Switching Directions

Out with the old, in with the new. On the surface, this saying seems like an exciting prospect: the chance to bring in new advanced services for subscribers. But the saying also brings its share of skepticism and questions.

Suddenly wireless carriers must reinvent themselves as merchants of bandwidth and providers of everything from video confer-encing to vehicle tracking. Just now settling into the role of established public utilities, they are thrust into the untrammeled marketplace of mobile computing and confronted with the uncertainties of predicting public acceptance of new services.

Actually, third-generation proposals are manifestations of a larger upheaval taking place throughout public networks. According to some studies, text and multimedia data account for more than 50% of telephone traffic in the United States. The Yankee Group predicted that percentage will grow to 98% within a few years. Soon, telcos will be promoting high-speed data services aggressively and ultimately will have the capability to transmit high-quality, full-motion video-on-demand.

Third generation also will expand and develop services such as roaming, 1-number portability and radio location. It not only will enhance network capacity to make high-speed data possible, but also accommodate voice traffic on an unprecedented scale. According to its prophets, third-generation technology will enable wireless to achieve the long-awaited parity with wireline.

"We believe wireless is coming into the mainstream," said Bob Salinger, director of marketing and business development for Lucent Technologies. "We'll see prices of under a dime a minute fairly soon. At that point, data use will start to grow."

Although there are skeptics, the technology to achieve third-generation goals is within reach, according to many industry representatives. But, it is still in the planning stages. Standards have yet to be hammered out, and technology has yet to be tested. Nevertheless, the general direction in which wireless telephony is proceeding is clear.

Data is strongly emphasized and not consigned to an ancillary subnetwork as in current CDPD within the AMPS standard, nor limited to short messaging as in second-generation digital systems. Overall throughput per allotted spectrum will be augmented substantially. Current distinctions between mobile service and wireless local loop will be erased. Third-generation wireless telephony aims to offer full mobility with the capacity to function as a primary phone service.

This is where the infrastructure segment of the industry is tending, but how will such changes manifest themselves within the wireless realm? More importantly, what do those changes mean to your infrastructure, central office and specifically the switching platform?

"The first thing you're going to see in regard to the switch is more functions migrating off of it," said Jim Olson, vice president of marketing for DSC's Intelligent Network Division. "Already, second-generation switches are running out of subscriber capacity, and if third-generation switches are to handle still greater capacity, then they can't be burdened with new functions. You'll see new functions and services moving to auxiliary or adjunct elements."

Keith Shank, Ericsson's director of product management for IS-136, predicted that the industry will phase out switching as we know it today.

"The trend is already under way as we move toward wireless intelligent networks, and it will continue through the third generation," Shank said. "Switches won't go away overnight, but new functions will be implemented off the switch because it's really the only way to preserve any backward comparability with the existing infrastructure. Eventually, though, when the transformation is complete, that older infrastructure won't be needed."

Dave Robertson, assistant vice president of Wireless Strategic Product Development for Nortel, agreed.

"In third generation, you won't be able to tell what's the switch and what's the network," he predicted.

But what precisely does such a switchless or off-switch network entail in terms of equipment and software? How much will it cost?

The specifics of the transformation still must be worked out, which means it's too early to predict costs. But Olson pointed out that it's always dangerous for a manufacturer to require that carriers make a forklift replacement to acquire a new technology.

Although exact cost paramters are not yet known, you will have to consider some refurbishment as you move toward third-generation systems. In order to handle high-speed data (the 64kb/s up to perhaps 2Mb/s already mentioned in standards discussions), a network will have to provide for packetization of data and voice. It also will have to allow for bandwidth on demand to individual users of the network. Such requirements have been stated clearly in position papers that vendors such as Ericsson and Motorola have produced.

The packetization issue has implications far beyond the wireless networks. The current organization of networks represents inefficient network use and must be abandoned gradually for data traffic to increase. Because switches and base stations are intended to switch between channels rather than routing packets, existing infrastructure is fundamentally unsuitable there.

