WO2002093850A2

WO2002093850A2 - System for, and method of, synchronizing events in asynchronously operating communications systems

Info

Publication number: WO2002093850A2
Application number: PCT/IB2002/002806
Authority: WO
Inventors: Martin Götzer
Original assignee: Marconi Communications Gmbh
Priority date: 2001-05-16
Filing date: 2002-05-16
Publication date: 2002-11-21
Also published as: AU2002344804A1; WO2002093850A3; DE10124070A1

Abstract

The invention relates to a system for synchronizing events in asynchronously operating communications systems having first means for generating first events and second means for generating second events taking into account the first events, the time when the first events are generated being selected taking into account the intended generation of the second events. The invention also relates to a method for synchronizing events in asynchronously operating communications systems.

Description

SYSTEM FOR. AND METHOD OF. SYNCHRONIZING EVENTS IN ASYNCHRONOUSLY OPERATING COMMUNICATIONS SYSTEMS

The present invention relates to a system for, and a method of, synchronizing events in asynchronously operating communications systems, such as a packet based communication system.

Synchronous and asynchronous communications networks are known. In synchronous networks, which are mainly used to transmit voice, mechanisms are known which synchronize the network elements and the digital sources to a network clock. The synchronization is carried out at a byte level, that is at a digitization sample level. Given a typical sampling rate of 8 kHz, this ensures that no perceptible delay is generated owing to the formation of packets because any delay lies in the range of the sampling period, that is to say in the range of approximately 125 ms.

In asynchronous networks, for example ATM (Asynchronous Transfer Mode) networks or IP (Internet Protocol) networks, there are mechanisms for transporting a large number of sampling results in the form of packets or cells. The time which has elapsed between a first and the last sampling operation in a packet is referred to as the packetization delay. Typical packetization delays are of the order of 6 ms for ATM networks and 20 ms for more compressed links or for IP networks. It is to be noted that, although samples are taken periodically, coding mechanisms can result in variable packet rates, for example as a result of the use of silence suppression. In point-to-multipoint (PMP) transmission systems a plurality of terminals communicate with a central node using a shared communication medium, for example using radio or using a passive optical network (PON). The terminals can share the transmission medium on the basis of time, frequency, coding or wavelength. Since the nature of packet traffic is asynchronous, the use of asynchronous allocation, for example one based on chronological allocation, is preferred. Typically a TDM (Time Division Multiplex) method is used in PMP transmission systems, for transmission from the central node (base unit) to the terminals (subscriber units) in a "downstream" direction and a TDMA (Time Division Multiplex Access) method used for the transmission from the subscriber units to the base station ("upstream") direction.

A TDMA system can be controlled, for example, in the base station by means of a time-based allocation device (hereinafter termed a scheduler). The subscriber units are synchronized to a MAC (Media Access Control) time base. Each time-based allocation (hereinafter termed a grant) issued by the scheduler is directed to one subscriber unit, providing the subscriber unit with the possibility of transmitting a data packet at the time specified in the grant. An algorithm in the scheduler determines the packet rate of each specific subscriber unit and ensures that collisions do not occur in the transmission medium (eg. A frequency band) which is jointly shared by the subscriber units.

In contrast to point-to-point transmission systems, the subscriber units PMP systems cannot transmit with a constant packet rate because the flow of the packets is controlled by a central scheduler. Problems can arise when it is desired to transport synchronous services over , asynchronous transmission systems, such as for example in PMP transmission systems. In PMP systems a synchronization problem can occur which results in an undesirably high delay or in delay jitter for the synchronous service in an upstream direction. The source, that is to say the sampling device, is typically not synchronized to the MAC scheduler.

One known measure to counteract the synchronization problem comprises the scheduler reacting to an enquiry (request) from the subscriber unit. However, this leads to an additional MAC delay owing to the time between the arrival of a packet and the actual transmission of the packet, and to possible jitter because it is possible for a plurality of subscriber units to have a plurality of packets waiting in them for transmission. Another known measure is to allocate grants periodically. However, if the allocation of grants takes place too early this can result in a failure to transmit the next packet within the allocated time slot.

