Chipsets provide the fundamentals of mobile phones but in today’s marketplace, basic functionality is no longer enough. Catherine Haslam reports on the ever-increasing demands being put on the producers of chipsets and the impact these are having on the market structure.
All silicon manufacturers face the same pressure — to produce more for less. However, the mobile terminals market is even more exacting in its demands. As a result, silicon prices are falling at an average of 6% per quarter and every new technology needs to be introduced at the at break-neck speed. Differentiation and growth therefore comes from adding value in terms of new functionality and greater integration.
Whereas this challenge would once have been the concern of OEMs, the structure of the market means that OEMs and indeed, ODMs, are putting more pressure on the chipset players to provide complete solutions.
Those requirements manifest themselves in two clear trends. The first fits easily into the traditional remit of chipset developers as it is all about reducing the baseband processing power required by the RF functions to free it up for other functions. The second is integrating those other functions into a system design to create a complete platform.
On the first point, Julian Hildersley, managing director of TTPCom’s Silicon Business Unit, explains that there is a desire to change the partition of the analogue interface between the RF and baseband sub-systems. “The analogue interface is being challenged…by a digital interface which will split the mixed signals into baseband and RF. It’s in the early days — about 18–24 months away — but it represents a significant change.”
Getting connected
There are also a number of new connectivity technologies that are impacting on the handset market, such as Bluetooth and 802.11, which are often pitched as complimentary to cellular systems. As a result cellular devices which support these as well are hitting the market. However, supporting these functions needs to be done with as little impact on the overall functionality of the terminal. After all, no-one will want WLAN functionality if it doubles the cost of the device and cuts the battery life in half.
The cheapest way to deliver such new functionality, in monetary and power terms, is to integrate the new chips with existing ones for GSM/GPRS. However, that means that the cost and power consumption of the integrated chips are higher than the GSM/GPRS-only version. Therefore, it only makes sense to do this once the functionality is required by a significant proportion of the market.
According to Phillips’s marketing director for 3GSM, Ton van Kampen, “When features become mass market you can combine the chips as this is a lower cost option than keeping them separate.” While experts vary in their definition of what makes up a critical mass, 40–50% seems typical. However, in the case of Bluetooth which is now a prerequisite for high-end terminals at least, there are some who question the sense of integration. “The investment to integrate is still quite substantial and it is still not being pushed by the market,” says Hildersley.
Motorola’s Mike Philips goes a stage further and states, “It doesn’t always follow that integrating air interfaces is the best option. We are hearing that OEMs don’t want to integrate these because the evolution of Bluetooth standards are asynchronous with those of cellular technologies. This means that there would be more changes required and handset manufacturers are trying to keep these technologies separate. Multiple air interface integration is therefore about GSM, GPRS, EDGE and UMTS.”
In the past, multiple RF has been about supporting additional GSM bands — 850MHz, 900MHz, 1800MHz and 1900MHz. GPRS has been added and UMTS has made its first entry into the market. It has done so in the knowledge that, with the exception of PCMCIA card chipsets, UMTS RF and baseband will need to work in a dual environment with GSM/GPRS. As a result, one of the first issues is whether or not the UMTS and GSM/GPRS functions should be combined. According to Werner Sievers, ceo of Zyray Wireless, “Ultimately, high volume dual-mode single chips are the aim and will be cheaper.” But, he suggests that time is not here yet, “Right now single chip are in the USD50-60 price range. GPRS chips are around USD10 each.”
The Zyray solution is a UMTS (FDD) co-processor which has been demonstrated working with Infineon’s GPRS chip and is easily integrated into the existing chipset because it looks like a piece of memory. Qualcomm, on the other hand is already on its third incarnation of UMTS/GPRS/ GSM chips — the 6250, which boasts a highly integrated baseband to reduce the size and power requirements.
However, as all are agreed, the RF capability is just the beginning, which brings us on to the second trend — supporting new application functions. This is not so straightforward and has created two distinct approaches to market. The first sees some of the larger developers, including Motorola and Samsung, developing the entire platform in-house, while the second sees a range of fluid and formal partnerships, aimed at providing the same complete solution.
These new radio technologies have been created for a reason and that is to carry data services effectively. That means that terminals need to support a step change in functionality. Even in the entry level market, polyphonic sound and colour are becoming standard. Java games and downloads are being added and all of these functions have implications in the chipset — on power, on processing requirements and therefore on the system design, the size of the unit and the price.
