GB2228097A - Single mode optical fibre star coupler - Google Patents

Abstract

An N x N single mode fibre star coupler is produced by forming two bundles 30, 31 of close-packed single mode fibres in e.g. a regular hexagonal array with an appropriate spacing Z between the facing ends of the two bundles. Spacing Z contains a transparent medium e.g. air which precludes multipath transmission. Alternatively, one such bundle may be used in conjunction with a mirror. Preferably the individual members of the two bundles are tapered in order to expand the modal spot size in relation to the overall fibre diameter at the facing ends. It is also preferred to form the bundles in the tapered regions of the fibres and to choose the taper angle and the bundle spacing so that all members of a bundle point directly at the end face of the central member of the other bundle. Preferably such tapers are made by a progressive stretching method in which a length of fibre is repeatedly transversed axially through a micro-torch flame. <IMAGE>

Description

Optical Fibre Star Coupler This invention relates to optical fibre star couplers. Optical networks, for instance for the distribution of high speed data or digital video signals, will require the employment of passive optical branching components, and such components may be required to operate over a relatively wide range of optical wavelengths, typically serving the two low-loss fibre 'windows' respectively in the region of 1300 nm and 1500 nm. Ideally such a component should be low cost, low loss, and should achieve a uniform power distribution among all its output ports. Additionally, since the dispersion of multimode fibres makes them unsuitable for high bandwidth signal traffic, the component must be suitable for use in a single mode system.

A relatively low cost passive optical branching network for multimode fibre, which achieves relatively good uniformity of power distribution, is provided by a star coupler. Such a coupler has the ends of a first set of multimode optical fibres arranged side-by-side in a close packed array terminating at one end of a transparent mixer rod, while a similar second set of multimode fibres similarly terminate at the opposite end of the mixer rod. Light launched from any one of the multimode fibres into the mixer rod quickly spreads out in that rod, and as a result of reflection at the side walls of the rod, provides a relatively uniform distribution of light at the far end of the rod, provided that that rod is of adequate length.

Because this star coupler employs multimode fibre, it is not suitable-for use in a single mode fibre optical system. Using-a -mixer rod in association with - single mode fibres appears -unattractive, at least when first considered, format least two reasons First,: there is liable to be considerable loss resulting -from the fact that the modal spot size of a conventional single mode fibre is considerably smaller than the cross-sectional area of the fibre, and hence, if the light were uniformly distributed over the far end face of the mixer rod, only a very small proportion would be coupled into the zero order modes of the fibres butted against that far end. This is essentially a packing fraction problem.Second, whereas light launched into a mixer rod from a typical multimode fibre will relatively quickly produce a substantially uniform distribution of light across the cross-section of the rod, this does not happen when monochromatic light is launched into the rod from a single mode fibre, but instead an interference speckle pattern is produced. This difference in behaviour is due to the fact that the light launched into the rod from a single mode fibre excites only a limited population of modes in the rod, whereas light launched from the multimode fibre excites thousands of modes, as a result of the fact that the fibre itself will be guiding hundreds of modes.

In principle a multiway passive branching component can be formed by concatenating a set of two-way passive branching elements. 2 x 2 couplers are known in both integrated optics format and all-fibre format in which 3dB coupling is effected between two identical waveguides by bringing them close enough together over a specific distance to effect the requisite strength of lateral coupling. Such a device is however intrinsically wavelength sensitive because the coupling strength of laterally coupled waveguides is a function of wavelength.It is possible, by arranging to use waveguides with appropriately related dissimilar propagation characteristics in the coupling .region,..-to achieve a coupler with a 'wavelength flattened' response which provides close to 3dB coupling over a significantly broader spectral range 'than --is- produced when the lateral coupling is between optical waveguide with identical propagation characteristics.

