re: Deutsche Bank Securities on 3G (2 of 2)
Continued from prior post.
Part 1 of this report summary is in the previous post:
Message 17928089
>> The Rise of the 3G Empire - Part 1 B
[Part 2 will continue in the September/October issue of Base Station Earth]
Brian T. Modoff
Michael W. Thelander
Daniel D. Kaplan,
Deutsche Bank Securities
August 2002
base-earth.com
Don’t Forget Memory Requirements
In addition to the need for extra horsepower and battery life, memory requirements with 3G also increase. It is estimated that the memory needs of a DSP Read Only Memory (ROM) for an EDGE handset could increase by a factor of at least five times and that DSP Random Access Memory (RAM) will at least double. These estimates are even worse for WCDMA. The memory requirements for the microprocessor increase as well. It is estimated that Flash memory will have to increase from 16 – 32 Mbits to 64 Mbits and SRAM will have to increase from around 2 – 4 Mbits to as high as 32 Mbits.
Memory chips are another drain on the battery and can hog precious space in the compact handset as well. At the same time, the handset architecture needs to be flexible to support changing standards or improvements. To address these issues, a flexible software approach is preferred instead of a strictly hardware solution (in a following section, we discuss how a software solution can be taken to a further extreme–a software–defined radio). Thus, efficient coding is needed to minimize memory requirements and memory devices are required that can optimize the performance of the processor(s).
Intel leads the pack in providing Flash memory, having recently introduced a 1.8V (remember the advantages of lower voltage requirements) Flash memory chip, which is reported to be four times faster than existing solutions. Based upon Intel data, the typical phone in 1995 had roughly 2.5 Mbits of flash memory. Over the next four years, the amount of flash doubled every 24 months. Since 1999, the average amount of flash memory doubled every 18 months, with Japanese phones having over 40 Mbits of Flash in 2000, prior to the release of FOMA.
Software–Defined Radio. With a Software–Defined Radio (SDR), a number of RF functions that were previously performed by hardware are performed entirely by software. Conventional radios require an analog front end, analog-digital converter (ADC) and digital signal processor (DSP) for each protocol in order to convert the incoming analog RF signal to the digital domain, or vice versa. In the software radio, most of these radio functions are implemented in the software domain.
The software radio, which has already seen widespread use in military applications, offers an enormous advantage over existing radios since the hardware can be simplified and reduced considerably, resulting in lower costs. Of course, the military is more concerned about preventing jamming and eavesdropping than costs, so military SDR solutions are not necessarily less expensive. Additionally, the software can be modified and downloaded into the radio (base station/mobile phone) for greater versatility–new frequencies, channel-coding schemes, power control, smart antenna technology, different protocols. These parameters can also be adaptively changed to reflect the varying quality of the radio channel.
SDRs can be used in the base station as well as in the handset. As an example, the evolution to EDGE can be as simple as a software download from one location in the network to all of the base stations in the network. Although these radios are not necessarily full-up SDRs, this example illustrates some of the advantages of the SDR approach. In the handset, SDR-like functions already include DSP functions such as vocoders, which we discussed in an earlier section of the report. However, this software cannot be modified through a simple network download.
Additionally, some of the functions requiring higher speed and power requirements, such as A/D conversion, are still being performed by hardwired solutions. This cost can be significant since it is estimated that the RF portion of the handset can represent half the cost of the handset–obviously more with multi-mode phones. Further, it is estimated that the number of processing gates (think intersections in a circuit or the circuit’s complexity) jumps from nearly 50,000 gates with GSM to nearly one million gates with WCDMA. Go back and take a look at our “Physics equation” to see how the increased number of gates impacts power consumption.
SDR comes with its own share of challenges that have to be addressed. Of most significance are issues concerning dynamic range and very high power consumption.
Figure 8: Conventional Radios versus Software Defined Radios (Source: Nikkei)
The Handset Industry
Figure 9 highlights some of the major players in the mobile handset industry and how they interact plexity and magnitude of the industry, we do not claim that the chart depicts the complete picture or that all of the inferred relationships exist. However, the picture at least sheds some light on who some of the players are and how they interact with each other in the industry.
