WO1996009727A1

WO1996009727A1 - Optical crossbar switch

Info

Publication number: WO1996009727A1
Application number: PCT/GB1995/002245
Authority: WO
Inventors: William Alden Crossland; Robert Joseph Mears; Stephen Thomas Warr
Original assignee: Cambridge University Technical Services Limited
Priority date: 1994-09-23
Filing date: 1995-09-21
Publication date: 1996-03-28
Also published as: AU3527895A; GB9419278D0

Abstract

There is provided a switch for connecting one or more data inputs to one or more outputs, in which the interconnections are made optically, through the use of light beams directed between the inputs and outputs. An input device is illuminated by a source array such that the selected output depends upon the angle of incidence of the illuminating beam.

Description

OPTICAL CROSSBAR SWITCH

The present invention relates to a switch for connecting one or more data inputs to one or more outputs, in which the interconnections are made optically, through the use of light beams directed between the inputs and outputs.

A basic switch well known in communications theory for connecting a plurality of N inputs to a plurality of N outputs is the crossbar switch, as illustrated schematically in Figure 1 for an example where N = 4. The crossbar switch comprises an array 2 of crosspoints CP.. .. CP.., which represent on/off switches; with four inputs INPUT .. INPUT on input lines 4, and four outputs OUTPUT- .. OUTPUT on output lines 6. In general, for a crossbar switch having N inputs and N outputs the number of crosspoints required to connect any one of the N inputs to any one of the N outputs is N 2, ie here there are sixteen crosspoints. In this example each of the input lines 4 is connected to four crosspoints and the crossbar switch therefore has a fan-out of four.

Similarly each of the output lines 4 is connected to four crosspoints and the crossbar switch has a fan-in of four.

In general, a crossbar switch having N inputs and N outputs has a fan-out of N and a fan-in of N. The crossbar switch is a single-stage type, since there is only a single interconnection in^'the path between any particular input and output. The crossbar switch is also known as a non-blocking crossbar switch, since provided a given input or output is not already in use a connection can always be made between the given input and any output. In the example of Figure 1 INPUT is connected to OUTPUT by crosspoint CP- _ , INPUT is connected to OUTPUT- by crosspoint CP_- , INPUT is connected to OUTPUT- by crosspoint CP.-, and INPUT is connected to OUTPUT by crosspoint CP .

Although such a switch has been implemented electronically, the requirement for N 2 crosspoints becomes a problem as N becomes large since it becomes difficult from practical considerations to implement due to its physical size. Against this, the non-blocking nature of the switch gives the advantage that minimal control and arbitration algorithms, which resolve connection contentions where for example two or more inputs attempt to connect to a single output, are required.

It has also been proposed to provide electronic interconnection between N inputs and N outputs through a multistage network which has a lower fan-out per stage than the fan-out of N of the crossbar switch and therefore requires a much reduced number of crosspoints. Although this reduced number of crosspoints reduces the physical size of the switch such that it can be electronically implemented relatively easily from a practical point of view, the network requires complex arbitration control.

It has previously been proposed in British Patent No 2243967 to implement the crossbar switch optically, whereby the connections are defined by optical paths in free space. Such a device reduces the practical problems associated with implementing a large number of crosspoints electronically. Such a switch is shown in Figure 2. Referring to Figure 2, the optical crossbar switch comprises a linear vertical array 8 comprising N (where here N = 4) input illumination sources arranged at an input plane, a linear horizontal array 10 comprising N

(four) output detectors in an output plane, and a shutter matrix 12 comprising N 2 (sixteen) shutters which constitute the crosspoints arranged as an N by N (four by four) array. The shutter matrix 12 is a spatial light modulator, as will be described further hereinbelow. A first cylindrical lens (not shown) is positioned between the linear array 8 and the shutter matrix 12, and a second such lens is positioned between shutter matrix 12 and linear array 10. By means of these lenses each input source in the linear vertical array projects light onto each of the shutters in the corresponding row of the shutter matrix. In a similar manner, each output detector in the linear horizontal array may receive light through each of the shutters in the corresponding column of the shutter matrix. The shutter matrix 12 may be a non-emitting device such as a pixellated spatial light modulator (SLM) .

The shutter matrix is controlled by electronic means (not shown) such that a connection can be made between any one of the four input sources of the linear vertical array 8 and any one of the four output detectors of the linear horizontal array 10. In the example of Figure 2 the shutter 16 of the shutter matrix 12 is open such that the input source 14 is connected to the output detector 18. It will be appreciated that as with the crosspoints of the crossbar switch of Figure 1, each of the shutters of the shutter matrix is associated with a unique path between an input and an output.

Such switches provide a form of interconnection in which not only is free-space optics used to greatly increase the connectivity between interfaces to avoid the bandwidth and pin-out limitations associated with electronics type switches, but in which the actual switching operations occur in free-space optics. In this way the optical interconnections are both "massively parallel" and reconfigurable. Such reconfigurable optical interconnections can be advantageously utilised in a variety, of applications, as diverse as telecommunications switches, processor interconnections and parallel computing, opto-electronic implementations of neural networks, image processing (e.g. motion compensation) and arithmetic processing (e.g. matrix manipulation) .