'Wavelength flattened' 3dB couplers have been constructed in all-fibre format and, as stated previously, it is in principle possible to produce an N x N branching component by concatenating an appropriate number of such couplers. This becomes an increasingly unattractive approach as the size of N increases. Thus to produce a 4 x 4 branching component this way involves the use of 4 couplers inter-linked by means of 4 splices, whereas to produce an 8 x 8 branching component involves the use of 12 couplers inter-linked with 16 splices, while a 16 x 16 branching component involves the use of 32 couplers inter-linked with 48 splices. This shows how complexity, and hence also cost, rapidly escalates.An attempt to implement a corresponding structure in integrated optics format, for instance using silica on silicon, would in principle allow a monolithic approach avoiding any requirement for splices, but this advantage is offset by the need to provide cross-overs, and by the fact that the size of device becomes rapidly relatively large because of radius of curvature limitations of bends in integrated optics single mode waveguides. Moreover there is the additional problem of interfacing the input and output ports of the integrated optics structure with individual single mode optical fibres. Thus it can be seen that although concatenation of 2 x 2- couplers is in principle usable to provide a multiway passive branching element, this approach becomes less and less practical as the requirement for the number of output channels increases.

Previous reference has been made to the problems of adapting the design of a multimode optical fibre star coupler-to--be-sui-tabl-e-~for~operåtion~~with~~~--- single mode fibre. In certain circumstances these problems are capable -of being overcome, and in the specification of United Kingdom Patent Application No. 2,207,525A, to which attention is directed, -there-is described how this can be achieved.The specification describes how the packing fraction problem can be ameliorated by tapering the fibres to produce enlarged modal spot sizes at the interface with the mixer block, and describes how, by the use of an annular mixer block, the interference effects resulting from the launching of a restricted number of modes can be turned to advantage and used to produce a regular pattern of spots at the far end of the mixer block that register with the ends of the output fibres. This is a multiway branching element capable of dividing optical power substantially equally between ten or possibly up to a few tens of single mode output fibres. Dimensional tolerance and fibre alignment are however increasingly critical as the number of desired outputs is increased. Furthermore, since the operation of the device relies upon interference effects, it is intrinsically wavelength sensitive, and so will operate satisfactorily over a narrow range of wavelengths, typically only a few tens of nanometres wide. Similar considerations also apply if the tubular mixer block is replaced by a holographic imaging element similarly designed to produce a pattern of spots imaged on the ends of the output fibres from illumination of the element with a single spot of light from a single one of the input fibres.

Substantial insensitivity to wavelength changes can be provided in integrated optics single mode Y-junction power dividers. The principle of operation of such a device is that zero order mode optical power launched into the stem of the Y proceeds substantially adiabatically across the coupling region and into the two branches of the Y without exciting, to any substantial extent, anyhigher order modes. In contrast with the 2 x 2 coupler, the couplingSdoes-not-rel-y--upon interference effects between different modes,- -and hence is wavelength insensitive. The same principle can be employed to construct a' fibre format version of the wavelength insensitive coupler, and an example of such a coupler has been described by Minelly and Hussey in a paper entitled 'A Single Mode Fibre Y-Junction Beamsplitter', Electronics Letters Vol. 23, p. 1087, 1987. The principle of operation can be extended to produce a device which divides the power substantially equally between, not just two single mode output fibres, but a small number of single mode output fibres.

Concatenation of couplers can then in principle be employed to produce a wavelength insensitive single mode 1 x N branching component. The principle of operation is however not suited to the production of an N x N branching component. The reason for this is that although the basic building block of the component, the Y-junction beamsplitter, is efficient as a beamsplitter, it is not efficient when operated in the reverse direction. If light is directed into one of the branches of the Y, it will proceed to the coupling region, at which point optical power is concentrated to one side of the guide, namely the side of the energised branch. This means that in this coupling region the power is launched substantially equally divided between the zero order mode and the first order mode.The zero order mode power is successfully launched into the stem of the Y but, since this stem is a single mode waveguide, all the first order power is lost. Thus in the reverse direction there is at least 3dB loss of power.

