|
As late as last February, irrationally exuberant telecom execs were renting $40,000-a-week yachts to promote their upcoming 3G wireless wares at the 3GSM World Congress in Cannes.
With all that headiness, market motion sickness may have prompted the president of Japan's NTT DoCoMo to say 10 million pets and 60 million bicycles would sport 3G data gadgetry by 2010.
At the vortex of the buzz, 3G is billed as a broadband service which features the packet-based transmission of text, digitized voice, video, and multimedia.
| TDMA: TIMING WHEN TO YELL |
|
By: Jaquelin Spong
Among the greatest spoils of being first is ability to establish standards to which followers must adhere. Ericsson of Sweden, flanked by Nokia of Finland, led the Viking invasion by propounding a new cellular system for Europe, named Groupe Speciale Mobile (GSM) after the commission that conceived it.
With more than 100 million users in 147 countries by 2000, GSM became a global standard and its acronym changed to Global System Mobile. The US Telephone Industry Association (TIA) in 1991 adopted the IS - x4 standard similar to the European GSM, opening the way for eager European firms to sell equipment in America.
GSM uses an access method called time-division multiple access (TDMA). Reflecting the time-sharing methods of centralized computers with large numbers of users, TDMA stems from time-division multiplexing employed by phone companies to put more than one call on each digital line. That is, multiple users share the same channel by dividing it into time slots. Each user gets exclusive use of its assigned slot.
Traditional wisdom would suggest that to be sure your message heard in a crowded room, yell louder. The use of low frequency, high power transmitters is a vestige of the loud-yelling days of analog transmission, in which the higher the amplitude, the better the signal-to-noise. TDMA persists in this paradigm but reduces the need to yell by requiring all others in the room to be silent during the time you want to be heard. The time-division multiplexing yields a threefold improvement over standard channel technology.
|
Unlike existing 2G and planned 2.5G macrocell networks, 3G deployment locations will include the deployment of microcells and picocells, which are part of the Bluetooth scheme (see the PANdemonium companion story). More importantly, 3G cellular networks will evolve into truly all-IP network architectures with improved access to Internet-based applications. For this type of new all-IP network, some sort of VoIP or voice-over-packet standard, such as H.323, will be used for voice services. As a result, true all-IP 3G systems will have huge infrastructure and management planning implications for major end-user organizations.
The sheer magnitude of this technology leap is what led Olav Ostin, who is the UK managing director of the global venture capital firm ETF Group Half, to throw cold water over the crowd at the World Congress. He solemnly predicted that 50% of exhibiting companies there would go bust within a year, with wireless content providers as the most at risk. Can you spell WAP?
A few months later, Europe-supposedly a generation ahead of the US in advanced mobile wireless system deployment-was struggling to get "2.5G" networks on the air. Even the rosiest of industry projections from Orange, Europe's number two mobile operator, puts 3G profitability far off into the future. Orange says the ink will run red for at least four to seven years after its service launch in 2002, which may slip to 2003. Even the wildly successful NTT DoCoMo has hiccupped on its scheduled rollout dates in its bid to become the world's key provider of 3G high-speed wireless data services.
For much of the media, NTT DoCoMO is seen as an overnight success. Nothing's further from the truth. From a cold start on February 22, 1999, NTT DoCoMo introduced i-mode always-on Internet access via 2G-cellular technology. By August 2000, DoCoMo's customers numbered more than 10 million. And contrary to conventional wireless Net revenue wisdom, NTT DoCoMo is raking in billions of un-WAPped yen per year.
What has essentially gone unnoticed is that NTT DoCoMo's i-mode always-on handheld Internet terminals are based on TRON-specification technologies dating back to 1982. From this perspective, NTT DoCoMo's 3G-service rollout is simply a continuation of 20 years of infrastructure planning in Japan.
Notably, the Business TRON (BTRON) computer architecture specification was unveiled in 1988 prior even to the existence of the World Wide Web. BTRON was designed for platforms with limited resources and incorporated a unified model for user operation, data management, and program execution, which is a construct not unlike today's PDAs. As a result, DoCoMo was able to establish a criterion for wireless data success, be it 2G i-mode, 2.5G, 3G, or any other type of user service elsewhere in the world.
What's more, 3G is supposedly an all-encompassing, worldwide single standard, but it's clear that geopolitical/market/technology forces will prevent any single worldwide standard. So the question is, what will 3G network services look like technologically? The answer to that question largely depends on who and what end up winning the 3G race, an outcome that becomes more uncertain with each passing day.
For now, all signals point to a prolonged pit stop at 2.5G, which is the last step in mobile communications for traditional cellular voice operators.
