re: Deutsche Bank Securities on 3G (1 of 2)
Link courtesy of Peter Ecclesine.
Part 1 of this report summary continues with next post.
See link below for exhibits.
>> The Rise of the 3G Empire - Part 1 A
[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
DSPs And Microprocessors–Better Eat Your Wheaties
Back in the good old days of first-generation technology, DSPs operated at 5 volts and only had to handle 10 million instructions per second (MIPS),2 primarily because AMPS is an analog system. Then along came digital technology and the MIPS requirement quickly jumped by a factor of at least two to four. As wireless systems move from second-generation voice-centric circuit-switched networks to third-generation voice-and-data, packet-switched networks, the required processing jumps dramatically by as much as 1,000 times.
Next-generation phones will need to support multiple bands (e.g., 2000 MHz, 1800 MHz), multiple modes (WCDMA, GSM) and multiple modulation techniques (GMSK, QPSK), and handle multiple time slots with GPRS and EDGE in order to increase the data rate. Each of these functions places an additional burden on the DSPs (yes, there is more than one), the microprocessor, plus about 1 million additional ASIC gates for WCDMA processing.3 Motorola and Nokia both estimate that the radio channel processing requirements alone for WCDMA could reach 200 MIPS and that total MIPS requirements could reach as high as 12,000 MIPS. Although we have not seen any published numbers for MIPS, we believe similar processing power will be required for CDMA2000 (EV/DV).
Market Forecast
In May 2002, we published a 275-page report on 3G that updated our first report from September 2001. In our latest report, we provide expanded coverage of various next generation technologies, discuss recent market developments and provide our forecast for mobile handsets and infrastructure capital expenditures. We would like to share a few sections of our report with the readers of BSES.
While the global Deutsche Bank mobile handset estimate remains at 410 million units in 2002 and 460 million units in 2003, our current estimate for handset unit sales is 386 million units and 432 million units respectively. Our current estimate for handset unit sales is 386 million in 2002 (up 0.9% year over year), and 432 million in 2003 (up 11.8% year over year). We estimate the mobile device market growing to 576 million units in 2007. In this report, we discuss the market drivers that could lead to this growth, including new color screens and more feature-rich devices, as well as the factors that could influence our estimate over time. We also highlight some new product offerings that are, or will be available, in the coming months and years.
In our report, we break down our infrastructure forecast into regional and technology forecasts, as well as discuss the underlying assumptions in our models. We present a methodology for conducting sensitivity analysis on future 2G, 2.5G and 3G spending. We examine potential 3G network rollouts under a number of scenarios that take into consideration network coverage, capacity requirements, subscriber growth and the efficiencies of the underlying technology being used. Through our analysis, we believe that readers will develop an appreciation of all of the variables that can have an impact on the future growth of our industry. We currently estimate a 14.4% decline in 2002 over 2001, followed by a year-over-year increase of 3% in 2003. We believe WCDMA infrastructure spending in Western Europe will not ramp until 2004 as operators wait for the technology to prove itself in test networks during 2003. Subsequent growth in WCDMA is heavily dependent on the business economics of 3G, the uptake amongst consumers and their willingness to pay a premium fee for the service.
Given our outlook, especially for slowing subscriber growth and a more cut-throat pricing environment among carriers, we often wonder aloud and in our weekly Signals to Noise Newsletter: Do we really need six major operators in the United States, each blowing through billions of capital expenditure dollars? Do we really need seven major OEMs, several of which are currently struggling with their bond rating status? Do we really need a half dozen or so RF IC suppliers, all of whom are facing price pressure?
In a high-growth industry, such as personal computers during the early to mid 1990s, several major suppliers can exist and thrive. However, once the dollar spigot gets cut off and the flow of money runs dry, changes need to occur. In the near term, OEMs can offer heavy discounts and subsequently squeeze their suppliers in an attempt to limit the impact to margins.
The current pricing trend in the industry, however, cannot continue in our view. As any bug-eyed, newly anointed MBA graduate will tell you, sustainable growth, at the expense of cut-throat pricing and operating losses, is a recipe for disaster. One could argue that once industry growth returns, pricing strategies and operating profits will become the norm. We, however, believe that it will be too little, too late.