In a paper by C. Dragone entitled 'Efficient N x N star Coupler Based on Fourier Optics', Electronics Letters 21 July 1988 Vol. 24 No. 15 pp 942-4, to which attention is directed, there is described an integrated optics format of star coupler designed for use with single mode fibre. -In this starcoupler is formed a first set of waveguides which terminate on one side-of a portion of slab waveguide, while a second set of waveguides terminate on the opposite side. Optical power launched into the slab waveguide from the end of one of the first set of waveguides radiates within the slab waveguide, with the result that each of the members of the second set of waveguides intercepts a proportion of radiated power.In this way there is produced an N x N optical branching component which, with the appropriate arrangement of waveguides, will divide the incoming power substantially equally between all the output waveguides, and moreover will achieve this power division in a substantially wavelength independent manner. The device is described in the context of a requirement for N x N couplers with N as large as 100.

Some difficulty will however be encountered in packing 100 integrated optics waveguides side-by-side in a linear array, and there is the additional not-insignificant problem of interfacing those two sets of 100 integrated optics waveguides with 200 single mode optical fibres. Accordingly, though the underlying concept of the device has a certain elegant simplicity, its practical implementation presents not inconsiderable difficulties.

The present invention is also directed to a form of star coupler suitable for incorporation into single mode optical fibre networks, but in which recourse is not had to an integrated optics format. In this way the need for an interface between integrated optics waveguides and optical fibres is avoided, and a consequent saving in complexity and cost can be made.

Additionally a more compact and potentially more optically efficient assembly of optical fibre waveguides at the two ends of the mixing element is possible because these fibre waveguides can be assembled into a two-dimensional array instead of the one-dimensional array of the integrated optics N x N star coupler.

According to the present invention there is provided an optical fibre star coupler having a first set of optical fibres assembled with their first ends in a substantially close-packed array of substantially circular cross-section with their first ends terminating in a first surface spaced from a second surface at which the first ends of a second set of optical fibres terminate with said first ends of said second set of fibres assembled in a substantially close-packed array of substantially circular cross-section, wherein the first and second surfaces are separated by a substantially transparent intervening medium, wherein the optical fibres of said first and second sets of optical fibres are single mode fibres and wherein said intervening medium is configured to preclude multipath transmission of light from any member of either one of said sets of fibres via said intervening medium to any member of the other set.

The invention also resides in a method of manufacturing star couplers.

There follows a description of optical fibre star couplers embodying the invention in preferred forms, and of a method by which such couplers can be made. The description refers to the accompanying drawings in which: Figure 1 schematically depicts the optical fibres across the diameter of two bundles of fibres arranged in a regular hexagonal close-packed array with a spacing between the facing ends of the two bundles so as to provide a star coupler, Figure 2 depicts a plot of the computed 'worst case' loss of the coupler of Figure 1 plotted as a function of fibre offset distance for specific values of spacing of the end faces of the two fibre bundles, Figure 3 depicts the star coupler of Figure 1 modified by the use of tapered fibre bundles whose taper angle is chosen in relation to the inter-bundle spacing so that each fibre points-towards-the end--of -the-central fibre of the opposite bundle, Figure 4 depicts a plot corresponding to the plot of Figure 2, but in respect of the coupler of Figure 3, Figure 5 is a schematic diagram of the apparatus used to produce the tapered fibres for use in the coupler of Figure 3, Figure 6 is a schematic diagram of one of a pair of jigs employed to form the close packed fibre bundles of the star coupler of Figure 3, and Figures 7 to 10 depict further modified forms of star coupler.

The essential components of the star coupler comprise two bundles of single mode optical fibres separated by a gap which may be air or some other transparent medium, and which is configured to preclude multipath transmission of light from any member of either one of the bundles to any member of the other bundle. The avoidance of multipath transmission, for instance constituted by a direct path from the one fibre to the other and an indirect path by way of a single reflection in a side wall of the intervening medium, is a requirement in order to preclude interference effects from introducing any significant wavelength sensitivity into the operation of the device.

