8/7/2019 T-Packtiming
1/30
Contact: Stefano RuffiniEricsson
Italy
Tel: + 39 06 7258 9892Fax:
Email: [email protected]: This is not a publication made available to the public, but an internal ITU-T Document intended only for use by the
Member States of ITU, by ITU-T Sector Members and Associates, and their respective staff and collaborators in their ITU related
work. It shall not be made available to, and used by, any other persons or entities without the prior written consent of ITU-T.
INTERNATIONAL TELECOMMUNICATION UNION STUDYGROUP15
TELECOMMUNICATION
STANDARDIZATION SECTOR
STUDY PERIOD 2005-2008
TD 147 (WP 3/15)
English only
Original: English
Question(s): 13/15 Geneva, 16-27 May 2005
TEMPORARY DOCUMENT
Source: Editor Recommendation G.pactiming
Title: Latest draft of G.pactiming
This document contains the latest draft of the Recommendation G.pactiming. This version of
the Recommendation was approved during the SG15/13 meeting, held in Geneva the 16-20May 2005.
ITU-T Recommendation G.pactiming
Timing and Synchronization aspects in Packet Networks
Summary
This Recommendation analyses synchronization aspects in packet networks and outlines theminimum requirements for the synchronization function of network elements.
Keywords
Synchronization, jitter, wander, clock
Introduction
1 Scope
This Recommendation analyses the synchronization aspects in packet networks and outlines the
minimum requirements for the synchronization function of network elements.
The packet networks that are in the scope of this recommendation will initially be limited to the
following scenarios (with increasing complexity):
Point-to-point Ethernet [IEEE 802.3] [16] Multipoint (bridged) Ethernet [IEEE 802.1] [15] Connection-oriented MPLS [IETF RFC 3031, ITU-T G 8110] [30]
8/7/2019 T-Packtiming
2/30
- 2 -
TD 147 (WP 3/15)
Connectionless IP [IETF RFC 768 and RFC 791]Editors note: that last two bullets may not be covered in the first release of the G.pactiming.
2 ReferencesThe following ITU-T Recommendations and other references contain provisions, which, through
reference in this text, constitute provisions of this Recommendation. At the time of publication, the
editions indicated were valid. All Recommendations and other references are subject to revision;
users of this Recommendation are therefore encouraged to investigate the possibility of applying the
most recent edition of the Recommendations and other references listed below. A list of the
currently valid ITU-T Recommendations is regularly published.
The reference to a document within this Recommendation does not give it, as a stand-alone
document, the status of a Recommendation
[1] ITU-T Recommendation G.703 (2001), Physical/electrical characteristics of hierarchical
digital interfaces.
[2] ITU-T Recommendation G.783 (2004), Characteristics of synchronous digital hierarchy
(SDH) equipment functional blocks.
[3] ITU-T Recommendation G.801 (1988), Digital transmission models.
[4] ITU-T Recommendation G.803 (2000), Architecture of transport networks based on the
synchronous digital hierarchy (SDH).
[5] ITU-T Recommendation G.810 (1996), Definitions and terminology for synchronization
networks.
[6] ITU-T Recommendation G.811 (1997), Timing requirements of primary reference clocks.
[7] ITU-T Recommendation G.812 (2004), Timing requirements of slave clocks suitable for
use as node clocks in synchronization networks.
[8] ITU-T Recommendation G.813 (2003), Timing requirements of SDH equipment slave
clocks (SEC).
[9] ITU-T Recommendation G.822 (1988), Controlled slip rate objectives on an international
digital connection.
[10] ITU-T Recommendation G.823 (2000), The control of jitter and wander within digital
networks which are based on the 2048 kbit/s hierarchy.
[11] ITU-T Recommendation G.824 (2000), The control of jitter and wander within digitalnetworks which are based on the 1544 kbit/s hierarchy.
[12] ITU-T Recommendation G.825 (2000), The control of jitter and wander within digital
networks which are based on the synchronous digital hierarchy (SDH).
[13] TR 101 685 - Timing and synchronization aspects of Asynchronous Transfer Mode (ATM)
networks
[14] IEEE 802, IEEE standard for Local and Metropolitan Area Networks: Overview and
Architecture
[15] IEEE 802.1, MACBridging and Management
[16] IEEE 802.3, CSMA/CD access method and physical layer specifications
8/7/2019 T-Packtiming
3/30
- 3 -
TD 147 (WP 3/15)
[17] ITU-T Y.1413, TDM-MPLS network interworking User plane interworking
[18] ITU-T Y.tdmip, TDM-IP network interworking User plane interworking
[19] ITU-T O.171, Timing jitter and wander measuring equipment for digital systems which are
based on the PDH
[20] ITU-T O.172, Timing jitter and wander measuring equipment for digital systems which arebased on the SDH
[21] TS 145 010, Digital cellular telecommunications systems (Phase 2+), Radio subsystem
synchronization
[22] TS 125.104, Universal Mobile Telecommunication Systems (UMTS), UTRA BS FDD, Radio
Transmission and Reception
[23] TS 125.105, Universal Mobile Telecommunication Systems (UMTS), UTRA BS TDD, Radio
Transmission and Reception
[24] TS 125.402, Universal Mobile Telecommunication Systems (UMTS), Synchronization in
UTRAN Stage 2
[25] ITU-T Draft Recommendation V.90, A digital modem and analog modem pair for use on
the Public Switched Telephone Network (PSTN) at data signalling rates of up to 56 kb/s
downstream and up to 33.6 kb/s upstream
[26] ITU-T T4 Standardization of Group 3 facsimile terminals for document transmission
[27] 3GPP2 C.S0010-B, Recommended Minimum Performance Standards for cdma2000 Spread
Spectrum Base Stations
[27] 3GPP2 C.S0002-C, Physical layer standard for cdma2000 Spread Spectrum Systems
[28] ITU-T G.8010, Architecture of Ethernet layer networks[30] ITU-T G.8110, MPLS Transport Network Architecture
[31] ITU-T G.701, Vocabulary of digital transmission and multiplexing, and pulse code
modulation (PCM) terms
[32] ITU-T Y.1411, ATM-MPLS Network Interworking - Cell mode user plane interworking
3 Definitions
Editors note:
Candidate terms to include here are for instance: Network-Synchronous Operation, ServiceSynchronization, Packet Network, TDM, IWF, CES, Packetizer, Depacketizer.
