Time distribution strategies in cellular networks

White Paper
Time Distribution
Strategies in Cellular
Networks

Abstract
This paper reviews the various methodologies currently available for
ensuring Time of Day (ToD) synchronization in cellular networks. It also
introduces RAD’s revolutionary Distributed GMTM
scheme, designed to
deliver superb ToD accuracy at a lower cost in LTE and small cell
networks, by bringing Grandmaster functionality closer to the base
station in a small form factor device.

Time Distribution Strategies in Cellular Networks
© 2013 RAD Data Communications Ltd 1
Contents
1 Introduction ......................................................................................................................2
2 Using GPS for Time Distribution in Cellular Networks.................................................4
3 Backup to GPS using Sync-E ...........................................................................................5
4 Transition to IEEE 1588-2008 (PTP)..............................................................................5
5 Centrally Located PRTCs/PTP-GMs .................................................................................6
6 Access Located Distributed GMs....................................................................................9
7 Joint GPS-PTP ..................................................................................................................11
8 RAD’s Distributed GM Solutions...................................................................................12
9 Summary..........................................................................................................................13
References..............................................................................................................................15

2 © 2013 RAD Data Communications Ltd
1 Introduction
Having overcome the challenge of precise frequency distribution, time distribution (or Time-Of-Day
{TOD} as it is sometimes referred to) is the next hot thing when it comes to synchronization of
cellular base stations-and a worthy challenge it is indeed.
3rd
-generation cellular base stations, such as the UMTS-TDD and the TD-SCMA, require provisioning of
a time reference that deviates from the Universal Time Coordinated (UTC) by no more than 1.5
microseconds [1]. Future LTE cellular networks (regardless of their duplexing method: FDD or TDD)
will have even stricter requirements in order to enable new features, such as Multiple Input Multiple
Output (MIMO) and Location Based Services (LBS). End-to-end time accuracies here are likely to be in
the order of few hundreds of nanoseconds!
Time synchronization is so challenging mainly because, unlike frequency synchronization, it cannot
solely rely on a stand-alone, specific physical phenomenon (such as the hyperfine energy level
transitions of the Cesium element). Time synchronization, although usually based on accurate
frequency distribution, requires some additional things.
Figure 1 below presents the conceptual difference between a frequency and a time lock from the
point-of-view of a simple signal scope. With frequency lock (upper figure), the two clock signals are
completely “standing waves” relative to each other, with some arbitrary fixed time (phase) offset in
between. Since in a frequency lock we are only interested in having the locked signals pace at the
same rate, this arbitrary fixed time offset is of no importance. Nevertheless, with time lock, we want
both signals to be completely time-aligned. That is, their signal rise events should occur in exactly the
same instant, which essentially means that the fixed offset must be zero.

Figure 1: Example of frequency lock (upper) and time (phase) lock (lower)
Thus, time synchronization is mandated by the following requirements:
• A stable enough (good frequency stability) primary counter that counts time units based on a
given standard timescale (e.g., UTC – Universal Time Coordinates), relative to an arbitrary
predetermined epoch (e.g., 1 nanoseconds elapsed from 1st
January 1970, achieved by a 1
GHz-driven frequency counter).
• A method (protocol) that measures delay between the primary counter and the client that
requires the time information (zeroing the fixed time offset).
The first requirement is quite straightforward and easy to implement nowadays, using a Global
Navigation Satellite System, or GNSS (e.g., GPS). A GNSS essentially disseminates the same time (up
to some very small inaccuracies) to every point on earth. Thus, a decent GNSS receiver, backed with a
very precise frequency reference (to maintain the progress of the time between GNSS updates or
allow for holdover in case of a GNSS failure), can be used as such a primary counter at every point on
earth. This apparatus is usually referred to as a Primary Reference Time Clock (PRTC). Contrary to the
SDH/SONET/Sync-E Primary Reference Clock (PRC), whose stand-alone frequency source could have a
residual fractional frequency error of up to ±10-11
compared to the UTC, a PRTC is always disciplined
to a GNSS (under normal operation), and, thus, its frequency output is always perfectly in-line with its
time output. We usually refer to this attribute as time-frequency coherency.