In Figure 1 the first and second bundles of single mode fibres are represented respectively at 10 and 11. Each bundle is of substantially circular cross-section, and consists of a close-packed regular hexagonal array of fibres. The central fibres of the two bundles lie on a common axis. Light emerging from any one of the fibres into the space between the two bundles radiates into that space so that, at the far side of the space, distant z from the near side, the optical field covers not only the end of the fibre directly opposite, but also the ends of the fibres surrounding that fibre. - -Clearly-the fibre directly opposite intercepts the greatest proportion of the light, while the fibre most remote from the directly opposite fibre intercepts the least.The 'worst case' is that which results for light launched into the intervening space from a fibre at one vertex of the first bundle to be intercepted by the fibre at the diagonally opposite vertex of the other bundle. The proportional difference between the power intercepted by the 'best case' fibre, the fibre directly opposite, and that intercepted by the 'worst case' fibre is reduced as the separation between the ends of the two bundles is expanded. On the other hand the absolute value of the power intercepted by the 'best case' fibre is monotonically reduced with increasing separation. In the case of the power intercepted by the 'worst case' fibre, this increases from zero with zero separation, through a maximum at an intermediate separation, and then reduces again asymptotically to zero as the separation increases towards infinity.Generally it is found that the optimum separation is that which produces the maximum interception of power by the 'worst case' fibre.

A typical single mode fibre for use at a wavelength of 1530 nm has an overall diameter of 125 Vm, a core diameter of about 8 pm, and an index profile providing a modal spot size of about 10 um. (Modal spot size is defined as the diameter at which the optical field has dropped to l/e of its peak value at the centre of the spot). Clearly the efficiency of the coupler would be increased if this modal spot size could be increased so as to be a larger proportion of the overall fibre diameter. This is conveniently achieved by starting with conventional single mode fibre, and drawing it down in a manner providing an adiabatic taper.By the time the fibre has been drawn down to an overall diameter of about 50 Fm, the modal spot size has increased to about 23 m.- -Such fibre should not be drawn down any further than this. because then the optical field would penetrate to an-appreciable extent beyond the physical confines of the fibre, and thus give rise to wavelength dependent optical coupling between adjacent fibres of the two bundles.

In Figure 2 there is shown the computed loss involved in the coupling of light from one 23 um modal spot size fibre of bundle 10 (Figure 1) to one 23 modal spot size fibre of bundle 11, this loss being plotted as a function of fibre offset distance for specific values of separation (z) between the facing ends of the two bundles. From this Figure 2 it is seen that, with a fibre offset distance of 200 Fm, the computed loss is about 29dB for a bundle separation of z = lOmm, and is slightly greater both for a bundle separation of both z = 8 mm and z = 12 mm. This 200 pm offset distance is the 'worst case' offset for a regular hexagonal close-packing of nineteen fibres having an overall diameter of 50 pm.Splitting power nineteen ways involves an intrinsic branching loss of about 12.8dB, and so the 'worst case' excess loss introduced by the inefficiencies of this particular design of 19-way star coupler is approximately 16dB.

The corresponding value of loss for a 300 pm offset distance, corresponding to a worst case offset for a regular hexagonal close-packing of thirty seven 50 pm overall diameter fibres is about 33dB for the optical bundle separation z 12mm. The intrinsic branching loss is in this instance about 15.7dB, and so the 'worst case' excess loss is about 17dB. Further calculations reveal that this 'worst case' excess loss for optimised bundle separation (z) remains at about 17dB in respect of larger regular hexagonal close-packed arrays at least as far as the array of 127 fibres. It is also found that in each instance there is about 4dB difference between 'best case' and 'worst case' coupling efficiencies at optimum separation.

