OPTICAL REVIEW Vol. 8, No. 4 (2001) 294-300

Design Criterion for Micro-Machine Scanner Optical SWitch Based on Beam Steering Angle Error Analysis Hideaki OKAYAMA* R&D Department, Optical Components Division, Oki Electric Industry, Co., Ltd., 550-1 Higashiasakawa, Hachioji,

Tokyo, 193-8550 Japan (Received January 30, 2001 ; Accepted Aprn 26, 2001)

The micro-machine scanner optical switch is gaining attention for its ability to switch over 100 ports. A beam steering error for devices using tilted mirrors or a moving lens is analyzed and schemes to overcome the limit are reported. The channel number is limited by the achievable beam steering error. A device with several hundred ports is possible with the beam steering error under I %. Optimized focal length or fiber mode divergence angle is beneficial for increasing the channel number. To minimize the thermal expansion effect, Iarge focal length is required indicating need for a compensation scheme to attain larger channel number. A new type of scanning optics is proposed to overcome the limit set by the beam steering error, and

a new switch network described can further expand the channel number. Key words: optical switch, optical cross-connect, scanner, angle error, micro-machine, micro electro-mechanical sys-

tem, micro-mirror, moving-1ens

1 . Introduction

2. Bipolar Drive Configuration In the conventional structure, without any lens system other

The MEMS (micro electro-mechanical system) scanner optical switchl-8) is gaining attention for its scalability over 100

ports realizing high-density optical cross-connect. The basic structure of the scanner optical switch is shown in Fig. I . The input port scanner directs the light to the desired output port and the output scanner selects the light from the desired input port.

than for the optical scanner and using an equivalent optical scanner for all ports, only single polarity (side) of the scanning angle is used near the margin of the optical scanner array or substrate. Several methods can be used to enable bipolar drive to overcome the inconvenience. In the first method, the light beam at zero steering signals is directed to the center

There are two types of optical scanners as shown in Figs. 2 and 3.1-7) The first typel~) (Fig. 2) uses a rotating or tilting

output scanner 4) Either positional shifting of the fiber or lens

micro-mirror and the second a moving lens5~7) (Fig. 3). The

tical fibers are shifted against the center of collirnator lens in

micro-mirrors can be integrated on a substrate or packaged in-

accordance with the scanner position in the array or substrate,

can be used for this purpose (Fig. 4). The input and output op-

dividually (Fig. 2). There can also be two configurations for a

so that at zero scanning angle the light beams from input scan-

moving-1ens device similar to the micro-mirror device. How-

ners are directed to the center output scanner (Fig. 4(a)). This

ever, the structure difference is not as large as in the micro-

arrangement generates the same one to one correspondence

mirror device which requires refiected light path configuration. The scalability is determined by several factors. Not reported so far but most importantly, the beam steering angle error tolerance sets the upper boundary of the attainable

between output port and scanning angle for all optical scanners. The focusing effect of a lens can also be used to direct the light toward center output scanner as in Fig. 4(b). In the following sections, all the entire performance is analyzed in an architecture based on this improved configuration.

channel number N, especially when we do not want to rely on a sophisticated control method.9) The analysis of the beam steering error and its effect on switching performance is reported here for the first time to our knowledge. In Sect. 2 the bipolar drive configuration is described which

is assumed throughout the paper. The beam steering error analysis is reported in Sect. 3. In Sects. 4 and 5, subsystem architecturero) and an optical system design implementing a

As another scheme, which we propose, arrangement placing a convex lens in the middle of input and output optical scanners can also enable the usage of both polarities of the scanning angle. A parallel collimated light beam entering the input-side concave lens is diverged and launched into a convex lens. The convex lens focuses the light onto the outputside concave lens. The light is converted to a parallel beam by

larger scale device to overcome the limit set by beam steering

the output-side convex lens. The convex lens refracts the light

error are discussed.

beams with different incident angles into light beams propagated to different positions. The positions are set according to the incident angles almost regardless of the input positions.