Network-synchronous operation: synchronization of the physical layer (usually by a timing
distribution of a timing signal traceable to a Primary Reference Clock PRC, see ITU-T G.811).
Editors note : to be checked if this definition is already covered in other relevant recommendation
(i.e. G.810, G.803)
Service Synchronization: synchronization of a specific service carried over a packet network.
8/7/2019 T-Packtiming
4/30
- 4 -
TD 147 (WP 3/15)
TDM: See ITU-T Rec. G.701 [] [31]. Editors note: this is a preliminary definition. Further
contributions are invited.
Interworking Function (IWF): See ITU-T Rec. Y.1411 [32].
Differential Method: synchronisation of the end points of a packet based network using signalstraceable to a Primary Reference Clock (PRC) see ITU-T G.811 source. The PRCs at the end
points may be traceable to the same PRC thus operating in a synchronous mode of operation or may
be traceable to different PRCs thus operating in a pseudo-synchronous mode of operation.
Editors Note : a more generic term used for these methods is Methods based on Common Clock
which may not necessarily need to be a PRC. A formal definition is missing and should be added in
the G.pactiming.
4 Abbreviations
ATM Asynchronous Transfer Mode
CES Circuit Emulation Service
IWF Interworking Function
LLC Logical Link Control
MAC Medium Access Control
NTP Network Time Protocol
PDH Plesiochronous Digital Hierarchy
PRC Primary Reference Clock
RBS Radio Base Station
RTP Real-time Transport Protocol
SASE Stand Alone Synchronization EquipmentSDH Synchronous Digital Hierarchy
TDM Time Division Multiplex
5 Conventions
6 General
Packet switching was originally introduced to handle typically asynchronous data.
However, for new applications such as the transport of TDM service and the distribution ofsynchronization over packet networks, the strict synchronization requirements of those applications
must be considered.
In fact the ongoing evolution in telecommunications increases the likelihood of hybrid
packet/circuit environments for voice and voice band data services. These environments combine
packet technologies (e.g., ATM, IP) with traditional TDM systems. Under these conditions, it is
critical to ensure that an acceptable level of quality is maintained (e.g. limited slip rate).
Synchronization in TDM networks is well understood and implemented. Typically, a TDM circuit
service provider will maintain a timing distribution network, providing synchronization traceable to
a Primary Reference Clock (i.e., clock compliance with ref[6], ITU-T G.811). These arrangements
are described in industry requirements whereby network synchronization is derived from a PRC
8/7/2019 T-Packtiming
5/30
- 5 -
TD 147 (WP 3/15)
source and distributed through network elements with lesser stratum clocks. An alternative solution
is to follow a distributed PRC approach (e.g. by using GPS).
Although the goal of the document is to cover the relevant packet network technologies, the timing
and synchronization aspects addressed in this document will initially concern with Ethernet based
networks with protocol layers as defined in IEEE 802.
Functional architecture for Ethernet networks are defined in ITU-T G.8010 [28].
In the context of the present document, the highest layers (e.g. layer 7 in OSI model) refer to
applications transported over the packet networks. Real time applications have relatively tight
timing requirements concerning delay and delay variation. Some applications might resolve their
timing issues within higher layers (e.g., MPEG2), other applications rely on the timing support
provided by one or more of the lower layers (e.g. physical layer).
The present document aims at describing different methods to obtain the synchronization related
requirements.
Additionally, the requirements for interfaces and equipment that are part of the Ethernet network,
are described. It also gives recommendations when to apply different types of synchronizationmethods.
Some considerations on applicable synchronization requirements in a packet based network are
summarized in the following subsections.
6.1 Packet Network Synchronization Requirements
The nodes involved in packet oriented transmission technology, (e.g. ATM network nodes) do not
require any synchronization for the implementation of the packet switching function. In fact at any
entrance point of a packet switch, an individual device shall provide packet timing adaptation (for
instance cell timing adaptation in case of ATM switch) of the incoming signal to the internal timing.
For instance in case of ATM network the principle to cater for frequency differences is to use idlecell stuffing. Transmission links therefore in principle need not to be synchronized with each other.
However, as network shall be able to integrate TDM-based application, when transporting a CBR
stream over a packet network and when interworking with PSTN networks, the packet network
shall provide correct timing at the service interfaces.
This means that the requirements on synchronization functions in packet networks, especially on the
boundary of the packet networks, are depending on the services carried over the network. For TDM
based services, the IWF may require network-synchronous operation in order to provide acceptable
performance.
6.2 TDM and SDH timing requirements
The transport of TDM signals through packet networks requires that the signals at the output of the
packet network still comply with TDM timing requirements, this is crucial to enable interworking
with TDM equipments.
These requirements are independent of the type of information transported by the TDM signal,
voice or data.
The adaptation of such signals is sometimes called Circuit Emulation Services (CES).
The timing requirements that are applicable are: jitter and wander limits at traffic and/or
synchronization interfaces, long term frequency accuracy as this can influence the slip performance
and total delay (which is critical for real time services as for instance voice service).
8/7/2019 T-Packtiming
6/30
- 6 -
TD 147 (WP 3/15)
6.2.1 PDH timing requirements
The PDH timing requirements are mainly related to the jitter, wander and slip performance. In
particular the relevant requirements refer to the jitter and wander tolerance applicable at the input of
the network element at the boundary of a packet network that receives the TDM data and jitter and
wander generation applicable at the output of the network element generating TDM traffic at the
other end of the packet network. These values are specified in ITU-T G.823 for the network basedon 2048 kbit/s hierarchy and ITU-T G.824 for the network based on 1544 kbit/s hierarchy.
In addition the ITU-T G.822 specifying the slip rate objectives may be applicable. This is the case
when the clock of the equipment that generates the TDM signal and the clock used in the equipment
recovering the TDM signal from the packets are different.
6.2.2 Synchronization interfaces requirements
In case the PDH signals are defined as synchronization interfaces, the synchronization requirements
are more stringent than for 2048 kbit/s and 1544 kbit/s traffic interfaces. The synchronization
interface requirements for PDH interfaces are also defined in ITU-T G.823 and ITU-T G.824.
