GB2312137A - CATV for video/telecomms - Google Patents

Abstract

A distribution network comprises a distribution hub 120 and a shared transmission medium 130 extending from the hub to a plurality of terminals T1-T80 for distributing video services to the terminals and for carrying telecommunications signals. Preferably upstream and downstream transmissions of a connection are paired, receipt of a downstream transmission triggering an upstream transmission. On-going connections can be repositioned by varying the timing of the downstream transmission. Telecommunications signals are transmitted in both the downstream and upstream directions over a common portion of the bandwidth of the bus. Preferably signals are unmodulated. The upstream and downstream transmissions can be paired, separated by a time period which is known by terminals T1-T80 so as to minimise collisions. Connections of different lengths, such as speech and data connections, can be supported together.

Description

DISTRIBUTION OF COMMUNICATIONS SERVICES OVER A DISTRIBUTION NETWORK TECHNICAL FIELD This invention relates to distribution networks having a transmission medium which is shared by a plurality of terminals and to the provision of communications services to terminals in such networks.

BACKGROUND OF THE INVENTION Broadband distribution networks, such as Cable Television (CATV) networks are becoming increasingly commonplace. A CATV network typically comprises a headend which has access to a number of video sources such as broadcast television and taped video material. The headend is linked to subscribers via a distribution network, which is often a hybrid combination of optical and coaxial cabling having a treeand-branch configuration. While the networks are primarily used for distributing video services to subscribers, often two-way telecommunications services such as telephony and data are also supplied.

Telecommunications signals may be distributed via a twisted-pair cable laid alongside the coaxial cable, or via the coaxial cable itself with the telecommunications signals occupying a portion of the available cable bandwidth. It is known to time division multiplex the telecommunications signals upon RF carriers, with separate frequency bands being assigned for transmissions in the upstream and downstream directions. It is preferable to supply telecommunications signals over the same cable as the video services as this leads to simpler maintenance of the network.

However, using RF bands requires subscribers to be equipped with RF modems to receive and translate received signals to baseband. -RF: modems considerably increase the cost for a subscriber war. > tfng telecommunications services. Such modems include circuitry to upconvert signals to, and down-convert signals from, the assigned RF carriers which carry the telecommunications information. A system of this type is described in UK Patent Application GB 2,263,041 A.

There is also a desire to supply both voice and data services over the network. An example of a CATV network which carries voice and data, with separate frequency bands being used for upstream and downstream transmissions, is described in an article entitled 'Voice and Data on a CATV Network' published in the IEEE Journal on Selected Areas in Communications, Vol. SAC-3, No. 2, March 1985.

SUMMARY OF THE INVENTION The present invention seeks to provide an improved method of distributing communications services over a distribution network.

According to a first aspect of the present invention there is provided a method of operating a communications distribution network, which network comprises a distribution hub and a transmission medium extending from the hub to a plurality of terminals, the medium being shared by the plurality of terminals, the method comprising: - supporting a plurality of communications connections between the hub and the terminals, each connection comprising a sequence of pairs of a downstream transmission from the hub to a particular one of the terminals and an upstream transmission from that terminal to the hub, which upstream transmission is triggered by the terminal receiving the downstream transmission.

The feature of pairing the upstream and downstream transmissions, such that the upstream transmission is triggered by receiving the downstream transmission, allows the hub to efficiently control activity on the transmission medium. It avoids contention between on-going connections without requiring a structured TDMA channel. It is particularly useful in networks where the shared transmission medium inhibits transmission of signals directly between terminals, such as CATV networks. The method has a further advantage of minimising complexity of terminal equipment, which only has to respond to a downstream transmission at the appropriate time.

Operating a distribution network in this manner also has the advantage of allowing traffic of different types, such as voice traffic and data traffic, to be carried together, in a manner that is difficult on rigidly-structured TDMA systems.

Preferably the connections occur within a series of time-contiguous frames. the method further comprising repositioning on-going connections by varying the position of the downstream transmissions within the frame. The repositioning can be performed in response to termination of connections, so as to maximise the contiguous length of idle time within the frame. This allows the network to accept new connections with minimal delay, and with minimal inconvenience to ongoing connections.

The connections can comprise transmissions of different lengths. For a voice connection the upstream and downstream transmissions will usually be of the same length, but for data connections and other nonsymmetrical services the upstream and downstream transmissions of a connection can be of different lengths. Connections of different types, such as voice connections having relatively short transmissions, and data connections having longer transmissions, can be carried together.

A data connection may carry an ATM cell to link a terminal to a wide area network (WAN).

A data connection for conveying data from a terminal to the hub comprises a downstream transmission and an upstream data transmission. The downstream transmission conveys signalling information for the data connection, and can be used to instruct the terminal to end data transmissions if the medium becomes full.

Preferably a terminal matches a continuous input or output data rate to the terminal (such as voice data) to a varying transmission position within a frame. An elastic buffer can be used.

Preferably there is an access period during which a terminal can request a connection. This can comprise a downstream transmission from the hub and a period during which terminals can reply.

Further aspects of the invention provide a method of operating a distribution hub and a method of operating a communications terminal in a communications distribution network.

