`lhe voltage applied to tb.e electrostatic cell. ln a bipolar
`drive, one actuator electrode is driven by a drive clock; the
`other, by the latcb. The MEMS clement may be a tillable
`plate supported in its middle by a torsion beam. Comple(cid:173)
`mentary binary signal'> may drive two capacitors formed
`across the axis of the beam. The register and comparison
`logic for each cell may be formed by a content addressable
`memory.
`
`29 Claims, 14 Drawing Sheets
`
`11 2
`
`132
`
`130
`
`134
`
`110
`
`117
`116 ~ 118
`~
`(r
`128
`
`114
`
`0001
`
`Capella 2016
`Ciena/Coriant/Fujitsu v. Capella
`IPR2015-00816
`
`
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 1 0f 14
`
`US 6,543,286 B2
`
`11
`
`124
`122
`
`
`112
`
`
`
`0002
`0002
`
`

`

`US. Patent
`
`Apr. 8,2003
`
`Sheet 2 0f 14
`
`US 6,543,286 B2
`
`Ememwm
`
`defifioo
`
`m.03
`
`mm:
`
`omr
`
`IomoHE
`
` mammmoomm
`
`0003
`0003
`
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 3 of 14
`
`US 6,543,286 B2
`
`
`
`0004
`0004
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 4 0f 14
`
`US 6,543,286 B2
`
`LEVEL
`
`SHIFTER
`
`2-..}
`
`4%
`1315-:
`Elk.
`.43
`m.
`
`0005
`0005
`
`192
`
`19g
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 5 0f 14
`
`US 6,543,286 32
`
`>o§<
`
`|90992550099..-@0
`
`
`co
`:1:
`
`CD
`
`D:
`I
`I‘-
`D:
`
`4:
`b”
`
`CD
`
`Di.
`‘
`'41
`>
`
`CD
`
`Di
`I
`[2D
`>
`
`FIG.6
`
`0006
`0006
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 6 0f 14
`
`US 6,543,286 B2
`
`
`
`mofimmmbzmmommmuommomog
`
`
`
`_HEW;ZEDQOUBoaELDQ_
`
`
`
`
`
`mamflzmman?Nmafiamohbfim
`
`
`
`mm;Emlm
`
`v50
`
`mMHZDOo
`
`lEms/Hm
`
`mm¢mzou
`
`mmmoommmmmmmmafl
`
`$0.53Eml
`
`m
`
`wmm
`
`Emim
`
`meaqumfioo
`
`&.UE
`
`$0.53Emlfi
`
`
`
`0007
`0007
`
`
`
`

`

`US. Patent
`
`mm8.,wA
`
`Ha...mhS
`
`US 6,543,286 B2
`
`
`
`49PAH.IIHUHH1moMwmflom
`
`mdIona
`
`am?
`
`mdlwfi
`
`Ofim
`
`3
`
`.1
`
`
`
`mmmmoummgoo
`
`mmmmoomm.400Lil!
`
`
`
`>533.mmEmQ>m
`
`0008
`0008
`
`
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 8 of 14
`
`US 6,543,286 B2
`
`
`
`
`
`0009
`0009
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 9 0f 14
`
`US 6,543,286 B2
`
`.65
`
`56
`
`mEmHUmm
`
`mmmmmm
`
`RS
`
`mmm835
`
`0mm£415
`
`0mmcom
`
`Aa,
`
`gE
`
`comgig
`
`0010
`0010
`
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 10 of 14
`
`US 6,543,286 B2
`
`0011
`0011
`
`