This paper was originally presented at the 2nd International Conference

on Optical Design and Fabrication, ODF2000 which was held on November 15-17, 2000 at the International Conference Center, Tokyo, Waseda University, Japan.

*E-mail: okayama575 @oki.co.jp

A concave lens placed in front of the micro-mirror scanner enlarges the scanning angle (Fig. 5). The focal length of the center convex lens is denoted as fcv'

The concave lenses having focal lengths of fcc are placed at dLL from the center of the convex lens. The distance between

294

H. OKAYAMA 295

8, No. 4 (2001) OPTICAL REVIEW Vol.

Input fiber Lens ( ~- \\)()

Opto-mechanical scanner Input

Output

Output fiber

---O-- - -- ---- ---

~~

////

-~

- --- :O// _ O- -

(a)

••> O

L

-----1-

eeee e'e o o e.eee

eeee Fig. 1.

///~

-----~-

O

~\ Lens ~l

(b)

Fig.

4. Bipolar scanning angle configuration I.

Beam steer ng type optical switch using optical scanners. i

Concave lens

~' Output scanner Input scanner \ .._\\.,~- ~•/l ' /

Optomechanical

Micro-mirror

Substrate

'~.

~'\+ _ _,.><~\

scanner

-> ->

Convex lens

;'i. ~~-:~1: ~ ~ ~ ~

;~ ~ ~

( a)

- ' - l- ' -~~~ ,-

\\

- - - - -~ "...

\

-~

dLm dLL

Micro-

Light bearn

mirror ----_ l~

/

'.'..~L

Lens

Fiber

\

Fig.

\Mirror

Od

(b)

Lens

2. Micro-mirror optical scanner. Fig.

5. Bipolar scanning angle configuration II.

Opto-mechanical scanner

Moving lens e)d

->

/

Fiber Fig.

by the convex lens. In the limit dLm << dLL the beam position

is approximated as O (dLL + 2dLm)' The beam diameter at the concave lens becomes dLL/(2fcv) = 1/2 - dLm/(2fcc) of

~'~\ ~~~\~_ \~+. ..:~ ~.~~**

3. Moving-1ens optical scanner.

micro-mirror or moving lens and concave lens is denoted as dLm' The light beams with various propagation angles should be converted to parallel propagated beams after the convex lens. The parallel beam entering the input-side concave lens should be converted to a parallel beam in the middle of the convex lens.

Analysis using a ray matrix shows that conditions fcclfcv = I + dLmlfcc and dLL = 2fcv fcc Should be es tablished. The light beams with scanning angle G) before the input concave lens ejected from the micro-mirror or moving lens are focused into the output concave lens at position Ofcc

the diameter at the convex lens. The diameter of the beam at the micro-mirror becomes almost 1/2 of that required for the collimator system and the beam scanning range also becomes 1/2 of the system without bipolar drive. The channel (or port) number increases compared at the same total length and maximum scanning angle, with the system without bipolar drive. Decreasing the diameter of the micro-mirror leads to a larger scanning angle, further increasing the channel number.

3. Beam Steering Error Asnalysis The effect of the beam steering error is different for the two

types of optical scanner shown in Figs. 2 and 3. The distance L between scanners should be long enough to scan all the out-

put ports in the limited scanner beam steering angle. A Iens collimates the light from the input fiber and the light beam is focused into an output fiber. Elements for steering the light beam, such as micro-mirrors, are inserted into the collimated

light beam. For the moving-1ens device, the collimator lens

H. OKAYAMA

296 OPTICAL REVIEW Vol. 8, No. 4 (2001)

Input scanner

-~::::""'~•-•••~.,, ~; . : ",~-

/~-' -,.

Light beams output scanner ~(1) i

' ' Collimator e d Fig. 7. Calculation scheme using overlap of light beams.

lens Fig. 6. Features of micro-mirror scanner.

=0.1 rad Om 105

itself works as the scanner. The diameter ~) of the light beam

should be large enough for the light to overcome diffraction as in Eq. (1), where ~ is the wavelength (1.55 um).

