GB2172174A

GB2172174A - Switching systems

Info

Publication number: GB2172174A
Application number: GB08605574A
Authority: GB
Inventors: Dr Alan Michael Hill
Original assignee: British Telecommunications PLC
Current assignee: British Telecommunications PLC
Priority date: 1985-03-07
Filing date: 1986-03-06
Publication date: 1986-09-10
Also published as: GB8505942D0; GB8605574D0; GB2172174B

Abstract

A bi-directional, rearrangeable switching system is described in which the number of switching points (A-H) required is reduced close to the theoretical minimum. For an N port network, where N is a power of 2, the number of switching points required by systems according to the invention is (N/2)log2(N/2). Each port P1-P8 is connectable to each other port, and in a modification, a port may be selectively connected to itself. Each switching point (A-H) may connect corresponding inputs and outputs directly of via crossed connections. <IMAGE>

Description

SPECIFICATION Switching systems The invention relates to switching systems, for example systems for switching optical signals.

Communication networks with more than two access points usually require the facility for selective point-to-point connection between these access points. For example, in a telephone network each telephone subscriber expects to be able to communicate with any other subscriber, usually without any interference to or from simultaneous communication over the same network between third party subscribers.

Even for a moderate number of access points, simply the facility for connecting any two access points across the network in itself requires a comparatively large number of switching points. With the additional requirement that communication across the network should be nonblocking, the number of necessary switching points increase even further, because "non-blocking" means, in essence, that any given switching point can be by-passed via one or more alternative switching points.

In contrast to this it is desirable to reduce the cost of the network by reducing the number of switching points.

In respect of conventional, i.e. electrical, telephone systems, extensive analyses as to the required number of switching points for a given number of access points can be found, for example, in: C.Clos, "A Study of Non-Blocking Switching Networks" BSTJ, Vol. 32, pages 406-424 (1953); M.C. Paull, "Reswitching of Connection Networks" BTSJ, Vol. 41, pages 833-855 (1961); and V.E. Benes, "On rearrangeable Three-stage connecting Networks", BSTJ, Vol. 41, pages 1481-1492 (1962).

The work of these three authors allows accurate comparisons between network designs, and provides guidance on how conventional switching networks. can be optimised so as to employ a minimum number of switching points.

As previously mentioned, both traditional switching structures and the non-blocking structure discussed in the aforementioned articles, require large numbers of switching points for a given number of access points.

In respect of optical networks attention has now turned to beam steering using holographic techniques, as discussed for example by P. Gravey and J. le Rouzic, "Optical Switching Technologies for High Capacity Exchanges", ISS 1984, Florence May 1984, Session 41A, Paper 5.

These techniques however require much development work.

In accordance with one aspect of the present invention, a switching system comprises N communication terminals connected to respective ports of a rearrangeable switching network, characterised in that the network is operable bidirectionally whereby any of the N terminals can have two way communication with any one of the other N-1 terminals.

In accordance with a second aspect of the present invention, we provide a method of bidirectionally operating a system according to the first aspect of the invention.

It has already been shown, as discussed above, that a significant reduction in the number of switching points can be achieved by using a rearrangeable network. We have discovered that further significant reductions can be achieved with operating a rearrangeable switching network bidirectionally.

In this specification, a switching point is herein defined as a switch having four ports, the switch being operable to connect pairs of the ports whereby any port may be connected with one of at least two other ports. Thus, in the case of optical signals switching points may include lithium niobate switches and Afi switches.

We are primarily concerned with systems for switching optical signals but the systems described are equally applicable to the switching of non-optical signals. In these latter cases the switching points could be constituted by electrical gates. For example, systems according to the invention may be used in electronic and electrical space switching networks.

In general, systems according to the invention may be applied in telephone or other communication systems, computers, e.g. data transfer networks in multiprocessor computer systems, data sorting systems, self-repairing multiprocessors and cellular arrays.

A rearrangeable network exhibits blocking, but blocked calls can be connected by rerouting some existing connections through the network, thus making room for the new connection.

Unblocking can be implemented in two ways: 1. By employing very fast-switching switching points, so that transient and crosstalk effects are unnoticed in the relatively lower signal bandwidth, or by synchronising the switching with digitised signals.

