W H I T E PA P E R

www.albedotelecom.com

PTP Testing Overview

Accurate frequency distribution through packet Table 1
switched networks can be understood as an exten‐ Summary of ITU-T PTP Telecom Profiles main settings
sion of the TDM synchronization network based on a
ITU‐T ITU‐T
few new building blocks like the Synchronous Ether‐
G.8265.1 G8275.1
net Equipment Clock (EEC) and the Packet‐based
Equipment Clock (PEC). However, for time and phase Frame structure UDP Ethernet
distribution applications, where most of the interest Addressing mode Unicast Multicast
resides today, this approach does not work. One way / Two way Both Two way
One step / two step Both Both
Time and phase synchronization require new syn‐ Path delay mechanism End‐to‐end End‐to‐end
chronization architectures. The challenges to deliver Domain 4 ~ 23 24 ~43
the required accuracy level are especially important Priority 1 range ‐ 128
in these new scenarios. There is also a renewed in‐ Priority 2 range ‐ 0 ~ 255
terest in synchronization testing related both with Local priority range 1 ~ 255 1 ~ 255
network commissioning tasks and troubleshooting. Class 80 ~ 110 6, 7, 135,
140, 150,
160, 165,
This paper reviews existing testing techniques appli‐ 248, 255
cable to frequency distribution and introduces the Time scale Arbitrary, PTP
new techniques for phase and time applications. A PTP
minimum description of the technologies that en‐ BMCA Static Alternate
able accurate phase and time distribution, including BMCA BMCA
the Precision Time Protocol (PTP) is also addressed. Sync message rate 1/16 ~ 128 16
Delay request message rate 1/16 ~ 128 16
Announce transmission rate 1/16 ~ 8 8
1. PTP PROFILES Announce receive timeout 2 3 ~ 10

PTP plays a central role in most time and phase syn‐ As stated in IEEE 1588‐2008, the purpose of a PTP
chronization architectures and is the key technology profile is to allow organizations to define specific se‐
in all current applications requiring a high degree of lections of attribute values and optional features of
accuracy, usually in the range of a few microseconds PTP that, when using the same transport protocol,
and sometimes in the sub‐microsecond range. inter‐work and achieve a performance that meets
the requirements of a particular application.
The IEEE 1588‐2008 standard, where PTP is defined,
is flexible enough to allow for very different ways of Typical profile examples are the Default profiles de‐
using the protocol, including both frequency and fined in IEEE 1588‐2008 (two basic general purpose
WP-PTPTESTINGGUIDE 06/18

time distribution applications. Due to its flexibility, profiles), the Power Profile (IEEE C37.238‐2011), the
PTP may operate in different profiles. These profiles Enterprise profile (currently an IETF draft) and the
are not interoperable with each other; they offer dif‐ Telecom profiles (ITU‐T G.8265.1, G.8275.1 and
ferent performance levels and involve different re‐ G.8275.2). Discussion about the PTP Telecom pro‐
quirements for the network.

All rights reserved. No part of this document may be stored, copied or transmitted, by any means, without the permission in written of the Legal Owner

© 2018 ALBEDO Telecom

 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 2 / 27

files together with the Default profiles are the main but to extend the existing network to include PTP

A L B E D O - WHITE PAPER
subject of this paper. as a protocol suitable to carry synchronization with
a minimum impact in the installed infrastructure
ITU-T G.8260
based on TDM technology (or Synchronous Ether‐
BASICS

Definitions & metrics

Frequency Time
net).
ITU-T G.8261 ITU-T G.8271
Basic topics Basic topics One interesting feature of the ITU‐T G.8265.1 pro‐
Network limits file is the ability to operate in one‐way mode. PTP
ITU-T G.8261.1
masters use the Sync message flow to share time
ITU-T G.8271.1
Network limits Network limits
stamps with their peers (slave clocks, boundary
NETWORK
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

clocks). If time synchronization between the mas‐

LIMITS

ITU-T G.8271.2
Network limits (PTS)
ter and its peers is required, then the time it takes
for the remote end to receive the time stamp has to
be compensated for in some way. This is done
ITU-T G.8262 ITU-T G.8272
EEC specification PRTC specification
through either the end‐to‐end or peer‐to‐peer path
delay mechanisms. If no time synchronization is re‐
IITU-T G.8262.1 ITU-T G.8272.1 quired, there is no need to apply any delay com‐
eEEC specification ePRTC specification
pensation and the message flows associated to the
ITU-T G.8263 ITU-T G.8273 path delay mechanism could be removed. This
DEVICE SPECIFICATION

PEC-S-F specification Framework one‐way operation mode is allowed by IEEE

ITU-T G.8266 IITU-T G.8273.1 1588‐2008 and it is optional within ITU‐T G.8265.1.
PEC-M-F specification T-GM specification

ITU-T G.8273.2 PTP operation has to be compatible with existing

T-BC, T-TSC spec telecommunication networks which may not in‐
IITU-T G.8273.3
clude specific support for PTP. Actually, it is as‐
T-TC specification sumed that the network may be completely
IITU-T G.8273.4
unaware of PTP. This requirement restricts the way
APTSC specification the protocol has to be deployed in several aspects:

ITU-T G.8264
• UDP over IPv4 (IEEE 1588‐2008, Annex D) is the
ITU-T G.8275
SyncE requirements Requirements
chosen transport protocol rather than Ethernet
REQUIREMENTS

or other protocols. This is because of the uni‐

ITU-T G.8265
PTP / NTP requirements
versal availability of IPv4.

ITU-T G.8265.1 ITU-T G.8275.1

• Unicast is the only allowed transmission mech‐
PTP Profile (freq.) PTP Profile (FTS) anism. Multicast may be more efficient but pro‐
visioning multicast is also more complex and it
In force ITU-T G.8275.2
PTP Profile (PTS)
may not be available, or even if it is available
Under development the network administrator may decide to re‐
strict its use for security reasons. In ITU‐T
Figure 1 ITU-T packet switched network G.8265.1 networks, PTP slaves must request
synchronization standards and the relationship among
them.
permission from the master to exchange PTP
messages through the signaling mechanism de‐
fined by IEEE 1588‐2008 and complemented by
ITU‐T G.8265.1.
The ITU-T G.8265.1 Frequency Profile
• No on‐path support through boundary or trans‐
The aim of the ITU‐T G.8265.1 PTP profile for fre‐ parent clocks is used. Actually, PTP masters
quency synchronization is to adapt PTP to common (PEC‐M) and slaves (PEC‐S) are the only PTP en‐
telecom network synchronization practice. The tities considered by ITU‐T G.8265.1 profile. To
purpose of this profile is not to provide better per‐ compensate for the lack of support from the
formance than any previous protocol or to define network, the ITU‐T G.8265.1 standard allows
new functionality in the synchronization network for message rates higher than in other profiles

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 3 / 27

(up to 128 messages/s for Sync and Delay re‐ • Definition of a new clock reference selection al‐

A L B E D O - WHITE PAPER
quest messages). Another consequence of the gorithm. Telecom networks must behave in a
lack of on‐path support is that the path delay very predictable way. Synchronization slaves
mechanism cannot be peer‐to‐peer and there‐ choose their time reference using static value
fore if a path delay mechanism is used it has to configured in the device (the Local priority), the
be end‐to‐end. QL if available, and certain alarms detected in
the clock interface such as Loss Of Signal (LOS)
One of ITU‐T G.8261 most important requirements or the Alarm Indication Signal (AIS). These
is the need for smooth inter‐operation with existing mechanisms are replicated by (1) defining a Lo‐
synchronization networks. Some features added for cal priority for PTP to replace Priority 1 and Pri‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

this purpose are: o r i t y 2 , ( 2 ) re p l a c i n g t h e d y n a m i c a n d

somewhat unpredictable Best Master Clock Al‐
• Re‐use of the ITU‐T G.781 Quality Level (QL). gorithm (BMCA) defined in IEEE 1588‐2008 by
ITU‐T synchronization networks and IEEE PTP
the more simple and more deterministic Static
networks both have their own way to signal the
BMCA which works in the same way as the nor‐
quality level they are supplying. The ITU‐T mod‐
mal decision algorithm used in telecom net‐
el is based in QL codes included in the Synchro‐
works to choose the synchronization reference,
nization Status Message (SSM), PTP uses the
(3) defining a new Packet Timing Signal Fail
clockClass attribute. These mechanisms are not
(PTSF) alarm to distribute information about
compatible, and therefore the clockClass can‐
synchronization faults such as loss of announce
not be directly used by ITU‐T networks. To add
messages, loss of timing messages or excessive
compatibility between both models ITU‐T
Packet Delay Variation (PDV).
G.8261 defines a range within the clockClass to
add the ITU‐T G.781 QL values. With this modi‐
fication, network elements can handle a PTP The ITU-T G.8275.1 Phase / Time Profile
reference in the same way as with SDH / SONET
or Synchronous Ethernet inputs. Unlike ITU‐T G.8265.1, the purpose of ITU‐T
G.8275.1 is to enable the deployment of accurate
Table 2 phase and time distribution in a telecommunica‐
ITU-T G.784 and G.8265.1 Quality Levels tions network based on the PTP protocol. Also un‐
like ITU‐T G.8265.1, the ITU‐T time profile requires
ITU‐T G.781 ITU‐T extensive on‐path support; actually, all network
Option I Option II
QL G.8265.1 QL equipment that must deal with PTP traffic must be
0 82 ‐ QL‐STU PTP aware. Switches must implement the Telecom
1 80 ‐ QL‐PRS Boundary Clock (T‐BC) function and endpoints may
2 84 QL‐PRC ‐ be, depending on their role, Telecom Grandmasters
3 88 ‐ ‐ (T‐GMs) or Telecom Time Slave Clocks (T‐TSCs).
4 90 QL‐SSU‐A QL‐TNC Transparent clocks are not used. All these are man‐
5 92 ‐ ‐ datory requirements within ITU‐T G.8275.1 and
6 94 ‐ ‐ they justify the name of PTP telecom profile for
7 86 ‐ QL‐ST2 phase / time synchronization with full timing sup‐
8 96 QL‐SSU‐B ‐ port from the network.
9 98 ‐ ‐
10 102 ‐ QL‐ST3 / Why is the Full Timing Support (FTS) profile so de‐
QL‐EEC2 manding? The answer is that modern applications
11 104 QL‐SEC / ‐ (most often, applications related to cellular and
QL‐EEC1 wireless communications) require highly accurate
12 106 ‐ QL‐SMC time and phase synchronization, usually in the mi‐
13 100 ‐ QL‐ST3E crosecond range. Another answer to the same
14 108 ‐ QL‐PROV question is that requirements are so strict because
15 110 QL‐DNU QL‐DUS technology exists that is capable of fulfilling these

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 4 / 27

requirements, even if it is at a considerable engi‐ Announce messages; it is appended to the messag‐

A L B E D O - WHITE PAPER
neering cost. es received in the relevant port and it is then used
in the decision algorithm. Actually, the decision al‐
Table 3 gorithm as been modified: In the Alternate BMCA,
Cellular interfaces and their phase accuracy the clockClass attribute has the strongest weight.
requirements

Application Requirement
(a) Deployment case 1 (b) Deployment case 2
GSM, WCDMA‐FDD, LTE‐FDD None
Network time reference Network time reference
TD‐SCDMA ±1.5 s
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

CDMA2000 ±3 s, ±10 s

WiMAX‐TDD ±1 s, ±1.5 s PRTC PRTC

LTE‐TDD ±1.5 s, ±5 s A A

LTE‐A ±1.5 s, ±5 s
T-GM T-GM

B B
The ITU‐T G.8275.1 profile is designed for maxi‐

PTP-aware packet switched network

PTP-aware packet switched network

mum performance and efficiency and hence the T-BC T-BC
frame structure is Ethernet (IEEE 1588‐2008) with
multicast addressing. This is well suited for the cur‐
rent architecture of Metropolitan Area Networks
(MANs) that are based on Ethernet pseudowires
and other related technologies.