Another aspect of the present invention provides a communications distribution network, which network comprises: - a distribution hub; - a transmission medium extending from the hub to a plurality of terminals, the medium being shared by the plurality of terminals; - means for supporting a plurality of communications connections between the hub and the terminals, each connection comprising a sequence of pairs of a downstream transmission from the hub to a particular one of the terminals and an upstream transmission from that terminal to the hub, which upstream transmission is triggered by the terminal receiving the downstream transmission.

According to another aspect of the present invention there is provided a method of operating a communications distribution network, which network comprises a distribution hub and a transmission medium extending from the hub to a plurality of terminals, the medium being shared by the plurality of terminals and having transmissive properties which inhibits transmission of signals between terminals, the method comprising transmitting communications signals in the downstream direction, from the hub to a particular one of the terminals, and in the upstream direction, from that terminal to the hub, as a pair over a common portion of the medium's bandwidth, the transmissions of the pair being separated by a time period which is known by the plurality of terminals, whereby a terminal can determine when the medium is in use by another terminal.

Pairing the upstream and downstream transmissions in this manner has the advantage of minimising collisions between terminals. Such collisions would otherwise occur since a terminal cannot receive the upstream transmissions of another terminal due to the directional properties of the medium. This is a particular problem in a CATV system where the taps which couple signals between a bus and a terminal are directional. This aspect of the invention is useful for networks having a shared transmission medium where terminals access the medium by a carrier-sensing mechanism.

The above aspects of the present invention are explained in the following detailed description in terms of providing telecommunications services over a shared transmission medium which is part of a CATV distribution network. However, it will be apparent that it has application to other kinds of distribution networks employing a shared transmission medium.

According to another aspect of the present invention there is provided a method of operating a cable television (CATV) distribution network, which network comprises a distribution hub and a cable bus network extending from the hub to a plurality of terminals for distributing video services to the terminals and for carrying telecommunications signals, the method comprising transmitting telecommunications signals in both the downstream direction, from the hub to the terminals, and in the upstream direction, from the terminals to the hub, over a common portion of the bandwidth of the bus.

Using a common portion of the cable bandwidth for transmissions in both directions, rather than assigning particular portions for each direction, has the advantage of allowing the network to efficiently accommodate different traffic patterns. Considering some of the different telecommunications services, telephony makes symmetrical use of upstream and downstream capacity while access to the Internet makes asymmetrical use, with short upstream requests for new pages and associated long downstream bursts of that page data. Home working typically involves occasional downstream transfer of data information from a company's file-server and periodic upstream transfers to save the information produced during the day. The proportion of upstream and downstream traffic will vary during the day.

Preferably the baseband portion of the cable bandwidth is used with the telecommunications signals being unmodulated. This overcomes the need for expensive RF modems and simplifies the equipment required at each terminal. Transmissions can use line coding techniques to encode data such that they have a spectral content which corresponds to an available band on the distribution bus.

BRIEF DESCRIPTION OF THE DRAWINGS Embodiments of the invention will be described, by way of example only, with reference to the accompanying drawings, in which: Figure 1 shows a cable distribution network; Figure 2 shows a part of the cable distribution network of figure 1 in more detail; Figure 3 shows the spectral content of an exemplary coding scheme for use in the network of figure 1; Figure 4 shows the structure of a speech packet carried over the network of figure 1; Figure 5 is a table illustrating features of packets carried over the network of figure 1; Figure 6 shows the structure of a data packet carried over the network of figure 1; Figures 7A-7E illustrate typical bus activity according to a first manner of using the network of figure 1; Figures 8A-8E illustrate typical bus activity according to a second manner of using the network of figure 1; Figures 9 and 10 show parts of the network of figure 1 in more detail; Figure 11 shows how connection capacity varies with length of the network; Figure 12 is a state diagram used in illustrating performance of the network.

DESCRIPTION OF PREFERRED EMBODIMENTS Figure 1 shows a cable television distribution network. A headend centre 100 is equipped with various equipment to support video services, such as satellite receivers, terrestrial broadcast television receivers and video players. Headend 100 also has switches and other equipment which is necessary to support a telecommunications service to subscribers. Trunking 105 connects headend 100 to other networks, such as the PSTN. Headend 100 modulates services for transmission over a high-capacity network 110, such as an optical fibre network. One branch of the optical network terminates at a hub 120 which is located close to the group of subscribers who are to be served. Hub 120 is often called an 80-home point (or some similar figure representative of the number of homes which it serves) or an optical network unit (ONU).

Hub 120 includes equipment to convert signals received over the optical network for transmission in eiectrical form over the final part of the path to the subscribers. A coaxial cable network 130 extends from hub 120 to subscriber terminals T1, T2 . . T80. The back-bone of the coaxial cable network is a bus 150. The bus can be a single length of cable or a tree-and-branch arrangement. Each subscriber terminal connects to the bus via a matched passive tap 140, 141, 142, 143 and a relatively short branch cable 160 of the order of tens of metres in length. Taps can be dedicated to an individual terminal or can be shared by a group of terminals, with each terminal in the group coupling to the tap by a short branch cable. A two-way tap 141, four-way tap 142 and eight-way tap 143 are shown.