`

`US. Patent
`
`Apns,mm3
`
`Sheet 11 of 14
`
`US 6,543,286 B2
`
`LEO
`
`m4
`
`*mzo
`
`.moh<2
`
`0012
`0012
`
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 12 0f 14
`
`US 6,543,286 B2
`
`310
`
`K Vcc
`
`Q
`
`325
`LD
`
`o
`
`a
`
`320
`
`I
`
`
`
`5
`322 ‘ ~
`
`~—|
`
`316
`
`Q‘
`
`328
`
`:
`
`R3;
`
`318
`
`324
`
`Rep
`
`312
`
`\ Vcc
`
`MATCH:
`
`I—I—I
`
`LD
`
`*
`
`CMP
`
`I—I-
`_-
`" —I_-
`
`T
`
`R D
`
`R“ D*
`
`0013
`0013
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 13 of 14
`
`US 6,543,286 B2
`
`CE
`
`RE
`
`LD
`
`H
`D
`COMCLK
`(MCLK
`)
`
`II
`__.____g
`I_.__W_____+_s
`1
`’
`E
`
`
`~—
`
`———8
`
`360
`
`15:2
`
`n
`
`
`RESET
`- REGISTER
`
`I
`
`
`
`~
`
`363
`
`364
`
`n
`
`EDGE
`
`START
`
`I
`
`DETECTOR
`
`COUNTER
`
`|
`
`366
`
`368
`
`VA
`
`VB
`
`FIG. 16
`
`0014
`0014
`
`

`

`US. Patent
`
`Apr. 8, 2003
`
`Sheet 14 0f 14
`
`US 6,543,286 B2
`
`CE
`
`RE
`
`i
`I—8
`,
`‘8
`
`I
`
`——'
`1’ “+1
`
`RESET
`
`352
`
`REGISTER
`
`360
`
`} D
`
`I
`I
`I I
`
`|
`
`5
`
`I
`
`363
`
`DIRECTION
`
`n 364
`
`DEEEgEOR
`
`COUNTER
`
`I
`
`VA 4
`
`1
`
`0
`
`1
`
`366
`
`FIG. 17
`
`B
`
`0015
`0015
`
`+1
`
`
`LD
`3 D
`MCLK
`
`COMCLK
`
`n/
`
`