6.2.3 SDH timing requirements
Any STM-N signal must be compliant with ITU-T G.825. The relevant requirements refer to the
jitter and wander tolerance applicable at the input of the network element at the boundary of a
packet network that receives the STM-N data and jitter and wander generation applicable at the
output of the network element generating STM-N traffic at the other end of the packet network.
In case of STM-N signals there are no distinctions between traffic and synchronization interfaces as
all STM-N signals are defined to be as synchronization interfaces.
Editors note: STM-N CE is to be considered but probably this will not be done in the first release
of the G.pactiming. This is an important issue as STM-N is by definition a synchronization
interface. The impact on the existing assumptions on the hypothetical network reference model hasto be considered.
6.3 Synchronization Network Engineering in Packet Networks
The driving force for much of this work is to cater for the synchronization needs of the application
or in general the need of certain technologies (for instance Radio Base Station in GSM and
WCDMA networks). In order to achieve such goal, operators have therefore to distribute a
Reference Timing Signal of suitable quality to the Network Elements processing the application or
in general requiring accurate timing signal.
One approach is to follow a distributed PRC strategy (for instance by means of GPS technologies).
An alternative approach is based on a Master Slave strategy. The engineering rules for designing thesynchronization network in these cases are well understood and documented (see for example ITU-
T G.803, ref.4) when the underlying transport of the packets (e.g. IP frames) will run over the
existing synchronous technologies (PDH or SDH networks). On the other hand when the
underlying transport is based on non-synchronous technologies (i.e. Ethernet), alternative
approaches shall be considered. This will be further analysed in section 8, Reference Timing
distribution over Packet Network.
6.4 Timing requirements at edge versus timing requirements in core networks
Different performance can be requested in case the packet network is part of an access network or is
the underlying layer of the Core Network.
8/7/2019 T-Packtiming
7/30
- 7 -
TD 147 (WP 3/15)
The distribution of a synchronization reference over a portion of a core network may be requested
to comply with strict jitter and wander requirements (i.e. G.823, G.824 for synchronization
interfaces and G.825).
On the other hand in the access network, requirements may be related to allow a distribution of a
timing reference signal with performance sufficient to support the timing requirements of the end
node (for instance a Radio Base Station, or a V90 modem).
6.4.1 Synchronization requirements for GSM, WCDMA and CDMA2000 Base Stations
(RBS)
The timing requirement applicable to the GSM radio interface can be found in the ETSI technical
specification TS 145.010 (ref.xx). The basic requirement is to fulfil 50 ppb frequency accuracy on
the radio interface.
The timing requirement applicable to the WCDMA radio interface can be found in the technical
specifications and TS 125.104 (FDD mode) (ref. xx) and TS 125.105 (TDD mode) (ref xx). Also for
the WCDMA the basic requirement is to fulfil 50 ppb frequency accuracy on the radio interface.
The WCDMA TDD mode puts additional requirements on the phase accuracy. The requirement on
the relative phase difference between Node Bs is to not exceed 2.5 s (see TS 125.402, ref xxx).
It should be noticed that for in case of GSM and WCDMA radio access network there are not so
strict frequency accuracy requirements related to limit the slip rate.
In fact in these cases the data of a single user is stored in relatively large buffer (from 10 to 30 ms)
and also assuming 50 ppb frequency accuracy the data would be lost (buffer empty or full) after
long times, much longer if compared with classical switching network elements where buffers
handling the data are much smaller (125 microsec.).
The relevant CDMA2000 standard is the 3GPP2 C.S0010-B. Concerning synchronization
requirements, it states that:
the timing of the base station shall be within 10s of UTC. the average frequency difference between the actual CDMA transmit carrier frequency and
specified CDMA transmit frequency assignment shall be less than 50 ppb
In addition according to 3GPP2 C S0002-B specification, in order to support the CDMA System
Time, all the base station digital transmissions are referenced to a common CDMA system-wide
time scale that uses the Global Positioning System (GPS) time scale, which is traceable to, and
synchronous with, Universal Coordinated Time (UTC).
6.4.2 Synchronization requirements for facsimile and voce band Modem applications
The timing requirements to be considered for the Facsimile and Modem applications are certainlynot so strict as for the GSM and WCDMA RBS, nevertheless these have to be considered when
moving the access network towards non-synchronous (e.g. Ethernet) environment.
There are many types of voice band modems but the high-speed types are likely to be most sensitive
to jitter. Two significantly different types exist: the QAM based modems e.g. V.34bis and the PAM
based V.90 and V.92.
The V-Series recommendations provide guidance regarding the maximum error allowed in the bit-
rate of the various voice-band modems used in the general public switched telephone network. The
limit is generally 0.01% (see for instance ITU-T Recc. V.90).
Assuming that modems can recover the timing from the incoming data, the frequency accuracyrequirement applicable to Network Elements connected to modems can therefore be specified to be
8/7/2019 T-Packtiming
8/30
- 8 -
TD 147 (WP 3/15)
in the range of several ppm (up to 100 ppm) depending on the specific modem technology that shall
be supported.. However it should also be considered that an Access Device with frequency accuracy
deviating from the service frequency would create unacceptable data loss (buffer full or empty) and
frequency accuracy better than 100 ppm is therefore required.
The actual value is left for further study.
Editors note: The value of +/-50 ppb applicable to Access Device connected to modems is
proposed as initial proposal for discussion.
With respect to standards applicable to Facsimile application, ITU-T has published many
recommended standards for facsimile (FAX) transmission.
In particular ITU-T Recommendation T0 describes four FAX apparatus groups; Groups 1 through
4.
The T4 standard can probably be considered the most modern and the most important one. The T4
standard states that the modulation is based on ITU modem standards like V.27ter, V29 and V.34.
The assumption is therefore that the synchronization related requirements for modem applicationsare applicable for Facsimile applications as well.
Editors note: requirements for Access Device supporting fax and modems is probably covered in
other VoIP recommendations (IWG, H series); section 6.4.2 is therefore to be confirmed.
7 Packet Networks Reference Models
The packet network reference models that have been used to characterize the performance of the
packet networks in terms of packet delay variations are shown in the following figures:
model A in Figure 1 is related to applications with very strict delay and delay variation
requirements, model B in Figure 2 refers to scenarios with less strict packet delay variationrequirements.
