GB2214746A - Optical interconnection - Google Patents

Abstract

An optical interconnection arrangement is described consisting a plurality of waveguides with optical switching means attached to associated electronic devices. The optical switching means allows routing of light signals through said waveguides, possibly according to a specific regime such as perfect shuffle, until the light signals are incident upon the electronic device desired. Preferably the switching is effected using reflective couplers and waveguide switch elements. <IMAGE>

Description

OPI1CAL IN"IERCONNECIION The present invention relates to optical interconnection.

In electronic arrangements, devices or groups of devices require connection in order to provide a specific function. There are problems with electromagnetic interconnection bowever with increased semi-conductor areas and reduced device size in that such connections may introduce undesirable time delays. In addition, there are electromagnetic interconnection problems of layout due to consideration of possible mutual noise induction and actual provision by etching.

A proposed solution is to perform such device interconnection using optical waveguides. It is also appreciated that by repeated application of a static function such as "Perfect Shuffle" distribution of signals between a group of devices or simple active components that comparators, adders and switches could be constructed.

The theory of "Perfect Shuffle" is given by "Parallel Processing with the Perfect Shuffle" H.F. Stone, IEEE Trans. Computers C-20, 153-161, (1971).

The perfect shuffle is illustrated graphically in Figure 1.

The usefulness of the Perfect Shuffle stems from the observation that many important complex and sophisticated mathematical functions, including Fast Fourier Transform, polynomial evaluation, matrix inversion and sorting, can be implemented by repeated applications of the shuffle operator, coupled with relatively simple two argument arithmetical operations. In particular, the general permutation interconnection between the ports of a general network of size N x N (N = 2m) can be implemented using 2 x 2 exchange elements (switches) in conjunction with shuffle operations.

This operation is of obvious importance in the field of switching.

Figure 2 shows the functional representation of a Shuffle Exchange network of a size N = 8. - Signals applied to inputs 1o - I7 are redistributed to outputs 0o - 07, in accordance with the shuffle algorithm, as expressed in Equation 1, above. The shuffled outputs are applied sequentially in pairs to the inputs of a set of N/2 2x2 switches, the outputs of which are fed back to the system inputs via the return lines R0 - R7. It has been shown that this arrangement is capable of realising every arbitrary permutation of interconnections in 3.log2(N) 1 passes.

A known optical implementation of the perfect shuffle operation is illustrated in "Optical Interconnections for VLSI Systems", J.W. Goodman, F.J. Leonberger, S-Y Kung and R.A. Athale, Proc. IEEE 21, 850-866, (1984).

This comprises a waveguide array represented as a spatial mapping of the shuffle operation, with a further array of parallel waveguides providing a direct 1:1 return mapping of outputs to inputs. The waveguide arrays are interfaced to two linear optoelectronic arrays of alternating and electrically coupled sourceldetector pairs. These serve to detect an optical signal propagating along an element of the forward, "shuffle", waveguide array and to generate a corresponding signal propagating along the corresponding direct reverse path; this is indicated schematically in Figure 3. As described, this configuration relies on optical/electrical/optical conversions at each waveguide interface and makes no provision for optical implementation of the 2x2 exchange function.

The provision of a perfect shuffle arrangement having optical waveguides with electro-optic device termination and switching ends would present problems with a monolithic substrate. A first problem would be to interface the semi-conductor substrate with a planar waveguide with sufficient adhesion and precision. Secondly, there is a problem to produce optical linear arrays of sufficient yield. Finally, it is not possible to produce optical sources and detectors of the same material consequently hybrid materials must be formed.

It is an objective of the present invention to provide an interconnection arrangement wherein the above identified problems are substantially relieved.

According to an embodiment of the present invention there is provided optical distribution means for distributing an optical signal between at least one input node and a plurality of output nodes or between a plurality of input nodes and at least one output node, the distribution means comprising a network of waveguide elements, optical switching means whereby an optical signal input at the or one of the input nodes may be routed, through a plurality of the waveguides, to the or a desired one of the output nodes, and control means for controlling said routing.

Embodiments of the present invention will now be described by way of example only with reference to the accompanying drawings wherein:- Figure 4 illustrates an optical interconnection arrangement according to the present invention having a "Perfect Shuffle" distribution regime and downstream reflective coupler/switches; Figure 5 illustrates an optical interconnection arrangement as shown in Figure 4 with additional upstream reflective couplers; Figure 6 illustrates an interconnection arrangement as shown in Figure 5 with additional input or upstream gating, and, Figure 7 illustrates a "Perfect Shuffle" regime timing diagram for the above arrangements.

In the present invention a network of waveguides is provided between a plurality of input and output nodes. Optical light signals are switched between waveguides in a purely optical manner using 2 x 2 switch elements. With the present arrangement, any one-to-one interconnect pattern can be set-up and, assuming availability of appropriate control circuits. This allows interconnect patterns to be altered at a clock rate approaching the waveguide substrate roundtrip time.

It will be appreciated that sources and detectors may be either directly hybridised active devices or be butt-coupled optical fibres leading from/to remote active devices.