2. By employing two relatively slowly switching rearrangeable networks in parallel, but provide ing fast or synchronised switching between them.

Preferably, the system is arranged such that no part is normally connectable back to itself. In other words a terminal is not allowed to communicate with itself. This requirement enables a further reduction in the number of switching points which need to be used.

If a self-monitoring i.e. self-connection facility is required, the N port network may further comprise N/2 additional switching points arranged to enable ports selectively to be connected back to themselves. This arrangement also enables idle ports not to be connected to other idle ports.

We have found that when N is a power of 2 it is possible in an N port network to reduce the total number (C) of switching points to a value substantially equal to (N/2)log, (N/2).

In one example, the system comprises a first subsidiary network connected to N/2 ports and being arranged to connect any of the N/2 ports to any of N/2 intermediate ports; N/4 central switching points each connected to a respective pair of intermediate ports of the first subsidiary network; and a second subsidiary network connected to the other N/2 ports and having N/2 intermediate ports which are connected in respective pairs to the central switching points, the arrangement of the second subsidiary network and the central switching points being such that any of the N/2 ports connected to the second subsidiary network can be connected to any of the other (N/2)-1 ports via the central switching points, the central switching points also being arranged selectively to return signals from the first subsidiary network to the first subsidiary network, or to connect intermediate ports of the first and second subsidiary networks together whereby any of the N ports can be connected with any of the remaining N-l ports.

In this example, the N port network is divided into two subsidiary portions connected together by a number of central switching points. The first subsidiary network is relatively complex in enabling any of the N/2 ports to be connected to any of the N/2 intermediate ports. The second subsidiary network is less complex, and indeed for an eight port network, may simply comprise two switching points. The second subsidiary network, however, has only a limited choice of central switching points to which it connects pairs of its N/2 ports.

A much simpler system from that just described comprises an N/2XN/2, rearrangeable subnetwork; and (N/2)-1 additional switching points connected to (N/2)-1 pairs of the N ports, each switching point having two outputs connected to either side of the sub-network, the additional switching points being arranged such that any of the ports of the (N/2)-1 pairs may be connected to either side of the sub-network, the remaining two ports being connected one to each side of the sub-network.

Preferably, the sub-network has substantially (N/2)Iog2(N/2)-N/2+ 1 switching points where N is a power of 2. This results in a minimum number of switching points being used.

Some examples of switching systems in accordance with the present invention will now be described and compared with conventional examples with reference to the accompanying drawings, in which: Figure 1 illustrates graphically a comparison between the numbers of switching points in different prior art systems and in some systems according to the invention; Figure 2 illustrates a four port switching network for use in one example; Figure 3 illustrates one example of an eight terminal bi-directional rearrangeable system according to the invention; Figure 4 illustrates a second example of an eight terminal bi-directional rearrangeable system according to the invention with the terminals omitted; Figure 5 illustrates a sixteen terminal system based on the principle of Fig. 4 with the terminals omitted;; Figure 6 illustrates an example of an eight terminal system according to the invention with the terminals omitted based on a different principle from that shown in Fig. 4; Figures 7 to 9 illustrate three examples of six terminal systems with the terminals omitted; and Figure 10 illustrates a modification of the Fig. 6 example.

Fig. 1 illustrates the number of switching points required for various alternative forms of unidirectional and bidirectional switching networks. A very basic network is represented by a line 1 in Fig. 1. The optimum non-blocking network proposed by Clos as described above has a number of switching points represented by a line 2 in Fig. 1.

C. E. Shannon, "Memory requirements in a telephone exchange" BSTJ, Vol. 29, pages 343-349 (1950) and V. E. Benes, "Optimal rearrangeable multistage connecting networks", BSTJ, Vol. 43, Pages 1641-1656 (1964) showed that a two-sided (undirectional transmission), rearrangeable network requires at least 2Nlog2N memory elements, (e.g. relays), for N a power of 2. He achieved this with 4(2N-l)log,N switching points (relay contacts).

V. E. Benes, "Optimal rearrangeable multistage connecting networks", BSTJ, Vol. 43, pages 1641-1656 (1964) showed that the number of switching points could be reduced to 4N(log2N/2). This is plotted by a line 3 in Fig. 1. He also showed that the optimal rearrangeable networks (having minimal cost, defined as the number of switching points per terminal) have the general properties: i) As many stages as possible.

ii) Switches as small as possible.

iii) Largest switches must be in the middle stage.