`

`US 6,543,286 B2
`
`1
`HIGH FREQUENCY PULSE WIDTH
`MODUlATION DRIVER, PART!CULARLY
`USEFUL FOR ELECTROSTATICALLY
`ACTUATED MEMS ARRAY
`
`RELATED APPLICATIONS
`
`This application claims benefit of U.S. Provisional Appli(cid:173)
`cations No. 60/ 264;267, filed Jan. 26,2001, and No. 60/267,
`285, filed Feb. 7, 2001.
`
`FfELD OF THE INVEN110N
`
`'fbe invention relates to electrical driving circuits. In
`particular, the invention relates to electrical driving circuits
`configured to drive an array of electrostatic actuators, Cor
`example, micro electromechanical systems used for optical
`switches.
`
`BACKGROUND ART
`
`The technology of micro electromechanical systems
`(MEMS) originates from technology developed over
`decades for Lbe fabrication of silicon integrated circuits.
`MEMS permits the fabrication of large arrays of microac(cid:173)
`tuators Lbat can serve as mirrors, valves, pumps, etc. for a
`variety of applications. Although the invention is not so
`limited, an important application is an array of liltable
`mirrors integrated in a single substrate and used for switch(cid:173)
`ing of a large number of optical beams. Each mirror is pact
`of a separately controlled actuator. These actuators arc
`typically electrostatic in nature and require actuation volt(cid:173)
`ages near lOOV to operate.
`An example of one cell of an electrostatically controlled
`MEMS array is illustrated in plan view in FIG. 1 and in
`cross-sectional view in FIG. 2 . The cell is o ne of many such
`cells arranged typically in a two-dimensional array in a
`bonded s tructure including multiple levels of silicon and
`oxide layers. The cell includes a gimbal structure of an outer
`frame 110 twislably supported in a support s tructure 112 of
`the MEMS array through a first pair of torsion beams 114
`extending along and twisLing about a minor axis. The cell
`further includes a mirror plate 116 having a reflective
`surface 117 twistabl y supported on the outer frame 110
`through a second pair of torsion beams ll8 arranged along
`a major axis perpendicular to the minor axis and twisting
`thereabout. In Lbe favored MEMS fabrication technique, the
`illustrated structure is integrally formed in an epita xial (epi)
`layer of crystalline silicon. The process bas been disclosed
`in US Provisional Application, Serial No. 60/260,749, filed
`Jan. 10, 2001, incorporated herein by reference in its
`entirety.
`The structure is controllably tilted in two independent
`dimensions by a pair of electrodes 120 under the mirror plate
`116 and another pair of electrodes 122 under the frame 110.
`The electrodes 120, 122 are symmetrically disposed as pairs 55
`across the axes of their respective torsion beams 18, 114. A
`pair of voltage signals V A• V 11 are applied to the two mirror
`electrodes 120, and anot her pair of voltage signals are
`applied to the frame electrodes 122 while a common node
`voltage signal V c is applied to both the mirror plate 116 and
`the frame UO. Tbc driving circuitry for these and similar
`voltage signals is the central focus of this invention.
`Horizontally extending air gaps 124, 126 arc formed
`respective ly between Lbe frame 110 and Lbe support stn1cture
`112 and between the mirror plate 116 and Lhe frame 110 and
`overlie a cavity or vertical gap 128 [ormed beneath the [rame
`110 and mirror plate 116 so that the two parts can rotate. The
`
`0016
`
`tO
`
`15
`
`2
`support structure 112, the frame 110, and the mirror plate
`116 are driven by the common node voltage V c• and the
`frame 110 and mirror p late 116 form Q OC sci of plates for
`variable gap capacitors. 1\Jthoug h FEG. 2 illustrates the
`5 common node voltage V c being conlllectcd to the mirror