The excess loss is accounted for in part by the fact that part of the energy-launched into the intervening space between the two bundles from any single member of one of the bundles will entirely miss all the members of the other bundle. Another factor is that not all of the light that is intercepted by the end face of a receiving fibre will be launched into the zero order propagating mode of that fibre. Such light as is not launched into this mode is not guided, and hence is lost.It is calculated that if a plane wave of uniform power density is incident normally upon the end face of a single mode receiving fibre having a modal spot size of 23 rm and an overall diameter of 50 pm, then the proportion of the incident power upon the regular hexagon that circumscribes the 50 pm diameter end face that is not launched into the zero order mode, and hence is lost, corresponds to a loss of about 3.8dB. Thus it is clear that the majority of the 17dB 'worst case' loss is attributable to light launched from a member of one bundle entirely missing the end of the other bundle.

This suggests that some improvement of layout should be possible.

Figure 3 depicts an improved layout which, like the star coupler of Figure 1, has two bundles 30 and 31 of single mode optical fibres separated by a gap which may be air or some other transparent medium. As before, the proximal ends of the fibres are arranged in a substantially circular cross-sectional close-packed array, preferably an array in the form of a regular hexagon, and the central fibres of the two bundles are arranged to lie on a common axis with their two ends separated by a distance z. In this instance the fibres of a bundle are not parallel with each other, but are inclined so that the axis of each fibre of the bundle passes through the centre of the end face of the centre fibre of the other bundle.This angular orientation is conveniently provided-by arranging for the individual fibres to be linearly tapered from their 125 jim overall diameter to 50 pm diameter at the appropriate angle to provide line contact between adjacent fibres in a bundle over substantially the full length of the taper.

In Figure 4 there is shown the computed loss involved in the coupling of light from one 23 pm modal spot size fibre of bundle 30 to one 23 pm modal spot size fibre of bundle 31, this loss being plotted as a function of offset distance of the end of the receiving fibre from the axis of the central fibre of that bundle for specific values of fibre bundle separation (z) between the facing ends of the two bundles. This computation takes into account not only linear displacement, but also an obliquity factor arising from the fact that the axis of the launching fibre is inclined at an angle to that of the receiving fibre. In order to make this a 'worst case' obliquity factor, it is assumed that the launching fibre is the one whose end face is displaced from the central axis in the same direction, and by the same amount, as the end face of the receiving fibre.

From Figure 4 it is seen that, with an offset distance of 100 pm, the computed loss is about 24dB for an optimum bundle separation distance (z) of between 4 and 6 mm. With this arrangement of tapered fibre bundles, this offset distance of 100 pm is the 'worst case' offset distance for a regular hexagonal closepacking of 19 tapered fibres with 50 pm diameter end faces. As mentioned previously, the intrinsic branching loss for 19-way branching is about 12.8dB, and so the 'worst case' excess loss introduced by the inefficiencies of this particular design of 19-way coupler described with reference to Figure 5 is about lldB.

The corresponding value of loss for a 150 pm offset distance, corresponding to a 'worst case! offset distance for a regular hexagonal close-packing of 37tapered fibres with 50 pm diameter end faces, is about 26dB for the optimum bundle separation distance (z)- of about 8 mm. The intrinsic branching loss is in this instance about 15.7dB, and so the 'worst case' excess loss is about 10dB.

Further calculations reveal that this 'worst case' excess loss, for optimised bundle separation distance (z) of angled tapered bundle couplers of the type described with reference to Figure 3, remains at about 10dB in respect of larger regular hexagonal close-packed arrays at least as far as the array of 127 fibres. It is also found that in each instance there is about 4dB difference between 'best case' and 'worst case' coupling efficiencies.

