In many ways, the requirements of the telecommunications field appearat odds with the design objectives for the original IEEE 1588 standarddiscussed in earlier Part 1 , Part 2, Part 3 and Part 4 in this series.
In particular, IEEE 1588 was designed forlocalized industrial and instrumentation networks, hardly a descriptionof a telecommunications network. Yet, at the very first IEEE 1588conference in 2003, a paper was presented by Glenn Algie of Nortel ,outlining how IEEE 1588 could be used to solve some of the emergingproblems in telecommunication systems.
Since this initial presentation, there have been discussions andpresentations of IEEE 1588 at several telecommunications forums,including papers by Algie, Algie and Ouellette, Rodrigues, Tonks, and Zampetti.
There have been several field trials of various implementations ofIEEE 1588, but as yet no announced commitments by anytelecommunications standards body, service provider, or equipmentmanufacturer to provide IEEE 1588-based technology intelecommunications.
This situation is expected to change but for now this discussionremains highly speculative. Much of the material in this section isbased on private conversations with Glenn Algie of Nortel, andunpublished memoranda from Doug Arnold of Symmetricom, SilvanaRodrigues of Zarlink Semiconductor, and Dave Tonks of Semtech.
|Figure9.1. Proposed uses of IEEE 1588 in telecommunications (courtesy ofNortel)|
Background on using IEEE 1588 inTelecom Systems
Algie envisioned the system illustrated in Figure 9.1 above . Shown in thisfigureare the metropolitan core and collector networks that form thebackbone of the telecommunications systems found in most largemetropolitan areas.
These networks typically are formed with SONET and othercircuit-switched network protocols. As everywhere else, there areeconomic forces that make the use of Ethernet-based technologyattractive for this application.
Indeed, there is a trade association—the Metro EthernetForum1—formed with the explicit purpose of promoting the use ofEthernet in this, and other related applications. The primary drivingforce is the switch from circuit-based to packet-based communicationsfor voice-over-IP (VoIP) applications.
Unfortunately, the packet-based networks do not provide the samelevel of timing services to enterprises as those currently derived fromthe circuit-switched networks.
Connections to various users are made via routers and switcheslocated at points on the collection rings. Typical services arerepresented by the lettered squares A, B, C, E, and K in the figure.Each of these is separated from the core networks by a boundary termedthe “user-network interface”, or UNI.
In the original presentation of Algie, a number of services wereidentified that could benefit from timing distributed using IEEE 1588.These proposals, and others that have emerged from the presentationsand discussions cited earlier are summarized here.
Algie proposed providing a stable source of IEEE 1588 timing packetsto be distributed over the metro core networks. These source devicesare illustrated in Figure 9.1 by the items X, Y, and Z that representstratum 1 traceable clocks implementing IEEE 1588 master clockfunctionality.
These devices could be owned by the telecommunications serviceproviders, as illustrated by items X and Y, or could be provided fromthe enterprise side of the UNI demarcation, as shown by item Z.
The timing information would be recovered at edge points G, B, C, E,and K. These can be on either side of the UNI demarcation. Therecovered timing signals would then be used by the associatedenterprise services.
The actual timing requirements for the proposed services vary. Somewill require only highly accurate frequency information, whereas othersrequire epoch information as well.
IEEE 1588 was not designed to distribute timing information over thenetworks topologies proposed for telecommunications. Even within ametropolitan area, the core networks will encompass many switches, eachof which introduces appreciable timing fluctuations.
The method IEEE 1588 specifies for overcoming router and switchtiming fluctuations is the boundary clock, and depending on the outcomeof the current standardization efforts, the transparent clock.
It is unrealistic to simply assume that telecommunications companiesare going to replace existing networking equipment with routers andswitches that implement IEEE 1588.
If IEEE 1588 is to work as proposed, something must be done toovercome the network timing fluctuations. If epoch is important, thenit will also be necessary to somehow control the symmetry properties ofthe network.
As the following discussion will make clear, some of these measuresrequire special configurations of network properties such as routingtables.
Is it reasonable to expect telecommunications providers to engineertheir networks for the purpose of meeting the requirements of IEEE1588? The likely answer is yes, if they are convinced that IEEE 1588will actually provide the needed timing using the packet-basednetworks.
