WP Sync IP Mobile Networks

8/3/2019 WP Sync IP Mobile Networks

1/13

Synchronizing IP Mobile Networks

WHITE PAPER


2/13

THE BUSINESS CHALLENGE

As operators tested 3G killer applications a mobile internetgeneration emerged, taking or granted high-capacityaccess irrespective o time or location. Text messaging nolonger satisfes the user needs, being replaced by bandwidthintensive social networking, picture messaging and streamingvideo applications. Unlike voice services, mobile broadband isalways-on, and must always be available.

An estimated 500 million subscribers will have access tomobile broadband by the end o 2011, exceeding the number

o wireline subscribers1. Supported by generous data plans,smart handsets and wireless modems, the demand or mobilebroadband continues to grow exponentially and is the newbusiness opportunity or service providers.

But that opportunity is not without challenges.

Fiscally, data demand is growing exponentially, but theadded revenue is incremental, placing pressure on capitaland operating budgets. Since taris cannot be increasedin proportion to demand, the cost per bit delivered must bereduced!

Technically, the data must be transported over the air, andthen delivered via the backhaul (oten reerred to as theRadio Access Network) to the high capacity core network. Airinterace protocols satisy the current data demand, but thebackhaul does not have the transport capacity needed or thedata transer.

With the largest part o the operating cost being attributed tothe Radio Access Network, the linear relationship betweenbandwidth and cost or channelized E1/T1 circuits does notmeet the fnancial goals. This is also true or leased circuits,despite alling prices. Ultimately, mobile operators need totransport more data or less money, and TDM backhaul is nota viable long term solution.

IP/Carrier Ethernet is a practical and thereore inevitablechoice or the backhaul.

When looking at data orecasts, it was also evident thatexisting radio schemes wont meet uture demands, and

engineers were tasked with designing lower cost, exible andmore efcient wireless systems.


A Practical Guide to Synchronization For Long Term Evolution Wireless Systems

2

Mobile broadband has become reality. It has changedall our paradigms.

IP/Carrier Ethernet meets the mobile business needs increased bandwidth and for less money.

Voice Dominant

Traffic

Revenues

Data Dominant

Time

Revenues

& Traffic

De-Coupled

1 Inonetics Research, Inc., Fixed and Mobile Subscribers annual worldwide market orecasts October 2008

FIG 1: Revenue/Trafc Relationship


3/13

3


The result Long Term Evolution (LTE) and IEEE 802.16 WiMAXmobility. Many articles have been published on the relativebenefts o LTE and WiMAX, but concisely stated, they oerimportant benefts or both consumers and operators:

Radio transmission rate increases, with download links up to200 Mbit/second (LTE).

Low round trip latency (10 milliseconds).

Gains in spectral efciency, allowing more simultaneoususers or a given bandwidth. This includes Frequency DivisionDuplex (FDD) and Time Division Duplex (TDD) modes.

Underlying all IP transport networks.

What is important to us is the IP transport network. IP meetsthe backhaul investment drivers increased bandwidth andreduced cost. Combining the IP oundation with protocolsthat meet tomorrows air interace demands makes LTE acompelling technology.

As we drive or more perormance and lower cost, ourinnovation sometimes depends on secondary, enablingtechnologies. This is especially true or IP transport platormsin the mobile architecture. Data can be transported through anIP network without synchronization, but by the same token, IPnetworks cannot transport synchronization naturally (as wasthe case or SONET and SDH).

What does this have to do with the IP backhaul? The answerlies in the origin o the base stations requency reerence. Thehandset (or UE) derives its requency reerence rom the basestations air interace and operates at that recovered requency.I the requency dierence between adjacent base stations istoo big, the handset will not lock to the new BTS as the usermoves into the adjacent cell, and the call will be dropped.

The air interace stability allows the user equipment to hand o

calls between cell towers without interruption, and is central tothe Quality o Service (QoS).

Data rate increases, latency reduction, spectral

efciency and backhaul bandwidth collectively drivethe need for LTE.