The manufacture of a star coupler as described with reference to Figure 3 requires the use of single mode optical fibres with linear tapers having specific taper angles. Such fibres are conveniently made from standard parallel-sided fibre by adapting the progressive stretching method of fused fibre tapered coupler manufacture described in United Kingdom Patent Specification GB 2,150,703A so as to produce controlled tapering of a single mode fibre, rather than the controlled simultaneous tapering of two or more stranded single mode fibres. According to this manufacturing method, a length of parallel-sided single mode fibre, complete with plastics protective coating 50, is secured by clamps 51 and 52 which are mounted on independently driven linear movement carriages (not shown) that operate along a common direction aligned with the axial extent of the fibre between the two clamps.Either before or after the mounting of the fibre in the two clamps 51 and 52, the plastics protective coating is removed from the region between the two clamps to expose a region 53 of bare fibre. A micro-torch 54 is located beneath the fibre between the two clamps in order to heat-soften the short zone of bare fibre 53 that lies within the flame of the micro-torch sufficiently to allow it to stretch under tension provided by moving the two carriages in the same direction, with the leading carriage constrained to move slightly faster than the trailing carriage. A traverse of this kind produces a controlled drawing-down of the fibre by an amount determined by the difference in speeds of the two carriages. The location of the drawn-down region is determined by the movement of the two carriages relative to the micro-torch.In order to produce two substantially linear tapers in the fibre, these tapers being separated by a reduced diameter portion of substantially constant cross-section, a number of such traverses are performed.

Typically the traverses are alternately in one direction and then in the other. The traverses are controlled in extent so that the small neck in diameter produced at the beginning and end of each progressive traverse is located at the appropriate distance inside the two corresponding necks produced by the preceding traverse. In this way there is produced a succession of necks whose spacing is such as to approximate to the required linear taper. In practice the necks are so gradual that the approximation is close.

It is found that this progressive stretching method is capable of providing highly reproduceable results, so that, once an appropriate schedule has been determined for producing a first tapered fibre of the appropriate profile, the schedule can simply be repeated to produce a whole set of fibres with substantially identical profiles. The individual members of such a set are assembled in two stranding jigs which hold the ends of the fibres in a regular hexagonal array. Such a jig may take the form depicted in Figure 6, and comprises a disc 60 provided with slots 61 wide enough to freely accommodate entry of the optical fibres 62, complete with their plastics protective coatings 63.

Shims 64 are inserted between adjacent fibres to space them apart in the requisite hexagonal array. The individual fibres are individually secured (by means not shown) above the first jig, which is the upper jig, and hang vertically down from the first jig and through the second jig, which is the lower jig, vertically below.

Slight tension in the fibres is provided by individual weights (not shown) attached to their lower ends.

At this stage the sections of bare fibre which have been tapered are spaced from each other in the region between the two jigs. They are brought into contact with each other by rotating one of the jigs with respect to the other. This causes the fibres to become stranded together but, because each fibre is free to rotate about its own axis in the lower jig, such stranding produces substantially no twisting of any of the fibres about its own axis.

Following stranding of the fibres, the bundles are clamped and the jigs rotated back to their original relative orientation. A relatively low viscosity adhesive is applied to the bare fibres, and is allowed to wick in between them, before being cured to a solid.

The two jigs are removed, and the array of fibres is sawn in half between the two taper bundles. A ferrule (not shown) is fitted around each tapered bundle, and is secured with adhesive before the ferrule-terminated end is polished to provide a plane polished surface at the point along the length of the taper where the individual fibres have an overall diameter of 50 pm. Next the two assemblies are secured by their ferrules to a supporting substrate (not shown) which holds them in alignment with the required intervening spacing. If desired, this intervening space may be filled with an appropriate transparent medium before the resulting assembly is sealed within a protective housing (not shown).

