Shosteck Group White Paper on TDMA 3G Migration Paths  |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
GSM or CDMA: The Commercial and Technology
Challenges for TDMA Operators
4 Comparing the Migration Paths 4.1 Introduction In our previous chapter, we described four potential barriers and the ensuing uncertainties that currently face TDMA/IS-136 operators who may choose GSM as a migration path to 3G. We consider two of these especially important.
In this chapter, we examine further the time that may be needed to develop and deploy infrastructure and handsets for GSM 800. We introduce the sometimes overlooked issue of the backward compatibility of cdmaOne base stations with legacy TDMA/IS-136 switches and how that may impact the cost of the 3G transition. Finally, we discuss plausible handset costs. In some cases, we find apparent advantage for cdmaOne. In other cases, we find apparent advantage for GSM. Overall, we maintain the conclusion of our previous chapter, that the uncertainties of the transition warrant TDMA operators reviewing CDMA as an alternative path to 3G. 4.2 The Time Required to Develop and Deploy Infrastructure In February 2000, Nokia announced that it would begin “system deliveries…[of GSM technologies at 800 MHz] during the second half of [2001].”46 Similar announcements from Ericsson, Motorola, and Nortel followed. This new GSM alternative expanded the migration options for TDMA/IS-136 operators using the 800 MHz frequencies. However, the deployment of GSM 800 faces challenges. All vendors must decide how to allocate their finite resources. The wave of recent staff reductions has not eased this decision. This raises two questions.
These are not trivial questions. If vendors fail to deliver and integrate GSM 800 infrastructure in a timely manner, the operators left waiting lose competitive position to their cdmaOne- and GSM-based rivals. 4.2.1 Delivering the Infrastructure Nokia points out, correctly, that the GSM standard is exceptionally rigorous. Nokia characterizes “down banding” from 900 to 800 MHz as “fairly trivial.” However, Nokia cautions that subsequent system testing takes time. As of May 2000, Nokia anticipated that it would deliver GSM 800 base stations during the fourth quarter of 2001. This is not inconsistent with its original announced date of deployment. Given that Nokia announced GSM 800 only in February 2001, this timeframe may seem optimistic. However, while Nokia will not divulge the timing of its R&D cycles, it acknowledged that it initiated the GSM 800 program during 1999 and that it was in a “fairly major stage” of development by the February announcement.47 Given this, Nokia’s target of a fourth quarter infrastructure delivery is not unrealistic. Ericsson’s perspectives are not dissimilar from those of Nokia. Ericsson points out that a great deal of R&D has already been invested in defining EDGE for 800 MHz TDMA/IS-136. Most of that R&D is applicable to GSM 800. As of May 2000, Ericsson, in common with Nokia, anticipated delivering initial volumes of GSM 800 infrastructure during the fourth quarter of 2001 and larger volumes during the first quarter of 2002.48 Other industry sources have corroborated these initial deployment dates.49 On the basis of these reports, it appears that Nokia and Ericsson, at least, will be shipping GSM 800 infrastructure by the end of 2001. 4.2.2 Deploying the Infrastructure However, as we observed earlier, shipping GSM 800 infrastructure in a timely manner resolves only half of the infrastructure challenge. The remaining half centers on successfully overlaying that infrastructure onto established TDMA/IS-136 networks. This may take longer than vendors foresee. The reason would center on the unique complexity of RF engineering in the 800 MHz spectrum. In contrast to mobile spectrum allocated to other frequencies (450, 900, 1500, 1700, 1800, and 1900 MHz), the 800 MHz spectrum in North and South America differs in two regards.