Nevertheless, we still believe that 3G will happen. Love it or hate it, next-generation wireless technologies are already available in some markets and should eventually become commercially available throughout most regions of the world. In the interim, while you are waiting to place your first mobile video phone call (don’t hold your breath), sit back, relax and take a read through our 3G report.
Critical Success Factors
We believe those companies that have “all the right STUFF” will benefit with the 3G evolution.
Scaleable. One of the biggest drivers for infrastructure spending, and a compelling reason for 3G, is network capacity. Already, a number of cellular networks are close to reaching their maximum capacity and will soon be unable to accommodate additional subscribers or increased minutes of usage without degrading the performance of the network. At the same time, wireless operators are reluctant to add unnecessary capacity or purchase equipment that cannot support increased traffic loads as network usage increases. Thus, we believe that winning equipment solutions will need to be able to handle the current demands of the wireless operator and will be able to support increased capacity in order to meet the growing capacity requirements of the network.
Transportable. Network traffic has to be moved (or transported) from the individual base stations back to the mobile switch (or between switches and/or other network access points), where it is routed to its intended destination. Currently, the transport cost can represent as much as 30% of the operator’s total network operational costs. With 3G applications and increased voice traffic, the amount of estimated data throughput can easily exceed current capacity constraints, or at a minimum make it cost prohibitive to use current transport solutions. As a result, carriers are turning to more cost-effective means to transport traffic throughout their network, with microwave radios and fiber optics being the two most popular solutions.
Upgradable. The evolution to 3G is not one giant leap, but a series of sequential steps that vary by technology or by the business strategy of the wireless operator. Obsolescence is a four-letter word and carriers want to avoid at all costs deploying equipment that cannot be quickly and cost effectively upgraded to offer enhanced services. Additionally, as wireless operators select their ultimate 3G technology, they will base their decision, in part, upon the required upgrade path that takes them to their final network solution.
Flexible. Networks and network equipment have to accommodate numerous technologies along the path to 3G. In particular, networks have to support 3G applications, while at the same time provide 1G, 2G and 2.5G service–perhaps even from the same location or from the same mobile device. Major Original Equipment Manufacturers (OEMs) and suppliers (including multi-mode/multi-band handsets, Radio Access and Core network equipment and components) need to provide flexibility in their architecture in order to meet the demands of 3G while supporting the services available today.
Footprint. Ask any corporation or individual looking to lease property and they will tell you that property rental rates are at a premium, when and if they can find a location that satisfies their needs. Wireless operators face the same predicament. As they roll out their 3G networks, operators will require new locations where they can install their equipment (sometimes sharing a location with another operator). In many instances, the amount of available floor space will be limited, and in almost all cases it will cost them a king’s ransom. Thus, operators demand equipment that is compact and that can fit in a small place–in other words, the equipment needs a small footprint.
You’ve Got the Whole World in Your Hands–The Future of Mobile Handsets
No discussion on 3G would be complete without a discussion of mobile handsets. Our discussion begins with a glimpse at the future of handsets and some of the innovative applications these handsets will support. We next turn our attention to some of the handset challenges that are being addressed today in order to provide expanded functionality tomorrow. Finally, we will present our forecast for the handset market and the underlying assumptions that can influence the accuracy of our forecast.
Future Mobile Phone Capabilities
Do you want to know what the future holds for the wireless handset? You need to look no further than the Japanese or Korean market today, blend in a few more existing technologies like Bluetooth, and then sprinkle on top some of the wireless applications that are currently in their infancy, but becoming commercially available in the near future.
You may not believe that mobile teleconferencing will ever become the norm (or, for that matter, affordable)–a point we might concede, at least in the near term. However, there are still a number of new mobile phones and whiz-bang features that will take advantage of a packet-based network and be enhanced by higher data rates–even if the actual data rates fall short of the marketing hype.
Buy a cellphone for a pet. In Japan, people are able to buy a small cellphone that they can attach to their pet’s collar in order to find them using location-based tracking when they wander off from their owner.