Theoretically the smallest number of two-state switching elements (switching points) required in a unidirectional switching network is of course log.2 (N!), Because N! is the number of different configurations between N inputs and N outputs.

A Waksman, "A permutation network", J.Assoc. Computing Machinery, Vol. 15, pages 159-163 (1968) and L. J. Goldstein and S. W. Leibholz, "On the synthesid of signal switching networks with transient blocking", IEEE Trans Electronic Computers EC-le, pages 637-641 (1967) and W. H. Kautz, K. N. Levitt and A. Waksman, "Cellular interconnection arrays", IEEE Trans Computers, Vol. C-17, pages 443-451 (1968) showed that Benes' network is essentially one form of an inter-connection network requiring [NIog2N-N+ 1] 2-state, double-pole, doublethrow, reversing switches (2X2 changeover switches) (see line 4 in Fig. 1).Since switching points in integtrated optics are of this type, this is the minimum number of switching points that would be required in a uni-directional, rearrangeable NXN switch implemented in integrated optics.

The use of bi-directional, rearrangeable switching systems enables a further reduction in the number of switching points to be achieved. In one series of examples, systems are built up using a four-port switching network forming a basic building block which is operated bidirectionally. This is illustrated in Fig. 2. The building block has four external ports a, b, c, d and a pair of 2-state switching point switches, 5, 6. The ports a, b, are connected to the switching point switch 5, and the ports c, d, are connected to the switching point switch 6. The switching point switches 5, 6 are connected together via intermediate ports e-h.

In one state, the switching point switch 5 connects the ports a, b, together and connects the ports e, f, together. In the other state, the ports a, f, are connected together, and the ports e, b, are connected together. In a similar way, the switching point switch 6 can connect either the ports c, g, and d, h, together or the ports c, h, and d, g.

It should be appreciated that since the building block shown in Fig. 2 can be operated bidirectionally, it is possible for any of the main ports a,- d, to communicate with any of the other main ports by suitably actuating the switching point switches 5, 6. Table 1 below indicates the state of actuation of each of the switching points 5, 6 and the resulting connections.

TABLE 1 Cross-point Cross-point Connected Switch 5 Switch 6 Ports 0 O a-b; c-d 1 0 a-d; b-c 0 1 a-b; c-d 1 1 a-c; b-d In this Table, "0" means that the switching point switch connects ports in parallel (for example, a-b, and e-f for the switching point 5) while "1" indicates that the switching point switch connects ports in a cross-wise manner.

It should be noted from the table that there are two ways in which the connections a-b, and c-d- can be achieved. Thus only three configurations are required in practice. Since only two connections (conversations) can exist simultaneously, and since for a new connection to be made both existing connections must first cease, the network shown in Fig. 2 is, in fact, nonblocking.

Fig. 3 shows a bi-directional rearrangeable system including six building blocks 7 of the form shown in Fig. 2. This eight-port switching network does possess blocking, and must be reconfigured sometimes to allow a new connection to be made.

Each port of the network is connected to a respective terminal T1-T8 each of which can receive and transmit optical signals. In the remaining examples the networks only are illustrated with the terminals omitted for simplicity.

The number of switching points shown in Fig. 3 is twelve. In fact, it has been found that this cen be reduced to eleven. However, the minimum number of switching points which are theoretically required for a bi-directional rearrangeable network is Iog2[(N-1)(N-3)(N-5).. .]. For N=8 this suggests that just seven switching points should be sufficient.

An alternative eight port switching network in which the number of switching points has been reduced to eight is illustrated in Fig. 4. This is a tremendous reduction (by more than a factor of two) on the conventional number of switching points required in a uni-directional rearrangeable network.

The network shown in Fig. 4 can be considered to be divided into three portions. A first subsidiary network 8 having four of the eight ports, a second subsidiary network 9 having the other four of the eight ports, and two central switching points 10, 11 connecting the two subsidiary networks together.