With this optical crossbar switch arrangement particular difficulties arise with the optics used for the fan-out and fan-in. As will be apparent from Figure 2 the first cylindrical lens positioned between the linear vertical array 8 and the shutter matrix 12 requires a large numerical aperture in the horizontal direction and a minimal numerical aperture in the vertical direction. Conversely, the second cylindrical lens, positioned between the shutter matrix 12 and the linear horizontal array 10, requires a large numerical aperture in the vertical direction and a minimal numerical aperture in the horizontal direction. This requirement to have a high numerical aperture in one direction and a very low numerical aperture in another direction leads to an inefficient use of optics. Moreover, there is a tendency for the image of a particular input not to be replicated precisely on each row of the shutter matrix but to be somewhat smeared which can create problems for the accurate shuttering at a particular shutter position.

In an attempt to overcome this, it has been proposed to replace the cylindrical lens arrangement with a Fourier Plane Array Generator (FPAG) . Such a proposal is discussed in A G Kirk, W A Crossland, and T J Hall, 'A compact scalable free space optical crossbar', Proc. Third International Conference on Holographic Systems, Components and Applications, pages 137-141, Edinburgh UK, 16-18 Sept 1991, IEE London, Cong Publication No. 342. This enables an array of images to be generated from a single light source, the images being discrete and not smeared. The principle of operation of the FPAG is illustrated in Figure 3. Figure 3 shows a source of light 20 which is converted into a discrete array of light sources 24 by means of a Fourier grating or hologram 22. A first lens (not shown) is positioned between the source of light 20 and the grating 22, and a second lens (not shown) is positioned between the grating 22 and the discrete array 24. It will be appreciated that the optical requirements of the fan-out arrangement of Figure 3 are improved over the fan-out arrangement of Figure 2 since light spreads out symmetrically in two dimensions and therefore the lens system requires the same numerical aperture in both the horizontal and the vertical direction, whilst the "discrete images of the array 24 can be accurately centred on the respective shutters of the shutter matrix.

Where instead of a single discrete input the input itself comprises an array, the array itself is replicated, as illustrated in Figure 4. This shows an input array 30 comprising four sources of light, a shutter matrix 32 comprising sixteen shutters arranged as a four by four array, and an array 34 of light detectors. An FPAG is positioned between the array 30 of light sources and the shutter matrix 32, and a set of fan-in optics positioned between the shutter matrix 32 and the array 34 of output detectors. Here, both the fan-out and the fan-in of the optical crossbar switch are improved, relaxing the numerical aperture requirements of the optics.

As well as the switching or connection of inputs which are discrete input sources the inventors have also sought to provide connections of inputs where each input is an input "image" made up of a number of data bits or pixels representing a data block, and this image is then connected en bloc to an output capable of resolving the image. Such an arrangement is discussed in pending British Patent Publication No. 2269296. This image I is still passed through only a single shutter in the shutter matrix, but it is found that because the image generally has an increased extent compared to a discrete point input the size of that shutter must be increased to accommodate the image, and this also increases the overall physical size of the shutter matrix.

It has been considered to utilise an FPAG arrangement for achieving fan-out, positioned between the input array and the shutter matrix, additionally providing a degree of demagnification of the image. By reducing the size of the image in this way the extent to which the size of the shutter matrix must be increased to accommodate the image is limited, but there is still a significant increase in the shutter matrix size, depending on the image size. In such- an optical crossbar switch the optical apparatus for achieving fan-in, positioned between the shutter matrix and the output array, will be required to incorporate a magnification factor to compensate for the demagnification on the fan-out. Such a demagnification system also requires more complex lens structures on both the fan-out and fan-in which may also result in degradation of the image between the input and output.

The present invention seeks to provide a switch which overcomes the problems identified above.

According to the present invention there is provided an optical interconnection switch for optical connection of one or more inputs to one or more outputs, comprising: a) an input device comprising an array of elements on which input data is optically displayed; b) an output device comprising an array of optical detector elements; and c) means for illuminating the input device in one of a plurality of illumination patterns to give a resultant illumination pattern modified by the displayed input data and detected by the output device, said illumination means being adapted to illuminate selected input elements with a beam or beams incident thereon at one or more of a plurality of possible angles of incidence according to the output detector element or elements to which that input is to be optically connected.

Preferably the input data displayed on each element of the input device is a single data bit or an array of data bits spatially arranged in a block, and each optical detector element of the output device is either a single detector or an array of detectors arranged in a block

This arrangement represents a radical departure from the prior art arrangements described above, in that instead of providing a routing arrangement between the input and output planes, the routing is effectively performed before presentation of the input data by means of providing illuminating beams at a plurality of angles. The input data is used to modify the illuminating beams and is thus projected to the output or outputs to which each input is to be optically connected. Moreover, the switching of data blocks can be readily effected. Because the routing pattern between inputs and outputs is controlled by the illumination pattern, and the data image is not required to illuminate any shutter type device, the size problems associated with shutter matrices used for handling images are avoided. So too are the image degradition problems associated with the FPAG arrangement described above. Use of control of the angle of incidence of illumination of the displayed input data provides an extremely simple effective means of providing routing control, which can be readily utilised to provide N to N routing even for large N.

