Snagging Statistics

RF engineering is about more than blasting a signal to all points of the compass. So it's no surprise that RF analysis is about more than looking at a couple of parameters and pronouncing the network good or bad.

Case in point: Switch statistics are a good source of data for benchmarking and catching problems such as a site that's dropping a lot of calls. But in many cases, they don't provide information specific enough to pinpoint the cause. Switch statistics usually provide the what, while drive-testing provides the why.

"It's conceivable that you could rectify the problem without going out there," said Jay Noceto, Cellular One - San Francisco RF-engineering director. "You might be able to look at other information and determine that you've got some bad radios or an interference problem. But if someone is complaining about a stretch of freeway or a secondary-road intersection, that would be tougher to do just off of switch statistics."

C/I and bit error rate (BER) show analog and digital voice quality, respectively. Better than 17dB C/I is an industry rule of thumb; for BER, it's better than 3%. Although received-signal strength (RSS), C/I and BER usually give a good indication of what subscribers are experiencing, they don't provide a complete picture. Just because a signal is so solid that subscribers could walk on it doesn't mean that call quality is great.

"Interference is sometimes difficult to catch," said Dean DiSalvo, Safco Technologies senior RF engineer. "C/I isn't always easy to go out and measure. You need to drive-test, place calls and see what the system is doing."

That's partly because switches don't log all events that a subscriber might experience. Suppose that two subscribers are in an area plagued by uplink co-channel interference, and one of them hits "end."

"The other cell site can intercept the end signal and terminate the other user's call," said Greg d'Entremont, NBTel Mobility technical associate, infrastructure planning, systems-engineering department. "In both instances, the switch will believe that both calls were intentionally terminated when in fact one user actually had a dropped call."

WHAT DOES IT REALLY MEAN? RF-related parameters alone don't tell the whole story. If subscribers complain about call quality in an area, but BER and RSS are good, electronic mean-opinion scoring might identify the cause as a bad echo canceller or a bad T1 line between the site and the switch.

RF parameters themselves can be misleading. Suppose that a site suffers from a lot of dropped calls, but drive-testing shows good signal strength. Interference is the most likely culprit, but you'd have to look at BER to narrow down the potential causes.

"An RSS measurement without C/I would be pretty misleading," DiSalvo said. "For example, in an AMPS system, if I measure a nice, hot signal from a certain channel, I could still have really poor quality just because my C/I isn't very good."

If the interference is strong enough, the call can drop, which might make poor coverage seem like the real culprit. But in AMPS and TDMA, the desired signal needs to be several decibels stronger than other co-channel signals.

"If several signals on the same frequency arrive at a mobile at the same signal level, the FM receiver doesn't have the ability to demodulate the intended signal," d'Entremont said. "However, if there is a difference in signal strength, the stronger signal will be demodulated even if the other signal on the same frequency is of a usable strength. If the mobile's demodulator runs free and locks onto the wrong site's signal, the SAT (supervisory audio tone) signal presented back to the network will be from a wrong site, and the switch will terminate the call."

The solution is to optimize the network so that each area has just one dominant signal. But on co-located sites, that's not always easy. An 800MHz wireless carrier, for example, can experience a second harmonic from a 400MHz paging carrier. If the paging carrier is running high power, the view from a spectrum analyzer can be misleading.

"The technician looks at it, sees a strong signal and says, 'I've got a good signal here, so I guess I don't have a problem,'" said Yuenie Lau, Anritsu field-solutions business-unit marketing manager. "What he's seeing is the harmonic of a pager. The problem is, (you need) a trained technician. To see the spectrum, he may have to spread the signal out and see if it's an FM-type signal."

Multipath also can create false readings on a spectrum analyzer. In the right environment, as little as 100 yards can make a significant difference between readings.

"It depends on the foliage and the buildings," Lau said. "The reflection can cause multipaths and give you a false reading."

BIGGER & BETTER PICTURES As the airwaves become more crowded and the technologies more complex, it's handy to have T&M tools that make troubleshooting less of a chore. Mapping tools such as MapInfo's deciBel Planner provide the ability to overlay trouble reports and parameters such as BER and signal strength onto a map. That makes it easy to identify problem areas in a glance and helps non-technical personnel digest complex information.

"It sure beats looking at a spreadsheet or a graph plot," said Steve Fortin, MapInfo telecommunications consultant. "They can see exactly where their problem spots are and pass those to their customer service and salespeople to raise the awareness of service quality throughout the organization."

One area for improvement is in-building coverage. Today, subscriber input is key to determining whether a call suffered indoors or out. When subscribers report problems, CSRs should ask whether they occurred inside a building and, if so, which building. That can save time and legwork tracking down the source of a problem.

"Unless I have a specific request from a customer or it's a public facility, (such) as a convention hall, I can't differentiate between in-building and outdoor problems off of the switch," said Cellular One's Noceto. "The in-building issues tend to come more directly through (our) sales force. They'll get a telecom manager or a fairly large user base, and they're having problems in building. That's generally how I get that kind of input."

Perhaps the biggest asset for optimizing both indoor and outdoor coverage will be E-911 Phase II, where handsets could report their locations sothat carriers would know within 125 meters where problems occurred. In effect, subscribers would continually drive-test the network.

"That would be fantastic to be able to aggregate (information) to see that we had X number of dropped calls on this cell site, but the bulk of them were within this 1/8-mile radius," Noceto said. "That's something that I'd love to see developed."

Given the amount of additional traffic on the SS7 network that approach might create, it's doubtful that all subscribers would be doing drive testing all the time. One possible option is to collect that information and then upload it during low-traffic periods.

"I think it probably would be most germane to get information related to performance issues," Noceto said. "So you might not trigger a location request unless the call was seeing, say, a class 5 or higher BER. If there was some way to get that information just prior to a dropped call, I think you would be selective about how much of this data you got."