The rapid increase in importance of data communication has led to a corresponding increase in demand for bandwidth in communication systems. Wavelength division multiplexing (WDM) has been developed to expand the capacity of new and existing optical fibre systems, in which multiple wavelengths of light simultaneously transport <BR> <BR> <BR> information through a single optical fibre. Each wavelength operates as a single channel carrying a stream of data and accordingly the capacity of the fibre is multiplied by the number of WDM channels available.

It has been proposed to use polarisation techniques to increase the capacity of WDM systems but in general this approach is not favoured as polarisation maintaining fibre is too expensive and suffers high losses compared to single mode fibre. Polarisation maintaining fibre also suffers from the production, and installation, difficulty that it is difficult to join fibres so as to maintain polarisation. Additionally, any system relying on polarisation in this way is incompatible with existing single mode fibre systems and would therefore be unlikely to find general market acceptance.

In all communication networks, there is a need to connect individual channels to individual destination points such as another network or end customer. In traditional telecommunication systems, these are referred to as cross-connects and are implemented using electronics. However, due to the inherent advantages of all optical systems, there is great : interest in developing cross-connects which operate at a wavelength level using photonic network elements.

Most ; conventional optical cross-connects are essentially physically reconfigurable switches, m known optical cross-connects, the incoming light stream must generally first be demultiplexed into its constituent wavelengths, each on an individual fibre. Each individual wavelength is then directed to its destination using an optical space switch, generally composed of a large number of switch elements. The wavelengths must then be remultiplexed or recombined before continuing onto the destination fibre. For full connectivity, wavelength conversion may be required. A typical 40 channel WM system will require thousands of switch elements to cross connect all the wavelengths.

The complexity and loss of optical cross-connects is further increased by the need to provide access to parallel redundant switching planes to provide reliable operation.

Consequently, such optical cross-connects are highly complex and expensive systems and there exists a need to maximise the number of available channels and the bandwidth of each channel as cost-effectively and reliably as possible.

An alternative approach is the wavelength router. Instead of a space switch which is optically transparent to all wavelengths, a wavelength-selective optical interconnect provides alternative routes between input ports and output ports according to the wavelength of the optical signal. A wavelength-tunable optical transmitter or wavelength converter at each input port can thereby selectively access each output port, providing a cross-connect function. A suitable wavelength-selective optical interconnect can be constructed by configuring an optical combiner with a'wavelength demultiplexer. The' optical loss is then large for significant channel numbers, dominated by the loss of the combiner. A better wavelength-selective optical interconnect can be constructed by combining the functions of a wavelength multiplexer with a wavelength demultiplexer in a single component. The loss for the overall NxN function is then reduced. Suitable technologies for constructing an NxN single stage wavelength-selective optical interconnect include the array waveguide grating (AWG) (Figure la) and the free-space grating (Figure lb). Components currently available commercially with NxN function are limited to 40x40.

The present invention seeks to provide a wavelength router with an increased capacity over known wavelength routers.

According to the invention there is provided a wavelength router comprising a plurality of input ports, a plurality of output ports and a routing means for routing optical signals between the input ports and the output ports, at least one input and output port being polarisation selective and the routing means being polarisation maintaining wherein the router is adapted to separate a light signal into two polarisations at the polarisation selective input ports and the router comprises data writing means adapted to write a data signal onto a respective polarisation of each wavelength of the light signal.

In a first preferred embodiment, each input port has two transmitters connected thereto and each output port has two receivers connected thereto, the signal from each transmitter being combined using a polarisation maintaining optical coupler and the signal from the router being polarisation resolved by a polarisation beam splitter. Full connectivity is therefore provided between transmitters and receivers. Preferably, contention resolution is provided for the case when more than one transmitter requires access to the same at the same time. For M channels, M tunable lasers, M data modulators, M receivers, M/2 combiners, M/2 polarisation beam splitters and I M/2xM/2 polarisation maintaining wavelength router are required The tuning requirement of a laser diode is a function of channel width and the bandwidth of each channel. However, the tuning range of any given laser diode is limited for physical reasons. The invention advantageously permits the effective number of channels to be increased by using the polarisation of light to encode different data signals on different polarisations of the same wavelength without any commensurate increase in wavelengths. By making the routing means polarisation maintaining, any polarisation control at the receiver can be avoided, which, as such controls are typically slow, will facilitate the routers use in high capacity communication systems. The polarisation maintaining routing means or interconnect thus has a different routing characteristic for each polarisation and wavelength combination, thereby simplifying the design and manufacturing costs for the device.

