Design Challenges: Avoiding the Pitfalls, winning the game - Xilinx
Design Challenges: Avoiding the Pitfalls, winning the game - Xilinx
Design Challenges: Avoiding the Pitfalls, winning the game - Xilinx
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
interfaces available, Microsoft intentionally<br />
designed a flexible solution that allows<br />
rapid changes to <strong>the</strong> back-end vehicle<br />
interface without affecting <strong>the</strong> underlying<br />
architecture and performance of <strong>the</strong> system.<br />
For example, in <strong>the</strong> future it would be<br />
possible to adapt <strong>the</strong> FPGA solution to suit<br />
<strong>the</strong> needs of <strong>the</strong> end application with automotive<br />
buses such as MOST, IDB-1394, or<br />
ano<strong>the</strong>r digital vehicle network.<br />
Voice Recognition System<br />
Central to <strong>the</strong> Microsoft Telematics<br />
Platform is <strong>the</strong> voice recognition (VR) system.<br />
The audio signal path within any VR<br />
system is analog biasing/filtering, digitization,<br />
and digital filtering before <strong>the</strong> signal<br />
is finally presented to <strong>the</strong> VR engine for<br />
speech processing.<br />
Within this path, multiple opportunities<br />
exist for unwanted noise to be introduced<br />
into <strong>the</strong> system (both onboard <strong>the</strong><br />
electrical platform and within <strong>the</strong> vehicle<br />
environment even before <strong>the</strong> electronics).<br />
Both <strong>the</strong> product developer and <strong>the</strong> vehicle<br />
manufacturer must ensure that <strong>the</strong> microphone<br />
position and type are correctly suited<br />
to <strong>the</strong> application and environment.<br />
In a perfect world, <strong>the</strong> VR engine will<br />
receive clean, consistent speech signals –<br />
but given <strong>the</strong> dynamic nature of <strong>the</strong> vehicle<br />
environment, acceptable voice recognition<br />
implementation is not a straightforward<br />
exercise. Factors such as vehicle speed, window<br />
position (open/closed), road noise,<br />
and wea<strong>the</strong>r conditions (rain/wind) only<br />
add to already difficult VR problems such<br />
as languages, accents, and gender. These<br />
added factors have increased <strong>the</strong> importance<br />
of preconditioning using highly<br />
adaptive digital filtering algorithms before<br />
<strong>the</strong> signal is presented to <strong>the</strong> VR engine.<br />
Microsoft chose to implement this signal<br />
preconditioning in hardware and take<br />
advantage of <strong>Xilinx</strong> parallel DSP processing.<br />
Spartan-3 FPGAs, with as many as<br />
104 embedded 18-bit multipliers, are ideal<br />
for implementing compact DSP structures<br />
such as MAC engines, distributed arithmetic<br />
FIR filters, and fully parallel FIR filters<br />
in a low-cost device.<br />
Microsoft also offloaded processorintensive<br />
software filtering into hardware.<br />
Of course, this pre-processing is possible in<br />
ASSPs such as dedicated DSP chips. But<br />
<strong>the</strong> benefits gained through high levels of<br />
integration in o<strong>the</strong>r parts of <strong>the</strong> platform<br />
would be lost.<br />
The combination of telematics and VR<br />
allows implementations of adaptable and<br />
upgradeable VR engines and DSP filters<br />
tailored to suit certain types of users and<br />
environments (Language: English, Accent:<br />
Scottish, Gender: Female).<br />
The importance of designing automotive<br />
products (especially in <strong>the</strong> infotainment<br />
section of <strong>the</strong> vehicle) with sufficient<br />
spare bandwidth to cope with new and<br />
unexpected future upgrades also applies to<br />
<strong>the</strong> FPGA. It is now becoming clear to<br />
automotive OEMs that architectures that<br />
allow for flexible and scalable firmware are<br />
a necessity in future platforms.<br />
Although not currently implemented<br />
in <strong>the</strong> Microsoft platform, it would be<br />
possible to easily add soft processors to<br />
act as system co-processors. Just as <strong>the</strong><br />
DSP processing was offloaded from <strong>the</strong><br />
main processor in Microsoft’s design, it<br />
would also be possible to use embedded<br />
processors (such as <strong>the</strong> <strong>Xilinx</strong> 32-bit<br />
MicroBlaze soft processor or 8-bit<br />
PicoBlaze microcontrollers) to take<br />
some of <strong>the</strong> processing load from <strong>the</strong><br />
main system processor.<br />
FPGAs for Automotive Applications<br />
In-car electronics have seen tremendous<br />
growth in recent years, not only in traditional<br />
body control and engine management<br />
but in <strong>the</strong> new areas of driver<br />
assistance systems and telematics applications.<br />
Figures recently published by <strong>the</strong><br />
IEEE indicated an annual increase in car<br />
electronics of 16%, with a prediction that<br />
by 2005 electronics will account for 25%<br />
of <strong>the</strong> cost of a mid-size car.<br />
Telematics systems exhibit characteristics<br />
more like those of consumer products –<br />
short time to market, short time in market,<br />
and changing standards and protocols.<br />
These issues impact <strong>the</strong> way engineers<br />
approach designs and select <strong>the</strong> hardware<br />
needed to quickly create, iterate, and support<br />
future upgrading.<br />
FPGA technology can now solve <strong>the</strong>se<br />
requirements. <strong>Xilinx</strong> is committed to serving<br />
telematics and car infotainment applications<br />
through its <strong>Xilinx</strong> Automotive (XA)<br />
family, which delivers:<br />
• Extended temperature ranges – up to<br />
125°C<br />
• Full production part approval process<br />
(PPAP) support<br />
• Industry-recognized AEC-Q100 devicequalification<br />
flow<br />
• Compliance with <strong>the</strong> worldwide automotive<br />
quality standard ISO TS 16949,<br />
as well as Pb-free packaging to meet <strong>the</strong><br />
RoHS directive<br />
These devices, based on our Spartan<br />
family of FPGAs, are ideal for digital<br />
designs requiring low cost per logic cell<br />
(system gate), low cost per I/O, and<br />
advanced features such as multiple I/O<br />
standards on a singe device and embedded<br />
multipliers for high-speed DSP.<br />
Conclusion<br />
Backed by a commitment from supporters<br />
such as <strong>the</strong> Microsoft Automotive Business<br />
Unit and <strong>Xilinx</strong> Automotive, <strong>the</strong> vision of<br />
Microsoft’s Telematics Platform is now<br />
becoming a reality. The convergence of key<br />
technologies is being adopted today by<br />
first-tier automobile manufacturers in a<br />
platform that enables:<br />
• A valuable and affordable telematics solution<br />
• Reliable connectivity through wireless<br />
networks<br />
• High-quality voice recognition<br />
• A broadly supported operating system<br />
for application developers<br />
• Low-cost hardware<br />
This is giving rise to a “virtuous cycle” of<br />
continuous investment by developers, who<br />
will use <strong>the</strong>se platforms to create even more<br />
value for end users.<br />
For more information, visit<br />
www.microsoft.com/automotive/<br />
windowsautomotive/about.mspx/ and www.<br />
xilinx.com/automotive/.<br />
Third Quarter 2005 Xcell Journal 43<br />
$<br />
TOTAL COST