Each terminal includes a receiver which demodulates video signals carried by network 130 in a known manner. Each terminal also includes a transmitter and a receiver to transmit and receive telecommunications signals between the terminal and hub 120 (which are carried over optical network 110 to the headend.) Figure 2 shows a section of coaxial network 130 in more detail. The taps are arranged such that the total signal power transmitted by hub 120 is substantially equally divided among subscribers. For example, with a total of 80 subscribers connected to bus 150, the first downstream single-way tap ideally divides the power in the ratio of 79/80 continuing along the bus, 1/80 channelled along the branch to terminal T1. This ensures that downstream transmissions So from the hub appear at each of terminals T1 - T80 with similar power levels. Loss LA shown in figure 2 determines the proportion of power that is tapped from bus 150.

In the upstream direction Su, transmissions from each of terminals T1 T80 appear at the hub at similar power levels. However, upstream transmissions from a particular terminal are received by other terminals at greatly attenuated levels. This is because the value of loss LB for the path between bus 150 and branch 160 is high. The level at which a terminal will receive a transmission from another terminal will vary according to the distance by which the terminals are separated, but for all distances this will be a very low level. This effectively prevents terminals from monitoring the activity of other terminals to determine when it is possible to initiate a transmission on the bus, in the manner which is performed in an IEEE 802.3 (EthernetTM) type of system.

A baseband portion of the available bandwidth on bus 150 is reserved for providing a telecommunications service. In some networks the low frequency portion of the cable bandwidth is reserved for carrying a power supply to repeaters which are periodically inserted along the cable path to boost the signal level. Telecommunications signals can be carried in a frequency band separate from that used to supply power.

Taps used in CATV networks ac couple subscriber branches to the bus and generally offer a bandwidth in the range of 5MHz to 600MHz or 10MHz to 1000MHz. Communications traffic therefore uses a balanced line code with no dc content and minimal spectral content in the band below 5MHz. Also, to prevent interference with existing RF channels positioned above 50MHz, the line code should have minimal or no spectral content above 50MHz. Line codes having a spectral content which matches the available band of 5 to 50MHz are preferred.

Line codes based upon Walsh functions offer a balanced code with good clock timing content to enable receiving stations to synchronise with incoming packets. Figure 3 shows the spectral content of WAL 2 coding where f is frequency and T the duration of one data bit. In this coding scheme a binary 1 is encoded into the codeword '0110' and a binary 0 into the code word '1001'. This coding scheme has no significant spectral content below 0.2fT. Hence, if f is made equal to 50MHz, the upper frequency limit of the transmitted signal, frequencies below 5MHZ may be ignored resulting in a spectrum matching the available bandwidth. The corresponding value of T is 40ns which is equivalent to a data transmission rate of 25Mbps.

It is proposed that two types of packets are conveyed over the network; speech packets to support a telephony connection and data packets to support a data connection.

One example of the structure of a speech packet is shown in figure 4.

Each terminal samples speech at the standard 8kHz sampling rate to generate 8 speech bits per sample (for each of the two B channels).

These are added to 2 signalling bits (the D channel) to form an 18 bit packet (28 + D) of data each sampling period. A fixed number (n) of such 18 bit samples are inserted into the data field DATA of a speech packet for transmission. Synchronising field SYNC enables receiving terminals to obtain bit synchronisation. A 56 bit synchronising field SYNC of the type specified in IEEE 802.3 is preferred. The Start of Packet Delimiter SFD marks the start of the packet once synchronism is achieved. A Control field CONTROL of 4 or 8 bits in length is used to indicate the type of frame (speech or data), a request for a speech or data connection, granting of a data connection request and that an expected packet has not been received (because of collision). Other uses may also be made of this field.

As indicated earlier, packets are not transmitted between terminals.

Packets are only transmitted between a terminal and the hub, or viceversa. This means that upstream packets need only inform the hub of their source and downstream packets, their destination. Hence packets need only contain a single address field which contains the address of the associated terminal. This means that, from a control consideration, the network may be likened to a logical star configuration. That is, packets effectively only pass from the hub to one terminal and vice versa.

With speech being sampled at a fixed rate of 8kHz, there is a design consideration as to how many samples to group together into a packet.

This determines the length of a packet, the packet repetition rate and the delay which other users may experience. Some efficiencies can be gained with longer packets, as the overhead information appended to each packet remains the same regardless of packet length. The table shown in figure 5 gives a summary of the main features of packets, based on a sampling rate of 8kHz, for various values of n.

The structure of a data packet is shown in figure 6. Data packets are transmitted over the bus along with speech packets. Similarly to speech packets, data packets comprise a synchronising field SYNC of 56 bits in length, a Start of Packet Delimiter SFD of 8 bits in length, and a control field CONTROL of 2 or 4 bits in length. The data payload DATA is preferably one ATM cell of 53 bytes in length, but other lengths of payload and other data packet structures are possible.

Two alternative methods for a terminal to initially access the bus are proposed: (i) a technique whereby terminals contend for access to the bus during a special access period; (ii) a Carrier Sense Multiple Access (CSMA) technique in which terminals monitor the bus for activity, and attempt to access the bus during any quiet periods. The basic technique which is widely used on local area networks is specially adapted for use on a CATV network.

