Bluetooth gear could take years, say confabers
Source URL: eet.com
By Stephan Ohr
EE Times
(04/07/00, 9:43 a.m. EST)
GENEVA ÿ The long-promised proliferation of short-range personal-area wireless networks
based on the Bluetooth standard could take years due to the nagging costs of embedding the
technology in cellular phones and other mobile devices. That was the conclusion of a debate at
this week's Bluetooth Conference.
The debate was over the cost of a wireless connection. It started when Roger Gush, marketing
manager for Wireless Data Services (Poole, U.K.) opened the conference by mentioning the $5
figure that the initial members of the Bluetooth Consortium ÿ Ericsson, IBM, Nokia and Toshiba
ÿ said they wanted for a Bluetooth connection. "You run a piconet, connecting up to eight users,
and it costs $5,"Gush said.
But semiconductor manufacturers here ÿ those entrusted with integrating Bluetooth radio
transceivers and link-level controllers ÿ were divided over whether and how quickly that price
could be reached. The concern deepened as the usage models continued to expand from simple
point-to-point cable replacements, to wireless Internet ports for laptops in hotels and airports, to
shopping list managers for consumers raiding Costco with their Bluetooth-enabled PDAs. While
the requirements of a radio transceiver were relatively fixed, each scenario called for a more
complex software stack and faster, more power-hungry controllers.
"The magical $5 figure is not doable within the next three years," said Randy Giusto, vice
president for worldwide mobile research at the International Data Corp. (Framingham, Mass.).
Though he predicted that Bluetooth add-ons will begin to roll out this year, he said that the cost of
hardware alone will be in the $20-to-$25 range, meaning that embedded implementations would
be limited to laptops for mobile professionals and other high-end devices that could absorb the
extra costs of Bluetooth. Though the price of hardware will drop to $15 next year, the cost of
application firmware and software will likely keep the cost of implementation in the $20 range.
Bluetooth will not likely be immediately incorporated in cellular handsets, Giusto said, since the
margins of handset makers are already considerably stressed. In fact, the cost of Bluetooth is so
troubling to handset makers that it will not likely be embedded, but rather would be available in
snap-on modules for which users pay extra, predicted Ivar Johansson, director of technical
marketing for wireless systems at Infineon Technologies. The basic Bluetooth wireless service
would be something that cellular service providers would need to subsidize; that is, they will need
to encourage users to sign up for Bluetooth service at an extra cost, in the same way they
encourage subscribers to add voice and text messaging to their service.
Thus, 70 percent of advanced service providers such as UMTS/WCDMA will include Bluetooth
services by 2005, said Johansson, but only 30 percent of the GSM service providers will include it
by then. But even that 30 percent would depend on a Bluetooth bill of materials that totals less
than $10, he said.
Costly delays
Stefan L”f, senior product manager of Ericsson, told implementers not to focus their cost
analyses on bills of materials. "That's a trap," he said. "Look out for hidden costs." These can
range from semiconductor yield losses to the time-to-market delays imposed by regulatory
agencies requiring RF type approval. A delay of one quarter could cost a company 10 percent in
overall volume shipments, but up to 25 percent of a year's profits in a competitive market, he
said.
A Bluetooth implementation has four major components, L”f said. These are applications
software, a Bluetooth stack, hardware/firmware and an antenna ÿ and "antenna knowledge is
not that easy to find." In the modules currently marketed by Ericsson, this includes three ICs, a
link controller, a baseband processor and an RF transmitter/receiver IC and 30 to 50 discretes for
tuning. Even with increasingly integrated RF ICs, some parts cannot be integrated, L”f said. The
single-chip radio would still require from 20 to 30 discretes, he said, and yield losses for such a
device would impose costs. If a 0.35-micron BiCMOS device costs $4 with a 99 percent yield, it
will cost $12 with a 97 percent yield and $20 when yields drop to 95 percent. Thus, Bluetooth
radio costs are likely to average $27.20 in 2000 and $10 in 2003, according to L”f. "I never said
$5."
"It's not a $5 socket now," said Mark Lambert, the product manager for digital enhanced cordless
telecommunications (DECT) at Cambridge Consultants (Cambridge, U.K.), a subsidiary of Arthur
D. Little Co. "There are similar applications that provide less cost now," he said. Lambert's
presentation compared Bluetooth with other wireless connections such as IrDA, IEEE 802.11,
DECT and HomeRF. IrDA was identified as the best overlooked possibility for wireless
connectivity. "But Bluetooth devices will be out there anyway," he said.