20000 -

E E

2000~ ~

* I04 (D

J2 E:SC 1000

,: e) C:

20

C: I OO

L ~ JT~)2/~. (1)

c(D

200

o ~s o c c: e)

(:l

~:

O 10

In the following, the conditions at the boundary are considered so that L = Jc~)2/~ and likewise.

~ e)

0.2

1

0.2

o o

2

O 4 0.6 0.8 1

~c

3

Error (o/o)

3.1 Micro-Mirror Scanner This type of device uses the reflection angle change by controlling the tilt angle of the mirror. The features of the micro-

Fig. 8. Calculated channel number versus error for nucro mnror scanner.

mirror scanner are shown in Fig. 6. 8G) is the mirror angle error, 8S the focus position error, ~)d the diffraction angle of the optical fiber and f the focal length of the collimator lens.

8S is given as

and optical fiber mode width co are the same.

For the micro-mirror device the attainable channel number is obtained from Eqs. (1)-(4) as

8S = f80 (2)

The incident angle error is given as AG)t = 80L/f. The angle AG)t Should be smaller than a certain value (1.5 deg for standard single mode fiber) to prevent loss increase. The fo-

cus position error dominates when f > (L8S/A~)t)1/2. The area of the mirror is assumed to be large enough so that the beam shift in the output mirror does not contribute to the loss.

The focal length f is related to ~)d as

~) = fG)d. (3) When the micro-mirror is fabricated on a substrate, N1/2 elements are placed in rows and columns for a device with N ports. The size of the substrate becomes A ~>N1/2 if we place elements at a distance of A~) (A ~ 4) to avoid crosstalk. The

maximum steering angle required is

G)m = A~)N1/2/(2L)

= A[~N/(4JTL)]l/2 (4) The effect of beam steering angle error can be given by

N = 4JTL[8S/(A8f)]2/~

= 4[7T~)d8S/(A8~)]2 (5) with 8 being the fractional maximum tilt angle error 8 = 8~)/~)*. Calculated channel number versus error 8 is shown in Fig. 8. The loss ipcrease due to focus position error dominates over fiber incident angle error when f > (L8S/A G)t) 112

so that G)d < L~AOt/(JT8S)]l/2. As will be shown in Sect. 3.4, the channel number takes maximum value at this

boundary and is given as N = 4JcA~)t8S/(A8)2. For 8 = 0.5% and with e)d the output fiber diffraction angle of O. I rad,

a device with N ~ 400 is attainable. From Eqs. (2) and (3) the lens focal length f = 2 mm, beam diameter ~) = 200 pam and substrate size becomes 16 mm. The length between scan-

ners becomes L = 80mm from Eq. (4).

3.2 Moving Lens Scanner This type of device uses the refraction angle change by controlling the position of the lens against the light beam. The features of the moving-1ens scanner are shown in Fig. 9. The

making 8S equal to affordable displacement of the focal spot,

deflection angle O changes with the lens displacement S as

which is typically I um for a single mode fiber. In another method, the effect of beam steering angle error can be analyzed by the overlap of light beams from and to the input and output scanner in the middle as in Fig. 7. The latter is valid when the directions of the light beams are almost the same. The two schemes give similar results when the percentages

The lens diameter ~)L Should be large enough to cover the required lens maximum displacement and ~). (~)L =

of the light field area overlap to gain the desired loss for the

beam waist Lco/(2do) (do: distance between fiber and lens)

G) = Slf. (6)

y (2S + (~?) with y being a constant and S the maximum displacement. The size of the actuator is denoted as clS with cl being a constant. The size of a scanner becornes ~)L + 2clS. When the scanner is fabricated on a substrate, Nl/2 elements

H. OKAYAMA 297

OPTICAL REVIEW Vol. 8, No. 4 (2001)

-E1000

N=500

E

OcS

stb

Actuator

cI ~2

.~ d,...T::...C~.;"'

/

OO

-//

,:

o o e)

c: 10

Moving lens

c:

o o * o ~c

Input fiber :'1'1']']L~~) =>

@ '/"~" ~~~~~~~~

/

_l

02

:!