Figure 1 Packet network reference model A (switched Ethernet network)
8/7/2019 T-Packtiming
9/30
- 9 -
TD 147 (WP 3/15)
Figure 2 Packet network reference model B (switched Ethernet network)
Following cases have been considered:
Scenario 1: Switched Ethernet network, Best effort with over-provisioning (single queue) Scenario 2: Switched Ethernet network, Quality of service according to IEEE 802.1q, IEEE
802.1p (at least 2 queues, with one dedicated queue for handling real-time data and WFQ,
Weighted Fair Queuing, discipline)
Scenario 3: Switched Ethernet network, Quality service according to IEEE 802.1q, IEEE802.1p (with one queue dedicated for the handling of data used for timing recovery,e.g.Time stamps)
Editors note: It has to be confirmed if similar scenarios can be provided for IP Router network. In
this case similar models and approaches could be applicable.
Editors node: the traffic models have to be defined (see Appendix I for initial input). Contributions
are invited.
Based on the above models, the following parameters apply for the different cases:
Network model A Network Model B
Scenario 1 tbd tbd
Scenario 2 tbd tbd
Scenario 3 tbd tbd
Table 1/G.pactiming: Parameters for the relevant network models
Editors note: Contributions are invited to fill the table with the relevant parameters (e.g. Packet
delay variation, etc.).
8/7/2019 T-Packtiming
10/30
- 10 -
TD 147 (WP 3/15)
8 Network Limits
The jitter and wander network limits currently specified in the relevant ITU-T recommendations
(i.e. ITU-T G.823 and ITU-T G.824) have to be fulfilled in all the scenarios that are relevant for this
recommendation.
In particular following jitter and wander limits apply to the TDM signal carried over the CES in thescenarios described in Annex A.
Case 1: is for further study. Case 2: Application A
The jitter and wander budget of the CES in this case is the difference between the 2048 kbit/s
network limit (see Fig1/G.823) and the 2048 kbit/s synchronization interface network limit
(see Fig10/G.823). As an example the wander budget over 24h for 2 048 kbit/s
synchronization interface reaches 12.7 s in this case.In case of 1544 kbit/s interfaces, the jitter and wander budget of the CES in this case is the
difference between the 1 544 kbit/s network limit (see Table2/G.82) and the 1544 Mbit/s
synchronization interface network limit (see Fig 3/G.824).
Case2 application BThe wander budget of the CES is only limited by the timing quality requested by the
application and not by the ITU-T G.823 and ITU-T G.824 specifications.
Note: this application is valid only for application with a single signal; if 2 signals are
received there might exist a differential jitter and wander between one signal and the clock
extracted from the other signal.
Case 3:When retiming is implemented at the output of the SDH islands, the amplitude of noise on thePDH output is that of a synchronization interface; this allows to increase the wander budget
up to that of case 2 application A in some configurations
9 Reference Timing Signal Distribution over Packet Networks
In order to fulfil the applicable synchronization requirements, it should be possible to distribute a
reference timing signal with proper phase stability and frequency accuracy characteristics.
Two main classes of methods are identified in this recommendation:
Plesiochronous and Network Synchronous methods, Packet based methods.
9.1 Plesiochronous and Network Synchronous Methods
The first class of methods refers to PRC distributed method (for instance based on GPS), or Master-
Slave method using a synchronous physical layer (e.g. STM-N), see figure 3. These methods are
widely implemented to synchronize the TDM networks.
It can be noticed that the latter option is not available in case of Ethernet networks, as the physical
layer in this case is not defined to be synchronous.
8/7/2019 T-Packtiming
11/30
- 11 -
TD 147 (WP 3/15)
Figure 3- Distributed PRC and Master Slave methods
Editors note: it has been recognized that when G.811 quality is required, in some cases and for
limited parts of the network, it would be useful to have also the option to distribute timing via a
synchronous Ethernet; further study is needed (e.g. traceability, etc.).
9.2 Packet based Methods
The second class of methods relies on timing information carried by the packets (e.g. sending
dedicated Time Stamp messages as shown in figure 3; methods using two-way transfer of timing
information are also possible such as NTP).
In some cases (for instance transport layer not synchronous) this is the only alternative to a PRC
distributed approach.
Packet based Methods and related performances are under study.
Figure 4 Example of packet based method with timing distribution of the Reference Timing
Signal via Time Stamps
10 Timing Recovery for Constant Bit Rate services transported over Packet Networks
CBR (Constant Bit Rate) services (e.g. Circuit Emulated TDM signals) require that the timing of
the signal is identical on both ends of the packet network: this is handled by the IWF responsible to
deliver the constant bit stream.
The following operating methods are identified within this recommendation:
Time Stamp
Master Packet NetworkTime Stamp
Processing
Time Stamp
PRC Reference
Recovered Reference Timing Signal
Time Stamp
Master Packet NetworkTime Stamp
Processing
Time Stamp
PRC Reference
Recovered Reference Timing Signal
8/7/2019 T-Packtiming
12/30
8/7/2019 T-Packtiming
13/30
- 13 -
TD 147 (WP 3/15)
Adaptive Methodsin this case the timing can be recovered based on the inter-arrival time of the packets or
on the fill level of the jitter buffer. It should be highlighted that the method preserves
the service timing (Asynchronous Circuit Timing).
Figure 7 Example of Adaptive Method
Reference Clock available at the TDM end systemsthis is a trivial case, in fact in this case both the end systems have direct access to the timingreference, and therefore there is no need to recover the timing, the only aspect to consider is
to guarantee a proper interworking with the IWFs, for instance via a loop timing scheme
(see figure 8). An example when this scenario might apply is when two PSTNs are
connected via a packet network.
Figure 8 PRC reference timing signal available at the TDM end systems
8/7/2019 T-Packtiming
14/30
- 14 -
TD 147 (WP 3/15)
Editors note: above figure shows the TDM transmitter and receiver being traceable to a PRC. This
is the case where transmitter and receiver are for example digital switches where there is a need to
control slips. However there are cases where the transmitter and receiver are not traceable to G.811
clocks, for example E3 or DS3 multiplexors.
Editors note : Further details on the above mentioned clock recovery methods should be provided
in the Appendix. Contributions are invited.