Ln a second preferred embodiment, each input port has a transmitter connected therewith, which transmitter is adapted to produce two signals having the same wavelength but transversely differing polarisation. Preferably, the transmitter includes first and second modulators, each adapted to write a data signal at different polarisations. Preferably the modulators are in series with one another and a polarisation rotator is located between the first and second modulators. Alternatively, the modulators may be parallel to one another with a polarisation beam splitter located in front of the modulators and a polarisation beam combiner located after the modulators. In this latter case, a polarisation beam rotator may be included between the splitter and one modulator. Although reduced connectivity is thereby provided between transmitters and receivers, (also. preferably subject to contention resolution when more than one transmitter requires access to the same receiver at the same time), a reduced number of tunable lasers is required to obtain the capacity increase. For M channels, M/2 tunable lasers, M data modulators, M receivers, M/2 combiners, M/2 polarisation beam splitters and 1 M/2xM/2 polarisation maintaining wavelength router are required.

Exemplary embodiments of the invention will now be described in greater detail with reference to the drawings in which: Figure la schematically shows a known wavelength interconnect ; Figure lb schematically shows an alternative known wavelength interconnect; Figure 2 shows a first embodiment schematic diagram of a wavelength router according to the invention ; Figure 3 shows a second embodiment schematic diagram of a wavelength router according to the invention ; Figure 4 shows a schematic diagram of the polarised signal generator ; Figure 5 shows an alternative polarised signal generator to Figure 4 ; and Figure 6 shows an alternative arrangement for the modulator component.

Figures. la and Ib show known wavelength selective interconnects. The function of the wavelength selective interconnect is that for any input port, a choice of output ports can be made by selecting appropriate optical wavelengths. Each of N discrete wavelengths will select one of N ports. There is no intrinsic splitting loss between input and output ports.

In the arrayed waveguide grating (AWG) implementation of Figure 1 a, N input ports are combined at a mixer 100, split between an array of waveguides whose length incrementally increases before recombination and splitting at a second mixer 101 between output ports. The length differences in the waveguide between the mixers 100 and 101 correspond to phase changes for different wavelengths, resulting in different output routing.

The equivalent function is available in bulk form (see Figure 1b) where a diffraction grating provides wavelength selectivity in the routing. The routing means comprises incident lens 111, a diffraction grating 110 and output lens 112.

Figures 2 and 3 show an NxN wavelength router according to the invention having a plurality of input ports 1-8-N, a plurality of output ports 1'-8'-N'and routing means 20.

The routing means comprises a known arrayed waveguide grating (AWG) as described above, which is passive and also adapted to maintain the linear and orthogonal polarisation of the light signal passing through the waveguide. Again, it would be possible to use alternative passive polarisation maintaining structures such as a bulk grating in appropriate circumstances. In Figure 2, a system wit'full connectivity is shown in which each channel has an associated tunable laser, within a transmitter 30, to generate a signal. Figure 3 shows an alternative embodiment in which a single tunable laser, within a transmitter 35, is provided for each pair of channels. This has reduced connectivity but also requires only half the number of tunable lasers.

'In use, use, two effective optical transmitters TX are connected to each input port and two effective optical receivers RX to each output port. Typically, the transmitters and receivers will comprise optical transceivers or wavelength translators connected to an optical communications network based on single mode fibre.

