TECHNICAL FIELD

This invention relates to optical networks, and more particularly, to optical cross-connect switches used for routing multi-wavelength optical signals.

BACKGROUND OF THE INVENTION

Optical networks are widely used to transport large volumes of telecommunications traffic. For example, systems employing wavelength division multiplexed (WDM) technology are capable of supporting dozens of communications channels transported at different wavelengths on just a single optical fiber.

In a multi-fiber network, some of the many channels on individual optical fibers may need to be selectively routed to other fibers. Selective routing may be required, for example, to balance communications traffic, or to avoid an out-of-service leg in the optical network. Such routing can be facilitated by interconnecting the individual optical fibers via an optoelectronic cross-connect switch. However, these switches suffer the disadvantage of requiring the multiple conversion of WDM signals first from optical form into electronic form and then back into optical form. It would be advantageous if the optical switching could be performed without these conversions.

Some cross-connect switch fabrics have been devised that enable WDM signals to be optically switched (see, e.g., U.S. patent application Ser. No. 09/123,085 now U.S. Pat. No. 6,067,389, entitled WAVELENGTH-SELECTIVE OPTICAL CROSS-CONNECT, assigned to Lucent Technologies Inc., having a filing date of Jul. 27, 1998). These fabrics typically use various optical filter technologies to “select” the optical channels to be routed (such filters also being referred to as “wavelength-selective elements”). However, filter performance factors such as insertion loss, “blue” wavelength loss and tuning range effectively limit the number of communications channels that can be supported by each fabric. For example, in wavelength-selective cross-connect (WSXC) fabrics employing tunable fiber Bragg gratings as wavelength selective elements to select optical channels with 50 gigahertz spacing, experience suggests a practical limit of about ten wavelength-selective elements per path through the fabric. As a result, optical WSXC fabrics have not been used to support large-scale optical cross-connect switch applications.

Thus, in order to employ current optical switching components (such as WSXC fabrics) in large-scale optical cross-connect switch applications, an optical cross-connect switch architecture is required that is capable of switching optical signals with a large number of optical channels while requiring only a small number of wavelength-selective elements on each signal path through the switch.

SUMMARY OF THE INVENTION

The number of wavelength-selective elements required on each signal path through an optical cross-connect switch is substantially reduced in a novel optical cross-connect switch architecture employing multiple WSXC fabrics. One or more optical channel distributors each receive a multi-wavelength optical signal as input. Each signal contains a plurality of channels, each associated with one or a plurality of wavelengths. Each distributor distributes each channel in its associated multi-wavelength input signal to one of a plurality P of WSXC fabrics. Each WSXC fabric is arranged to receive channels from the one or more distributors that are associated with a unique subset of the plurality of wavelengths.

Upon receiving the distributed channels, each WSXC fabric employs wavelength-selective elements on each of a plurality of WSXC fabric cross-paths to route each received channel to one of a plurality of optical combiners. Each combiner then combines the channels it receives from each WSXC fabric to produce and output a multi-wavelength optical signal with a full complement of channels. By employing this architecture, the number of channels carried by each WSXC fabric cross-path is reduced by a factor of P over the number of channels present in the multi-wavelength input and output signals.

In an exemplary embodiment of the invention, the distributors comprise optical slicers and the wavelength-selective elements comprise tunable fiber Bragg gratings (FBGs). The FBGs are tunable to reflect or pass optical signals in associated channels. The optical slicers operate to cause the optical channels associated with each multi-wavelength input signal to be allocated to the individual WSXC fabrics such that spacing between adjacent channels on each fabric is increased over spacing between adjacent channels in the input signal. This added spacing provides a “parking space” to which FBGs associated with adjacent channels may be tuned, thus enabling signals to pass in the adjacent channels. In addition, the added spacing reduces the effects of overlapping “blue” wavelength losses contributed by the FBGs on each path.