`plate 116, in practice the electrical co ntact is made in the
`support strucwre 112 and electrical leads are formed on top
`of the torsion beams 114, 118 to apply the common node
`voltage signal to botb the frame 110 and the mirror plate U6,
`which act as top electrodes. Tbe electrodes 118, 120 are
`formed at the bottom of the cavity 128 so the cavity forms
`the gap of the four capacitors, two between the boltom
`e lectrodes 118 aod tbc frame lJO, a.od two bctwcco tbc
`bollom electrode 120 and mirror plate 116.
`The torsion beams 114, 118 act as twist springs attempting
`to reStOre the outer frame ll0 and tht: mirror plate 116 tO
`neutral tilt positions. Any voltage applied across opposed
`electrodes exerts a positive force acting to overcome the
`torsion beams 114, 118 and to close the variable gap between
`20 the electrodes. The force is approximately linearly propor(cid:173)
`tional to the magnitude of the applied voltage, but oon(cid:173)
`Iincarities exist for large deflections. 1 fan AC drive signal is
`applied well above the resonant frequency of the mechanical
`elements, the force is approx.imately linearly proportional to
`25 the root mean square (RMS) value of the AC signal. lo
`practice, the precise voltages needed to acbieve a particular
`tilt are experimentall y determined.
`Because the capacitors in the illustra.ted conLiguratioo arc
`paired across the respective torsion beams 114, 118, the
`30 amount o( tilt is determined by the diiierencc of the RMS
`voltages applied to the two capacitors of the pair. 1be tilt can
`be controJicd io either direction depending upon the sign of
`the difference between the two RMS vollages.
`As shown in fiG. 2, the device has a large lower subs trate
`35 region 130 and a thin upper MEMS region 132, separated by
`a Lbio insulating oxide layer 134 but bonded together in a
`unitary s tructure. The tilting actuators arc etched into the
`upper region, each actuator suspended over the cavity 128
`by several telhers. The electrodes are patterened onto the
`40 substrate, which can be an application specific integrated
`circu it (ASIC), a ceramic plate, a printed wiring board, or
`some other substrate with conductors patterned on its sur(cid:173)
`face. The actuators io
`the upper region form a single
`electrical oodc called the "common node". Each actuator is
`45 suspended above four electrodes, each electrode being iso(cid:173)
`lated from every other electrode. l b cause the actuator to Lilt
`in a specific direction, an electrostatic force is applied
`between the actualor and one or more of its e lectrodes by
`imposing a poteotial difference between the common node
`so and the desired electrode. Each actuator has two pairs of
`complementary electrodes, one causing tilt along the major
`axis and the other causing tilt along tbc minor axis. Fabri(cid:173)
`cation details are supplied in Lbe aforementioned Provisional
`Application No. 60/260,749.
`One drawback of electrostaLic actuation used for this
`micro mirror is a phenomenon known as '·snapdown" .
`Because electrostatic force is inversely proportional to the
`distance between the electrodes, there comes an angle at
`which the auractive force increases very rapidly with greater
`60 electrode proximjty. Beyond this angle, a small decrease in
`distance leads to an enormous increase in force, and the
`electronic control loop becomes unstable, causing the elec(cid:173)
`trodes to soap together. With such an actuator io whicb the
`electrodes comprise a fiat plate suspended over a cavity by
`65 small tethers, a rule of thumb states that the plate will begin
`to snap down at a deflection corresponding to approximately
`four oinths the depth of the cavity. Hence, in order to achieve
`
`

