WO1997034428A2

WO1997034428A2 - Optical switch architecture

Info

Publication number: WO1997034428A2
Application number: PCT/GB1997/000607
Authority: WO
Inventors: Robert Spagnoletti; Terry Bricheno; James Wilson Parker
Original assignee: Northern Telecom Limited
Priority date: 1996-03-13
Filing date: 1997-03-06
Publication date: 1997-09-18
Also published as: WO1997034428A3; GB2311180B; GB2311180A; GB9605298D0

Abstract

A switch architecture in which a set of n smaller switches (10) are interconnected by means of a core controller (12) and an optical shuffle (16) to form a big switch. Each smaller switch has transmitter block (14) with n laser diodes (14a) connected by an n-fibre ribbon (15) with the shuffle, and a detector block (18) similarly connected with the shuffle. The core controller selects not more than one laser diode of any smaller switch to transmit at any one time.

Description

OPTICAL SWITCH ARCHITECTURE

Background of the Invention

This invention relates to switch architectures in which a set of up to n smaller switches are interconnected to form a big switch. An example of such an architecture is described for instance in GB 2 289 813A. The architecture described in that specification involves transmission of information from the smaller switches to a central optical shuffle on a 'broadcast' basis, and selection of the broadcast information for the appropriate destination small switch is performed on the information returning from the optical shuttle to the smaller switches.

Summary of the Invention

The present invention is directed to an alternative form of architecture having a number of advantages that include the reduced cross-talk susceptibility and the possibility of reduced optical loss and hence of improved power budget enabling for instance operation of the switch at a higher bit rate.

According to the present invention there is provided a switch architecture in which a set of up to n smaller switches are interconnected with a core controller and an n x n optical shuffle to form a big switch, wherein each of the smaller switches includes a transmitter block provided with a set of n optical transmitters optically coupled with the optical shuffle via an associated first n-pathway optical connection, an associated buffer store, and a detector block optically coupled with the optical shuffle via an associated second n-pathway optical connection, and wherein the interconnections provided by the optical shuffle are one-to-one interconnections between individual pathways of the first n-pathway optical connections and individual pathways of the second n-pathway optical connections arranged to provide, for each one of the smaller switches, optical coupling between its associated detector block and an optical transmitter of each one of the other smaller switches.

The invention also provides a method of switching in which n smaller switching are interconnected, using a core controller and an n x n optical shuffle, to form a big switch, wherein each of the smaller switches includes a transmitter block provided with a set of n optical transmitters optically coupled with the optical shuffle via an associated first n-pathway optical connection, an associated buffer store, and a detector block optically coupled with the optical shuffle via an associated second n-pathway optical connection, and wherein the interconnections provided by the optical shuffle are one-to-one interconnections between individual pathways of the first n-pathway optical connections and individual pathways of the second n-pathway optical connections arranged to provide, for each one of the smaller switches, optical coupling between its associated detector block and an optical transmitter of each one of the other smaller switches, in which method the core controller controls the buffers of the smaller switches to ensure that only one transmitter of any particular smaller switch is enabled to transmit at any one time, allowing different transmitters of different smaller switches to transmit at the same time while ensuring that no two transmitters are transmitting simultaneously to the same destination smaller switch.

Brief Description of the Drawings There follows a description of an architecture of very high capacity ATM switch embodying the invention in a preferred form. The description refers to the accompanying drawings in which:

Figure 1 is a schematic block diagram of the switch architecture, and Figure 2 is a schematic representation of the optical coupling arrangement between one of the n-pathway optical connections of the architecture of Figure 1 and an associated photodetector. Detailed Description of a Preferred Embodiment

The switch architecture of the preferred embodiment employs a time- space-time configuration interconnecting n smaller switches 10 (10.1 , 10.2, .... 10.n). Typically there may be sixteen such smaller switches interconnected by the architecture.

Each switch 10 has a buffer having an input port 11a and an output port 1 1b respectively for receiving ATM cells from, and transmitting ATM cells to, the outside world. The buffers operate under the control of a core controller 12 established via control links 13, the controller 12 being common to all n smaller switches 10. Under the control of the core controller 12, ATM cells are transferred from the input buffer 1 1 a to a transmitter block 14 where, according to the destination of the particular ATM cell concerned, a particular one of a set of n laser diodes 14a is energised. The individual members of the set of laser diodes are optically coupled with the individual single mode optical fibres of an n-fibre optical fibre ribbon 15 connecting the transmitter block 14 with a n xn optical shuffle 16 to a receiver block 18 of the small switch 10 in which a combiner unit 18a optically couples each of the n optical fibres of the ribbon 17 with a single photodetector 18b. The output of the photodetector 18b feeds ATM cells into the output port 11 b of the buffer for onward transmission under the control of the core controller 12 to the outside world.