BRIEF DESCRIPTION OF THE DRAWING

The invention will be more fully understood from the following detailed description taken in connection with the accompanying drawing, in which:

FIG. 1

depicts a typical prior art large-scale optical network employing an electronic cross-connect switch;

FIG. 2

provides a schematic diagram for a prior art wavelength-selective cross-connect fabric;

FIG. 3

presents a schematic diagram of a preferred embodiment of the present invention;

FIG. 4

depicts the operation of an optical wavelength slicer suitable for use in the present invention;

FIGS. 5A

,

5

B illustrate a first advantage of optical wavelength slicers used in the present invention by showing channel spacing for sliced and non-sliced signals;

FIGS. 6A

,

6

B illustrate a second advantage of optical wavelength slicers used in the present invention by showing “blue” wavelength losses for sliced and non-sliced signals; and

FIG. 7

shows a bi-stable reconfigurable fiber Bragg grating suitable for use in the present invention as a wavelength-selective element.

For consistency and ease of understanding, those elements of each figure that are similar or equivalent share identification numbers that are identical in the two least significant digit positions (for example, WSXC input

206

of FIG.

2

and WSXC input

306

of FIG.

3

).

DETAILED DESCRIPTION

In order to better understand the principles of the present invention, several key aspects of the cross-connect switch prior art are first reviewed.

FIG. 1

illustrates a typical architecture for an optical network

100

. The network

100

includes long-haul optical fibers

102

, a first metro ring

104

, a second metro ring

106

, and add/drop lines

110

. Optical signals directed to and from these elements of the network

100

are routed by a cross-connect switch

108

.

Cross-connect switches supporting large-scale optical networks must have significant capacities. In network

100

, for example, cross-connect switch

108

is configured to switch multi-wavelength signals among three fiber pairs

102

carrying eighty channels each, two bi-directional metro rings

104

and

106

carrying forty channels each, and a variety of single-wavelength add/drop lines

110

. To date, such large-scale optical cross-connect switches typically operate to transform optical signals into electronic signals that are switched and routed electronically. For example, in optical network

100

, cross-connect switch

108

receives optical signals at switch input ports

112

, transforms these optical signals into electronic signals, routes the electronic signals within switch

108

, transforms the routed electronic signals into optical signals, and delivers the transformed optical signal from switch output ports

114

.

Although such optoelectronic optical cross-connect switches provide significant switching capacities, they suffer the disadvantage of requiring the multiple conversion of multi-wavelength WDM signals first from optical form into electronic form and then back into optical form. It would be advantageous if the optical switching could be performed without these conversions.

Some cross-connect switch fabrics have been devised that enable WDM signals to be optically switched (see, e.g., U.S. patent application Ser. No. 09/123,085 now U.S. Pat. No. 6,067,389 entitled WAVELENGTH-SELECTIVE OPTICAL CROSS-CONNECT, assigned to Lucent Technologies inc., having a filing date of Jul. 27, 1998).

FIG. 2

illustrates the operation of an optically switched wavelength-selective optical cross-connect (WSXC) fabric

200

. In

FIG. 2

, WSXC fabric

200

is arranged to switch a multi-wavelength optical signal of two channels arriving at input

206

. Optical coupler

202

routes copies of the WDM input signal received at input

206

to links

210

,

220

. Link

210

interconnects to wavelength-selective element

212

, which further interconnects to optical combiner

250

via link

214

, a second wavelength-selective element

216

and link

218

. Wavelength-selective elements

212

and

216

are each controllable either to filter one of the two WDM channels in the optical signal received at the input

206

such that the filtered channel does not reach optical combiner

250

and output

254

. Accordingly, by appropriately controlling elements

212

and

216

, neither, one or both channels may pass on to optical combiner

250

and output

254

.

Similarly, link

220

interconnects and establishes a path to optical combiner

252

via wavelength-selective element

222

, link

224

, wavelength-selective element

226

and link

228

. Wavelength-selective elements

222

and

226

are each controllable to filter one of the two WDM channels in the optical signal received at the input

206

. Again, by appropriately controlling elements

222

and

226

, neither, one or both channels may pass on to optical combiner

252

and output

256

.