`

`US 6,543,286 B2
`
`tO
`
`3
`a deflection of eat the end of the cantilever. the cavity must
`be approximately 2.25 e deep. Electrostatic MEMS mirror
`arrays have been used as video display drivers, but they
`operated at two voltage levels, zero and full soap-down. ln
`contrast, tbe mirrors described above must be nearly con(cid:173)
`tinuously tillable over a sigoif:lcaot angular range.
`Optica l constraints determine the dellection distance
`requiremeot for the electrostatic micromirror. The RMS
`voltage level required for a given amouot of deflection
`results from a combination of actuator size, tether spring
`constaot, and cavity depth. The cavity depth required to
`avoid snapdown generally dictates the use of relatively high
`voltages, typical ly in excess of 40V, the upper limit for many
`standard IC processes. The generation of sucb voltages
`requires an electronic system composed of high-voltage JS
`(!IV) semiconductor components, either olf-tbe-sbelf or
`cw;tomized, which are fabricated by specialized H V
`processes, such as the IIVCMOS process available from
`Supertex, lac.
`1l1e application for which the invention was developed 20
`requires a 12x40 array of micromirrors, and the mirrors mm;t
`be independently til table in both directions along two axes.
`Each tilt axis requires its own actuator pair so the driver
`array is 24x40. The size of the array is dictated by the
`switching of 40 wavelength-separated channels in a wave- 25
`length division multiplexing (WDM) optical network being
`switched between 6 input fibers and 6 output fibers with a
`folding mirror optically coupling paired input and output
`mirrors. Switching is accomplished by selective tilling about
`a major axis; and, power tuning by selective tilting about a 30
`minor axis. The MEMS structure accomplishes
`bi-directional tilt using two electrodes that are symmetri(cid:173)
`cally placed about the central tether of each axis. Hence,
`there are four electrodes per microactuator, for a total of
`384() electrodes thai must be independently cootrolled. 35
`Optical techniques such as " interleaving" may be used to
`split the array into two 12x40 chips, but even with tbis
`amelioration, each MEMS chip will have L920 high-voltage
`inputs and outputs (1/0s). While l/0 counts of several
`thousand are commonplace in certain low-voltage digital 40
`technologies such as memories. But, when the inputs here
`are bigb-voltage analog signals, as in the described mirror
`switching array, bigh l/0 counts present a significant pack(cid:173)
`aging problem.
`Conventional metbocls for silicon chip l/0 include wire
`bonding and die-to-substrate attachment known as " flip(cid:173)
`cbip". lt is generally accepted thai wire bonding becomes
`impractical at about 800 VO 's, due to the large chip perim(cid:173)
`eter required to contain the bond pads. Integrated circuits 50
`with higher 1/0 counts arc typically attached to a substrate
`with solder bumping, and signals are routed to discrete
`drivers that are !lip-chip bonded to the same substrate, but
`this solution becomes difficull in the intended application
`due to the very large number of high-voltage (HV) signals
`and the size of conventional IIV circuitry.
`MEMS actuators often exhibit a charging e[ect that
`builds up over Lime and, when Lhe driving voltage is DC,
`eventually disables operation. Charging therefore dictates
`thai the driving voltage bas alternating polarity with zero DC 60
`bias. Also, MEMS microactuators may display significant
`operational variation from actuator to actuator or the opera(cid:173)
`tion may depend upon environmental conditions.
`
`45
`
`4
`that used for a micro electromechanical system (MEMS). In
`an e lectrostatic actuator, a variable gap capacitor is formed
`between electrodes fixed oo two mechanical elements, one
`of which is movable with respect to the other against a
`s restoring force, sucb as a spring. The relative position of the
`two elements is controlled by pulse width modulation
`(PWM) in which the pulse width of a repetitive drive signal
`determim:s the RMS value of the applied voltage. The
`frequency of the drive signal is preferably at least teo times
`the mechanical resonant frequency of the mechanical ele-
`meots.
`Preferably, for electrostatic actuators, the drive s ignal is a
`bipolar signal having a zero DC component. Such a bipolar
`drive sigual is achieved using digital circuitry by applying a
`first high-voltage signal synchronized to the drive frequency
`to one electrode and a second high-voltage signal to the
`other electrode at the same drive frequeocy but delayed from
`the first high-voltage signal.
`The MEMS element may be a tillable plate symmetrically
`formed about the axis of a torsion beam supporting it with
`two variable gap capacitors formed on opposing sides of the
`beam axis. Advantageously, a first binary bigh-voltage sig(cid:173)
`nal is applied to a first electrode spanning the beam axis, a
`delayed binary second high-voltage signal is applied to a
`second electrode opposed to one side of the first electrode,
`and a binary third high-vol!age signal complementary to the
`second high-voltage signal is applied to third electrode
`opposed to the other side of the first electrode.
`Alternatively, the delayed high-voltage signal is applied
`to a selected one of the paired capacitors while a high(cid:173)
`voltage clock signal is applied to the unselected one.
`The inveotion is advantageously applied to an array of
`MEMS actuators formed in top level of a bonded multi-level
`silicon structure. A control cell is associated with each
`actuator. Preferably, a high-voltage section, for example,
`having a power bus of 40 VDC or gr~ater, of each control
`cell is positioned below tbe actuator it drives, and an array
`of such high-voltage sections are arranged on a same pitch
`as the actuators. The PWM control may be effected using a
`low-voltage logic section, for example. having a power bus
`of 110 more than 5 VDC. The high-voltage and low-voltage
`sections are distinguished by a ratio of power supply volt(cid:173)
`ages of at least 8. Tbe low-voltage section supplies a
`low-voltage version of the delayed drive signal, which the
`associated bigb-voltage section converts to a high-voltage
`drive sigual. The low-voltage sections may be disposed
`below its corresponding actuator or may be disposed on a
`side of an array of actuators and corresponding high-vollage
`sections.
`The control cell may be implementetl as a counter driven
`by a master clock at a multiple of atleast8 of the drive clock
`to wbich the bipolar drive signal is locked and supplying iLs
`multi-bit output to many control cells. Each control cell
`includes a register for selectively storing a value correspond(cid:173)
`ing to the desired delay. A multi-bit comparator compares
`the counter value with the register. When tbe two agree, a
`bipolar polarity signal oscillating at tbe drive frequency is
`latched until the corresponding time of !be nex1 half cycle.
`The latched signal is delayed from tbe drive frequency by
`the delay stored in tbe register. Data is stored in a selected
`one of the control cells by a multi.plexiog architecture
`including address decoders and a shared multi-bit data bus.
`Such logic is advantageously implemented in a content
`65 addressable memory (CAM) baving multiple CAM bits,
`each of which both stores a bit and compares ii to tl1ecoun!er
`bit. Wbcn the two agree, its output is combined wiib that
`
`55
`
`SUMMARY OF THE INVENTION
`The invention includes the method and circuitry for
`driving an electrostatic or other type of actuator, particularly
`
`0017
`
`