Firstly, the technique where a special access period is provided will be described. During each frame there is an access sub-frame during which the hub transmits a downstream access packet AD and waits for a reply from a terminal. A flag in the control field of the access packet indicates that it is an access packet. This access packet has two functions: (i) to alert a terminal of an incoming call; (ii) to invite an idle terminal to initiate a connection.

All terminals monitor access packets to check for incoming calls. When an access packet carries an alert of an incoming call, the alerted terminal responds by returning an upstream packet within the access sub-frame. Terminals recognise that they have an incoming call by comparing an address of the alerted terminal, carried in the access packet, with the terminal's own address. In the following frame, the position that was previously occupied by the access sub-frame is assigned as a connection with the alerted terminal. The following subframe period then becomes an access sub-frame for further alerts or terminal accesses.

A terminal wishing to initiate a connection makes a request by transmitting an upstream packet, indicating the identity of the terminal, within the reply period of the access sub-frame. In the following frame the position that was previously occupied by the access sub-frame is reassigned as a connection with the terminal, and, as above, an access sub-frame follows in a new position.

Activity on bus 150 is described with reference to figures 7A-7E. In the time domain there is a series of time-contiguous frames, with the frame length being chosen in consideration of n (the number of speech samples inserted into each speech packet). Frame length is typically 250us or a sub-multiple thereof.

Figures 7A-7E illustrate how bus activity can develop from no connections (fig. 7A), through a number of speech connections (figs. 7B- 7D), to the granting of a data connection (fig. 7E).

Figure 7A shows a frame with just an access sub-frame comprising the downstream access packet AD transmitted by the hub, to alert terminals or to allow terminals to initiate a connection, and a period AU in which a terminal can reply by transmitting upstream. The access sub-frame is present in every idle frame, and is generally present in a frame except when full capacity is reached, when the space occupied by the access sub-frame can be used by a further speech connection (as described below. ) Figure 78 shows a frame with N on-going speech connections plus the access sub-frame. Each connection comprises a pair of adjacent transmissions in time - a downstream transmission addressed to a particular terminal and, immediately following that, an upstream transmission from that same terminal to the hub. The upstream transmission is triggered by the reception of a downstream transmission addressed to that terminal. No terminals, other than the terminal identified in the hub's downstream transmission reply to the hub, thereby avoiding contention between on-going connections. A typical connection between the hub and a terminal lasts for a large number of frames, with one pair of upstream and downstream transmissions occuring within each frame. The pair of transmissions for a particular connection will usually occupy the same position within a frame until repositioning is necessary. The downstream transmission of a connection comprises a packet which carries speech information for the call in the direction from the hub to the terminal and the upstream transmission comprises a packet which carries speech information for the call in the direction from the terminal to the hub.

Figure 7C shows the situation immediately after terminal 2 has disconnected, leaving an idle period between the on-going connections for terminals 1 and 3. In order to maximise the contiguous length of idle time within the frame, the on-going connections 3 to N are repositioned within the frame, so as to fill the idle time left by the terminal 2. Figure 7D shows a frame following the repositioning operation. Repositioning is effected by the hub varying the timing of the downstream transmission for each terminal's on-going connection. Repositioning minimises blocking of future data connections. The hub and each terminal have an elastic buffer. A first packet of a received transmission is delayed by one frame period. Thereafter this delay may be varied to accommodate the position of the terminal's sub-frame within the frame. This enables received packets to vary in time (within a frame boundary), and yet the output information from a terminal's receiver to retain real-time synchronisation.

Now consider a terminal wishing to establish a data connection. It makes a request by transmission of an upstream packet in the access sub-frame. This hub may grant the request providing that there is at least sufficient idle time within the frame to accommodate both a data connection and an access sub-frame. The access sub-frame is necessary to ensure that data cannot block (have higher priority than) the telephony service. The idle time within a frame should occur as a contiguous period, because of the repositioning operation described above. Alternatively, if the idle time within a frame occurs as a number of separate periods, whose aggregate sum is sufficient to accommodate a data connection plus an access sub-frame, it is possible to vary the position of the downstream transmissions of on-going speech connections within the frame in response to a request for a data connection in order to accommodate the data connection. The hub indicates a grant by means of the control field of the downstream packet DD in the access sub-frame's position in the next frame. The terminal immediately responds by transmitting an upstream data packet Du after which the hub transmits the downstream access packet. A data connection sub-frame therefore comprises one 'speech' packet and an associated data packet, as shown in fig. 7E.

A situation may arise whereby one, or more, data connections exist, and a new speech connection is required. Where this would lead to insufficient idle time to accommodate a further speech connection, data packets are suspended by signalling from the hub over the downstream 'speech' packet. A speech connection may now be established in the normal manner. When sufficient time becomes available transmission of data packets is resumed under control of the hub.

Where bus activity reaches its limit, in which case all active connections are speech only, one further speech connection may be supported, if required. This is achieved by allocating the access sub-frame to speech. Thereafter no access sub-frame is transmitted until a connection ceases. The network is therefore blocked to all further connection requests.

The protocol described above is akin to master-slave operation where most of the intelligence resides in the hub. This is a technical and economic attraction. It is apparent that once connections are established they remain undisturbed by all other users throughout their connection.