Any time distribution chain must therefore start with such PRTC. However, from this point on the
“game” opens up and different strategies for time distribution exist, based on service-provider
CapEx/OpEx preferences and GNSS geopolitical view.
2 Using GPS for Time Distribution in Cellular Networks
The challenge of time distribution to the base stations can be easily and quite effectively solved by
deploying a “PRTC” on each and every end-application. This essentially means installing a GNSS
receiver plus antenna on every base-station. Thus, assuming a clear sky view is available at each such
site, each base station would directly get its time (and probably also frequency) reference directly
from the GNSS. This strategy is mainly used today in North America, where almost all time (and
frequency) supplied to cellular base-stations is GPS driven.
Indeed, as long as it is operational, GPS is capable of delivering extremely accurate time reference in
the order of ±50 nanoseconds that is more than enough even for the most stringent cellular
technology requirements. However, GPS (and GNSS in general) has its drawbacks.
To begin with, putting a GPS antenna on every cellular base station has problematic consequences in
terms of both CapEx and OpEx. It complicates the initial installation process of the base station
(additional antennas, wiring etc.), mandates having an unobstructed sky-view (a major problem for
the emerging small cell antenna technologies that are mainly targeting building walls and closed
spaces such as shopping malls) and wastes expensive technician time whenever the outdoor antenna
requires maintenance. But this is just the beginning…
GPS is controlled by the U.S Department of Defense. Ever since GPS became fully operational in 1994,
it has become such a prominent tool in our daily civilian lives that we often tend to forget this.
Nevertheless, cellular service providers around the world (other than in North America) do take that
into account and recognize that under certain circumstances, the GPS service could be summarily
terminated. Thus, relying on GPS has strong geopolitical factors attached to it and many countries in
Europe and Asia are reluctant to place their strategic telecommunications assets in foreign hands.
This is mainly the reason why new GNSS systems like the European Galileo project, the Russian
GLONASS and the Chinese Beidou navigation system were initiated. Nevertheless, the only fully
operational, GNSS system with full world coverage existing today – and for the foreseeable future – is
GPS.

Nonetheless, all of the above is just a prelude to the scariest problem of all, GPS jamming. Being a
passive radio technology element, a GPS receiver can be easily jammed using a $5 piece of equipment
that can be easily bought on Ebay. Such an active jammer can disrupt the operation of a base station
and even cause it to crash temporarily when it is operated somewhere nearby. The problem even
worsens in metro areas having a dense concentration of cellular base stations as well as moving
vehicles. Some of these cars could have active GPS jammers, used by the drivers to block the car’s
speed/position log recordings. This, in principle, could cause occasional disruption to nearby base
stations. GPS vulnerabilities have been at the center of a few recent conventions dealing with
frequency and time, as the European ITSF and American WSTS.
3 Backup to GPS using Sync-E
Putting aside the prohibitive cost issue of installing/maintaining a GPS antenna on every cell-site, a
backup to GPS at each cell-site must be applied. Such a backup can be effectively realized by
supplying the base-stations with an accurate frequency source so that they will be able to keep their
time ‘ticking’ at the right rate once GPS is lost. For networks that already employ and distribute it to
the end-applications, Sync-E would be a natural choice1
. However, many cellular networks today (e.g.,
wholesale networks) are not supporting Sync-E. Furthermore, the introduction of small cells and the
massive role these small antenna technologies – expected to be mainly installed in dense
urban/indoor locations – are going to play in 4G is driving the search for an alternative, less GPS-
dependent, solution.
4 Transition to IEEE 1588-2008 (PTP)
The only time synchronization alternative today to GPS is IEEE 1588 (the 2nd
version of the standard
termed IEEE 1588-2008, or PTPv2, to be exact) [2]. With PTP, the time (and, possibly, also frequency)
distribution is carried using dedicated packets that are exchanged between a PTP Grandmaster (PTP-
GM) and a PTP slave device (PTP-slave). The PTP-GM is usually directly connected to a PRTC, receiving
accurate coherent time and frequency references, and uses the on-going packets exchange with the
PTP-slave to convey the time (and frequency) information to it. It is the PTP-slave’s job to recover the
time (and frequency) information back from the received packets.
1
Though one needs to be certain the base-stations are capable of using the Sync-E ref. for the time holdover
work, rather than just for controlling the frequency of the RF transmission.

