Professional Documents
Culture Documents
RTC Magazine January 2006
RTC Magazine January 2006
Dual-Core Processors
The magazine of record for the embedded computing industry Merging Model-Based and Code-Based Design
DATA
Acquisition
Goes
DEEP
and
FAST
An RTC Group Publication
"OBUPNZPGB1SPDFTTPS"EWBODFE.$
4#4LOPXT.PEVMBS$PNQVUJOH$BSFGVMEFTJHOZJFMETTQFFEUIFSNBMFGmDJFODZ
4#4LOPXT
'JOEUIF"EWBODFE.$CPBSEZPVSFMPPLJOHGPSBUXXXTCTDPNPSDBMM4#4&.%&%%&%
Departments
11 Industry Insider
66 Products&Technology
Features
A major trend in high-performance, real-
Technology in Context Data Acquisition time data acquisition boards and subsys-
tems is the move toward more flexible
14 Data Acquisition Subsystems Getting Faster, Interface-Agnostic
interfaces. For example, the 400 Mbyte/s
Ann R. Thryft
StreamStor Amazon SATA Disk Control-
18 Standardizing Digital IF Data Transfer with VITA 49 ler board (bottom) from Conduant uses a
Stephen M. Pereira, Mercury Computer Systems, Inc. modular, mezzanine approach to external
interfaces for direct-to-disk recording:
24 Shared Memory Network Targets Video-Centric Data Acquisition optional interchangeable daughtercards
Ralph Barrera, Curtiss-Wright Controls Embedded Computing, Data Communications (top) for interfaces such as FPDP, Serial
FPDP, FPDP II or the PCI bus. • Pg. 14
28 Data Acquisition Systems Track Signal Processing Technology
Andrew Reddig, TEK Microsystems
Executive Interview
53 RTC Interviews Mercury Computer Systems’ James “Jay” R. Bertelli
January 2006
January 2006
Publisher
PRESIDENT
John Reardon, johnr@r tcgroup.com
VICE PRESIDENT, European Operations
Zoltan Hunor, zoltanh@r tcgroup.com
EDITORIAL DIRECTOR/ASSOCIATE PUBLISHER
Warren Andrews, warrena@r tcgroup.com
Editorial
EDITOR-IN - CHIEF
Tom Williams, tomw@r tcgroup.com
SENIOR EDITOR
Ann Thr y f t, annt@r tcgroup.com
MANAGING EDITOR
Marina Tringali, marinat@r tcgroup.com
STAFF EDITOR
Craig Choisser, craigc@r tcgroup.com
COPY EDITOR
Rochelle Cohn
Art/Production
CREATIVE DIRECTOR
Jenna Hazlett, jennah@r tcgroup.com
GRAPHIC DESIGNER
Melissa Gaeta, melissag@r tcgroup.com
PRODUCTION DESIGNER
Kirsten Wyatt, kirstenw@r tcgroup.com
DIRECTOR OF WEB DEVELOPMENT
Jan Garcia, jang@r tcgroup.com
WEB DEVELOPER
Marke Hallowell, markeh@r tcgroup.com
Advertising/Web Advertising
CALIFORNIA COASTAL ADVERTISING MANAGER
Diana Duke, dianad@r tcgroup.com
(949) 226 -2011
WESTERN REGIONAL ADVERTISING MANAGER
Lea Ramirez, lear@r tcgroup.com
(949) 226 -2026
EASTERN REGIONAL ADVERTISING MANAGER
Nancy Vanderslice, nancy v@r tcgroup.com
(978) 443 -2402
EMEA SALES MANAGER
Anna Brown, annab@r tcgroup.com
(949) 226 -2025
ASIAN ADVERTISING MANAGER
John Koon, johnk@r tcgroup.com
(858) 755 - 4999
Billing
Mag gie McAuley, mag giem@r tcgroup.com
(949) 226 -2024
Editorial Office
Warren Andrews, Editorial Director/Associate Publisher
39 Southport Cove, Bonita, FL 34134
Phone: (239) 992-4537 Fax: (239) 992-2396
Discover how Dan and the QNX team deliver the shortest
migration path to multi-core. Call 1 800 676 0566 or
visit www.qnx.com/innovate.
