Timing Technologies for Next-generation Wireless Networks

By: Ullas Kumar, Zarlink Semiconductor
( 1 Jul 2008 )

Wireless networks are becoming the predominant infrastructure for data and voice communication. Third generation (3G) wireless technologies and beyond aim to provide voice, data and video to end users. However, 2G wireless networks have been typically designed based on the voice network infrastructures, and are not optimized for data or video communication.

The growth of personal communication devices in recent years has stimulated the faster evolution of the packet-based wireless networks. As these networks evolve, there is an increasing need to provide accurate timing and frequency information over packet-based architectures. The ability to better support time-sensitive services over these new networks may spur the development of future applications.

Synchronization is a major requirement for the wireless network elements; both the radio interfaces as well as the network interfaces. Synchronization was inherent in circuit-switched voice networks, with accurate timing achieved through technologies like global positioning system (GPS) or long range navigation (LORAN) and frequency information delivered through the plesiochronous digital hierarchy (PDH) links connected to the base stations. With the migration towards packet-based transport technologies, these methods are no longer available.

Circuit switched networks traditionally work on Time Division Multiplexing technology and carry timing transparently though physical layers and follow well engineered synchronization schemes. Packet switched networks are engineered to be statistically multiplexed and any timing information embedded on the traffic is lost as soon as it enters the packet switched network.

A typical 2G or 3G mobile infrastructure hierarchically is seen in Figure 1. There are number of synchronization elements involved in this system, with this paper focusing on the challenges in the implementation of Node-B synchronization.

SYNCHRONIZATION CHALLENGES ON THE NODE-B
A typical Node-B architecture is illustrated in Figure 2. The Node-B should be able to meet the network interface-timing requirement, the radio interface timing requirement and inter node-timing requirements.

Network interfaces connect the Node-B to the radio network controller (RNC), typically over E1/T1 lines or SDH/SONET. G.703 defines 50 parts per million (50ppm) accuracy for E1 and G.823 defines the jitter and wander requirements of the interface. For SDH interfaces, G.813 specifies the performance characteristics of the slave equipments clocks (SEC).

Radio interfaces aim to achieve inter node synchronization by having a common time reference primarily relating to the timing of the radio frame transmission. In time division duplexing (TDD) based systems, such techniques reduce cross-interference and also allow simpler and efficient implementation of procedures like handover. For frequency division duplexing (FDD) based systems, synchronization ensures the end terminal receives frames synchronously from different cells, reducing the need for extra buffers.

Another important aspect is frequency synchronization. The frequency accuracy helps smooth handover of the end terminal from one cell to another. A moving mobile unit could experience Doppler Effect. To accommodate speeds over 100km/h, frequency accuracy of 50 parts per billion (50ppb) is defined in most cellular systems. In most applications, the cellular base stations need to extremely be accurate to ensure no disruption with the neighboring base stations. This is required to allow a minimum channel spacing to accommodate maximum channels.

Table 1 summarizes the air interface requirements of various wireless networks standards.

BASE STATION SYNCHRONIZATION OPTIONS
Global System for Mobile Communications (GSM) and Universal Mobile Telecommunication System (UMT) base stations relied on the recovered backhaul T1/E1 clock, traceable to a primary reference clock (PRC) and locked to an embedded oscillator, to provide the long-term accuracy requirement of 50ppb. As there was a reliable and accurate clock available, a lower cost solution could be implementation.

However, the advent of 3G systems and higher bandwidth requirements has made the T1/E1 connections obsolete. TDM-based T3/E3 infrastructures are expensive and less flexible for bandwidth provisioning, creating a natural migration to packet-based backhaul technologies. To realize transport cost-advantages, base stations that are still using T1/E1 backhaul interfaces are moving to circuit emulation technologies.

Code division multiple access (CDMA) based systems had GPS modules with clock extraction embedded into the systems because the phase alignment reference was needed for the handoff between cells. The highly accurate Stratum 1 traceable clocks provided clocks well suited for packet-based backhaul technologies. There are limitations to GPS as a synchronization solution. Some of these are country-specific political issues, while some in-building, underground and underwater applications can't rely on a global navigation satellite system. Technologies like Personal Handy-phone Systems (PHS) make use of the time information through GPS, but the frequency information is still derived through other timing distribution methods.

A third method used in North America is a radio-based terrestrial transmission system called LORAN. First deployed as a radio navigation system, the LORAN C and its latest derivatives provide extremely accurate time and frequency information and are projected as a complimentary system and a backup to the Global Navigation Satellite System (GNSS). However, coverage is limited primarily to North America.

Another option is to use rubidium-based oscillators that are free running with 1 e¨C09 accuracies over a 10-year lifespan. These oscillators are extremely expensive and are only appropriate for applications where frequency accuracy is a requirement. Such robust solutions are very convenient as that they don't require any maintenance or calibration over the entire life of the equipment.

There are currently two proposed methods for the synchronization of packet networks. In the first method a server sends timing information in the form of packets. Timing-over-Packet (ToP) technology refers to the mechanism of Clients interacting with the server on a defined protocol, extracting the information from the packets and process the information to generate a local clock.