Although PTP is capable of both distributing frequency and time, a specific service provider might
choose, for various reasons, to take advantage of the existing physical layer’s frequency distribution
infrastructure (e.g., TDM or Synchronous Ethernet) and use the PTP service for time only. Everything
said in this paper is applicable to either case.
Practices of distributing time using PTP in cellular networks can be divided into two main strategies:
1. Small number of PRTCs/PTP-GMs at the cellular backhaul core/aggregation, each servicing a
large number of PTP-slave devices integrated within the base station or colocated with it.
2. Larger number of PRTCs/PTP-GMs at the aggregation/access, each servicing a relatively small
number of PTP-slave devices integrated within the base station or colocated with it.
5 Centrally Located PRTCs/PTP-GMs
The first strategy is more or less based on existing SDH/SONET and Synchronous Ethernet (Sync-E)
frequency distribution principles. That is, a primary reference followed by a relatively long distribution
chain of 10 and more nodes. This strategy is depicted in Figure 2. The advantages of this approach
include lower total cost spent on PRTCs/PTP-GMs2
(fewer of them are needed) as well as an easier
and more efficient fault protection scheme (as each PRTC/PTP-GM is responsible for more PTP-slaves
and has better visibility of the other slaves in the network not under its direct responsibility during
normal operation).
2
A practical implementation is likely to integrate the PRTC and the PTP-GM within a single piece of equipment.

Figure 2: Example of centrally located PRTC/PTP-GM time distribution
The main problem of this approach is the relatively high number of intermediate network elements
(e.g., switches and routers) that will need upgrading to facilitate the PTP messages exchange in order
to bring the end-to-end Packet Delay Variation (PDV) to a minimum. Such PTP on-path support
mechanisms include the Boundary Clock (BC) and Transparent Clocks (TC). Meeting the most stringent
time distribution requirement (and giving Service Level Agreement {SLA} for it) would probably
mandate that all intermediate network elements will offer some kind of on-path PTP support. This
understanding was the main drive to the on-going development efforts for the new ITU-T G.8275.1
1st
PTP Telecom Profile for time distribution with full network support. Current expected completion
date is middle or end of 2013.
Even though many cellular service providers understand that at the end of the day they will probably
need to implement some kind of network forklifting to support PTP, they do not necessarily want to
do it from day one. Some would very much prefer to take a more gradual approach and delay the
required network modification to a date as close as possible to when they can realize a real payback
for those services that require the precise time (e.g., LTE network MIMO or Location Based Services).
In the meantime, they would go for a less expensive working solution, even though true SLA could
not be guaranteed at any given moment.

The nice thing about PTP is that, contrary to other sync distribution techniques such as Sync-E, it will
benefit from having more on-path network support but does not mandate it. Thus, different schemes
of partial on-path support can be used in order to improve the level of performance while keeping
CapEx under tight limits. These could later on be supplemented with more on-path network support
to yield an even better level of performance.
A popular example for the use of partial support is depicted in Figure 3. Here, an intermediate BC is
placed at a strategic point in the time distribution path between the core-located PTP-GM and the
PTP-slave in the base station. The job of the intermediate BC is to divide the PTP distribution chain
into two parts (e.g., core/aggregation and access). The BC will terminate the time information after
the core/aggregation cloud, dealing with PDV introduced on that section only. The regenerated PTP
flow would then traverse the access, terminated by the PTP-slave within the base station that will
need to mitigate PDV introduced by the access only. Such a scheme can allow better PTP end-to-end
performance3
at the additional cost of just one PTP intermediate function (or two if a more secure
fault tolerant scheme is pursued). Nevertheless, as already stated, true SLA guarantee would still be
very difficult to deliver.
As the time distribution following this approach is more ‘end-to-end’ in nature, the principles of the
existing ITU-T G.8265.1 PTP Telecom Profile for frequency only [3] could also be used here. This is the
scope of the work currently unfolding in the ITU-T SG15/Q13’s group of timing experts. The aim is to
start working on a 2nd
Time Telecom Profile for partial support (designated number G.8275.2) as soon
as the work on the 1st
full-support one is finished.
3
Placing an intermediate BC would result in better overall end-to-end performance in many cases, but certainly not all. The
merits of this approach mainly depend on the PDV profile of the core-aggregation cloud. Taking into account this network
section is comprised of high capacity links (10GB), hardware driven network elements and high QoS for the PTP flows, this
approach would probably work well. Moreover, the intuitive assumption that adding more PTP support, by placing more
intermediate BCs, would give even better performance might not always hold true. This is due to the inherent noise
accumulation characteristics of BCs. Of course, when a BC is implemented in every node along the chain (full network
support), PDV will no longer exist and performance would be optimal. TCs, on the other hand, do not have this problem and
the end-to-end performance will be directly proportional to the number of elements that support TC. As in the BC case, full
network support will guarantee optimal performance.