Without a stable air-interface frequency reference,wireless mobility cannot be supported.

+/- 50 ppb

+/- 50 ppb

Time

}

}

Mobile cannot lock to BTS2and call is dropped

BTS2 drifts outside 50ppb window

BTS2

BTS1

F1 +F

F1

T1 T2

FIG 2: Mobility Requirements


4/13

4


From the table below, it is apparent that all base stationsmust support an air interace stability o 50ppb2 or better,irrespective o the mobile protocols or technology generation.At frst glance, this may seem a daunting task, but basestations traditionally sourced their requency rom the E1/T1 backhaul (assuming they meet the synchronization masksdefned by the ITU-T G.823 or Telcordia GR.253). To meet the50ppb air interace requirement, the span line stability mustbe approximately 15ppb.

To get to the heart o the matter, the E1/T1 links traditionallyprovided the requency reerence to base stations. Whenreplaced with Ethernet/IP, the link to the requency sourceis lost, and the timing chain is broken. This is the designchallenge or mobile backhaul planners.

The air interace requency requirement is ever-present, butthere are cases where precise Time o Day is also required atevery base station.

The frst is driven by technology. Time Division Duplex (TDD)mode improves the spectral efciency o the allocatedbandwidth. Phase synchronizing base stations eliminateinter-cell intererence by ensuring that adjacent uplink and

downlink transmissions are coordinated. This can only beachieved by synchronizing all the base stations to the sameTime o Day reerence.

The second is driven by application. Applications such as SFN-MBMS transmit video rames rom multiple base stations,and accurately timing them allows the handset to rebuildthe video stream coherently. In addition, Location BasedServices that use Time o Arrival (TOA) to triangulate positionalso need precise time synchronization.

The means to deliver the required time accuracy hastraditionally been to install GPS receivers at every BTS, andthis will be discussed urther as we explore next generationsynchronization distribution.

REBUILDING THE SYNCHRONIZATION CHAINS

While the transport vendors built packet network elements,the timing community worked on methods to deliver Timingover Packet cost-eectively. The obvious goals were to keep itsimple, cost-eective and predictable. Simple suggests usingthe network to deliver time and requency (at the physical

layer or in-band). There were many resultant methods todistribute precise time and requency through the network, buthose selected by standards organizations were SynchronousEthernet (SyncE) and IEEE 1588-2008 (also reerred to as PTPv2). Although not a packet technology, the use o GPS is alsoconsidered.

Global Positioning System (GPS)

Technology advances and the widespread adoption o GPShave resulted in cost reductions, allowing GPS timing receiversto be deployed without the cost penalties o ormer versions.GPS is a high-perormance solution, providing time, requencyand location inormation independently o the network,

and has traditionally been used to deliver microsecondrequirements at base stations.

One obstacle to GPS is service availability in metropolitan andindoor installations resulting rom weak and reected signals.The second concern is that GPS is not autonomous, and manywireless operators preer not to be dependent on widespreadGPS. Finally, the deployment and maintenance costs o GPSare not trivial, particularly in urban applications.

New Global Navigation Satellite Systems (GNSS), suchas Galileo, are being deployed and will satisy the sameapplication objectives as GPS. The political barriers to thesesystems may be lower in dierent geographies, but the urbancanyon, autonomy and cost challenges are the same.

2 Fractional requency oset (/) o 5 x 10-8

IP interrupts the frequency distribution neededby base stations. Synchronization must now beengineered into the network.

Political barriers to GPS are declining, but widespreaddeployment demands a recovery contingency for GPSoutages.

IEEE 1588 and SyncE are the two standards basedmethods for distributing frequency through packetnetworks.

