SAN DIEGO ( ChipWire) -- The market for Bluetooth radio chips will get even hotter this month as a pair of startups begin peddling their reference designs.
Silicon Wave Inc. here claims its Odyssey radio modem IC is ready for production. Pairing the Odyssey with a Texas Instruments Inc. baseband IC, Silicon Wave is showing a reference design suitable for Palm handhelds and other portable devices.
Competitor Cambridge Silicon Radio Ltd. in Cambridge, England, meanwhile, is promoting its own integrated radio and baseband solution and hopes to be in full production later this month. The company, which is setting up U.S. operations in Richardson, Tex., claims its part requires no special packaging and eliminates many of the passive components associated with RF tuning circuits.
Both startups are vying to be first in a crowded field that includes IC contenders like Philsar (now part of Conexant Systems Inc.), Innovent (acquired by Broadcom Corp.), Atmel Corp. and National Semiconductor Corp. All want slots in a Bluetooth device market that could be as big as 103 million units in the United States and 450 million units worldwide, according to research firm International Data Corp.
A key to Bluetooth's success, market participants agree, is cost. For this short-haul wireless network to become ubiquitous, the price of hardware and firmware for an entire Bluetooth module must aim for the $5 mark, according to the initial members of the Bluetooth Consortium. This necessitates a single-chip Bluetooth radio, if not a solution that integrates radio and baseband processor.
The cost of a version from Ericsson Components, housed in a low-temperature cofired ceramic package, is in the $27 range. Because of the number of passive components that must be included in the ceramic, the company is not optimistic that the price of this ARM-based module will drop much below $14 by 2003.
Thus, to approach the $5 goal, chip makers must not only come up with single-chip solutions, but ones that use low-cost plastic packaging and eliminate all the extra passives. Cambridge Silicon Radio and Silicon Wave are among the most ambitious in this regard, claiming their radio chips can be mounted on low-cost FR4 pc-board materials, with inexpensive chip antennas and minimal passives. Both companies are offering development samples now, promising full production shortly and providing road maps for future integration. But both are predictably cagey about how a
frequency-hopping spread-spectrum radio operating in the 2.45-GHz ISM band can be implemented on a single chip for $5.
In principle, a direct-conversion architecture would allow the capture of an RF signal, convert it (untuned) to digital and use a digital signal processor to segregate the data-carrying modulation pattern from the RF carrier frequency. In practice, even the high-end data converters used in cellular basestations do not have the sample rate necessary to capture RF signals -- and their cost and power consumption make them unsuitable as yet for handheld consumer devices.
A superheterodyne architecture extracts a modulation signal from a radio-frequency carrier by mixing the captured RF signal with an exact inverse of the transmitted RF carrier. The mixing must be completed in successive stages to bring a 2.4-GHz signal down to baseband frequencies, and dozens of inductive "tank circuits" are typically required to lock the desired signal and to filter harmonics and intermodulation by-products.
Speculation suggests that semiconductor makers like Cambridge Silicon Radio and Silicon Wave are using some combination of superheterodyne and direct intermediate-frequency conversion to shrink the radio down to size. If, for example, the captured RF signal were mixed with an ultra-precise inverse of the RF carrier, the modulation pattern would fall out with a minimum of extra tuning components, and this could be fed directly to an A/D converter and DSP to extract the voice or data patterns.
Cambridge Silicon Radio chief executive officer John Hodgson and applications vice president Eric Janson would neither confirm nor deny that this was the technique employed. But they did note that the company's technical staff includes mathematicians from Cambridge Consultants Ltd. who have many years of experience in synthesizer design.
Hodgson and Janson said that the BlueCore01 -- which includes the radio, baseband processor, microcontroller and scratchpad memory -- is a "near-zero-IF" design, and it requires just a few external
capacitors and a crystal frequency reference. "Just a crystal, not an oscillator," Janson said. The inductive filter elements and baluns can be incorporated into the printed-circuit board traces and, when implemented in a GSM phone, even the 13-MHz crystal can be eliminated, he said. "RF-to-UART is what the user sees," said Janson.
Cambridge Silicon Radio is looking for the lowest-cost design, the lowest bill-of-materials cost and the lowest production costs, said Hodgson. By using an all-CMOS design, he believes he can sell the
BlueCore01 design for $8 this year, and a BlueCore02 design -- a single-chip solution that incorporates flash memory for protocol stack dissection -- for $5.50 next year.
The BlueCore01 design is fashioned in 0.35-micron CMOS and will be shipped in 100,000 quantities later this year. A Bluetooth module can be implemented with about 15 cents' worth of pc-board materials, a
10-cent copper trace antenna and 15 cents' worth of additional components, the company said. The company will build the BlueCore02 in 0.25-micron CMOS.
Meanwhile, Silicon Wave's engineering chief, Steve Brown, said that an intelligent, self-trimming frequency synthesizer is the key element in that company's single-chip BiCMOS transceiver design. Silicon Wave does not use a fractional-N synthesizer, because it uses too much current, said Brown, but rather, a very precise self-calibrating voltage-controlled oscillator. To keep current consumption low, the frequency is synthesized in what Brown called "small bursts."
Silicon Wave's reference design uses pc-board traces as the inductive filter elements. "There is some debate as to whether lumped or distributed inductances are better," Brown acknowledged. The applications circuit he devised requires no external inductor components. It does need decoupling capacitors, but their number depends on the application, he said.
TV design techniques define the optimum number of decoupling capacitors via trial and error: The designer starts with as many as 20 capacitors and progressively removes them until the design becomes unstable, he explained. The goal is to find a stable design that uses the minimum number of external components.
The big advantage of BiCMOS is that it offers better performance in interference-prone environments like cell phones, Brown said. "It's a solution that's easy to use, offers good performance and iscost-effective," he said.
Other startups in the Bluetooth arena will have big-company support in their efforts to catch up to Cambridge and Silicon Wave. Innovent in El Segundo, Calif., with expertise in digital tuning and calibration, is an arm of Broadcom, while Philsar in Nepean, Ontario, has been acquired by Conexant Systems Inc. Philsar's PH2401 Bluetooth RF transceiver is implemented in BiCMOS with silicon germanium bipolar transistors.