Following a similar path, two channels in a WDM signal arriving at an input

208

may be selectively filtered or passed on to output

254

over a path including links

230

,

234

and

238

, wavelength-selective elements

232

and

236

and optical combiner

250

. In addition, these channels may be selectively filtered or passed on to output

256

over a path including links

240

,

244

and

248

, wavelength-selective elements

242

and

246

and the optical combiner

252

. In this manner, each of the WDM channels at inputs

206

and

208

may be filtered or passed on to each of the outputs

254

and

256

, respectively.

Generally, the WSXC fabric

200

will receive signals at inputs

206

and

208

comprising channels centered at equivalent wavelengths &lgr;. For example, each of the signals at inputs

206

and

208

may comprise a first channel centered at a wavelength &lgr;

1

and a second channel centered at a wavelength &lgr;

2

. By operation of WSXC fabric

200

, at most one of each of the input signal channels centered at wavelengths &lgr;

1

and &lgr;

2

will typically be forwarded to outputs

254

and

256

. Accordingly, each of the signals at outputs

254

and

256

will contain at most two channels, and these two channels will be centered at wavelengths &lgr;

1

and &lgr;

2

, respectively. For example, WSXC fabric

200

may operate such that the &lgr;

1

channel from the signal at input

206

and the &lgr;

2

channel from the signal at input

208

are forwarded to comprise the signal at output

254

.

As the number of channels in an associated WDM signal grows, the complexity of WSXC fabric

200

increases. Each additional input channel requires that an additional wavelength-selective element be placed in each WSXC path between inputs

206

,

208

and outputs

254

,

256

. For example, a WSXC fabric supporting four input signals each containing 20 channels would require 20 wavelength-selective elements on 16 paths, or 320 wavelength selective elements in total. Significantly impacted by filter performance factors such as insertion loss, “blue” wavelength loss and tuning range, such large fabric configurations are currently impractical.

However, a large-scale optically-switched optical cross-connect switch can be realized, for example, by incorporating a plurality of conventional optical WSXC fabrics in a novel optical cross-connect switch architecture

301

shown in FIG.

3

. This architecture is the subject of the present invention, and is described in further detail below.

Included in the cross-connect switch architecture

301

is a series of optical WSXC fabrics

300

,

386

and

388

that each receive optical signals provided by a series of optical branch elements

372

,

374

and

376

. Optical WSXC fabrics

300

,

386

, and

388

of this type operate according to the principles outlined for the WSXC fabric

200

of

FIG. 2

, and are well-known in the art (see, e.g., Daniel Y. Al-Salameh et al., “Optical Networking,” Bell Labs Technical Journal, Vol. 3, No. 1, January-March 1998, pg. 57). Optical branch elements

372

,

374

and

376

may be optical band demultiplexers commercially available, for example, from JDS FITEL Inc.

Optical branch element

372

receives a first multi-wavelength optical input signal I

1

, at input

373

. Signal I

1

comprises N channels. Optical branch element

372

demultiplexes the signal I

1

, into a plurality of signals each containing one or more of the N channels. Each of plurality of signals is delivered to one of the WSXC fabrics

300

,

386

or

388

over links

306

,

359

or

365

, respectively.

A similar arrangement is configured for each additional signal I

2

through I

M

. For example, the signal I

2

is received at an input

375

and demultiplexed by the optical branch element

374

into a second plurality of signals each containing one or more of N channels Each of the second plurality of signals is delivered over one of the links

308

,

361

or

367

to the WSXC fabrics

300

,

386

or

388

, respectively.

Also included in the optical cross-connect switch architecture are a series of optical combiners

378

,

380

and

382

that each receive optical signals outputted by the WSXC fabrics

300

,

386

and

388

, respectively. Optical combiners

378

,

380

and

382

may comprise, for example, optical combiners, optical wavelength multiplexers, inverse optical band demultiplexers or inverse optical slicers. Optical wavelength multiplexers are well-known in the art and are commercially available, for example, from JDS FITEL Inc.