`

`US 6,543,286 B2
`
`5
`register's other CAM bit outputs in an AND circuit. This
`may be e[ected by precharging a single line that is dis(cid:173)
`charged by any of the CAM bits connected to it. That single
`line enables a latch to latch the current value of the drive
`clock.
`
`BRIEF DESCRIPTJON OF THE DRAWINGS
`
`5
`
`6
`small electrically contro lled mechanical syste ms. The
`rnicromirror array 140 may be formed of a large number of
`cells illustrated in FIGS. 1 and 2 arranged on a regular pitch
`in two dimensions. The mirrors may have s izes of about400
`,um arranged on perpendicular pitches of about 650 )lm and
`1000 11m, allowing the entire 12x40 mirror array to be
`fabricated on a chip having dimensions of about12 mmx26
`mm. Each of tbe mirrors 142 includes two microactuators,
`each driven by a respective driver 144 in a driver integrated
`10 circuit 146. Tbe driver 144 applies a high-voltage (IN)
`signal to electrodes forming variable gap capacitors with the
`tillable mirror and effecting an electrostatic actuator (ESA).
`Tbe figure indicates only a single drive for each mirror 142.
`However, the drive circuitry is easily extended to a two-axis
`15 tillable mirror by including separate and independent drivers
`144 for the two axes.
`Advamageously, the driver in tegrated circuit 146 is fab(cid:173)
`ricated on an application specific integrated circuit (ASIC)
`fabricated by a process which, if desired, can accommodate
`both the HY drivers 144 and lower-vollage control circuitry
`for the HY drivers. The driver integrated circuit can be
`interfaced directly to bouom of the micro mirror array 140 by
`chip-on-chip solder bumping, frit bonding, or similar means
`leaving the top surface including the mirrors 142 exposed.
`At least the high-vollage drivers 144 are preferably posi(cid:173)
`tioned below tbe corresponding mirror microactuator 142
`and are directly and vertically connected to tbe correspond(cid:173)
`ing electrodes. As a resu.lt, the high-voltage drivers l 44need
`to be small enough to be arranged on the same pitch as the
`30 mirrors 142. Typically, the mirror chip 140 is smaJJer than
`the driver chip 146 wilh bonding pads and perhaps the
`lower-voltage circuitry in the driver chip 146 being exposed
`to the side of the mirror chip 146. Alternatively, the low(cid:173)
`voltage circuitry is formed in one or more chips connected
`35 by bonding wires or solder bumps to the high-voltage ASIC
`in a multi-chip module (MCM) configuration. 'Ibe single
`electrical connection to the common node forming the top
`electrodes of tbe electrostatic microactuators can be accom(cid:173)
`plished by eutectic bondi ng, polymer bonding, or wirebond-
`40 ing from the top side of the mirror chip 140.
`Low-voltage control circuitry is readily available for DC
`power supplies of 5 VDC although lower voltages are
`becoming prevalent in digital integrated circuits. On the
`other band, the electrostatic actuation of MEMS devices
`45 usable as optical switches require much higher voltages,
`generally a minimum of 20 VDC, but at least 40 VDC is
`preferred, and at least 100 VDC eases the overall design.
`According ly, when high-voltage circuitry is distinguished
`from low-voltage circu itry, the DC power buses of tbe two
`50 circuits supply voltages ditier by at least a factor of four and
`preferably by a factor of eight.
`The control system is a completely digital system based
`on a microprocessor 150 operating at a clock rate, approxi(cid:173)
`mately 25.6 Ml-lz in the embodiment to be described later in