QNX, Momentics, and Neutrino are trademarks or registered trademarks of QNX Software Systems GmbH & Co. KG and are used under
license by QNX Software Systems International Corporation. All other trademarks belong to their respective owners. 301786 MC339.05
SHARE YOUR VISION.
WE’LL SHARE OUR INNOVATION.
w w w. v a d a t e c h . c o m 702.896.3337
Editorial
January 2006
I
will say up front that I have not seen the movie, March of among remote experts who will need access to their data. Indus-
the Penguins, but I have seen one image of penguin behavior trial companies will increasingly need to distribute design data
that seems indicative of where one part of our industry stands. and organize teleconferences.
That’s where a bunch of penguins are crowded up at the edge of The specialized nature of many of these applications will push
the ice trying to decide whether it’s safe to go into the water. If the value-add from merely supplying high-speed pipes to provid-
there are sharks or orcas down there, of course, everybody will ing services closer to the application in the form of middleware or
take a pass. What usually happens is that there is such a crush that protocol conversion or lower-volume, more specialized hardware.
one penguin eventually gets pushed over the edge. If he doesn’t That makes it attractive to build out the bulk of the basic infra-
get eaten, the others decide it’s OK to get into the water. The wa- structure using open standards hardware. Of course, the coun-
ter, of course, is where the food is. The question is simply who is ter argument to this always is that such hardware will be driven
going to be the food. to commodity status where only huge volumes will be profit-
This seems to be where the industry is at this moment in able, putting it in the hands of only a few very big players—if
terms of ATCA and MicroTCA. A number of CEOs I have talked it is successful at all. Commoditization will surely happen. The
to recently keep saying things like, “We see an enormous poten- question is whether in such a potentially large market there is
tial,” and, “Our customers are taking a close look and testing the room for makers of specialty hardware in the same form-factor or
market.” A number have even jumped into the water, and while for makers of configuration modules (e.g., AMC and MicroTCA)
they haven’t exactly been gorging themselves, they haven’t been that will, of necessity, sell at higher margins but fit into the com-
eaten either. The last telecom debacle was traumatic and there is modity backplanes and carrier boards.
still a good deal of bad-mouthing of telecom going on in some Everybody takes glee in dissing Windows, but Microsoft’s
quarters today. But is the outlook really as discouraging as some near monopoly has provided fertile ground for a huge number
would have us believe? of successful companies, spreading wealth in all directions. By
For one thing, let us remember once again that what we re- the same token, if there is a base infrastructure founded on the
fer to as “telecom” today is a different animal than it was in the products of a relatively few large commodity suppliers, it is cer-
past. The telecom of the future is a fully digital, IP packet-based tainly possible that more specialized suppliers of software and
broadband communications network whose infrastructure is still hardware will be able to build on such a foundation.
in transition from the old POTS world of the late Ma Bell. The de- Instead of wringing our hands about whether the new
mands that will be placed on this network would have completely “telecom” industry will take off, perhaps we should think about
overwhelmed the plain old telephone system of yesteryear. where to attach ourselves when it does. The water is full of fish
The recent Consumer Electronics Show in Las Vegas may and not all of them are appetizing to orcas. There will be many
be one indication of what is starting to pull at the telecom in- specialized needs that can be served by a brand new digital
dustry. There the talk was of high-definition video and multi- telecom infrastructure and it will require imagination and daring
media, wireless connectivity, movies on demand and pervasive to identify and serve them, which will demand new hardware and
connectivity. Products are starting to emerge that will depend on software that can play in the arena. If you are the penguin on the
the existence of pervasive broadband networking. The consumer ice that gets pushed in first, there is a risk you might be eaten, but
market, however, is only one indicator. The needs of the finan- there is also a very good chance you’ll get first choice of the best
cial sector will drive the build-out of the infrastructure as will fish. Then again there are some companies (if not penguins) who
advances in medical imaging equipment. The latter can produce jump in with the sole intention of being eventually swallowed.
