Your browser is not supported. Please update it.

19 February 2008

Baseband chipmakers court Nokia with low power and strong integration

Nokia shook up the handset baseband processor sector last year when it named several preferred suppliers alongside the incumbent Texas Instruments, giving Broadcom, STMicro and Infineon a position in the world’s largest customer. Meanwhile, Motorola has fallen in and out and in again with Qualcomm, raising questions marks for its former chip unit and lead supplier Freescale, and Samsung is pursuing an accelerated agenda of its own, along with Qualcomm. How the business of the handset majors, especially Nokia, will be divided up among the various contenders is still far from certain, and some still regard the Nokia second-sourcing strategy primarily as a stick with which to beat TI, rather than something that will generate vast revenues for the other suppliers. Whatever, the politics, the chipmakers used the recent Mobile World Congress and ISSCC (International Solid State Circuits Conference) to set out their stalls and try to court the handset giants. And they know that the way to the hearts of Nokia and the others will be proven breakthroughs in two areas – low power and high degrees of integration, in order to deliver the key growth driver for the phonemakers, the low cost, high function mediaphone.

While TI and Qualcomm battle it out at the top of the handset rankings, with Broadcom desperately nipping at their heels, the most difficult challenges are faced by the European chipmakers and the beleaguered Freescale. The Europeans are mainly spin-offs from former national conglomerates ‘ NXP from Philips, Infineon from Siemens, STMicro originally a venture of France’s Thomson and Italy’s SGS. They generally have very wide product remits, large amounts of legacy business to support, and can rarely be regarded as agile. All are looking to steal more share of the cellphone from the US giants though, and are seeking to streamline their products, become more cost competitive and boast of some technical headstarts.

At MWC, Holland’s NXP had an uncharacteristically flashy stand heavily geared to multimedia and was aiming to convince the market that its often confusing wireless strategy is now clear and geared to growth. The wireless unit has had a shake-up in management and is now headed by Marc Cetto, who joined from Texas Instruments almost a year ago. Under his management, the unit has axed its cordless phone and multimedia coprocessor activities, and Cetto says there is now a clear focus, on single-chip solutions for ‘mainstream multimedia phones that require no multimedia coprocessors’. A major focus will be supporting ‘rock-bottom GSM to 3G handsets’, as Cetto put it at MWC, with single-chip devices integrated with RF, baseband and a power management unit.

On the connectivity side, NXP has defocused on Wi-Fi and is targeting technologies that it perceives as more mainstream to low cost or midrange handsets, notably Bluetooth, GPS, USB and FM. This is a company that is clearly avoiding the blue sky, dismissing UltraWideBand and NFC as ‘long term’, although it has resurrected its efforts in mobile TV (see Wireless Watch February 15 2008). At MWC, Cetto laid out three priorities for his stretched R&D budget – more aggressive development in cellular systems; a focus on mainstream multimedia features; and connectivity technologies with high attach rates in cellphones.

However, this is not all about safe, currently existing markets. NXP has bet considerable resource on one key product that aims to convince the handset makers that the Dutch supplier can be leading edge too ‘ and perhaps lure Nokia into the fold. This is the multimode, programmable LTE modem, the basis of NXP’s most important future-oriented product, its software defined radio system, Embedded Vector Processor (see Wireless Watch February 15 2008). NXP believes LTE is a chance to join the frontrunners, rather than lagging behind as it did in 3G and HSDPA.

Infineon is even more focused than NXP on the midrange and low cost handset market, hence its selection by Nokia as a second source for 2G products, and its strategy is bearing some fruit, with the communications unit seeing a turnaround in mobile and fixed markets last year, and delivering 51% year-on-year revenue growth in first quarter of fiscal 2008 (partly helped by the acquisition of LSI’s mobile chipset products last August). Overall, Infineon reported massive losses, and a 25% sales decline to €1.6bn ($2.37bn), but this was mainly down to poor performance at the Qimonda memory chip venture, in which Infineon holds a 77% stake, a level it intends to reduce dramatically, or it may sell its share altogether. Without Qimonda, Infineon achieved sales of €1.09bn, up 14%. With Qimonda, EBIT loss increased by 53% to €256m, but without that factor, operating results reversed a 2006 loss of €25m in the past quarter to a profit of €65m.