T-BC T-BC
One of the key features of ITU‐T G.8275.1 is grand‐
master selection and protection. The Alternate C C
BMCA defined by this standard is a compromise be‐
tween the determinism required by telecom net‐ T-TSC T-TSC
works and the flexibility of the default BMCA from
D
IEEE 1588‐2008.

The Alternate BMCA is a dynamic protocol, in the E

same way as the IEEE 1588‐2008 BMCA. However,
unlike the IEEE 1588‐2008 BMCA, the Alternate E
BMCA defines a fixed role for each PTP entity:
T‐GM, T‐TSC or T‐BC. The IEEE 1588‐2008 Ordinary
Clocks (OCs) which may become masters or slaves
depending on the result of the BMCA are not al‐ Figure 2 ITU-T G.8271.1 / G.8275.1 network reference
model. The PTP endpoints are T-GMs and T-TSCs. All
lowed within the ITU‐T G.8275.1 framework. The packet transmission / switching network elements
purpose of the Alternate BMCA is to let the slave implement the T-BC function.
clocks decide which grandmaster to use and to al‐
low for a dynamic, loop‐free architecture. With this
objective in mind, the ITU‐T phase / time protocol The clockClass attribute from ITU‐T G.8265.1 can
defines a new port‐specific attribute, notSlave, that not be reused in this profile because the ITU‐T
is set to true in the T‐GM, false in the T‐TSC and con‐ G.781 classes are usable for frequency references
figurable to true or false in the T‐BC. only. Instead, the phase / time profile defines a
new QL scale. The new clockClass scale allows for
The ways the priorities are managed by the Alter‐ T‐GMs or T‐BCs that may operate in partial holdover
nate BMCA also differ from the IEEE 1588‐2008 state without a usable time reference but still
BMCA. Priority 1 is not used and it is statically con‐ traceable to some frequency source such as a PRC
figured to 128. There is a new port specific attri‐ or a Synchronization Supply Unit (SSU).
bute, the Local Priority, that is not delivered in

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 5 / 27

Table 4 relaxed set of requirements to the network com‐

A L B E D O - WHITE PAPER
ITU-T G.8275.1 clockClass attribute pared with the FTS architecture. The important
point is that PTS does not require all transit nodes
F.trac
Class Entity Meaning from the grandmaster to the slave to be PTP aware.
eable
In other words, FTS becomes PTS if at least one T‐BC
6 T‐GM True Connected to a PRTC in locked is replaced by a non‐PTP aware device.
mode (e.g., PRTC traceable to
GNSS)
(a) PTS architecture (b) APTS architecture
7 T‐GM True Holdover, within holdover specifi‐
cation, traceable to QL‐PRC / Network time reference
QL‐PRS frequency source
7 T‐GM False Holdover, within holdover specifi‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

Network time reference

cation, non‐traceable to QL‐PRC / PRTC
QL‐PRS frequency source
135 T‐BC True Holdover, within holdover specifi‐ PRTC
cation, traceable to QL‐PRC /

PTP-aware island
T-GM
QL‐PRS frequency source
135 T‐BC False Holdover, within holdover specifi‐
cation, non‐traceable to QL‐PRC / T-GM
QL‐PRS frequency source T-BC-P
140 T‐GM True Holdover, out of holdover specifi‐
cation, traceable to QL‐PRC /
QL‐PRS frequency source
150 T‐GM False Holdover, out of holdover specifi‐
cation, traceable to QL‐SSU‐A / Packet Switched Packet Switched
QL‐ST2 frequency source Network Network

160 T‐GM False Holdover, out of holdover specifi‐

cation, traceable to QL‐SSU‐B / GNSS
Backup
QL‐ST3E frequency source
PTP-aware island

T-BC-P
165 T‐BC True Holdover, out of holdover specifi‐
False cation APTSC
248 T‐GM True Without time reference since
T‐BC False start‐up T-TSC-P
255 T‐TSC True Slave only OC (does not send
False Announce messages)

PTS / APTS Telecom Profile

The strong on‐path support requirements from the Figure 3 Architectures considered in ITU-T G.8275
FTS profile limits its applicability to Greenfield de‐ and ITU-T G.8275.2 for time and phase distribution
ployments or to networks simple and modern through networks with partial timing support.
enough to allow for a deep re‐engineering. For this
reason, the ITU‐T has published a new PTP profile
for phase / time synchronization, but requiring Non‐PTP aware devices or islands are still expected
only partial timing support. This new profile has to provide good performance through mechanisms
been released as the ITU‐T G.8275.2 standard. such as packet prioritization, congestion avoidance
and control or by any other mechanism. At least
In order to understand why the ITU‐T G.8275.2 they are expected to do so under moderate traffic
standard is relevant, it is first necessary to highlight load.
the advantages and disadvantages of the Partial
Timing Support (PTS) and Assisted Partial Timing The APTS architecture evolves from deployments
Support (APTS) architectures. that rely entirely on GNSS. The advantage of these
architectures is that they do not require any syn‐
Both PTS and APTS are time and phase distribution chronization support from the network but on the
architectures. PTS is the result of applying a more other hand they require massive GNSS facility in‐

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 6 / 27

stallation at the network edges and they are vulner‐ APTS has emerged as a GNSS‐assisted architecture

A L B E D O - WHITE PAPER
able to GNSS signal jamming or spoofing. that uses PTP for backup rather than physical layer
synchronization. The main advantage is that full
(a) Distributed PRTC architecture timing support from the network is not required.
PRC
The ITU‐T G.8275.2 profile could be reused for
APTS. In order to keep the PTP synchronization
quality under control the path from the T‐GM that
Packet or Circuit
Switched Network generates the backup clock to the APTSC node
must be as short as possible. This is a difference be‐
PHY Sync
tween APTS and physical layer synchronization ar‐
GNSS
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

chitectures where the timing source location does

not really matter and it could be installed in the
PRTC core network, far from the edge. There are two pre‐
Endpoint

ferred locations for the PRTC in APTS:

1. At the aggregation sites. The PTP timing is then
transmitted to the APTSCs though the backhaul
network.
(b) APTS architecture

T-GM PRTC 2. At selected endpoint sites. These special sites

are then used to deliver the PTP timing to other
Packet switched
endpoints through the backhaul network.
network

Table 5
GNSS PTP Sync Summary of ITU-T PTS / APTS Profile

ITU‐T G8275.2
APTSC
Frame structure UDP
Endpoint

Addressing mode Unicast

One way / Two way Both
One step / two step Both
Path delay mechanism End‐to‐end
Figure 4 Comparison between the legacy distributed
Domain 44 ~63
PRTC and APTS architectures.
Priority 1 range 128
Priority 2 range 0 ~ 255
Older GNSS‐assisted architectures may be Local priority range 1 ~ 255
equipped with some kind of physical layer frequen‐ Class 6, 7, 135, 140, 150, 160,
165, 248, 255
cy synchronization for backup purposes. With some
Time scale PTP
legacy transport technologies such as
BMCA Alternate BMCA
SDH / SONET, frequency synchronization is inher‐
ent and for this reason this architecture is suitable Sync message rate 1 ~ 128
to deployments that still rely on circuit switched Delay request message rate 1 ~ 128
networks for backhaul and transport. The same ap‐ Announce transmission rate 1 ~ 8
proach can also applied to packet switched net‐ Announce receive timeout 2
works through Synchronous Ethernet technology
but in this case the backhaul network is required to An advantage of APTS and other GNSS‐assisted ar‐
implement Synchronous Ethernet in all interfaces chitectures is that they enable the network opera‐
used to carry synchronization. These architectures
tors to monitor the endpoint nodes. This is because
have been used mainly for frequency distribution
at least two references are available in slave clocks
applications but they are compatible with phase and they could be used to mutually control their
and time applications because GNSS provides both
performance. One application is GNSS spoofing de‐
frequency and time.

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 7 / 27

tection which is one of the main weaknesses of time the local oscillator is locked to another

A L B E D O - WHITE PAPER
GNSS‐assisted architectures. clock reference. In this case, the local oscillator
inherits some of the properties from the refer‐
The ITU‐T G.8275.2 profile is a mixture of the ITU‐ T ence. A typical situation is to discipline the in‐
G.8265.1 and G.8275.1 profiles. The encapsulation ternal oscillator with a GNSS source. It is then
is UDP with unicast transmission like in ITU‐T expected that the local oscillator gets the long
G8265.1 but on the other hand it allows for the term frequency / time accuracy of the GNSS
same flexible master clock selection as ITU‐T source.
G.8275.1 though the Alternate BMCA. The clock‐
Class attributes are also closely related to the FTS
• Clock reference: Sometimes, such as in jitter
tests, the clock reference can be recovered
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

profile to report holdover and time / frequency

from the signal under test by some kind of fil‐
traceability in the same way this profile does.
tering processes, but more commonly the clock
reference constitutes an independent input in
An interesting point about the PTS / APTS profile is
the test. Two main alternatives are used in
that the one‐way operation is not forbidden. This
practical scenarios: Primary Reference Clocks
fact makes it impossible to apply delay compensa‐
(PRCs) / Primary Reference Time Clocks (PRTCs)
tion in the slave clock. At first glance, this feature
and Global Navigation Satellite System (GNSS)
could seem unnatural in a profile that is designed
signals.
for accurate time and phase distribution. The ex‐
planation is that one‐way operation could be useful • Network emulator: The test set has to be con‐
in some APTS configurations. Ahead of a GNSS sig‐ nected to a device or network to measure and,
nal degradation or outage, the APTSC would react to some extent, it has to be compatible with
by switching over to the backup PTP reference. the system where it is connected. Sometimes it
However, if the one‐way operation is used, the ref‐ is enough to achieve interface compatibility like
erence switchover could be replaced with a partial with TDM or 1PPS testing, but in other situa‐
holdover event: frequency would be obtained from tions both interface and protocol compatibility
PTP (the backup reference) but time is kept from is required. The most typical example of proto‐
the lost GNSS input. This partial or time holdover col compatibility is PTP testing that requires the
operation mode could be declared whenever all test set to be interoperable not only with the
time references become unavailable but there are physical interface (usually Ethernet and IP) but
still some usable frequency references. The advan‐ also with the PTP protocol itself. Specifically,
tage of the partial holdover is that it maintains an most often, the test set has to implement some
accurate time for much longer than a total or fre‐ of the functionality of a PTP slave. It has to not
quency holdover which is declared when the clock only decode timing information from remote
relies only on its local oscillator. PTP entities but also generate different kinds of
PTP messages such as signaling messages and
delay request messages.
2. TESTING CHALLENGES • Test engine: The purpose of any test equipment
is to measure and generate a result based on
In any synchronization test set, there are at least this measurement. In a synchronization tester
five critical blocks or functions: the local oscillator, the measurement results consist of a sequence
the clock reference, the network emulator, the test of numbers computed by comparing a relative
engine and the post‐processing unit. Each block in‐ or absolute time associated to the device under
volves its own challenges for reliable and accurate test and the time from the clock reference. For
tests: a typical wander measurement the test unit
• Local oscillator: The local oscillator constitutes may generate several tens or hundreds of test
an internal frequency source required by any results per second. The exception to this rule is
synchronization test equipment. The local oscil‐ jitter testing, as a high frequency phase impair‐
lator is expected to be accurate within certain ment. Measurement bandwidth for jitter is in
limits. It could theoretically be used as an au‐ the range of kHz or beyond and it requires a dif‐
tonomous (internal) reference but most of the ferent approach.

