Reducing churn is a key business priority for wireless network operators. It is well accepted that acquiring a new customer can cost significantly more than retaining an existing customer. Additionally, new number portability laws have made it easier for users to switch network providers when they have a dissatisfactory "quality of experience." As of late, wireless network operators have invested heavily to complete cellular network coverage to reduce dead spots. As a result, the competitive spotlight has shifted from comprehensive coverage to network stability as dropped calls remain one of the key reasons users cite when changing service providers.

Calls are dropped as a result of poor synchronization between base stations when handing off calls. Under the influence of various timing transients, a base station may be forced to adjust frequency and timeslot parameters multiple times. To facilitate call hand-off and maximize efficient bandwidth usage, precision timing/synchronization is required: GSM networks require that base station frequency synchronization accuracy be better than 50 parts per billion (ppb).

Backhaul timing

In an ideal world, the most cost-effective means for synchronizing base stations would be the transport of timing signals over the backhaul link. This backhaul link transports all calls to the PSTN and is increasingly being used for high-bandwidth data services. In the early days, the backhaul transport was comprised of a tightly controlled network that provided reliable transport timing signals from a stratum one primary reference source (PRS) (see Figure 1). Each node between the base station and PRS would lock and track the clock signal, maintaining its accuracy. Of course, jitter and wander would be added in transit, but these would be filtered out at each transit point.

Such a network provides the accurate and stable frequency reference required to ensure successful call hand-off. This reference is used to calibrate or "discipline" a precision quartz oscillator that provides internal timing signals and RF carrier reference for the base station. If for some reason the timing signal is lost, quartz oscillators offer good short-term stability but will eventually drift out of spec in the long-term, on the order of six to 12 months. Locking the clock to the network timing feed keeps the quartz oscillator calibrated. However, if the timing signal is degraded before it is received, then the base station will recalibrate its internal quartz oscillator inaccurately. Thus, any network degradation will directly and immediately manifest itself as a reduction in synchronization precision.

With the introduction of high-speed data services, operators are facing a need for increased backhaul capacity. Many operators are planning to transition to alternative backhaul mechanisms to reduce costs. However, timing signals must negotiate a complex maze of backhaul providers employing varied transport topologies that degrade the reliability of network timing feeds significantly:

• Circuit emulation services, such as encapsulating a T1/E1 line over ATM, introduce delay variations that can cause synchronization issues at the base station. These links are very sensitive to traffic loading, and as traffic is dynamic, timing flows may experience delay variations in outgoing queues that cannot be filtered appropriately
• TDM networks degrade timing signals because of ambiguities in SONET pointer adjustments that introduce nondeterministic wander and jitter
• IP and Ethernet networks cannot pass synchronization feeds because of a lack of ability to control the path the timing signal takes through the network and/or time-delay variations greater than in ATM networks

Over the long term, as low-cost, high-capacity IP and Ethernet backhaul pipes become more prevalent; there will be no reliable way to transport timing signals over the network. In other words, these low-cost backhaul networks will eventually isolate a base station from its source of synchronization.

As the quality of the backhaul network is entirely out of the control of wireless providers, it is becoming increasingly more difficult to acquire accurate and stable synchronization over the network. If wireless providers want to assure high quality of service and reduce their dropped-call rates, they will need to take direct control of synchronization at the base stations.

Retiming

To maintain precise synchronization over an unreliable network backhaul requires a mechanism for recovering the stability of the network timing signal. Because the effects of jitter and wander are transient in nature, timing signal instability is itself transient and comes and goes depending upon the state of the network.

One approach for recovering the stability of the network timing signal is through the use of a retimer. A retimer removes instability introduced to the timing signal by buffering incoming traffic and clocking it back out with precise timing by locking it to a clock with PRS accuracy (see Figure 2). This means that the retimer itself must have access to an accurate timing or frequency signal.

Acquiring an accurate clock signal for the retimer requires overcoming two obstacles: the remote nature of base stations in general and the fact that they already have a method in place for deriving the clock signals. Note that if it were easy to run a stable and accurate clock signal to the retimer, this same clock signal could be used to clock the base station in place of the unstable timing feed.

A retimer that acquires its timing from GPS addresses both of these obstacles elegantly. Additionally, a GPS-based retimer can be transparently introduced to an existing base station. In this case, transparently means that the retimer is an independent unit and the base station is unaware of its presence. The retimer is placed on the backhaul feed logically before it arrives at the base station. As can be seen from Figure 2, the timing signal the base station now receives is precise and stable, enabling accurate synchronization between base stations.

In the case that GPS is not available, the retimer implements a cut through to preserve communications. In this way, the retimer is not a point of failure; the base station reverts back to the original backhaul timing scheme. The incidence of dropped calls will revert to its original rate as well, but operating conditions will be no worse off than they were without the retimer.

The introduction of retimers to base stations results in a dramatic reduction in dropped calls. A recent five-cell field trial conducted in September 2004 with a major GSM operator compared dropped call rates when the backhaul synchronization signal was retimed and resulted in a 25.5% reduction in dropped calls (see Table 1). The results are based on one full week of measurement before installation of the retimer and for two full weeks after installation.

It is important to note that the trial sites had been serviced prior to the field trial and that the backhaul links had been verified as being within acceptable operating parameters by the network operator. In other words, the comparison was made over links that already had good transmission quality. If the links were unstable, the measured improvement would be even more substantial. Note that precise synchronization does not completely eliminate dropped calls; rather, it eliminates timing feed instability, which is one of the primary sources of dropped calls.

A second field trial based on a detailed study on synchronization of UMTS Node B base stations conducted by a multinational GSM/UMTS operator monitored dropped call performance under various operating scenarios. Again, the reduction in dropped calls was tremendous (see Table 2).

It is worth noting that these measurements include call hand-offs with base stations not using retimed synchronization feeds. While there was still significant improvement in these cases, the highest improvement was seen when both base stations were retimed. As a consequence, the operator recommended adding GPS-based synchronization to all base station sites that were synchronized from traditional E1 leased lines.

Deployment

Work is in process to create mechanisms for reliably transporting synchronization feeds over IP networks from numerous industry organizations such as the IEEE, IETF, ITU, and ATIS. However, while these technologies are promising, they are unproven and introduce dependencies upon new master clocks in the network as well as network engineering constraints.

The advantage of retiming is that it is proven, effective, and available today. GPS-based retimers are relatively cost effective, simple to install, and require little ongoing maintenance. Additionally, because retimers can be easily retrofitted to existing base stations, wireless network operators are able to roll them out to troubled areas first to proactively improve quality of service in those areas where customers are at the highest risk of changing providers. Because operators already track the number of dropped calls in an area, prioritizing roll out is a straightforward process. Measuring results is a simple matter of comparing before and after dropped call rates.

Wireless network operators can no longer take synchronization reliability and stability of their base stations for granted. As the pressure to reduce costs leads operators to lower cost backhaul alternatives, synchronization of the base stations will become compromised. By retiming the backhaul feed, operators can cost-effectively and reliably maintain synchronization, reducing dropped calls and retaining customers.

Visit Symmetricom online.