Figure 2. shows a first embodiment, where at each of the transmitters 30, for example TRIA, a data signal, DiA, is encoded on to the light signal, 1b1A, at a particular polarisation, plia, and the polarized signal is passed down a polarisation maintaining fibre 21 to a combiner 40 to effectively combine the signal with that from transmitter'TXis at wavelength #1B and data D1B encoded ori phase P1B, which combiner 40 is connected to an input port 1-8. The signal is then routed by the routing means 20 to the appropriate output port I'-8'for the specific wavelengths #1A and #1B.. At the output port, for example 1', the signal, for example #1A with polarisation P1A, is passed, via polarisation maintaining fibre 21, to a polarisation beam splitter 50, which routes the light signal to a receiver 60, for example RX1'A or RX1'B according to its polarisation. Similarly, the other wavelength #1B with polarisation P1B due to its differnt wavelength will be routed from input port 1 to a different output port say 4', where it will pass, via polarisation maintaining fibre 21, through a polarisation beam splitter 50 to be passed to a receiver 60, RX4, A or RX4B according to its polarisation. Analogously, a signal arriving at input port 8 with wavelength B8B, say, with polarisations P8B would be passed to the appropriate output port 1' if #8B equalled #1A. In the circumstance where the two signals at port 1'are co-phased, and the encoded data is on mutually exclusive phases, so the data will be successfully routed to the appropriate receivers RIA or RXi'B. Contention resolution will be required to eliminate mutual interference if more than one data signal concurrently reaches a common receiver 60. Similarly contention resolution will be required to eliminate non co-phased signals reaching a common output port of the (wavelength dependent) router 20, as these would interfere Figure 3, shows a second embodiment where each input port of the router 20, receives, via polarisation maintaining optical fibre 21, data from a transmitter 35, in which two sets of data, for example, D1 and D2, are encoded on transverse phases of the light signal of wavelength i, from the associated laser of this example. The wavelength Al is routed to a specific output of the router, say output 1', where it is conveyed via polarisation maintaining optical fibre 21, to a polarisation beam splitter 50, and thence the two encoded data signals D1 and D2, are routed to respectively receivers 60 being RX1'A or RX1'B, dependent upon their respective encoding polarisatiions. Analogously, data signals on a wavelength will also reach output port 1'if ? b8 equals Xi. Hence, as with the first embodiment, contention resolution will be required to eliminate mutual interference if more than one data signal concurrently reaches a common receiver 60. Similarly contention resolution will be required to eliminate non co-phased signals reaching a common output port of the (wavelength dependent) router 20, as these would interfere.

In the most general case, the transmitter 30 or 35, comprises an information encoder or modulator such as an. electro-optic modulator to encode a data signal at a particular wavelength. The data signal will be derived from an external device. The wavelength is determined by system characteristics and will typically be generated within the transmitter using a tunable laser, although the invention does not preclude nominally fixed wavelength lasers e. g. Distributed Bragg Reflector (DBR) lasers being used within the transmitter. For each wavelength, the, light signal can have two orthogonal polarisations. In common usage, these are designated the (transverse electric) TB and (transverse magnetic) TM modes (neglecting the forward component of the vector in each case). Each of these modes can be used to encode separate data. signals at the same' wavelength.

Figure 4 shows-a schematic diagram of the means for generating the polarised light signal. In use, the'tunable laser 200 will generate a linearly polarized light signal. A A/4 waveplate 201 converts this to an equal quantity of TE and TM light. The TE mode polarisation is modulated at 202 with a first data signal, the light is then passed through a ,/2 waveplate 203 and the modulated polarisation component, previously TM, now TE, is modulated at 204 with a second data signal.

In the most general case the laser transmitter comprises an information encoder, also known as a modulator, to encode a data signal at a particular wavelength. Suitable modulation means may be either direct modulation of the laser, or by means of an external modulator.