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 8 / 27

• Post processing unit. This building block com‐ Table 6

A L B E D O - WHITE PAPER
putes synchronization performance metrics Frequency accuracy of oscillators commonly used for
Telecom applications
from the raw measurement results. Many im‐
pairment sources are either random or difficult Frequency accuracy
to predict (variable waiting time in queues, os‐
TCXO 2 ppm ‐ 0.2 ppm
cillator noise, variations on GNSS coverage,
temperature fluctuations). For this reason, the OCXO, DOCXO 10 ppb ‐ 0.1 ppb
associated performance metrics are statistical Rubidium 5x10‐11 or better
in nature. Some common statistics are general Cesium 10‐12 or better
purpose metrics like averages or standard devi‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

ations while some others have been defined

specifically for synchronization applications terms, the crystal behaves like a tunable electrical
such as the Allan Deviation (ADEV) or the Time circuit of a very high Q‐factor.
Deviation (TDEV). Randomness of synchroniza‐
tion test results is a challenge in terms of re‐ The accuracy in these kind of oscillators is limited
peatability. For example, estimations of the by the sensitivity to temperature changes in the
standard deviation of some kinds of phase crystal’s natural oscillation frequency. Temperature
noise does not converge to any specific value, Compensated Crystal Oscillators (TCXOs) have bet‐
even in very long tests; there is no way to mea‐ ter performance in terms of temperature sensitivi‐
sure (or even to define) an standard deviation ty. They are based on a Voltage Controlled Crystal
for such processes. In other cases, the impair‐ Oscillator (VCXO) and a temperature sensitive cir‐
ment processes involved in synchronization cuit that applies a voltage that corrects the fre‐
tests have a very low frequency (hours, days, quency of the VCXO at any temperature within the
weeks) or they are not periodic at all. Measure‐ operating temperature range.
ments involving non‐stationary processes may
be very long and even in this case may not be A different approach to temperature stabilization is
totally repeatable. implemented by Oven Controlled Crystal Oscillators
(OCXOs). This type of oscillator has a temperature
controlling circuit to maintain the crystal and key
The Local oscillator components at a constant temperature. Double
Oven Controlled Crystal Oscillators (DOCXOs) are a
Synchronization testers are equipped with accurate refinement on the same technology that uses two
(or not so accurate) local oscillators. The alterna‐ separate heating circuits coupled together. DOCXOs
tives for this important component are described in are even better than OCXOs and their frequency ac‐
the following lines. curacy could be a fraction of a part per billion. The
inconvenience of OCXOs / DOCXOs is that they are
In order to understand oscillators, it is useful to rate more expensive and they consume more power
the accuracy of a standard wristwatch clock that than TCXOs.
has a quartz Crystal Oscillator (XO). The most intui‐
tive way to qualify the accuracy of a clock is the The accuracy of OCXOs / DOCXOs is the best that
fractional frequency offset measured in parts per can be achieved through a piezoelectrically induced
million or other units. For the wristwatch the accu‐ vibration. Increasing the performance requires us‐
racy is on the order of 10 ppm (10‐5). We will see ing atomic clocks that use frequencies of specific
that there exist technologies that enable improve‐ electron state transitions of certain atoms or mole‐
ments many orders of magnitude better than this cules. In the telecommunication industry two
basic accuracy. atomic frequency standards are commonly used.
One is the Cesium atomic beam standard that uses
Some crystals like quartz are capable of storing or the transition between two ground levels of the hy‐
supplying electrical energy depending on the me‐ perfine spectrum of the 133Cs atom equivalent to
chanical stress applied to them. This is known as 9,192,631,770 Hz. The second frequency standard
piezoelectricity and enables the crystal to couple is based on Rubidium vapor cells that use an hyper‐
mechanical and electrical vibrations. In practical

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 9 / 27

fine transition of the 87Rb isotope to generate a fre‐ Cesium tubes are by no means the best atomic

A L B E D O - WHITE PAPER
quency of 6,834,682,610.904 Hz. clocks available today. Research in accurate time
sources is a very active field and current accuracies
achieved in new engines are in the range of 10‐15 or
Physics
better. Of the new techniques used to improve the
RF Synthesizer
package performance of basic beam devices the most im‐
portant is probably laser cooling of atoms. Using
cold atoms reduces the contributions to error from
Loop Clock the Doppler effect, atom collisions and thermal ra‐
control output
Oscillator
diation, thus increasing the device accuracy by sev‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

eral orders of magnitude.

Figure 5 Simplified block diagram of an atomic clock.
The feedback from the physics package is used to tune
a conventional oscillator such as an OCXO. Clock References

The correct way to assess how good a clock refer‐

The operation of an atomic clock is based on an in‐ ence is depends on the metric to be measured, the
terrogation‐correction mechanism. A conventional measurement frequency band and, if disciplining is
crystal oscillator generates a frequency that is used used, the local oscillator specifications. For exam‐
to interrogate a “physics package” that contains the ple, in jitter tests the measurement band is usually
Cesium tube, the Rubidium vapor cell or any other in the kilohertz range. To get a valid clock reference
device based on atomic resonances. The “physics it is enough to apply a 10 Hz low pass filter to the
package” generates an error signal that depends on test signal. The filtered test signal is perfectly suit‐
the de‐tuning of the test frequency from the atomic able to be used as a clock reference in a jitter test.
resonance. The error signal is processed and the re‐ If the measurement band is to be extended to low‐
sult is used to control the frequency generated by er frequencies, this mechanism becomes more dif‐
the crystal oscillator that is also the clock output. ficult to implement because the cutoff frequency in
The key piece of this setup is the “Physics package” the low pass filter has to be reduced beyond practi‐
that behaves like an extremely high Q‐factor band‐ cal limits. In wander tests, the measurement band
pass filter. often starts in the millihertz or microhertz range.
For this reason, the clock reference is an additional
Aging effects are smaller in Cesium tubes than in input to the test. The reference signal phase is ex‐
Rubidium clocks. For this reason the Rubidium fre‐ pected to be stable in the measurement band‐
quency standard is not suitable to operate in PRCs. width.
However, Rubidium is well adapted for SSUs that
are usually disciplined by a primary source or GNSS. If we now focus on external clock references, there
These clocks are also perfectly suited to test appli‐ are three popular alternatives used in practical ap‐
cations due to their low power consumption, com‐ plications: PRCs, PRTCs and GNSS references. This
pact size and relatively low price. On the other section deals mainly with PRCs and PRTCs. As the
hand, Cesium tubes may offer frequency accuracies performance of GNSS references is strongly depen‐
of 10‐12 or better and good long term frequency dent on how they are used to discipline the local os‐
stability. These oscillators are therefore ideally suit‐ cillator, these are discussed in a section devoted to
ed to be installed in PRCs. oscillator disciplining.

A special type of vapor cell atomic clock is the so For many years, the best clocks available for tele‐
called Chip‐Scale Atomic Clock (CSAC). Current communication applications have been PRCs. The
CSAC implementations have the advantages of low PRC performance is described in ITU‐T G.811 in
power consumption and short warm up period. terms of three interface independent metrics: The
Performance in terms of fractional frequency accu‐ Maximum Time Interval Error (MTIE), the Time De‐
racy is around 0.1 ppb, in line with the best avail‐ viation (TDEV) and fractional frequency offset. Ac‐
able OCXOs. tually, the specification for the fractional frequency
offset is embedded in the MTIE and therefore all

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 10 / 27

that can be said about the PRC wander is packaged mance. PRCs are expected to always work in a free

A L B E D O - WHITE PAPER
in specifications for MTIE and TDEV. There is a lot to running state.
be explained about MTIE and TDEV but it in simple
terms, the MTIE defines the maximum phase excur‐ (a) ITU-T G.811 PRC wander (MTIE)
MTIE (s)
sion within an observation window and the TDEV
1000
measures the noise level of a clock source in the
given observation window. Both the MTIE and
100
TDEV depend on a time parameter, the observation
window. Because of this, their limits are expressed
10
though masks specified in terms of a variable ob‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

servation window. Unlike MTIE and TDEV, the frac‐

tional frequency offset is just a number, 10‐11 for a 1
0.3
PRC. We have seen that Cesium beams are able to
0.1
achieve this accuracy level.
0.025
0.01
10-1 1 10 102 103 104 105 106 107 108
Low pass Observation interval  (s)
filter

(a) ITU-T G.811 PRC wander (TDEV)

Phase Post-processing
Input

TDEV (ns)
detection unit
Test result

Phase 1000
comparison

(a) Closed loop measurement (jitter)

100

30

Clock 10
reference
3
Test result

Phase Post-processing 1
Input

detection unit
0.1 1 10 100 1000 10000
Phase
comparison Observation interval  (s)

(b) Open loop measurement (wander) Figure 7 ITU-T G.811 performance limits of a PRC
expressed in terms of MTIE and TDEV.
Figure 6 Simplified block diagrams of a jitter and a
wander test equipment. (a) Closed loop measurement
typical of jitter measurements. (b) Block diagram A PRC locked to a time reference with a good long
corresponding to an open loop test required to measure
wander.
term accuracy, typically derived from a GNSS inter‐
face, is a possible realization of a PRTC. While PRCs
are frequency references, PRTCs are designed to
PRCs are not expected to be locked to any external provide time outputs. Sometimes, PRTCs have 1
reference. They are devices designed to provide pulse per second (1PPS) / Time of Day (ToD) out‐
synchronization but they are accurate enough to puts which can be used to feed other devices re‐
not require synchronization to an external refer‐ quiring accurate time synchronization but often the
ence. For this reason, ITU‐T G.811 does not contain PRTC function is packed together with a PTP grand‐
any specification about locked or holdover perfor‐ master; it is quite common to see commercial PTP
grandmasters with one or several GNSS inputs. Be‐
fore addressing the detailed description of PRTCs, it