The optical shuffle interconnects the fibres of the ribbons 15 and 17

so that the photodetector of each smaller switch 1 is optically coupled by the individual fibres of its associated ribbon 17 with a laser diode 14a from each one of the smaller switches 10. This interconnection may follow the pattern in which the n fibres of ribbon 15 associated with the first smaller switch are connected by the optical shuffle with the first fibres of each of the ribbons 17; in which the n fibres of the ribbon 15 associated with this second smaller switch are connected by the optical shuffle with the second fibres of each of the ribbons 17, and so on.

Under the control of the core controller, the buffer of any particular smaller switch 10 is enabled to transmit from only one of its laser diodes 14a at any one time, but different laser diodes 14a on different smaller switches 10 may transmit at the same time provided that they are transmitting to different detectors 18b on different smaller switches 10. Control of this sort may for instance be exercised in the core controller 12 after the manner described in the specification of Patent Application No 9507871.3.

Each transmitter block 14 may take the form of a monolithic bar of laser diodes with a multi-way laser driver IC co-packaged with a low impedance track between them. Cross-talk is not an issue here because only one laser diode 14a of an individual transmitter block is on at any one time. This is in direct contrast with the situation pertaining in the architecture described in GB 2 289 813A to which previous reference has been made.

Obviously, from power budget considerations, it is desirable for the combiner unit 18a of each receiver block 18 not to introduce unnecessary loss. The function of the combiner unit 18a is to receive an input from any one of the n fibres of its associated fibre ribbon 17 and direct that input into the photoresponsive area of its associated photodiode. A similar function is required in the architecture of GB 2 289 813A where this function is provided by a 1 to 16 integrated single mode waveguide radiative star arrangement which is operated in reverse direction so as to act as a combiner rather than as a splitter. It could alternatively have been provided by a tree of bifurcating splitters, similarly operated in reverse. In either instance the combining function is achieved with no insignificant loss: in the case of the tree structure this would be not less than 12dB for a 1 to 16 tree. Such loss provided by a combiner operating in the optical regime can, at least in principle, be avoided by choosing instead to perform the combining operation in the electrical regime. This is not our preferred option in this instance because it means that each detector block 18 would have to have a further (n-1 ) detectors 18b which not only adds to the component costs but complicates adjustment procedures and also feed-through procedures in the case of hermetically packaged detectors. A preferred form of optical combiner 18a with significantly less loss than the equivalent tree or radiative star combiners described above will now be described with reference to Figure 2. The basic components of this combiner comprises an adiabatically tapered bundle 20 of fibres 21 produced by the controlled stretching of a bundle of standard single mode fibres, a length of step-index multimode fibre 22, and an aspheric lens 23 for focusing the output of the multimode fibre as a reduced size spot upon the photosensitive surface of the photodetector 18b. Such a lens 23 may be a lens designed for launching light from a semiconductor laser diode into single mode fibre.

The photodetector 18b is a fast detector capable of operating, for example, at 10 Gbit/s, having a photosensitive area of about 25μm diameter, and can accept light from quite high angles. The individual fibres of the ribbon 18 are 125μm in diameter and have a spot size of about 10μm diameter, and a divergence N.A. of about 0.1.

Considering by way of example a switch formed by the interconnection of n smaller switches, where 16 < n < 19. The fibres of a ribbon can be hexagonally close-packed to provide a bundle in which the distance between centres of any further apart pair of fibres is between 400 and

500 μm.