According to a preferred embodiment of the invention the illumination means comprise an optical assembly having a source or sources of illumination, means for generating a plurality of beams from said illumination source or sources directed onto respective input data elements at each possible angle of incidence, and shutter means disposed between said source or sources of illumination and said input device to interrupt selected beams, to thereby control the pattern of connections between inputs and outputs. The shutter means, which may be a shutter matrix device embodied as a spatial light modulator, is required to shutter the illumination beams for the input device, rather than shuttering the data-carrying beams themselves, as occurs with the prior art devices discussed above. Preferably, the illumination means further comprise an assembly of lenses which direct the beams through optical paths in which at a point in each path each beam is substantially focused, said shutter means associated with that beam being disposed in the region of this point. This facilitates accurate shuttering and allows for a more compact shutter device.

The means for generating the plurality of beams may comprise a fixed fan-out hologram, and the lens assembly may comprise a Fourier lens which focuses said beams, followed by a lens array directing said beams onto respective elements of the input device. Alternatively, the means for generating the plurality of beams may comprise a fixed fan-out hologram array, having holograms associated with each element of the input device, and the lens assembly may comprise a first array of lenses which focuses said beams, followed by a second lens array, with each array having a lens associated with each element of the input device. This arrangement is particularly advantageous in that each hologram of the array is required for fan-out to only N points, followed by an array of only N lenses.

In an alternative arrangement for illuminating the imput device, the illumination means comprise a source of illumination, and a dynamic hologram array, which is generated by a computer and which produces from the illumination source a plurality of beams incident on the input device at selected positions and at selected angles of incidence. Here the input device is located immediately adjacent the dynamic hologram array.

In a still further arrangement for illuminating the input- device, the illumination means comprise an array of light emitting devices, such as light emitting diodes, and a lens array to direct the emitted light onto the elements of the input device. Each diode provides a beam directed onto one element of the input device at one particular angle of incidence.

Although appropriate design could ensure that the resultant pattern resulting from the illumination incident on the input device as modified by the displayed data is received directly on the output device, preferably there is provided an arrangement of lenses between the input device and output device. This allows for appropriate optical routing of the resultant pattern to allow maximum inaturisation of the various components. Preferably, the lens arrangement comprises a single long-focal length lens followed by a lens array in which there is provided a lens associated with each output detector.

The input device may be a transmissive-type spatial light modulator (SLM) in which case the incident illumination is allowed to pass through a "light" region of the displayed data, and is blocked by a "dark" region, whereby the illumination is modified by the input data to give a resultant transmitted pattern. As an alternative, the device may be a reflective-type SLM in which appropriate beam splitters are used, the resultant pattern being that reflected from the SLM. Use of a reflective-type SLM allows folding back of the beam paths, through the optics so that the number of optical components can be reduced, and the overall switch dimensions reduced accordingly.

In a further embodiment additional optics may be provided to correct errors in registration between beams incident on a particular output detector. These registration errors arise as a result of the spatial separation of the input data images, such that the further an input is from the optic axis the greater its misalignment on the output plane. In one form there is provided registration compensation optics disposed between said input device and said illumination means or between said input device and said lens arrangement. Alternatively, there is provided a registration compensation phase grating between said input device and said illumination means or between said input device and said lens arrangement.

For a better understanding of the present invention and to show how the same may be carried into effect, reference will now be made by way of example to Figures 5 to 13 of the accompanying drawings.

In the drawings:

Figure 1 is a schematic diagram of the architecture of a basic crossbar switch,

Figure 2 is a schematic diagram illustrating how the basic crossbar switch of Figure 1 may be implemented in free-space optics,

Figure 3 illustrates how a two-dimensional array of light points may be generated from a single input light point using a Fourier phase array generator,

Figure 4 illustrates how the principle of the Fourier phase array generator may be extended to an array of inputs in an optical crossbar switch,

Figure 5 illustrates how the principle of an optical crossbar switch may be adopted in accordance with the present invention,

Figures 6 to 9 illustrate different arrangements for illumination of the input plane of an optical crossbar switch according to the present invention,

Figure 10 illustrates an embodiment of an optical crossbar switch according to the present invention,

Figure 11 illustrates a development of the optical crossbar switch of Figure 10,

Figure 12 illustrates schematically the switching of data blocks, Figure 13 illustrates schematically an application of the optical crossbar switch of the present invention in a high-level packet switch architecture.

Turning to Figure 5 of the drawings, the basic principle of the switch according to the invention is described, before the various specific means for implementing the invention are described in reference to Figures 6 to 11.