The second subsidiary network 9 is, in fact, a four port building block as shown in Fig. 2. The subsidiary network 9 is connected to the switching points 10, 11 in such a way that signals from any of the ports labelled P5-P8 in Fig. 4 can be returned to one of the remaining ports P5-P8 via the switching points 10, 11. Alternatively, the signals can be passed through the central switching points 10, 11 in pairs to the first subsidiary- network.

The first subsidiary network 8 has four switching points labelled A, B, D and F connected together that any of the four ports labelled P1-P4 can be connected to any of four intermediate ports i-l of the first subsidiary network 8. Since these ports i-l are connected in pairs to either side of the central switching points 10, 11, any of the four ports P1-P4 can be connected to either of the central switching points 10, 11 and to either side of the central switching points. This allows full access for the four ports P1-P4 to the four ports P5-P8. In addition, communication between the four ports P1-P4 is also possible by making use of the central switching points 10, 11.

This is a useful network since the second subsidiary network can have a relatively simple form, the first subsidiary network providing the necessary complexity to allow full communication between all the ports.

It will be seen from Fig. 4 that the number of ports is (N/2)10g2(N/2). This function is plotted as a line 12 in Fig. 1. This line is very close to the theoretical minimum number of switching points required, as defined above.

It will now be described how in the Fig. 4 example, an existing connection (or call in the case of a telephone system) can be rerouted to enable a blocked connection to be made. Assume that the following connections exist: P1-P7, P2-P6, P3-P5, and P4-P8. To achieve these connections, the states of the switching points A-H may be as set out in table 2 below.

TABLE 2 A B C D E F G H 1 1 0 0 0 1 1 1 The figures "0" and "1" have the same meanings as in Table 1.

Now assume that the connections P1-P7 and P2-P6 are broken independently and then P1 wishes to connect with P2. This also means of course that P6 must connect with P7. In the present configuration, these new connections are blocked because the remaining connections P3-P5 and P4-P8 employ both central switching points 10, 11 to transfer connections between the two subsidiary networks 8, 9. In order to allow the new connections, the two existing connections must be allocated to the same central switching point leaving the other central switching point free to return the new connections P1-P2 and P6-P7 on the same subsidiary networks. This can be achieved, for example, by rerouting the connection between ports P3-P5 from switching point 10 to switching point 11. The new connections can now be made, and the required switching point states are then as set out in Table 3 below.

In this Table, "-" means that the connection state of the switching point is irrelevant.

TABLE 3 A B C D E F G H - 1 1 0 1 0 0 1 The network shown in Fig. 4 can be used as a basis for much larger networks, and as an example, Fig. 5 illustrates a sixteen port network which employs twenty four switching points.

This network has a first subsidiary network 8', a second subsidiary network 9', and four central switching points 13-16. The properties of the first and second subsidiary networks 8', 9' are the same as in the Fig. 4 example. The second subsidiary network 9' is in fact an N/2 port (eight port in this case) network as in the Fig. 4 example. In this case, however, N/4 internal links are brought out to the central switching points 13-16. Alternatively, the second subsidiary network 9' could consist of an N/2 port network of the type shown in Fig. 6 to be described below. Once again, the total number of switching points is very close to the theoretical minimum.

Another example of an eight port network is illustrated in Fig. 6. This also requires eight switching points. The eight ports are labelled P1-P8. Three pairs of ports P2,P3; P4,P5; and P6,P7 are connected to respective switching point switches 17, 18, 19. The output ports of these switching points 17-19 are connected to either side of a 4X4 rearrangeable network 20.

This network 20 has five switching points 21-25. The network 20 is a conventional N/2XN/2 rearrangeable network. The remaining ports P1, P8 are connected to opposite sides of the network 20.

Fig. 7 illustrates a six port switching network based on the principles of the Fig. 6 example.

This example comprises two switching points 26, 27 connected to the ports P2, P3, P4, and P5 and a 3X3 (N/2Xn/2) unidirectional rearrangeable network 28. The network 28 is connected to the switching point 26, 27 and directly to the remaining ports P1, P6.

Two alternative structures for switching systems having six ports are illustrated in Figs. 8 and 9 both of which need only five switching point.