A contention situation occurs where two, or more, terminals may attempt to make a connection request during the same access sub-frame.

Where there is contention, contending terminal's upstream packet transmissions will mutually interfere and will be unintelligible at the hub.

The hub makes no response in the next frame which is recognised by the requesting terminals as a failure. A random back-off retry strategy minimises the possibility of repeated collisions and ensure that one terminal quickly succeeds in a request in a subsequent sub-frame.

Thereafter any remaining terminals seeking to request an access will do so over access sub-frames in subsequent frames.

An alternative strategy may be used to eliminate contention. The hub, at any time, may hold a list of all inactive terminals. A polling technique may be utilised whereby the addresses of each inactive terminal are transmitted in the access sub-frame, one at a time, in turn. Any terminal seeking to make a request waits for its address to appear in an access sub-frame and when it does so, responds immediately by return of an upstream packet within the sub-frame. This differs from the contention strategy above in that, for a network of 80 homes, a terminal may have to wait for up to 79 access sub-frames, and hence frames, before an access may be made. However, with frame periods of 125ups, or a multiple thereof, delay in establishing a connection is still relatively small. Polling does offer a marginal improvement in bus utilisation and hence throughput efficiency.

In addition, the hub could transmit multiple access sub-frames within a frame. For instance, it could fill up the idle time as far as is possible with access sub-frames. This would reduce terminal contention, or raise the mean rate of polling, and hence reduce delay in granting connections. It would also speed up incoming calls to the network in that the hub could in general signal to several terminals per frame. The use of multiple access sub-frames slightly increases the complexity of the hub, but not the complexity of the terminals.

The alternative technique of accessing the bus, is now described.

Terminals use Carrier Sense Multiple Access (CSMA) to reduce the risk of collision with a packet from another terminal (or from the hub) attempting transmission at a similar moment in time. Collision of speech packets occurs if two or more terminals transmit an initial speech packet, or a new data packet, simultaneously. Terminals only 'see' downstream packets because of the directional properties of taps 140 of the network.

To ensure that an initial packet is not transmitted in the time interval occupied by an existing connection, terminals employ a conventio upstream and downstream transmissions with those of other terminals according to rules which are known by all terminals. By knowledge of such pairing terminals may, by monitoring downstream activity, determine periods when upstream packets are being transmitted and so avoid collision with existing packets when attempting transmission of an initial packet.

When two or more terminals attempt to access the bus at the same time, the packets which they transmit collide, corrupting the information carried by each of the packets involved in the collision such that the hub is unable to decode the packets. In this situation the hub fails to reply to any of the terminals which attempted transmission and does not return a packet. Terminals suffering a collision become aware of it by the absence of a response from the hub. A simple time-out mechanism in a terminal can be programmed to wait for a reply for a predetermined period of time after transmitting a packet to determine whether it has been successful.

In the time domain, there is a set of time contiguous frames. Terminals listen to bus activity for the timing of packets. When a terminal has a packet to transmit it does so when it has determined that the bus is idle and at such a time that, as far as possible, the remaining idle time after transmission is as large as possible. In this way best use is made of the available capacity of the bus, so as to leave room for carrying data packets, as explained later.

The first few packet transmissions from a terminal or the hub are usually either indicating, via the D channel, off-hook upstream or ringing downstream. Any delay in establishing packet transmissions due to contention between terminals should be relatively short and appreciably less than a second. Hence the effect of such delay will have an almost imperceptible effect on a telephone user. A residential line typically offers 0.07 Erlang of traffic and has a mean holding time of 90 seconds per connection. Therefore, a network consisting of 80 homes, generates a mean of 5.6 Erlang, with calls originating at a mean interval of 16.1s.

Given that packets originate at 125tis intervals (or, if n > 1, a multiple thereof), the probability of two stations attempting to transmit at the same time is very low. Once a packet is successfully transmitted, all subsequent packets occur at regular intervals.

Once a terminal has succeeded in transmitting an initial packet, no subsequent speech packets will experience a collision with other speech packets since all other terminals are aware of their presence and avoid transmitting new packets at such a time that would lead to a collision.

This then means that each terminal is successfully transmitting packets, all at the same rate, but at different intervals in time. This scenario may be regarded as analogous to a TDM system with a number of active channels, or slots. The total number of sub-frames which may be supported is a function of transmission rate, bus length and packet size.

In order to transmit a data packet, a terminal makes a data connection request using the Control field CONTROL (possibly in conjunction with information contained in the data field DATA) of a speech packet. The hub responds and, if sufficient idle time exists on the bus to accommodate a data packet, grants permission to the requesting terminal to transmit data packets. The terminal then sends data packets, at the same rate as that of speech packets. These are positioned, as far as possible, so as to avoid collision with any existing packets. Data requests for downstream transmission to a terminal can also be accommodated.

A data connection request can be granted providing there is an adequate interval in time (in aggregate across the frame and not necessarily as a continuous period of time) to accommodate a data packet and it's corresponding 'speech' packet. In addition, sufficient idle time should also exist to support at least one further speech connection.

If sufficient idle time is not left then blocking of telephony services will result. If a data request is granted and subsequently there is insufficient idle time within a frame to support a further speech connection the hub commands that one data packet transmission be suspended (via the Control field of the downstream 'speech' packet).