Mobility Standard Frequency Time/Phase

CDMA2000 50 ppb Range: 3s to 10sGSM 50 ppb

WCDMA 50 ppb

TD-SCDMA 50 ppb 3s inter-cell phase

LTE (FDD) 50 ppb

LTE (TDD) 50 ppb +3s inter-cell phase LTE SFN-MBMS 50 ppb +6.4s inter-cell phase

WiMAX (TDD) 50 ppb inter BTS Typically 1-1.5s

Femtocell (FDD) ~250 ppb

+ Standards being consolidated

TABLE 1: Mobility Air Interace Stability Needs


5/13

5


Synchronous Ethernet (SyncE)

Synchronous Ethernet is a scheme that distributessynchronization over the Ethernet physical layer without

compromising the asynchronous data switching unctions.Based on the IEEE 802.3 standard or Ethernet, it issynchronous at layer one with the higher layers beingasynchronous. There is no dierence between a Synchronousand Asynchronous switch in the way the data is handled. Theonly dierence is at the clock distribution and recovery layer.

Asynchronousswitches receive dataat the incoming linerate, and in adherenceto IEEE 802.3, transmitdata using a ree-running clock o100ppm (poor stabilityin synchronizationterms, but suitableor the switchingunction).

SyncE switches, bycontrast, use a moreaccurate 4.6ppmoscillator disciplinedto the RX (incoming)line rate. There isa Sync relationship

between the RX andTX, allowing theincoming clock to bepropagated.

By adding an external sync port to the SyncE switch, a Stratum1 reerence can be introduced to, and distributed through, apacket network independently o the trafc.

Unortunately cascading synchronous and asynchronous(traditional) switches will interrupt the originating (andaccurate) sync ow, making SyncE a point-to-point requencydelivery method. This means that cascaded synchronous andasynchronous switches will not deliver the source requency

through the network. The 100ppm reerence will be substitutedin the asynchronous switch, breaking the timing chain.

There are many similarities to SDH/SONET distribution, andthis was by design. SyncE distributes requency rom a sourceto a destination, and Primary Reerence Clocks are neededat the source. SyncE accumulates jitter and wander over thepath just like SDH/SONET, and SSUs must be used to flter the

jitter and wander per the ITU-T G.823 and Telcordia GR.253guidelines.

The 4.6ppm SyncE Equipment Clock (EEC) is the same onedefned or SDH/SONET Equipment Clocks (SEC/SMC).

And just like SDH/SONET, SyncE only distributes requency.SyncE does not distribute a Time o Day or phase reerence.

In summary, SyncE has the advantage o being a deterministicrequency distribution method that is independent o the dataow. However, all the switches in a path must support SyncE,and widespread adoption will be governed by the cost and easewith which installed asynchronous switches can be upgraded.

IEEE 1588-2008 (PTP v2)

The IEEE 1588-2008 protocol (also called Precision TimeProtocol or PTP) is a standardized method to distributeaccurate time and requency over IP networks. The basiso operation is that packets carry timestamp inormationbetween a master (sometime called a server) and slave(sometimes called a client), and the slaves use the timestampsto synchronize to the master. Bidirectional ows eliminate the

round trip delay to enhance the accuracy. Frequency can, inturn, be recovered rom the disciplined Time o Day clock.

The 1588 timing and management messages are transportedin-band with the mainstream trafc, eliminating the need or adedicated timing plane.

100 ppm

TX

RX

TX

RX

4.6 ppm

TX

RX

TX

RX

4.6 ppm

TX

RX

TX

Ext. sync

RX

Cascaded SyncE and asynchronous switches willtransport data, but cannot not distribute the sourcefrequency.

IEEE 1588 is an in-band distribution system,eliminating the need for a dedicated timing plane.

Planning for SyncE Jitter and Wander ltering followsthe ITU-T G.823 guidelines.

4.6 ppm

Accurate TX

RX

TX

Ext. Sync

SyncE Switch Asynchronous Switch

RX

100 ppm

TX

RX

TX

RX

Inaccurate

FIG 4: Broken Synchronization Chain


6/13

6


IEEE 1588 was initially developed or industrial automationover Local Area Networks, but a second version, tailored orconstrained telecommunication environments, was publishedin 2008. The pending ITU-T G.8265 Telecom Profle simplifesthe diversity o confgurable parameters needed to supportWANs, improving the protocols interoperability.