Figure 9: Putting Together Pieces of the Puzzle—The Mobile Handset Industry
Our Handset Market Forecast
When we published our first 3G report in September 2001, we commented that the task of forecasting 3G handsets is akin to being “a visually impaired gambler taking part in a crapshoot.” By that comment, we meant that it is extremely difficult to develop a forecast for a product that you cannot see or touch, let alone buy at your local retailer. In the last seven months, and in particular during the last three months, Motorola and Nokia have displayed (not demonstrated) 3G (WCDMA) handsets, as well as announced availability dates. In addition, we have had the chance to view a number of compelling, feature-rich 3G CDMA2000 1x models that should become available later this year in the U.S. market that are based upon similar models that have been available in Asia for the last several months.
Figure 10: Handset Methodology (DB)
Besides vendors introducing several new models of phones, the overall economy and the financial strength of major carriers has changed, but not for the better. As a result, we see a potential conundrum between the supply and demand curves developing.
We want to believe that subscribers will rush to adopt new technologies and purchase the feature-rich phones that will be flooding the market later this year. However, we also have serious doubts about the operators’ willingness to subsidize these more expensive phones and the subscribers’ desire to purchase the new phones, let alone subscribe to the data services possible with GPRS, CDMA 2000 1x and eventually WCDMA.
As we pointed out in our last report, our forecast is likely to change on numerous occasions, and it already has changed on three separate occasions since September (all downward revisions). It should, therefore, be evident that we have become even more conservative in our views about handset unit sales in 2002 and the near-term growth of the industry, both with respect to replacement sales and subscriber growth.
Before we jump to our forecast, we would like to present our methodology, which is more involved than our earlier forecast.
Our forecast divides the world into eight regions: North America, Western Europe, Eastern Europe, developing Asia, developed Asia, China, Latin America and Africa and the Middle East. For each region, we estimate the installed base for each technology (e.g., WCDMA, CDMA2000, GSM). The installed subscriber base from the previous year, combined with our assumptions regarding new subscriber adds, churn) and the percentage of used phones that are reused as replacement phones, then determines the number of new phone sales in the forecasted year. For the observant reader, you will notice that used/reused phones is a new category that we did not capture in our earlier forecast. We added this category after observing a meaningful percentage of new subscribers, especially in developing markets like China, purchasing phones on the second-hand market or reusing their phone when they switch to a different mobile operator.
Figure 11 plots our installed subscriber base forecast through 2007.
By 2007, we estimate the installed base will reach 1.7 billion subscribers.
Figure 11: Forecasted Installed Base by Region (DB estimates)
The annual change in the installed base is obviously due to new people joining the throngs of the wireless world. However, it is readily apparent that the forecasted number of new subscribers added each year (Figure 12) declines for nearly every single region of the world, although the total installed base, as depicted in Figure 11, continues to increase. We estimate that the CAGR for new subscribers is -12.3% (2001-2007), suggesting that growth in the industry will have to come from other contributors such as replacement sales or the adoption of new technologies like WCDMA and CDMA2000.
Figure 12: Subscriber Edition by Region (DB estimates)
Luckily for the industry, handsets are not a durable good and are replaced on a periodic basis. The question then becomes, “how often will a consumer replace his or her handset to upgrade to a new model, switch to a new carrier (churn), or to replace a handset that becomes lost, stolen or fails to meet the expectations of the subscriber?” (We have done that a few times ourselves.) Ideally, one would like to believe that “cool” new phones, like those on display recently at Cannes and Orlando, will encourage consumers to upgrade their existing plain vanilla phone and purchase an exotic flavored phone with lots of additional toppings (MMS, Bluetooth, color screen). Our view is that consumers will gradually alter their perception about mobile devices and be willing to spend more to get more, but it will not be an overnight paradigm shift (remember the early days of Palm?).
Since the replacement rate varies by technology, region and year, it becomes rather cumbersome to present all of the information. However, in Figure 13 we present the following information.