3D layered images of minute detail in color and some in real That doesn’t have to be a bad thing either.
time. There is a growing need to share these expensive machines
January 2006
7 ( !4 - ! + % 3 ! 0 2 / $ 5 # 4 % 8 4 ) . # 4
6KQN,"*LNK@Q?P?=J>A?KIAATPEJ?PHKJC>ABKNAEPOPEIA QJHAOOUKQD=RA=HKJCPANI?KI
IEPIAJPBNKIUKQN0 RAJ@KN3ANO=)KCE?CQ=N=JPAAOLNK@Q?P=R=EH=>EHEPUBKN=PHA=OPBERAUA=NO
BNKI EPO EJPNK@Q?PEKJ @=PA 1D=P O=RAO UKQ BNKI PKOOEJC BQPQNA O=HAO EJPK PDA F=SO KB ?KILAPE
"5
PKNO>A?=QOAUKQOQ@@AJHU?=JJKPCAPUKQND=J@OKJPDA>K=N@OUKQJAA@
SDAJUKQJAA@PDAI
3ANO=)KCE?O"TPAJ@A@)EBA U?HA-KHE?UEOFQOPKJAKBPDAOANRE?AOPD=PDAHLA@
QOA=NJPDALNAOPECEKQO3! -H=PEJQI3AJ@KN=S=N@#NKIKQN?HKOAPEAOSEPDGAU
-
RAJ@KNOPKKQNLNA@AOECJ?KILKJAJPOPQ@EAO
SAI=GAEPKQN>QOEJAOOPKATPAJ@
"EFORE +RISTIN +OONS JOINED
ENDANGERED SPECIES LIST
-H=PEJQI3AJ@KN=S=N@>U
3AJPQNA!ARAHKLIAJP KNLKN=PEKJ Ġ
01! -H=PEJQI3AJ@KN =J@ SSS3ANO=)KCE??KILR
Industry Insider
January 2006
January 2006 11
Get Connected
with companies and Get Connected
products featured in this section.
Industry Insider
with companies mentioned in this article.
www.rtcmagazine.com/getconnected www.rtcmagazine.com/getconnected
Event
www.rtcmagazine.com/getconnected
status allows Brooktrout’s
customers and partners to test
12 January 2006
Industry Insider
Mentor Graphics. Through the QuickMIPs and QuickPCI. Beta for its Leesburg, Virginia and market access as a result of a
agreement, FPGA designers get support is also available for the Ottawa, Ontario operations. standardized process system
immediate access to a broad range company’s newly announced AS9100 is overseen by the recognized across the aerospace
of performance and productivity PolarPro devices. International Aerospace Quality industry. Because AS9100
benefits using the Mentor QuickWorks supports Windows, Group and standardizes quality drives ongoing improvements
Graphics Precision Synthesis Sun Solaris and Linux-based management system requirements in products and processes, it
tool within the QuickLogic operating systems and provides a and delivers quality assurance in reduces errors and returns and
QuickWorks FPGA development design environment ranging from design, development, production, increases customer satisfaction,
software environment. Using schematic and HDL-based design installation and servicing. resulting in reduced transaction
a highly interactive graphical entry, HDL language editors and The standard also drives cost costs.
environment, designers gain the tutorials, logic synthesis, place reductions throughout the AS9100 compliance exceeds
flexibility to cross-probe between and route, timing analysis and aerospace industry supply chain. the ISO9001:2000 quality
HDL and schematic views. The simulation support. QuickLogic Curtiss-Wright’s accreditation standard on which it’s based,
tool performs “what if” timing has partnered with leading program was directed by Gerry with additional quality system
analysis with instant feedback, software vendors to provide Bellehumeur, Quality Director, requirements such as independent
enabling designers to make industry-leading Synthesis and Curtiss-Wright Controls Embedded validation of materials and
confident decisions for fast and Simulation tools, as well as Computing, Ottawa. The AS9100 processes. It adds approximately
accurate timing closure. provide an interface to other audit was performed by TUV 80 additional requirements
Precision Synthesis is industry standard EDA tools. America. and 18 amplifications to the
scheduled to be fully integrated Curtiss-Wright has announced ISO 9001:2000 Standard. The
into the QuickWorks FPGA Curtiss-Wright Earns that it has developed plans to standard addresses the unique,
development software in early AS9100 Aerospace Quality implement the AS9100 standard complex and highly regulated
2006. The new Precision at additional operational sites nature of the defense aerospace
Systems Certification
Synthesis QuickLogic Edition throughout the United States and industry.