Figure 3: Example of centrally located PRTC/PTP-GM time distribution with intermediate BC
(partial support)
6 Access Located Distributed GMs
An alternative strategy to the centralized PTP-GM deployment would be to locate a relatively large
number of distributed PRTCs/PTP-GMs in the access network, each servicing a smaller number (a few
dozens usually) of PTP-slave devices. The benefits of this approach are obvious. Positioning the
PRTCs/PTP-GMs closer to the PTP-slaves would result in much smaller time distribution chains and
would dramatically cut the number of intermediate network elements that need to be enhanced with
PTP on-path support. Furthermore, no timing distribution capability is demanded for the mobile
network preceding the distributed GMs. This is particularly important for mobile service provides
leasing transport services from wholesalers. On the other hand, more PRTCs/PTP-GMs would be
required. An example for such a PTP deployment strategy is given in Figure 4.

Figure 4: Example of Access located PRTC/PTP-GM time distribution
The dramatically shorter time distribution chains together with the desire to meet, at the end of the
day, the stringent cellular time accuracy requirement will probably drive many service providers
adopting this strategy to incorporate full PTP on-path support from day one. Nevertheless, as the
number of hops is now much lower, the gradual migration path concept for the end-to-end case we
saw in the previous chapter can be even more attractive here, by gradually adding on-path support
between the distributed GM and its PTP-slave devices.
The traction of distributed approach to the cellular market is mainly conditioned on two factors:
1. The new distributed GM would need to have a markedly reduced cost than its older “brother”,
the big central GM.
2. The means for backup are still required to protect against GPS failure.
The latter point can be solved using Sync-E or any other accurate frequency source that can be
supplied to the distributed GM unit. In cases where Sync-E is not applicable, PTP could also be used as

an effective and economical source for backup. Such apparatus is described in details in the next
section.
7 Joint GPS-PTP
The ever growing quest for “cost-effective” and “good enough performance” solutions recently gave
traction to yet another time distribution strategy, which can be referred to as “Joint GPS-PTP”. The
notion is quite straightforward. If we do not have Sync-E deployed in our network we can still have a
plausible backup to revert to in case the GPS fails, by taking advantage of the central GM that might
already be installed in our network. An example of this strategy is depicted in Figure 5.
Figure 5: Example of Joint GPS-PTP