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 11 / 27

is worth looking at time references and their prop‐ Table 8

A L B E D O - WHITE PAPER
erties more carefully. Leap seconds added

UTC Date Amount TAI to UTC offset

A time reference is made up of an (accurate) fre‐
quency reference plus a counting device that in‐ 1972‐06‐30 +1 second 11 seconds
creases as new cycles in the frequency reference 1972‐12‐31 +1 second 12 seconds
are recorded. Actually, as it would be unpractical to 1973‐12‐31 +1 second 13 seconds
count cycles one by one, the standard time unit is 1974‐12‐31 +1 second 14 seconds
defined to be an integer number of cycles taken 1975‐12‐31 +1 second 15 seconds
from a specific frequency reference. For the sec‐ 1976‐12‐31 +1 second 16 seconds
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

ond, the frequency reference is the same atomic 1977‐12‐31 +1 second 17 seconds
transition from the 133 Cs that is used in Cesi‐ 1978‐12‐31 +1 second 18 seconds
um‐beam atomic clocks. One second is thus de‐ 1979‐12‐31 +1 second 19 seconds
fined to contain exactly 9,192,631,770 cycles of this 1981‐06‐30 +1 second 20 seconds
atomic reference. 1982‐06‐30 +1 second 21 seconds
1983‐06‐30 +1 second 22 seconds
Table 7 1985‐06‐30 +1 second 23 seconds
Commonly used time Scales 1987‐12‐31 +1 second 24 seconds
1989‐12‐31 +1 second 25 seconds
Time Scale Definition
1990‐12‐31 +1 second 26 seconds
TAI International Atomic Time. Weighted 1992‐06‐30 +1 second 27 seconds
average of the time kept by about 200
atomic clocks in over 50 national labora‐ 1993‐06‐30 +1 second 28 seconds
tories worldwide. 1994‐06‐30 +1 second 29 seconds
UTC Coordinated Universal Time. Atomic scale 1995‐12‐31 +1 second 30 seconds
compensated by an integer number of 1997‐06‐30 +1 second 31 seconds
seconds so that the difference with UT1 is
less than 0.9 seconds. 1998‐12‐31 +1 second 32 seconds
UT0 Uncorrected UT as derived from astro‐ 2005‐12‐31 +1 second 33 seconds
nomical observations or from measure‐ 2008‐12‐31 +1 second 34 seconds
ments carried out from the GPS system. 2012‐06‐30 +1 second 35 seconds
UT1 UT0 corrected for the polar motion of the 2015‐06‐30 +1 second 36 seconds
Earth.
2016‐12‐31 +1 second 37 seconds
UT2 UT1 corrected for the regular slowing
down and speeding up of the Earth in
winter and summer. It is now considered
obsolete. Ephemeris Time (ET) uses certain astronomical
GPS time This time scale was designed to match events that are supposed to happen at regular in‐
UTC in the period from 1980‐01‐01 to tervals. The accuracy of the UT is limited because of
1981‐06‐30 but as no leap seconds have irregularities in the Earth’s rotation. The Earth’s ro‐
been added since that date, the GPS time
is now (2017) 18 seconds ahead UTC and
tation axis and speed are known to change with
19 seconds behind TAI. time and different adjustments to UT time are nec‐
essary in order to maintain accuracy. That’s the rea‐
son there are UT0, UT1 and UT2 time scales
The specification of the time reference must also depending on the correction added to account for
contain an origin or epoch. The epoch is defined as these irregularities. Being independent of the
the point of time that contains zero units of time. Earth’s rotation, ET is more accurate. Actually, the
The specification of the time unit, together with definition of a second based on ET replaced the
the epoch make up a time scale. It can therefore be definition based on the mean solar day in 1956 and
concluded that a time reference is a frequency ref‐ was used until 1967 when it was replaced by the
erence with a specific time scale. atomic second. The main drawback of ET is that it is
necessary to wait for astronomical events to hap‐
Historically time scales (the calendar) have been pen to adjust the time. Atomic time is readily at any
based in astronomical observations. Universal Time time.
(UT) is based in the concept of mean solar day. The

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 12 / 27

Time scales based on astronomical observations ence between a given time and a reference time,

A L B E D O - WHITE PAPER
were replaced by time scales based on atomic time both expressed in the same time scale. TE is de‐
at the end of 1950s. Coordinated Universal Time fined so that it is positive if the test signal is ahead
(UTC) is an atomic time scale defined to match the of the reference and negative otherwise.
UT2 time (and later UT1) within a certain error mar‐
gin. Before 1972 the adjustment mechanisms in‐
Phase reference
cluded slight modifications in the standard second
length and small phase adjustments of 1/10 of sec‐
ond. From 1972 onwards the leap second mecha‐
nism was agreed upon; through this mechanism, 0 Tr 2Tr 3Tr 4Tr 5Tr
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

one day (always chosen to be January 1st or June

Phase offset
30th) is allowed to have one more or one less sec‐
ond than an standard day. The leap second mecha‐
nism has been applied 27 times to compensate for
the offset measured from UT1. Closely related with 0 Tr 2Tr 3Tr 4Tr 5Tr
the UTC time scale, there is the International Atom‐
ic Time (TAI) scale which is exactly the same as UTC Frequency offset

time but it contains no leap seconds. The TAI time

was adjusted 10 seconds ahead of UTC at the begin‐
ning of 1972. This means that the current offset 0 Tr 2Tr 3Tr 4Tr 5Tr
(2017) is 37 seconds. This difference accounts for Time
the slow down of Earth’s rotation in the last half Tr
century. TE

A full description of the epoch in use for different offs

et
uency
time scales would be quite complex. For our pur‐ Fre
q
poses it is enough to state that PTP uses the TAI Phase offset
time scale and the epoch is 00:00:00 01/01/1970. Tr
This selection is done so that the POSIX algorithm 0 Tr 2Tr 3Tr 4Tr 5Tr
applied to the PTP 0 seconds time stamp gives the Time
date and time mentioned before. PTP also allows TIE
the use of arbitrary time scales to account for net‐
work administrators willing to use a different ep‐ offs
et
ency
och. The GPS system starts counting time from Fre
qu

00:00:00 06/01/1980 (6th of January) but as no leap

seconds are applied to the GPS time, this time scale Tr Phase offset
is currently 18 seconds ahead of UTC. 0 Tr 2Tr 3Tr 4Tr 5Tr
Time
Specification of a PRTC is the subject of ITU‐T
G.8272. Unlike PRCs, PRTCs are expected to be dis‐ Figure 8 Differences between TE and TIE. The TE is
the difference between two times and the TIE is the
ciplined by at least one time reference. For this rea‐
difference between two time intervals
son, PRTC specifications are given not only for free
running status but also for locked and holdover.
When locked to a GNSS or other reference, the The TE is to be compared with the Time Interval Er‐
PRTC specifications are similar to the PRC but the ror (TIE). While the TE is the difference between
requirement about fractional frequency offset is re‐ two absolute times, the TIE is the difference be‐
placed by a new requirement for Time Error (TE). tween two time intervals and it is therefore a rela‐
Specifically, the total TE is expected to remain un‐ tive metric independent of the epoch. The TIE is a
der 100 ns for the entire test duration. very useful performance metric for frequency de‐
ployments (the MTIE and TDEV are derived from
The TE is the performance metric typical of time the TIE) but in time and phase applications the TIE
sources. It is defined in ITU‐T G.810 as the differ‐ has to be replaced by the TE.

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 13 / 27

There is also a relationship between TE / TIE at one Oscillator Disciplining

A L B E D O - WHITE PAPER
hand and fractional frequency offset at the other. A
positive (negative) frequency offset makes the A synchronization test set is expected to measure
clock to run faster (slower) by a factor that matches the performance level of accurate timing sources
the frequency offset per unit time. For this reason, such as Cesium PRCs that are often used to supply
the TE and TIE increase (decrease) depending on synchronization to large networks. The question is,
the sign of the fractional frequency offset. how can a synchronization tester measure the ac‐
curacy of a clock that is potentially much better
The |TE| < 100 ns requirement is related with a fur‐ than its own local oscillator? This is done through
ther MTIE < 100 ns limit. Actually, the ITU‐T G.8272 an external reference, or still better, by a combina‐
tion of an external reference and the local clock
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

MTIE mask is the intersection of the MTIE < 100 ns

region and the ITU‐T G.811 PRC mask. The TDEV re‐ achieved through a process known as oscillator dis‐
quirement is exactly the same for a PRC and a PRTC. ciplining. Even with oscillator disciplining it is not
From this point of view, it can be said that the PRTC uncommon that the test signal is of the same nom‐
requirements are stronger than the PRC operation inal accuracy level as the local (disciplined) oscilla‐
limits but it must not be forgotten that PRTCs are tor. A typical example is the measurement of a PRC
specified when they are operating in locked status using another PRC clock reference. In this case, a
and PRCs are specified in free running status. The pass result is certainly reliable but the same cannot
requirements for both are thus not directly compa‐ be said about a fail because there is no way to sep‐
rable. arate the phase / frequency degradations in the
test signal and in the reference. As a result, we
A recent new category in the hierarchy of clock ref‐ must be prepared for uncertain results in telecom
erences is the enhanced PRTC (ePRTC) defined in synchronization tests.
ITU‐T G.8272.1 with the purpose of providing the
timing performance required by 5G cellular net‐
works. The TE requirement for an ePRTC is 30 ns TDEV (ns)
(|TE| < 30 ns) in locked mode and 100 ns in a 100
14‐day holdover period. The MTIE and TDEV limits
are also tighter than for PRTCs.
Result #1
ePRTCs are equipped with at least one GNSS input 10
(PASS)
in the same way as most PRTCs but they also have
ore or more frequency inputs that are expected to Result #2
(uncertain)
be feed with references from co‐located PRCs. The
trick to the improved performance provided by the 1
ePRTC is that GNSS still supplies time and date but
1 10 100 1000 10000 100000
stability is left to the frequency references. In other Observation interval  (s)
words, ePRTCs merge the timing information from
the frequency reference and time references to get
the best of each. Figure 9 Two real test results run at the output of the
same PRTC. Compliance with the ITU-T G.8272 TDEV
mask is verified. One test is pass the other is uncertain.
An additional advantage of ePRTCs is that they are
much less exposed to GNSS jamming and spoofing
than PRTCs. On the other hand, the required co‐lo‐ In test applications, it is quite common to use GNSS
cated PRCs are bulky and expensive. For this reason to discipline a Rubidium or an OCXO local oscillator.
ePRTCs are not suitable for mass deployments It is a notable fact that none of the
where OCXO and Rubidium PRTCs are still the pre‐ Rubidium / OCXO or GNSS references alone are
ferred options. compliant with ITU‐T G.8272 but a carefully de‐
signed Rubidium / OCXO oscillator locked to a
GNSS reference may be perfectly suitable to oper‐
ate as a PRTC.

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 14 / 27

It is a common misunderstanding to think that a injected into the receiver.