(~)L

/

04

(a) l

0.2

0.4

0.6 0.8 1

3

:$;:::"II"

Error (olG)

""""'1". \)~

dS

E E

ccS

104

~ tb

Actuator

c

J2

1000

c:

o 15 o o c c o o L 'D ~c

Fig. 9. Features of movmg lens scanner

1 oo

10

0.2

0.4

0.6 0.8 1

are placed in rows and colurnns for a device with N ports. The size of the substrate becomes (~>L +2cl S)N l/2 . The maximum

3

Error (o/o)

steering angle required is

Fig. 10.

Calculated channel number versus error for moving-lens

scanner.

~)* = (~)L + 2clS)N1/2/(2L). (7) The steering angle error effect can be analyzed using the

scheme of Fig. 7 with 8 = 8G)lG)* (8~) = 8S/d*: 8S =

f = I .7 mm, beam diameter ~> = 169 pam and substrate size

rco), where r is the allowable displacement fraction against the light field diameter and d. the lens to the fiber distance

becomes 8 mm. Channel number in the optimum condition

(~ f). For the moving-1ens device, the attainable channel

N = [4Jc8SAG)t/(~82)]lL2(y + a)S[JT/(~L)]l/2 + yl2.

equating spot position and incident angle errors is given as

number N is related to the interconnection length

3.3 Thermal Expansion Effect

L = (JTN/~){2S(y + ot)/[(2Jc/A)(8S18)G)d - yNll2]}2 (8)

with S being the maximum displacement and S = 8S18. The movement of the focused light spot at the output fiber is equal to the lens displacement. The allowable spot shift is equal to allowable lens displacement error. To attain controllability, N1/28S/S << I is required. The incident angle

error is given as AOt

= 80L/f. The angle AOt Should

be smaller than a certain value (1.5 deg for standard single mode fiber) to prevent loss increase. The effect of focus position error dominates over fiber incident angle error when f > (L8S/AG)t)1/2 so that G)d < [~AG)t/(JT8S)]l/2. As will

be shown in Sect. 3.4, the channel number takes maximum

The effect of thermal expansion is analyzed using Fig. 1 1 . When the deflection angle is fixed, the position of the light

beam entering the output collimator lens moves due to the thermal expansion. However, the focal point does not move when the incident angle to the collimator lens remains the same. The input or output fibers and collimator lens are as-

sembled into a sub-assembly. The thermal expansion of the collimator lens amay and the fiber array should be the same to prevent the shift of the light beam center against the lens.

After being focused by the collimator lens, the incident angle into the output optical fiber changes with the light beam shift leading to insertion loss increase.

With fixed light beam angle from a scanner, the light bearn

value at this boundary and the interconnection length is given

shifts for amount St = (X - X')P + ((T - X')P with p being

as L = (JTN/~){2S(y +ce)/[2(JT8SAG)t/~)112/8 - yNll2]}2

the lateral relative position of input and output beam scanner

Calculated channel number versus error 8 is shown in Fig. 10.

array, P the lateral position difference between output and input scanner in the array, X the lateral expansion factor of a chassis holding the scanner array, X ' Iongitudinal expansion factor of a chassis holding the scanner array and cr the lateral

N versus f is given as in Eq. (9) (S = 8S18). Short f leads to a large channel number.

N = (4JTL/~)(Slf)21[2(y +cl)S[Jc/(~L)]l/2 + y]2. (9) The device design should be of a reasonable size L. Smaller L is attainable with the same N with larger ~)d. For Od =

expansion factor of the substrate holding the scanner. Using steel for the chassis and Si for the substrate, the light beam shift due to 100'C temperature change becomes several tens

0.1 rad, 8 = 0.5% and cl = y = 1, a device with N ~ 200

of micrometers when p and P are several centimeters. The

for L = 58 mm is attainable. Relatively long lens-traveling-

fiber incident angle change is given as St/f with f being the

length S = 200 pcm may require a scratch drive type ac-

collimator lens focal length. Long focal length (f > I mm)

tuator.6) From Eqs. (1), (3) and (6) the lens focal length

is beneficial for preventing insertion loss increase due to tem-

H. OKAYAMA

8, No. 4 (2001) OPTICAL REVIEW Vol.