Infineon has been increasingly focusing on high volume GSM, backing away from the massive R&D burden of taking on Qualcomm and others in 3G and beyond. With volume the key to this game, it does need to bulk up its market share, and the €330m ($447m) cash acquisition of LSI’s baseband processor operations brought, LSI’s largest customer for mobile chipsets, within its sights. As well as the pressures to scale up in a commodity market, Infineon has an additional reason for buying market share ‘ the loss of BenQ Mobile. This company, formerly Siemens Mobile and the world’s fourth largest handset maker, was Infineon’s largest mobile chip customer before its insolvency under its new Taiwanese ownership. This was a disaster for Infineon that has only been partially offset by the Nokia success and a deal with new customer Apple for the iPhone digital baseband, RF transceiver and some power management components (Infineon controls 6% of the unit’s cost).

Now the German company needs to build on its progress in low cost mediaphones, and it recently introduced its new generation of handset chips, which bring multimedia features such as MP3 player, digital camera and mobile internet to low cost devices. The range consists of two options, X-Gold 113 and X-Gold 213, which are part of the vendor’s third generation ultra-low cost platforms XMM 1130 and XMM 2130. The 65nm chips integrate the baseband functions with power management, audio, an FM radio tuner and, in the case of the X-Gold 213, a digital camera interface for devices resolving up to 3megapixel and a driver for QVGA screens. An integrated 260MHz ARM 11 microprocessor core runs application software such as Java, messaging or other multimedia applications and the baseband also uses a Teaklite DSP. According to Infineon, handset vendors need only to add the memory devices and display. Infineon said its greatest challenge had been interference problems arising from signal coupling and crosstalk issues, highlighting an area of concern for the whole sector as handsets shrink and integrate more functions. ‘Digital crosstalk effects are becoming more significant as the technology scales down, and digital noise will become increasingly significant,’ said the company’s cellular RF team. Broadcom claims to have found one solution to receiver noise ‘ while quad-band 3G designs can require up to seven SAW (surface acoustic wave) filters, which can add to board size and cost significantly, Broadcom has described a 65nm W-CDMA transmitter using a feedback filter to suppress receiver noise while consuming just 65mW of power, and eliminating the need for a SAW filter.

While Infineon looks to the low cost market, STMicro is pushing its cellphone strategy in the opposite direction, trying to take on the giants in the 3G and 4G markets and targeting advanced smartphones. It gained the most valuable of the new Nokia baseband contracts, as this included a joint development initiative in 3G/HSPA that will help offset some of the STMicro R&D costs, and probably give it a guaranteed customer for a highly specified product. It has also recently turned its attention to LTE and WiMAX, and at ISSCC it described early prototypes on the way to creating a single chip that will bring 802.11n Wi-Fi and WiMAX into a cellphone, and even put the WiMAX/Wi-Fi transceiver and synthesizer on a single die, which would dramatically reduce power consumption. The chip would deliver 1Mbps data rates over a 10MHz channel, and could be tailored to handle LTE. Such products, STMicro hopes, will allow it to tap into the new market for WiMAX devices, with Motorola and Samsung clear targets for new business.

The tribulations of the European majors are sometimes mirrored at the US’ Freescale, which has also been going through the transition from being a subsidiary of a larger company to surviving on its own merits, and in private ownership. It recently replaced former CEO Michel Mayer with Rich Beyer, former CEO of chipmaker Intersil. The new CEO faces a major challenge – Freescale’s 2007 sales were down 10% to $5.72bn and its cellular chip sales, which totaled $1.22bn, were down 18% on the previous year. The main problems are slackening demand from Motorola, whose obligations to buy from its former subsidiary ended last year, and which has been spreading its favors to Qualcomm and TI, plus itself becoming a less attractive customer overall, with the crisis in its own handset sales. Freescale is even facing a slowdown in its other key market, automotive chips.