The present invention is based on a system in which the time when the first events are generated is selected taking into account the intended generation of the second events. This provides the asynchronous system with synchronization which alleviates the problems of the prior art. In particular, a short delay and a small degree of jitter can be provided, which is advantageous particularly with delay-sensitive and jitter-sensitive data traffic such as, for example, voice traffic. The principle of the invention is based on the fact that the processing of the data in a communications system takes place in a plurality of stages or in a plurality of layers. The data, or items more generally the events, are passed on from one layer to the next layer where events are in turn generated taking into account the events which have been passed on. In principle, the further processing of the events is therefore dependent on the supply of events from a preceding layer. In systems of the prior art, this goes so far that the transmission of data from one layer into a following layer is completely dependent on processes in the following layer. However, in the present invention the time of the transfer of data from one layer into a following layer depends on the events in the following layer.

In one particularly preferred embodiment of the system according to the invention, the system is developed in that the second means transmit triggering events to the first means in order to trigger the first events. In this way, means in a subordinate layer generate an additional event which can also be referred to as a trigger event. This informs means in superordinate layers that events can be taken up for further processing.

The system according to the invention advantageously further comprises means which, with the respect to the generation of events, have relationships with one another or with the first means or the second means which correspond to the relationship between the first means and the second means. A multiplicity of layers which are located one above the other can thus be provided, the generation of events by higher layers being initiated by triggering events in lower layers.

In a particularly preferred embodiment the time when the first events are generated depends on time slots which are available for the dispatching of data which is associated with the first events. For example, as soon as a time slot for the dispatching of specific data becomes available, a MAC layer can cause corresponding data from superordinate layers to be made available.

Advantageously the time slots are allocated as a function of upper delay limits. A scheduler can be informed about the various services of the subscriber units, and upper delay limits can be taken into account as a function of the specific services.

Alternatively the time slots are allocated as a function of events which are generated at

I predictable times. This variant of the invention can be useful if the source cannot be synchronized with respect to the scheduler, for example if a coding scheme with packets of fixed length is being used. In such a system the scheduler can be synchronized to the source. This is conveniently implemented if the source generates predictable deterministic output data. The scheduler is then arranged to predict when a new packet has to be provided by the subscriber unit, and reserves a time slot for this time.

It is particularly preferred that the events comprise the transmission of data packets. The present invention can be used for any communication system with a plurality of layers in which events which are dependent of one another "in one direction" are triggered "in the other direction". However the invention is particularly advantageous for transmitting data packets in asynchronous communications systems, for example in the case of voice transmission and the like. Advantageously, the system of the present invention comprises: a physical layer, an ATM (Asynchronous Transfer Mode) layer which is located above the physical layer, an AAL (ATM Adaptation Layer) which is located above the ATM layer, the AAL layer having an AAL2-CF sublayer (CF = Common Function) and an AAL2-SSF sublayer (SSF = Service Specific Function) which is located above the AAL2-CF layer, and in which the generation of events by a specific layer are initiated by events of layers located below it. The invention can thus be used, for example, in a layered ATM architecture with AAL layers.

The invention finds particular application in a point-to-multipoint (PMP) system. PMP systems which operate, for example, using TDM in a downstream direction and with TDMA in an upstream direction are by their very nature tied to a packet data transmission and are in that respect asynchronous systems. For this reason, the present invention can be used in a particularly useful way in such systems.

According to a second aspect of the invention there is provided a method in which the time when the first events are generated is selected taking into account the intended generation of the second events. In this way, the advantages of the system according to the invention are also implemented in the method.

In one particularly preferred embodiment of the method triggering events are transmitted in order to trigger the first events.

Preferably the further events are generated which have relationships with one another or with the first events or the second events which correspond to the relationship between the first events and the second events.

Advantageously the time when the first events are generated depends on time slots which are available for the dispatching of data which is associated with the first events.

Advantageously time slots are allocated as a function of upper delay limits.

Alternatively time slots are allocated as a function of events which are generated at predictable times.

It is particularly preferred that the events comprise the transmission of data packets.

Preferably the method is applied to a system which comprises: a physical layer, an ATM (Asynchronous Transfer Mode) layer which is located above the physical layer, an AAL layer (ATM Adaptation Layer) which is located above the ATM layer, the AAL layer having an AAL2-CF sublayer (CF = Common Function) and an AAL2-SSF sublayer (SSF = Service Specific Function) which is located above the AAL2-CF layer, and the method being characterised by events of a specific layer being initiated by events of layers located below it. The invention is particularly useful in such a development if the communications system is a PMP system.

The invention is based on the recognition that a short delay and a small degree of jitter, in particular with respect to voice transmission or even video transmission, can be achieved by the method of synchronizing higher layers of a system with events in lower layers. Synchronization can take place explicitly on the basis of synchronization protocols, synchronization events or on the basis of a handshaking operation between the various layers, or implicitly, in which case the functions in lower layers are predicted.