298

Moving mirror substrate

Micro mirror plane

/

\

Projected micro mirror plane

Lens ~~ Output fiber

Input sub-

/

Focal plane

(a)

assembly

Lens Output sub-

Input fiber

~

assembly

l

Lens ------/

\

Focal plane

Deformation by thermal expansion

Moving mirror substrate

!\

/

(b)

Lens Fig.

Projected micro mirror plane

13. Interconnection length reduction.

1 1 . Thermal expansion effect.

Fig.

L > f2AOtl8S. The maximum channel number is at1 ooo

(a)_

~

_/'

e) J:)

tained at this boundary. Above this length, the available channel number decreases due to increment of fiber inci-

__~ __/ *\ __\ _.\

e =0 .-_\ 4% ~'\__/

f=1 mm --_

dent angle error. The beam position error at the output collimator lens, Ieading to fiber incident angle error, increases

em =0.25rad A=1 . 55kLm

E = c: ~5 1 oo c c (Q

with longer collimator length L . The channel number is derived as N = [4JT/(~L)][f AG)t/(A8)]2 for micro-mirror and N = [4JT/(~L)]( f Ae)t/8)2/[2(y +ce)S[Jc/(~L)]l/2 + y]2 for

~=1% em =0, ~

J:

o

moving-1ens devices, respectively, when the incident angle er10

10

1

ror is larger. For f = I mm, Ae)t = 0.03 rad and 8S = I um,

optimum length becomes L = 30 mm. The beam diarneter is ~) = 100 pLm. When 8 = 0.5% the available channel num-

Interconnection length (cm)

ber is N = 608 for a micro-mirror device. For a moving-1ens

1 ooo

f=1 mm

(b)

~ o

~E

device with 8 = 0.5% and S = 200 kLm, channel number N = 170.

A=1 . 551~m

4 e=0.%

:5 c:

Long focal length f tends to require a long interconnection

em =O.2)5 adi -- --

o c I OO

length. It is seen in Eqs. (5) and (9) that L is the limiting fac-

..);p

c~'

~:

O li

/ /

8=1%

tor for the attainable channel number at given focal length f ,

em =0. 1

errors 8 and 8S. The interconnection length L can be reduced by configurations5,6) shown in Fig. 1 3 using telescope optical

10

10

1

Interconnection length (cm)

Fig. 12. Calculated channel number versus interconnection length for device with focal length f = I mm for (a) micro-mirror scanner and (b) moving-1ens scanner.

perature change.

3.4 Designfor Given Focal Length Calculated micro-mirror device channel number N is given

systems. The light beam-scanning angle can be made larger at the interconnection stage than the micro-mirror or moving lens scanning angle. The scanning angle magnification factor is equal to focal length ratio Rf Of two lenses. The pro-

jected image of the scanner array is reduced. The required dis-

tance between input and output telescopes becomes I IR~ of the conventional system's interconnection length at the same channel number. However, the attainable channel number due to the error remains the same. The bipolar drive configuration described in the introduction can be used together with the system shown in Fig. 13. The telescope optical system is inserted between the convex and concave lenses.

in Fig. 12 for f = I mm 8S = I um and AOt = 0.03rad. Figure 12(a) is for the micro-mirror device and Fig. 12(b) the

moving-1ens device, respectively. The maximum steering angle Om is related to the focal length f as G)m = 8Sl(ef). The channel number N is a function of length L as in Eq, (4)

with ~)m being a parameter. When ~)m is over an achievable value the channel number in Fig. 12 is not attained even if error 8 is small. When error 8 is smaller than the achievable value the channel number in Fig. 12 is not attained even if angle G)m is large. The fiber incident angle error becomes larger than focal spot position error when

4. Improvement Schemes As we have shown the beam steering angle error tolerance sets the upper boundary of the attainable channel number N, especially when we do not want to rely on a sophisticated control method. The sophisticated control circuit limits the switching speed and increases the cost. Wear (creeping effect) of the moving-mirror may degrade the precision. Some designs to improve the tolerance against beam steering error are desirable.

