article - Philips Research

Powerful processing
Until affordable, low-power AD and DA (digital-to-analog) converters are fast
enough, the part of the radio implemented purely in software is confi ned to the
processing of the baseband signal – the voice or data. This includes modulating
and demodulating, encoding and decoding according to the transfer protocol
– roles that are hard-wired in application-specifi c integrated circuits (ASICs) today.
Because software algorithms can adapt to different standards, such a ‘software
modem’ makes effi cient reuse of the hardware platform.
The heart of the software modem is the embedded vector processor (EVP).
In the late 1990s, Philips researchers started looking at what a software modem
needed to achieve. This resulted in the fi rst communications vector processor
(CVP), which was key in developing the hardware and software architectures that
led to the more powerful EVP. Unlike traditional processors that manipulate one
number at a time, the EVP operates on multiple vectors of 16 numbers each in
parallel. This lets it manipulate vast quantities of data quickly.
To prove and fi ne-tune the concept, Philips Research is currently concentrating on
UMTS (Universal Mobile Telecommunications System) telephony, WLAN (Wireless
Local-Area Network), DVB (Digital Video Broadcasting), and the TD-SCDMA (Time
Division – Synchronous Code Division Multiple Access) data standards. This
selection covers the major challenges for designing SDR, and over the next year
or two should help fi nalize the software and hardware architectures. Then an
application programming interface (API) should make it possible for software
Antenna
system
RF signal
processing
Analog/
digital
conversion
Receiving
Transmitting
Baseband
Signal
Processing
Feature technology Feature technology
Generic radio architecture Software-defined radio (SDR) architecture
Protocol
Media
Access
Control
Radio 1
Radio 2
Radio N
A new medium for
radio-frequency (RF)
design
CMOS (complementary metal-oxide
semiconductor) technology makes
it easy to integrate RF and digital
circuits on a single chip
(and reap the advantages of the rapid
developments in coming CMOS
generations). However, implementing
RF on CMOS is not straightforward.
For example, the inductors needed
by RF circuitry take up a lot of silicon
‘real estate’, and their size is absolute
(because it is determined by the
operating frequency). As CMOS
technology shrinks further, the cost
per unit area increases, so the
inductors not only become relatively
bigger, they also become more
expensive. The challenge for RF
designers is to provide all the
fl exibility and adaptivity needed
by SDR, with the smallest possible
number of inductors.
Virtual
Machine
Monitor
(VMM)
SDR manager
Hardware
Abstraction
Layer
(HAL)
Interconnect
(buses,
bridges, etc.)
An overview of the components of a software-defi ned radio architecture, capable of handling
multiple standards and multiple channels. For example, for a cellular handset, Radio 1
might be a UMTS transceiver, Radio 2 a WLAN transceiver, and Radio 3 a GPS receiver.
RF
frontend
timers
EVP
SWcodec
accelerator
ARM
shared
RAM
suppliers to set up other radio standards.
Running different radios simultaneously on the same, generic hardware is the
ultimate goal of the current research. This minimizes the hardware cost and the
power consumption of the overall device, but it means ‘virtualizing’ the radios so
they can run independently. Virtualization has been around since the 1960s, but
SDR needs to allocate the shared hardware and processing blocks used to emulate
the radio, from the antenna to the user, while isolating the various incoming and
outgoing signals from one another. Encapsulating the radio pipes and switching
between them every few microseconds has to guarantee uninterrupted real-time
performance – not a trivial matter.
Industry implications
SDR will roll out slowly, taking over slowly from existing technology as this reaches
technical and economic obsolescence. The fi rst generation of SDR mobile phone
products is expected around 2008. Using a minimum set of parallel, softwarecontrolled
bi- and tri-standard RF blocks with a single processor block, these
should be able to cover standards from cellular telephony to networking to radio
and TV. Signifi cantly, they should be able to do this at close to the same cost as
today’s devices.
A major goal of SDR is to serve many markets, many customers, and many
standards with a minimum number of architectures and chips. This will not only
reduce R&D costs, but also greatly simplify logistics for both the semiconductor
manufacturers and handset makers. But it is a big step. So big that no party can go
it alone. The established companies are already talking together about open designs
that should ease programming specialized applications. Apart from Philips, other
major players include Infi neon, Samsung, Benq and Nokia. For handset
manufacturers in particular, the greater fl exibility will shorten development cycles
and make it easy to patch performance or upgrade algorithms, and through this
reinforce the positive value of their brand. For operators, SDR makes it easier to
provide new, fashionable and above all competitive services. Reconfi gurable
handsets also have longer useful lifetimes – an important consideration in Europe
where operators subsidize most handsets. Open, industry-standards for
SDR APIs (application program interfaces) could open up a whole new market for
radio software.
By reducing development effort and component costs and by improving fl exibility,
responsiveness and ease of product diversifi cation, SDR should benefi t everybody,
not least the end-user. Even before it fi nds applications in other, and yet
unthought-of areas, SDR will be a key technology in how we communicate
in the future.
Dr Neil Bird l Philips Research l neil.bird@philips.com
Prof dr ir Kees van Berkel l Philips Research l kees.van.berkel@philips.com
Dr ir Albert van der Werf l Philips Research l albert.van.der.werf@philips.com
Extra info www.research.philips.com/password l software-defi ned radio l RF IC design
Philips’ unique
position
Philips’ expertise in the various areas
touched by SDR places the company
in a uniquely favorable position for
developing the fi rst generation of SDR
equipment. Philips Research has already
designed and simulated an antenna
for use over fi ve cellular bands, and
demonstrated a multimode receiver for
GSM, UMTS and CDMA2000 standards.
The DSP Innovation Centre (Philips
Semiconductors) has developed an EVP
capable of 100 parallel operations at
speeds of 300 MHz. Also, Philips
Semiconductors have already developed
and productized a multi-standard
baseband chip for WLAN and DVB-H,
based on the predecessor of the EVP.
This established position on both sides
of the RF divide places Philips ideally to
make the most of SDR.
Researchers are working closely with
Philips Semiconductors to make sure
their designs are not just functional, but
also feasible, and, of course, attractive
for Philips customers of both complete
radios and chipsets for further software
development.
18 Philips Research Password 26 l February 2006 Philips Research Password 26 19
l February 2006