In

FIG. 3

, multiplexer

378

produces a first multi-wavelength optical signal O

1

over an output

379

that is formed from a plurality of N channels. The N channels are delivered by the WSXC fabrics

300

,

386

and

388

over links

354

,

360

and

366

, respectively. Multi-wavelength optical signals O

2

through O

R

are produced in a similar fashion by multiplexers

380

through

382

, respectively.

By means of the cross-connect switch architecture

301

of

FIG. 3

, a component signal in a selected channel in one of the input signals I

M

may be directed to a selected channel in one of the output signals O

R

, such that the selected input channel and the selected output channel are both associated with a common wavelength. For example, a component signal in signal I

2

associated with a wavelength &lgr;

2

may be directed over link

361

, through WSXC fabric

386

, over link

360

and through optical combiner

378

to a channel in output signal O

1

associated with wavelength &lgr;

2

. Alternatively, a component signal in the input signal I

M

associated with a wavelength &lgr;

3

may be directed over link

369

, through WSXC fabric

388

, over link

368

and through optical combiner

382

to a channel in output signal O

R

associated with wavelength &lgr;

3

. Significantly, any number of channels N in each of the input signals I

M

and output signals O

R

is effectively accommodated so long as:

P>N/Q  (1)

Where P represents the number of WSXC fabrics incorporated in the cross-connect switch architecture

301

of

FIG. 3

, and Q is equal to the a maximum number of wavelength-selective elements associated with each of the WSXC fabrics

300

,

386

and

388

.

In the cross-connect switch architecture

301

of

FIG. 3

, the number of input signals I

M

may differ from the number of output signals O

R

. This asymmetry may result, for example, in cases where additional signals of one or more channels are added or dropped at add/drop ports are incorporated in the WSXC fabrics

300

,

386

or

388

.

The advantages of the present invention are particularly apparent in large-scale multi-wavelength cross-connect system applications (for example, as depicted in FIG.

1

). Here, the number of optical channels that can be transported over a single optical fiber far exceed the limited number of wavelength-selective elements that can be practically placed in each path of a WSXC fabric. These limitations arise, for example, from performance characteristics of the wavelength-selective elements such as insertion losses, “blue” wavelength losses and channel spacing.

Insertion losses accumulate additively according to the number of wavelength-selective elements present on a path. Per-element insertion losses may vary, for example, from about 0.1 dB for tunable FBGs to as much as 8 dB for arrayed waveguide gratings (see, e.g., D. Sadot et al., “Tunable Optical Filters for Dense WDM Networks,” IEEE Communications Magazine, December 1998). Because of their low insertion loss and ease of tuning, FBGs are employed in a preferred embodiment of the invention as the wavelength-selective elements of the WSXC fabrics

300

,

386

,

388

of FIG.

3

.

Although favored because of low insertion losses, the number of tunable FBGs that can be placed in a path through a WSXC fabric is limited by other characteristic losses. FBGs are designed to attenuate optical signals at wavelengths within a sharply-defined interval centered at a characteristic wavelength. While attenuation at wavelengths above this range is generally negligible, FBGs may attenuate optical signals by as much as 5 dB across a broad range of wavelengths below the attenuation interval (in a region referred to as the “blue” wavelength region).

To further illustrate the effect of “blue” wavelength losses,

FIG. 6A

shows an attenuation profile for a series of five FBGs on a single transmission path. The five FBGs are centered at wavelengths &lgr;

1

through &lgr;

5

, respectively. “Blue” wavelength losses are shown in a region

610

below an attenuation interval

612

. Losses in this region are attributable to FBGs centered at wavelengths &lgr;

2

through &lgr;

5

, and accumulate to produce a between-channel loss

602

in the region

610

. Accordingly, this “blue” wavelength loss limits the number of FBGs that can effectively be placed in a single WSXC path. Experience suggests a current limit of about ten FBGs.