When the marketing dust settles, the so-called 2.5G networks are primarily differentiated from older 2G setups by higher data throughput. Anything that hits a data rate above 56 kbps apparently earns the moniker 2.5G, although always-on service is often touted as being part of the 2.5G market fabric. 2.5G networks use six conflicting and incompatible systems to attain these higher numbers: HSCSD, GPRS, EDGE, EGPRS, IS-136B/HS, and cdma2000-1x. As of late, no less than four of these-HSCSD, GPRS, EDGE, and cdma2000-1x-have been getting most of the telco attention.
High Speed Circuit Switched Data (HSCSD) is a slight modification of the GSM coding scheme coupled with multi-slot (channel) operation. Second-generation GSM systems use Time Division Multiple Access (TDMA) with dedicated channels for each side of a voice conversation and a network optimized for voice traffic. Accordingly, 2G GSM networks bear a strong resemblance to the wire-line telephone infrastructure. Data users are treated as wasteful, second-class citizens in the 2G world and are given a measly 9.6 kb/second per time-slot on their GSM phones.
HSCSD allows concatenation of multiple time-slots to increase data throughput, but it still retains the original dedicated voice channel architecture of GSM. Basically, combining these two techniques permit user data rates to double to 14.4 kb/second per slot, or even shoot up to 57.6 kb/second if four dedicated time slots are combined. Thus, a three or four-time slot HSCSD network is a 2.5G rig, even though it's not a packet network overlay system.
Nonetheless, time slots are limited and thus valuable for the bread-and-butter business of cellular operators: voice traffic. Moreover, the use of a dedicated channel, although a natural extension of wired telephone networks, is a poor choice for the short, bursty nature of a data network. As a result, HSCSD service is typically available only where cells are not fully loaded or during off-peak hours. General Packet Radio Service (GPRS) has come along to counter these thorny issues.
GPRS gets rid of the dedicated slot scheme of HSCSD and substitutes shared channels capable of transmitting data packets from multiple users on each channel. GPRS is a separate packet network overlaid on the original circuit-switched voice network, which is capable of adjusting itself to varying signal propagation and interference conditions. As a consequence, four possible data rates are available per channel, ranging from 9 kbps to 21.4 kbps. Like HSCSD, multi-slot operation is also permitted, providing, theoretically at least, data rates as fast as 171.2 kbps-assuming you live next to a cellular tower and are the only person in the world on the network-by using eight time slots under the best possible conditions.
Live GPRS simulations done by the industry- indicated maximum user data rates under
realistic conditions are said to be only about 40 kbps. Worse, in heavily loaded cells, GPRS has delivered only about a 2x improvement in data rates compared to GSM, primarily due to the use of packet-switched channels.
British Telecommunications Cellnet unit launched the world's first commercially available GPRS service and it provided a miserable data throughput of just 26 kbps. That performance was far removed from the 115 kbps to 170 kbps numbers that GPRS backers breathlessly promised users, even though everyone close to the situation knew that the high rates were totally off the wall.
In the harsh light of day, GPRS stalwart Nokia had to admit that maximum data rates of about 43 kbps were going to the norm, while Ericsson thought 56 kbps might be achievable. Both revised estimates should still be met with some skepticism, since GPRS cells quickly become overloaded, making it unlikely that a single data user would be allocated all eight time slots.
Interestingly, 43 kbps also approximates the throughput achievable from HSCSD when using three channels. At least one operator, UK-based Orange, plans to offer both HSCSD and GPRS as new 2.5G services. Orange currently offers HSCSD, although only at 28.8kbit/s, and intends to roll out GPRS later this year. It's easy to see the business logic behind Orange's thinking. Unlike the simple base station modifications required by HSCSD, GPRS requires both hardware and software modifications that delve much deeper into the infrastructure.
Significantly, GPRS, unlike HSCSD, enables always-on service, a feature in a packet-switched network that NTT's highly successful i-mode has shown to be more important to users than raw data rate. GPRS delivers faster connection-set-up time than GSM, permitting data niblets to be continually downloaded to a user's device. In addition, the packet-switched nature of GPRS allows billing either by minutes of use for voice or by volume of data transferred.
A form of GPRS will be at the packet core of the base stations providing the always-on capability for 3G networks. That always-on feature may be the deciding GPRS blow to HSCSD.
With GPRS now proven to be only slightly better than HSCSD, a path was cleared for a third combatant to step into the 2.5G ring: Enhanced Data Rates for GSM Evolution (EDGE). EDGE is a higher data rate, logical extension of GPRS whose claim to fame is its supposed ability to deliver maximum user data rates in excess of 384 kbps. This feat is accomplished by changing the modulation format to pack more bits of information into each slice of frequency spectrum. In theory, that scheme should retain full system capacity while increasing achievable data rates. The fly in the EDGE ointment is error rate correction, better known as maintenance of an acceptable quality of service (QoS).