H. OKAYAMA 299

OPTICAL REVIEW Vol. 8, No. 4 (2001)

Micro-lens array

Collimator lens

'~~Movmg-mirror Lens ///1/1w!~;~~}/;;

Moving-mirror x

' > ~~.'*"=~',~'~;**i'j~~~'(*'~~*;'~"'*"~'~*=i*

j . ~ens Input

~;;

i" Lens :

; ""/*:~'*:'~'~~*'>,~*;*=,~,~~';i=.".="'*'~;;::~:.;

-"I . ~}•-

////

.. *~"*""' ~i'~;~~

- " Diverging element Multi-Port coupler

i ii~i ;! '~

i'!i"';;/if / /

{i/{f j,)// (/' /

Moving-m:S::ii;;irror ~ fl:

Input

'. "="'~~~~'~':~~~'~='~~':'i OPtical fiber

' Optical fiber

/ i~i/f:i/'i /1' : I Output

}! /' Lens

(b)

(a)

Fig. 16. Reduction of lens to optical fiber distance by (a) diverging optical element and (b) multiple lenses.

Output (b)

(a)

Output

Fig. 14. Digital beam scanner (a) I x N and (b) N x N.

Moving-mirrors

/j~/////

Cylindrical MMI Lens""' "'~\'

,):

~

/

f)\,!' //'//

IL

~

/ ll

-;;r ~

Output fiber

.~f :,\,/.: " Lens

~

\

'L

.. ' .. .'