Curtiss-Wright Controls
supports QuickLogic’s leading the United Kingdom. The benefits
Embedded Computing has been
programmable logic devices, of AS9100 accreditation for
awarded AS9100 certification
including Eclipse I, Eclipse II, Curtiss-Wright include expanded
To Learn more about AdvancedTCA (ATCA) and DTI’s Products based on this standard
January 2006 13
TechnologyInContext
Data Aquisition
Data Acquisition
Subsystems Getting Faster,
Interface-Agnostic
More and more embedded applications rely on next-generation,
high-performance data acquisition subsystems to gather large
quantities of data at high speeds and convert it into usable form.
The use of switched fabrics, FPGAs and network topologies, along
with a move toward flexible interfaces and vendor interoperability,
d
are helping designers of these subsystems keep pace.
exploration by A
nn R. Thryft
her your goal
peak directly Senior Editor
al page, the
t resource.
A
chnology,
s high-performance embedded com-
and products
puting systems continue to achieve
new levels of functionality and per-
formance, larger amounts of data must be
gathered, processed and analyzed. Ap-
plications such as radar data acquisition
and video-centric imaging, for situational
awareness on the battlefield or high-speed
mpanies providing manufacturing
solutions nowlines, have vastly increased
the amount
oration into products, of and
technologies datacompanies.
and theWhether
speed your
at which
goal is to research the latest
plication Engineer, or jump to a company's technical page, the goal of Get Connected is to put you
it must be processed.
vice you require for whatever type of technology,
nies and products you For example,
are searching for. next-generation radar
data acquisition systems are being built
for advanced aircraft such as Northrop
Grumman’s E-2D Hawkeye. These will
help the U.S. Navy’s Sea Strike offen-
sive capabilities by increasing battlespace
awareness, providing theater air missile
defense capabilities, improving detection
and tracking, and narrowing the link be-
End of Article
tween sensor and shooter for more agile
Figure 1 A major trend in high-performance, real-time data acquisition boards
and subsystems is the move toward more flexible interfaces.
response to time-sensitive targets. For example, the 400 Mbyte/s StreamStor Amazon SATA Disk
Controller board (bottom) from Conduant uses a modular, mezzanine
approach to external interfaces for direct-to-disk recording: optional
Get Connected interchangeable daughtercards (top) for interfaces such as FPDP,
with companies mentioned in this article. Serial FPDP, FPDP II or the PCI bus.
www.rtcmagazine.com/getconnected
14 January 2006
Get Connected with companies mentioned in this article.
. www.rtcmagazine.com/getconnected
TechnologyInContext
Manufacturers of data acquisition while also delivering high performance. data capture and distribution, in addition
boards and subsystems are leveraging Meanwhile, real-time video imaging to the high I/O throughput rates needed to
existing hardware off-the-shelf building has arrived on the data collecting stage handle large video streams and preserve
blocks—such as intelligent I/O control- in a big way. Video is playing an increas- data accuracy.
lers, embedded switched fabric intercon- ingly central role in capturing additional Distributed shared memory network
nect and high-speed fiber interface PMC data and monitoring operations. On the architectures based on a ring topology are
modules—and adding existing software to factory floor as well as on the battlefield, being utilized to construct high-speed I/O
keep costs and risk down. Combined with data must be processed and fused in real networks that not only meet these needs,
the customization capabilities of onboard time and made available simultaneously but also enable remote processing far from
FPGA-based IP, these components are to all of the system’s nodes. Here, the the harsh environments where the gather-
producing subsystems that meet aggres- challenge for design engineers is to main- ing of data occurs. Data is captured at
sive size, weight and power constraints tain the low latency required for real-time multiple stations and sent to several pro-
cessors, each of which processes different
pieces of that data simultaneously.