As shown in Figure 5, an input array 36 is provided at an input plane, at which input data from input data lines is presented or displayed optically as an array, for example on a spatial light modulator which receives the data as electronic signals and displays these optically. The arrangement shows four input elements I ... I . An output array 38 is an array of optical detector elements, 0....0. which detect incident optical beams and pass these to respective output lines as electronic (or possibly optical) signals, constituting the output of the switch. In a radical departure from the prior art switches as described above, connection of an input to an output is effected by illuminating each input element with separate non-data carrying light beams incident at particular controllable angles. The illumination beams are modulated by the input data such that the input data is projected towards the ouput array. The routing pattern of connections between the input and output elements is controlled by both spatial and angular control of the illumination incident on the elements within the input array, as now described. The arrangement for illumination of the input array 36 comprises an array 40 of light sources S. 1...S4. and a device 42 for control of the illumination reaching the input array 36, such as a shutter matrix. An optical arrangement between the source array 40 and the shutter matrix 42 (such as a Fourier plane array generator or FPAG) is utilied to replicate the light sources S. .. S over the shutter matrix 42. An additional optical system is positioned between the shutter matrix 42 and the input array 36, such that each of the four shutters in the shutter matrix 42 onto which the light source S is projected will, when the respective shutter is open, illuminate the input element I.. The one of the four output detector elements 0....0 onto which the image I is projected will depend upon which of the four shutters associated with the light source S is open. This selection of the connection of input I onto one of the four output detectors occurs since each of the four shutters associated with the light source S- will pass beams having a different angle of incidence on the input array 36. In a similar manner each of the inputs I , I , I can be projected onto one of four output detectors depending upon which of the four shutters associated with each of their respective light sources S_, S , S. is open. Thus the projection of an input onto a selected one of a plurality of output detectors is determined by selecting the angle at which the illumination source is incident on the input array. Thus, the routing pattern through the optical crossbar switch (and hence the reconfiguration of the switch) is a function of the spatial and angular criteria of the illumination of the input array.

Although the arrangement for illuminating the input array 36 employing a source array 40 and shutter matrix 42 could be utilised, more preferred arrangements for providing the illumination of the input array are now described in detail with reference to Figures 6 to 9. Each of the examples shown in Figures 6 to 9 relate to an optical crossbar switch having only two inputs, but it will be apparent to a person skilled in the art that these examples may be extended to optical crossbar switches having a larger number of inputs.

Referring to Figure 6, there is illustrated a first illumination scheme which comprises an array of emittive illumination sources disposed in the focal plane of a lens array. In the illustrated embodiment there is an array 50 of four emittive illumination sources 52..58 such as light emitting diodes and an array of two identical lenses 60 and 62. The four illumination sources 52..58 are switchably controllable such that they effectively replicate the function of the shutter matrix 42 of Figure 5. The array 50 is positioned such that the illumination sources 52..58 lie in a plane which passes through the focal points of both the lenses 60 and 62. The illumination sources 52 and 54 are preferably positioned substantially equidistant from the focal point of the lens 60. The lens 60 forms two collimated beams 66 and 68 from the illumination sources 52 and 54, which cross and then diverge. Similarly, from the illumination sources 56 and 58 the lens 62 forms two collimated beams 70 and 72 which cross and then diverge away from each other. At a distance from the array of lenses 60,62 the beams 66 and 68 coincide, whilst the beams 70 and 72 will coincide. The plane where this coincidance occurs is utilised as the illumination plane 64, and it is in this illumination plane 64 where the two inputs of the input array (not shown) are located. One input of the input array is arranged at the region 74 where the collimated beams 66 and 68 coincide and the other input of the input array is arranged at the point 76 on the illumination plane where the collimated beams 70 and 72 coincide.

For N to N routing (ie where each input is connected to a single output only) one of the two illumination sources 52 and 54 associated with each lens is emitting at any one instant. As will be apparent from Figure 6, the input positioned at region 74 on the illumination plane will be projected in one of two directions according to which of the two illumination sources 52 and 54 is emitting. Similarly, the input positioned at point 76 on the illumination plane will be projected in one of two directions according to which of the two illumination sources 56 and 58 is emmitting. In this way, the spatial separation of the four illumination sources 52..58 is transformed into an angular separation (or multiplexing) of the collimated beams 66..72. Although the illumination sources 52 .. 58 are shown arranged in a linear array, in general (for larger N) these would be arranged in a 2-dimensional array, with a corresponding 2-dimensional array of lenses. In the general case of N inputs, there are provided N lenses, and N 2 emitters.

Referring now to Figure 7, an illumination arrangement using holograms is presented. The arrangement comprises a phase modulating device 80 on which which can be displayed a set of reconfigurable diftractive phase holograms, in this example two holograms 90 and 92, but in the general case of N inputs and outputs, a set of N holograms. These are pre-generated by a computer.

The generated and displayed holograms 90 and 92 are configured such that a blanket of incident illumination 78 either interrupts the illumination or creates beams diffracted away from the optic axis through some angle according to the angular multiplexing required. In this example, the hologram 90 will be configured in one of two configurations such that the incident light may be interrupted, or deflected in one of two directions either as a collimated beam 82 or as a collimated beam 84. The hologram 92 will also be configured in one of two configurations such that incident light may be interrupted or may be deflected in one of either two directions as a collimated beam 86 or as a collimated beam 88.

It will be apparent from Figure 7 that the point at which the collimated beams 82 and 84 coincide is their point of origin, and therefore the illumination plane at which the input array is to be located is immediately adjacent the hologram array, on the opposite side to the illumination source. One of the inputs is positioned immediately adjacent to the hologram 90 of the array. Similarly, the other one of the inputs is positioned adjacent to the hologram 92. The input provided at the hologram 90 may be projected in one of two directions depending on the configuration of the hologram 90, and the input at the hologram 92 may be projected in one of two directions depending on the configuration of the hologram 92.

As the illumination scheme of Figure 7 has-its illumination plane (i.e. the plane in which the inputs are placed) directly behind the array 80, it is possible to combine, onto a single modulating device, the phase modulation property of the holograms of the array (which control the directions of the collimated beams) with an amplitude modulating function representing the input data.

