GB2178271A - Statistical multiplexing - Google Patents

Abstract

In a variable bit rate switching network which can handle both speech and data, and which uses statistical multiplexing, the activity state of the channels within the multiplex are conveyed by bit map signalling. This uses a burst of signalling information sent during the cycle of the system prior to that to which it relates, thus minimizing delays in the system. In the system shown, two levels of multiplexing are used, with separate switch trains in each switching node. One or more low order multiplexes after demultiplexing from a high order multiplex are switched by the low order multiplex switch train, and one or more channels after handling by the low order switch train are remultiplexed into the high order multiplex.

Description

SPECIFICATION Statistical multiplexing This invention relates to an automatictelecommunication system, in which so-called statistical multiplexing is used.

Statistical multiplexing is a form of multiplexing in which the information rate of the transmission medium ontowhich traffic is multiplexed islessthan the sum ofthe information ratesofthe (input) channelsthatare being multiplexed. Ittherefore depends on the statistical nature oftraffic origination (i.e. it in effect applies Erlang'sformula), and implies that a given input channel does not have a predetermined position in the multiplex structure, contrary to what is generally the case for conventional multiplexing.

Because channel assignments are not fixed, a statistical multiplexer needs to provide signalling information to the demultiplexeron the current state of channel allocations. Thus signalling information can be included with the data which is then sent as a "packet", or it can be sent separately. The term "statistical multiplexing" is sometimes used in the sense of covering both possibilities, but in the current context is only intended to cover the separate signalling case.

According to the invention there is provided an automatic telecommunication system, which includes switching nodes serving time division multiplex systems each ofwhich itself serves a number of users ofthe system, in which the bandwidth of some at least ofthe channels allocated to the users of the system can be varied by varying the number of bits in the TDM cycle allocated to said users, in which for each multiplexframeofasaid multiplexsystem there is a burst of signalling information which includes bits which together form a bit map indicative ofthe allocation ofthe intelligence-conveying channels to the respective system users, the allocation of said channels to some at least of the users being varied automatically in accordance with the current bandwidth requirements ofthe respective users.

An embodiment of the invention will now be described with reference to the accompanying drawing, in which Fig. lisa highly simplified block diagram of a switching node embodying the invention.

Figs. 2 and 3 are explanatoryttming diagrams.

With thevaried services which are expected of modern telecommunication systems a considerable degree offlexibility is needed. Th us these services range from speech, and broadband services such as video,to high speed and low speed data. Hence variable bit rate systems are fashionable.

Time division multiplex of the so-called statistical multiplexer (SM) type is used in the system to be described herein. Such a system uses both low order and high order multiplexes, and we assumethata low order multiplex has a line rate of 2.048 Mb/s, while a high order multiplex might have a line rate of 34 Mb/s, or more. A low order multiplex can befitted into, or "super-multiplexed" into, higherorder multiplexes on the basis ofan integral numberofhigherorder multiplex channels.

In the low order multiplex, the "basic" frame rate is 8KHz (125us), with a "super-frame" rate of 250 Hz (4 msecs), channel bandwidth being 8 Kb/s. Eight channels can be used, inter alia, as a single speech circuit, but the bandwidth can be allocated in integer multiples of 8 Kb/s, and also in binary sub-multiples of 8 Kb/s.

Signalling is handled in a bit map mannerwhere bursts of information have to be conveyed, using a signalling framewhich would, in effect, "tell"the recipience of the bit stream how the various channels in the superframe are allocated. Such a signalling burst may conveniently be sent at a superframe rate, using a "header" of bits in each complete superframe dedicated to signalling. Sending signalling informa tion at otherthan superframe rate is possibly but is generally less convenient and transmision cutthe superframe rate is assume unless otherwise stated.

The low order multiplex has an unpacketised and a packetised capability, with packets sent transparently.

Message signalling, such as CCITT No. 7, is used for call processing information.

In the case of a higher order multiplex, basic frame and superframe rates are as forthe low order multiplex, with a channel bandwidth in one example of 64 Kb/s. The bandwidth can be allocated in integer multiples of 64 Kb/s, and an integral numberofthe 64 Kb/s channels is dedicated to the low order multiplex.

The same burst signalling arrangement is used as in the low order multiplex with the low brder multiplex carried transparentty in channel. As in the case ofthe low order multiplex, there is an unpacketised and a packetised capability.

With such a system there is no delay for speech or video sent unpacketised, and packetisation is used for asynchronous and non-network synchronised data.