In a first further embodiment, the signal is passed to a first modulator 202 adapted to eneode a daa signal at the wavelength of the'l-i-ght signal., which modulator encodes the signal in the TM polarisation mode with no substantial effect on the TE mode. The signal is then passed to a second modulator 204, which encodes a second data signal at the wavelength of the light signal in the TE polarisation mode with no effect on the TM mode. In a case such as this, a material with linear elecso-cptic tensor coefficients with appropriate zero terms, such as gallium arsenide must be used Crystals of cubic Zinc-blende structure, such as III-V semiconductors are optically isotropic by default and owe their electro-optic coefficients to their non-centrosymnietiic nature. The electro-optic effect is described by a 6x3 tensor whose elements 41, 52, and 63 alone are non-zero. The zero-populated upper-half of this tensor implies that none of the primary crystallographic axes provide electro-optic change for applied E-fields in their own direction for light polarised in the same direction. The lower-half tensor diagonal non-zero elements provide an electro-optic effect for in-plane, 45°polarised light to an E-field applied perpendicular to the plane, where the planes are defined by major crystal axes.

Thus Ey produces a maximum effect for-light polarised in the x-z plane at 45° to x and z axes, wkere x, y, z are crystallographic axes. These 45° directions also define"cleavage- planes in GaAs, thereby defining preferred propagation directions and polarisation states also. TE-polarised light reacts optimally to normal (into-the-plane) fields. TM polarise light sees no effect since (as noted above) the polarisation direction is that of both the field and a major crystal axis.

As the TE and TM modes are orthogonal, a polarisation rotator 203 can be included-. between the first and second modulators, so that these are co-planar, which simplifies the manufacture of the modulators. Although it would be possible to dispense with the polarisation rotator by aligning the modulators orthogonally to one another, this will typically prove more complex and expensive than including an intermediate polarisation rotator.

Figure 5 shows an alternative embodiment to Figure 4. It is possible to have the two modulators 202 & 204 in parallel to each other by incorporating a polarisation beam splitter 210 in the'optical path prior to the modulators and a polarisation beam combiner 212 after the modulators 202,204. In this case it would also be advantageous to include a polarisation beam rotator 214 between the splitter 210 and one modulator 204, and another beam rotator 216 between the modulator 204 and the combiner 212 so that the modulators can again be co-planar. Alternatively, a simple power splitter could be used, followed by a polarisation rotator in one am This arrangement would be suitable for modulator materials such as lithium niobate, which do not have the aforementioned particular properties of gallium arsenide, that one polarisation can be modulated without significant effects on the other polarisation.

For both the embodiments of Figures. 4 and 5, it would be possible in a further embodiment to include a fast active polarisation changer in the optical path, so that once a first data signal has been written in the TE mode, the TE and TM modes are swapped and the second modulator cvan also write a second data signal in the TE mode. In general, signal modulation is easier to achieve with the TE mode than with the TM mode. This arrangement also benefits from the production advantage that the modulators will be co-planar.

In general changing polarisation state cannot be achieved instantaneously and therefore polarisation state changes will typically only be made between system data blocks.

This applies to all embodiments of the invention.

Figure 6 shows an alternative arrangement for the modulator component. In this case, the modulators based on gallium arsenide have additional control circuitry mounted on the gallium arsenide comprising a simple FET or HEM : T type GaAs switch 220 adapted to switch the data path between the first and second modulators. In this case, a data signal can be routed to either modulator and be written at either polarisation depending on its destination using a conventional semiconductor switching arrangement. This arrangement provides a simple cost-effective control over the routing of the signal.

As an alterative to the basic structure having a single routing means which is adapted to maintain both TE mode and TM mode polarisation, it would be possible to adapt the router to comprise two parallel routing means, each adapted to maintain a single p. plarisation mode. Although mis inevitably increases the number of components and hence complexity of the router design, due to the significant cost component of the polarisation maintaining routing means, it may prove to have lower production costs.

For the avoidance of doubt, it is to be understood that the expression"polarisation maintaining"used herein is intended to include (at least in the broadest aspects of the invention) the polarisation being substantially maintained.

Although the invention has been particularly described using a single laser to generate a single signal having two polarisations, it would be possible to use two lasers to generate the different polarisations. If separate transmitters are used for each respective polarisation, then either fast polarisation switching in the optical domain is required or the ability to choose which polarisation is modulated. Although the light signals have been described as being generated by a tunable laser, it would also be possible to use other known wavelength-selectable sources.

SUBSTITUTE SHEET (RULE 26)