Figure 8 illustrates a third arrangement for generating the angularly controllable illumination, comprising a fixed fan-out hologram 94 which is illuminated by a

2 blanket of illumination 78, which generates four (N in the general case) diverging beams brought to four points

98..104 in a plane 106 termed the selection plane by a

Fourier lens 96. The fixed fan-out hologram may be carried on a conventional etched glass plate.

The selection plane passes through the focal point of Fourier lens 96. It will be appreciated that the four points 98..104 are directly equivalent to the four illumination sources 52..58 described hereinabove with reference to Figure 6, such that a similar lens array 60,62 can be utilised to produce the collimated beams which coincide at the illumination plane 64. Where each of the light sources 52..58 of the Figure 6 arrangement can be individually controlled to emit or not, here the beams are permanently directed at the points 98-104. It is here therefore necessary to place a device in the selection plane 106 which can 'block' the illumination at each point 98..104, such as a shutter matrix, embodied as a spatial light modulating device. Such a shutter matrix has a shutter for each point 98..104. The shutters of the shutter matrix are electronically controlled through means well known in the art to achieve the desired illumination of the input array to thereby achieve the desired routing pattern between input and -output.

Figure 9 illustrates a fourth embodiment of an arrangement for generating the angularly controllable illumination. This arrangement is similar to that of Figure 8 except that the fixed fan-out hologram 94 and the Fourier lens 96 of Figure 8 are replaced by a fixed fan-out hologram array 108 and a lens array comprising two (N in the general case) identical lenses 110 and 112. The fixed fan-out hologram array comprises two (N in the general case) holograms 114 and 116. The hologram 114 generates two (N in the general case) collimated diverging beams which are incident on the lens 110 which focusses these at points 98 and 100 in the focal plane of lens 110. Similarly, the hologram 116 generates two collimated beams which are incident on the lens 112 focused at points 102 and 104 in the focal plane of lens 112. The selection plane 106 is positioned such that it passes through the focal points of both the lenses 110 and 112. As was described above with reference to Figure 8, a device such as a shutter matrix is placed in the focal plane of the lenses 110 and 112 which constitutes the selection plane.

The arrangement of Figure 9 has the advantage over that of Figure 8 in that each hologram 114 and 116 fans out to only two points (in the general case each of N holograms fans out to N points) , followed by the array of two (in the general case N) lenses to complete the fan-out focused to four points (in the general case N 2 points) in the selection plane. Conversely the apparatus of Figure 8 has a single hologram, regardless of the number of inputs, required to fan out as four (in the general case N 2) points; as N increases, the high degree of fan-out required can create problems of beam uniformity and aberration problems.

The apparatus of Figure 9 also has the advantage that if the focal length of the lenses 110 and 112 are the same as the focal length of the lenses 60 and 62, the optics are symmetrical about the selection plane. This symmetry allows for the optical axis to be folded back on itself by placing a reflective shutter device (such as a reflective type SLM) in the selection plane 106 and placing a beam splitter in front of a single, rather than double, lens array. In this manner the number of components required is significantly reduced.

In the arrangements discussed with respect to Figures 6 to 9, it can be seen that the illumination plane 64 has a significant amount of surface area which is not used. The unused surface area is the area between the points at which the inputs are placed, and is termed 'dead-space'. It is generally desirable to reduce the amount of dead space in the illumination plane to a minimum, and in this way the inputs can be positioned closer together and therefore the device which holds the inputs can be reduced in size. The amount of dead-space is determined by the aperturing effect of the lens array comprising the lenses 60 and 62 on the points in the selection plane (Figures 8 and 9) or emitters (Figure 6) . One solution to this problem is to rearrange the N 2 points m the selection or emitter plane into closely packed groupings of N points, each of the groupings being close to the optical axis of one of the lenses 60 or 62. The dead-space will thereby be shifted from the illumination or input plane to the selection plane or emitter plane, where it is less likely to be a problem. In addition, this 'packing*" of points closer to the optic axis of the lenses 60 and 62 relaxes the numerical aperture requirements of the lens array comprising the lenses 60 and 62.

A second solution to this problem, which can be applied in the schemes illustrated in Figures 8 and 9, is to increase the focal length of the lens array comprising the lenses 60 and 62 to be greater than the focal length of the lens 96 in Figure 8, or the lens array comprising the lenses 110 and 112 in Figure 9. This method will also have the secondary effect of shifting the illumination or input plane closer to the lens array comprising the lenses 60 and 62.

With reference to Figure 10, a preferred embodiment of an optical crossbar switch in accordance with the invention based on the angular illumination scheme of Figure 9 will be described, where like reference numerals are used in Figure 10 for the identical parts already shown in Figure 9. The apparatus includes as its central portion the arrangement of Figure 9 comprising fixed fan-out hologram 108, a first lens array 110,112, a shutter matrix 146 at the selection plane and a second lens array 60,62.