In addition, the number of tunable FBGs that can be placed in a WSXC path is limited by wavelength spacing. Tunable FBG's operate to reflect light at a specified wavelength. Light at this wavelength can be made to pass through the FBG by changing the properties of the FBG (“tuning”) so that light is reflected instead at an alternate wavelength. Methods for tuning FBGs include applying thermally-induced or piezoelectrically-induced strains (see, e.g., S. Jin et al., “Broad-range, latchable reconfiguration of Bragg wavelength optical gratings,” Applied Physics Letters, Vol. 74, No. 16, Apr. 19, 1999). If a multi-wavelength signal on a WSXC path is intended to pass over the path unfiltered, each of the FBGs on the WSXC path must be tuned to a wavelength not present in the signal. FBGs are typically limited to a tuning range of less than 10 nanometers (see, e.g., D. Sadot et al., “Tunable Optical Filters for Dense WDM Networks,” IEEE Communications Magazine, December 1998). For multi-wavelength signals containing adjacent channels with limited wavelength spacing between channels, each FBG on a WSXC path must be tuned to a wavelength outside of the aggregate pass band in order for the a signal in an associated channel to pass. Where individual channels may be spaced, for example, 1.5 nanometers apart, the number of channels (and FBGs) on a path is effectively limited to six.

However, limitations on the number of FBGs inserted in each WSXC path may be accommodated by the inventive optical cross-connect switch architecture of FIG.

3

. By employing the demutiplexers

372

,

374

and

376

, the multiplexers

378

,

380

and

382

and the additional WSXC fabrics

386

and

388

, for multi-wavelength signals with a given number of channels, the number of WSXC paths may be increased in order to decrease the number of WSXC elements required on each path. In particular, for a system configuration switching N optical channels on M incoming fibers to R outgoing fibers, M 1xP optical branch elements, P WSXC fabrics and R Px1 multiplexers may be interconnected as shown in

FIG. 3

to reduce the number of FBGs required on each cross path from N to N/P. As a result, the number of multi-wavelength channels that can be switched on each incoming optical fiber is increased over prior systems by a factor of P.

Notably, in addition, filter limitations imposed by channel spacing requirements are eased in a preferred embodiment of the inventive optical cross-connect switch architecture

301

of

FIG. 3

in which the optical signal distributors

372

,

374

and

376

are selected to be optical wavelength slicers. Optical slicers are well-known in the art (see, e.g., see, e.g., U.S. Pat. No. 5,694,234, entitled WAVELENGTH DIVISION MULTIPLEXING PASSIVE OPTICAL NETWORK INCLUDING BROADCAST OVERLAY, having an issue date of Dec. 2, 1997) and are commercially available, for example from Chorum Technologies Inc. Optical slicers typically inject smaller insertion losses in the signal stream than other branch elements (for example, slicer insertion losses may be on the order of 2 dB while band demultiplexer insertion losses may be on the order of 8 dB). In addition, optical wavelength slicers effectively allocate optical channels to slicer outputs in a manner that maximizes spacing between adjacent channels on each slicer output.

FIG. 4

illustrates this spacing effect by examining the operation of a 1x4 optical slicer

400

. Input

402

to optical slicer

400

carries a multi-wavelength optical signal including channels centered at adjacent wavelengths &lgr;

1

through &lgr;

12

. Optical wavelength slicer

400

allocates the channels at adjacent wavelengths &lgr;

1

, through &lgr;

4

to outputs

404

,

406

,

408

and

410

, respectively. This allocation pattern is repeated for subsequent adjacent channel sets &lgr;

5

through &lgr;

8

and &lgr;

9

through &lgr;

12

. As a result, adjacent channels placed on the outputs

404

,

406

,

408

and

410

have a spacing between wavelengths associated with adjacent channels that is that is four times as large as the spacing between wavelengths associated with adjacent channels on the input

402

. More generally, in a 1xP optical wavelength slicer, spacing between wavelengths associated with adjacent channels is increased by a factor of P.