Location-Based Services (LBS). Today in Japan, all three major carriers offer some form of location-based services. J-Phone has J-Navi (yellow pages) and J-Skystation, which is a pushed-message service (e.g., nearby sales inside a shopping mall). In fact, all J-Phone’s phones are automatically equipped to receive an emergency LBS message such as a building evacuation notice. NTT DoCoMo launched i-area, which allows users to find out where they are located and then follow a series of links to explore the area (museums, restaurants, etc). Finally, KDDI already offers its EZnavigation service, which allows users to check on local traffic conditions based upon their location using Qualcomm’s SnapTrack solution.
Download video and audio clips. Today, on the FOMA network, and using the N2002 phone, subscribers can download movie clips, sample music from the hottest music CD and watch recent news clips.
Text recognition is a man’s best friend. Despite the claimed simplicity of T9, we have personally grown accustomed to our Palm-based Kyocera phone, which uses a stylus. With the stylus, we can use the “hunt and peck” method to type a message or preferably we can use graffiti, which is relatively easy to learn and much faster once you have it mastered. Even more advanced than graffiti are the smartphones, which have character recognition software.
Nearly every handset manufacturer already offers some form of speech recognition. Currently available speech recognition features primarily support voice-activated dialing and answering of phone calls. However, some smartphones/PDAs could eventually read your e-mail to you or dictate an e-mail message.
Figure 1: A Glimpse at a Few Futuristic Concept Phones (Source: Company Info)
3G Handset Challenges– A Few Hurdles Still Have to Be Cleared
In the seven months since we first published our 3G report, it seemed like barely a week went by without another 3G licensee announcing that it was delaying its 3G launch because a commercial handset was not available. At the same time, some European carriers have launched their networks without handsets in order to comply with the requirements of their 3G license.
Since February, Motorola and Nokia have unveiled their 3G WCDMA handsets. At the same time, most data points suggest that it will be at least 2003 or more likely 2004 before a 3G handset is available in commercial volumes that is acceptable to users. Based upon our personal experience with the Motorola handset, it still has a ways to go before it is ready for prime time (we witnessed it locking up on two separate occasions, even though it was not even connected to a network). Additionally, the Nokia 3G phone will initially be only available within a portion of Sonera’s network, which is entirely provided by Nokia.
Depending on how you look at it, tomorrow’s handsets represent today’s challenges, or today’s opportunities, for companies that are targeting products and solutions for the handset. Let’s take a few minutes to look at some of the demands being placed on handsets and how new technologies are addressing the challenge.
Handsets 101
Before we can discuss 3G handset challenges, we need to provide you with some detail on how handsets actually work. In the interest of saving trees, we are going to skim over the details and focus on what we consider to be the most important aspects of the handset.
Figure 2: The Innards of a Multi-band, Multi-mode Handset (Source: DB)
We can divide the handset into three functional areas: the baseband, the IF (intermediate frequency) and the RF (radio frequency). Here is a very short synopsis of each functional area.
Baseband. In order for the handset to transmit your voice, an audio digital converter (ADC) converts your voice to a digital signal (1s and 0s). Next, a digital signal processor (DSP) is used to compress your voice according to the appropriate network protocol (GSM, CDMA, etc.), of which each may have more than one compression standard. The DSP(s) also performs speech decoding/encoding (the vocoder), echo cancellation and synchronization. Next, the appropriate channel-coding scheme is applied to the digitized voice signal, along with redundancy and some form of encryption. The digitized signal is now ready for the IF portion.
In addition to processing analog voice signals, the baseband section is responsible for other attributes of a handset. The power management function controls battery charging and monitoring, as well as regulating the voltage for all of the phone’s circuits. A separate microprocessor provides the PDA and smartphone functions, such as PIM (Personal Information Manager and multimedia applications). Flash memory stores all of the code for the microprocessor, as well as storing Java and BREW applications. Finally, the Static Random Access Memory (SRAM) or, in some cases, Dynamic RAM (DRAM) stores information that can be more rapidly accessed, but loses its memory when power is not supplied.