Although packetization already is enshrined in the CDPD standard, it applies only to text data. In third-generation systems, it will apply to voice and every sort of content.

Such a change will have a major impact on the infrastructure and the switch and ultimately might lead to the elimination of the switch proper. In the latter case, you no longer will need the switch because the network itself won't be switching amongst channels and establishing connections. Because the transmission of packets is "connectionless," with individual packets following different paths to arrive at the same destination, a network router is needed to manage packet traffic through a network.

"You definitely need separate routers on top of the switch," Nortel's Robertson said. "You don't want to twist a narrowband switch into those directions because it has to match SS7 service requirements."

Routers and servers performing general network computing functions already are playing an increasing role in wireline voice telephony, in long-distance connections and in private internal telephone systems. But it has not yet reached the local exchange. SS7 protocols operating over Class 5 switches still manage most of the nation's phone traffic, but the migration trend toward more services and calling features to off-switch adjuncts using general computing platforms is clear, and is not without parallels in the wireless realm. Billing and home- location registers have long resided on general-purpose computers, and more recently intelligent network functions also are administered off the switch. So, the move toward packetization is advancing on all fronts, and if it is to advance any further, you will face major cost implications.

The next requirement for high-speed data and multimedia is flexible allocation of bandwidth. It will require numerous changes in the physical plant and in the software it uses.

Bandwidth on demand is necessary to achieve throughputs exceeding the T1 range -- that is, exceeding 1Mb/s -- within the narrow slices of spectrum assigned to cellular and PCS, especially when those slices are further subdivided into voice channels. With high-speed data, this flexibility is a necessity. You can achieve temporary expanded bandwidth in a number of ways, depending upon air interface.

Channel aggregation allows for increased bandwidth but may not be spectral efficient. The advantage is that it represents an incremental upgrade.

Simple channel aggregation represents an easy migration path for GSM and digital AMPS carriers because the core standards-based switching function only needsto be modified, not supplanted. Still, simple channel aggregation has to be considered transitional because of its effect on capacity.

Unfortunately, other more efficient and sophisticated methods only can be achieved with spread-spectrum air interface technology. This may be an area of concern if you are committed to time- division interfaces. It also is one of the main reasons that overall support for CDMA is strong within the third-generation standards groups. Technical possibilities include code aggregation and variable channel width.

Code aggregation, peculiar to CDMA, might be considered a form of virtual channel aggregation in that normally one code is assigned to one session. Aggregating codes would involve two or more codes to one user simultaneously. But because the spreading codes pack spectrum more efficiently than simple time division, the assignment of more than one code to a user makes less demand on spectrum per a given data rate than the combination of time slots or frequency bands. According to Qualcomm, code aggravation could be achieved within existing IS-95 systems with minimal modification of the switch or base station.

Codes themselves reside within broad frequency bands. The precise width depends upon the standard. You can adjust these in bandwidth to permit more codes to be assigned to a given user, which would permit higher throughputs. According to Ericsson, this technique will be important for high-speed wireless data services in the future.

A variable channel-width feature, the ultimate goal of third-generation planners, constitutes a fundamental modification of the air interface, which would require a significant hardware upgrade at the switch and the base stations. But the extent of the upgrade is a moot point because both CDMA and GSM standards committees are proposing new broadband CDMA air interfaces that differ from the existing IS-95 in many other respects as well and would entail extensive replacement of existing infrastructure. Some vendors have suggested that the new air interface could be imposed as an overlay, permitting older interfaces to continue to be used for established customers. Another vendor is still suggesting equipment upgrades.

Within a given air interface, you can implement several measures to improve throughput, all involving digital signal processing (DSP) prior to the final modulation, which establishes the air interface itself. These include higher digital compression rates, improved error correction and advanced modulation techniques.

