April 1, 2007

Synchronization

Avoid Timing Meltdown

By Jeremy Bennington, Symmetricom

Synchronization of time and frequency has always been crucial to cable networks since the development of DOCSIS, the first and still current cable-modem interface standard.

Synchronization remains essential to cable networks for two reasons. The first reason is that because the physical transmission medium is shared by all cable modems on the network, basic connectivity is likely to cause high levels of transmission interference unless synchronization is precise. As shown in Figure 1, every cable modem within a network connects to the cable modem termination system (CMTS). As a cable modem is turned on, it must first go through a ranging process to synchronize its frequency and timing to the CMTS. This process ensures that all cable modems sharing the HFC plant and the CMTS do not interfere with each other.

Figure 1: CMTS overview

In asynchronous time division multiple access (A-TDMA) mode, each cable modem gets a specific timeslot to transmit, and all timeslots are aligned among hundreds of cable modems so that no two modems on a given channel transmit data during the same timeslot (except for contention slots). In synchronous code division multiple access (S-CDMA) mode, cable modems are perfectly aligned to transmit simultaneously on the same RF channel during the same time slot. Perfect alignment is mandatory to ensure that the CMTS correctly demultiplexes the bursts to determine the data transmitted from the various cable modems. In either mode, if cable modems are not properly synchronized, transmissions will be completely lost.

The second and most recent reason that synchronization is essential to cable networks is that new specifications for modular CMTS (M-CMTS) architectures and new services like T-1 or E-1 circuit emulation require extremely precise synchronization in cable networks.

With regard to M-CMTS architectures, the synchronization interface, named DOCSIS timing interface (DTI), ensures that the M-CMTS core, edge QAM modulator and upstream are synchronized to support the existing DOCSIS requirements for frequency and timestamps that existed in the traditional CMTS. In an M-CMTS architecture, shown in Figure 2, a cable modem receives its synchronization from the edge QAM modulator so that it is synchronized to other cable modems to properly transmit to the upstream burst receiver. Additionally, the M-CMTS core is synchronized to the edge QAM modulator to schedule, correct and insert MPEG timestamps for video.

Figure 2: M-CMTS system row

In an M-CMTS topology, a DTI server contains system intelligence; it also controls frequency and time for the headend or distribution hub. Each M-CMTS device (edge QAM modulator, upstream burst receiver and M-CMTS core) contains an integrated DTI client required by the specification. The DTI client is a low-cost digital transceiver of the DTI protocol consisting of a small field programmable gate array (FPGA), inexpensive oscillator and supporting circuitry. The DTI protocol generated at the DTI server replicates the precise time and frequency at each DTI client within 2 ns to support the existing ranging requirements. DTI has a robust feature set including automatic cable delay compensation, early fault detection, path traceability and all the DOCSIS 1.0/2.0/3.0 requirements. If desired, the application also enables time-of-day services, hitless protection switching, redundancy, enhanced management and Stratum 1 traceable commercial services timing. From a synchronization and timing perspective, the M-CMTS devices are on a common, virtual backplane, analogous to the integrated CMTS.

Timing tiers

There are three synchronization tiers in a cable telecommunications system - traceable network timing, M-CMTS and integrated CMTS. (See Figure 3.) The business motivation for multiple tiers of synchronization is to achieve economies of scale, conserve capacity and throughput, and offer new services that are competitive with those provided by the telephone companies, especially to business customers.

Figure 3: Synchronization tiers in a cable system

Historically, cable telecommunications has not synchronized its network timing to Universal Coordinated Universal Time (UTC) or any other standard time source, and thus has been operating at Tier 3. It uses a 10.24 MHz clock internal to an integrated CMTS for symbol generation in its modulation schemes. The DTI TimeStamp (DTS) is a 32-bit counter that increments every 10.24 MHz root clock cycle.

M-CMTS defines a mode of operation at Tier 2. It radically changes how cable systems are synchronized. With modular CMTS, timing information is exchanged between physically separate network elements. Timing information that used to be internally moved over a backplane must now be exchanged among separate equipment units. This information must be synchronized within 5 ns for exchanges between servers and clients in the same building.

Operation at the traceable network timing of Tier 1 synchronization becomes necessary when network elements are geographically separated, such as when the CMTS core is at a headend and QAM modulators are at a hub. Traceable network timing at each location is critical to synchronization within 100 ns of network element communications interfaces between locations.

The consequences of loss of synchronization are dire. Data throughput in high-speed data applications drops radically because of severe increases in packet errors. Worse still, voice quality is lost, and digital video service may not be available.

Having the architecture synchronized to UTC that is accessed through Global Positioning System (GPS) receivers provides the solution that makes precise synchronization possible. UTC is the international time standard that has been in effect since 1972. UTC is maintained by the Bureau International de l'Heure (BIH), which forms the basis of a coordinated dissemination of standard frequencies and time signals. UTC is the time and frequency standard that is the source for the Traceable Network Time shown in Figure 3. It is used for all of telecom. GPS (the U.S. military refers to it as NAVSTAR GPS) is a satellite navigation system used for determining one's precise location and providing a highly accurate time reference almost anywhere on Earth or in Earth orbit. The GPS satellites continuously transmit digital radio signals that contain data on the satellites' location and the exact time to the Earth-bound receivers. The satellites are equipped with atomic clocks that are precise to within a billionth of a second.