Intermediate Frequency. Prior to transmitting the signal, the baseband signal needs to be stepped up to the carrier frequency (e.g., a PCS frequency). Generally, this is a two-step process that requires an intermediate step. A similar analogy would be to have a ledge six feet up a 12-foot wall that makes it easier for climbers to scale the wall. To maintain the desired signal and to prevent unwanted noise, various IF filters, such as SAW filters, are also used.
Qualcomm’s radioOne technology, also called ZIF (zero IF), converts the RF signal directly to a baseband signal without the need for first converting to an IF signal. As a result, the IF circuitry and a SAW filter are eliminated from the design. According to Qualcomm, this feature reduces overall handset costs, reduces the area required to house the entire transceiver function by approximately 50%, improves talk time by an estimated 20% and increases standby times nearly fivefold over current CDMA handsets. For a number of years, companies have sought to eliminate the IF portion of a radio, and it remains to be seen whether or not Qualcomm’s solution effectively accomplishes that task. However, we have reviewed its approach, and we believe that it will do the job. Figure 3 illustrates the top and bottom sections of a mobile phone, both before ZIF and after ZIF. Although we have included the MSM, power management and memory, we want to draw your attention to the radio frequency (RF) section. It should be apparent that Qualcomm has reduced the area and bill of material (BOM) required to house RF functionality with the new ZIF architecture.
Figure 3: Qualcomm’s radioOne Architecture (Source: Qualcomm and DB)
Radio Frequency. The RF portion includes the antenna, duplexer (allows simultaneous transmit and receive to occur) or transmit/receive switch, the transmitter and the receiver. The transmitter modulates the IF or baseband signal to the appropriate carrier frequency and then uses power amplifiers (PAs) to amplify the signal so that it reaches the base station. Due to the amplification process, transmitting a signal places a significant drain on the battery in comparison to receiving a signal. Recall that one of the advantages of CDMA 2000 1x is improved power control. This ensures that the most efficient transmit power is used between the base station and the mobile devices (both near to and far from the base station), thus adding capacity to the network by minimizing interference and improving spectrum efficiency.
Turning to the receiver, low noise amplifiers (LNAs) are used to boost the signal that is received at the antenna. Additional filters are used to remove spurious and unwanted noise. One of the key aspects of a good receiver is the ability to maintain a good signal-to-noise (S/N) ratio at both low and high signal levels.
Qualcomm Chips–What Do They Really Do?
Given that the general perception is that Qualcomm and CDMA are so closely linked together, you may wonder why the company does not change its ticker symbol from “QCOM” to “CDMA.” However, although investors recognize the synergy between Qualcomm and CDMA, we believe investors generally do not really understand what the company actually provides. [1]
[1, To be fair, Qualcomm is one of a myriad of companies that make ASICs for mobile phones (e.g., Texas Instruments). However, within the mobile world, Qualcomm seems to gets the most notoriety for its CDMA ASICs.]
Qualcomm designs and manufactures ASICs (Application Specific Integrated Circuits) as well as licenses its CDMA technology to other companies who in turn manufacture their own ASICs or use CDMA technology in their products. Qualcomm’s primary ASICs are its modem chips that are located in the handset and in each base station. The handset modem chip product name begins with MSM (mobile station modem), the microprocessor begins with MSP (mobile station processor) and the cell station ASIC begins with CSM (cell site modem). Figures 4 & 5 provide a sampling of the various modem chips and their functionality. Qualcomm offers Wireless Internet Launchpad software applications consisting of multimedia (Qtunes, QTV, Compact Media Extension for text and animation) User Interface functions (SIM/UIM card interface, PureVoice Audio, voice recognition), connectivity (Bluetooth, Java engine, Secure Sockets Layer), positioning with Snaptrack/gpsOne, and storage.
Figure 4: A Sampling of Qualcomm’s Family of MSM ASICs (Source: Qualcomm)
Figure 5: A Sampling of Qualcomm’s Family of CSM ASICs (Source: Qualcomm)
In addition to all of the baseband and IF processing, the MSM provides a host of features, to include providing an interface to the keypad and external peripherals (PC), converting (modulating) analog signals from the microphone into digital signals (vocoder) and managing power requirements.