What makes IEEE 1588 very attractive is the microsecondaccuracy (and associated 1ppb requency stability) that can berealized over managed Ethernet. This allows PTP platorms tosupport a wider range o applications than any other solution,addressing both the FDD and TDD modes o LTE.

Being packet based, IEEE 1588 is sensitive to the networkbehavior and the accuracy depends on the clock recoveryalgorithm and the packet jitter (also called Packet DelayVariation or PDV). In general, meeting the requencyrequirement is moderately easy, but phase synchronization ismore sensitive to PDV and requires added planning.

Fortunately the protocol designers oresaw this, and includedon-path support in the specifcation. On-path support consistso Transparent and Boundary clocks that reduce the packet

jitter, improving perormance over long hop counts. On-pathsupport will be discussed in more detail later.

Network Time Protocol (NTP)

NTP is the most popular protocol or distributing timeover LANs and WANs, and when looking at IEEE 1588, the

similarities cannot be ignored. The reasons NTP is not used inhigh perormance applications are the low transaction rate (1per 64 seconds) and the sotware nature o the solution. NTPpackets pass through the Ethernet PHY and Media AccessControl (MAC) layers like any other packet. CPU processing orthe timing requirements is not addressed until the sotwarestack is ully processed. The NTP packets are thereore delayedby an indefnite time depending on the operating systemlatency, limiting the assured accuracy o the solution.

NTP will continue to be used in LTE or Call Detail Records andbilling, just as it was in earlier mobile generations. NTP timesynchronization is also used by switch and router elementsin IP networks to monitor perormance and optimize routingtables (with one way latency methods being notable). Carrier-class NTP servers distributed at mobile switching ofcesremains a undamental component o any wireless system.

It is worth mentioning that a Femtocell assumes the userto be essentially stationery, and the 50ppb air interace orthis technology was relaxed to ~250ppb. As a result, someFemtocell FDD solutions have embedded NTP clients that

recover requency rom the Time o Day.

Embedded

Slave

External Slave

Grandmaster

(Server)1588 Packet Flow

1588

IEEE 1588 over Ethernet designed for QoS can delivermicrosecond accuracy.

Call Data Record management and routing tables arestill dependent on NTP services.

FIG 5: IEEE 1588 Architecture


7/13

7


IEEE 1588-2008 (PTP) & SyncE In Perspective

How are IEEE 1588 (PTP) and SyncE similar? How are theydierent? And what does this mean to you?

Both SyncE and IEEE 1588 are standards-based methods totransport requency (and time in the case o PTP) through thenetwork to heal the broken synchronization chains. Where theydier is in implementation and unction.

The table below summarizes key dierences between thetwo, but in essence SyncE is a conscious decision to add the

eature to every switch between the source and the destination.To recap, cascaded synchronous and asynchronous switcheswill not transport synchronization (even though they canroute data). IEEE 1588 is largely independent o the transportelements; largely because boundary and transparent clocksmay be embedded in switching elements, but this is not a pre-requisite. This allows PTP networks to be built independentlyover diverse transport systems.

It is important to note that SyncE and 1588 are not mutuallyexclusive. SyncE unctions at layer one independently o the

trafc. IEEE 1588 unctions at higher layers (UDP/Ethernet)independently o the transmission rate. It is entirely possiblethat both technologies could be used on the same path - SyncEor requency and PTP or time/phase. In reality, slaves thatsupport both can converge on an accurate time very quickly byusing the SyncE requency to discipline the local oscillator.

ENGINEERING FOR COST, PERFORMANCE & SIMPLICITY

Selecting A Synchronization Strategy

Assuming a standards-based timing solution is chosen,synchronization will be based on either Synchronous Ethernet(SyncE) or IEEE 1588 (PTP), or the two together. Both thesetechnologies have distinct advantages and disadvantages overeach other. SyncE is deterministic and the perormance isindependent o the trafc. PTP can unction over asynchronousswitches, and distributes requency and time with the datatrafc.