Most compression schemes have focused on voice and have involved the refinement of the vocoder, a predictive algorithm that models the human vocal tract and can decrease the throughput necessary to transmit speech. Today, wireless vocoders go as low as 8kb/s, but third-generation advocates are discussing devices in the 3kb/s range.

Pressure for increased data throughput and overall system capacity will push the industry in the direction of lower rates, but in the short term, concern for voice quality will exert other possibilities.

"I don't think you'll see a reduction in vocoder rates right away," Salinger said. "In fact, there's been some backing away from the lowest rates. What you will see is the increasing use of adaptive vocoders, which vary the rate according to conditions on the network."

In addition to compression for voice, parallel compression schemes exist for text and for video, all exploiting the predictability of the material to avoid the transmission of complete representations. In addition, most methods achieve further economy by transmitting only changes. For example, in video, compression image content transmitted for one frame would not be retransmitted if it appeared unaltered in the next frame.

Advanced modulation schemes intrude on the air interface itself and represent overall increases in speed and capacity irrespective of message content. In a digital radio system, state conditions representing numerical values are represented as either a discrete number of amplitude levels in respect to a carrier wave, or as sidebands of given frequencies representing sum and difference products of the carrier. The more amplitude levels or sidebands relative to a carrier, the more information it can carry within a given allotment of spectrum. Current cellular and PCS systems recognize at most only four sidebands per carrier, but that figure easily could be increased to 16. Amplitude modulation could increase information density substantially more, although noise rejection characteristics of amplitude modulation are relatively poor.

Advanced modulation schemes change the way that radio signals are detected and involve the replacement of the radio and modifications of signal-processing algorithms. This replacement would have to occur at all levels, including handsets, base stations and central office.

Still another agent of change in the central office is the evolution of wireline telephone networks, which are moving toward packet voice as well. When a change occurs on the local level, the wireline network will undergo a physical transformation, and routers and servers will begin to replace conventional switching equipment. It would make sense for the wireless networks to install appropriate matching routers to transfer calls and to keep them within the appropriate protocol in order to maximize speed and efficiency.

A move to packetization in the wireline sector would vastly increase the capacity of the networks and would speed Internet traffic while opening the Internet to the kinds of multimedia transmissions that are still impractical today. Considering Internet growth and the demand to run more bandwidth-intensive operations over it, a changeover from circuit telephony to packet telephony is a logical development. It also would be beneficial to wireline carriers that already have adopted packet protocols.

"It would flatten the total telephone network," Robertson said. "You wouldn't have to go through (several) SS7 gateways."

But wireline telcos have reason to hesitate before adopting this scheme. Without local competition, regulated tariffs still apply under the law. There are still legal uncertainties about pricing voice as data. Local wireline carriers don't want to spend tens of millions of dollars on infrastructure upgrades without a clear confirmation that enhanced revenues will recoup those expenses.

Ultimately, you face a similar conundrum to that of their wireline brethren in contemplating the move toward a packet-based, third-generation infrastructure. Can high-speed data and multimedia be sold profitably to a mass subscriber base? Third-generation boosters point to the still explosive growth of the Internet, and predict steep and continuing acceleration of that growth in a sort of network analogy to Moore's Law where computer processing power for a given price doubles every 18 months.

Perhaps, but most current Internet service providers are caught in a ruinously intense price war where they must meet lower price points for unlimited monthly service and then contrive to upgrade their infrastructures. If that trend persists, it doesn't bode well for the third-generation early adopter or any other mass-market vendor of high-speed data.

"Right now, third generation is really being driven by technology, not the market," said Bill Frezza, a consultant with Wireless Computing Associates. "I'd be scared to death if I were an operator because the real market potential has yet to be demonstrated. Will the operators be able to pay back the cost of infrastructure with increased revenues? I'm not sure."

The path to packetization is not a straightforward one. At least three overall standards for packetizing data already are established in the data networking world: Internet protocol, ATM and frame relay. Although they are not strictly incompatible, each imposes different equipment requirements on carriers.