As with the other competing systems, to get the fastest possible EDGE connection you have to live on top of the cell tower and be the only one on the system. The higher modulation levels of EDGE-a TDMA system-require higher signal strength at the receiver than GPRS.
Move away from the base of the cell tower, and EDGE's link quality control protocol will reduce the data rate to maintain an acceptable QoS. In particular, co-channel interference is the primary rate-killing culprit.
Nonetheless, simulations by Ericsson indicate EDGE's link-quality control is significantly better than that offered by GPRS. Even when measured at a comparable GPRS signal strength, EDGE delivers twice the data rate performance per slot. More importantly, since EDGE utilizes higher signal strengths, data rates will keep climbing after the weaker GPRS signal fade out. Based on the real-world BT Cellnet data, this probably means an average single slot EDGE data user will see about 45-60 kbps performance.
In comparison to GPRS, EDGE does enjoy an edge, but this performance is just about comparable to a typical analog dial-up line. The broadband beef is still out there somewhere roaming the high-frequency range. In GSM-based systems, an upgrade to EDGE will involve mainly software upgrades.
Going beyond the 2.5G-technology pit stop, the 3G version of GSM is technically called Universal Mobile Telecommunications System (UMTS), but also goes by the moniker IMT-2000. Wideband-CDMA (W-CDMA) was chosen as the basis for 3G/UMTS by existing GSM equipment suppliers, who are mostly European and Japanese manufacturers. But in the standards process, another major contender to W-CDMA emerged, called cdma2000.
The architecture of cdma2000 is fundamentally different from that of GPRS/EDGE, which use specialized, cellular-specific protocols. The new cdma2000 uses PPP to link users to a packet-data serving node (PDSN) and Mobile IP to support customers roaming between cdma2000 networks. If it lives up to the hype, the Mobile IP option will allow customers with unique IP addresses to actively roam in cdma2000 networks.
Another strength of cdma2000 is backwards- compatibility with the IS-95 standard, which is widely deployed in North America and is the archrival to GSM-TDMA. The most discussed and debated parameter between the two dueling 3G systems is the system chip rate.
W-CDMA uses a chip rate value of 4.096 Mbps, while cdma200 uses 3.6864 Mbps. That claim W-CDMA supporters improves capacity by as much as a 10% over cdma2000. Proponents also claim that W-CDMA provides more efficient use of a 5-MHz channel. The W-CDMA protocol was also designed to provide data transfer rates of 2.3 mbps. Nonetheless, nothing in 3G W-CDMA is backward-compatible with a 2G system, be it GSM or CDMA. As a result, transition costs to a 3G system built on W-CDMA are in the billions.
NTT DoCoMo's delayed 3G cellular service is reported to have already cost $10 billion dollars.To ensure its first-to-launch status, a limited (just under 4,000 customers) pilot project was announced for kickoff. Such financial bloodletting has pushed back 3G W-CDMA deployments until 2003 and maybe 2005. None of this made for a sunny day on the Cote d'Azur, especially after Qualcomm's Chief Executive at the February World Congress said that W-CDMA would not be commercially viable for several years.
On top of that, the U.S. seems determined to be the last major country to turn on the 3G lights. It was only in January 2001 that the FCC finally got around to asking for comment on the best frequencies to be designated for 3G services. The US doesn't plan to have 3G-spectrum available for carriers until at least September 2002. At that tortoise pace, it may not be until 2004 or 2005 before that first U.S. 3G cell phone starts its annoying beeps.
So where's the good news? For those "lucky" auction winners of 3G-spectrum, the standard allocates spectrum in existing TDMA 2G systems, which enables current network providers to preserve and incorporate their legacy infrastructure investments. On the other hand, that also means legacy operators will not feel compelled to build new all-IP networks and 3G handsets will have to support multiple formats-at least 3G, GPRS, EDGE and maybe HSCSD, in turn a major impediment for manufacturers trying to design small, low-cost battery-efficient handsets.
In contrast, existing handsets will be interoperable with a cdma2000 network, as it is backward-compatible with 2G IS-95 systems.
Making the 3G-picture even fuzzier, interim solutions that are based on TDMA or hybrid techniques will also be evolving into 3G for some time to come.
Most of these solutions will probably utilize EDGE. AT&T, for example, has already announced its intention to use EDGE for its new 3G networks, proclaiming its maximum data rate of 384 kbps as 3G-proof positive. Nonetheless, achieving that 384 kbps rate in the real world is suspect.