Figure 4 illustrates an embodiment of the present invention wherein a network of interconnective waveguides is arranged to connect a plane 3 of source/input nodes (So-S7) and detector/output nodes (Do-D7) to a plane 5 of optical directional coupler switches (0o- 07) and reflective directional couplers. The switches employed are usually of the 2 x 2 type. The reflective directional couplers act to provide the reverse signal without the need for opto-electronic conversions whilst the 2 x 2 directional coupler switches perform the programmable exchange function. It will be appreciated that such a waveguide network could be readily fabricated in titanium diffused lithium niobate or in a III - V semiconductor waveguide system.

In Figure 5, a second embodiment of the present invention is illustrated. A further array of reflective couplers 15 is incorporated in the input plane 3 as compared to the embodiment illustrated in Figure 1. These additional reflective couplers 15 allow optical signals to circulate repetitively through the shuffle processor. Each additional coupler 15 is formed for additional signal injection and signal extraction. This additional coupler 15 is achieved by arranging for the reflectors 19 to exhibit less than 100% reflectivity allowing continuous light leakage to a suitably gated photodetector interfaced to the reflectors 19. It should be noted only alternate couplers 15 are illustrated in Figure 5, in a practical embodiment each input Io-I7 has a coupler 15.

In Figure 6, a further embodiment or configuration of the present invention is illustrated. This configuration includes the features of the embodiments shown in Figures 4 and 5 with additional switching (2 x 1 switches) to permit appropriately gated signal injection and extraction of the return signals after the appropriate number of passes through the processor. It will be noted that only alternative reflective coupler/2xl Gating Switches are illustrated in the input plane.

It will be appreciated waveguide configurations as shown in Figure 4 to 6 can be constructed upon substrates of the order of 100mm in length with optical round trip times approximating lns giving a clock rate of 1GHz. The size of substrate is determined by the following factors:- the number of element or process stages; the essential maintainance of high angles between waveguides in order to provide low interaction; and, at present, the practical ability to produce - niobate substrates of particular sizes.

The operation of the above interconnection arrangements as a perfect shuffle processor is essentially re-entrant and systolic with a stage period equivalent to the optical round-trip propagation period.

The stages of a perfect shuffle processor operation would include: 1: Pulsed optical inputs, of pulse width less than or equal to the single-pass propagation time, are applied in parallel to the system inputs 1o - I7 which, in the configuration of Figure 6, have been previously enabled. The output cross-over switches are configured to provide the appropriate first-pass routing. Once the pulses have cleared the upstream gate network, the gates are reconfigured to enable the input reflective couplers.

2: The optical pulses traverse the waveguides, enter the downstream reflector/crossover network and are routed to the appropriate return waveguides dependent upon switch setting.

3: The pulses enter the "return lines". Once they have cleared the downstream reflector/crossover system, the crossovers are reconfigured for the next pass.

4: The optical pulses traverse the return waveguides, enter the upstream reflector/crossover network, by now reconfigured as reflectors, and are coupled to the corresponding "shuffle" lines for a further traverse of the network.

5: Stages 2, 3 and 4 are repeated the appropriate number of times, with downstream crossover reconfiguration between passes of the optical pulse front.

6: When data extraction is required the upstream gating switches in Figure 6 are enabled once the optical pulses have cleared the upstream termination region on their final passage round the loop; data extraction is continuous in the configurations shown in Figures 4 and 5.

The cycle of events is indicated in the schematic timing diagram shown in Figure 7. Note that the total processing time is dependent on the number of passes round the loop, which is defined by the size of the network and the processing operation being performed. As described above, pulses occupy either the "shuffle" or "return" lines alternately. A two-fold increase in processing capacity can be achieved by interleaving two pulse systems with a time separation equal to the end-to-end propagation time; the temporal relations implied by this modification are indicated in the timing diagram.

It will be appreciated that the waveguides could be of any type or even a network of optical fibres although planar waveguides are preferred. In addition it will be noted that the present interconnection arrangement can be fabricated upon a monolithic substrate with all switching being conducted without electro-optic interfacing. The removal of electro-optic interfacies reduces the inherent time delay of interconnection and problems of source/detector fabrication and alignment upon a monolithic substrate.

Claims (7)

1. Optical distribution means for distributing an optical signal between at least one input node and a plurality of output nodes or between a plurality of input nodes and at least one output node, the distribution means comprising a network of waveguide elements, optical switching means whereby an optical signal input at the or one of the input nodes may be routed, through a plurality of the waveguides, to the or a desired one of the output nodes, and control means for controlling said routing.

2. An arrangement as claimed in claim 1 wherein the waveguide elements are constructed upon a monolithic substrate.

3. An arrangement as claimed in claims 1 or 2 wherein said routing is a perfect shuffle operation.

4. An arrangement as claimed in claims 1, 2 or 3 wherein the optical switching is performed by a respective combination of reflective couplers and waveguide switch elements arranged to switch light signals between an adjacent pair of the waveguides at a downstream end thereof the waveguides.

5. An arrangement as claimed in any preceding claim wherein optical gating switches are provided at an upstream end of the waveguides to re-enter light signals into the network of waveguides unless the switch is set to release said light signals to an associated electronic device.

6. An arrangement as claimed in claim 2 wherein the substrate is lithium niobate.

7. An optical interconnection arrangement substantially as hereinbefore described with reference to the accompanying drawings.