A L B E D O - WHITE PAPER
GNSS input alone may behave as a primary refer‐
ence source. GNSS modules provide time referenc‐
es in the form of 1PPS / ToD outputs but these are
TDEV (ns)
not explicitly visible. We can therefore think about
GNSS references as special kind of 1PPS / ToD inter‐ 100

face and we can measure the performance of the

GNSS module in terms of TE, MTIE and TDEV in the
GNSS 1
same way we do for other 1PPS interfaces. If we
could measure the output from the GNSS module, 10
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

we would see many short and medium term varia‐

tions and strongly jittered pulses. The amplitude of
GNSS 2
GNSS derived phase fluctuations could easily reach
tens of nanoseconds but on the other hand the 1
output exhibits very good long term stability based 1 10 100 1000 10000 100000
on the accuracy from GPS, GLONASS, BeiDou, Gali‐ Observation interval  (s)
leo or other satellite constellations.
Figure 10 The same TDEV measurement carried over
with two different GNSS references and the same local
With Rubidium / OCXO references it happens to be oscillator (an OCXO).
the opposite: they are stable in the short term but
they drift when they are free running. By disciplin‐
ing the Rubidium / OCXO with the GNSS, the local
oscillator inherits the long term frequency accuracy The receiver itself is also important. A general pur‐
from the satellite system while keeping good per‐ pose GNSS receiver can be used but it is probably a
formance in shorter observation windows. More‐ better choice to use a receiver specifically designed
over, with the disciplining process, the local for timing applications.
oscillator gets a time scale and thus becoming a
time source. If the design is good enough, the TE of One of the most important differences between
the disciplined clock can be made smaller than Rubidium and OCXO oscillators is the way they be‐
100 ns and the MTIE and TDEV could be con‐ have when they are locked to a GNSS reference.
strained within the ITU‐T G.8272 pass region. The The longer the time constant corresponding to the
disciplined oscillator then effectively becomes a filter applied to the GNSS reference the better the
PRTC. ability to remove undesired drift from the refer‐
ence. To be efficient, the local oscillator has to be
Using the GNSS receiver in the best possible way is stable during a period equivalent to the filter time
important for wander result accuracy. The first constant. Rubidium oscillators remain stable for
point to be considered is the antenna. Fixed anten‐ much longer times than OCXOs and therefore their
na installations tend to offer better performance ability to filter the GNSS reference is larger.
than small portable GNSS antennas. There are at
least three operation conditions to be considered: The second advantage of Rubidium oscillators and
other atomic time / frequency references is the
1. Proper operation requires the antenna to see ability to operate in holdover mode for long periods
the largest possible portion of the sky to of time. As a general figure, the error in TE estima‐
receive a signal from as many satellites as possi‐ tions due to the drift of a Rubidium reference in
ble. holdover could be smaller than 100 ns in a two
hour test and smaller than 1000 ns in a 24 hour
2. Multi‐path signal propagation caused by close
test. OCXOs have a very limited holdover capability.
buildings or other elements may damage the
Typical phase errors after a holdover period of two
GNSS signal quality and it should be avoided.
hours is 1 s or more. For this reason, OCXO are al‐
3. Signal strength is also important. If necessary,
the GNSS signal has to be amplified before it is

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 15 / 27

most of no use for long MTIE and TDEV tests with important point about 1PPS / ToD references is that

A L B E D O - WHITE PAPER
no external reference. they are slow signals. They can be used to adjust
the local oscillator only once per second, which is
MTIE (ns) the 1PPS frequency. The local oscillator must re‐
OCXO main stable during the time period between two
1000
consecutive adjustments (1 second). This is feasible
to Rubidium oscillators and OCXOs but not TCXOs.
Frequency references can be used with all kinds of
100
local oscillators, including TCXOs but they cannot
25
be used for time and phase applications, unfortu‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

nately.
10 Rubidium

3. BASIC TESTING SCENARIOS

1
10-1 1 10 102 103
273
Synchronization tests may be classified as emula‐
Observation interval  (s)
tion and monitoring tests. In an emulation test, the
test set behaves as specific network element (or a
Figure 11 Rubidium and OCXO holdover performances group of elements) and sometimes it replaces this
in a 2 hour MTIE test. Both tests have been run at the entity. Usually the test set is not required to repli‐
output of a PRTC. Only the Rubidium gives a pass
cate all the functionality of the emulated equip‐
result. The OCXO measures an MTIE larger than the
mask for observation windows longer than 2 - 3 minutes. ment, but on the other hand the tester is able to
carry out some diagnostics that are beyond the em‐
ulated equipment capabilities. As an example, a
test unit may be unable to manage hundreds of si‐
Most of the discussion about oscillator disciplining multaneous unicast PTP sessions but it may carry
has been focused on Rubidium / OCXO disciplining out advanced TE, MTIE and TDEV tests over a re‐
with GNSS but this is by no means the only possibil‐ duced set of these sessions. The purpose of a mon‐
ity. Virtually any reference could be used to disci‐ itoring test is to get information about the tested
pline an oscillator. Using a 1PPS / ToD reference is entities without disturbing them. A network moni‐
also a popular alternative and disciplining to a fre‐ tor should not generate any traffic and it should not
quency reference (either periodic such as 1544 kHz, disturb existing traffic. It must rely on the traffic
2048 kHz or 10 MHz or non‐periodic such as 1544 captured from the devices under test through one
kb/s 2048 kb/s) is possible as well. or various interfaces.

Table 9 Many PTP tests could be run both in endpoint or in

Performance level of different combinations between monitoring modes, but generally, gathering the
clock references and local oscillators data required to compute all the important perfor‐
mance metrics is more difficult in passive monitor‐
TCXO OCXO Rubidium ing mode because the tester has to intercept
PRC (frequency ref.) High High High various PTP message flows including the mas‐
PRTC (1PPS / ToD ref) Low High High ter‐to‐slave and slave‐to‐master transmission di‐
GNSS Low Medium High rections. Monitoring also assumes that there is a
network already in operation, which may not be al‐
ways true. On the other hand, active endpoint em‐
Disciplining to 1PPS / ToD references has many sim‐ ulation may disturb the network but due to the
ilarities with GNSS disciplining. The main difference small amount of traffic involved this is not signifi‐
is that 1PPS / ToD may be the output of a high per‐ cant most of the time and access to the testing data
formance network clock such as a PRTC. These sig‐ is more straightforward in this case. An advantage
nals are normally “cleaner” that GNSS references of active synchronization testing is the possibility to
and therefore they do not require the same level of generate background traffic to see how the net‐
sophisticated filtering applied to GNSS inputs. An

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 16 / 27

work reacts under specific load conditions. For all delay request messages (in some profiles, the

A L B E D O - WHITE PAPER
these reasons, this paper focuses mainly on active path delay mechanism may be disabled and
testing. therefore not even delay request messages are
received), which do not carry enough informa‐
Tester
PTP Master
tion to do any detailed performance analysis.
The basic application of the master emulation
GNSS mode is to verify that remote slaves are capable
of communicating smoothly with the master.
xGenius

IEEE 1588
This mode could be used to see how the slaves
respond to some uncommon operation condi‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

tions: processing of TAI and arbitrary time

Ethernet / IP Ethernet / IP
scales, interworking with 1‐step or 2‐step
clocks, behaviour under different time and fre‐
IEEE 1588
quency traceability conditions, conformance
with different message rates and the ability to
process leap second events. The ability to gen‐
PTP Slave
xGenius

erate simultaneous PTP and background traffic

Tester requires special mention. This is important to
(a) Master emulation (b) Slave emulation
check the tolerance to high traffic conditions in
the slave and other PTP‐aware or
PTP Master PTP
slave non‐PTP‐aware network elements. The master
emulation mode could also be applied to Syn‐
chronous Ethernet and other physical layer syn‐
chronization technologies, but in this case
background traffic generation becomes irrele‐
Ethernet / IP vant. The ability to generate wander signals to
Ethernet / IP
verify how the phase impairments are accumu‐
lated is still an important feature though.
IEEE 1588
PTP 2. Slave emulation: In this case, the slave is
slave
replaced by the test unit. The tester processes
GNSS
1PPS / frequency the information received from the master and
xGenius
it tries to track the timing signal in the same
Tester GNSS way as any other PTP slave clock. This
operation mode can be used to get message
(c) Pseudo- slave emulation
xGenius

statistics, verify basic conformance and to get

Tester
(d) Clock monitor test
some PDV metrics such as the packet delay
variance, standard deviation and range. The
slave emulation mode is not suitable for more
Figure 12 Basic PTP testing scenarios in endpoint sophisticated performance tests involving
mode: (a) Master emulation mode, (b) Slave emulation MTIE, TDEV and TE. It is difficult to measure
mode, (c) Pseudo-slave emulation mode, (d) Clock
monitor mode
MTIE and TDEV in this mode because the test
unit is disciplined by a device that is at the
same time the device under test. This means
Depending on the equipment connected to the that in the long term all recorded phase
network and the entity being emulated there are fluctuations are zero and gives unrealistic MTIE
four basic test setups: and TDEV results. In other words, test unit
configured in slave emulation behaves as a
1. Master emulation: The test unit replaces a PTP low‐pass filter; it absorbs slow phase
master. The main purpose of this operation fluctuations and it filters out fast impairments.
mode is not to do any measurement but to For this reason, higher frequency phase
stress the network, including the slave. Actu‐ fluctuations are the only ones that can be
ally, in this mode, the test unit only receives

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 17 / 27

detected in this mode. The explanation for the over the packet interface. It is a good idea to

A L B E D O - WHITE PAPER
limitation to measure TE is different and more compare results from a 1PPS / frequency
involved. IEEE 1588 slaves assume that they are output in the slave and the packet test to
operating over symmetrical transmission qualify the slave. Some of the main
media. Specifically, IEEE 1588 slaves run an disadvantages of clock monitoring are the lack
iterative algorithm to minimize the path delay of active traffic generation in this mode and the
asymmetry. In a steady, noiseless channel the difficulty to access some key statistics about
asymmetry will always be close to zero in slave latency, path delay asymmetry and delay
emulation mode. A different kind of test dispersion.
available in slave emulation mode is the
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

background traffic generation test. The Table 10

purpose of this test is the same that in master Comparison between packet and clock monitoring tests
emulation mode but in this case the
background traffic flows in the slave‐to‐master 1PPS IEEE 1588
direction. Message exchange statistics No Yes
3. Pseudo‐slave emulation: This mode is similar to Latency and asymmetry No Yes
the slave emulation mode but now the test unit Ethernet / IP traffic statistics No Yes
keeps an independent synchronization source. Master / slave emulation No Yes
Typically, a GNSS reference is used but the test Background traffic generation No Yes
equipment could use any other reference such TE Yes Yes
as 1PPS / ToD, frequency inputs or even an MTIE, TDEV Yes Yes
internal oscillator in holdover / free running Floor delay population No Yes
states. From the outside, the pseudo‐slave and
slave emulation modes are indistinguishable
Strictly speaking, virtually all commissioning tests
but internally they are different. Now the
required in both frequency and time distribution
reference and test signals can be compared
deployments could be done in clock monitor mode.
and the measurement bandwidth could be
Clock interfaces are good enough if all that is re‐
extended to very low frequencies involving
quired is to qualify the network to support specific
phase variations of hours or days typical of
PTP profile, but the information they supply is quite
MTIE and TDEV tests. If a time reference is used
limited for troubleshooting applications. While the
(1PPS / ToD, GNSS) the TE could be computed
description of commissioning tests is addressed in
as well. Finally, the pseudo‐slave operation
several standards such as ITU‐T G.8261.1 and
mode is also compatible with background
G.8271.1, troubleshooting is largely forgotten by
traffic generation. This feature could be used to
the main standardization bodies, but advanced
check any change in the TE, MTIE and TDEV
testing carried out at different points in the distri‐
depending on the traffic load.
bution and access networks is essential for these
4. Clock monitor: It is good for a test set to kinds of applications.
support at least some kind of passive test
mode. This mode could be the monitoring of
clock interfaces. Monitored signals should 4. VERIFICATION OF FREQUENCY
include both frequency (2048 kb/s, 2048 kHz,
1544 kb/s, 1544 kHz, 10 MHz) and time
DISTRIBUTION DEPLOYMENTS
(1PPS / ToD). The performance metrics in these
interfaces are similar to that in Ethernet / IP It has already been stated that ITU‐T G.8261 and
ports. Traditional MTIE and TDEV are used G.8261.1 extend the applicability of ITU‐T G.823
rather than their versions for packet interfaces and G.824 to packet switched networks. With this
and TE could be reused almost with no purpose in mind, these standards define two new
modification. Clock monitoring tests run over kinds of network clocks, the EEC and the PEC. PECs
the clock recovered by some network may refer to NTP or PTP clock equipment. Both the
equipment, typically a PTP slave. This is EEC and the PEC are expected to interwork with
conceptually different to a PTP test run directly other synchronization network entities such as