11 Impact of impairments in the packet network on timing distribution and service clock
recovery
Editors note: this section should provide the analysis of the impacts of the impairments in the
packet network (i.e. packet loss, burst packet loss, bad sequencing of packets, delay variation,
temporary network congestion, routing changes, low frequency components of packet delay
variations and errored packets) on the timing distribution and on the service clock recovery.
Contributions are invited on these topics (the work done within ANSI T1X1.3 and ETSI on theATM networks could be used as basis for this analysis).
Editors note: There is a common understanding thatPacket loss and packet misordering within
certain limits to be specified do not significantly affect clock recovery performance for any of the
methods presented in G.pactiming. Specifically, at levels at which the TDM transport service is
useable, packet loss (both uniform and bursty) and packet misordering have negligible effect on
clock recovery performance. Further contributions are required. There is also a common
understanding that Packet delay variation should not affect clock recovery performance for the
Common Clock method presented in G.pactiming. Further contributions are invited.
12 Impact of Network Clock Impairment on timing distribution and service clockrecovery
Editors note: this section should provide the analysis of the impacts of Network Clock Impairments
on the service clock recovery (e.g. Differential methods) and on the timing distribution.
Contributions are invited on these topics (the work done within ANSI T1X1.3 and ETSI on the
ATM networks could be used as a basis for this analysis).
12.1 Impairments for the Network Synchronous Operation methods
The clocks involved in the transport of TDM signals through a packet network are shown in
figure 9.
8/7/2019 T-Packtiming
15/30
- 15 -
TD 147 (WP 3/15)
Figure 9: clocks involved in the transport of TDM signals through a packet network
These are:
- The clock that generates the TDM signal, PDH or SDH (clk1 in the figure)
It is considered that most signals are now synchronous; if not, the use of a network clock reference
in the depacketizer (i.e. clk3 in the figure) will raise severe problems.
-The clock that is used to generate the packets (clk2 in the figure)
This clock has no direct effect on the recovery of the right timing; it simply changes the duration ofthe packets but not their periodicity. In the example shown in figure this clock is traceable to a PRC.
-The clock at the depacketizer which retimes the TDM signal after the packet network
In order to have a correct timing in the output TDM signal, the clocks generating (i.e. clk1) and
retiming (i.e. clk3) the TDM signal have to be identical, otherwise slips will be generated.
In normal operation these clocks are locked to a reference timing signal that is traceable to a PRC.
However during failures conditions in the synchronization network these clocks may be locked to a
reference timing signal that is traceable to a clock operating in holdover mode. The requirements
on clocks that during failure conditions shall provide a suitable holdover are to be based on the
ITU-T G.822 slip performance objectives, and on ITU-T G.823 and ITU-T G.824 jitter and wander
network performances.
In case of the packetizer and depacketizer clock, the clock providing holdover function during
failures in the synchronization network may be either integrated in the IWF itself or available in the
site (for instance integrated in a transmission network element or in a SASE). It is the responsibility
of the network planner to provide the most suitable solution.
Summarizing, the network synchronous mode of operation requires either the introduction of
expensive clocks in the sink IWF or a system that allows to switch to another, better, mode in case
of loss of synchronization from the network clock (PRC).
In order to detect the periods of loss of synchronization some kind of supervision of the traceability
is needed (e.g. SSM).
8/7/2019 T-Packtiming
16/30
- 16 -
TD 147 (WP 3/15)
13 Equipment synchronization related requirements
13.1 Traffic Interfaces
Following requirements have been taken from existing recommendations (eg. ITU-T G.823, ITU-T
G.824, etc.).
Editors note: the SDH interfaces are covered in the following subsections,although these may not
be in the scope of the first version of the G.pactiming.
13.1.1 Physical, electrical and optical characteristics
Physical and electrical characteristics of E0 (64 kbit/s), E11 (1544 kbit/s), E12 (2048 kbit/s)
interfaces, all PDH interfaces, 51840 kbit/s (STM-0) and ES1 (STM-1) interfaces shall comply with
requirements of ITU-T Recommendation G.703.
Physical and optical characteristics of OS1 (STM-1), OS4 (STM-4), OS16 (STM-16) interfaces
shall comply with requirements of ITU-T Recommendation G.957 (Tables 2/G.957, 3/G.957).
13.1.2 Jitter and wander generationEditors note: this has been recognized to be an issue being related to the budget that can be
allocated to the CES segment. The upper bound is defined in G.823 and G.824. The actual budget
that can be allocated to the CES segment is expected to be a fraction of the total budget in most of
the cases (see also section 8).
For information, details on the total network limits are provided below.
The peak-to-peak output jitter for networks based on 2048 kbit/s hierarchy at E0, E12, E22, E31,
E4 traffic interfaces shall comply with requirements of ITU-T Recommendation G.823 (clause 5.1).
The peak-to-peak output jitter for networks based on 1544 kbit/s hierarchy at E11, E21, 32064
kbit/s, E32, 97728 kbit/s traffic interfaces shall comply with requirements of ITU-T
Recommendation G.824 (clause 5.1).The peak-to-peak output jitter for SDH-based networks at ES1, OS1, OS4, OS16 traffic interfacesshall comply with requirements of ITU-T Recommendation G.825 (clause 5.1). The peak-to-peak
output jitter at 51 840 kbit/s traffic interface shall comply with requirements of ITU-T
Recommendation G.703 (clause 16.2).
The output wander limit for networks based on 2 048 kbit/s hierarchy at E12, E31, E4 traffic
interfaces shall comply with requirements of ITU-T Recommendation G.823 (clause 5.2).
The output wander limit for networks based on 1 544 kbit/s hierarchy at E11, E32 traffic
interfaces shall comply with requirements of ITU-T Recommendation G.824 (clause 5.2).
The output wander limit for SDH-based networks at 51 840 kbit/s, ES1, OS1, OS4, OS16 traffic
interfaces, according to ITU-T Recommendation G.825 (clause 5.2), shall comply withrequirements of ITU-T Recommendation G.823 (clause 6.2) for networks based on the 2048 kbit/s
hierarchy and ITU-T Recommendation G.824 (clause 6.2) for networks based on the 1544 kbit/s
hierarchy. These requirements are defined for synchronization interfaces (PRC, SSU, SEC) because
STM-N traffic interfaces are considered as synchronization interfaces.