\\"''

~~~"

Input

MOving-mirrors

Lens Fig. 15. Digital beam scanner using MMI coupler.

4. I Digital Beam Scanner The spot movement can be minimized by an optical system with an image plane located at a moving mirror (Figs. 14(a) and (b)) even for long focal length f. For a N x N switch,

2N - I concave lenses in rows and columns, respectively,

Fig.

17. Optical switch network.

are placed in the middle of light paths. The bipolar drive ar-

chitecture described in the last part of Sect. 2 is adopted in Fig. 14(b). The tolerance for the input angle AG)t (1.5 deg for

single mode fiber) becomes less for long focal length. The steering angle error results in a shift of small fiber input an-

gle and minimum spot position shift. The optical system also compensates for the thermal expansion effect. A small focal spot movement for fabrication error adjustment by the moving-mirror can be attained by several designs, A moveable lens or additional mirror near the lens can be used to adjust the fabrication error. A slight displacement of the

mirror position from the focal point produces small movement of the focal position at the output fiber. A precise alignment of the focal spot and the fiber is attained.

allowable displacement error increases channel number as indicated in Eqs. (5) and (8).

Two other possible methods are shown in Fig. 16. The first method uses a diverging optical element and the second uses multiple lenses. A Iarge light beam is divided into small diameter bearns and focused into a coupler to combine the light. The spot shift due to error is reduced by short fiber-1ens dis-

tance with fiber incident angle effect remaining the same.

5. Optical Switch Network Several devices can be connected9) to each other to increase

the channel number over which can be attained by a single

4.2 ScannerDesign A digital switching curve for the output power against the mirror angle is achieved by the structure in Figs. I and 2 or 3 using standard single mode input fiber and output fiber

connected to a multi-mode interference (MMI) couplerll)

device (Fig. 17). The structure of an optical switch is similar to that of Fig. 1

except that multiple output ports are provided for an output micro-mirror scanner. The output scanner selects a desired

output port among the multiple output ports. The optical

(Fig. 1 5), at the expense of a loss increase due to mode mis-

switch network is strictly non-blocking9) when using n micro-

match. The concept is similar to that commonly used in ar-

mirrors at the input and more than 2n - I micro-mirrors at the

ray waveguide devices 12) The flat top field distribution is at-

output. When q outputs are possible for a single micro-mirror at the output, qn total network channel can be attained. Bundled fibers are used at the interconnection to reduce the num-

tained using a cylindrical MMI coupler as revealed by BPM13)

(beam propagation method) simulation. As long as the input light field is in the position overlapping with a much wider light field of MMI, the insertion loss remains low. Increasing

ber of connectors.

300

H. OKAYAMA

OPTICAL REVIEW Vol. 8, No. 4 (2001)

Switch fabric

6. Conclusrons Beam steering error for devices using tilted mirrors or a moving lens has been analyzed and schemes to overcome the limit have been reported. The attainable channel number in the conventional device is limited by the light beam steering

Output

Input

Monitor

Power monitor

angle error to several hundred ports. Difference in the limit of light beam steering angle due to actuator type leads to different results for micro-mirror and moving-lens devices. Several possible schemes to increase the channel number, such as

inserting dedicated optical components, using novel optical system designs or optical switch networks, are proposed. A device with several hundred ports is possible with a beam steering error under 1%. To minimize the thermal expansion effect large focal length is required indicating need for a com-

pensation scheme to attain a larger channel number. There is

Controller Fig. A- I . System architecture with power monitoring and standby optical switch.

an optimum design for attaining maximum channel number at a given focal length.

Schemes described in Sect. 4.2 are an improvement for each scanner and 4.1 for the total optical system design. A drawback is that new optical components are needed for the scheme in Sect. 4.2. Using a new total optical system design

may overcome the limitation set by Eq. (1). The improved optical switches can be connected as in Sect. 5 to increase the channel number above that is attainable with a single device. More detailed analysis of improvement schemes will be reported elsewhere .

7) B. H. Lee and R. J. Capik: Proc. ECOC2000 (2000) p. 95. 8) H. Okayama: Patent appllcation (1997). 9) H. Okayama and T. Ushikubo: Technical Dfgest ofthe 2nd International

Conference on Optical Design and Fabriccttion (ODF2000) (2000) p. 26.

lO) H. Okayama, Y. Okabe, T. Aral, T. Kamijoh and T. Tsuruoka: J. Lightwave Technol. 18 (2000) 469. 1 1) E. C. M. Pennings, R. van Roijen, B. H. Verbeek, R. J. Deri and L. B. Soldano: Technical Digest LEOS ( 1 993) Paper 102. I .

12) M. K. Smit and C, van Dam: IEEE J. Select. Top. Quantum Electron. 2 ( 1 996) 236.

13) H. J. W. M. Hoekstra: Opt. Quantum Electron. 29 (1997) 157.

References 1) D. T. Neilson, V. A. Aksyuk, S. Arney, N. R. Basavanhally, K. S. Bhalla, D. J. Bishop, B. A. Bole, C. A. Bolle, J. V. Gates, A. M. Gottlieb, J. P.

Hickey, N. A. Jackman, P. R. Kolodner, S. K. Korotky, B. Mikkelsen, F. Pardo. G. Raybon, R. Ruel, R. E. Scotti, T. W. Van Blarcum, L. Zhang

and C. R. Giles: Technical Digest OFC2000 (2000) PD12. 2) H. Laor: Technical Digest LEOS99 ( 1999) Paper WK1. 3) R. Giles: Technical Digest CPT2001 (2001) p. 15. 4) A. NeukenTlans: Techniccd Digest CPT200] (2001) p, 19. 5) H. Toshiyoshi, J. G.-D. Su. J. LaCosse and M. C. Wu: Tc'chinical Digest

MOEMS99 ( 1999) Talk26. 6) H. Toshiyoshi, G.-D. J. Su, J. LaCosse and M. C. Wu: Proc, of the IEEE/LEOS International Conference on Optical MEMS (2000) Paper PD- I .

Appendix System architecture as shown in Fig. A-1 can be used to monitor the light beam position and place a redundant standby switch. The monitor light is fed into the signal light by a cou-

pler. The coupler also divides the signal light to the redundant

switch. At the output, a coupler combines the signal lights from working and redundant switches. The monitor light is coupled out into the photodiode for monitoring. Adjusting the beam angle reduces the insertion loss and overcomes the channel number limit due to temperature change. Improved reliability by redundancy and 1-t0-2 multicasting are attained.