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 18 / 27

PRCs, SSUs or SDH Equipment Clocks (SECs) (follow‐ Table 11

A L B E D O - WHITE PAPER
ing the ANSI terminology, Stratum 1, Stratum 2 and Performance limits in frequency distribution applications
Stratum 3 entities). More specifically, in a frequen‐
Interface Limit (ITU‐T)
cy distribution deployment we can find three dif‐
ferent kinds of technology: TDM network G.823, G.824
TDM equipment G.811, G.812, G.813
• TDM synchronization equipment: Includes all Synchronous Ethernet network G.8261
equipment related with SDH / SONET synchro‐ Synchronous Ethernet equipment G.8262
nization. This equipment commonly has syn‐ PTP network G.8261, G.8261.1
chronization inputs and outputs based on the
PTP equipment G.8263
2048 kb/s and 1544 kb/s interfaces.
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

• Synchronous Ethernet equipment: Synchronous studied for PRCs, PRTCs and ePRTCs are examples
Ethernet could be understood as a mixture be‐ of device operation limits but no specific limit for
tween TDM and packet synchronization. It is a networks has been described so far in this docu‐
technology that works in the same way as TDM ment. The next paragraphs deal with this subject.
synchronization but it operates in an Ethernet The main reference for Synchronous Ethernet net‐
interface. It is capable of supplying potentially work operation limits is ITU‐T G.8261. Limits for
the same performance level as TDM synchroni‐ packet‐based networks are described in ITU‐T
zation. One Synchronous Ethernet drawback is G.8261.1.
that it requires on‐path support. TDM synchro‐
nization and Synchronous Ethernet are the two
existing L1 synchronization technologies. Synchronous Ethernet
• Packet‐based Synchronization equipment: Pack‐
et based synchronization protocols carry the Synchronous Ethernet is an ITU‐T standard that
timing information in departure / arrival times provides mechanisms to transfer frequency over
of certain protocol messages and in time the Ethernet physical layer or L1, which can then be
stamps carried by these or other messages. made traceable to an external source such as a net‐
Packet based synchronization is independent of work clock. As such, the Ethernet link may be used
the physical transmission layer. The most im‐ and considered part of the synchronization net‐
portant packet based synchronization protocols work. Currently, Synchronous Ethernet is seen as
are PTP and NTP but this document deals exclu‐ an important building block for accurate frequency
sively with PTP, by far the most accurate of the over packet switched network. A limitation of Syn‐
two. PTP works better with on‐path support but chronous Ethernet is the inability to transfer time.
it may work without it. This is a big advantage if It can be used only for frequency synchronization.
packet based synchronization is to be deployed
in existing networks. A key topic in Synchronous Ethernet is the defini‐
tion of the mechanisms necessary to achieve inter‐
Verification of frequency distribution deployments working between SDH / SONET and Ethernet
in packet switched networks is pretty much the equipment. These mechanisms and procedures are
same as in circuit switched networks. Most opera‐ found fundamentally in three different recommen‐
tion limits and masks are re‐used and some others dations: ITU‐T G.8261, G.8262 and G.8264:
are only slightly modified. For example, the frac‐ • Extension of the synchronization network to in‐
tional frequency accuracy for a PRC is 10‐11 and a clude Ethernet as a building block (ITU‐T
free running SSU (ITU‐T G.812 Type II clock) has fre‐ G.8261) enables Synchronous Ethernet net‐
quency accuracy of 16 ppb or better. These are al‐ work equipment to be connected to the same
most psychological operational limits to rate the synchronization network as SDH / SONET. Syn‐
operational performance of network clocks. These chronization for SDH / SONET can be transport‐
limits are still relevant in packet switched networks. ed over Ethernet and the opposite is also true.

Standards define performance limits both for iso‐ • ITU‐T G.8262 defines the EEC to be compatible
lated devices and for networks. The limits we have with other SDH clocks. EECs are based on ITU‐T

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 19 / 27

G.813 clocks and they are defined in terms of al card is still able to use the clock from an external

A L B E D O - WHITE PAPER
accuracy, noise transfer, holdover performance, subsystem (for example the CPU) for data transmis‐
noise tolerance, and noise generation. While sion but data reception is not coupled to the trans‐
the IEEE 802.3 standard specifies Ethernet mitter clock and it is also uncoupled to other
clocks to be within ±100 ppm ITU‐T G.8262 transmitters in the network. This last feature is the
specifies EEC accuracy to be within ±4.6 ppm. one that defines IEEE 802.3 Ethernet as an asyn‐
Additionally, PRC traceability of the interface is chronous technology.
achievable by disciplining the EEC.
Synchronous Ethernet’s ability to accept or give
• ITU‐T G.8264 extends the usability of the ITU‐T timing signals makes this technology suitable for hi‐
G.781 SSM by Synchronous Ethernet equip‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

erarchical synchronization. Here, the key element is

ment. The SSM contains an indication of the
the EEC which enables Ethernet nodes to accept or
quality level of the clock that is driving the syn‐
supply synchronization to other Ethernet or TDM
chronization chain. The Ethernet Synchroniza‐
equipment. Thanks to this property, Synchronous
tion Message Channel (ESMC) is used for
Ethernet becomes a new building block of the syn‐
propagation of the SSM through the Synchro‐
chronization network.
nous Ethernet network.

SSU SSU MTIE and TDEV

Local osc. Local osc. MTIE and TDEV are the most important perfor‐
±4.6 ppm ±4.6 ppm
mance metrics in Synchronous Ethernet and PTP
frequency distribution deployments. If there are
ITU-T G.8262 ITU-T G.8262 fractional frequency offset requirements, these can
EEC EEC
be built into the MTIE mask.
Central timing card Central timing card

MTIE and TDEV network limits for Synchronous

ETH ETY ETY ETH Ethernet are given in ITU‐T G.8261. MTIE and TDEV
are computed in the same way as in any TDM inter‐
SyncE SyncE face but the Synchronous Ethernet test is carried
timing timing
out over a 1000BASE‐T, 1000BASE‐X or any other
Synchronization Backplane
Synchronization Backplane

SyncE card SyncE card Ethernet interface compatible with this technology.
Actually, ITU‐T G.8261 extends the applicability of
ETH ETY ETY ETH ITU‐T G.823 and G.824 to Synchronous Ethernet.
Performance of Synchronous Ethernet deploy‐
Local osc.
±100 ppm
Local osc.
±100 ppm
ments do not depend on the load and therefore the
measurement could be run without worrying about
Conventional card Conventional card traffic conditions.

For packet synchronization the situation is quite

similar. Different limits apply if the packet network
Figure 13 Synchronous Ethernet Architecture and
comparison with conventional Ethernet
is to totally or partially replace a TDM segment or if
packet synchronization is to be used to deliver tim‐
ing to specific application. In the former situation,
packet synchronization is expected to provide the
The basic difference between a conventional Ether‐ same performance as Synchronous Ethernet and
net and a Synchronous Ethernet network interface therefore the same operational limits in terms of
card is that the Synchronous Ethernet card is pre‐ MTIE and TDEV apply. If packet synchronization is
pared to accept external timing or to supply timing aimed to deliver timing to specific applications,
to other subsystems. On the other hand, the con‐
ventional card is relegated to operate with its own
±100 ppm internal clock. Note that the convention‐

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 20 / 27

then the limits are given by the application require‐ Synchronous Ethernet the limits from ITU‐T

A L B E D O - WHITE PAPER
ments themselves. G.8261 apply. If the network is TDM, then the
(a) Deployment case 1 (a) Deployment case 2 limits from ITU‐T G.823 / G.824 apply. If there is
no synchronization distribution network and
PRC the PEC‐M is directly connected to a PRC then
PRC
ITU‐T G.811 applies.
(a) ITU-T G.8261 EEC output wander Option 1
MTIE (ns)
L1 synchronization
L1 synchronization network 10000
network
5330
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

A
A
2000
PEC-M
PEC-M
1000
B
B

Packet switched
Packet switched network 250
network
100

0.01 0.1 1 10 100 1000 10000 100000

C2
2.5 20 2000
C1
Observation interval  (s)
PEC-S

E (b) ITU-T G.8261.1 Modified G.823 output wander

D
MTIE (ns)

100000
E

16 ppb
Figure 14 (ITU-T G.8261.1 reference model for
frequency distribution applications. It includes both L1
synchronization and packet-based synchronization.
10000
9000

When PTP (or NTP) is used to supply frequency syn‐

chronization to a remote application the limits from 2300
ITU‐T G.8261.1 apply. This standard defines refer‐
ence models for frequency delivery deployments, 1000
0.01 0.1 1 10 100 1000 10000 100000
reference test interfaces, performance metrics and 0.05 0.2 32 64 1125
operation limits based on these metrics. Some of Observation interval  (s)
the reference test interfaces are packet based and
some others may be based on a number of differ‐ Figure 15 (a) ITU-T G.8261 EEC output wander option
ent technologies (TDM, Synchronous Ethernet, 1. Applies to the output of an EEC-1 but also may apply
etc). MTIE and TDEV are expected to be measured to a PEC-M and PEC-S. (b) ITU-T G.8261.1 modified
in non‐packet interfaces. ITU-T G.823 mask that applies to PEC-S output.

This is a summary of the ITU‐T G.8261.1 operation‐ • Network limits applicable at the output of the
al limits in terms of MTIE and TDEV: PEC‐S (Reference point D): Again ITU‐T G.8261
• Network limits applicable at the input of the applies when the PEC‐S output is Synchronous
PEC‐M (Reference point A): If the PEC‐M is syn‐ Ethernet. If the output is 2048 kb/s, the perfor‐
chronized through a network, then the opera‐ mance in terms of MTIE is provided by a modi‐
tion limits from that network apply at the fied mask that results from the intersection of
PEC‐M input. For example, if the network is the ITU‐T G.823 2048 kb/s mask for traffic inter‐

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 21 / 27

faces and the 16 ppb MTIE straight line. These • Bandwidth filtering: The bandwidth filter is a

A L B E D O - WHITE PAPER
measurements are to be done over the PEC‐S linear averaging filter. This filter removes high
recovered clock. frequency impairments in the signal under
analysis so that only slow variations are taken
Despite being defined for a packet network, the into account.
PEC‐M and PEC‐S limits are given in terms of tradi‐
tional TDM metrics. For example, there is no re‐
quirement at the output of the PEC‐M. In some Floor Delay Population Test
deployments, reference point D may not exist.
Then, if ITU‐T G.8261 is literally followed there is Floor Delay Population is the only real packet met‐
ric required for frequency deployment commis‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

only one wander measurement to be run, the one

at the input of the PEC‐M! sioning. The floor delay population test attempts to
implement a mechanism to measure the number
Network operators willing to improve the control of synchronization messages suitable for slave syn‐
they have on the network performance may be in‐ chronization. With this objective in mind, the test
terested in running at least two more tests. One at defines an acceptable end‐to‐end delay range. The
the output of the PEC‐M (reference point B) and a lowest delay is defined to be the floor delay for the
second one at the input of the PEC‐S (reference path under test. In other words, it is the smallest la‐
point C). The traditional MTIE and TDEV metrics tency recorded for a packet as it is transmitted
could not be reused here but they could be re‐ through the test path. The highest delay allowed is
placed by the pktfilteredMTIE and pktfilteredTDEV, obtained by adding a fixed time to the floor delay.
both defined in ITU‐T G.8260. In some respects, the Samples are rated as conforming if they are found
pktfilteredMTIE and pktfilteredTDEV behave like between the minimum and the maximum allowed
packet interface equivalents of MTIE and TDEV. The delay values. Non conforming packets exhibit an
main difference being that the packet metrics re‐ end‐to‐end delay larger than the maximum. By
quire the input sequence to be preprocessed. definition, there are no packets with end‐to‐end
delay below the floor delay.
Packet preprocessing is necessary to avoid unnec‐
essarily pessimistic results. Raw PTP TE samples

Estimation excess
Settling period Testing period
contain a certain amount of PDV that is easily fil‐
Sync PTD

tered out. Packet preprocessing is defined in ITU‐T

G.8260 and it consists of two filters to be applied 

sequentially to the raw TE:

x(t) x’(t) y(t)

Floor delay
Packet Bandwidth corrected estimation
Stability Floor delay
selection filtering metric initial estimate

Figure 17 Illustration of the floor delay population test.