Combined, these two factors pose considerable challenges to deploying any overlay technology, but especially GSM. The crowding of the 800 MHz spectrum stems from differences in spectrum allocation among countries and services. In the U.S., and most countries of North and South America, the allocation for 800 MHz mobile service is relatively narrow (2 × 25 MHz). In comparison, the European allocation for mobile services at 900 MHz provides more than half again as much spectrum (2 × 39 MHz). The PCS allocation at 1900 MHz provides more than twice as much (2 × 60 MHz). The crowding at 800 MHz increases the RF engineering challenges of introducing a new technology to supplement or replace TDMA/IS-136. This would hold for either GSM or cdmaOne. To make room for the new technology, the cellular reuse structure of the TDMA network must be disassembled while that of the new technology must be put in place. The challenge would be greatest during the initial deployment. Due to issues of interference and required “guard bands,” CDMA would need an initial 1.8 MHz while GSM would need an initial 2.5 MHz.50 In theory, CDMA’s need for less spectrum would make its initial deployment less difficult. The engineering difficulties of such a deployment translate directly into a marketing challenge. By removing TDMA/IS-136 channels to support a new RF technology, operators run the risk of degrading service. Degraded service will dissatisfy customers and risk stimulating churn. Depending on the bandwidth an operator has available, the initial spectrum needed to deploy cdmaOne or GSM may or may not be important. The issue of multiple RF technologies at 800 MHz may prove as challenging as the issue of crowded spectrum and possibly more so. With the exception of systems at 1900 MHz, GSM has always been exclusively deployed on dedicated spectrum at 900 and 1800 MHz. No other RF technologies co-share the spectrum—and, thereby, generate the potential of cross-technology interference. GSM engineers who work in the 900 and 1800 MHz frequencies have ample experience in dealing with interference among adjacent GSM channels. They have no experience in dealing with interference from adjacent TDMA/IS-136 and/or cdmaOne. An argument can be made that GSM, TDMA/IS-136, and cdmaOne occupy the same spectrum at 1900 MHz. Because of this, GSM engineers have experience in deploying GSM in a TDMA and CDMA world.51 Up to a point, this argument is valid. However, as we observed earlier, less spectrum is available at 800 MHz than at 1900 MHz. Because of this, the GSM experience at 1900 MHz has not dealt with the frequency congestion common to 800 MHz. In addition, the specialized mobile radio (SMR) bands are adjacent to the 800 MHz mobile frequencies in the U.S. and Canada and in some cases elsewhere in the Americas. These introduce the further challenge of dealing with interference from the iDEN technology deployed on those bands. Nor should operators overlook the continuing prevalence of 800 MHz AMPS and its potential for interference. The deployment of TDMA/IS-54 (the predecessor of IS-136) serves as an example of initial promise versus eventual reality. At the time of TDMA’s introduction, operators were told that its deployment would require only replacing a 30 kHz AMPS radio channel with a 30 kHz TDMA radio channel at the base station. The reality proved different. RF engineers discovered that useful TDMA signals attenuate more rapidly than AMPS signals. This phenomenon required a time consuming re-balancing of the networks as TDMA was introduced. In addition, TDMA experienced adjacent channel interference with AMPS. Overcoming these problems of network re-balancing and interference minimization proved a long and costly process. The same phenomena occurred in the UK and Europe as network operators made the transition from TACS to GSM.52 If past experience serves as a guide, problems of interference and network re-balancing will lead to a more difficult integration of GSM onto TDMA networks than vendors anticipate. Should this be the case, the successful deployment of GSM 800 infrastructure would extend well beyond the fourth quarter of 2001. This is not to say that cdmaOne will not pose similar challenges. It will. However, in contrast to GSM, CDMA engineers have five years of experience in deploying CDMA in the 800 MHz frequencies. GSM engineers have none. 4.3 The Time Required to Develop and Deploy Handsets If the availability of GSM 800 infrastructure is critical for the transition plans of TDMA/IS-136 operators, the availability of GSM 800 handsets is even more so. Without handsets, networks cannot function. Idle networks strand operator investment. When handsets are available, they must be fully functional. If not, end-users will reject them, and revenues to operators will fail to materialize. 