GB2243967A - Optical crossbar switch - Google Patents

Abstract

The switch includes an array (3) of spatial light modulators (4) comprising a crossbar matrix mask. Associated with each optical source (1) is a respective photodiode (8) at the matrix which extracts routing information from optical signals applied to the mask by the sources. This information is decoded and employed to address the SLM columns and rows to configure the matrix as appropriate for the desired routing. <IMAGE>

Description

OPTICAL CROSSBAR SWITCH This invention relates to optical crossbar switches and in particular to such switches incorporating spatial light modulators (SLMs).

An optical crossbar switch is a device for connecting N inputs to one or more of M outputs Light from a linear array of N input sources 1 (Fig. 1) and forming input vector i is spread out by means of optics (not shown) such that each source 1 illuminates a respective column 2 of an M x N array (matrix) 3 of optical shutters (modulators) 4. In the example illustrated M = N = 4. The matrix 3 is a crossbar mask matrix A (a spatial light modulator SLM) which controls the routing of inputs from sources 1 to outputs comprised by a linear array 5 of M detectors 6. The light passing through each row 7 of the SLM is collected by optics (not shown) and falls on one of the M detectors 6, thus forming output vector o. This configuration is known as a matrix vector multiplexer as the operation o = iA is performed.Typically the light travelling through the system carries a high bit rate data stream and the optical crossbar performs the function of a telephone exchange. In wFiber-Optic Crossbar switch with Broadcast Capability1 A.R. Dias et al SPIE Vol 825 p 170-177, two types of SLM technologies are referred to in this context, namely PLZT and magneto-optic.

According to the present invention there is provided an optical crossbar switch including an array of spatial light modulator elements comprising a crossbar matrix mask, and wherein associated with said spatial light modulater elements are photodetector means, serving to extract routing information from optical signals applied to the mask, and means serving to configure the matrix in response to said extracted routing information whereby to route data information carried by the optical signals appropriately.

Embodiments of the invention will now be described with reference to the accompanying drawings, in which: Fig. 2 illustrates an optical crossbar switch based on the matrix vector arrangement of Fig.

1 and including a detector array, Fig. 3 illustrates an example of an arrangement of functional sections of the switch of Fig. 2 when provided as a monolithic device, Fig. 4 indicates a cross-section through an SLM element, Figs. 5 illustrates an array of SLMS and associated FETs Figs 6 and 7 illustrate possible arrangements of optics for transmissive and reflective SLM devices respectively.

The information concerning the destination of a data stream passing through an optical crossbar switch SLM can be transmitted from the particular source along with the data. This information can be transmitted on the same wavelength as the data as in a conventional packet switched system, or in one or more different wavelengths. By adding photodetectors to a liquid crystal SLM, routing information can be extracted from the data passing through the crossbar matrix 3 which information can then be used to configure the crossbar matrix to produce the required routing.

It is thus proposed to place a linear array (row) 8 of photodetectors 9 (Fig. 2) along one side of the liquid crystal SLM matrix 3, each detector monitoring a respective source 1. This allows the optics directing light from each source 1 on to a column 2 of elements (modulator, shutters) 4 to be used to illuminate the respective photodetector 9. The outputs from the photodetectors are directed to a crossbar controller (not shown) which controls row and column addressing and causes the modulators to be "closed" (opaque) or wopenedw (transparent) as appropriate. The crossbar may form part of an array of crossbars each having associated photodetectors.

The array of input optical sources 1 may be comprised by the output ends of a plurality of polarisation maintaining optical fibres, which output ends are disposed in a linear array, laser diode sources of defined polarisations being coupled to the input ends of the fibres. The intensity of the output light can be directly modulated by controlling the power applied to the laser diodes. The SLM may be a reflective array of liquid crystal elements on the surface of a silicon integrated circuit, in which case the detector array 5 is in the same side of the SLM as the source-array. The detector array 5 may be a linear array of photodiodes, as can the detector array 8.The photodiode array 8, destination decoding circuitry 10, which employs the routing information, and the SLM matrix controller including SLM column addressing 11 and row addressing 12, may be integrated on the same silicon substrate as the SLM liquid crystal elements (Fig. 3).

The output of the arrangement may include a linear array of high numerical aperture multimode optical fibres (not shown) feeding the array of photodetectors 5. When connected by the SLM crossbar matrix, the sources and detectors are linked to form a high bandwidth communication channel. Typically the crossbar matrix may be 64 x 64 for such purposes.

The construction of the integrated circuit providing the SLM elements, the photodiode array and the decoding and addressing circuitry will now be considered. Each SLM element may be based on the structure whose cross-section is illustrated in Fig. 4.