Samples are classified as conforming or not conforming
depending on the latency the experience from the
Figure 16 ITU-T G.8260 pktfilteredMTIE and grandmaster computed from the floor delay.
pktfilteredTDEV preprocessing.

• Packet selection: It is a non‐linear filter that

The acceptable delay range given in ITU‐T G.8261.1
samples the TE sequence looking for values in
at the input of the PEC‐S (reference point C) is
specific ranges to highlight certain properties in
150 s. It is expected that at least 1 % of the syn‐
the result. For example, packet selection could
chronization messages will fall into this range for
be used to discard packets with a potentially
any 200 s observation window. As the expected de‐
high amount of delay variation. These values
lay variation generated in most currently available
are then eliminated before they can be aver‐
PEC‐M is in the range of nanoseconds, it can be
aged with more accurate samples and degrade
concluded that degradation in frequency deploy‐
result accuracy.

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 22 / 27

ments is expected to happen due to variable delay is 100 ns. For an ePRTC, the maximum TE is 30 ns,

A L B E D O - WHITE PAPER
in network elements. equivalent to the propagation delay of an electric
signal over 5‐ 6 m of coaxial cable. Measurement of
As a performance metric, the floor delay popula‐ these minute times requires a highly accurate
tion has several inconveniences which limit its ap‐ time / phase reference and carefully designed mea‐
plicability: surement engines.
• It depends on one‐way delay computations and The fractional frequency offset and floor delay pop‐
it therefore requires an external clock refer‐
ulation are not relevant in phase / time deploy‐
ence. Actually, the measurement does not re‐
ments. Actually, a frequency offset of 1 ppb
quire a time / phase external clock but at least
generates a phase error equivalent to around 90 s
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

a frequency reference is necessary.

in one day, much more than the accuracy figure re‐
• It requires knowledge of the minimum path de‐ quired in this kind of deployment. Together with
lay or floor delay. The floor delay could be a test the MTIE and TDEV, the most important perfor‐
input parameter or it could be computed in a mance phase / time metric is the TE.
training period before the start of the real test.
If the network is not stationary the floor delay Table 12
may be difficult to compute. If the floor delay Performance limits in time distribution applications
changes during the test (due to re‐routing for
example) results may not be accurate. Interface Limit (ITU‐T)
PTP network G.8271.1, G.8271.2
• In many deployments the Floor Packet Percent‐
PTP equipment G.8272, G.8272.1, G.8273.x
age is just 100%, which means that all messages
have been received within the 150 s range re‐
quired by the standard. This is a clear pass re‐ Network limits for phase and time applications are
sult but it gives no more insight about the way defined in ITU‐T G.8271.1; limits for isolated enti‐
the network is behaving. It is always possible to ties (PRTCs, ePRTCs, T‐GMs, T‐BCs, etc) are spread
run the test with a different delay range but throughout different standards. This document
this requires previous knowledge of the net‐ deals mainly with network limits and the applica‐
work performance. tions considered in ITU‐T G.8271.1. Among the
most important of these applications are 3G and 4G
Due to the floor delay population test limitations it cellular communications systems requiring a phase
is interesting to consider some complementary accuracy of 1.5 s or better.
performance metrics available in frequency distri‐
bution deployments. Among these metrics we can In order to guarantee that the phase offset be‐
highlight the classical dispersion metrics, which in tween any two base stations or enhanced Node‐B’s
this case have to be estimated over the delay prob‐ (eNBs) is to remain within specified limits, the
ability density function. Some of the most import‐ whole network has to be carefully engineered. Each
ant statistical dispersion metrics are the variance, network element (and also the transmission medi‐
standard deviation and range. None of them re‐ um!) is constrained to specific operational limits in
quire of an external reference, pre‐testing training terms of TE. The network operator is expected to
periods or previous knowledge about network per‐ consider also the TE already present in the
formance. PRTC / P‐GM, variable TE due to random phase
noise processes and TE generated in the application
end. In case some critical equipment loses all exter‐
5. VERIFICATION OF TIME nal clock references it will enter in holdover status
DISTRIBUTION DEPLOYMENTS and it will start drifting. This condition should be
planned from the beginning and some margin
should be reserved to accommodate temporary
Time and phase testing is where most of the inter‐
holdover in the timing distribution equipment. All
est resides today, but it is also an area with import‐
ant testing challenges. The TE threshold for a PRTC

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 23 / 27

these considerations about performance make up of accuracy, the pulse rise times have to be con‐

A L B E D O - WHITE PAPER
the so called TE budget for the deployment. trolled (< 5 ns in the 50  1PPS interface) and
Network time reference
hence a wide‐band transmission medium is re‐
quired. Bandwidth requirements limit cable lengths
in this kind of link (< 3 m in the 50  1PPS inter‐
PRTC face).
A 100 ns PRTC 1PPS / 50 , single- ended interface
Amplitude
50 ns
T-GM 50 ns
50 ns
B 100 ns ~ 500 ms < 5 ns
50 ns
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

50 ns
Network
T-BC 50 ns elements 1.2 V ~ 5.5 V
Packet switched network

50 ns (11 x 50 ns)
Network TE (1.1 s)
#1 50 ns
50 ns
50 ns
50 ns

250 ns 1 s ± 5 ns
T-BC Link
asymmetries
#10 -0.3 V ~ 0.3 V
Time

T-TSC Dynamic
200 ns TE Figure 19 1PPS pulse shape specification in 50 ,
C single ended interfaces. This interface is described in
ITU-T G.703 and G.8271.
Application TE (0.4 s)

End Holdover,
application 250 ns network
rearrangements For some time, many 1PPS implementations have
D
coexisted but now two standard interfaces have
150 ns
End application been defined in ITU‐T G.703 and G.8271. One of
TE
them is an unbalanced interface designed to oper‐
ate over a 50  coaxial cable and the second is a
Figure 18 Time Error budget in a phase / time delivery balanced interface based on the ITU‐T V.11 / RS‐422
application. TE control requires careful planning in these data communications standard and designed to op‐
applications. erate over 100  wire with RJ‐45 connectors. 1PPS
interfaces operating over RS‐232 are still quite
though.
The 1PPS Interface
The main difference between the unbalanced and
In phase / time application commissioning, the test the ITU‐T V.11 / RS‐422 interfaces is that the latter
interface could either be the packet interface or can accommodate a data communications channel
1PPS interface. Using 1PPS for testing is popular be‐ to distribute Time‐of‐Day (ToD) messages. The ToD
cause it allows for more simple testing tools but on message adds a time scale to the 1PPS signal, which
the other hand these interfaces may not be always considered alone carries information about phase
available. In some other situations it may be useful but not absolute time.
to compare performance results in 1PPS and packet
interfaces to rate certain network elements. Unlike for 1PPS, there is not a unique or at least
there is not a single clear candidate for ToD mes‐
At first glance, 1PPS looks like a quite simple inter‐ sage formatting. The National Marine Electronic
face. The 1PPS source generates a pulse once per Association (NMEA) has developed specification
second. This pulse is transmitted at accurate times that defines the messaging interface between ma‐
and it can then be used to signal transitions from rine electronic devices including compasses, RA‐
second to second. In order to achieve a high degree DAR equipment, computers and many others.

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 24 / 27

One common version of NMEA is version 0183 that Table 14

A L B E D O - WHITE PAPER
uses a simple ASCII character formatting to distrib‐ NMEA-0184 ZDA message structure
ute data from a single transmitter to one or several
Field(s) Meaning
destinations. All NMEA‐0183 talker messages have
a similar structure. They all start with the “$” char‐ ZA Talker identifier, timekeeper, atomic clock
acter followed by a variable number of fields: ZDA Date and time information
152713 UTC time, 15:27:13
• Two characters to identify the talker equip‐ 01 Day, 1st
ment, the entity that generates the message. 07 Month, July
For example HC is used if the message is gener‐
2016 Year, 2016
ated by a magnetic compass; GP is for a GPS re‐
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

00 Local zone offset from UTC (0 hours)

ceiver.
00 Local zone offset from UTC (0 minutes)
• Three characters identify the message type. For *3F Checksum
example CGA messages contain GPS fix data,
GLL is used for latitude / longitude geographic
This message format is perfectly suitable to be used
position, MTW is for water temperature mes‐
by timekeeping equipment to share time scale in‐
sages, ZDA contains time and date information
formation in a 1PPS / ToD interface.
with local time zone information, etc.
• It has a variable number of numeric or alphanu‐ Finally, the NMEA specification could also used to
meric fields separated with commas. The maxi‐ generate queries to certain device types. More‐
mum length of a NMEA‐0184 message is 80 over, the structure is extensible with proprietary
characters plus the start of message and the messages. NMEA‐0184 queries and proprietary
end of line sequences. messages have their own specific syntax.
• A checksum code that uses the “*” character as
a separator. Path asymmetry and TE
• The NMEA message finishes with an end of line
character sequence. The basic performance parameter for phase and
time deployments is TE. The TE tells how much
An example of a NMEA‐0184 talker message gener‐ time is ahead or behind a network clock compared
ated by a GPS receiver to communicate geographi‐ with a reference clock. TE is generated in
cal position is: PTP‐aware and non‐PTP‐aware network entities.
Moreover, the transmission medium could also
$GPGLL,4130.00,N,210.52,W,162012,A*1D contribute to the TE. There are two mechanisms
that could potentially generate TE:
Table 13 • Due to limited PRTC performance, the time dis‐
NMEA-0184 GLL message structure
tributed through the network may not be accu‐
Field(s) Meaning rate. If the PRTC is in holdover status an
additional phase offset is expected to happen.
GP Talker identifier, GPS receiver
This offset will be propagated to all the equip‐
GLL Geographic position, latitude and longitude
ments locked to the PRTC.
4130.00,N Latitude 41 deg. 30.00 min. North
210.52,W Longitude 2 deg. 10.52 min, East • Due to path delay asymmetry the mas‐
162012 Fix taken at 16:20:12 UTC ter‐to‐slave and the slave‐to‐master propaga‐
A Data valid (A) or invalid (V) tion delays may not be the same. It is not
*10 Checksum difficult to see that the TE generated by path
asymmetry is one half of the value of the asym‐
metry. For example if the master‐to‐slave laten‐
The ZDA message could be used to distribute infor‐ c y i s 1  s d i ffe re nt co m p a re d w i t h t h e
mation about time as in the following example:
$ZAZDA,152713,01,07,2016,00,00*3F

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 25 / 27

slave‐to‐master latency, then the induced TE in Given a TE result, there is no way to know if it is

A L B E D O - WHITE PAPER
the PTP slave will be 500 ns. caused by path asymmetry or PRTC limited accura‐
cy. Not even looking at the master‐to‐slave and
MTIE (ns) (a) ITU-T G.8271.1 reference point C MTIE limit
slave‐to‐master delay results help determining the
1000 TE origin as these metrics are computed based on
both the PRTC time and the local time reference.
580

