Field Trial & Error

Sprint PCS' CTO takes on 3G, evolving what is known and preparing for the unknown.

Sprint PCS began conducting pre-commercial 3G field testing more than a year ago. Part of the ongoing trial is to make sure 1X equipment works with the existing network architecture. Part of it is to test hardware and software in a pre-commercial state. Part of it is to simulate how the network will perform with customers.

The goal is to be prepared for every eventuality before going commercial. However, how do you prepare if there are no precedents, no models and no predictability in equipment, performance or customer demands?

That's Oliver Valente's job as Sprint PCS CTO: Make 3G work in a nationwide network, come what may. And that may be a gradual ramp-up of customers, or it could be a flood. It may be a doubling of monthly minutes of network use from today's multibillion to two times that, three times that … who knows? It may be a lot of things, none of which is certain until you are in the throes of it.

First Steps

At its Integration Lab in Lenexa, KS, Sprint PCS launched its first cdma2000 1X trial efforts with Samsung Telecommunications America (www.samsungusa.com), 3Com (www.3com.com) and Qualcomm (www.qualcomm.com). The trial tested the integration of voice and data in one cell at the same radio frequency. The carrier wanted to know if it could achieve 144kb/s with 35 full-rate voice calls at the same time.

The system included three fully integrated ATM-ready base stations, a BSC, base-station management and an MSC.

In a cdma2000 1X network, the BSC connects directly via a radio-port interface to a packet-data serving node (PDSN). The PDSN performs the traffic-integration function similar to that handled by circuit switches. It uses authentication, accounting and administration servers for user authentication and traffic management, and then forwards traffic to the Internet.

3Com provided the PDSN, authentication/billing servers and network-management system. Samsung provided handsets, as well as V.4 technology for doubling voice capacity. Qualcomm provided handsets, as well as the MSM5000 chipset and system software.

After the first phase of testing, Sprint PCS brought in Lucent (www.lucent.com), Motorola (www.motorola.com) and Nortel (www.nortel.com) to conduct additional testing. So far, it's all pre-commercial products. Vendors are expected to deliver commercial hardware and software later this year. If all goes according to schedule, Sprint PCS will start rolling out 3G 1X later this summer and have nationwide coverage rolled out by June 2002. (See Figure 1 on page 16.)

Through the testing, Valente said Sprint PCS wanted to verify the three main benefits that it needs from 3G 1X: improvements in voice capacity, data rates and battery performance.

According to Valente, “The testing is coming out as expected.”

In the pre-commercial phase, voice capacity increased 1.8 to two times. The data rates reached peak performance of 144kb/s consistently, whether close to a cell site or far away.

As for battery performance, Sprint expected to reach a 50% improvement. However, the tests revealed 30% to 50% improvements.

“We just haven't had the devices to test en masse yet,” Valente said. “However, between now and 2002, there'll be more 1X devices to test the (battery) performance.”

This performance improvement will help when color screens begin populating the 1X landscape. The laws of physics predict that color screens will draw more battery power, but Valente hopes the improvements will offset any customer perceptions of drain.

Voice Capacity

The major limitation of lab tests is that they aren't commercial. But Sprint PCS has pushed the network hard in the pre-commercial phase. Among the many tests it has conducted, it has simulated multiple calls on the system and added interference to simulate what you would see as capacity increases.

So how far can you push a system? Valente said that, unfortunately, you can't simulate millions of calls or stress the system as much as you would like. He added, however, that there are other ways to stress it.

“Although we can't do capacity, we do lots of what we call destructive testing,” he said. “In the middle of running a series of test calls, we introduce an error into the system and see how it reacts and see if it's able to recover.”

To test the system from a capacity standpoint, Valente said the company has to rely on commercial field trials.

“We do testing in our integration center. Then we do a first-market application, where we take it out to a field location and run virtually the same set of tests we ran in the lab. The difference is you have commercial traffic on it, and you're really able to stress it out from a load perspective. Then we make sure it works well in the field trial. Once that's done, we roll it out nationwide.”

Were there any surprises? Valente explained that Sprint PCS and its vendors went through learning curves together and experienced the normal growing pains you would expect to see in pre-commercial testing.

“The good side of that is they got worked out,” he said. “It has caused us to find things we could work out sooner rather than later. You really want to make sure you have found the problems before it gets near a customer.”

Blink Speeds

Later this year, when Sprint PCS rolls out 1X in the initial markets, which the company hasn't yet released, it will offer data rates of 144kb/s. So what does Valente think is more important — the data rate or the performance?

“You have to be careful on the whole data rate because you don't want to overhype the capability,” he said. “When we've talked about the data speeds, we've always said ‘up to’ 144kb/s. When you're into 1X, say it's ‘up to’ 307kb/s; 1X-EV-DO gets you ‘up to’ 2.4Mb/s, then beyond that, 3Mb/s to 5Mb/s. Those are peak data rates.”

According to Valente, average data speeds will be more important to consumers.