`55 detail, set by an oscillator 152. Other types of microcon(cid:173)
`trollers may also be utilized. Preferably, lhe microprocessor
`150 and oscillator 152 together with the assembled ESA
`array 140 and driver integrated c ircuit 144 and perhaps a
`separate low-voltage control integrated circuit are mounted
`60 on a standard substrate carrier, typically formed of plastic or
`ceramic, witb a small number of wire bonds connecting the
`microprocessor 150 and the periphery o[ the driver chip 146.
`The microprocessor 150 receives commands from the sys(cid:173)
`tem controlling the optical switch and through a multiplex-
`us ing sequence controls a large number of actuator cells
`lhrough a small number of cootrol lines. These commands
`include most importantly the desired positions of the mirrors
`
`FIG. I is a plan view ol' a cell of an array of micro
`electromechanical actuators inclucling a mirror tiltable in
`two perpendicular directions.
`FTG. 2 is a cross-sectional view of the cell of FIG. L taken
`along view line 2- 2.
`FIG. 3 is a schematic diagram of a control system
`architecture for an array of MEMS mirrors.
`FIG. 4 is a timing diagram for pulse width modulated
`vollagc driving signals combined with a schematic of lhe
`electrostatic actuator they are driving.
`FIG. 5 is a circuit diagram for the high-voltage drive
`circuit which translates logic level pulse width modulation 20
`(PWM) signals to high-voltage signals.
`FIG. 6 is a timing diagram illustrating the generation of
`tbe PWM signal.
`FIG. 7 is a block diagram of a logic driver circuit, which 25
`may be implemented
`in content addressable memory
`(CAM).
`FIG. 8 is a floor plan o( a mixed high-voltage and
`low-voltage integrated circuit driving 480 two-axis mirrors.
`FIG. 9 is a floor plan of one logic column of the integrated
`circu it of FIG. 8.
`FIG. 10 is a schematic diagram of the address decoders
`cootrolling the CAM register.
`FIG. 11 is a circuit diagram of circuitry used to groom
`control signaLs used to control the CAM register.
`FIG. 12 is a liming diagram of signals in tbe grooming
`circuit o f FIG. U .
`FIG. 13 is a circuit diagram of lhe CAM register.
`RGS. 14 and 15 are schematic diagrams respectively of
`the RAM bii cell and the CAM bit cell in the CAM register
`of FIG. 13.
`FIG. 16 is a block diagram illustrating alternative cir(cid:173)
`cuitry for implementing the logic drive circuit of FIG. 7 .
`FIG. 17 is a block diagram of a modification of tbc
`circuitry of FIG. 16 usable when net force is applied to only
`one of two electrode pairs.
`
`DETAILED DESCRIPTION OF 11-IE
`PREFERRED EMBODIMENTS
`
`The drive circuitry of the invention is advantageously
`combined with otbcr clements to form a micromirror switcb
`array and control system illustrated schematically in FIG. 3.
`Although the invention most directly concerns the driver
`control circuitry and in particular its use of pulse width
`modulation, the invention is not limited to driving micro(cid:173)
`mirrors. llowever, the mirror array as implemented in an
`optical switching system will be described first to provide
`specificity to the description of the control system. Further,
`some of the features of the micromirror array and its
`implementation in a bonded mu lti-level structure are advan(cid:173)
`tageously combined with features of the comrol system.
`A micromirror array 140 of FIG. 3 includes a number of
`tillable mirrors 142 fabrica ted as a micro electromechanical
`systems (MEMS) by techniques originally developed in lhe
`semiconductor industry but now further developed for very
`
`0018
`
`