Early movers
As Mark Frankel, senior director of product management of Qualcomm QCT explains, Qualcomm is a long way down this road with Launchpad, an integrated chipset solution for multimedia applications. This includes a music synthesiser for MIDI, an MP3 player, JPEG encoder/decoder and MPEG4 decoder (an encoder is in development), all of which run on an ARM and DSP core. From this it is clear to see that Qualcomm is determined to integrate high-level functionality early, something that is epitomised by the company’s early inclusion of location technology.
According to Arnold Gum, senior product manager for position location, this early move (at least for the European market) has enabled the company to gain a clear cost advantage. Citing an incremental cost of USD3–4 for adding the GPSOne chip, he states that, “It is cheaper than putting in an MP3 chip.”
However, these decisions to integrate functionality are reactions to the market. They may be reactions to requirements in CDMA markets but they are far from being gambles on the future. Legislative requirements for location information in the US and the video and demands of the CDMA markets in the Far East, mean that a great deal of experience has been gleaned in integrating functions that are now wanted here in Europe. Yet this does not mean that it will instantly port to the requirements of Europe, or that others cannot meet Europe’s needs with equally cost-effective solutions. For example, Qualcomm has integrated its own applications platform, BREW, but runs Java, Windows CE and Symbian on top of that, whereas Nokia’s Series 60 fully integrates the Symbian OS.
The dominant trend is clearly towards platforms rather than chipsets. As Motorola’s Philips explains, “Handset manufacturers want complete platforms, rather than developing these themselves. This is because they have fewer people and also because the value of the terminal is in the look and feel and applications. They don’t see platforms as their differentiator …The trend towards platforms is extremely strong and if you want to be a major chipset supplier, you have to be able to play in this arena.” However, he suggests there are few companies capable of doing this single-handedly.
While this is not disputed by the like of Wavecom and TTPCom, the value of the single provider is. “Developing all the elements yourself is costly,” says Guy Lanrezac, marketing director at Wavecom. He continues, “We have taken an approach which provides a complete solution to OEMs. This is based on the best baseband and RF on the market. At no one time has anyone got all the best products.”
Motorola’s Philips is, not surprisingly, adamant that this is not the case. “Only a few players can do it all and this allows us to make key architectural decisions. Others will only be as strong as the weakest link…Partnerships are good because they can deliver a total solution but you have to get a number of companies working together and that slows things down,” he suggests.
A further problem in the multiple supplier approach is that it can create a tangle of Intellectual Property (IP), characterised by claim and counter-claim. For example, a single terminal may well have four technology providers for Java with four IP claims. Multiply this by the number of handsets functions and the situation is messy to say the least. However, it is not an issue that has passed Wavecom by as Lanrezac explains, “Wavecom sorts out all the IP issues for the OEM and presents it collectively for the platform.”
The market must decide which approach best meets its needs but, whether one or several companies are involved, the task of integrating new functions is about making decisions on what should be integrated when and how. This, Philips claims, makes up 50% of the work. New hardware means new software and new integration, therefore developers don’t want to change the hardware that often. Yet, on the other hand, changing the hardware is the most efficient form of integration. Getting the balance right is far from easy.
There are constant choices to be made, as van Kampen illustrates with the relatively simple example of MPEG4. “This could be done in software on the baseband as this is the lowest cost option but the baseband chip has limited power and throughput. Therefore, sometimes it makes sense to have a companion chip with a separate interface. Alternatively, a half-way house can be achieved using an accelerator chip.”
MPEG4 is just one element of functionality and it is clear from van Kampen’s explanation that there is no simple answer. Every chipset developer has to decide what functionality will be supported for each range of chipsets but more than that, it must work out how to optimise this for the handset manufacturer.
As you then move up the handset scale, to picture and multimedia services and add in more complicated Operating Systems (OS) such as Symbian, Pocket PC or Palm, the integration between the lower layers, applications and OS becomes massively more complicated and ever-more important.
Designing these systems is a real black art, a constant balancing act which defines what should and shouldn’t be supported and one which is particularly time sensitive. Getting the right products to market at the right time is crucial. Lanrezac explains, “If the OEM misjudges the end of a handset’s life, it can lose up to 50% of the gross margin for that product.” Therefore, chipset vendors have to be able to respond with system designs that provide the flexibility to add new functions at a speed the market requires, while maintaining reliability and quality. The price of chips may be cheap but the right integration is priceless.