4.3.1 The “Reality Gap” in Handset Delivery Historically, the availability of handsets has always lagged the deployment of infrastructure. We call this the “reality gap.” It consists of two dimensions. On the one hand is the difference between the performance that handset vendors promise and the capabilities of their initial products. On the other hand is the difference between when vendors promise delivery and when they actually do so. As our examples illustrate, the reality gap is universal. The first GSM network was officially launched in Helsinki in July 1991. Due to chronic handset shortages and malfunctions, GSM was officially “re-launched” in Berlin in July 1992. The first TDMA/IS-54 (the predecessor of IS-136) handsets could not transmit calls to other TDMA handsets. The first cdmaOne handsets were characterized by an embarrassingly rapid battery drain.53 More recently, WAP handsets have proven a universal disappointment, failing to provide anything near the performance promised by the marketing hyperbole. British Telecom (BT), among the most enthusiastic promoters of WAP, has suffered. Purloined internal figures show total visits to BT’s WAP site fell from 115 million in January 2001 to 40.5 million in April 2001, a decline of 65 percent. Over the same period, total Web use fell from 36 million minutes to 12 million, a decline of 67 percent.54 No information has been reported regarding how many customers have churned because of their poor WAP experiences. GPRS is repeating the process of delayed availability and disappointing performance. In July 1999, Nokia promised that “GPRS services will be launched…in the second half of…2000, at which time GPRS terminals will also be available.”55 As of May 2001, Nokia planned to “begin shipping commercial GPRS phones during Q3 [2001] and in volume (multi-millions) during Q4.”56 This would be at least nine months later than originally announced. GPRS handsets that are already available from other vendors, most prominently Motorola, are not performing well. Anite Telecoms, a vendor of network performance equipment, has independently monitored BT’s GPRS network. Anite finds that current handsets are delivering data rates of only 8 kbps. This is but one-third of the 30 kbps expected and less than the 9.6 kbps provided by GSM.57 Overall, the reality gap suggests that, in common with all other technologies, GSM 800 handsets will come to market later than vendors first promise. 4.3.2 The Delivery of GSM 800 Handsets In discussing handset delivery, we must first ask what kind of handsets? In concept, TDMA/IS-136 operators, such as AT&T and Cingular, could construct their GSM networks rapidly and provide their subscribers with nation-wide coverage through roaming on VoiceStream’s GSM network. Under this scenario, TDMA operators could, in concept, require only GSM 800/1900 MHz handsets. This, however, is unlikely. More likely, TDMA operators will require dual-mode/dual-band handsets. Such handsets will enable their subscribers to hand off from GSM to TDMA networks and back again. This is the path that AT&T will follow.58 The advantage of the dual-mode approach is that it provides subscribers with continuous coverage, giving TDMA/IS-136 operators time to construct their GSM networks. The disadvantage centers on the time it will take for vendors to develop and deliver dual-mode/dual-band handsets and, as we suggest later, higher handset costs. As of May 2001, only Nokia would go on record regarding when GSM 800 handsets might be available. Nokia observed that given its experience in producing 900/1800 MHz GSM, “it’s almost a trivial thing” to down band from 900 to 800 MHz.59 A separate Nokia source, while making no commitment, pointed out that handset delivery usually follows infrastructure delivery by “6 to 12 months.”60 By inference, this suggests that, at present, Nokia intends to deliver GSM 800 handsets somewhere between the second and fourth quarters of 2002. However, the reality gap raises the question of whether Nokia (or any other vendor) can devote the resources to deliver GSM 800 handsets when, by inference, it has promised them. This question becomes especially acute because providing dual-mode/dual-band TDMA-GSM handsets requires developing and integrating two, and possibly three, separate technologies.