This comprises a silicon substrate 21, an FET comprised by source and drain regions 22, 23 and a gate 24 in a silicon dioxide layer 25 which is apertured to provide electrical contact between back electrode 26 and drain 23. Another (front) electrode 27 which is transparent is provided by a coating such as of gold or indium oxide on a transparent panel 28. A liquid crystal, for example a ferro-electric chiral smectic C phase liquid crystal material 29 is disposed between the electrodes 26 and 27. The element may be operated by applying a control signal to the gate 24 such that the electrode 26 is driven to a positive or negative voltage relative to the other electrode 27 so as to switch the ferro-electric liquid crystal material between its two stable states, i.e. cause it to appear light or dark, transparent or otherwise.In the case of a reflective SLM element the back electrode is a mirror (aluminium pad) provided on the substrate which serves to reflect incident light when the liquid crystal material is transparent.

In the case of an array of such SLM elements as for a matrix A, the gates of the FETs in each row of elements are connected together, in each column of elements the sources of the FETs are connected together, and the drain of each FET is coupled to its respective reflective back electrode 26 (Fig. 5). By sequentially applying a voltage to each of the rows of FET gates, each row of elements can be addressed at a time. Only during this addressing time will voltages present on the matrix columns, less the transistor threshold voltage, be transferred to the element pads (back electrodes).

These voltages switch the liquid crystal material over the respective back electrode between its stable states. The front electrode may be held half way between the higher and lower voltages appearing on "ON" and wOFFw elements.

By applying suitable surface treatment, a uniform wbookstackw arrangement of tilted planes of molecules will be induced in the liquid crystal. In this state the ferro-electric dipole of the molecules will align anti-parallel with the electric field applied across each element. The field across each element will decay as charge leaks from the back electrodes after the row address period. However the memory associated with each of the two stable states possible in the liquid crystal layer will allow the optical properties of the layer to remain substantially constant. The layer thickness of the liquid crystal cell should be such that the rotation of the optic axis between the two stable states will switch between a reflection maximum and minimum for the polarised light at the wavelength used.

Input polarisation will be defined by the laser diodes and the orientation of the single mode polarisation maintaining optical fibre.

The integrated circuit providing the SLM elements, the photodiode array and the decoding and addressing circuitry may comprise a silicon integrated circuit manufactured by a merged bipolar (CMOS) process. This can allow the production of both npn bipolar and p and n-type MOS transistors on a p-substrate. The FETs at the SLM elements would be n-MOSFETs in this case. The photodetectors (photodiodes) may be manufactured using an n region normally used for the n-MOSFET drain/source regions as the cathode. The anode would be formed from the p-type substrate. The decoding (packet header decoding) and SLM control (row/column addressing) circuitry would employ the conventional CMOS logic gates provided by bipolar/CMOS process.With regard to the photodiodes, for a binary system, calculations suggest an incident power of 50nW would be sufficient to achieve a bit error rate of 10 9 using a 100/um detector area and a 10V reverse bias. Each photodiode is connected to an integrated transimpedance amplifier on the integrated circuit to produce a voltage output to be further processed in the decoder circuitry. Areas of silicon not required to be photosensitive should be protected by an earthed metal layer.

Thus data will be sent through the system in packets, each packet being preceded by a header indicating its destination. The photodiode associated with each source receives the data stream and is connected to a respective part of the decoder circuitry, which reads the packet header and instructs the SLM, by appropriate row and column addressing, to take the correct configuration to establish the connections required. It is not necessary for the photodiode to be capable of responding at the same speed as the data rate of the link as a slower clock rate may be used for the packet header. Indeed the slower clock rate may be used to distinguish packet header and data.

Whilst the above arrangement is based on silicon technology, alternatively III -V semiconductor technology could be employed, in which case the modulators may be III - V as well, for example, multiple quantum well devices, rather than liquid crystal devices.

The necessary optics between the sources, SLM and detectors of the overall arrangement will now be considered. Figs. 6a and b show elements for the transmissive case in a side view and plan view respectively. Five optical fibres 31 direct light from respective laser diodes 32 to a spherical lens 33, spherical lens 34 and a lenslet array 35 and thence to SLM array 36, another lenslet array 37, spherical lenses 38 and 39 to optical fibres 40 and respective photodiodes in an array 41. The fan-in optics (to the left of the SLM array 36) spread the Gaussian intensity profile of the single mode input fibre vertically and image it horizontally. The lenslet array 35 not only allows all the light to pass through the centre of each SLM element but also provides a focussed spot at the SLM to be imaged on to the output fibre by the fan-out optics (to the right of the SLM).Fan out is performed by a similar system to the fan-in optics but rotated by 0 90 , the light being imaged in the horizontal direction and focussed in the vertical. The two lenses could be combined into one holographic optical element or could be separated to form a 4f system allowing smaller numeral apertures to be used for the output.