To collect all the light emitted from the cores of all the n fibres on to the photosensitive area of the photodetector using standard imaging optics is not possible because the light emitted from the fibre ends is too divergent for it to be collected by any standard imaging optics providing the requisite approximately twenty-fold demagnification factor. The adiabatically tapered bundle 20 of fibres 21 provides a solution to this problem. This bundle is created by forming a hexagonally close-packed fused together assembly of n 125μm diameter fibres, created for instance by bringing together into the array individually separated members of a ribbon of fibres. This assembly is then drawn down in a controlled manner to form two tapers joined by their smaller ends. Preferably these tapers are made using the progressive stretching technique described in GB 2 150 703 in which the bundle is longitudinally traversed several times through a localised hot zone using two translation stages, the leading one of which is moved at a controlled rate faster than the trailing one so as to promote plastic flow strain in the fibre where it is locally softened by the heat of the localised hot zone. By this means an adiabatic taper is formed over which the diameter of the bundle is reduced to about 90μm. The precise profile of the taper is not critical provided that it is slow enough to be adiabatic. It is found that along the length of the taper from its large end to its small end the spot size associated with any individual fibre 21 gradually evolves, starting at about 10μm and ending up between 10 and 15μm. These spots are now much more closely spaced but, because the taper is adiabatic, the brightness N.A. is unchanged, and the 15μm spot size provides a value of less than 0.08 for the N.A. of the emergent light. In consequence of the closer spacing, conventional imaging optics can be employed to collect all the light emitted from the small end of the taper 20 on to the photosensitive area of the photodetector, but it is generally preferred not to do so because, using lenses that are readily available for this purpose, involves positioning the photosensitive surface no more than about 400μm from the lens. This is an inconveniently short distance, particularly if the detector is contained in an hermetic package and the lens is outside that package. Additionally there is a potential problem of lack of uniformity of response of the photodetector to inputs from different fibres due to local variations in photosensor sensitivity, and having regard to the fact that the imaging system will image the output of the different fibres 21 on different portions of the photosensitive area. Accordingly it is preferred to couple the emergent light into the length 22 of step index multimode fibre. This multimode fibre has a core diameter slightly larger than the diameter of the small end of the fibre taper so as to be able to accept all the light emitted. One advantage of using this fibre is that it can form a convenient feed-through in the wall of an hermetically packaged detector. This is particularly the case if the fibre outer diameter is 125μm to match that of standard transmission type single mode fibre for which well-proven feed-through technology is readily available. Another advantage of using this fibre is that, because it is multimode, light launched into one end over a small area of its core relatively rapidly evolves in its propagation along the fibre into a relatively even distribution over the whole area of the core.

Using a 125μm diameter multimode fibre that has a 93μm diameter undoped silica core surrounded by a fluorine doped cladding providing the fibre with a numerical aperture in the range 0.25 to 0.28, it has been found that a metre length of the fibre can be coiled without inducing any significant amount of mode coupling such as would significantly impair light collected by the lens, and that over that length of fibre the dispersion penalty is not significant for 10Gbit/s signals. A preferred way of providing optical coupling between this fibre and the bundle is to form a fusion splice between the two components.

Claims

CLAIMS:

1. A switch architecture in which a set of up to n smaller switches are interconnected with a core controller and an nxn optical shuffle to form a big switch, wherein each of the smaller switches includes a transmitter block provided with a set of n optical transmitters optically coupled with the optical shuffle via an associated first n-pathway optical connection, an associated buffer store, and a detector block optically coupled with the optical shuffle via an associated second n-pathway optical connection, and wherein the interconnections provided by the optical shuffle are one-to-one interconnections between individual pathways of the first n-pathway optical connections and individual pathways of the second n-pathway optical connections arranged to provide, for each one of the smaller switches, optical coupling between its associated detector block and an optical transmitter of each one of the other smaller switches.

2. A switch architecture as claimed in claim 1 , wherein each detector block has a single photodetector with which the n pathways of the associated second optical connection are all optically coupled.

3. A switch architecture as claimed in claim 2, wherein the n pathways of each optical connection are optically coupled with the photodetector of the associated detector block via an adiabatically tapered bundle of single mode optical fibres and a lens.

4. switch architecture as claimed in claim 2, wherein the n pathways of each optical connection are optically coupled with the photodetector of the associated detector block via the series combination of an adiabatically tapered bundle of single mode optical fibres, a length of multimode optical fibre, and a lens.

5. A switch architecture as claimed in claim 4, wherein each photodetector of each detector block is contained in an hermetic enclosure one wall of which is penetrated by its associated length of multimode optical fibre.

6. A method of switching in which n smaller switching are interconnected, using a core controller and an n x n optical shuffle, to form a big switch, wherein each of the smaller switches includes a transmitter block provided with a set of n optical transmitters optically coupled with the optical shuffle via an associated first n-pathway optical connection, an associated buffer store, and a detector block optically coupled with the optical shuffle via an associated second n-pathway optical connection, and wherein the interconnections provided by the optical shuffle are one-to-one interconnections between individual pathways of the first n-pathway optical connections and individual pathways of the second n-pathway optical connections arranged to provide, for each one of the smaller switches, optical coupling between its associated detector block and an optical transmitter of each one of the other smaller switches, in which method the core controller controls the buffers of the smaller switches to ensure that only one transmitter of any particular smaller switch is enabled to transmit at any one time, allowing different transmitters of different smaller switches to transmit at the same time while ensuring that no two transmitters are transmitting simultaneously to the same destination smaller switch.