We referred to GAIT in our last chapter. GAIT is a proposed network standard
that translates MAP network signals into ANSI-41 signals and vice-versa. By
so doing, GAIT would enable TDMA-GSM handsets mobile subscribers to access their
full portfolios of personalized services regardless of whether they would be
on a TDMA/IS-136 or GSM network. Without GAIT, subscribers to GSM systems could
use only their voice and possibly SMS capabilities when handing off to TDMA
systems. This would abrogate a major attraction of GSM—its ability to access
a rich portfolio of personalized subscriber Given the complexity of a dual-mode/dual-band GSM 800 handset, the likelihood of a reality gap between promised and actual delivery appears substantial. That said, one could posit that AT&T and Cingular,63 as the second and third largest operators in the U.S., will command attention from the handset vendors. This is possible, but we think that it is too U.S.-centric a view. The established world market—primarily devoted to GSM 900 and 1800—is five to six times as large as that of the U.S. Given this size difference, handset vendors will continue to place the needs of Europe and Asia-Pacific first. These center on GSM 900/1800 for Europe and Asia and cdmaOne and CDMA2000 1X for Asia (and the Americas). AT&T and Cingular, demanding GSM 800/1900, will fall to the end of the handset supply line. Smaller TDMA/IS-136 operators, such as those in Latin America, will fall with them. Other frequencies, unique to the Americas, show the diminished attention that handset vendors pay to them. In early 2001, while Motorola was shipping its GPRS-enabled Timeport 260 to Europe, it was not shipping GPRS models to the U.S.64 A glance at a manufacturer’s website illustrates that fewer models for the Americas are available. Ericsson, for example, lists over 50 models each under its categories of GSM 900 and GSM 1800, but only 19 models under its category of GSM 1900 phones.65 Fewer models may actually be sold. Nokia, the world’s largest vendor of handsets, sells only five GSM 1900 models in the U.S.66 We see little likelihood that the forces favoring GSM 800 will strengthen. In 12 to 18 months, the pressures on handset vendors to focus on GPRS, EDGE, and/or UMTS will be greater than at present. In light of this, resources to produce an initial multi-mode TDMA-GSM 800 handset, let alone multiple models, will be more constrained than they are now. In sum, under best of circumstances, handsets for GSM 800 will not be available until the second to fourth quarters of 2002. However, if the historical reality gap serves as a guide, they will not be delivered until later. 4.4 The Delivery of CDMA2000 1X Handsets By comparison, handsets for CDMA2000 1X are in production and being used on all three Korean networks. As of early May 2001, as many as ten different handset models were available from four different manufacturers. Samsung led the pack with its SCH X100, X110, X120, X130, X200, and X1000 models, followed by two models from SK TeleTech, one from LG, and one from Motorola.67 And vendors are introducing more models. Samsung has begun marketing what it calls “the world’s first [CDMA2000 1X] mobile handset with the capability to receive motion pictures in color” and capable of reproducing “clear motion picture images in 200,000 different color shades.”68 This has increased the number of available models to 10. By the year-end 2001, SKT expects additional vendors to enter the market, including Nokia. Together, they will produce 26 additional CDMA2000 1X models, bringing the total available to 36.69 This is a meaningful lead over GSM 800, for which models will not be available until Q2 to Q4, 2002, at the earliest and, given the reality gap, not until much later. More handset variety stimulates market demand. In addition, more handset manufacturers drive down handset prices.70 This further stimulates demand. We will examine handset prices in greater detail later in this chapter. 4.5 Backward Compatibility with Legacy Networks and the Full Cost of Infrastructure Unlike the original North American standards—AMPS, TDMA/IS-136, 71 and cdmaOne—Europe’s GSM standard has always specified the interface between mobile switches and base stations.72 This means that GSM base stations and switches produced by any one vendor can interoperate with the base stations and switches produced by any other. In addition, although the total number of CDMA infrastructure vendors may be greater, more major vendors—notably Nokia, Siemens, and Alcatel—supply GSM infrastructure than supply CDMA infrastructure. As a consequence, measured on a hardware-to-hardware basis, the nominal price of GSM infrastructure is less than that of CDMA infrastructure. Discounting the substantial arguments concerning the relative performance benefits of the two technologies, this provides an apparent pricing advantage to GSM as a migration path from TDMA/IS-136 to 3G. This apparent advantage may remain, albeit to a lesser degree, even when GPRS and EDGE are incorporated into the GSM infrastructure pricing. That said, cdmaOne may hold infrastructure price advantages that are not as readily apparent. Two appear plausible.