Nevertheless, Cambridge Consultants' spin-off company, Cambridge Silicon Radio, claimed it has
the formula for a $5 Bluetooth implementation. A Cambridge Silicon Radio spokesman brushed
off competitors' claims that Bluetooth will cost more. "It's a familiar story," he said. "It is in the
interests of the industry to manage expectations when you don't have a solution to comply with
pricing hopes."
Cambridge Silicon's entry is a single-chip transceiver and link controller. It is based on what the
company calls a BlueCore and BlueStack, and includes an entire Bluetooth module on one chip,
said the company. It requires no external SAW filters, ceramic capacitors or inductors, said
Robert Davis, the Bluetooth product manager for Cambridge Consultants. The cost would be
something like $8 when the device comes out later this year, and will drop to $5 next year, the
Cambridge Silicon spokesman said. "We're the only ones with a single-chip solution," he said.
"You would need very dense processes," said Bob Tait, strategic marketing manager for Silicon
Systems Ltd. (Dublin, Ireland), an intellectual property provider and integrator whose offerings
include Bluetooth and GSM. "With 0.18- or 0.15-micron processes, you can put the baseband into
the system chip for a nominal cost. It's just a couple of square millimeters, depending on memory.
All the stress [to reduce die size and costs] is on the radio guys," he said.
Costs beyond silicon
While Silicon Systems' current implementation schemes for the Bluetooth radio focus on silicon
germanium, Tait believes that CMOS will foster higher levels of integration and bring costs in line
with expectations. "The problem with the traditional heterodyne radio architectures is that no one
can afford the filter and tank circuits," he said, referring to the 30 to 50 passive components now
required to tune a 2.4-GHz radio. "It's $5 silicon, you're still using $6 worth of passives. And you
need to add the antenna costs. The $5 price is not doable until 2004 or 2005," Tait said. "But we'll
get under $5 in pure silicon much earlier than that."
The second-generation Bluetooth assembled by Silicon Systems will promote complete solutions
on a single chip, Tait insisted. "One of the things, if you're going to be doing Bluetooth, you have
to get as close to a complete solution as possible. We prefer to use a highly-integrated approach
(rather than rush products), but what we come up with will have a small footprint, code size and
cost."
On the semiconductor front, manufacturers like National Semiconductor (Santa Clara, Calif.) and
Samsung Electronics (Seoul, South Korea) were demonstrating or discussing multichip sets here,
with separate transceivers and baseband processors.
Ki W Lee, vice president and general manager of corporate research at Samsung, saw a great
potential for Bluetooth devices in home appliances and consumer electronics. Their overlap with
802.11 and HomeRF solutions would depend on bandwidth and distance considerations, he said;
for example, he added, HomeRF and 802.11 transceivers would dominate where there was more
than 50 meters between connected devices, and Bluetooth would dominate for shorter distances.
But such a solution would still require a module with IF components eliminated by direct
conversion, an RF SAW filter embedded in the printed-circuit-board substrate, a fully integrated
RF IC (BiCMOS, with frequency synthesizer and voltage-controlled oscillator) on chip and
antennas that ensured a wide variety of propagation patterns. The baseband processor ÿ
handling traffic between multiple devices and loading packet data onto the Gaussian frequency
shift-keyed modulator ÿ would be based on a proprietary low-power, 16-bit RISC architecture.
"Chip size doesn't tell you anything until you get to assembled costs," Lee said, partially agreeing
with L”f of Ericsson. However, he saw semiconductor integration reducing the bill of materials to
close to the total semiconductor costs. But such a chip set would be $15 in 2000, with a total
implementation cost of $26, said Lee. That chip set might cost $5 by 2003 and cost several
dollars to solder down, said Lee. His conclusion: the total cost of a Bluetooth chip set won't drop
below $5 until 2005. And in discussions after his presentation, L”f was sanguine about the $10
price tag he saw for Bluetooth in 2003. "Time does not stop then: We will get to $5 for the total
solution."
Still, even if it does not come until 2005 the magical $5 price tag remains an attraction for
Bluetooth implementers and service providers. It may say a lot about why 1,000 technical
professionals showed up at this conference, and why the Bluetooth special interest group has
more than 1,600 members. "At $5, it is not a problem," agreed Eugenia Cimatti, a business
analyst with Italy's No. 1 service provider, Telecom Italia. "You can begin to think about
imaginative applications and see new possibilities," she said. |