Selecting one over the other is an economic reality as muchas a technical reality, but setting aside the cost o upgradingthe switches to SyncE, deployments with the ollowingcharacteristics are likely to be based on IEEE 1588:

Need time/phase such as LTE TDD and SFN-MBMS

Do not have end-to-end SyncE switches (Ethernet)

Function over diverse transports (microwave, Ethernet,SHDSL..)

Share leased network sections, unless SyncE or a SDH/SyncEhybrid can be assured over the ull path

Keeping these limitations in mind, an operator planning todeploy LTE (TDD mode), or SFN-MBMS would typically selectIEEE 1588 as the primary synchronization method. SyncE wouldonly address requency synchronization requirements. Thisdoes not preclude using a combination o SyncE or E1/T1 tosynchronize legacy base stations.

For LTE specifcally an IEEE 1588 strategy requires:

Carrier-class Grandmaster clocks installed at strategiclocations

A managed Ethernet network designed or QoS

Base stations with integrated 1588 slaves or with external

sync ports, including 1PPS/ ToD or phase synchronization.By defnition LTE base stations will not have external syncports, but these may be available or early deployments.

SyncE and IEEE 1588 are complementary technologiesand can co-exist in the network.

SyncE in its current form cannot fulll the phaserequirements of LTE TDD or SFN-MBMS.

Selecting a Sync strategy is an economic and technicalreality, determined in part by the mobility protocols.

Attribute IEEE 1588 SyncE

Capability Frequency, Phase, Time Frequency

Layer Ethernet/UDP Physical

Distribution In-band 1588 Packets Physical layer

Schema Point to multi-point Point to point

Transport Media Ethernet, SHDSL, Native Ethernet,Microwave Other in development

Interoperability Standards-based Standards-based SyncEGrandmaster & slave switches only

Sensitivity Packet Jitter / Bandwidth Asynchronous switchesutilization

Standards IEEE 1588, ITU G.8261/3/5 G.8261/2/4

TABLE 2: IEEE 1588 / SyncE Comparison


8/13

8


Dening the Sync Objectives

The air interace requirement defnes the stability that must bemaintained between the base station and the handset, typically

over 1 time-slot. For LTE TDD, the requency stability must be50ppb, and the phase alignment 3S. The air interace needsmust not be conused with the synchronization that needs to bedelivered to the base station (on the backhaul interace).

We also know that when TDM network synchronizationobjectives are met, 50ppb or better can be maintained on theair interace. The TDM network perormance objectives aredefned in ITU-T G.823 and G.824 in the orm o masks.

Masks defne the MTIE limits (quality o synchronization) that anetwork should conorm to over 100,000 seconds (~28 hours).Depending on where the clock signal is measured, there aretwo masks, the trafc and the synchronization mask. I theclock signal is being measured at an end point, it mustcomply with the trafc mask.

I it is measured along the synchronization chain, it shouldcomply with the more demanding synchronization mask.

The BTS is an end point, and by defnition, a carrier-gradeservice can be delivered i the network jitter is lower than(below) the trafc mask. This is also true in practice, but

some operators are more cautious, and have selected thesynchronization mask as the minimum perormance to be metat their BTS interaces.

How do we reconcile IEEE 1588 to the TDM mask? Packetnetworks dont transport synchronization, but the TDMinteraces are still synchronous. Ultimately, the 1588 packetsmust be converted to an analog source, and the quality othe recovered clock at the BTS must pass the MTIE test (asmeasured with a traditional SDH analyzer). I the measuredresult is under the mask, the 1588 system has worked. I not, ithas ailed and some level o re-engineering is required.

Because TDM masks only defne the perormance over ~28hours, it is not uncommon or operators to defne a long termstability requirement. A long term stability o 15ppb is typicallyspecifed and adequate or mobile base stations.

Finally, or TDD applications, the phase alignment (usuallymeasured on a rising or alling edge) o the source(Grandmaster Clock) and slave must be less than the specifedvalue, 3S or LTE with TDD support. This is also moreaccurate than the 6.4S or SFN-MBMS, addressing thatapplication well.

Sync Masks are used to dene goals and determinewhether those objectives are being met.