On another front, a major break-
through has occurred in the IF data
transfer interface. The emerging VITA
49 standard defines a standard way of
transferring IF data in a digital, link-ag-
nostic format between analog front-ends
g and DSP subsystems. Instead of depend-
w offerin debug
No ware sors! Arium offers robust JTAG emulation ing on application- and/or equipment-
OS a roces specific interfaces to transmit digitized
Linux -based p ay! and development tools for today's
RM d incoming analog signal data to system
for A Call us to embedded software engineers using
targets with ARM7™/ ARM9™/ARM11™, elements, the new format defines a data
Intel XScale®, and TI OMAP™ cores and structure for the transmission of digital
Intel® Pentium® processor families. IF data between multiple sources and
destinations for both receive and trans-
• Full symbolic, source-level Linux kernel mit paths. The standard’s methodology
debug and source-level process debug; for representing digital IF data can be
seamless debug between them! No layered on top of any transport protocol
other vendor offers this powerful or physical communications link.
feature at any price. The ramifications are clear: OEMs
• Real-time, integrated ETM will no longer be tied to vendor-specific
trace data collection at 640 interfaces, but can select interoperable
MHz and a GByte of trace system and subsystem components from
memory. many vendors based on which ones best
fit their applications. As an added bonus,
• Highly integrated designers no longer must rework system
SourcePoint™ IDE with pow- hardware and software each time a source
erful, flexible code editing or destination component changes, thus
with debug integration. speeding time-to-market.
• Real-time performance
analysis for faster, more
accurate results.
• Fast, easy, intuitive run control with
robust C-like command language
facilities.
• SourcePoint debugger available for
Microsoft® Windows® and Linux hosts.
16 January 2006
An Entire Family of High-Speed, Low-Power Universal Intel®
Pentium® M Processor Boards
Standardizing Digital IF
Data Transfer with VITA 49
Intermediate frequency (IF) data normally passes between system elements
in analog format over coaxial cables. VITA is developing a new interconnect
standard for passing IF data between analog front-ends and DSP subsystems
in a digital, link-agnostic format.
by S
tephen M. Pereira, Chairperson,
VITA 49 Working Group, Mercury
Computer Systems, Inc.
M
exploration any communications systems Standardizing the Digital IF “plug and play” those components. Ven-
her your goal digitize incoming analog signal Interface dors would no longer be required to re-
peak directly information with a high-speed A/ Standardizing the digital IF interface write their digital IF interconnect logic to
al page, the D converter and then route the digitized across receiver/transmitter equipment, yet another data format, saving resources
t resource.
chnology, information between system elements signal digitization and conversion equip- and increasing time-to-market. Standard-
and products for processing and analysis. Typically, ment and signal processing equipment izing digital IF could make system de-
the digitized information is intermedi- would clearly benefit both OEMs and ven- ployment faster and technology refresh
ate frequency (IF) data sent from a radio dors. System manufacturers would no lon- easier for both system manufacturers and
frequency (RF) downconverter to digital ger have to rework their systems each time their vendors.
signal processing equipment or sent from they upgraded a component. In 2004, Mercury Computer Sys-
digital signal processing equipment to an In addition, instead of being locked in tems and DRS Signal Solutions (DRS-SS)
RF upconverter. to a particular vendor, OEMs could pick formed an informal industry group that
Until
mpanies providing solutions now
now, the interface for transmit- and choose the best component for the solicited participation and input from the
ting the digitized IF data stream between application at hand from a marketplace signal acquisition and processing com-
oration into products, technologies and companies. Whether your goal is to research the latest
system
plication Engineer, elements
or jump to a company'sover thepage,
technical communica- of interoperable
the goal of Get Connected is to put youproducts and essentially munity and its OEM customers for the
vice you requiretions linktype
for whatever hasof technology,
been application- and/or
nies and products equipment-specific.
you are searching for. Often, it has also Notional Signal Receiving System
Fabric Fiber/Copper Fabric
been proprietary: the system’s digitiz- IF
ing source packages the IF data into a CN x N
End of Article
Clock Compute Compute Compute
for passing digitized data between them also Module Node Node Node
Node
changes, and new software must be written
to achieve or restore interoperability. VME VME
Figure 1 In today’s digital intermediate frequency (IF) interface, the boards on
Get Connected either side of the fiber/copper link provide the application-specific and
with companies mentioned in this article. generally proprietary logic that encodes and decodes the digitized data for
www.rtcmagazine.com/getconnected transmission across the link.