The second method, called Synchronous Ethernet (SyncE), emulates circuit-switched synchronization techniques with transceivers in the packet network upgraded to support extraction and filtering of clocks.

TIMING OVER PACKET NETWORKS
Traditionally, packet-based networks are designed to operate in a statistical multiplexing fashion, compared to the time division multiplexing applied in the circuit-switched networks. When circuit-switched network services are mapped to packet-switched networks, the timing information is lost at the inter-working function. The nodes in the packet-switched networks are independently timed, and separate mechanisms for timing recovery need to be implemented to support time-sensitive applications.

To transmit timing in packet networks, a server sends timing information to one or more multiple clients in a unicast or multicast fashion. The server and client follow specific protocol to interact and the client recovers frequency and timing information based on contents and other information, such as packet arrival times. Examples of protocols that can be used to send timing information are Network Timing Protocol (NTP) and IEEE-1588 Precision Clock Synchronization Protocol for Networked Measurement and Control Systems (PTP).

Designed to synchronize data networks, NTP has accuracies in the order of 10s of milliseconds. It is conceived by IETF and currently undergoing its fourth revision.

IEEE 1588 was designed by the IEEE test and measurement community, primarily for industrial automation applications, and is currently being finalized on the second revision. The protocol contains information exchange part and the configuration control part. The inaccuracies due the nature of the various network elements can be compensated by the time-correction mechanisms.

ITU-T consented G.8261 "Timing and Synchronization Aspect in Packet Networks". It defines the minimum requirements for the synchronization function of the network elements and the performance requirements. It also provides an appendix of measurement guidelines, and test cases that provide useful guidance on synchronization performance over packet networks. IETF is also proposing another working group, Transmitting Timing over IP connections and Transfer of Clocks (TICTOC), with the objective to identify and satisfy time and frequency needs of various services that need to be supported over IP and MPLS-based networks.

EXTENDING TRADITIONAL SYNCHRONIZATION TO ETHERNET
The second method of transporting timing across packet networks, in particular Ethernet-based networks, is Synchronous Ethernet (SyncE). In this mechanism the usually free running Ethernet physical layer clocks are synchronized to a network clock. At the inter-working function, the primary reference clock (PRC) is used to drive the Ethernet PHY. The clock is extracted from the PHY layer at each packet-switching node, filtered and used for system and downstream clocking to extend the idea of synchronous networks to the Ethernet. ITU-T consented G.8262, which describes the timing characteristics of the Synchronous Ethernet Equipment Slave Clock (EEC). The EEC clocks are specified to perform in a manner consistent with existing synchronous equipment slave clock in the SDH domain.

Only frequency transfer is possible today with the intrinsic Synchronous Ethernet technique, but this can be done with a degree of accuracy comparable to circuit-switched networks. The time transfer may be done using additional packet-based time transfer methods. The combination of techniques would greatly improve the overall performance of solution.

G.8262 outlines minimum requirements for timing devices used in synchronizing network equipment supporting the Synchronous Ethernet architecture. The specification outlines requirements for clock accuracy, noise (Jitter and Wander - transfer, tolerance and generation) transient and holdover performances.

Similar to G.813 which describes SDH Equipment Clock, the Ethernet Equipment Clock also contains two options, namely EEC option 1, which applies to Synchronous Ethernet equipments that are designed to inter-work with networks optimized for the 2,048kbps hierarchy and EEC-Option 2, applies to Synchronous Ethernet equipments that are designed to inter-work with networks optimized for the 1,544kbps hierarchy.

SYNCHRONIZATION OF PACKET NETWORKS (MOBILE SYSTEMS)
As previously described, depending on the radio technology used, mobile base stations need time and frequency information from the network to operate reliably and efficiently. ToP technology provides the time and frequency features required by the base stations.

TOP CLIENT EXAMPLE
In architectures where all of the intermediate nodes support physical synchronization, SyncE becomes a robust way to distribute timing. With current techniques in SyncE, the limitation is that it can be used in applications where only the frequency accuracy is required for the operation of the wireless interface.

It is proposed to have a combination of SyncE and Timing over Packet technology to service applications that requires frequency accuracy and phase alignment.

SUMMARY
As mobile networks migrate to an IP-based infrastructure, synchronization is an increasing technical challenge. Depending on the radio technology used, mobile base stations need only frequency accuracy or accurate frequency and timing. Packet-based timing transfer mechanisms can be used to address the requirements of base station synchronization, depending on the radio technology requirements and architectures.

About the Author
Ullas Kumar is an Applications Engineering Manager ¨C Asia Pacific with Zarlink Semiconductor, http://timing.zarlink.com

Figure 1: Next-generation mobile infrastructure.
Figure 2: A typical Node-B architecture.
Figure 3: A typical server and client applications.
Figure 4: Timing-over-Packet server example.
Figure 5: A typical SyncE application example.
Table 1: Air interface requirements of various wireless networks standard.

Click here for the illustrations:

Figure 1, Figure 2, Figure 3, Figure 4, Figure 5, Table 1