For a reflective system the optics is as in Fig. 7 which shows a plan view. In this case a polarising beam splitter 42 is used to separate the input and output beams. In Fig. 7 reference numeral 43 refers to spherical lenses, 44 refers to cylindrical lenses and 45 refers to cylindrical lens arrays. The polarising beam splitter acts as the analyser in the system. Typically the arrangement may be operated at 850nm.

With current liquid crystal materials switching times of the order of 1 to 10 used are obtained, although this is expected to improve to less than 1 rsec with future generations of materials. Contrast ratios of up to 200:1 are presently obtained routinely, allowing crossbars of 128 x 128 to be achieved. Once again, this figure is expected to improve as the quality of the liquid crystal aligning layer improves, to yield contrast ratios of the order of 1000:1.

So far it has been assumed that there is one photodetector per column of SLM elements. This is not the only possibility and, for example, there may alternatively be several devices operating on different wavelengths. In some circumstances more than one channel of information may be associated with each SLM element. Instead of a single input fibre/source, groups of eight fibres could be used, the respective beams passing through a single SLM to be imaged on to a group of eight outputs. Such a system would allow the transmission of information one byte at a time. In such a system each element would consist of several sub-elements responding to component bits or combinations of bits within the byte.

The photodetector need not be associated only to the pattern present on the column of the crossbar it is monitoring. By combining the output of several detectors additional information may be obtained. For example one may wish to detect and prevent the condition of two or more inputs falling on one detector.

The majority of prior art optical routing systems have been multistage devices passing each input through a number of individual routing elements. In such systems the light may be converted into an electrical signal, processed and reconnected to an optical signal, which increases the time to travel through the switch. The routing elements are often in the form of 2 x 2 exchange bypass units, a large number of which are required to make a system of a useful size. The overall latency is then the sum of the delay at each stage. Whilst each stage may have a delay in the nanosecond range, the overall system may have a latency of the order of multiseconds. The optical crossbar switch of the present invention is however a single stage device whose latency is not dependent on the number of input and output channels. Thus the switch elements of the crossbar matrix may have a relatively long switching time. In most practical known multistage switching networks it is possible for a required routing between an input and an output to be blocked due to a previously established route. Blocking is not possible with the single stage crossbar network of the present invention.

Attention is directed to our GB patents 2149555B and 2149176B and GB Applications 2166256A and 2188742A which describe various aspects of the structure and operation of ferro-electric liquid crystal displays, and also to co-pending Application No 8914453.9 (Serial No ) which describes so-called smart pixels incorporating chiral smectic liquid crystal light modulators and the contents of which are incorporated herein by reference.

Claims (13)

1. An optical crossbar switch including an array of spatial light modulator elements comprising a crossbar matrix mask, and wherein associated with said spatial light modulater elements are photodetector means, serving to extract routing information from optical signals applied to the mask, and means serving to configure the matrix in response to said extracted routing information whereby to route data information carried by the optical signals appropriately.

2. A switch as claimed in claim 1 wherein the spatial light modulators comprise chiral smectic liquid crystal modulators disposed on a semiconductor substrate, said photodetector means and configuring means including electronic circuitry formed in the substrate.

3. A switch as claimed in claim 1, wherein the liquid crystal comprises chiral smectic ferro-electric C phase material.

4. A switch as claimed in claim 2 or claim 3 wherein the substrate is of silicon and the modulators and electronic circuitry are formed by VLSI techniques.

5. A switch as claimed in claim 4 and formed by means of merged bipolar/CMOS processing.

6. A switch as claimed in claim 1 and manufactured by means of group III-V semiconductor technology.

7. A switch as claimed in any one of the preceding claims wherein the routing information is transmitted at the same wavelength as the data information.

8. A switch as claimed in any one of claims 1 to 6 wherein the routing information is transmitted at a different wavelength to the data information.

9. A switch as claimed in any one of the preceding claims wherein the optical signals are produced by an array of optical sources comprising laser diodes whose outputs are directed to the matrix via respective polarisation maintaining optical fibres.

10. A switch as claimed in claim 9 wherein the routed data information is detected by an array of photodetectors.

11. A switch as claimed in claim 10 wherein the routed data information is collected by respective high numerical aperture multimode optical fibres and directed thereby to said array of photodetectors.

12. A switch as claimed in claim 10 or claim 11 wherein said array of sources and said array of photodetectors are disposed on the same side of the matrix, the spatial light modulator elements comprising reflective elements.

13. An optical crossbar switch substantially as herein described with reference to Fig 2 or Fig 3 with or without reference to any one of Figs 4 to 7.