Activity on bus 150 is described with reference to figures 8A-8E. In the time domain there is a series of time-contiguous frames, with the frame length being chosen in consideration of n (the number of speech samples inserted into each speech packet). Frame length is typically 250ups or a sub-multiple thereof.

Figures 8A-8E illustrate how bus activity could develop from no activity (figure 8A), through a number of speech connections (figures 8B-8D) to the granting of a data request (figure 8E). This set of figures shows a protocol where upstream transmissions precede downstream transmissions but this could be reversed so that downstream transmissions precede upstream transmissions.

Figure 8B shows a frame having one active connection between a terminal and the hub. Upstream transmission 1U and downstream transmission 1D are adjacent in time, forming a pair which can be regarded as a sub-frame. Terminals join the start of an idle frame.

Other terminals wishing to access the bus see only the downstream transmission 1 D, but are aware of the upstream transmission because of the pairing rule, which is known to all terminals. A terminal that is attempting to access the bus monitors downstream transmissions on the bus and stores a record in a memory. It then calculates, from the known time period separating upstream and downstream transmissions in a pair, busy times when upstream transmissions corresponding to those downstream transmissions will occur. The terminal avoids both the times when downstream transmissions occur and the calculated busy times when upstream transmissions will occur when attempting to access the bus.

Figure 8C shows a frame which has three active connections. Further connections 2U, 2D and 3U, 3D have joined the frame since the situation shown in figure 8B. Sub-frames are initially contiguous in time at the start of the frame until some connections are released. Once terminals have joined the bus, they retain the same position relative to the start of the frame.

Figure 8D shows the position after N terminals have joined the bus, with some of the earlier connections having been released. This has produced a situation where the sub-frames are no longer contiguous, but where they are fragmented. The total idle time within the frame is the sum of idle periods t1, t2 and t3. It will be appreciated that many other fragmentation patterns are possible. If this total idle time is greater than that required for carrying a data packet (plus sufficient idle time to allow a further speech connection) then a grant is issued in reply to a data request. A grant is made in reply to a data request even though, as in this example, the data packet is longer than any single one of idle periods tl, t2 or t3.

The terminal which is granted a data request now has the options of transmitting after 1D, 9D or ND. The case where the terminal transmits its data packet after 1D will be considered. When the terminal associated with upstream transmission 9U attempts to access the bus it senses that the bus is now busy, because the data packet now occupies that part of the frame which it previously used to support a speech connection. The terminal waits, continually monitoring the bus, until it becomes free and then, based upon its previous record of bus activity, transmits in what it knows to be the idle time, either after the new data packet (as shown) or after ND. If the speech packet does not collide all subsequent packets are transmitted in this new position within the frame.

To overcome jitter, which could otherwise result when speech packets are delayed, all terminals have an elastic input buffer which delays initial speech packets in time by one frame. Any subsequent packet which is delayed has the buffer delay reduced by the amount of the delay.

Thereafter the buffer delay returns to that of the frame length since all following packets which are output from the buffer will now be periodic.

Although not shown in the drawings, a fragmentation scenario may develop such that two or more speech packets are delayed by a data packet. Stations transmitting speech packets may then all transmit at the same time at the end of the data packet and collide. However, corresponding receiving terminals would fail to receive such packets and signal this fact, via the Control field, in their corresponding transmitted packets (which may require re-timing to avoid contention with the new data packet). Stations can then randomise their next transmission, within the available idle time upon the bus, and so reduce the risk of further collisions. In a similar manner, any contention between an existing data packet and an initial speech packet can also be resolved by signalling a lost packet, upstream or downstream, via the Control field of a corresponding downstream or upstream 'speech' packet.

The loss of a subsequent speech packet from two or more active terminals due to contention after an initial data packet is transmitted is expected to be a very isolated incident and such occasional loss may be tolerated. Loss of subsequent data packets, again comparatively infrequent, would in practice be accommodated by a link management protocol implemented at layer 2 of the OSI Reference Model.

Where downstream transmissions precede upstream transmissions the hub can decide the order in which remaining terminals transmit by issuing downstream packets which incorporate the unique address of each of the terminals. For example, consider a case where two terminals are delayed because a new data packet occupies the time which those terminals previously used. Directly after the data packet the hub issues a first downstream packet containing the address of the first delayed terminal, which replies with an upstream transmission, and then the hub transmits a second downstream packet containing the address of the second delayed terminal. In this manner both delayed terminals are repositioned within the remaining time In the frame without collisions occurring and packets being delayed.

The 'speech' packets of a data connection can be used as a low-speed data channel or as a return channel for interactive applications. They can also be used for acknowledgement purposes at the Data Link layer.

Under light loading of the bus it may be possible for the bus to support more than one data packet within a frame at a time. If multiple data packets are supported within a frame, those data packets can be transmitted to or from the same or different terminals. A data packet may include multiple ATM cells, making particularly efficient use of the medium.