FIGS. 5A

,

5

B illustrate some of the benefits derived from the additional spacing generated by the 1x4 optical slicer

400

of FIG.

4

.

FIG. 5A

shows signal reflectivity as a function of optical wavelength for a non-“sliced” optical path incorporating a plurality of FBGs. Envelope

502

characterizes optical signal reflectivity for a first FBG with an attenuation interval centered at a wavelength &lgr;

a

. Envelope

512

similarly characterizes optical signal reflectivity for a second FBG with an attenuation interval for an adjacent channel centered at a wavelength &lgr;

b

. In order to transmit a signal in a channel associated with the wavelength &lgr;

a

(i.e., to pass the signal through the first FBG), the first FBG must be tuned to shift the envelope

502

so that reflectivity is low near &lgr;

a

. Envelope

504

illustrates this shift.

Because channel spacing

508

between &lgr;

a

and &lgr;

b

is limited, the first FBG must be tuned carefully. If envelope

504

is shifted a distance

506

that is too far, it will attenuate the transmission of a channel associated with &lgr;

b

. If distance

506

is too little, the shift will be insufficient to enable the transmission of the channel associated with &lgr;

a

.

Channel spacing limitations are largely eliminated by selecting optical slicers (such as the optical slicer

400

of

FIG. 4

) as the optical signal distributors

372

,

374

and

376

of FIG.

3

.

FIG. 5B

illustrates reflectivity as a function of wavelength for one of the output signals of optical slicer

400

of FIG.

4

. In

FIG. 5B

, channel spacing

510

is increased by a factor of four over channel spacing

518

for the non-sliced input signals of FIG.

5

A. As a result, envelope

504

can more easily be shifted a distance

508

away from the region defined by envelope

502

without impinging on the region for adjacent channel &lgr;

c

as characterized by envelope

514

.

In addition to generating additional spacing between adjacent channels, optical slicers provide the additional benefit of reducing “blue” wavelength losses.

FIG. 6B

shows signal losses as a function of wavelength for a “sliced” WSXC fabric path containing three FBGs (not shown) centered at wavelengths &lgr;

1

, &lgr;

3

and &lgr;

5

, respectively. This path has been configured to receive a multi-wavelength signal input from a 1x2 slicer output.

FIG. 6B

can be compared to

FIG. 6A

, which shows signal losses for a WSXC fabric path containing five FBGs centered at wavelengths &lgr;

1

through &lgr;

5

, respectively. The WSXC fabric path of

FIG. 6A

is configured to receive a non-“sliced” signal input.

As compared to

FIG. 6A

, region

614

of

FIG. 6B

between adjacent FBG attenuation intervals

608

and

616

is twice as large as region

610

between the adjacent FBG attenuation intervals

608

and

612

of FIG.

6

A. Using a suitably large 1xP slicer, the region

614

provides an ample “parking lot” for shifting the attenuation interval

608

without affecting adjacent channels. For example, with channel spacing of 50 gigahertz and P equal to 8, a spectral separation of 400 gigahertz is created.

In addition, accumulated “blue” wavelength losses between attenuation intervals are improved in the configuration of FIG.

6

B. In

FIG. 6B

, an accumulated “blue” wavelength loss

604

is shown as the sum of losses

607

arising in the region

614

from FBGs centered at wavelengths &lgr;

3

and &lgr;

5

. This accumulated “blue” wavelength loss

604

is substantially less, for example, than the accumulated “blue” wavelength loss

602

shown in the non-“sliced” configuration of FIG.

6

A.

A variety of FBG configurations may be used in support of the present invention. For example,

FIG. 7

provides a cross-sectioned view illustrating a tunable, bi-stable FBG

700

(see, e.g., U.S. patent application Ser. No. 09/159,380 now U.S. Pat. No. 6,055,348, entitled TUNABLE GRATING DEVICE AND OPTICAL COMMUNICATIONS DEVICES AND SYSTEMS COMPRISING THE SAME, assigned to Lucent Technologies Inc., having a filing date of Sept. 23, 1998). While bi-stable FBG

700

represents a preferred embodiment for the wavelength-selective elements employed by the present invention, it will be clear to one skilled in the art that a variety of other wavelength-selective elements may alternatively be employed.