The invention will now be explained by way of example only by means of preferred embodiments and with reference to the accompanying drawings, in which:

Figure 1 is a block diagram of an ATM architecture in order to explain the present invention, and

Figure 2 is a schematic representation of a point-to-multipoint (PMP) network in order to explain the present invention.

Referring to Figure 1 there is shown a block diagram of a layered ATM architecture in order to explain the present invention. The ATM architecture comprises a physical layer 10, an ATM (Asynchronous Transfer Mode) layer 12 and an ATM adaptation layer 14 (AAL). The ATM adaptation layer 14 is divided into sublayers: an AAL2-CF layer 16 for common functions (CF) and an AAL2-SSF sublayer 18 with one or more service-specific functions (SSF).

As is known the operation of ATM systems are not only asynchronous in the ATM layer 12 but also at the interfaces between the layers. This means that transmissions between the layers can take place at any desired times. In the system illustrated in Figure 1, the function which converts, for example, analogue voice signals into digital signals uses either a free-running clock or a network clock. The AAL2-SSF layer 18 combines the voice sampled values to form packets and is synchronized for this purpose with the free-running clock or with the network clock. As soon as a predefined constant number of samples is collected, the AAL2-SSF layer 18 transmits the packet of sampled values as a data unit to the AAL2-CF layer 16. The AAL2-CF layer 16 collects AAL2 packets and transmits a 48 byte cell to the ATM layer 12 as soon as either a cell is full or as soon as a timer in the AAL2-CF layer 16 has run out. Finally, the ATM layer 12 carries out shaping and arranges it in a queue. The data is then passed on to the physical layer 10 as soon as the layer can take it which transports the cell in a defined bit stream.

If the mechanism described above is carried out without further measures, this can lead to significant delays or to a significant delay jitter, especially if there are a small number of voice channels (e.g. one) and/or a low bit rate. It is also problematic in a TDMA system where operations are carried out in the physical layer 10 with a bit rate which is not constant. A possible problem is that the content which is transmitted by the AAL2-CF layer 16 just misses a time slot in the physical layer 10 and then has to wait for the next possibility.

The present invention eliminates above problem in that the events of higher layers are synchronized with respect to events in lower layers. For example, an additional trigger event, which indicates that a cell or a packet is to be transmitted, can be generated for the AAL2-CF layer 16. This trigger event can originate from the lower layers, that is to say from the physical layer 10 and/or the ATM layer 12. If, for example, the AAL2-CF layer 16 is allocated a time slot in the physical layer 10 to transmit a cell, this can take place at the allocated time so that the physical layer 10 is capable of using the aforesaid time slot taking account of delay and priority schemes within the ATM layer 12.

Furthermore the AAL2-SSF layer 18 can be synchronized with respect to the AAL2-CF layer 16. In this case, data units in the AAL2-SSF layer 18 are generated in response to a demand from the AAL2-CF layer 16. In contrast, in the known systems, the AAL2- SFF layer 18 waits until a predefined number of sampled values have been collected before transmitting a data unit to the AAL2-CF layer 16. In the synchronized system according to the present invention, the AAL2-SSF.layer 18 also transmits a data unit to the AAL2-CF 16 if it receives a trigger from the AAL2-CF layer 16. Dispatching then takes place regardless of the packet size, that is to say even if the packet is smaller than a predefined value. The length of the packet can be indicated by the AAL2 length field.

A system which is synchronized in this way minimises delay and also minimises delay jitter.

The synchronization mechanism described above is neither restricted to the AAL2 nor ATM systems and can be applied to any packet-based transport systems of constant or variable bit rate data stream such as for example voice transmission or video transmission over IP.

An example of a synchronization system in accordance with the present invention will now be described with reference to Figure 2, which is a schematic illustration of a point-to-multipoint (PMP) network. The PMP network comprises a base station 20 which is capable of communicating with a plurality of subscriber units 22, 24, only two subscriber units are illustrated in Figure 2. The PMP network can be configured, for example, for operation as a passive optical networks (PON) as specified in ITU-T G.983.1, or for operation as a radio access networks as specified in the ETSI BRAN project. Typically all data traffic flows between the base station 20 and the subscriber units 22, 24, rather than between the subscriber units 22, 24. Data traffic which is transported from a subscriber units 22, 24 to the base station is referred to being transported in an "upstream" direction, and data traffic transported in the opposite direction is referred to as being transported in a "downstream" direction.