Another consideration is that of receiver synchronisation. It may be possible to arrange that the hub initiates downstream frames in a synchronous manner. From a receiver point of view this means that downstream bit timing, as seen in all passing frames irrespective of whether they be addressed to it, or not, is purely repetitive. Although there may be periods of no transmission within a frame, the repetitive nature of bit timing over all packets may mean that the clock recovery task is eased compared with upstream frames. In the latter case packets arrive at the hub asynchronously. Although upstream frames could have the same clock frequency (by synchronising from the hub via downstream packets as indicated above), their phasing varies due to the different path lengths experienced by packets from different physical positions in the network. In consequence, consideration may be given to a shorter synchronising field in downstream packets compared with upstream packets. This would slightly reduce frame overhead and improve throughput.

Figures 9 and 10 show equipment which is used at the hub and at the terminal to support the telecommunications service. Figure 9 shows equipment at hub 120. An optical interface 200 couples signals to and from optical line 110. Frames of data are composed and decomposed by block 205 for transport over bus 150.

Figure 10 shows equipment at a terminal. Branch 160 couples to the terminal to convey signals to and from bus 150. A transmit path of the terminal equipment comprises a packetisation block 230, buffer 225 and line encoder 220. A computer 280, telephone 270 and telephone interface 275, or some other signal source, feed signals to and from the terminal equipment. Packetisation block 230 forms packets for transmission, which packets include the control, address and payload fields described earlier. Buffer 225 stores packets until instructed to transmit them. Encoder 220 encodes the binary data with a line code such that the spectral output of the transmision is matched to the available channel on the bus. A receive path of the terminal equipment comprises a line decoder 240, depacketisation block 260 and a downstream transmission detector 250 for use with the embodiment where a downstream transmission from the hub triggers an upstream transmission. Downstream transmission detector 250 can comprise a comparator which compares the terminal's own address with an address retrieved from a received packet by depacketisation block 260.

Downstream detector 250 outputs a control signal 270 to control the release of stored packets from buffer 225.

The performance of the system, for speech-only connections, will now be examined. The maximum number of speech connections N which may be supported depends upon frame duration, data transmission rate R and the propagation delay of the bus, td. A new packet may only be transmitted after the previous packet has propagated along the bus.

Hence the medium occupancy for one packet equals the sum of packet duration and propagation delay. N may be determined as follows, remembering that each connection requires one upstream and one downstream packet per frame.

frame duration N = frame duration 2(packet duration + propagation delay) Where: frame duration = n x sampling period (2) = 125n AIs (3) packet duration = number of bits per packet (4) data transmission rate Where: packet duration = 76 s (5) 5 (5) R Substituting Eqs. (3) and (5) into Eq. (1) yields:

6.25 x 10~6nor (7 = (7) n + 0.18Rtd + 7.6 Fig. 11 illustrates N as a function of bus length for various n at a data transmission rate of 25Mbps.

The system is an example of a finite-source finite-server which follows an Engset distribution and may be represented in general form using Markov chains, as shown in figure 12.

The probability of being in state r follows an Engset distribution and may be found by means of Eqn. (8):

where p is the arrival rate of one source (subscriber), h is the mean holding time of a single connection, S is the number of sources (subscribers) and N is the number of servers or, in this case, connections which may be supported. From the traffic figures suggested earlier I is given by: traffic in Erlang mean holding time (9) 0.7 E (10) 90 s = 2.8/hour (11) Blocking probability has been computed using the above value for pand 80 sources and it is found that if N is 10 then Grade of Service is 0:0086.

This is better than the figure of 0.01 which is widely accepted as satisfactory. If N is 11 then Grade of Service improves further to 0.0034, or less than one call in 200 blocked, which is a very high quality standard. Inspection of fig. 11 reveals that if n equals 1 then, ignoring tails between tap and home, acceptable service may be provided with a bus of 300 to 400 m in length. The actual length possible in practice depends upon the worst case attenuation of line and tap loss. If n is made equal to 2 then lengths in excess of 1000 m may be considered.

This has the effect of reducing packet transmission rate and therefore probability of collision. Where a collision does occur the delay becomes slightly greater but remains trivial as seen by a user.

Under light loading of the network, consideration may be given to assigning more than 2 B samples per packet to one or more terminals.

This may be performed on a dynamic basis such that, if traffic loading increases, service is progressively returned to the standard 2B operation. Since increased bandwidth provision cannot be guaranteed to a particular terminal, dynamic assignment is only recommended for non-real time activity such as data communication applications.

Claims (28)

CLAIMS:

1. A method of operating a communications distribution network, which network comprises a distribution hub and a transmission medium extending from the hub to a plurality of terminals, the medium being shared by the plurality of terminals, the method comprising: - supporting a plurality of communications connections between the hub and the terminals, each connection comprising a sequence of pairs of a downstream transmission from the hub to a particular one of the terminals and an upstream transmission from that terminal to the hub, which upstream transmission is triggered by the terminal receiving the downstream transmission.

2. A method according to claim 1 wherein the connections occur within a series of time-contiguous frames, the method further comprising repositioning on-going connections by varying the position of the downstream transmissions within the frame.

3. A method according to claim 2 wherein the repositioning is performed in response to termination of connections.

4. A method according to claim 2 or claim 3 wherein the repositioning is performed to maximise contiguous length of idle time within the frame.