Here, the fallback mechanism for the distributed GMs is achieved using PTP. The distributed GMs
receive and terminate a PTP flow in addition to the time/frequency reference they receives from the
GPS. As soon as the GPS fails, the distributed GM would fall back to work as a PTP-BC relying on the
time reference it receives from the central GM, until normal GPS operation is restored. The transition
is done in a hitless manner to prevent unnecessary transients from occurring. Furthermore, in
contrary to the partial-support case, the very accurate GPS reference could be used to improve the
backup PTP service level of performance under normal GPS operating conditions4
. Thus, on GPS
failure, an even better PTP time reference could be provided. An important implication is that the PTP
time distribution chain could be made far simpler, having a very limited partial on-path support or
even none at all (pure end-to-end).
8 RAD’s Distributed GM Solutions
RAD’s solutions for mobile backhaul – the ETX-5300A Service Aggregation Platform and the new ETX-
205A Mobile Demarcation Device – feature advanced timing synchronization functionalities in addition
to their service demarcation and aggregation attributes. This combination allows backhaul operators
and wholesale providers to reduce the number of network elements, together with their associated
costs, that are require to ensure dependable, per-CoS service delivery. Both products are MEF CE 2.0-
certified and feature a distributed GM with Sync-E holdover capabilities (as well as external frequency
source backup), while the smaller ETX-205A also includes a built-in GPS receiver. As depicted in Figure
6, upon the loss of GPS, the system will automatically switch to “Sync-E holdover” mode if Sync-E is
supported in the network. Otherwise, the system is designed to fall back to BC mode, taking its time
and frequency reference from a predefined centrally located GM.
4
Such improvements can include mitigation of inherit network asymmetries that directly affect the PTP level of performance
and could not be solved for otherwise.

Figure 6: Details of a Joint GPS-PTP distributed GM solution
9 Summary
Delivering accurate time to the cellular base stations will certainly be one of the major challenges
facing the cellular providers as they start to deploy their new LTE networks. Over the coming years,
we will witness a constant struggle between the will to meet the very stringent time accuracy
requirement on one hand, and the need for a cost-effective migration path, on the other. In reality,
accomplishing this challenging task will probably assume a variety of implementations based on
geographical location, CapEx/OpEx considerations and fault-protection perspectives. The different
attributes of the most prominent approaches discussed in this WP are summarized in Table 1.

GPS on every site Centralized GM Distributed GMs
Number of hops
between GM and
slaves
N/R High. Mandates full
BC/TC support in the
mobile backhaul
Small. Only the
last mile
equipment.
Reliance in GPS High. A GPS receiver is
required on every
base-station
Low. One GPS
receiver covers
hundreds of base-
stations
Moderate. One
GPS receiver per
dozens of base-
stations
GPS backup
provisioning
Problematic as many
mobile network do not
support Sync-E to the
base-station
Achievable using
Sync-E or other
accurate frequency
source at the core
Achievable using
Sync-E or PTP
(from the core)
CapEx/OpEx High CapEx/OpEx to
install and maintain
the GPS antennas on
every base-station
(~1000$ per base-
station)
High CapEx due to
the required full
BC/TC support in the
backhaul net.
Low. Smaller
number of GPS
antennas and no
need for BC/TC in
the backhaul net.
Applicability for
small-cells
Problematic due to the
“sky view”
requirement
Applicable (assuming
full PTP support)
Ideal due to its
flexibility to place
the GM at the
optimal location
Table 1: Summary of time distribution strategies in cellular application
RAD’s products comprise all the different synchronization ingredients and offer our customers a full
suite of synchronization solutions to choose from. For more information, please contact
market@rad.com.

References
[1] 3GPP TS 25.402 version 5.2.0 Release 5.
[2] IEEE Std 1588™-2008, IEEE Standard for a Precision Clock Synchronization Protocol for
Networked Measurement and Control Systems
[3] ITU-T Recommendation G.8265.1 (10/2010), Precision time protocol telecom profile for
frequency synchronization.

The RAD name and logo is a registered trademark of RAD Data Communications Ltd.
© 2013 RAD Data Communications Ltd. All rights reserved. Subject to change without notice.
Version 6/2013 Catalog no. 802593
www.rad.com
North America Headquarters
RAD Data Communications Inc.
900 Corporate Drive
Mahwah, NJ 07430 USA
Tel: (201) 529-1100
Toll free: 1-800-444-7234
Fax: (201) 529-5777
E-mail: market@radusa.com
www.radusa.com
International Headquarters
RAD Data Communications Ltd.
24 Raoul Wallenberg St.
Tel Aviv 6971923 Israel
Tel: 972-3-6458181
Fax: 972-3-6498250
E-mail: market@rad.com
http://www.rad.com

Time distribution strategies in cellular networks

More Related Content

What's hot

Similar to Time distribution strategies in cellular networks

More from Nir Cohen

Recently uploaded

Time distribution strategies in cellular networks