The system givesfastsignalling, which gives excellent potential for bandwidth re-allocation of traffic pattern changes and overload. There is also a simple interface forvoice traffic to the old network.

Where traffic conditions warrant, the bandwidth can be divided between statistically multiplexed and packettraffic, with either a fixed or a variable partition in a basic frame. Packets might be fragmented, i.e. not confinedto a frame, buta connectionless ordatagram mode of operation could be allowed, e.g. using "fast select" packets on X25 networks.

At this point we refer to Fig. 1,which shows a system node using the above-discussed arrangements but in which the higher-order multiplexfunctions at 140 Mbits/secs. Here the higher order multiplex is referred to asthe broadband section andthe lower order multiplex as the wideband section. In this node the various elements are as follows: TS 1, of which there are two, are time switches which can handle 2000 channels each of 64 Kb/s, eight bits wide.

TS2, of which there are also two, are time switches for2000channels each of 8 Kb/s, one bit wide.

SS 1, is a space switch, 128 x 128 crosspoints, each an eight bit crosspoint.

SS2, is a space switch, 128 x 128 crosspoints, each a single bitcrosspoint.

HLDMisa high level demultiplexer,which demultiplexes 64 Kb/s channels out ofthe 140 Mb/s bit stream for switching purposes. As will be seen, at least one such channel goes to the LLDM, below.

LLDM is a low level demultiplexer, which demultiplexes the 2 Mb/s bit stream, and also the 64 Kb/s bit stream from HLDM to obtain the 8 Kb/s channels which are then switched in the low order multiplex.

HLM is a high level multiplexer, which multiplexes 64 Kb/s channels which have been handled by TS1-SS1-TS1, plus one or more 64 Kb/s channeisfrom LLM, below, onto the 140 Mb/s bit stream.

LLMis a low level multiplexer which multiplexes 8 Kb/s channelswhich have been handled by TS2-SS2 TS2 on to a 2 Mb/s bit stream.

Notethata node may handle morethan one2Mb/s multiplex, and also more than one 140 Mb/s multiplex.

Note also that a system in which the higher order multiplex is a 34Mb/s onewould be generallysimilar in essence to the system shown.

Initial call set up, end of call clearance, and "register recall" use messages, as in a system such asthat of ourApplication No. 2139852A (MT. Shortland 1).

However, the signalling strategy whereby the trans mitting end of a link "tells" the receiving end about channel allocation uses the repeated transmission of a "bit map" of the status ofall potential users.This assumes a channel allocation algorithm based on that status. The receiving end responds to changes of status occurring on consecutivesignallingframesof bit map information.

With this bit map approach to signalling, when call activation occurs, a user is assigned a number of bits, i.e. a certain bandwidth, appropriate to his maximum channel demand. Thus, for example, the far end is told that userA has been assigned bits 5-12 on the bit map, and that his destination address is B. Ideally the bit map is sent as aframe header, butinfact it has been found that due to the operations needed at a switching node it is necessary to send the bit map at some point in the previoussuperframe.

Fig. 2 shows an example, in which each bit is regarded as being one channel. Here user G has been assigned bit 1 and is active, and user H has been assigned three channels, 2-4, and is inactive. User A currently wants 4 of his 8 assigned channels (bits).

While these conditions exist, G isallocated channel 1, H is allocated no capacity, and A is allocated channels 2to 5. If H becomes active and needs all his three channels,then A is switched to the use of channels 5 to 8. The ratio of the bits on the bit map to the number of channels plus fractional channels is the signal interpolation (SI) gain, and is set, on the basis of knowledge of traffic behavious, to give a desired level of loss and/or delay.

Note thatthe signalling rate need not be at the superframe rate, it may be made largerorsmallerto correspond with the on-off speech pattern rate.

When a fractional channel is to be allocated, a sub-channel is specified in the initial call set up procedure, and the far end keeps a record of that sub-channel in the superframe and only looks for and handles information in the relevantframes of the superframe. The user's activity bit, which is unique to him, then signals his current activity state.

We now discuss overload, a starting assumption being thatthe product ofthe number of users and their peak data rate is limited. When this limit is reached, new calls are refused. This applies across the network, transmit nodes feeding back to originating nodes commands to refuse service even when an originating node is not itself overloaded. Any user with priority service who initiates a call during overload causes an equivalent lowprioritycall to be discontinued. In such a system overload is of a random nature and of short duration.