5. A method according to any preceding claim wherein the connections comprise transmissions of different lengths.

6. A method according to any preceding claim wherein a connection comprises a voice or a data connection.

7. A method according to claim 6 wherein the data connection carries an ATM cell.

8. A method according to claim 6 wherein a data connection for conveying data from a terminal to the hub comprises a downstream transmission and an upstream data transmission.

9. A method according to claim 8 wherein the downstream transmission conveys signalling information for the data connection.

10. A method according to claim 2 wherein a terminal matches a continuous input or output data rate to the terminal to a varying transmission position within a frame.

11. A method according to any preceding claim further comprising providing an access period during which a terminal can request a connection.

12. A method according to claim 11 wherein the access period comprises a downstream transmission from the hub and a period during which terminals can reply.

13. A method according to any one of claims 1 to 10 wherein the upstream and downstream transmissions of a connection are separated by a time period which is known by the plurality of terminals, whereby a terminal can determine when the medium is in use by another terminal.

14. A method according to any preceding claim wherein the distribution network is a broadband distribution network for distributing video services to the terminals and for carrying telecommunications signals.

15. A method of operating a distribution hub in a communications distribution network, which network comprises the distribution hub and a transmission medium extending from the hub to a plurality of terminals, the medium being shared by the plurality of terminals, the method comprising: - the hub supporting a plurality of communications connections between the hub and the terminals, each connection comprising the hub transmitting downstream to a particular one of the terminals and receiving an upstream transmission from that terminal in reply to the downstream transmission, a connection comprising a sequence of pairs of such downstream and upstream transmissions.

16. A method of operating a communications terminal in a communications distribution network, which network comprises a distribution hub and a transmission medium extending from the hub to a plurality of the terminals, the medium being shared by the plurality of terminals, the method comprising: - receiving a downstream transmission from the hub to the terminal; - transmitting an upstream transmission from the terminal to the hub, which upstream transmission is triggered by the terminal receiving the downstream transmission; wherein the connection comprises a sequence of such pairs of a downstream transmission and an upstream transmission.

17. A communications distribution network, which network comprises: - a distribution hub; - a transmission medium extending from the hub to a plurality of terminals, the medium being shared by the plurality of terminals; - means for supporting a plurality of communications connections between the hub and the terminals, each connection comprising a sequence of pairs of a downstream transmission from the hub to a particular one of the terminals and an upstream transmission from that terminal to the hub, which upstream transmission is triggered by the terminal receiving the downstream transmission.

18. A method of operating a communications distribution network, which network comprises a distribution hub and a transmission medium extending from the hub to a plurality of terminals, the medium being shared by the plurality of terminals and having transmissive properties which inhibits transmission of signals between terminals, the method comprising transmitting communications signals in the downstream direction, from the hub to a particular one of the terminals, and in the upstream direction, from that terminal to the hub as a pair over a common portion of the medium's bandwidth, the transmissions of the pair being separated by a time period which is known by the plurality of terminals, whereby a terminal can determine when the medium is in use by another terminal.

19. A method according to claim 18 wherein the upstream and downstream transmissions are adjacent in time.

20. A method according to claim 18 or claim 19 wherein a terminal attempting to access the medium monitors downstream transmissions, calculates, from the known time period separating upstream and downstream transmissions in a pair, busy times when upstream transmissions corresponding to those downstream transmissions will occur, and avoids accessing the medium during times when it is occupied by downstream transmissions and the calculated busy times.

21. A method according to claim 20 wherein each terminal has a predetermined address and wherein a terminal attempting to access the bus makes an initial upstream transmission which includes the terminal's address, and waits a predetermined period for a subsequent downstream transmission from the hub, the terminal comparing the address in that subsequent transmission with its own to determine whether access was successful.

22. A method according to claim 18 wherein a series of time contiguous frames exist, each frame accommodating a plurality of upstream and downstream transmission pairs.

23. A method according to claim 22 wherein if a terminal is successful in accessing the medium it maintains that same position for its upstream transmissions in subsequent frames.

24. A method according to any one of claims 18 to 23 wherein the distribution network is a cable television (CATV) distribution network for distributing video services to the terminals and for carrying telecommunications signals.

25. A method of operating a cable television (CATV) distribution network, which network comprises a distribution hub and a cable bus network extending from the hub to a plurality of terminals for distributing video services to the terminals and for carrying telecommunications signals, the method comprising transmitting telecommunications signals in both the downstream direction, from the hub to the terminals, and in the upstream direction, from the terminals to the hub, over a common portion of the bandwidth of the bus.

26. A method according to claim 25 wherein the telecommunications signals are unmodulated and wherein they are encoded using line codes.

27. A method according to claim 25 wherein the encoded telecommunications signals have a spectral content substantially matching an available band of the bus.

28. A method of operating a cable television (CATV) distribution network, which network comprises a distribution hub and a cable bus network extending from the hub to a plurality of terminals for distributing video services to the terminals and for carrying telecommunications signals, the terminals being capable of supporting speech and data connections, the method comprising: - supporting a plurality of speech connections between the hub and terminals, each connection maintaining the same position within each of a series of time-contiguous frames; - receiving a request for a data connection; - granting a data connection if the aggregate sum of idle time within a frame is greater than the length required by a data connection, and - re-positioning on-going speech connections within the frame around the data connection.