A DTI server has a GPS receiver that enables the M-CMTS architectures to stay synchronized on UTC and thereby avoid packet errors, poor voice quality and loss of video service. (For more background on UTC, see this month's Telephony column.)

DTI architecture

DTI is not only a protocol, but also an architecture of interconnected DTI servers and DTI clients. This dedicated synchronization distribution architecture ensures that all the clients and servers together maintain the time and frequency accuracy needed for M-CMTS and DOCSIS. The synchronization is required to ensure that the M-CMTS elements together perform as if they were a single shelf. Through this tight synchronization, both new and existing cable modems can be connected to an M-CMTS.

Initially a DTI network is rather simple to deploy. A DTI server is installed, and it contains the root DOCSIS frequency and timestamp. This first DTI server is called a root server and establishes the DOCSIS root frequency and time. Each headend or hub can only have one root DTI server. Any additional servers are slave servers. A slave server contains a DTI client and uses a DTI connection to the root DTI server to maintain synchronization.

A root server, slave server, and all client servers must together maintain 5 ns of synchronization. This 5 ns must be budgeted among root, slave and client. In order to keep client implementation low cost, the client receives 50 percent of the budget. Slave and root together must therefore be within the 2.5 ns remaining window, or 1.25 ns each, with respect to the root server oscillator.

In addition to the 5 ns requirement in a single location, there may also be a requirement to synchronize among multiple locations. This requirement may be attributed to commercial services or DOCSIS Path Verification (DPV). With GPS, DPV measurements can be made to measure the one-way delay across the Internet protocol (IP) network from the headend to the hubs or across the backbone accurate to 100 ns. GPS also provides Stratum 1 frequency traceability. The time budget for DTI is shown in Figure 4.

Figure 4: DTI time budget

The M-CMTS architecture not only provides the transmission architecture for DOCSIS broadband services; it is also the foundation for voice services and video on demand (VOD) through the converged edge QAM modulator. With a wide variety of services running over the M-CMTS architecture, great care must be taken to ensure that it is reliable. Without DTI, in normal operation the M-CMTS will not work. Deploying a reliable DTI architecture can be done economically and provide scalability and reliability.

Desirable traits

Choosing a scalable and reliable DTI server is critical. A DTI server should have the option for power and clock card redundancy to provide high availability, but also provide flexibility for low cost. A DTI server should also be able to support both root and slave configurations. This provides an operator flexibility to use the same product for both types of deployment and simplifies training, management and operations. A DTI server should also support GPS as a highly recommended option.

When a root DTI server is configured, the operator must choose how the DOCSIS timestamp is configured by setting the time-of-day mode. The DOCSIS timestamp is used by the M-CMTS elements and cable modems to communicate. Once the DOCSIS timestamp is configured, if it needs to be changed for any reason, it causes the M-CMTS elements and cable modems to reset. GPS mode is the preferred time-of-day setting. A properly designed DTI server may also have NTP server capability since it contains all of the complex synchronization components and in most cases is connected to GPS.

All M-CMTS elements should support two DTI client ports to ensure that a simple disconnection from the DTI server does not cause failure. As the need for DTI capacity increases, slave DTI servers need to be deployed to increase capacity. Figure 5 illustrates options for connecting root servers, slave servers and clients.

Figure 5: DTI connections

Best practice deployment is to install DTI servers with internal power and clock card redundancy; however, if lower capital cost is desired, a fully redundant root DTI server can be deployed, and when additional capacity is needed, a slave DTI server without redundancy can be used. If this is done, the M-CMTS element should connect one DTI client port to the root and one to the slave. This ensures path protection that keeps the M-CMTS elements in sync if the slave server fails.

Management of DTI is critical to ensuring proper operation and deployment. The DTI protocol ensures that the DTI server can monitor the performance of each DTI client and detect changes in DTI client performance. This information is available to report to a management system and can be used by the management system to reroute traffic to other M-CMTS elements before failure occurs.

Deployment options

Reliable and economic synchronization provides cable operators the calculated edge in reducing operating costs while increasing network scalability and flexibility for advanced, next-generation services including DOCSIS 3.0 and commercial services.

DTI lays a foundation for the existing and future network architectures to converge voice, data and video reliably and economically. The DTI server is a shared element among the M-CMTS devices making it economical, however potentially a single point of failure. Deployment schemes for redundancy should include protection of power and the active server elements in the DTI server. Moreover, path protection through dual links from the DTI client embedded in the M-CMTS devices to the DTI server help guard against inadvertent physical disconnection. Dual links can also be configured to originate from redundant DTI server output cards or redundant DTI servers for carrier class applications.

There are several architectural and operation considerations to deploying DTI. Each cable operators or region must decide how they want to support redundancy DTI, sparing, how they want to cable it in the headend/hub, what options they want, like GPS, and they must also plan for expanded capacity and how root/slave servers will be deployed. Lastly, they may also need to consider an overall migration plan for M-CMTS, which may include DTI and other synchronization techniques.

Jeremy Bennington is business development manager for Symmetricom. Reach him at jbennington@symmetricom.com.