13.1.3 Jitter and Wander tolerance
The input jitter and wander tolerance for networks based on 2048 kbit/s hierarchy at E0, E12, E22,
E31, E4 traffic interfaces shall comply with requirements of ITU-T Recommendation G.823 (clause
7.1).
8/7/2019 T-Packtiming
17/30
8/7/2019 T-Packtiming
18/30
- 18 -
TD 147 (WP 3/15)
13.2.3 Jitter and wander tolerance
The input jitter tolerance at T12, E12 synchronization interfaces, according to ITU-T
Recommendation G.823 (clause 7.2), shall comply with requirements of ITU-T Recommendation
G.812 (clause 9.2, Type I) for SSU interfaces and ITU-T Recommendation G.813 (clause 8.2,
Option 1) for SEC interfaces.
The input jitter tolerance at E11 synchronization interface, according to ITU-T RecommendationG.824 (clause 7.3), shall comply with requirements of ITU-T Recommendation G.812 (clause 9.2,
Types II and III) for SSU interfaces and ITU-T Recommendation G.813 (clause 8.2, Option 2) for
SEC interfaces.
The input wander tolerance at T12, E12 synchronization interfaces, according to ITU-TRecommendation G.823 (clause 7.2), shall comply with requirements of ITU-T Recommendation
G.812 (clause 9.1, Type I) for SSU interfaces and ITU-T Recommendation G.813 (clause 8.1,
Option 1) for SEC interfaces.
The input wander tolerance at E11 synchronization interface, according to ITU-T Recommendation
G.824 (clause 7.3), shall comply with requirements of ITU-T Recommendation G.812 (clause 9.1,
Types II and III) for SSU interfaces and ITU-T Recommendation G.813 (clause 8.1, Option 2) forSEC interfaces.
13.3 IWF synchronization function
In the context of the present document the IWF provides the necessary adaptations between TDM
and packets streams.
With reference to figure 10 below, the supported timing options for the Tx clock are:
Timing from recovered source clock carried by the TDM input (loop-timing or line-timing); Timing from the network clock (the network clock can be derived either from the physical
layer of the traffic links from the packet network or through an external physical timinginterface, e.g., 2 048 kHz);
Timing from a free running clock (it shall provide an accuracy according to relevantTDM/CBR service interface, e.g. 2 048 kbit/s shall comply to ITU-T Recommendation
G.703, 50 ppm);
Differential methods; Adaptive timing (including clock recovery using dedicated Time Stamps).
Depending on the services to be provided, a suitable subset of the listed timing options shall be
supported.
It is recommended to have slip control in the TDM Tx direction to control possible over/underflow
in the playout buffer. Slips shall be performed on n x 125 microsecond frames.
When TDM transmitter and/or receiver clocks are in holdover or are traceable to clocks in holdover,
and a synchronous clock recovery technique (Differential method or Network synchronous
operation) is used, slips (most likely uncontrolled) will occur. A back-up method based on the use
of adaptive method may be used in order to avoid an excessive rate of slips.
In the TDM-to-Packet direction, the synchronization related requirements are mainly depending on
the synchronization requirements of the physical layer. These are not detailed in the figure.
8/7/2019 T-Packtiming
19/30
- 19 -
TD 147 (WP 3/15)
Figure 10 - IWF synchronization functions
Editors note: other subsection will be included in section 13 in order to cover all relevant
equipment synchronization requirements (noise transfer, latency, etc.) . Contributions are invited on
these topics.
14 Results and consequences of the different synchronization methods over Packet
Network Reference Models
Editors note: this section shall provide basic recommendations on how the synchronization shall bedistributed/recovered for the different Packet Networks models. The recommendations would
depend on the requirements to be fulfilled (for instance physical layer requirements -e.g. wander
and jitter at the traffic interfaces- or reference timing signal stability for distributing synchronization
reference to access nodes as RBSs). Contributions are invited on these topics.
8/7/2019 T-Packtiming
20/30
- 20 -
TD 147 (WP 3/15)
Annex A
Network model underlying the network limits
For the transport of PDH signals the model in figure A1 of ITU-T G.823 is the starting point to
consider the insertion of CES segment (similar considerations are valid for G.824 limits, see figuresA1/G.824 and A2/G.824). The wander budget allocation for CES must be only a portion of the
entire wander budget as specified in G.823 or G.824, since the total wander budget has to be shared
with the rest of the network
Depending of where the CES is located different wander requirements may apply (see figure
below). Following types of CES have been identified:
Case 1: When the CES is located between the two switches of G.823 reference model, thewander budget is calculated based on model in figure A.1 below. The model is based on figure
A.1/G.823, from ITU-T G.823, where one of the SDH Islands is replaced by the CES network.Editors note: with this model the wander generated by the CES portion shall be not greater than
half of the G.823 limits. The assumption is to be confirmed through simulations using the packet
networks model from section 7 of this recommendation. The model is valid for networks based on
the 2048 kbit/s hierarchy. A similar approach could be provided for the networks based on the 1544
kbit/s hierarchy (see ITU-T G.824).
Figure A.1 - Network Models for data and clock wander accumulation (networks based on 2048
kbit/s hierarchy)
IWF IWFSDH
IslandSDH
Island
CES
Island
IWF
CES
Island
IWF
Wander and jitter
budget for the CES
Synchronization
networkSynchronization
network
PRC PRC
End
equipment
Wander and jitter
budget for the CES
IWF IWFSDH
IslandSDH
Island
CES
Island
IWF
CES
Island
IWF
Wander and jitter
budget for the CES
Synchronization
networkSynchronization
network
PRC PRC
End
equipment
Wander and jitter
budget for the CES
8/7/2019 T-Packtiming
21/30
- 21 -
TD 147 (WP 3/15)
Case 2: Application AWhen the CES is located outside the 2 switches (see Figure A.2), the retiming effect of the switch
has to be considered. At the output of the switch, the timing of the G.823 signal has the quality of a
synchronization interface, mainly 2 Mbit/s synchronization interface. The jitter and wander budget
of the CES in this case is the difference between the 2048 kbit/s network limit (see Fig1/G.823) and
the 2048 kbit/s synchronization interface network limit (see Fig10/G.823). As an example the
wander budget over 24h reaches 12.7 s in this case.In case of 1544 kbit/s interfaces, the jitter and wander budget of the CES in this case is the
difference between the 1 544 kbit/s network limit (see Table2/G.82) and the 1544 Mbit/s
synchronization interface network limit (see Fig 3/G.824).