The reason for this optimism is that these companies do this for thecurrent circuit-switched technology. In existing switched networks,there is an elaborate system for distributing frequency based onin-line signals in the data streams between network devices.
A system of primary reference clocks (PRCs) has been established,with elaborate provision for redundancy, alternate paths, and recoveryprocedures. This system has required the telecommunications companiesto not only engineer their networks, but also to use special equipmentto recover the timing.
However, to make the initial inroads, a technology such as IEEE 1588must be able to demonstrate adequate performance for an economicallyattractive set of applications using the existing network equipment.
Currently there are a number of promising proposed applications ofIEEE 1588 in telecommunications in wireless networks, linking SONETrings via Ethernet, Cable TV and Internet TV, central office systems,TDM circuit emulation, and for internal timing in a wide range oftelecommunication router and switching equipment.
Applying IEEE 1588 to WirelessNetworks
The synchronization of wireless networks is one of the more compellingand most difficult applications of IEEE 1588 in telecommunications. Allwireless cellular protocols require either frequency or epochsynchronization to prevent frequency interference and dropped calls dueto handover failures. Table 9.1 below lists the requirements for several wireless protocols in current use.
|Table9.1. Timing requirements for selected wireless telecommunicationprotocols|
TheW-CDMA and WiMAX protocols have both frequency division duplex(FDD) and time division duplex (TDD) modes of operation. There are twosynchronization techniques used in today's wireless networks.
The first is to install a GPS receiver at each cell site. The GPSreceiver, often in combination with a rubidium clock, is used toestablish the needed frequency and epoch timing.
A GPS system has a substantial installation and equipment cost, andis subject to lightning strikes. In many cases, the installation mustbe located where there is not a clear view of the sky, which furtherincreases the installation cost because lengthy cables must be run tothe GPS antenna.
GPS signals require averaging over several minutes to achieveaccuracies in the range of 108 . There is also considerablereluctance, particularly by non-American telecommunications companies,to rely on the GPS system. GPS synchronization is widely used inCDMA2000 base station applications.
The second technique, widely used for GSM and W-CDMA (FDD mode)applications, is to provide synchronization over the backhaul links tothe base station controllers. These links may be T1 or E1 links,microwave links, or, increasingly, some form of Ethernet.
If IEEE 1588 can adequately provide timing information over Ethernetlinks, then there will be considerable savings due to the highertariffs on the other forms of backhaul links. Since a backhaul isalways present, a successful application of IEEE 1588 could eliminatethe need for a GPS system at each base station.
Instead, the timing would be derived from a GPS installed at one ofthe base stations or at the base station controller, and distributedthroughout the cluster using IEEE 1588. Alternatively, the timing couldbe derived from the network terminating at the controller, anddistributed to the base stations.
|Figure9.2. IEEE 1588 timing distribution in wireless base station boundaryclusters (courtesy of Symmetricom)|
These alternatives are illustrated in Figure 9.2 above , adapted fromdrawings provided by Doug Arnold of Symmetricom. Figure 9.2A and Billustrate the cases where the GPS installation is at the base stationcontroller and at the base station itself. Figure 9.2C illustrates thecase where the IEEE 1588 timing information is derived from the largernetwork, as shown in Figure 9.1.
Rodrigues notes that the backhaul links between base stations and basestation controller are generally not simple single-hop links, butusually involve several links in the packet-based networks, asillustrated in Figure 9.3 below .While currently a variety of protocols are used in backhaul links,future installations will increasingly use Ethernet links, asillustrated.
The use of IEEE 1588 to provide the timing references for the basestations using any of the schemes of Figure 9.2 remains to bedemonstrated. The data presented later in this article are encouraging,particularly for applications such as GSM and W-CDMA (FDD). Timing forCDMA2000 and other applications that require epoch and frequency is yetto be demonstrated in a telecommunications environment.
|Figure9.3. IEEE 1588 in wireless backhaul boundary links (courtesy of ZarlinkSemiconductor)|
Using 1588 to link SONET Rings viaEthernet
This section is based on material provided by Doug Arnold ofSymmetricom. As noted earlier, telecommunications companies areincreasingly using Ethernet backbones to replace existing networklinks. However, it will be some time before all of the existingbackbone links, typically implemented with SONET rings, will bereplaced.