Bi-stable FBG

700

includes an optical fiber

702

incorporating a Bragg grating

704

, both disposed within guiding tube

706

. Fiber

702

is affixed to guiding tube

706

at a point

716

near distal end

705

of guiding tube

706

. Also disposed within and affixed near opposing distal ends

705

and

707

, respectively, of guiding tube

706

are fixed magnets

710

and

712

. An axially magnetized programmable magnet

708

is movable between magnets

710

and

712

, and affixed to optical fiber

702

at a point

718

. Point

718

is centrally positioned with respect to distal ends

705

,

707

, such that Bragg grating

704

is substantially positioned between the points

716

,

718

. Guiding tube

706

is further encapsulated by solenoid

714

.

As shown in

FIG. 7

, the adjacent poles of movable magnet

708

and fixed magnet

710

are of opposite polarity and the adjacent poles of movable magnet

708

and fixed magnet

712

are of like polarity. This causes an attractive force to be generated between movable magnet

708

and fixed magnet

710

and a repulsive force to be generated between movable magnet

708

and fixed magnet

712

. As a result, movable magnet

708

moves toward fixed magnet

710

, and a compressive strain &egr;

c

(not shown) is generated in optical fiber

702

within the region of Bragg grating

704

.

A pulse current may be applied to solenoid

714

to invert the polarity of movable magnet

708

so that adjacent poles of movable magnet

708

and fixed magnet

710

are of like polarity and adjacent poles of movable magnet

708

and fixed magnet

712

are of opposite polarity. As a result, an attractive force is generated between movable magnet

708

and fixed magnet

712

and a repulsive force is generated between movable magnet

708

and fixed magnet

710

. This causes movable magnet

708

to move toward fixed magnet

712

, generating a tensile strain &egr;

t

(not shown) in optical fiber

702

within the region of Bragg grating

704

. A second pulse current may be applied to invert the polarity of magnet

708

to return once more to its initial condition. Control units for pulsing solenoid

714

in this manner are well-known in the art.

Movable magnet

708

is made of a material that retains a large portion of an induced magnetization after an inducing field is removed (for example, an iron-chromium-cobalt alloy). Thus, a single pulse in solenoid

714

is sufficient to set the polarity of magnet

708

and place device

700

into one of two bi-stable states. The strains &egr;

c

, &egr;

t

induce a shift in the wavelength attenuation interval of FBG

700

that is related both to the strains and to a photoelastic constant representing the effect of the strains on the refractive index of the fiber.

The exemplary embodiment described above is but one of a number of alternative embodiments of the invention that will be apparent to those skilled in the art in view of the foregoing description. Accordingly, this description is to be construed as illustrative only, and is for the purpose of teaching those skilled in the art the best mode of carrying out the invention. Various other alternatives can be devised by a worker skilled in the art without departing from the teachings of this invention. For example, instead of tuning FBGs, optical “cross-bar” bypass switches may be employed to switch FBGs into and out of a signal path (see, e.g., U.S. Pat. No. 5,712,932, issued to Alexander et al. on Jan. 27, 1998). Alternatively, Fabry Perot etalons tunable by microelectromechanical switches (MEMS) may be used to replace FBGs as wave-selective elements (such devices are commercially available, for example, from CoreTek, Inc.). The inventive optical cross-connect switch architecture

301

of

FIG. 3

may also be adapted to include additional ports to provide an add/drop function, optical isolators to eliminate stray signals reflected by the FBGs, and optical amplifiers to overcome insertion and other losses. Also, the optical combiners

378

,

380

and

382

of

FIG. 3

may be replaced, for example, by optical multiplexers, star couplers, inverse optical band demultiplexers or inverse optical slicers