On higher-end multimedia phones, a microprocessor (MSP) interfaces with the MSM. The MSP is designed to bring PDA/smartphone functionality to the handset. Additionally, the MSP allows third-party developers the ability to develop applications (Web browsing, JAVA, MPEG-4) without having to deal with CDMA-specific operations.
The processing requirements for 3G phones are distributed across three types of processing: microprocessor (MPU), DSP and signal processing accelerators. The MPU supports the user interface, the communication protocol and the applications made possible by 3G. The DSP must support increasingly complex communications requirements, as well as voice code processing. In a January 2000 IEEE article, Texas Instruments states that only 10% of the processing power with 3G can be accomplished on a DSP, whereas with 2G it is 100%.4 Thus, 3G phones will require additional horsepower that is nearly 10 times beyond what today’s DSPs are actually capable of providing. The remaining 90% of physical layer processing requires application specific processing acceleration. Figure 6 illustrates the distribution of 3G functionality across processing implementations.
Figure 6: Distribution of 3G MIPS (Source: Morphics)
The increased processing complexity poses a significant challenge to power consumption, time-to-market and cost. OEMs and chipset vendors are striving to deliver the integration required to keep handset costs low and still maintain long battery life, but the challenges are significant. One of the biggest challenges is in the dedicated hardware accelerators. While ASICs could be used in the past, ASICs lacked flexibility, which is unacceptable in the face of shortened time-to-market and rapid standards evolution.
One private company, Morphics, is approaching this challenge with programmable accelerators that can be assembled with the other Intellectual Property (IP) blocks for flexible, low-power 3G chipset solutions. The Morphics IP core can support multiple 2G and 3G standards and can be reprogrammed to adapt to 3G standard evolutions. With a Morphics solution, OEMs and semiconductor vendors can focus on system differentiation, optimizing their platforms for different applications, while at the same time accelerating time-to-market and lowering development costs.
This application layer of a 3G platform is where the general purpose MPU and DSPs must be employed. Unlike voice only phones, the 3G applications cover a range from simple messaging to real-time video. Figure 7 from Intel provides a nice illustration of how added functionality increases the need for more MIPS.
Figure 7: Applications Drive MIPS (Source: Intel)
To address the need for more MIPS, new microprocessors are being introduced that add more horsepower while consuming less energy. The U.K.-based company ARM provides its processor core technology to a myriad of companies in the mobile industry. It is estimated that at least half of all mobile phones contain some form of an ARM processor. ARM technology serves as the technology behind Texas Instrument’s Open Multimedia Application Platform (OMAP), a platform that Sony Ericsson is using in its flashy P800 smartphone that utilizes the Symbian architecture.
Last year, Intel and Analog Devices announced a solution that combines DSP and microprocessor functions on one chip. The DSP operates at up to 400 MHz and incorporates dynamic power management to improve battery life. Intel also offers its StrongARM SA1110 microprocessor (used in the Compaq iPAQ) which operates from 1.75 volts when running at 206 MHz (the importance of frequency and low voltage will become apparent in the section on batteries) and can deliver 750 MIPS. Intel has also taken its microprocessor product offering to the next level with its XScale microarchitecture. The XScale builds upon the StrongARM architecture and features improved power control and higher performance (1,300 MIPS at 1 GHz, operating at 1.8 volts). Moore’s Law is clearly alive and well.
Qualcomm’s MSP1000 (Mobile Station Processor) is another example of ARM technology at work. Qualcomm uses an ARM720T microprocessor to provide extra horsepower to increase the data rate of its chip sets, as well as to support additional features such as MPEG-4 and MP3.
Leading DSP providers include Analog Devices, Infineon Technologies, Intel, Motorola, STMicroelectronics and Texas Instrument, with TI currently holding a commanding lead at over 60%.