These possible advantages stem from the fact that TDMA/IS-136 and cdmaOne employ a common network signaling (American National Standards Institute-41 or ANSI-41). Network signaling drives the subsystems and service platforms. However, GSM employs an incompatible network signaling (Mobile Application Part or MAP). MAP’s incompatibility with ANSI-41 precludes incorporating GSM into TDMA switches or other TDMA network elements. The extent to which the reuse of TDMA/IS-136 infrastructure may prove feasible will depend on the vendor and the willingness of the vendor to assure interoperability between its legacy TDMA switches and current cdmaOne base stations. Nortel appears to enable such adoption the easiest. “We can configure a single MTX [switch] to simultaneously support CDMA and TDMA cell site equipment. This is a viable method to transition from a TDMA network configuration to a CDMA network.”73 Lucent’s 5ESS switch, executive cellular processor/inter-message switch (ECP/IMS), and applications processor (AP) are capable of handling both TDMA and CDMA base stations simultaneously. However, to ensure full interoperability, Lucent would have to undertake integration testing. Lucent would do so if customers expressed sufficient demand.74 Ericsson can also support TDMA and CDMA base stations on the same switch, but not simultaneously. For this reason, Ericsson would use one set of switches to support TDMA and another set to support CDMA. In concept, Ericsson could provide “commercial credits” for operators making the migration.75 The extent to which new cdmaOne infrastructure could reuse the switches or other elements of a legacy TDMA/IS-136 network would also depend upon the extent to which specific vendors have designed interoperability into their products. Motorola, Samsung, and Ericsson, for example, produce base stations intended to operate on the switches produced by Lucent and Nortel. This is not to imply that incorporating cdmaOne into TDMA/IS-136 infrastructure will be without problems. It will not. No one has yet done it. Without the engineering experience, unexpected challenges will surely arise. However, as we observe below, their magnitude should not be as great as those that will arise in overlaying GSM onto TDMA. The discussion of backward compatibility reintroduces the issue of deploying GSM 800 onto a TDMA/IS-136 network and integrating the two technologies—in particular the MAP and ANSI-41 signaling systems. Even if a TDMA operator decided to build an entirely separate GSM 800 network, it would still need the two networks to communicate with each other. The most obvious need would be in the realm of billing. Can the MAP signaling system integrate into an ANSI-41 driven billing system? If not, the network operator will be forced to use two separate billing systems, one for its GSM 800 network and the other for its TDMA network. Nokia acknowledges this issue, characterizing the initial reality as being two billing systems.76 If this will be the case, how, if at all, will operators render a single bill to end-users? And, if separate bills are required, how will end-users respond? What about packet-based mobile-to-mobile calls? European and U.S. landline operators have more than 30 years of experience in developing translation gateways to carry circuit switched calls between incompatible signaling systems. They have 15 years of experience in carrying circuit switched calls between the incompatible signaling systems of mobile networks. There is no equivalent experience with packet calls. What little there is has centered on implementing GPRS. To date that has not been good. How then, will packet calls be carried between MAP-based GSM 800 and ANSI-based TDMA? True, GAIT is intended to deal with the MAP-ANSI translation. However, the issue remains—regardless of GAIT’s theoretical elegance or long-term success, engineers will first go through a long and arduous applied learning process. The above will prove central issues for TDMA/IS-136 operators who might deploy GSM 800. They will prove less so for those who might deploy cdmaOne. 4.6 The Costs of Handsets The cost of handsets has proven the Achilles heel of all new mobile technologies. For each new generation of technology, the high initial cost of handsets has slowed end-user adoption. Only when operators subsidized handset prices have they motivated end-users to adopt the new technology. And only as economies of manufacturing scale developed, have handset prices declined. This has held true for GSM, TDMA/IS-136 (originally IS-54), and cdmaOne. It is well recognized that for any new technology, the handset price is always greater than the handset price for established technologies. It is also well recognized that as technologies mature and gain economies of manufacturing scale, handset prices decline. However, it is less well recognized that the handset prices of the newest technologies tend to remain above those of the more mature ones. Table 4–1 illustrates this phenomenon. It compares the wholesale prices of GSM and cdmaOne handsets in the U.S. market from 1998 through 2000. GSM, first introduced in Europe in 1991–1992, is the more mature technology. CDMA, first introduced in Hong Kong, Korea, and the U.S. in 1995–1996, is the newer one. Based on surveys of the U.S. wireless market, they represent the average low wholesale price for all classes of CDMA and GSM handsets. (See notes in Table 4–1.) For years 1998 and 1999, they are the averages of four quarters. For 2000, they are the averages of the June and December quarters.