As a result, there will be cases in which two SONET rings are joinedby an Ethernet link. To preserve the timing, the two SONET rings mustagree in frequency to one part in 1011 . It is possible thatIEEE 1588 can be used to convey the timing information between the tworings.
Both rings will be equipped with high-quality oscillators, so thatlong averaging times may be used to reduce the timing fluctuationsintroduced by the Ethernet link that would otherwise preclude the useof IEEE 1588. This, plus careful design of the Ethernet links, will nodoubt be necessary to achieve one part in1011 .
How IEEE 1588 benefits the Cable TVInfrastructure
The current architecture of a cable TV distribution system isillustrated in Figure 9.4 below .Shown is a typical head-end installation. Digital data from theinternet are received from a wide-area network (WAN) connection to thehead-end, combined with VoIP traffic from the public-switched telephonenetwork (PSTN), and transferred to the cable modem termination system(CMTS) via the head-end switch.
|Figure9.4. Current cable TV architecture (courtesy of Zarlink Semiconductor)|
It is possible for the CMTS to be located in a master head-end ifthe CMTS functions can be shared by users connected to hubs in thenormal head-end locations.
Currently, the digital link between a customer's cable modem and thehead-end terminates at the CMTS. Digital data to be placed on the linkare transmitted via the quadrature amplitude modulator (QAM), and arecombined with video traffic by the distribution hub.
The physical link is either cable or hybrid fiber cable. These linksuse one of two modulation techniques specified in CableLabsspecification, Data-Over-Cable Service Interface Specification 2.0,each with a specification on timing jitter at the output of thedownstream transmission convergence sublayer.
For the Advanced Time Division Multiple Access (A-TDMA) technique,the jitter must be less than 500 ns peak-to-peak, whereas forSynchronous Code Division Multiple Access (S-CDMA) the jitter must beless than 2 ns peak-to-peak.
Both coding schemes require compensation for the latency between theCMTS and the individual modems. This is accomplished by rangingalgorithms that calibrate and analyze both the downstream and upstreamchannels. In the current architecture, all of these functions areimplemented on the line cards of the CMTS.
To simplify installation and reduce costs, the cable operators wouldlike to combine expensive functions of the head-end into fewerinstallations, and replace the current local head-end installationswith simpler devices.
The operators also want to utilize their cables to provide so-calledtriple-play service that combines voice, video, and data, and to makeuse of lower-cost Ethernet links wherever possible. To accomplish boththese goals requires a change in the current architecture to anarchitecture closer to the one illustrated in Figure 9.5 below.
|Figure9.5. Future cable TV architecture (courtesy of Zarlink Semiconductor)|
In this future architecture, the CMTS functions are located inregional head-end installations. Combined voice, video, and data aredelivered to local edge-QAM devices via gigabit Ethernet (GigE) links,and then to the home via hybrid fiber channel links.
The uplink for data is provided via a separate Ethernet networkdirectly from the modems to the CMTS. The synchronization requirementsremain as before, but are now separated between the CMTS-to- QAM link,the QAM-to-modem link, and the uplink. Synchronization of the Ethernetportions of these links is an application addressable by IEEE 1588.
Appling 1588 to Central OfficeTiming Distribution
In the current telecommunications system, the central office is thelocation where local access terminates and connects to trunk links toother central offices, or to long-distance networks.
Within the central office, it is customary to distribute frequencyvia the Building Integrated Timing Supply, or BITS. Most BITS derivefrequency from a GPS installation with backup from land-line-basedsources. The BITS system currently supports only frequencydistribution. Epoch or time is normally distributed using othertechniques such as NTP.
If more accurate epoch information becomes necessary within thecentral office, the use of IEEE 1588 to distribute both time andfrequency over the building LAN becomes more attractive.
Possible uses are for precise logging and billing of data inpacket-based services. As in both industrial automation and test andmeasurement, a precise time service will be used for timestamping andlogging anomalous events.