18 January 2006
Get Connected with companies mentioned in this article.
www.rtcmagazine.com/getconnected
TechnologyInContext
development of such a standard. The in- specifying a data representation standard. Meta-data support enables a multi-
formal industry group voted to associate The working group has discussed the channel source to add meta-data with a
with the VITA Standards Organization need for three packet types. A stream- channel number to a data sample and then
(VSO), and the VITA 49 Working Group ing data packet defines the base structure interleave the samples in one data pay-
was created to design a digital IF interface for representing digital IF data, a source load. The meta-data associated with the
standard for adoption by VITA. characteristics packet enables components data sample, which contains the channel
in a distributed system to communicate number, allows the destination to deter-
The Digital IF Data their capabilities to each other, and a sta- mine to which channel a sample belongs
Representation tus change packet conveys changes in the when the destination unpacks the data.
The Digital IF Data Representation system’s state. System event support enables a source
(VITA 49) defines a data structure for the The Digital IF Data Representation to communicate information about system
transmission of digital IF data between basic standard (VITA 49.0) defines the events, such as an A/D converter overload
one or more sources and one or more des- streaming data packet. Extensions to this or a radar antenna crossing north, to a des-
tinations for both the receive and transmit standard (VITA 49.x) will define the other tination. Events that affect the entire pay-
paths. The Digital IF Data Representation two types. load can be indicated, as can events that
is link-agnostic: it defines a methodology affect only a portion of the payload.
for representing digital IF data that can Streaming Data Packet Format The packet definition also specifies
be layered on top of any transport proto- The streaming data packet is de- a configuration key for linking a stream-
col and any physical communication link signed to minimize transmission over- ing data packet to other Digital IF Data
(Figure 2). head and maximize its applicability to a Representation packet types that will be
The goal of the Digital IF Data Rep- broad range of applications. It consists of defined in the future.
resentation is to define a data structure a header, a few optional header words, a A streaming data packet can contain
that can be used by a sensor source to variable-size data payload and a trailer up to one million words of data payload.
transmit digitized data to a signal process- (Figure 4). It allows an equipment manufacturer to
ing destination, or by a signal processing Because the Digital IF Data Represen- encode digitized IF data samples of real
source to transmit digital data to an emit- tation has been designed with multi-chan- (unsigned or signed) or complex format in
ter destination. Initially, it is focused on nel beam-forming and direction-finding a comprehensive range of widths, packed
radio IF to convey digitized analog radio applications in mind, the streaming data or unpacked, and tag the data samples (or
signals between RF communication re- packet provides features that address the not) for interleaving or other purposes.
ceivers/transmitters and digital process- requirements of these applications. These The manufacturer must specify how the
ing devices (Figure 3). include timestamp support, meta-data sup- data is formatted within the data payload,
Although the Digital IF Data Repre- port and system event support. including data sample width, type, data
sentation is intended for use in both mili- Timestamp support is critical for syn- packing method and whether or not the
tary and commercial applications, it is chronizing multiple channels of information data samples have meta-data.
particularly targeted toward beam-form- in beam-forming and SIGINT applications. The data payload can also be zero to
ing and direction-finding signal-process-
ing systems, as well as communications
and signal intelligence (SIGINT) systems. VITA 49 Standard
(Data Only)
Any communications system that needs
to change an analog signal to digital in- Data
formation and then send it on for process-
ing—such as police and fire department
communications systems—is a candi- Source Destination
January 2006 19
And you thought SHE was
sensitive?
What about your embedded calibration software with our low Ethernet, GPS, CAN, wireless,
system? drift specs designed to keep your and the operating system of your
Just like sensing the pea under a measurements constant over time. choice.
mattress, your A/D and D/A system And our wide selection of cost Call now and let our technical
must detect and measure accurate- sensitive A/D and D/A boards is team help smooth out those
ly too. So, start your system design something that will make any sensitive issues in your system.
with boards that handle real world princess happy.
irritants like temperature, drift and We invite you to check out our
electrical noise. 12, 14, 16 and 24-bit A/D and D/A
single board computers from
Sleep easier knowing your 12MHz to 533MHz,
measurements are accurate including VGA,
3730 Park Place, Montrose, CA 91020
because the effective number of CompactFlash,® Voice (818) 244-4600 Fax (818) 244-4246
bits for the whole system is close info@embeddedsys.com
to the ideal 12, 14, 16, or 24-bit
resolution of the A/D silicon. www.embeddedsys.com
Benefit from onboard fault pro-
tection circuits that ease system
integration by handling the noise
and power sequencing challenges
common to multi-board designs.