“What we think that'll be and how we're designing our network is under a loaded condition to be in the 70kb/s range,” he said.

At 70kb/s, will consumers be able to see a difference or have the perception of it being slow or fast?

“At 20kb/s to 30kb/s, e-mail works extremely well. Beyond that, you aren't really noticing what we would call a blink-speed difference,” he said. “If you're faster than 30kb/s, based on the application, you aren't really going to notice any improvement beyond 30kb/s.”

Valente said the company is applying that same logic to all of the applications — e-mail, Web browsing, audio streaming and video streaming.

“We're trying to get to the application speed where, beyond a certain point, the user wouldn't have a perceptibly different experience.”

The New Call Model

Beyond data speeds, Valente has had his hands full trying to apply reality to a strictly theoretical model.

“In the voice-call model, it's well-known from historical trends how much traffic goes on during the busy hour, what percent of the total traffic occurs during the busy hour, average hold times, busy-hour call attempts,” he said. “Those are all things that all of us operators have a good handle on.”

However, when it comes to mobile wireless data, there's a whole new set of variables at play.

“We're trying to do our best to understand what we think will happen there, but until we are launched and running, it'll be just that,” he added.

Valente examined wired Internet usage to see how people use it in terms of average session length. The average wireline Internet session is about 30 minutes.

According to Valente, the activity factor is an important variable in the data call model.

“If you think about how you browse the Web, and you're clicking on pages and spending quite a bit of time just reading what has popped up in front of you, you're not actually sending and receiving data. What the activity factor gets to is just that: For a given session, how many minutes are consumed sending and receiving data.”

Sprint PCS is developing what it thinks the activity factor will be for all of its applications and segments of traffic. Valente sees three distinct segments, each dictating different activity factors: Web browsing from your handset; Web browsing from a laptop (connectivity) using a data card and wireless phone; and telemetry services.

“When we look at those sorts of things, connectivity is the one we think will consume the most amount of bandwidth,” Valente said. “We made projections about what percent of our total data traffic is going to be doing a lot of these different things and then tried to design our network around that.”

Added in with the activity factor is time of day. In the voice-call model, peak and non-peak hours reflect rush hour and windshield time. And looking at the wired Internet model, Valente found Tuesday nights, Friday nights, all day Saturday and Sunday are big wireline Internet sessions. He expects different key days of the week and different kinds of usage with wireless data.

“The good thing is that right now we aren't anticipating it necessarily coinciding with the voice busy hour, which I think is a good complement to the network.”

Spectrum Stretching

At Wireless 2001, Charles Levine, Sprint PCS president & COO, said the wireless company had enough spectrum to last for 10 years. How can he be so sure? What if other carriers dropped off the face of the earth and everyone was coming to Sprint PCS for service? Valente backs up his boss but provides a more convincing testament.

Besides Sprint's having 100% digital in the same frequency, Valente credits CDMA for its efficiency and ability to add incremental improvements.

“Most of our competitors are encumbered by analog and digital technology. Some of them have less efficient digital technology than CDMA. They have multiple frequencies. Those are all challenges that we don't have.”

In addition, Valente expects vocoder advancements, 1X and smart antennas to deliver additional increases to maximize its current spectrum.

“With EVRC, you get 40% to 60% improvement in capacity. You're able to effectively carry more voice traffic with the same quality of service over less bandwidth. We're already deploying that in our markets today.”

“With 1X, you get another (potential) doubling effect,” Valente said.

That doubling assumes that there's 100% penetration of 1X devices. If on average, customers keep their devices for about 18 months, the coupling effect could be imminent.

Valente also identified select-mode vocoders (SMV), which are similar to EVRC.

“In the next two years, you'll see SMV vocoders introduced into the CDMA world, which will improve capacity potentially another 30% by taking the vocoder rate down to around 5kb/s to 6kb/s. That will be another voice-capacity improver.” (See Figure 2.)

In the next two to three years, he also sees smart antennas providing two to three times the capacity that is available today.

“All of these are additive,” he said. When you add all of these other capacity enhancements, you can offer improvements 20 times TDMA and GSM and 35 to 40 times analog capacity, Valente said.

With this type of capacity increase, Sprint PCS is looking to invest roughly $2 billion in its 4-stage upgrade.

“That would be getting us through 1X-EV-DV in 2003,” Valente confirmed. “That's probably about a 2- to 3-year forecast in terms of spending.”

Valente thinks it'll cost roughly $800 million to roll out the first stage later this year.

Invariably Variable

For someone who faces so many variables in the face of launching 3G services, Valente ironically seems quite calm. So what keeps him up at night? Without missing a beat, he said his children do, especially a daughter who was born during Wireless 2001.

Turning serious, he keys on making sure the vendor partners meet their commitments and schedules.

“It's a painstaking process of continual review of key milestones that are going to get you to the endpoint,” he said.

Perhaps the continuous review is what helps in a world so filled with variables. But Valente is philosophical about making the unknown known.

“We just need to make sure that to the extent problems come up, we're able to find them early and get a game plan in place to solve them.”