- The 2002 estimated replacement/churn rate, including used handsets, by region.
- The 2007 estimated replacement/churn rate, including used handsets, by region.
- The estimated replacement/ churn rate for developed Asia by technology for 2002.
- The estimated worldwide replacement/churn rate for 2001-2007.
Figure 13: Replacement/Churn Rate Assumptions (DB)
Figure 14 shows our current assumptions for the distribution of technologies for the installed base of subscribers through 2007. By 2007, we believe that 6.3% of the installed base will be WCDMA subscribers, 68.3% GSM subscribers and 23.6% CDMA subscribers.
Figure 14: Installed Base Forecast by Technology (DB)
Now that we have discussed most of our assumptions, we are ready to present our handset forecast. As Figure 15 illustrates, we currently estimate 386 million unit sales in 2002, 432 million unit sales in 2003 and a CAGR (2001-2007) of 7.0%. The global Deutsche Bank estimate remains at 410 million units in 2002 and 460 million units in 2003. In our September report, we forecasted 440 million units in 2002, so obviously our outlook has turned much more cautious. We also currently believe that WCDMA unit sales will grow from essentially no market share to 11% of all handsets sold in 2007. We also believe that CDMA unit sales will account for 27% of the market in 2007, with nearly all of these handsets being next-generation handsets.
Figure 15: Handset Unit Sales by Technology (DB)
Before we conclude the discussion of our handset forecast, we believe there are some more ways to present our forecast that are interesting. The first alternative, as depicted in Figure 16, is to show our handset forecast based upon new subscriber versus replacement sales. Based upon our assumptions, replacement sales, including churn, will represent 64% of the total handset market in 2002 and reach 86% of the total market in 2007.
Figure 16: Handset Unit Sales Forecast by Technology (DB estimates)
Figure 17: New Subscriber Sales versus Replacement/Churn Sales (DB estimates)
Figure 18: Voice Only Versus Data-enabled Phones (DB estimates)
We have also segregated our handset forecast by data-enabled phones and voice-only phones by technology. Obviously, WCDMA phones will all support data and eventually most GSM and CDMA phones will support GPRS and CDMA2000 1x, respectively.
However, for GPRS and 1x, the phones may support data services, but the subscriber may not actually use the phone for data. We believe this is especially true in the near-term since operators separately bundle the voice and data service, and make it a bit difficult to sign up for, and activate, the data service. However, we believe that at some point operators will bundle the voice and data services into one complete package.
Finally, we present Figures 19-22 that separate our GSM and CDMA forecast into its various components (GPRS and EDGE for GSM, and IS-95, CDMA2000 1x, and 1xEV-DO for CDMA). The subscriber distribution figures depict only those subscribers that actually use the data feature in their GPRS/EDGE (Figure 19) or 1x/1x-EV-DO (Figure 21) phone. Given our conservative view on data, our forecast depicts a wide separation between the number of data phones in the network and those that are actually used for data.
Figure 19: GSM Subscriber Distribution (DB estimates)
Figure 20: GSM Phones Sold By Technology (DB estimates)
Figure 21: CDMA Subscriber Distribution ((DB estimates)
Figure 22: CDMA Phones Sold By Technology ((DB estimates)
References
* MIPS is sometimes referred to as Meaningless Indicator of Performance, since it only provides a rough estimate for measuring performance.
* The primary difference between a DSP and a microprocessor is that a DSP is designed for tasks that require repetition and are numerically intensive (e.g., voice encoding). A microprocessor is more robust and flexible and is designed for control-oriented applications.
* “DSP-Based Architectures for Mobile Communications: Past, Present and Future,” Texas Instruments, IEEE Communications Magazine, January 2000.
* MPEG-4 encoding/decoding requires as much as 3.7 billion (with a “b”) operations per second for high-quality video if compression techniques are not used.
* Paul OuYang, Communication Systems Design, March 18, 2001.
Part two of this special report will continue in the September/ October 2002 Issue of Base Station/Earth Station. <<
###
- Eric - |