Selecting the short and long Sync terms goals is therst step in engineering an all IP network solution.


9/13

9


The quality o the synchronization delivered to the BTS or anLTE TDD service would typically be defned as:

Meet the Trafc mask as defned in ITU-T G.823 over the frst100,000 seconds

Maintain a long term average requency stability o 15ppb(1.5x10-8) or better

Maintain the phase dierence between the source and theslave o less than/equal to 3 microseconds.

What Affects IEEE 1588 Accuracy?

What aects IEEE 1588 clock recovery, and how it is mitigated?First, lets set aside the attributes that dont aect 1588. Packetlatency, packet loss and packet errors do not signifcantlyimpact the clock recovery protocols.

The quality o the recovered time and requency does dependon:

1) The stability o the local (slave) oscillator

2) The Packet Delay Variation3 (PDV) o the IEEE 1588 packets

3) The quality o the servo-loop algorithm in the slave.

It should be apparent that the Grandmaster has not beenmentioned, and or good reason. The slave design determinesthe achievable perormance. High-cost oscillators are notan option in base stations already under pricing pressure, soPDV (or the elimination o PDV) becomes the overwhelmingconsideration.

PDV is governed by the bandwidth utilization, upstream/downstream path symmetry, and the number o hops (storeand orward buer devices) between the Grandmaster and theslave. There is not much that can be done about variations inbandwidth use, but the 1588 packet PDV can be reduced in two

ways. First, by fltering PDV in the clock recovery algorithmand second, by reducing the number o switch and routingelements between the Grandmaster and the Slave during thenetwork planning stage.

On-Path Support: Pushing the Limits

As described earlier, one method used to reduce PDV is tolimit the number o switching and routing elements betweenthe Grandmaster and the slave. Where this is not possible ordesirable, a second method is to minimize the eect o switchtransit delays (the time that 1588 packets spend in the switch).To do that, the IEEE 1588 standard includes two new clocktypes, the Transparent Clock and the Boundary Clock.

The 1588 slaves PDV rejection capability governs thequality of the recovered clock.

Eliminating Packet Delay Variation (PDV) has thelargest effect on 1588 clock recovery.

3 Packet Delay Variation is similar to packet jitter, but includes the low requency wander components.

FIG 6: Perormance Masks


10/13

10


Boundary Clock (BC)

The Boundary Clock was defned in both version 1 and 2 o theIEEE 1588 standards, and is a special orm o network switch

or router. From the network perspective, a boundary clock is anormal switch element, passing network trafc seamlessly.

The clock plane is very dierent. The boundary clock has anembedded 1588 slave that synchronizes the internal clockrom an upstream IEEE 1588 ow. It also has a Grandmasterunction providing 1588 ows to downstream slaves within thenetwork. One port othe BC is a designated1588 slave port, and theremaining ports are 1588master ports.

The master/slaveunctions in the boundaryclock terminate a syncow4, and generate newow(s), thus reducingthe PDV that has beenaccumulated. Reducingthe PDV is important,but in telecommunication applications, the Grandmaster is a

carrier-class device with high stability holdover oscillators,management and resilience. The ability o the boundaryclock to provide the same capability will depend on the 1588implementation by switch vendors.

Transparent Clocks (TC)

Transparent Clocks were added to the second version o the1588 standard to support cascaded topologies. Like boundary

clocks, transparent clocks are a special orm o network switchor router. Unlike boundary clocks, timing is not recreated, butthe residence time (time that 1588 packets spend in a switch)is measured and reported to the slave. The transparent clockdoes this by insertingthe residence timeinto the correctionfeld in the sync,delay-request andollow-up messages.

Two transparentclock modes aredefned, peer-to-peer and the morewidely advocatedend-to-end version,accumulating theresidence time o allthe Transparent Clocks(switches) in the correction feld. This allows the highly variableresidence periods to be removed rom the round trip delaycalculation, reducing the PDV considerably.

We know that boundary and transparent clocks reduce the 1588message PDV, and can increase the achievable perormance ora fxed path length, or increase the path length or a fxed goal.