Case2 application BIn this case the application recovers timing through the TDM signal; therefore there is no
differential jitter and wander between the clock and the data other than within the bandwidth of the
clock recovery since the data and clock are extracted from the same signal. The wander budget of
the CES is only limited by the timing quality requested by the application (e.g. RBS requirements)and not by the G.823 specification.
Note: this application is valid only for application with a single signal; if 2 signals are received
there might exist a differential jitter and wander between one signal and the clock extracted from
the other signal.
Figure A.2: Case 1 and Case 2 scenarios
IWF IWFSDH
IslandSDH
Island
CES
Island
IWF
CES
Island
IWF
Case 2
Case 1
Synchronization
networkSynchronization
network
PRCPRC
End
equipment
IWF IWFSDH
IslandSDH
Island
CES
Island
IWF
CES
Island
IWF
Case 2
Case 1
Synchronization
networkSynchronization
network
PRCPRC
End
equipment
8/7/2019 T-Packtiming
22/30
- 22 -
TD 147 (WP 3/15)
Case 3: When retiming is implemented at the output of the SDH islands, the amplitude of noiseon the PDH output is that of a synchronization interface; this allows to increase the wander
budget up to that of case 2 Application A in some configurations
Figure A.3: Case 3 scenario
The noise transfer characteristics of the CES have been identified to be an important issue.
Details on the measurement reference points are provided in figure A.4 (synchronized clock
configuration) and figure A.5 (independent clock configuration).
IWF IWFSDH
Island
SDH
IslandCES
Island
IWF
CES
Island
IWF
Case 3
Synchronization
networkSynchronization
network
PRC PRC
Endequipment
Synchronization
network
PRC
R
R = Retimer
IWF IWFSDH
Island
SDH
IslandCES
Island
IWF
CES
Island
IWF
Case 3
Synchronization
networkSynchronization
network
PRC PRC
Endequipment
Synchronization
network
PRC
R
R = Retimer
8/7/2019 T-Packtiming
23/30
- 23 -
TD 147 (WP 3/15)
Figure A.4 Measurement reference points in the synchronized clock configuration
Figure A.5 Measurement reference points in the independent clock configuration
Editors note: the figures above do not cover all possible measurement scenarios. Further
contributions are invited.
Editors note: packet delay and packet delay variations formal definitions shall be introduced in the
G.pactiming. Contributions are invited.
8/7/2019 T-Packtiming
24/30
- 24 -
TD 147 (WP 3/15)
Appendix I Characteristics of Ethernet switches and networks
Editors note: Expansions on the topics referred to in the body of this recommendation shall be
included in dedicated appendixes (e.g. details on clock recovery methods, introductory information
on Ethernet technologies, etc.).Editors note: this appendix provides initial information needed in order to build a network
reference model (to be defined in section 7). Further contributions on this subject are invited.
I.1 Delay Characteristics of Ethernet Switches
I.1.1 Functional Operations within an Ethernet Switch
From a black box perspective, an Ethernet frame passes through four functional operations in a
typical Ethernet switch. These are shown in Figure I-1:
Output
Port
Input
PortAdmission
Control
Switch
Fabric
Output
QueueClassifier
Figure I-1: Typical Functions within an Ethernet Switch
a. Classification identification of the flow to which the frame belongs, and determination ofthe output port and priority
b. Admission control application of traffic management for the flow (policing, shaping, marking)c. Switching forwarding to the appropriate output portd. Output queue waiting for a transmission slot on the output port. Typically queuing
policies such as strict priority, weighted fair queuing or round robin areapplied
The following sections discuss the delay properties of the various functions within a switch.
I.1.2 Input Stage Delay
The time required for the classification and admission control stages should be approximately
constant in most cases. However, depending on the switch design and the traffic loading on the
switch, the delay through these functions may vary. For example, in some switches, both
classification and admission control may be performed in software on a network processor. At full
load, the software may not be able to keep up with the number of frames to be processed, hence the
delay may increase, and some frame dropping may occur. The same may also be true of somehardware-based designs.
Figure I-2 shows a simplistic view of the variation of input stage delay with switch loading. Under
low traffic loads, the switch can cope with the number of frames passing through it without addingto the delay. As the frame rate increases, while the overall processing capacity of the switch is not
exceeded, the instantaneous frame rate may exceed the available processing rate. This will cause
frames to be buffered awaiting processing, and some extra delay to be incurred. Finally at some
point the mean incoming frame rate may exceed the processing capacity, causing the delay to be
increased further, and in some cases frames to be dropped due to lack of buffering capacity.
8/7/2019 T-Packtiming
25/30
- 25 -
TD 147 (WP 3/15)
Intrinsic
propagation delay
Incoming
Frame Rate
Minimum intrinsic
propagation delay
No bufferingrequired
Buffering
required
Processing
overload
Figure I-2: Variation of Input Stage Delay with Loading
I.1.3 Switch Fabric Delay
The delay through the switch fabric itself is also dependent on both switch architecture and trafficloading. For example, many switches operate scheduling algorithms for switching of frames from
input ports through to output ports, and this may cause a small variation in delay to the frames,
depending on their arrival time relative to the scheduler tick. However, in most cases this
variation in delay is small due to the high frequency at which the scheduler works.
At very high incoming data rates, the switch fabric itself may be overloaded, and unable to cope
with the full volume of traffic requiring switching. This will result in frames being dropped.
I.1.4 Output Queuing Delay
The amount of delay added by the output queue will depend on the queuing policy employed, and
the priority of the traffic flow. For example, a high priority flow (such as might be used for apacket timing flow) in conjunction with a strict priority policy might experience head-of-line
blocking delay. This is where although a frame has highest priority, it arrives at the output port
just after a low-priority frame has started to be transmitted. The high priority frame then has to wait
until the other frame has finished transmitting.