The preferred technique for packetised data is to use delay in overload conditions, with data delayed to provide extra temporary bandwidth for speech. High speed data synchronised to the network clock should not, however be delayed to give preference to voice traffic. For speech, some bandwidth reductions, e.g.

by bit stealing, may be used as may be some limited amount oftime clipping at eitherfront end or mid spurt. In general, of course, speech should not be delayed.

For real time video information, an initial assumption is of a variable bit rate system in which transient changes in picture content demand an increased data rate. One method of overload control atthe periphery would be to delay changes, but at a transit node it may be necessary to throw information away.

Users afthe system connected via high speed links will need to communicate for low speed services with users who only have access via low speed links. Hence the 2 Mb/s multiplex structure is included in the high speed structure, either en bloc or on a channel by channel basis. Inthefirstcase a 2Mb/s complex is regarded as onefixed channel in a high speed structure, while in the otherthe 2 Mb/s structure is broken down into its constituent channels and interleaved with broadband service channels.

However, a compromise solution has been selected in which the less than 2 Mb/s services are allocated a number of channels in the broadband multiplex concomitant with demand. This assumes that broadband channels are less then 2 Mb/s wide. Such channels are given a fixed portion ofthe total frame, and signalling can either share a common channel/ bandwidth outband, orthe signalling for the lower speed services can be containedwithinthe information space allotted to in, inband.

We now give an example of a statistical multiplexer framing and signalling structures based on the preceding discussion.

Forthe 2.048 Mb/s low order structure, we assume that Channel 0 is usedforsynchronisation, but that Channel 16 is free for use as desired. The basic channel is 8 Kb/s, so Channel 0 is 8 channels (i.e.

sub-channels) in the present terminology, 8 Kb/s being chosen for the bandwidth reduction flexibility it allows with 64 Kb/speech. The lowest sub-channel rate is 250 b/s, which makes the superframe duration 4 ms. This allows 75 and 150 b/s data to be sent packetised on the 250 b/s sub-channel. The signalling frame rate is also 4 ms., as this gives adequate response to burst events. The frame rate duration is 125 uses, so speech is undelayed and is sent sample by sample to give a conventional interface to the old network.

The relatively low basic channel rate has a disadvantage th at 8-bit encoded PCM speech needs 8 channels, which is expensive is signalling channel bandwidth using the bit map signalling method. If 8-bit speech channels predominate, it may be preferred to use, Fig.

3, a 4-bitword for each channel which uses an integral number of channels, assuming that channels when allocated are packed according to a preset algorithm.

One bit represents the circuit's activity state and the otherthreethe numberofchannels in use, less one.

Thus an active 8-channel circuit has all four bits set.

Sub-channels are each represented by one bit. The integral channels as coded fill the signalling frame from thefront end and the sub-channels from the rear, normally with an unused gap in the middle (or nearthe middle). Thus where 45 simultaneous speech and 15 simultaneous users are allowed, we end up with a number of the order of 212 bits per superframe, or 7 bits per frame. This uses 7 ofthe 248 available channels.

Call set up, clear and recall (i.e. call processing) are signalled by messages using one or more permanent ly assigned channels with an ability to use upto 8 channels on a statistical basis to deal with peak demands.

The signalling information in this example is transmitted every of 4 ms, and as already indicated it is preferred to send the bit map information in the superframe prior to the one to which it relates. This allows time for the necessary computation and establishmentofcross-office pathstooccur.

Forthe higher order multiplex 64 Kb/s channels are used, resulting in the lower order multiplex being assigned up to 31 channels in the higher order multiplex, lowordersignalting being included. The channels assigned forthe low order multiplex can be set semi-permanently active in the higher order multiplex, i.e. treated as a private wire at the higher level.

We now consider the synchronisation methods used where a number of statistically multiplexed links are connected in tandem. Thus when oniy one links is involved, changes in condition will occur at different times, which is aggravated when two or more links are connected intandem, since link delays and trans-node delays are involved. This is dealt with by padding the delay at a receiving end to either a full superframe, or in more difficult cases to two superframes.

In order to reduce cross-network delay and the cost of storage associated with frame alignment, albeit at the cost of a somewhat increased processing load, instead of updating the burst information once per superframe for all channels in use, that information is updated in respect of a few of said channels updated in sequence in each basic frame. Thus with 128 such channels,f6urare updated perframe and the bit map is "distributed" between basic frames in a superframe. However, this distributed information is still sent in the previous superframe.