It is quite likely that future packet-based services to homes andenterprises will require time as well as frequency. VoIP services willbe enhanced by precise timing. There is also considerable discussion inthe home entertainment community about providing synchronizationservices within the home (see, for example, the tutorial of Teener etal. ).
Simplifying TDM Circuit Emulation
Circuit emulation is a technique for carrying existing time divisionmultiplexed (TDM) telecommunications traffic over packet-basednetworks. Carriers are replacing traditional T1 and E1 TDM links withpacket-based links, but must continue to support customers usingTDM-based services, such as existing T1 lines, PBX, and fax.
Circuit emulation works by creating a tunnel for TDM data throughthe packet network. Equipment is provided at each end of a packet-basedlink to packetize or de-packetize the TDM data. The difficulty is thatthe TDM links depend on synchronous clocks at both the transmitting andreceiving ends of the link.
This timing is lost in the packet-based link and, if uncorrected,will result in corrupted or degraded performance. For example, if theclock of a receiving T1 link is running slower than the clock in thetransmitting link, then packets will be dropped.
TDM networks have very tight requirements on the allowed variationbetween the transmitting and receiving clocks. These requirements aredefined in a number of ITU recommendations, including G.823 and G.824.
Frequency differences between these clocks are characterized byhigh-frequency jitter, which is fairly easy to filter, andlow-frequency wander, which is much more difficult to eliminate.Frequency wander between two clocks will lead to dropped packets(slips) unless sufficiently large data buffers are provided.
Buffers add cost, introduce latency, and can accommodate only afinite level of wander, leading the ITU to specify the maximum numberof slips such that no more than 125 microseconds of data is lost in any70-day period.
There are currently two techniques used to provide the neededfrequency synchronization. The first is called adaptive clocking. Inadaptive clocking, the receiver adjusts its clock to match the rate ofincoming information by averaging the arrival rate or the inter-arrivaltimes of incoming packets, or by monitoring the fill levels of thebuffers.
This technique is subject to the same difficulties facing IEEE 1588,namely,packet loss, latency fluctuations, and changing distribution offluctuations.
However, the circuit emulation services are unlikely to be able touse the very long-term averaging algorithms available to an IEEE 1588installation, either due to the lifetime of the circuit, or the need tomore closely track the actual information rates.
The second technique is called differential clock recovery. In thiscase, a high-quality clock, such as a primary reference source (PRS),is provided at both ends of the packet network.
The difference between the TDM service frequency and the PRS at thesending end is encoded in the TDM packets transmitted over thepacket-based network. At the receiving end, this differentialinformation is used to reconstruct the TDM timing based on the localPRS.
This will work in central offices, but not in the planned extensionsusing Ethernet links between the central office and the emulatedservices in customer premises. IEEE 1588 can provide the requiredtiming in these situations.
Uses in telecom equipment internaltiming
The last example to be discussed is to use IEEE 1588 to distribute timewithin telecommunications equipment. Current designs of this equipmentare a mixture of centralized functions and line cards specific to eachlink terminating at the equipment.
Internal synchronization often involves precision oscillators online cards as well as in the centralized functions. IEEE 1588 has beenproposed as a means of distributing time from a few very stable centraloscillators to less expensive but more numerous oscillators on the linecards. The IEEE 1588 traffic could be carried either on the controlplane or in-line with the data.
Next in Part 6: Enabling 1588innetworking and the future of the standard.
To read Part 1, go to “
To read Part 2, go to
To read Part 3, go to “
To read Part 4, go to “Achievingsubmicrosecond synchronization accuracy“
Used with permission of its publisher, Springer Science andBusiness Media, this series of articles is based on material from “
John C. Eidson, Ph.D., received a B.S. and an M.S. from MichiganState University and a Ph.D. from Stanford University, all inelectrical engineering. He held a postdoctoral position at Stanford fortwo years, spent six years with the Central Research Laboratory ofVarian Associates, and joined the Central Research Laboratories ofHewlett-Packard in 1972. When HP split in 1999, he transferred to theCentral Research Laboratory of AgilentTechnologies. Dr. Eidson was heavily involved in IEEE 1451.2 andIEEE 1451.1 and is the chairperson of the IEEE 1588 standards committeeand a life fellow of the IEEE.