Internet Protocol IP already is used for voice in private networks, but because it has scant provisions for assigning priority to different messages, it tendsto impose troublesome latencies on voice transmissions, especially as the network is loaded. Nevertheless, the ubiquity of IP in the computing world is a strong argument for its adoption in telecommunications.

"I see IP voice as a real factor," said Dave Robertson, assistant vice president of Wireless Strategic Product Development for Nortel.

The Second Protocol ATM, the second protocol, is not really a packet system at all because the individual segments of data are fixed length. Developed in the late '80s and early '90s by the ISDN group within the IEEE, ATM was designed specifically to permit a range of broadband services to coexist on the same network and then appeared to be uniquely qualified to become the universal protocol in the multimedia age. ATM permits numerous distinct levels of service to be assigned network users and guarantees speed and latency for critical applications. But it does so at the price of considerable processing overhead compared to the other protocols.

ATM is used increasingly in the high-speed, fiber-optic backbones employed by long-distance carriers and probably will be widely used in the DSL high-speed data services that local carriers are expected to offer in the midterm. It also is advocated for cellular and PCS backhaul, and ultimately even for use in the normal wireless traffic of the networks.

"Our proposal for IS-34 would incorporate ATM at the air interface," noted Bob Salinger, director of marketing and business development for Lucent Technologies. "It will definitely play a major role in the third generation."

Nevertheless, because of its high overhead and relative inefficiency, ATM has many detractors. It is rarely used in private LANs and is ill-suited to transmissions of less than 1Mb/s.

Frame Relay Frame relay, the last of the packet or quasi-packet protocols, is a sophisticated true packet system that offers fairly low overhead, high efficiency and some degree of service segmentation and differential control of speed and latency. It is widely used by high-speed packet data services as an alternative to dedicated leased T1 lines and is being used for long distance within corporate networks. In voice applications, the better implementations already are close to toll quality.

In the wireless environment, frame relay is used to transmit commands from the base-station controller to the base stations, according to Salinger.

Frame relay is a good compromise for bandwidth-limited wireless applications, but it lacks the broad base of support that the other two protocols have. It is used less in private networks than IP and used less in wireline telephony than ATM. Still, it can't be dismissed.

According to Keith Shank, Ericsson's director of product management for IS-136, all three protocols will coexist into the future.

Nortel's Robertson agreed.

"Frame is already in use in the backhaul, and we see ATM as the key enabler for new types of traffic," he said. "I don't see one over the other; I think all three protocols are compatible. It's real easy to ride frames and IP on top of ATM."

But, Bill Frezza, a consultant with Wireless Computing Associates, added that the movement toward third generation likely will prevail.

"It always comes back to data when they're looking for a new source of revenue, and past disappointments really don't matter," he said.

But translation from one to another clearly does not conduce to overall network efficiency nor do all three protocols support the same mix of services. Protocol uncertainty is apt to persist for some time. How well will existing modulation schemes lend themselves to packet transmissions? Packet-data standards already have been established for existing GSM, IS-34 and IS-95 air interfaces, but none of these standards includes provisions for frame relay or ATM. An international wireless ATM group that Motorola heads currently is drafting proposals for over-the-air ATM, but accommodating either that protocol or frame relay to wireless transmissions would be problematic because both forgo redundancy and error checking for speed.

At any rate, the consensus among third-generation proponents is that all three international digital standards -- D-AMPS, GSM and CDMA -- will have to be modified extensively to meet the overall goals for third-generation speed and efficiency. This is particularly critical in the GSM camp, where many are saying that broadband CDMA will have to replace the underlying time division. A composite TDMA-CDMA scheme has been proposed in Europe to allow for the phased upgrade of existing GSM systems to full third-generation capabilities, but the cost of implementation remains to be seen.

However, Julie Cunningham, vice president of investor relations for Qualcomm, suggested that the migration path for IS-95 would be simpler than for the other two modulation schemes.