The manufacture of a star coupler as described with reference to Figure 1 may proceed along essentially similar lines, with the proviso that the tapering of the individual fibres needs to be programmed to provide an adequate length of substantially constant cross-section 50 um diameter fibre between the two tapers. The length of this 50 um diameter portion must be sufficient to allow it to be cut in half, and still leave sufficient length for the regions of the two bundles of fibres which are required to have their individual fibres in parallel alignment.In contrast with this, in the manufacture of the tapered fibres for the star coupler of Figure 3 there is no need for the taper to be halted at the point where the diameter is reduced to 50 um, and if tapering is allowed to proceed beyond this point, there is no need for any appreciable length of untapered fibre between the two tapers because such a region will be entirely removed before the coupler is completed.

The performance of the N x N coupler of Figure 3 is better than that of Figure 1 because all the fibres of each bundle are angled to point at the centre of the end facet of the other bundle. A substantially equivalent improvement in performance may alternatively be achieved by providing, as depicted in Figure 7, the fibres of the two bundles with angled end faces 70 each inclined to its fibre axis by an amount to refract the emergent light so as to be centred on the centre of the end of the opposite bundle. The preparation of individual appropriately inclined and appropriately oriented planar end facets for each of the fibres is liable to add significantly to the -manufacturing cost.

One cheaper alternative is to'leave the fibres with normal end facets, as depicted in Figure 8, to arrange for these facets to lie in a common plane, and to face them with a plano-convex lens 80 whose focal length is substantially equal to the inter-bundle spacing. A different alternative is depicted in Figure 9 where the lenses 80 have been dispensed with, and refraction is instead provided by polishing the end surfaces 90 to a convex form of the appropriate radius of curvature. Yet another alternative is depicted in Figure 10 where once again the fibres terminate with normal end faces lying in two common planes between which is located a graded index lens 100 of the appropriate length and power to be a quarter wavelength graded index lens.

The foregoing specific description has related exclusively to N x N couplers in which N is a number which forms a regular hexagonal close-packed array. It should however be appreciated that the invention is applicable to substantially circular close-packed arrays in which the number of individual fibres forming the array is not such as will-give rise to a regular hexagonal structure. It should also be understood that the invention is applicable to star-couplers of an N x M format where N M.

Another feature of the foregoing specific description is that it has related exclusively to transmissive-type star couplers in which light is launched into the coupler via any one of a first set of optical fibres, the input fibres, in order for that light to be shared substantially equally between the members of a second set of fibres, the output fibres.

There is however another form of star coupler, alternatively known as a reflex- or reflective-type coupler, in which there is only one set of fibres, which function as both input and output fibres, and in which light launched into the coupler by way of any one of those fibres is reflected in a manner providing substantially equal sharing of the reflected light between all the members of the set of fibres. It will be observed that, an N x N star couplers of the formats depicted in Figure 1, Figure 3 and Figures 7 to 10, each have a plane of symmetry between the two fibre bundles.

The result is that each of these types of N x N star coupler can be converted into a reflex-type format by replacing one half of the structure with a plane mirror at this plane of symmetry. In the case of the reflex versions of the Figure 3 and Figures 7 to 10 formats of N x N star coupler, this will mean that all the fibres of the bundle point at the centre of the image of the end face of the fibre bundle formed in the mirror.

Claims (17)

CLAIMS.

1. An optical fibre star coupler having a first set of optical fibres assembled with their first ends in a substantially close-packed array of substantially circular cross-section with their first ends terminating in a first surface spaced from a second surface at which the first ends of a second set of optical fibres terminate with said first ends of said second set of fibres assembled in a substantially close-packed array of substantially circular cross-section, wherein the first and second surfaces are separated by a substantially transparent intervening medium, wherein the optical fibres of said first and second sets of optical fibres are single mode fibres and wherein said intervening medium is configured to preclude multipath transmission of light from any member of either one of said sets of fibres via said intervening medium to any member of the other set.

2. A star coupler as claimed in claim 1, wherein the members of said first and second sets of fibres are tapered so that each fibre has at its first end a reduced cross-sectional area and enlarged modal spot size compared with the cross-sectional area and modal spot size of its other end, and wherein the extent of the taper is insufficient to introduce any significant lateral optical coupling between adjacent fibres.