PMP networks have in common the feature that the transport medium is shared by subscriber units to transport data. Accordingly access to the transport medium must be regulated using a MAC (Medium Access Control) mechanism to avoid collisions occurring. In packet-based PMP networks, a TDD (Time Division Duplex) mechanism is typically used for this purpose. It is also possible to combine various multiplexing schemes such as frequency, wavelength, space or code-division multiplex for upstream and downstream directions, time-division multiplex for a downstream direction and time-division multiple-access (TDMA) for an upstream direction.

TDD and TDMA mechanisms are similar in so far that access to the medium is controlled centrally by the base station. In the case of TDD systems, the base station also transmits in accordance with the multiplexing scheme, while in the case of TDMA systems only the subscriber units use a multiplexing scheme. The mechanisms of the present invention can be applied both to TDMA and TDD systems.

The MAC mechanism for the upstream direction is administered or controlled centrally by a time-based allocation mechanism (scheduling mechanism) which is typically located in the base station. All the subscriber units have a common absolute time base which is obtained by means of a mechanism which is referred to as "ranging". The scheduler (time-base allocation device) issues grants to the subscriber units which permit them to transmit a specified amount of data at a specific time. The scheduler ensures that only one subscriber unit is granted access to the transport medium at a any one time. The algorithm of the scheduler, scheduling algorithm, can be based on the current global traffic conditions existing within the PMP network. With such an algorithm it is possible, for example, to grant more allocations to a subscriber unit which is congested.

Although there may be a system-wide common network clock, for example 2.048 MHz, the TDMA system will not necessarily be synchronized to this clock at a cell level. This means that the scheduler is not aware of the timing of the ATM Adaption Layer (AAL ), that is the layer 14 in Figure 1. This can result in the scheduler issuing a grant which is a too early for the cell to be output by the AAL 14. A considerable amount of time may pass until the next grant is issued because the other subscriber units also need to be served. Consequently, the reaction time of the scheduler is limited by the physical round trip time. Consequently, the scenario presented ultimately leads to significant delay jitter and to a large amount of delay at the service layer. In the system of the present invention the above problems are overcome by a lower layer outputting a trigger to higher layers to enable data to be transmitted at the right time to make the allocated time slot. If this mechanism is described by reference to the block diagram illustrated in Figure 1, this means that the MAC protocol can be considered as being part of the physical layer 10 in Figure 1, and the physical layer transmits burst-like data, instead of continuous data streams, in the physical medium. The allocation can be transferred to the higher layers, leading to a situation in which data is made available in the physical layer 10 at just the right time, thereby leading to the lowest possible delay.

Preferably, if the scheduler is aware of the services in the subscriber units, it issues grants at a time which takes account of an upper delay limit for each service. The allocations of grants can be specific for a particular service and can be related to a particular specific queue for ATM, IP, AAL2 or the like.

In order to reach the upper limit for the delay of a specific queue, the scheduler preferably includes a respective timer for each queue, in which each timer indicates the latest possible time for a grant to be issued for this queue. Whenever the queue has been served, the timer is reset.

It is also possible that a source cannot to be synchronized with respect to the scheduler. For example, this may be the case if the source is an AAL2-SAR (Segmentation and Reassembly) sublayer and if the scheduler operates, for example, with a coding scheme with packets of fixed length. In this case, the scheduler can be synchronized with respect to the source. This can then be successfully implemented if the source generates predictable or deterministic outputs. The scheduler is configured to predict when a new packet has to be supplied by the subscriber unit and issues a grant for that time. The protocol mechanism can allow the subscriber unit to report to the scheduler the length of the packet delay in the subscriber unit. In response the scheduler can then adjust its prediction and thereby minimise any delay. Jitter and wander can also be eliminated in this depending on the length of the reporting period.

Synchronization mechanisms in accordance with the invention in which data units with a variable length are generated were described in conjunction with Figure 1 and Figure 2. If these data units are generated in response to requests of lower layers, this leads to AAL2 data units with a short delay and low jitter.

At the point in the system where data is recombined it is necessary to provide buffering in order to eliminate jitter. The buffer must be capable of accommodating both a minimum delay (T_mi_n) and a maximum delay (T_max) in the transmission path. When establishing the connection the system estimates the length of the delay which a respective data unit has experienced. In other words, an estimate as to the delay is made which is between the maximum delay and the minimum delay. If the maximum jitter, i.e. Tmax-Tπύn, is known, a preferred method for dimensioning the buffer is to assume that the data unit has experienced the shortest possible delay. To ensure that a data unit which is subject to the maximum delay is handled on time the data unit must then be delayed (buffered) by the time T_π^-T_n^ . On the other hand, the delay could already be the maximum possible delay T_ma_x so that the buffer must be capable of handling twice T^ - T^.