The shutter matrix 146 is an amplitude type pixellated spatial light modulator. This may use opto-electronic integrated circuits fabricated in silicon VLSI technology and integrated with Ferro-electric liquid crystals, as described in the article entitled "Active backplane spatial light modulators using smectic liquid crystals" of W.A. Crossland et al in Proc SPIE, Liquid Crystal Materials, Devices and Applications 1665:114-127, 1992. Such a switch is particularly well suited to a communication switch application in terms of speed, reliability and ease of interfacing to peripheral electronics. In the illustrated embodiment the shutter matrix has four shutters 138..144 onto which the beams are focused by the lenses 110 and 112 as discussed hereinabove. The input data for the switch is displayed as an input array on a further pixellated SLM device positioned at the illumination plane 64. In the illustrated embodiment there are two inputs, A and B.

To generate the input illumination for the FPAG fan-out hologram array 108 there is provided a collimated beam, which provides the blanket illumination 78 incident on this fixed fan-out hologram array 108. In this embodiment, such an illumination is provided by means of a laser 120 which generates a narrow beam of light into a spatial filter 124. The spatial filter disperses this narrow beam into a broad diverging beam of light 128 which is collimated by lens 126 into a parallel beam. As discussed above in relation to Figure 9, the fixed fan-out hologram array has two holograms 114 and 116 (N in the general case) . The hologram 114 generates two collimated beams of light 130, 132 which are incident on the lens 110 and focused into convergent beams which form points at the points 138,140 on the selection plane 106. Assuming the shutter of the shutter matrix 146 associated with the point 138 is open, the light will diverge from the point 138, being focused by the lens 60 into a collimated beam 156. The collimated beam 156 illuminates the image A on the data input array in the illumination plane as discussed in more detail below. Assuming the shutter of the shutter matrix 146 associated with the light point 140, is open, the light will diverge from the point 140 being focused by the lens 60 into a collimated beam 158. The collimated beam 158 illuminates the image A on the data input array in the illumination plane at a different angle to the beam 156.

In a similar manner, the hologram 116 generates collimated beams 134,1"36 focused into convergent beams which form points at 142,144 in the plane 106. If the shutter of the shutter matrix 146 associated with the point 142 is open, the light diverges from the point 142, and is then focused by the lens 62 into a collimated beam 162. The collimated beam 162 illuminates the image B on the data input array in the illumination plane. If the shutter of the shutter matrix 146 associated with the light point 144 is open, the light diverges from the point 144 and is then focused by the lens 62 into a collimated beam 160. The collimated beam 160 illuminates the image B on the input array at the illumination plane at a different angle to the beam 162. Thus, depending upon which of the shutters are "open" input elements A and B may each be illuminated in two (N in the general case) different angles. The output of the switch comprises a further opto-electronic device 163 comprising a plurality of optical detectors arranged at an output plane. This may be a fibre array, light-sensitive photodiode array or charge-coupled device array interfacing back to the electronics. In the illustrated embodiment two detectors or detector elements 165,167 are required (N in the general case) . In order to conveniently bring the beams 156..162 onto one or other of the output detectors an output lens arrangement is provided comprising a lens 150 and lens array having two (N in the general case) lenses 152,154.

The lens 150 is positioned beyond the illumination plane 64 such that the collimated beams 158 and 160 are focused at a point 164 in the focal plane of the lens 150, which is also in the focal plane of the lens 152. The lens 152 produces collimated beams 168,172 which are detected by the detector element 165. Similarly the collimated beams 156 and 162 are focused through a point 166 in the focal planes of the lenses 15O and 154 into collimated beams 170 and 174 to be detected by the second detector element 167.

The routing pattern between inputs A,B at the input array 148 and outputs 165 and 167 is then controlled by the selective opening of shutters at the shutter matrix 146. For example, the input A can be projected onto either one of the two output detectors 165,167 by opening one of the two shutters associated with the points 138 and 140, and the image B can be projected onto either one of the two output detectors by selecting one of the two shutters associated with the points 142 and 144.

Thus the angularly multiplexed illumination generated and controlled separately for each input element, projects that input element to the output element to which it is to be connected.

It should be noted that in practice, for larger N, the angular multiplexing of the optical crossbar switch will occur in two dimensions rather than the one dimension illustrated with reference to Figures 6 to 10. Hence the illumination plane and the output plane will be filled with two dimensional arrays of data or images.

It will be noted that the image A represented in the collimated beams 168 and 170 is inverted with respect to the input image in the illumination plane 64. Similarly the image B represented in the collimated beams 172 and 174 is inverted which respect to its input image in the illumination plane 64.

Although the specific embodiment of Figure 10 utilises an angular illumination scheme such as that described with reference to Figure 9, it will be appreciated that any of the angular illumination schemes of Figures 6 to 8 could be utilised.

Referring to Figure 10, it can be seen that in the arrangement illustrated, the lens 150 generates two image points 164,166 (N image points in the general case) in its focal plane. The generation of N image points leads to some degree of spatial registration error between the final collimated outputs, 168,172 associated with lens 152 and outputs 170,174 associated with lens 154. This spatial registration error may cause difficulties for detection of the collimated beam, detector 165 being required to detect both beams 168 and 172. This effect is greatly exaggerated in Figure 10, and in the usual case where f_ (the focal length of lens 150) is very much greater than f- (the focal length of lenses 152,154) the effect is small so that any offset is tolerated by the detectors. The effect is caused by the spatial separation of the input image data A,B on the input plane 148, such that the further an input is from the optic axis, the greater the misalignment of its projection at the output plane. Figure 11 illustrates a modification of the arrangement of Figure 10 which uses additional optics 180 termed registration compensation optics, positioned between the illumination or input plane 64 and the lens 150 (or alternatively between plane 64 and lenses 60,62) . These can take the form of N lenses or phase gratings to cancel the input spatial separation. As can be seen from

Figure 11, the introduction of the optics 180 causes there to be four (in the general case N 2) image points in the focal plane of the lens 150. The two collimated beams 168 and 172 now coincide exactly at the output plane 182.