In the Fig. 8 example, ports P1-P4 are connected in pairs to two switching points 29, 30, one of the remaining ports from each of which is connected to a common switching point 31 while the other is connected to a respective switching point 32, 33. The switching points 32, 33 are connected together and to the switching point 31 while the port P5 is connected to the switching point 32 and the port P6 to the switching point 33.

The Fig. 9 example is based on the building block shown in Fig. 2 as indicated by the dashed line 34. A pair of central switching points 35, 36 are connected between the building block 34 and a switching point 37 connected to the ports P1, P2.

It is believed that the operation of these systems shown in Figs. 7 to 9 is self-explanatory.

The networks illustrated in the drawings can be used in a variety of applications, but in particular they may be used for switching optical signals in, for example, a telephone network. In this case, the switching points are 2-state optical switching points of a conventional kind.

When devising optimum bi-directional, rearrangeable networks, two principles should be borne in mind.

1) No port should be allowed to connect back to itself via the switching network. This facility (e.g. for self-monitoring) can be provided separately if required, but more efficiently.

2) Idle ports are always connected to other idle ports. In the case of optical signals, light sources should therefore either be turned off (or left unmodulated) electrically, or dis-connected in pairs by an additional N/2 switches. These switches cquld also provide the self-monitoring facility.

In some cases it may be desirable not to allow idle ports to be connected to each other. This can be achieved by placing a switching point somewhat along their common path. Because each additional switching point will allow two idle ports to be disconnected, there would only need to be N/2 additional switching points which is to provide this facility for the whole N-port network.

There is a simple rule for finding suitable links in which to place these additional switching points. They must all be positioned in links which cannot be physically connected to any other links containing any other such additional switching points. This will ensure that every additional switching point disconnects a different pair of idle ports. Fig. 10 illustrates the Fig. 6 example with four additional switching points 38-41.

A self-monitoring facility for the idle ports can be provided by placing a reflector (for example a mirror) in the other two ports of the additional switching points so that light can be reflected back along the same path to the same idle port.

Claims

1. A switching system comprising N communication terminals connected to respective ports of a rearrangeable switching network, characterised in that the network is operable bidirectionally whereby any of the N terminals can have two way communication with any one of the other N-1 terminals.

2. A system according to claim 1, adapted to switch optical signals.

3. A system according to claim 1 or claim 2, the network being arranged such that no port can normally be connected back to itself.

4. A system according to claim 1 or claim 2,, further comprising N/2 additional switching points arranged to enable ports selectively to be connected back to themselves.

5. A system according to any of the preceding claims, wherein when N is a power of 2 the total number of switching points is substantiall equal to (N/2)10g2(N/2).

6. A system according to any of the preceding claims, the system comprising a first subsidiary network connected to N/2 ports and being arranged to connect any of the N/2 ports to any of N/2 intermediate ports; N/4 central switching points each connected to a respective pair of intermediate ports of the first subsidiary network; and a second subsidiary network connected to the other N/2 ports and having N/2 intermediate ports which are- connected in respective pairs to the central switching points, the arrangement of the second subsidiary network and central switching points being such that any of the N/2 ports connected to the second subsidiary network can be connected to any of the other (N/2)-1 ports via the central switching points, the central switching points also being arranged selectively to return signals from the first subsidiary network to the first subsidiary network, or to connect intermediate ports of the first and second subsidiary networks together whereby any of the N ports can be connected with any of the remaining N-1 ports.

7. A system according to any of. claims 1 to 5, the system comprising an N/2XN/2, rearrangeable sub-network, and (N/2)-1 additional switching points connected to (N/2)-1 pairs of the N ports, each switching point having two additional ports connected to either side of the sub-nefwork and being arranged such that any of the ports of the (N/2)-1 pairs may be connected to either side of the sub-network, the remaining two ports being connected one to each side of the sub-network.

8. A network according to claim 7 wherein the sub-network has substantially (N/2)log,(N/2)-N/2+ 1 switching points where N is -a power of 2.

9. A bi-directional, rearrangeable, switching system substantially as hereinbefore described with reference to any of the examples shown in Figs. 3 to 10 of the accompanying drawings.

10. A method of bidirectionally operating a system according to any of the preceding claims

11. A method of operating a bidirectional rearrangeable switching system substantially as hereinbefore described with reference to any of the examples shown in Fig. 3 to 10 of the accompanying drawings.