Even more suspect is the great 3G hope that rests on widespread WAP deployment on the Internet. Given the disastrous dot-com business climate, this is not exactly the kind of planning premise on which to bank billions.
| CDMA's Hot SPOT: FINDING THE RIGHT VOLUME
|
|
By: Jaquelin Spong
The United States has a long and distinguished legacy of achievement in signal processing, fueled in large part by the space program and its need to retrieve signals from distant space probes transmitting through a sea of noise. Led by the likes of Claude Shannon, Andrew Viterbi, and Irwin Jacobs, very sophisticated algorithms and esoteric mathematics begat tremendous increases in data-handling capabilities of digital channels. And so Hamming, trellis, and turbo codes made their way from technical literature, to Martian landers, and then into computer hard disk drives.
CDMA, developed by the San Diego-based Qualcomm, takes a fundamentally different approach from the entrenched TDMA: Rather than turning up the volume, it has the transmitter and receiver speak a unique language. If everyone in the room can be persuaded to keep their voices down, listeners can detect and decode a message in their own language with good accuracy. The fidelity is good and everyone can talk at once, which achieves a stunning increase in overall data throughput. CDMA's advantages in power consumption, error-free transmission, range, and data rate are rapidly making it a new global standard.
Rather then compressing each call between three and 10 tiny time slots in a 30 kHz cellular channel, CDMA differentiates calls by multiplying their signal with a pseudo-random sequence that resembles white noise. It spreads the signal across a comparatively large 1.2 MHz swath of the cellular spectrum. This allows many users to share the same spectrum space at one time. Each phone is programmed with a pseudo-noise code key, which is used to spread a low-powered signal over a wide frequency band. The base station uses the same code in an inverted form to despread and reconstitute the original signal. All other codes remain spread out, indistinguishable from background noise.
A serious apprehension dogging CDMA was the scenario in which one person is yelling, while others keep their voices down. In contrast to TDMA, high power
signals seriously interfere with the integrity of simultaneous communications. For years, this problem crippled CDMA as a way of increasing the capacity of cellular systems. Nonetheless, spread spectrum had many military uses as it was difficult to jam or overhear.
In a cellular environment, cars continually move in and out from behind trucks, buildings and other obstacles, causing huge variations in power. As a result, CDMA systems would be regularly swamped by loud signals. Similarly, nearby cars would tend to dominate faraway vehicles. This was termed the near-far problem.
When you compound that problem with the static of multi-path signals causing hundreds of gyrations in power for every foot traveled by the mobile units-dubbed Rayleigh pits and spikes-you can comprehend the general incredulity toward CDMA among the cellular cognescenti.
Incredulity no more. By using digital signal processing, error correction, and other tools in such diverse areas as disk drive and deep space communications, engineers at Qualcomm have surmounted these problems with a series of feedback loops.
Automatic gain control in the handset and constant surveillance of the signal-to-noise and error rates from the base station, wattage spike, and pits are regulated by electronic circuitry that adjusts power at a rate of more than 800 times per second. This is a fancy way of saying that Qualcomm's phones and relay
stations are very, very smart.
While sounding complex, such signal processing techniques are well known and relatively easy to implement electronically, far easier than adjusting time slots. As a result, these devices can figuratively focus on one Swahili speaker at a party of 1,000 guests. What's more, CDMA, unlike TDMA, captures for use every pause or silence in the conversation.
Once the problem of the stentorian voice could be instantly abated, device power changed from a crippling weakness of CDMA into a commanding asset. Power usage is a major obstacle for the PCS future. All market tests show that either heavy or short-lived batteries greatly reduce the attractiveness of cell phones. Because the Qualcomm feedback system keeps transmit power always at the lowest feasible level, batteries in CDMA cell phones last longer than TDMA phones.
A further advantage comes in handling multipath signals, which bounce off obstacles and arrive at different times at the receiver. Multipath just adds to the accuracy of CDMA. The Qualcomm system combines the three strongest signals into one. Called a rake receiver and co invented by Paul Green of fiber fame, this combing function works even on signals from different cells and thus facilitates hand-offs. In CDMA, which does not chop up the signal into time slots, multipath signals come in time to strengthen the message.
Finally, CDMA allows simple and soft handoffs. Because all the phones are using the same spectrum space, moving from one cell to another is relatively easy. Qualcomm has reduced all the digital signal processing for CDMA into one application-specific chip. Supplanting the multiple radios of TDMA-each with a fixed frequency-are digital signal processing chips that find a particular message across a wide spectrum swath captured by one broadband radio.
|
|