Similarly the two collimated beams 170 and 174 coincide exactly at the output plane 182.

As an alternative to this registration compensation scheme means could be provided to rotate en-masse each group of N possible beam paths associated with a single input toward the optical axis through an angle proportional to the distance of the input from the axis. An alternative solution would be to modify the generation of the

2 angularly controlled illumination by arranging the N points in the selection plane in an appropriate non-regular manner (Figs 6 and 9) or by incorporating the appropriate compensation directly in the dynamic hologram design (Figure 7 embodiment) .

The optical crossbar switch according to the present invention finds applications in a wide variety of fields as mentioned above, but can be employed with particular advantage in communications applications owing to the possibility of providing massively parallel interconnections, to the high switching speed, and to the case of interfacing to peripheral electronics. One specific application is in so called packet switching communications using an asynchronous transfer mode (ATM) protocol, and this is discussed with reference to Figures 12 and 13.

As the optical switch allows for data to be transferred in parallel, serial input data in the time domain may be converted to parallel data in the space domain such that the optical switch can operate at a speed much lower than that at which the serial data is inputted. Such a technique would require some form of buffering in the input plane. In this way data entering the optical space switch in a serial form may be buffered and presented to the opto-electronic input plane as two-dimensional blocks of data. This principle is illustrated with respect to Figure 12.

Here, data arranged serially in packets (or ATM cells) on particular input fibres is communicated via the switch to particular output fibres. Figure 12 shows as an example, four optical fibres 252..256 each transmitting a serial data signal. Electronic buffering converts the packets of serial data into parallel form displayed on each of the input elements as a data block or image. Here, the data is displayed as four blocks 260..266 arranged in a grid. Each one of the blocks 260..266 represents an accumulation of data bits from a respective one of the input optical fibres 252.-256. Each collection of data bits or block in each input element can be considered as an image and is transferred in the switch en bloc, and to an output element or elements of the output array. The switching entity is therefore whole blocks of data rather than individual data bits. Because the data is switched as a parallel transfer of blocks, the switch is required to operate at a speed very much lower than that at which the data is received on a particular line i.e. the line bit rate.

If a contention arises during routing i.e. when more than one data packet (or ATM data cell) is destined for the same output, the packet in contention must be delayed until the next routing transfer period. Such a delay can be achieved by electronic buffering at the input, optical buffering at the input or switch feedback. With electronic buffering the packet in contention is delayed by the input electronics and not displayed on the optical crossbar switch input plane until the contention has been cleared. With optical buffering the packet in contention is displayed and held on the input plane but is not routed to the output plane until the contention has been cleared. With switch feedback the packet in contention is optically routed to a reserved external output slot which simply feeds the packet back to the switch input to be dealt with in the next routing transfer as a normal input packet.

Referring to Figure 13, a high level block diagram of a packet switch architecture employing the optical space switch of the present inveniton is shown by way of example.

Serial input data is transmitted on each of the N input optical fibres I ..I which form inputs to an input buffering and formatting block 208. The input buffering and formatting block 208 performs the conversion to parallel data blocks to be displayed on an input plane 202 of an optical space switch 200, as described hereinabove with reference to Figures 10, 11 and 12. The data blocks are transmitted from the input buffering and formatting block 208 to the input plane via connections 218. The input buffering and formatting block 208 also transmits the control and address information associated with the incoming data to a packet header processing block 212 via connections 234. The data blocks on the optical input plane 202 are spatially arranged in blocks as described with reference to Figure 12 and illuminated by means of a routing and illumination block 204. The routing and illumination block 204 generates the required spatially and angularly multiplexed illumination for the input plane via connections 228 as described hereinabove with reference to Figures 5 to 9. This illumination is used to project the input data on the input plane 202 to an output plane 206 of the optical crossbar via connections 226 as described hereinabove with reference to Figures 10 and 11. The routing and illumination block 204 is controlled via connections 230 by a control and arbitration block 214. The control and arbitration block 214 receives the control and address information from the packet header processing block 220, and accordingly controls the routing and illumination block 204. The control and arbitration block 214 also provides a packet header replacement block 216 with address and control information via connections 222.

An output buffering and decoding block 210 receives the data signals on the output plane 206 via connections 224 together with output signals from the packet header replacement block 216 via the connections 232. The output buffering and decoding block 210 then reconstructs the data blocks into serial packet form with appropriate control and address headers, and transmits serial data packets on output optical fibres 0...0... The header associated with a data packet may comprise not only an address identifying which one of the N optical fibres 0-..0„ the packet is to be routed to, but also addresses of other locations to which the packet is to be routed as it progresses from its source to its destination, as well as error checking bits, framing bits and control bits such as packet identification bits.

The communications application of the switch of the present invention described, provides for high capacity time-space-time switching with the input and output buffering performed electronically and having additional interfaces for routing control, the actual routing being performed in the spatial domain.

