A Question of Quality
By Jeff Atkins, Actix
The objectives and challenges involved in setting about optimizing network performance, maximizing quality of service (QoS) and ensuring that capacity and usage move forward in step with each other are, at first glance, no different in the CDMA world than in GSM, or for that matter any of the older analog standards.
In fact the objectives-defined by business reality itself-are no different. They are, first, to produce a service that exceeds those of competitors in coverage, capacity, voice quality and all elements of mean opinion scoring; secondly, to do so at a cost that does not sacrifice business viability for the sake of technical perfection; and thirdly to maintain a close watch on what the competition is up to in these areas. But peer more deeply into the challenges, and a new set of objectives starts to emerge. What, then are the QoS challenges?
The permanent state of capacity flux comes about because, as each additional user joins the network to participate in a call, so total coverage declines. Quite simply, each new user introduces new interference and conversely, when there are correspondingly few callers on the network, or discrete part of the network, there is relatively little interference.
Testing and measuring-via drive tests for example-needs to occur at the busy hours in order to gain an accurate, or at least a useful, picture of QoS at the times when it might be most under threat. A drive-round survey at dead of night will yield different results from those taken by day, in a way that is less true of a GSM network.
This goes back to the fundamental nature of CDMA signaling itself, as summarized best by Roke Manor Research's WHW Tuttlebee. Correlating the received broadband signal with an identical pseudorandom sequence demodulates the signal, he points out. This collapses the signal back to its original bandwidth and also spreads any narrowband radio signals present within the occupied spectrum so that they now appear noise-like to the receiver. By using many different pseudorandom code sequences, multiple users may be accommodated within the same spectrum.
The key issue, which for many years had been thought to preclude the use of CDMA for many applications, was the so-called "near-far problem". If a narrowband interfering signal falls within the spread bandwidth and its amplitude is significantly greater than the wanted (spread) signal, then the interferer would still remain sufficiently strong after despreading to prevent correct demodulation of the wanted signal. For CDMA to operate successfully as a multiple-access technique, so that many subscribers can access the network simultaneously, all the accessing signals must be received at a similar amplitude.
The solution to this problem has been the application of power control, whereby the cellular base station controls, in real time, the transmission power of the users' mobile terminals. In GSM, power control is a feature that has been added to increase system capacity by reducing average interference levels; for CDMA, however, power control is essential to the successful operation of the basic concept.
Rewriting the rules
The impact of these mechanisms is to increase the system capacity for a given infrastructure density, but at the expense of base station and infrastructure complexity and at the risk of greater interference.
There are solutions to this problem, and, like many solutions to CDMA QoS problems, they lie in adjustments to the RF and infrastructure configuration of the network. But to know how, when, where and by how much these adjustments must be made depends on measurement and analysis of complex data.
In a typical CDMA network, the test data that are used to analyze QoS come from two sources-the switches themselves, providing high-level statistical network performance data at a sector level resolution; and air interface data. The challenge many operators face is in translating the high-level problems shown by the switch data into actions designed to solve these problems.
Detailed knowledge of where RF and infrastructure configuration problems occur is an essential part of the translation process, whereby operators diagnose and troubleshoot the causes of high-level problems to maintain a high QoS.
This is where drive test and call trace data become extremely important. Drive testing of the network with test mobiles allows engineers to collect air interface and related data along with GPS information. By analyzing the collected data, engineers can determine where fundamental CDMA problems such as pilot pollution, poor coverage, and improperly configured neighbor lists are occurring. By virtue of having detailed position information, engineers can determine solutions to these problems.
sSince the limiting link in CDMA is not consistently one link or the other (although it may be one more than the other), it is extremely important to monitor drive tests from the forward link perspective as well as from the reverse link perspective. Figure 1 (above) shows a chart of reverse FER and forward FER binned into 1000 ms intervals. Notice that the Forward FER (blue trace) and Reverse FER (green trace) are not well correlated.
Drive tests may typically be monitored from the reverse link perspective using proprietary call trace measurement programs which run on top of the various infrastructure providers' equipment. Call trace measurement programs can collect data for test calls by a specific test mobile including reverse link Eb/No, forward link digital traffic channel gain, and air-interface messaging as recorded by the base station. None of these uplink data are available from the test mobile data.
It bears repeating that geographic references for collected data are essential for the effective use of this data to resolve problems. This can be a problem with call trace data, because GPS-type position information is not typically available in switch call trace logs.
To solve this problem, CDMA engineers can use tools which synchronize the data contained in call trace log files with the position information in test mobile log files, thereby allowing engineers to visualize uplink data geographically.
The synchronization of the call trace and test mobile logs is often complicated by timing errors between the time stamps of the respective log files. Figure 2 (below) is a plot of the primary serving PN as measured by the switch call trace and the test mobile. Notice that without synchronization, the two plots are offset by almost two minutes.
Tools which allow synchronization of the call trace and test mobile logs must provide automated correlation algorithms that minimize the timing error for mapping of reverse link data to be easy enough to be feasible for everyday use.
The screen shot in figure 3 shows a sample Eb/No data trail plotted on a map after synchronization with a test mobile data set has been performed.
One further QoS-impacting measurement phenomenon which is unique to the CDMA environment is pseudo noise and the place of the PN Scanner. In CDMA, each base station broadcasts the same digital PN sequence and the mobile handset uses that to synchronize to the system. In GSM and TDMA systems, it is fundamentally difficult to make large-area measurements of desired signals and co-channel interferers simultaneously without making time-consuming, temporary modifications to frequency plans to create special test frequencies.
PN Scanner technology is revolutionary in that it allows the user to quickly measure all PN sequences, thereby measuring all desired and interfering signals, despite the fact that they are co-channel. This makes PN scanners particularly useful for showing where action is needed to mitigate the effects of interference and improve coverage. By mitigating interference and improving coverage, the customer-perceived QoS is directly improved.
Another QoS challenge to CDMA service providers has been the operation of systems which utilize multiple vendors for both infrastructure and drive test tools. As a result, CDMA service providers have looked to data analysis tools that integrate call trace and drive test data from multiple vendors in one platform. By doing so, operators have been able to save money by purchasing one platform which integrates with all data sources rather than one platform for each data source. In addition, this approach allows operators to reduce training time and to correlate previously disjointed data sets in one platform for more effective decision making. The end result is the ability to troubleshoot and optimize the network performance more efficiently and at a lower cost.
Further complicating the multiple vendor issue has been the lack of an open standard for protocol interfaces between the various network elements of the CDMA system (i.e. BTS, BSC, MSC).
In GSM, most equipment providers adhere to an open protocol interface which has resulted in the widespread availability of protocol analyzers capable of interfacing with the A and Abis links. Because of the lack of this type of protocol analysis equipment for CDMA, CDMA service providers are even more reliant on call trace and other proprietary switch data sets for troubleshooting of performance problems.
The complex nature of the CDMA air-interface, the presence of multiple vendors, and the lack of open standards have made provision of a high QoS in CDMA networks a major challenge. Effective use of drive test equipment and data analysis software which provides support for call trace and PN scanner data sets has been a primary tool for operators to meet these challenges.
Another key factor has been the use of data analysis software which supports multiple vendors' and multiple infrastructure providers' data formats. Looking forward to 3G and the new millennium, the continued evolution of the CDMA air-interface and a move towards more open standards in CDMA should be a major focus. This, in turn, will help CDMA operators towards truly effective troubleshooting and optimization in their networks, and enable them to offer the highest QoS possible.