Battery Life–Sometimes the Energizer Bunny Cannot Come to the Rescue
In the mid 1990s Nokia was advertising its 638 phone, which provided 2 hours of talk time and 26 hours of standby time. An additional extended-life battery increased talk time to 3 hours, 20 minutes and standby time to 47 hours. Jumping ahead to more modern times, Ericsson claims its R520 phone provides 25 hours of talk time and 715 hours of standby time. Even the popular Sony Ericsson T68i has up to 13 hours talk time and 290 hours of standby time, despite having a color screen. On the CDMA side, CDMA2000 1x-enabled handsets can increase the standby battery life by up to 50%, although the power control features of 1x that are used to increase capacity/data rates actually reduce talk time battery life. The net effect is still longer battery life. To quote Bob Dylan, “the times, they are a-changing.”
However, emerging applications like MPEG-4 decoding5 and GPS can rapidly drain the battery, not to mention a color screen, which can consume far more power than its monochrome predecessor. Furthermore, one of the underlying principles behind GPRS and EDGE is operating for longer periods of time, which means the transmitter is transmitting for a longer continuous period of time or the receiver is receiving for longer continuous periods of time. In either instance, a tremendous demand is placed upon the microprocessor, which has an impact on battery life. Improved power control features and color screens that have low-power requirements, such as the Philips TFT-LCD, help, but new battery technologies may still be required.
Battery life, until the next recharging, is determined by its energy capacity, which is expressed as a multiple of watt hours (Wh). For example, last December Samsung SDI announced that it had developed a battery, based upon lithium-ion technology, that can store up to 350Wh. If we assume two batteries operate at equal voltages, then the battery with the greater milliamperes per hour (mAh) rating should have a longer battery life. For example, Nokia offers an Ultralife Polymer battery that has 2,300 mAh, although most batteries have a capacity between 600 to 800 mAh. Of course, the other mitigating factor is how efficient that phone consumes power, both in standby and in talk mode.
Today’s premium phones consume between 2 to 4mA in standby mode and 100 to 150mA in talk mode.6 We did the math, and that means a premium phone that consumes 3mAh with a 750mAh battery should have a battery life of 6 hours (750/125 = 6) in talk mode and 250 hours (750/125 = 6) in standby mode. Thus, the increased current required by faster DSPs and microprocessors is an important feature in addition to typical performance attributes such as processor power and memory capacity. One of the reasons Intel announced its one-chip solution was that by placing the DSPs and memory on a single chip, additional power would not be consumed to support the flow of information (via the bus) between different chips that housed separate DSPs and memory devices.
So far we have avoided using a physics equation in our discussion on 3G, but we are going to have to make an exception at this point.
Power = #Gates ¥ Capacitance ¥ Voltage2 ¥ Frequency
In order to preserve battery life, the handset needs to conserve power, sort of like Californians need to turn off the lights in unoccupied rooms in order to prevent rolling blackouts. All things being equal, a faster-running processor/DSP drains the battery more quickly than its predecessor does (the frequency increases). However, the saving grace is the voltage term, which is squared in the above equation. If the operating voltage is reduced by a factor of 2, power consumption falls by a factor of 4 (22 = 4). Perhaps now it is apparent why companies such as Intel stress power and operating voltage requirements in addition to raw processing power.
There have been a number of battery technologies that have evolved over the years. Today, the two dominant technologies are lithium-ion (LiON) and nickel metal-hydrides (NiMh), with LiON rapidly winning the battle due to its ability to maintain its capacity after frequent recharging. It also appears that next-generation batteries will be based around an enhanced lithium technology, such as lithium-polymer, which reduces the size and improves the talk and standby times. Major LiON suppliers include Samsung, Sony, Toshiba, Sanyo, NEC, Matsushita, Panasonic and Japan Storage Battery.
Although a bit futuristic, batteries that utilize fuel cell technologies may one day find a place in 3G phones. Companies, including Motorola and Samsung, are developing fuel cells that can provide much longer talk times than today’s lithium technology. The fuel is typically methanol, but even carbon- or hydrogen-based solutions are being considered. Don’t look for a fuel cell for your mobile phone any time soon. This technology, if it pans out, is still a few years away (but then again, so is UMTS).
Continued Next Post:
Message 17928092
- Eric - |