The high capacity of such a switch is provided for by the angular encryption of positionally encrypted input data. This dual multiplexing technique can be used in any parallel processing application where a high degree of interconnection is required. As a result of the non-interaction of crossing illumination beams, a very high degree of parallelism can be provided, allowing a degree of miniaturisation which is impossible to achieve in an electronic system. In the communications switch this parallelism can be utilised to route parallel blocks of data which are received serially, thereby allowing the switch to operate internally at a reduced data rate and allowing more time for reconfiguration of the switch without a reduction in the external serial bit rates.

The architecture of the optical crossbar swtich proposed by the present invention whereby the routing pattern is optically encrypted before performing the presentation of the input data in the input plane, requires reduced optical resolution requirements as compared to previous architectures having a high fan-out and fan-in, in particular where data blocks or images are being switched as individual units. The arrangement also ensures that all optical components within the architecture scale to the number of inputs N rather than the number of interconnectors N 2.

Moreover, as the input plane of the switch on which the input signals are displayed is optically separated from the actual optical fibres carrying the input serial data, the size and scalability of the switch is not limited by the optical power requirements of the optical transmission fibres.

Claims

1. An optical interconnection switch for optical connection of one or more inputs to one or more outputs, comprising: a) an input device comprising an array of elements on which input data is optically displayed; b) an output device comprising an array of optical detector elements; and c) means for illuminating the input device in one of a plurality of illumination patterns to give a resultant illumination pattern modified by the displayed input data and detected by the output device, said illumination means being adapted to illuminate selected input elements with a beam or beams incident thereon at one or more of a plurality of possible angles of incidence according to the output detector element or elements to which that input is to be optically connected.

2. An optical crossbar switch according to claim 1 wherein the input data displayed on each element of the input device is a single data bit or an array of data bits spatially arranged in a block, and each optical detector element of the output device is either a single detector or an array of detectors arranged in a block.

3. An optical crossbar switch according to claim 1 or claim 2 wherein the illumination means comprises an optical assembly having a source or sources of illumination, means for generating a plurality of beams from said illumination source or sources directed onto respective input data elements at each possible angle of incidence, and shutter means disposed between said source or sources of illumination and said input device to interrupt selected beams, to thereby control the pattern of connections between inputs and outputs.

4. An optical crossbar switch according to claim 3 wherein the illumination means further comprise an assembly of lenses which direct the beams through optical paths in which at a point in each path each beam is substantially focused, said shutter means associated with that beam being disposed in the region of this point.

5. An optical crossbar switch according to claim 4 wherein the means for generating the plurality of beams comprise a fixed fan-out hologram, and the lens assembly comprises a Fourier lens which focuses said beams, followed by a lens array directing said beams onto respective elements of the input device.

6. An optical crossbar switch according to claim 4 wherein the means for generating the plurality of beams comprise a fixed fan-out hologram array, having holograms associated with each element of the input device, and wherein the lens assembly comprises a first array of lenses which focuses said beams, followed by ^"a second lens array, with each array having a lens associated with each element of the input device.

7. An optical crossbar switch according to claim 1 or claim 2 wherein the illumination means comprise a source of illumination, and a dynamic hologram array, which is generated by a computer and which produces from the illumination source a plurality of beams incident on the input device at selected positions and at selected angles of incidence.

8. An optical crossbar switch according to claim 7 wherein the input device is located immediately adjacent the dynamic hologram array.

9. An optical crossbar switch according to claim 1 or claim 2 wherein the illumination means comprise an array of light-emitting devices, and a lens array to direct the emitted light onto the elements of the input device.

10. An optical crossbar switch according to claim 9 wherein the array of light emitting devices comprises an array of light emitting diodes, each light emitting diode providing a beam directed onto one element of the input device at one particular angle of incidence.

11. An optical crossbar switch according to any preceding claim, further comprising an arrangement of lenses disposed to direct onto the output dectors said resultant illumination pattern resulting from the illumination incident on the input device as modified by the displayed input data.

12. An optical crossbar switch according to claim 11 wherein the lens arrangement comprises a single long-focal length lens followed by a lens array in which there is provided a lens associated with each output detector.

13. An optical crossbar switch according to any preceding claim wherein said input device is a pixellated spatial light modulator on which the input data is optically displayed, and each element of the device associated with a particular input for connection to a particular output or outputs comprises a plurality of pixels.

14. An optical crossbar switch according to claim 13, wherein the spatial light modulator is a transmissive type spatial light modulator.

15. An optical crossbar switch according to claim 13, wherein the spatial light modulator is a reflective type spatial light modulator.

16. An optical crossbar switch according to claim 11 further comprising registration compensation optics disposed between said input device and said illumination means or between said input device and said lens arrangement, in order to improve coincidence of the beams directed on to respective output detectors.

17. An optical crossbar switch according to claim 11, further comprising a registration compensation phase grating between said input device and said illumination means or between said input device and said lens arrangement, in order to improve coincidence of the beams directed onto respective output detectors.

18. An optical crossbar switch according to claim 3 which said shutter means comprises a spatial light modulator.

19. An optical crossbar switch substantially as hereinbefore described and as illustrated in Figures 9, and 10 or 11 optionally in combination with Figures 6, 7 or 8 of the accompanying drawings.