`

`US 6,543,286 B2
`
`7
`142, which effect switching between optical ports of tl1e
`system. For the 12x40 mirror array discussed above, each
`mirror needs to be positionable in the major direction at, for
`example, six gross tilt ang les as well as at finer angular
`resolution corresponding to tuning around those positions
`and in the minor direction in a fine resolution providing
`power LUning. As a result, two actuators are required for each
`mirror 142. It is understood tha t !be invention can be applied
`to a different number of MEMS elements and is not limited
`to two-axis tilling.
`The overall system also includes equalization of energies
`between the wavelength channels, a.<; d isclosed in US Pro(cid:173)
`visional Application No. 60/234,683, filed Sep. 22, 2000.
`lbe mirrors 142 redirect incident beams 160, only two of
`which arc illustrated, into reflected beams 162 at angles
`determined by !be mirror positions. The tilt of each mirror
`142, as controlled by the drive voltages, is selected to
`redirect the incident beam 160 originating from a fixed angle
`to the reflected beam 162 at a selected output angle. The
`angle can be selected to correspond to different <Jut put ports
`or to tune the optical coupling to a particular output port,
`taking into account the uoillustrated optics included withi n
`the system.
`A small fraction of the power in each reflected beam 162
`is detected in a respective optical detector 164. The larger 25
`fraction is coupled to unillustrated output ports of the
`switching system. A multiplexer 166 under control of tl1e
`microprocessor 150 selects one of the detector outputs and
`an analog-to-digital (A/ D) converter 168 digitizes !be
`detected optical intensity aud suppucs it to the micropro(cid:173)
`cessor 150. Thereby, the microprocessor 150 monitors !be
`optical intensity of the reflected beams. Thereby, !be micro(cid:173)
`processor 150 can instruct the tuning of !be mirrors to either
`maximize !be coupling or, more preferably, to equalize !be
`intensity between multiple beams destined for the same
`WDM outpu t fiber. Such equalization is important when the
`signals originate from differen t sources of uncertain power.
`In ODe preferred implementation, separate input and out(cid:173)
`put mirrors are coupled through an intermediate folding
`mirror. Each time aD input optical sigDal is routed to a new
`output liber, the microprocessor 150 reads the optimum
`position scllings [or both axes or both the input and output
`mirrors associated with tltis routing combination and sets !be
`mirror positions accordingly. Optimum mirror scuings may
`have changed since this routing combination was last used
`due to changes in environmental cond it ions, such as
`vibration, thermal expansion, fiber stress, etc. so the micro(cid:173)
`processor 150 will !ben need to hunt for a new maximum in
`measured power by maldng small adjustments to the mirror
`settings, using, for example, a gradient descent algorithm,
`until the positions of peak intensity arc determined.
`Once the transmission coupling is optimized, the power of
`the outpu t signals may be intentionally degraded to obtain
`equalization or other adjustment of power with the other
`output signals. Equalization may be achieved by reducing
`the angle on the minor axis until equalizatioD is obtained
`following Newson's method in which the new minor-axis
`angle is estimated by computing the change in power
`necessary to obtain equalization divided by tbe angular
`derivative of power. This method is repeated until cquiliza(cid:173)
`tion is obtained. The derivative of the output power with
`minor-axis position must be Learned by !be microprocessor.
`Each time a mirro r is adjusted to a new position, the angular
`derivative is computed using ilie measured change in power
`divided by !be commanded change in angle.
`Equalization is an ongoing process since environment
`conditions, including laser power, may change. The micro-
`
`8
`processor will routinely monitor aU output power levels and,
`using i.ts most recent knowledge o( tbe power derivative,
`will adjust the minor-axis settings to maintain equalization.
`Likewise, it will make small adjustments to the major-axis
`5 settings to maintain optimum alignment despite changes in
`environmental conditions.
`The microprocessor 150 controls a time multiplexed
`storage of position control in the actuator array ASIC 144.
`In the pulse width modulation control. the position control
`tO is dictated by a multi-bit duty cycle. The position data and
`a row and column address for which the data is to be applied
`arc delivered to the actuator array 144 by tbc microprocessor
`150. A write enable signal WE causes the addressed cell of
`the actuator array 144 to store !be position data. Thereby, all
`15 cells are sequentially stored with position data, and tbe
`position data of any one cell can be updated as desired. A
`compare enable signal CE from microprocessor causes all of
`the cells in the actuator array to be simultaneously PWM
`controlled according to position data stored in the respective
`20 cell with a tin1ing referenced loa clock signal CLK supplied
`from the microprocessor 150 as derived from an oscillator
`152. 1o the described example, the electrostatic microactua(cid:173)
`tors arc subjected to a bipolar signal osciJlating at 50 kHz
`and the CLK signal is 512 tinles greater, tllat is, 25.6 Mllz.
`This design is faciutated by drive circuitry having several
`characteristics. It shou ld output RMS voltages as large as
`200Y with zero DC bias to obtain adequate cl.cctrostatic
`mirror deflection while avoiding chargjng effects. Any sub(cid:173)
`stantial deviation from zero DC