Figure I-3 shows the delay profile experienced by a population of high priority frames in
conjunction with a strict priority queuing policy. For the purposes of simplicity, this diagram
assumes frames experience an approximately constant delay through the other switch functions,
here termed intrinsic propagation delay through switch. A proportion of frames arrive at the
output queue at a time when there are no other frames currently being transmitted. These frames are
transmitted immediately. The remainder have to wait in the queue while the current transmission
completes. There may also be an additional delay due to other high priority packets also in thequeue.
8/7/2019 T-Packtiming
26/30
- 26 -
TD 147 (WP 3/15)
Frame delay
through switch
number of
frames
Transmission time of
maximum sized frame
Intrinsic propagation
delay through switch
Frames arriving when there are
no other frames being transmitted
subject to zero queuing delay
Head-of-line blocking - high priority
rames arriving when there is already a
low-priority frame being transmitted
Additional delay due to other
high priority frames in the queue
Figure I-3: Strict Priority Queuing: Head of Line Blocking
I.2 Characteristics of Switched Ethernet Networks
I.2.1 Topology of Ethernet Networks
Although there are many different possible network topologies, for the purpose of considering a
particular flow through a network, it can be modelled as a chain of Ethernet switches as shown in
Figure I-4. At each switch in the chain, an Ethernet frame has the potential to be delayed due to the
mechanisms described in section 0. This delay will be affected by the other traffic flowing through
the switch. Traffic directed to the same output port will affect the output queuing delay, while the
sum total of all traffic flowing through the switch (including that flowing to other ports) will affect
the processing and switch fabric delays.
Ethernet Switches
Network traffic routed to the
same output port
(affects output queuing delay)
Network traffic routed to
other output ports
(affects overload points)
Traffic flow ofinterest
Figure I-4: Data flows within an Ethernet Network
The length of the chain affects the total delay of the system; clearly the more switches there are, the
greater the total delay, also the greater the delay variation. However, in many Ethernet networks,
the length of the chain may be quite small. For example in a hierarchical network, there may often
be only two or three levels of hierarchy, yielding a chain length of up to five switches.
In some instances, a ring topology may be employed. Typically these may contain around ten
switches, giving a maximum distance around the ring of five switches. Occasionally,
interconnected rings may be used, which could double the distance to around ten.
8/7/2019 T-Packtiming
27/30
- 27 -
TD 147 (WP 3/15)
I.2.2 Traffic Patterns and Levels
With the exception of constant bit rate and real time traffic, most network traffic is extremely bursty
in nature. It has been observed that on almost any level one cares to look, traffic variation can be
observed. For example, at a very small level there is bursting due to the opening and closing of the
TCP window size. At a larger level, there may be bursting due to the nature of the application (e.g.
downloads of large files), while at a larger level still there may be bursting due to the time of day(e.g. higher activity levels during the day than at night).
When considering the delay performance of a TDM transport flow, the effects of other traffic in the
network have to be considered. For example, in Figure I-4, each of the network traffic flows may be
varying in some form, independently of the other flows.
ITU-T Recommendation G.1020 proposes the use of four-state Markov models for modelling
packet loss distribution. A similar technique could be applied to burst lengths in each flow, allowing
bursts and groups of bursts to be modelled. The longer term (e.g. diurnal variation) could then be
applied as a gradual variation in burst densities.
I.2.3 Disruptive Events in Ethernet Networks
There are several types of disruptive events that may cause sudden changes in delay in an
Ethernet network. The resulting delay changes may be permanent or temporary. These disruptive
events include:
Routing change, causing a permanent step change in delay Temporary network overload, causing a significant but temporary delay change Temporary loss of service, causing all packets to be lost for a period
8/7/2019 T-Packtiming
28/30
- 28 -
TD 147 (WP 3/15)
Appendix II Packet delay variation, Packet loss and congestion modelling
Editors note : the following information has been received from Q5/13 and has been agreed to be
the starting point for the modelling of packet loss and congestion:
Regarding network modelling for timing, the present state of the art is truncated Gaussian
distribution of delay variation, and extended Gilbert models for packet loss and congestion. The
former is somewhat unrealistic due to its high-pass nature in time, and can not model long range
correlations that make wander conformance difficult for purely adaptive recovery. The latter has
been shown to be a good match to true networks under a variety of conditions. Alternatively,
realistic network models that take routing mechanisms into account are also available, but require
precise definition of the network topology. To overcome this limitation one may define a small
number of network topologies as a standard test-bed.
For information, the Gilbert model is summarized in ITU-T G.1020 Appendix I.
8/7/2019 T-Packtiming
29/30
- 29 -
TD 147 (WP 3/15)
Appendix III Functional diagrams based onITU-T G805 and G.809
The basic idea behind the functional description of TDM transport over Packet Switched Networks
(PSN) is that it is network interworking defining a client-server relationship between the TDMclient and the PSN server layer network. The IWF is technically an adaptation that accepts the client
TDM characteristic information and processes it to enable its transfer over a trail in the server
network. Hence the PSN forms a link connection supporting the TDM trail, providing a function
that could also be filled by an SDH or ATM network.
There are differences between the various cases. ITU-T Y.1413 details the connection-oriented
(CO) case, while Ethernet and IP networks are connectionless (CL) and need to be described in
terms of ITU-T G.809 diagrams (this work is presently being carried out in Q7/13 for Y.tdmip). As
examples, we diagram the simplest structure-agnostic TDM-over-MPLS and TDM-over-IP cases
below.
MPLS/
TDM
AP
TDM
TDM Network
MPLS/
TDM
TDM CPTDM CP
TDM
AP
MPLS MPLS
MPLS TCP MPLS TCP
TDM Network MPLS Network
MPLS
CP = Connection point AP= Access Point TCP= Termination Connection Point
Figure III-1: TDM transport over a CO PSN
Editors note: this appendix is preliminary and shall be checked with the relevant Questions toensure alignment with the other Ethernet recommendations.
8/7/2019 T-Packtiming
30/30
- 30 -
TD 147 (WP 3/15)
IP/
TDM
IP AP
TDM
TDM Network
IP/
TDM
TDM CPTDM CP
TDM
IP AP
IP IP
IP TCP IP TCP
TDM Network IP Network
IP
CP = Connection point AP= Access Point TCP= Termination Connection Point
Figure III-2: TDM transport over a CL PSN
_____________