Brought in by Freescale’s private equity owners – Carlyle, Blackstone Group, TPG and Permira ‘ Beyer inherits an ongoing program to shore up the company’s finances, including a lay-off plan that has affected 700 so far. He will also spearhead a bid to broaden Freescale’s activities away from wireless, and to do this will seek acquisitions despite its $800m in debt, and being regarded as a risky credit. Under his predecessor Mayer, Freescale had already moved to expand its business, with actions like the $110m takeover of SigmaTel for its complementary analog ICs, which will open doors in the consumer electronics space. Over time, many observers believe Freescale will be broken up, something that Motorola considered when spinning off its chip business in 2004. All the three European majors could line up to buy some or all of the US company.

For now, however, wireless and automotive remain the core businesses and need to be strengthened, particularly in terms of gaining new wireless customers. Freescale was indicating two main approaches to this problem at MWC ‘ LTE handset chips and a focus on some low power digital home technologies.

Freescale hopes to differentiate its LTE handset offerings by also supporting base station reference designs, creating and end-to-end solution. The chipmaker has developed a development platform designed to speed and simplify the creation of equipment for LTE and other emerging wireless technologies, the chipmaker claims. The platform combines programmable data plane software with Advanced Mezzanine Cards featuring Freescale’s MPC8548 PowerQUICC III processor and the quad core MSC8144 StarCore DSP, to handle Layer 1 and Layer 2 LTE radio processing. It provides multi-gigahertz performance, low power operation, and interconnect technologies that will be needed to support high data rate air interfaces and IP packet handling in LTE base stations, claimed the company. This follows the creation of 802.16e reference platforms for base stations, part of an alliance with ArrayComm.

At MWC, Freescale demonstrated its LTE systems running streaming high definition video at low power, something that ‘illustrates both the growing maturity of the LTE specification and our unique implementation,’ according to Tom Deitrick, general manager of Freescale’s cellular products group. The company uses custom processors with microcoded engines for OFDM to boost data rates, and has also been optimizing its RF power amplifiers, programmable DSPs and PowerQUICC communications processors for LTE.

Meanwhile, Texas Instruments has to make some smart moves of its own to prevent the erosion of its market lead in baseband chips, which has been hurt by its loss of its monopoly position at Nokia and also at Ericsson and NTT, and further damaged by the current aggressive progress of Qualcomm. Qualcomm, of course, remains excluded from Nokia, the world’s largest baseband chip consumer by a rising margin, but the CDMA giant may be making moves to effect a truce, with several senior executives hinting heavily at MWC that they did not regard Nokia as a closed shop, despite the feud with Qualcomm’s licensing arm.

In 2006, TI held 42% of the wireless baseband business, according to Forward Concepts, followed by Qualcomm on 25%. In preliminary estimates for 2007, TI’s baseband share is projected to have slipped to 40%, and Qualcomm’s to have risen to 30%. As well as Qualcomm, TI is also under pressure from Broadcom’s ambitions to increase its currently tiny baseband market share, and also from Taiwanese companies moving into this space ‘ in particular MediaTek, which Forward Concepts says has been hugely strong against TI and other players in China.

TI’s response is to offer a new chip line, rely more heavily on outsourcing manufacturing ‘ with even the closure of its Richardson facility mooted ‘ and shift to a new R&D model. Responding to fabless chipmakers like Qualcomm, Broadcom and MediaTek, TI is far less noisy these days about the advantages of controlling advanced processes in logic (including baseband devices) and is starting to outsource more of this to contract manufacturers like TSMC, and focus on design excellence instead. Other baseband competitors that still rely on their own foundries include the three European giants and NEC but TI has been following a hybrid model for some years in its logic and digital devices units, though in analog, where it leads the market, it continues to rely on its own fabs and processes.

In product terms, TI is sampling a 45nm baseband design but has not released details, and so may be behind Qualcomm, which has already taped out such a design at foundry partner TSMC. Beyond 45nm, TI is moving to a hybrid approach in process R&D as well as fabrication, and has shifted its digital R&D to TSMC starting at 32nm, in a move to reduce risk and cost.