In the cell layer, numerous mechanisms can cause jitter, for example OAM (Operations and Maintenance), buffering, multiplexing and the like. Such jitter, which is also referred to as Cell Delay Variation (CDV), increases when the bit rate in the medium is low or if transmission is taking place in an upstream direction in a point-to-multipoint network. The growth of CDV with reduced bit rate is due to the fact that the average time between the arrival of a data packet in the physical layer and the transmission of the last bit increases at relatively low bit rates.

If the synchronization which has been described in conjunction with Figure 1 and Figure 2 is applied up to AAL2-SSF, which results in AAL2-SSF data units of a variable length, substantially all of the CDV which is caused by the physical layer, including OAM of the physical layer, can be eliminated because the AAL2-SSF layer is synchronized with the transmission of cells. In addition to the buffering described above, the length of the data unit must be taken into account for the buffer positioning because there may be data units with maximum length. If it is assumed that the delay of a cell is approximately constant through the network as soon as it has been transmitted in the physical layer, the delay jitter will be significantly lower than in the known systems.

The present description of the exemplary embodiments according to the present invention serves only for illustrative purposes and not for the purpose of restricting the invention. Various changes and modifications are possible within the framework of the invention without departing from the scope of the invention or its equivalents.

Claims

1. A system for synchronizing events in asynchronously operating communications systems having first means for generating first events and second means for generating second events taking into account the first events, characterized in that the time when the first events are generated is selected taking into account the intended generation of the second events.

2. A system according to Claim 1, in which the second means transmit triggering events to the first means in order to trigger the first events.

3. A system according to Claim 1 or Claim 2, and further comprising means which, with respect to the generation of events, have relationships with one another or with the first means or the second means which correspond to the relationship between the first means and the second means.

4. A system according to any one of the preceding claims, in which the time when the first events are generated depends on time slots which are available for the dispatching of data which is associated with the first events.

5. A system according to any one of the preceding claims in which time slots are allocated as a function of upper delay limits.

6. A system according to any one claims 1 to 4, in which time slots are allocated as a function of events which are generated at predictable times.

7. A system according to any one of the preceding claims in which the events comprise the transmission of data packets.

8. A system according to any one of the preceding claims, in which the system comprises: a physical layer (10), an ATM (Asynchronous Transfer Mode) layer (12) which is located above the physical layer (10), an AAL layer (ATM Adaptation layer) (14) which is located above the ATM layer (12), the AAL layer (14) having an AAL2-CF sublayer (16) (CF =

Common Function) and an AAL2-SSF sublayer (18) (SSF = Service Specific

Function) which is located above the AAL2-CF layer, and in which the generation of events by a specific layer are initiated by events of layers located below it.

9. A system according to any one of the preceding claims, in which the communications system is a point-to-multipoint system.

10. A method of synchronizing events in asynchronously operating communications systems comprising: generating first events and generating second events taking into account the first events, characterized in that the time when the first events are generated is selected taking into account the intended generation of the second events.

11. A method according to Claim 10, and further comprising transmitting triggering events in order to trigger the first events.

12. A method according to Claim 10 or Claim 11, and further comprising generating further events which have relationships with one another or with the first events or the second events which correspond to the relationship between the first events and the second events.

13. A method according to any one of Claims 10 to 12, in which the time when the first events are generated depends on time slots which are available for the dispatching of data which is associated with the first events.

14. A method according to any one of Claims 10 to 13, and comprising allocating time slots as a function of upper delay limits.

15. A method according to any one of Claims 10 to 13 and comprising allocating time slots as a function of events which are generated at predictable times.

16. A method according to any one of Claims 10 to 15, in which the events comprise the transmission of data packets.

17. A method according to any one of Claims 10 to 16, in which the system comprises: a physical layer (10), an ATM (Asynchronous Transfer Mode) layer (12) which is located above the physical layer (10), an AAL layer (ATM adaptation layer) (14) which is located above the ATM layer (12), the AAL layer (14) having an AAL2-CF sublayer (16) (CF =

Common Function) and an AAL2-SSF sublayer (18) (SSF = Service Specific

Function) which is located above the AAL2-CF layer, and characterised by generating events for a specific layer which are initiated by events of layers located below it.

18. A method according to any one of Claims 10 to 17, characterized in that the communications system is a point-to-multipoint system.