3. A star coupler as claimed in claim 2, wherein the close-packing of the members of said first and second sets of fibres extends along their tapers, which are dimensioned in relation to the spacing of said first and second surfaces, such that the first ends of all the members of each said set point to the centre of the end face of the other set.

4. A star coupler as claimed in claim 1 or 2, wherein in each close-packed array the fibres extend with substantially parallel axes, and the first ends of the fibres of each array are individually inclined so that light propagating in each fibre along its axis is refracted towards the centre of the facing end of the other array.

5. A star coupler as claimed in claim 1 or 2, wherein in each close-packed array the fibres extend with substantially parallel axes with the first ends of the fibres of the two arrays respectively terminating in planar first and second surfaces each of which lies adjacent the plane surface of a plano-convex lens associated with that array whose focal length is substantially equal to the distance separating the first and second surfaces.

6. A star coupler as claimed in claim 1 or 2, wherein in each close-packed array the fibres extend with substantially parallel axes with the first ends of the fibres of the two arrays respectively terminating in planar first and second surfaces spaced from each other by an intervening quarter wavelength graded index lens.

7. A method of making a star coupler as claimed in any claim of claims 2 to 6, wherein each member of the first and second sets of fibres is made from parallel-sided fibre by a progressive stretching method in which a length of the parallel-sided fibre is repeatedly traversed axially through a flame providing localised heat-softening.

8. An optical fibre star coupler having a set of optical fibres assembled with their first ends in a substantially close-packed array of substantially circular cross-section with their first ends terminating in a first surface spaced from a reflecting second surface by a substantially transparent intervening medium, which second surface reflects light launched from any one of the fibres of the set of fibres back substantially equally divided between all members of the set of fibres, wherein the optical fibres of the set of fibres are single mode fibres; and wherein said intervening medium is configured to preclude multipath transmission of light from any member of the set of fibres to any other member of that set.

9. An optical fibre star coupler as claimed in claim 8, where in the members of said set of fibres are tapered so that each fibre has at its first end a reduced cross-sectional area and enlarged modal spot size compared with the cross-sectional area and modal spot size of its other end, and wherein the extent of the taper is insufficient to introduce any significant lateral optical coupling between adjacent fibres.

10. A star coupler as claimed in claim 9, wherein the close-packing of the set of fibres extends along their tapers which are dimensioned in relation to the spacing of said first surface from said reflector such that the first ends of all the fibres of the set point at the image of the centre of the end face of the set formed in said reflector.

11. A star coupler as claimed in claim 8 or 9, wherein in the close-packed array the fibres extend with substantially parallel axes, and the first ends of the fibres of the array are individually inclined so that light propagating in each fibre along its axis is refracted towards the centre of the image of said first surface formed in said second surface.

12. A star coupler as claimed in claim 8 or 9, wherein in the close-packed array the fibres extend with substantially parallel axes with their first ends terminating in a planar first surface adjacent the plane surface of a plano-convex lens whose focal length is substantially equal to twice the distance separating the first and second surface.

13. A star coupler as claimed in claim 8 or 9, wherein in the close-packed array the fibres extend with substantially parallel axes with their first ends terminating in a planar first surface which is spaced from the second surface by an eighth wavelength graded index lens.

14. A method of making a star coupler as claimed in any claim of claims 9 to 13, wherein each member of the set of fibres is made from parallel-sided fibre by a progressive stretching method in which a length of the parallel-sided fibre is repeatedly traversed axially through a flame providing localised heat-softening.

15. A star coupler substantially as hereinbefore described with reference to Figures 1 to 4 of the accompanying drawings.

16. A star coupler substantially as hereinbefore described with reference to Figures 7 to 10 of the accompanying drawings.

17. A method of making a star coupler substantially as hereinbefore described with reference to the accompanying drawings.