But is it necessary?

As our requency and phase results show, boundary andtransparent clocks are not necessary or FDD mobile systems(assuming the number o links between the MME and eNodeBis typical). Depending on the slave clock perormance, TDDapplications with phase requirements can also be met withouton-path support.

GMC

Slave

Boundary Clock

4 Sync, Follow_Up, Delay_Req, and Delay_Resp messages.

The burden of holdover and traceability belongs to theBoundary Clock when it is used.

Transparent clocks preserve the integrity of the syncow, but are limited in encrypted networks.

On-path support is not always necessary for IEEE1588 applications, particularly for frequencydistribution.

PTP Packet PTP Packet

Transparent Clock

Residence Time =

Packet Egress Packet Arrival

Arrival

Time

Egress

Time

FIG 7: Boundary Clock

FIG 8: Transparent Clock


11/13

11


When unctioning over a managed, symmetric metro Gigabitnetwork, engineered or QoS, with 5 to 10 switching elementsand controlled load, the Symmetricom slave can typicallydeliver 1ppb requency stability and 1 microsecond o phaseaccuracy5. These values will vary as a unction o trafc load,switching jitter and path asymmetry. Having said that, a well-designed slave will beneft rom boundary and transparentclocks.

In a world characterized by rapid change, the ITU-T workinggroups decision to exclude on-path support rom the Telecomprofle is a welcome step in reducing early deployment risk and

complexity. But to ensure that your investment is protected, theslaves that you buy today must process the correction feld in thePTP messages. This will ensure that the 1588 slaves can takeadvantage o on-path support elements i added in the uture.

Deployment Made Easy

Simple deployment guidelines ensure the best 1588 results:

First select a high perormance, high-reliability Grandmasterclock that has capacity or the number o slaves expected,now and in the uture. A system that is scalable to 1,000slaves, at a transaction rate o 64 per slave is typical.

The Grandmaster capacity can be checked by dividingthe maximum transaction rate by the number o servedslaves and the number o transactions per second. I thedemand exceeds the capacity, additional modules (or evenGrandmasters) can be added to share the load. Served slavesis the sum o the slaves that are expected to obtain sync owsrom a Grandmaster routinely, and those that could requestthe service i their designated GMC clock ails.

Second, select a good quality 1588 slave that has beendesigned to work over the transport being used (e.g. NativeEthernet, Ethernet over SDH, Microwave and SHDSL).

Embedded slaves will become commonplace, but stand-alone slaves allow early deployment risks to be reduced, aswell as serving the large installed base o legacy devices thatcan only accept the DS1/E1 Sync reerences.

The next step is locating the Grandmasters based on themessage rate o the Grandmaster and the PDV that eachslave will experience. Cost considerations encourageewer Grandmasters unctioning through more switchesand routers. Robustness calls or more Grandmasters,with ewer links between the Grandmaster and slave.Telecommunication networks avor robustness or Quality oService.

Experience has shown that the best place to locate theGrandmaster is at the RNC. LTE does not have a dedicatedRNC, but it is reasonable to assume that the MME will be

High quality clock recovery algorithms havedemonstrated the extent to which PDV ltering iseffective.

Validation of the slaves ability to use transparentswitch delays is key to a future-proof solution.

Probing a representative sample of backhaul linksprovides insight into the PTP capability of path.

5 The solution reerred to is the Symmetricom slave. Not all slaves have the same perormance and results will vary.


12/13

12


located at the same locations as the RNC or aggregationnodes in the oreseeable uture. The number o hops betweenthe MME and eNodeB will typically be about fve. In addition,the Ethernet path is likely to be managed, and a reasonablylow PDV can thereore be expected.

A representative sample o links should be probed to validatethe Grandmaster location assumptions. Probing the networkin this case reers to characterization o the PDV using a 1588tool, and not the measurement o data packet latency or loss.

And fnally, the 1588 elements should be managed oraults, availability and synchronization perormance. This isespecially true or boundary and transparent clocks i theyare used. I working correctly, on-path support can improvethe result, but i it does not unction correctly, the impact onclock recovery can be devastating.