Claims (12)

1. An automatic telecommunication system, which includes switching nodes serving time division multiplex systems each ofwhich itself serves a number of users ofthe system, in which the bandwidth of some at least ofthe channels allocated to the users ofthe system can be varied by varying the number of bits in the TDM cycle allocated to said users, in which for each multiplex frame of said multiplex system there is a burstofsignalling information which includes bits which together form a bit map indicative ofthe allocation of the intelligenceconveying cha nnels to the respective system users, the allocation of said channels to some at least of the users being varied automatically in accordance with the current bandwidth requirements of the respective users.

2. Asystem as claimed in claim 1, in which when a call is set up the number of channels allocated to the caller isthe maximum availableto that caller, in which when the progress of the call indicates that a less number of said channels is needed, the number of channels allocated to that caller is reduced, and in which when a said reduction in the number of channels is effected the channels allocated to calls later in the basic frame are moved forwards in the frame, so that the basicframe tends to fill up from its front end.

3. Asystem as claimed in claim 1 or 2, in which the burstofsignalling information in respect of one superframe is sent during the preceding superframe to give the nodes ofthesystem time to set up and adjusttrans-node paths.

4. Asystem as claimed in claim 1,2 or3, in which sorneatleastofthechannelsusedforcommunica- tions can each be split into two or more narrower bandwidth sub-channels, and in which when such sub-channel allocation isto be used the initial call setting procedure specifies the sub-channel to be used.

5. A system as claimed in claim 4, in which integral channels as taken into use are allocated commencing fromthefrontendofthebasicframewhilesub- channels as taken into use are allocated commencing from the rear end of the basic frame.

6. A system as claimed in anyone of claims 1 to 5, in which call setting and call clearing messages are conveyed via one or more system channels dedicated to such messages.

7. An automatictelecommincation system, which includes switching nodes serving low order multi- plexes each ofwhich can handle a number of communications channels and high ordermultiplexes each ofwhich can handle a numberof communication channelsgreaterthana low order multiplex, said nodes being interconnected via high order multi- plexes or both high and low order multiplexes, in which each said node includesfirstswitching means for handling high order multiplexes, in which within each said node's switching means the multiplex or multiplexes being handled are switched afterthey have demultiplexed, in which each said node also includes second switching meansforhandlinglow order multiplexes, in which within each said node's second switching meansthe multiplexes being handled are switched after they have been demultiplexed, in which one or more channels demultiplexed from a said high order multiplex are switched by the second switching means instead of by the first switching means atthe node, in which one or more channels which have been switched by the second switching means are multiplexed into a high order multiplex for transmission from the node, in which the bandwidth of some at least ofthe channels allocated to the users ofthe system can be varied by varying the numbers of bit in the TDM frame allocated to the users, and in which for each multiplexframe of a said multiplex there is a burst ofsignalling information which includes bits which togetherform a bit map indicative ofthe allocation of the intelligence-conveying bits to the respective system users, the allocation of the bits to some at least ofthe user being varied automatically in accordance with the bandwidth requirements ofthe respective users.

8. Asystem as claimed in anyone of claims 1 to7, in which the information contained in said bit map is updated in a numberofsuccessive basicframes,the updating in respect of each such frame being in respect of some only of the channelsto which that information relates.

9. A switching node for use in a system as claimed in any one of the preceding claims.

10. An automatictelecommunications switching system of the TDM type, substantially as described with reference to the accompanying drawing.

11. An automatic telecommunications switching node substantially as described and as shown in Fig. 1 ofthe accompanying drawing.

New claims or amendments to claims filed on 5th December, 1985.

New or amended claims:

12. An automatictelecommunication system, which includes switching nodes serving time division multiplex systems each of which itself serves a numberof users ofthesystem, inwhich the band- width of some at least ofthe channels allocated to the users ofthe system can be varied by varying the number of bits in the TDM cycle allocated to said users, in which for each multiplex frame of a said multiplex system there is a burst ofsignalling information which includes bits which togetherform a bit map indicative ofthe allocation ofthe intelligenceconveying channels to the respective system users, the allocation of said channels to some at least ofthe users being varied automatically in accordance with the current bandwidth requirements ofthe respective users, in which when a call is set up the number of channels allocated to the caller is the maximum available to that caller, in which when the progress of the call indicates that a less number of said channels is needed,the number of channels allocated to that caller is reduced, and in which when a said reduction in the number of channels is effected the channels allocated to calls later in the basicframe are moved forwards in the frame, so that the basic frame tends to fill upfrom its front end.