Table 4–1 documents that from 1998 through 2000 the average wholesale price of GSM handsets fell from $117 to $95. During the same time period, the wholesale price of cdmaOne handsets fell from $191 to $120. The price difference between GSM and CDMA handsets diminished from $74 in 1998 to $25 in 2000. Between 1999 and 2000, the price of CDMA handsets declined by 11 percent. Assuming that this rate of decline continues, it points to an average wholesale price of $106 during 2001. Our estimate of $106 as the current wholesale price for cdmaOne handsets is consistent with those of other sources. Ericsson places the current price at $95 to $110.77 QUALCOMM places it at $110 to $140.78 Lucent sees it as $100 to $129.79 Using our own estimates and the midpoints of the Ericsson, QUALCOMM, and Lucent ranges, these prices average to $112. However, our core interest centers on the likely wholesale price of CDMA2000 1X handsets. A number of our sources were willing to share this information with us. Sprint placed the price at “several tens” of dollars more than that of cdmaOne and did not dispute our inference that this indicated an incremental increase of $30 to $50.80 Added to the $112 price of cdmaOne, this indicates that CDMA2000 1X handsets will range from $142 to $162. Consistent with this, Lucent reports a price of $149.81 Ericsson estimates $130 to $140. Using the midpoints of the ranges, these estimates average to $145. Assuming an annual 11 percent price decline, the price would fall to $129 during 2002. How does this compare with the prices of the dual-mode TDMA-GSM phones that AT&T, at least, intends to use as it makes the transition from TDMA/IS-136 to GSM? One source, speaking not for attribution, holds that “a phone like that will never be cheap enough” for end-users to afford or for network operators to subsidize.82 This source observed that any dual-mode phone, whether TDMA-CDMA or TDMA-GSM, would be an interim product that could never develop long-term volume. Without long-term volume, such phones would always demand a price penalty. Implicitly, this source differentiated such phones from dual-mode AMPS-digital phones, which did realize long-term volumes. Given a price penalty for dual-mode, what might it be? At this juncture, Nokia is most advanced in developing GSM 800 products. Nokia was extremely reluctant to provide specific prices. However, it affirmed “we know that we can” produce “in a profitable way” a dual-band GAIT phone for under $200. Such a phone would be similar to its present 5165. However, Nokia emphasized that, depending on functionality, the [initial] price of such a phone could be higher or lower.83 We infer from this that the initial price would be higher. In sum, the likely wholesale price of GAIT phones would be $200 or more compared to the likely wholesale price of CDMA2000 1X phones, which would be $129. The minimum price difference would be $71. Likely, it would be more, plausibly much more. Operators who choose the GAIT alternative would have to compete under this price penalty. 4.7 The Issue of Dual-Mode TDMA-CDMA Handsets At this point, readers must raise the issues of dual-mode TDMA-CDMA handsets. Would not the same challenges that arise with TDMA-GSM handsets arise with TDMA-CDMA handsets? The answer is yes. However, dual-mode TDMA-CDMA handsets may not be produced. As of this writing, QUALCOMM is studying developing TDMA-CDMA chips. However, it has not yet committed to them.84 This suggests that such dual-mode phones may not be needed. This viewpoint may make economic and commercial sense. It assumes that operators and end-users benefit more by the operators’ spending their money to deploy infrastructure rather than to subsidize handsets. Under this assumption, TDMA/IS-136 operators who may choose CDMA2000 1X would deploy a parallel CDMA2000 1X network as quickly as possible. Once it provided sufficient area coverage, they would begin to migrate their subscribers to it. In concept, there is no reason why TDMA operators who may choose GSM could not do the same thing. Doing so would avoid the price penalties of dual-mode phones. Rapidly building out a network may not be as onerous a task as may at first seem to be the case. Critically, the civil engineering in terms of cell sites is already in place. The backhaul transmission is in place, as well. The new network would have few initial subscribers. Thus, it would be initially designed to provide wide coverage, not high capacity. For this reason, the initial deployment of a parallel network, whether CDMA2000 1X or GSM, would require relatively few base stations. More importantly, it would be deployed quickly and at relatively low cost. In sum, the option of quickly deploying a fully parallel CDMA2000 1X network, and completely foregoing dual-mode handsets may prove economically more attractive to operators than shouldering the continuous stream of subsidies required for dual-mode handsets. This economic advantage would apply to the paths of cdmaOne and GSM, alike. 
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
 |
|
|
||||
|
||||
|
|