Reer to the Symmetricom document Deployment o PrecisionTime Protocol or a more detailed IEEE 1588 planning guide.

CONCLUSION

Although revenue is derived rom voice services today, mobilebroadband is the new business or service providers. To beviable, high bandwidth must be oered or less revenue, andbe augmented with value added services. LTE and IP mobilebackhaul are two signifcant steps in that direction.

The economics or IP backhaul is compelling, but IP createssynchronization discontinuities that must be consciously

addressed in the network. The international community hasaccepted two new methods to distribute synchronizationthrough packet switched networks. The frst is SynchronousEthernet, that transports requency at Layer 1 or a

deterministic result. The second is the IEEE 1588 precisiontime protocol that unctions over higher layers. IEEE 1588is more versatile, but is sensitive to network behavior. It isimportant to remember that SyncE and IEEE 1588 are notmutually exclusive and can be used together in some networks

Both methods are valuable to network planners or FDDmobility protocols, but IEEE 1588 is the preerred method orTDD and SFN-MBMS applications that require requency andphase.

On-path support elements (boundary and transparent clocks)are not necessary in most FDD applications, and depending on

the client perormance and network design, on-path supportmay not be necessary or TDD applications. An improvementin the quality o the extracted clock can be expected i on-pathelements are used and managed rom a time perspective.Minimizing the use o on-path support in the short term helpsreduce early deployment risks.

Ultimately, deploying a standards-based synchronizationsolution or next generation mobile networks is key toinvestment protection, as well as meeting long termrequirements. A universal synchronization platorm is mostcost-eective, more robust and plays a large part in theoptimization o the network.

To fnd out more about next generation synchronization,contact your local Symmetricom representative, or e-mail us [email protected]

Probing a representative sample of backhaul linksprovides insight into the PTP capability of the path.

Standards based solutions are interoperable and willgrow with tomorrows needs.


13/13


Copyright 2010 Symmetricom, Inc. All rights reserved. Symmetricom and the Symmetricom logo are registered trademarks o Symmetricom, IncAll other trademarks are the property o their respective companies. All specifcations subject to change without notice. 04-27-1

www.symmetricom.com Phone: +1 408 428 7907; toll-ree: +1 888 FOR SYMM | Fax: 408-516-9597 | E-mail: [email protected]

ABBREVIATIONS USED

BC...........Boundary Clock

BTS ......... Base Station Transceiver

CES ......... Circuit Emulation Service

DSL ......... Digital Subscriber Line

EEC......... Ethernet Equipment Clock

ETSI ........ European Telecommunications Standards Institute

FDD ........ Frequency Division Duplex

GigE ........ Gigabit Ethernet

GMC........ Grandmaster Clock

GNSS ...... Global Navigation Satellite System

GPS ........ Global Positioning System

IEEE ........ Institute o Electrical and Electronic Engineers

IP ............ Internet Protocol

ITU-T ...... International Telecommunication Union

LBS ......... Location Based Service

LTE ......... Long Term Evolution

MBMS ..... Multimedia Broadcast Multicast Service

MTIE ....... Maximum Time Interval

NGN ....... Next Generation Network

NTP ........ Network Time Protocol

PDV ......... Packet delay Variation (synonymous with PacketJitter)

PTP ......... Precision Time Protocol

PWE ........ Pseudo-Wire Emulation

SEC ......... SDH Equipment Clock

SDH ........ Synchronous Digital Hierarchy

SFN ........ Single Frequency Network

SONET .... Synchronous Optical Networking

SMC ........ SONET Minimum Clock

SyncE ..... Synchronous Ethernet

TC ........... Transparent Clock

TDD ........ Time Division Duplex

TDM ........ Time Division Multiplex

TOA ......... Time O Arrival

UDP ........ User Datagram Protocol

UE ...........User Equipment

UTC......... Universal Coordinated Time

WAN ....... Wide Area Network

Documents

WP Sync IP Mobile Networks