280 The TE accumulates through long transmission

200
paths. The way the TE is accumulated and the po‐
100 tential degradation it could cause depends on how
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

it is generated. The total TE could be classified as

slow TE and fast TE.
• Slow TE contains the TE components that are
10 immune to filtering. Such TE components are
-1
10 1 10 10 2 3
10 10 4
10 5
10 6
107
108 the result of, for example, asymmetry in the
1.3 2.4
Observation interval  (s)
transmission medium between network ele‐
ments or asymmetries within network ele‐
(b) ITU-T G.8272 PRTC wander in locked mode ments. It is often assumed that the slow TE
MTIE (ns) spans the frequency band between 0 and
1000 0.1 Hz. The continuous component (0 Hz) of the
slow TE, also referred as constant TE (cTE),
PRC
could theoretically be compensated through
static setting in the slave equipment, but slowly
varying TE with periods of hours or days is both
100 difficult to compensate or filter and it therefore
should be avoided as much as possible.
• Fast TE is related to random noise accumulation
25
due to T‐BC time‐stamping, packet‐delay varia‐
10 tion experienced by the timing signal packets or
10-1 1 10 102 103 104 105 106 107 108 due to any other phase noise source. The fast
273 TE power is spread out over the frequency
Observation interval  (s)
spectrum and the phase noise can be reduced,
(c) ITU-T G.8272.1 ePRTC wander in locked mode
to some extent, through low‐pass filtering. The
MTIE (ns) fast TE is referred in ITU‐T standards as the high
100 frequency component of the dynamic TE, dTEH
PRTC
The raw TE or any of its bandwidth filtered versions
30
(slow TE and fast TE) are not appropriate to qualify
ePRTC
15 many non‐PTP aware networks. These kind of net‐
10 works may generate large and quite unpredictable
TE due to variable buffering delay, congestion
avoidance and control mechanisms and other caus‐
4 es that could not be efficiently removed by linear
filtering. However, there are alternative non‐linear
1 packet selection methods capable of recovering
10-1 1 10 102 103 104 105 106 107 108 the original timing information to a good extent.
4x105
Observation interval  (s) The slow TE and fast TE do not take into account the
effect of packet selection methods and therefore
Figure 20 MTIE limits at the PRTC / ePRTC output and they provide unrealistically pessimistic figures. The
at the input of the T-TSC in a phase / time delivery
ITU‐T G.8260 pktSelected2wayTE is defined as the
application.
main metric to be used to rate networks where not

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 26 / 27

all equipments are PTP‐aware such as in PTS and cation and the application is the same for FTS and

A L B E D O - WHITE PAPER
APTS deployments. Basically, the pktSelected2way‐ PTS / APTS. Standard ITU‐T G.8271.2 define a limit
TE is the result of applying packet selection to the of 1.1 s for peak‐to‐peak pktSelected2wayTE
raw latency sequences. (APTS) and the max |pktSelected2wayTE| (PTS) at
forward and reverse path latency forward and reverse path latency the TSC input (reference point C). The 1.1 s figure
xf(t) xr(t) xf(t) xr(t) is the same that in FTS but the performance metric
is different. While FTS requires that all TE samples
Packet
selection
Packet
selection
met the 1.1 s limit, only a subset of these samples
xf’(t) xr’(t)
are required to be compliant in PTS / APTS. It is con‐
sidered that this more relaxed limit is enough to
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

Combination Combination provide an accurate timing signal in slave clocks so‐

x(t)
phisticated enough to include packet selection fil‐
tering in their PTP inputs. The TSC output
Bandwidth (reference point D) is not always accessible for test‐
x(t)
filtering
y(t)
ing but in deployments where testing is possible o
TE limit of 1350 ns (APTS) applies.
Stability Stability
metric metric
These limits provide an answer to the question: Is
the network performance good enough to support
(a) pktSelected2wayTE (b) slow and fast TE
phase and time delivery applications? However if
Figure 21 ITU-T G.8260 metrics derived from the raw the answer is not affirmative, they don’t tell why. To
TE: (a) pktSelected2wayTE, (b) slow and fast TE. The answer this question more tests are necessary and
bandwidth filter is a 0.1 Hz low-pass filter for the slow TE these will need to be carried out at other locations
and a 0.1 Hz high-pass filter for the fast TE different to the reference points listed in the stan‐
dard. The expected results at different locations in
The basic ITU‐T G.8271.1 TE operational limit is the network could be inferred from the TE budget
±1.5 s (reference point E) but if one has to focus planned by the network administrator. Using the TE
on the requirements for the timing distribution budget it can be verified which network elements
network only (reference point C), then the require‐ are generating more TE than expected.
ment is ±1.1 s applied to the slow TE (0 to 0.1 Hz).
while the fast TE limit, including all frequency com‐
ponents above 0.1 Hz, is set to 200 ns (peak to peak MTIE and TDEV
amplitude).
MTIE and TDEV are still important performance
There is also a ±100 ns requirement for the PRTC metrics in phase / time distribution deployments
output, in line with ITU‐T G.8272. If the PRTC is in‐ but the way they are used is slightly different in this
tegrated with the T‐GM, then the PRTC output is case. In the same way that a constant frequency
not available for testing and the ±100 ns limit ap‐ offset requirement could be added to the MTIE
plies to the T‐GM output instead. These measure‐ mask through a straight line with specific slope, a
ments could be done either over the packet phase requirement could be added through an hor‐
interface or in a clock monitoring output at the izontal line with the phase offset requirement be‐
T‐GM. The limit is referred to the whole frequency ing the distance of the line to the horizontal axis.
band, including the continuous component. Unlike This approach is used by standards ITU‐T G.8271.1
PRTC, PTP GMs providing a ±30 ns performance in and G.8272, among others. The MTIE at the PRTC is
line with the ePRTC specifications (ITU‐T G.8272.1) up to 100 ns, the T‐TSC MTIE contains components
don’t exist yet but it is not impossible that these de‐ up to 580 ns.
vices could become available in the future.
The ability of the MTIE and TDEV to qualify slow TE
One question that arises is about he operation lim‐ is explicitly used in some standards such as the
its applying to PTS and APTS architectures. These ITU‐T G.8273.2 which is fundamentally devoted to
limits are not too different to the FTS thresholds be‐ T‐BC and T‐TSC performance requirements. This
cause performance is driven only by the end appli‐ standard defines a separated limit for cTE (continu‐

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER
 Networking & Telecoms - P T P T e s t i n g O v e r v i e w 27 / 27

ous TE frequency component) and for “slowly vary‐ [12] ITU‐T G.8261.1 / Y.1361.1, “Packet delay variation

A L B E D O - WHITE PAPER
ing dynamic TE” which is termed as dTE L and it network limits applicable to packet‐based meth‐
includes all low frequency TE (usually up to 0.1 Hz) ods (Frequency synchronization)”, February 2012.
but without taking into account the continuous [13] ITU‐T G.8262 / Y.1362, “Timing characteristics of
component. The MTIE and TDEV are very well suit‐ synchronous Ethernet equipment slave clock”, Jan‐
ed to rate the dTEL both in constant temperature uary 2015.
(CT) and variable temperature (VT) environments. [14] ITU‐T G.8263 / Y.1363, “Timing characteristics of
packet‐based equipment clocks”, February 2012.
Table 15 [15] ITU‐T G.8264 / Y.1364, “Distribution of timing
T-BC performance limits from ITU-T G.8273.2 information through packet networks”, May 2014.
ALBEDO Telecom - Registered in Barcelona, Book 41613, Page 155, Sheet B-390886 - VAT : ESB6523022

Metric Class A devices Class B devices [16] ITU‐T G.8265 / Y.1365, “Architecture and require‐
ments for packet‐based frequency delivery”,
TE (peak) 100 ns 70 ns
November 2010.
cTE ±50 ns ±20 ns
MTIE (CT) 40 ns ( < 1000 s) 40 ns ( < 1000 s) [17] ITU‐T G.8261.1 / Y.1365.1, “Precision time proto‐
col telecom profile for frequency synchronization”,
MTIE (VT) 40 ns ( < 10000 s) 40 ns ( < 10000 s)
July 2014.
TDEV (CT) 4 ns ( < 1000 s) 4 ns ( < 1000 s)
fast TE 70 ns 70 ns [18] ITU‐T G.8271 / Y.1366, “Time and phase synchro‐
(peak‐to‐peak) nization aspects of packet networks”, February
2012.
[19] ITU‐T G.8271.1 / Y.1366.1, “Network limits for
6. BIBLIOGRAPHY time synchronization in packet networks”, August
2013.
[1] IEEE Std. 1588‐2008, “IEEE Standard for a Precision [20] ITU‐T G.8272 / Y.1367, “Timing characteristics of
Clock Synchronization Protocol for Networked primary reference time clocks”, January 2015.
Measurement and Control Systems”, 24 July 2008.
[21] ITU‐T G.8272.1 / Y.1367.1, “Timing characteristics
[2] ITU‐T G.703, “Physical/electrical characteristics of of enhanced primary reference time clocks”,
hierarchical digital interfaces”, November 2001. November 2016
[3] ITU‐T G.781, “Synchronization layer functions”, [22] ITU‐T G.8273 / Y.1368, “Framework of phase and
September 2008. time clocks”, August 2013.
[4] ITU‐T G.810, “Definitions and terminology for syn‐ [23] ITU‐T G.8273.2 / Y.1368.2, “Timing characteristics
chronization networks”, August 1996. of telecom boundary clocks and telecom time
[5] ITU‐T G.811, “Timing characteristics of primary slave clocks”, April 2014.
reference clocks”, September 1997. [24] ITU‐T G.8273.3 / Y.1368.3, “Timing characteristics
[6] ITU‐T G.812, “Timing requirements of slave clocks of telecom transparent clocks”, November 2017.
suitable for use as node clocks in synchronization [25] ITU‐T G.8275 / Y.1369, “Architecture and require‐
networks”, June 2004. ments for packet‐based time and phase distribu‐
[7] ITU‐T G.813, “Timing characteristics of SDH equip‐ tion”, November 2013.
ment slave clocks (SEC)”, March 2003. [26] ITU‐T G.8275.1 / Y.1369.1, “Precision time proto‐
[8] ITU‐T G.823, “The control of jitter and wander col telecom profile for phase/time synchronization
within digital networks which are based on the with full timing support from the network”, July
2048 kbit/s hierarchy”, March 2000. 2014.
[9] ITU‐T G.824, “The control of jitter and wander [27] ITU‐T G.8275.2 / Y.1369.2, “Precision time proto‐
within digital networks which are based on the col telecom profile for phase/time synchronization
1544 kbit/s hierarchy”, March 2000. with partial timing support from the network”, July
2016.
[10] ITU‐T G.8260, “Definitions and terminology for
synchronization in packet networks”, February [28] ITU‐T O.174, “Jitter and wander measuring equip‐
2012. ment for digital systems which are based on syn‐
chronous Ethernet technology”, November 2009.
[11] ITU‐T G.8261 / Y.1361, “Timing and synchroniza‐
tion aspects in packet networks”, August 2013.

Professional Telecom Solutions

TEST- LABOS - TAPS - WAN EMULATION - E1 - GBE - SYNCE - WLESS - LTE - 3G - IPTV - VoIP - QoS - SLA - ONEWAY - DATACOM - POLQA - PTP - JITTER - WANDER