Eliminate the need for cumbersome
and complicated layers of auto-
EPIC TM
EBX PC/104
24-bit A/D
Our MPC624 provides 24-bit A/D
for PC/104 with four channels of
instrument grade voltage measure-
ment and/or four channels of direct
connect to off-the-shelf sensors.
that reads and parses the incoming packet Figure 3 The Digital IF Data Representation passing from source to destination.
and hands off the data to the next process- The A/D converter and the digital IF logic are provided in a single board,
ing stage for data payload decoding. the 3U/6U tuner component. Some signal processing logic could also be
Both source and destination equip- provided on this board.
ment manufacturers generally embed the
digital IF interface logic in FPGAs. The streaming data packet format under design a snapshot of the outgoing data from the
source equipment manufacturer must also in the demo source equipment (DRS-SS) source equipment chassis approximately
publish its compliance with the Digital IF and destination equipment (Mercury). The once per second.
Data Representation’s data payload en- companies then separately developed their A bit in the streaming data packet
coding parameters, typically in the prod- respective source and destination digital had been previously identified to mark
uct’s data sheet. In addition, if the source IF logic based on these demonstration ver- a packet as a snapshot for display. When
equipment manufacturer uses meta-data sions. the tuners receive a request from the dis-
and events, it must publish its meta-data In the demonstration configuration play computer, they set this bit in the next
and event definitions. (Figure 5), two RF signal generators gener- packet they process to mark it as a snap-
Source and destination components ate an RF signal that continuously sweeps shot for the destination equipment. The
using the Digital IF Data Representation over a frequency band to two tuners. Each tuners then send this packet to the display
are immediately interoperable for digi- of these tuners converts the incoming RF computer over the Ethernet connection
tal IF transmission over communications to IF, digitizes it at 80 Msamples/s and en- and output it over the fiber.
links. Vendors never have to rewrite those codes the digitized IF data and timestamp On the other end, the destination
products’ packing or unpacking logic. For into the prototype Digital IF Data Repre- equipment chassis examines the snapshot
example, while the logic in a destination sentation streaming data packet format. bit in the incoming packets and sends the
component that decodes the data payload The tuners then transmit the streaming snapshot packets it receives over the Eth-
will change depending on the application, data packet output over 2.5 Gbit/s Serial ernet connection to the display computer.
the logic that unpacks it will not. Front Panel Data Port (SFPDP) fiber to The display computer shows the snap-
the destination equipment chassis, which shot data from the source equipment chas-
Demonstrating the Standard reads and parses the incoming streaming sis and the destination equipment chassis
In January 2005, Mercury and DRS- data packets. in side-by-side windows. The source win-
SS began preparing a demonstration of Meanwhile, a display computer con- dow shows what was transmitted over the
the prototype VITA 49.0 Digital IF Data nected to the source and destination equip- fiber, while the destination window shows
Representation, using a version of the ment chassis via an Ethernet hub requests what the destination equipment received.
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Packet
Header K TS Reserved Number of Data Payload Words (0 to 1,048,575)
Type
Optional (If ‘K’ is set in the Header) V49 Configuration Key
V49 Ext Hdr
Elements Optional (If ‘TS’ >0 is in the Header) V49 Timestamp
(3 words - 1 PPS Time Reference plus 2 words of Delta Time Reference)
Data
Payload
Data Payload (0 to 1,048,575 Words)
Trailer Trailer
Figure 4 Designed to minimize transmission overhead and maximize its applicability to a broad range of applications, the
streaming data packet structure is defined in the Digital IF Data Representation basic standard (VITA 49.0).
January 2006 21
TechnologyInContext
Concurrent Controller
ter of 2005. Extensions to the standard, for
Motorola Controller
Mercury PowerPC
10 MHz Ref
Mercury FPGA
Reference
example, the source characteristics and
Source
SI-9136C
SI-9136C
(1PPS)
source status change packet definitions,
2.5 Gb/s Fiber,
areLooking
in the planning
For More?stages and will likely to download
Visit www.rtcmagazine.com
additional technical information related to this article.
S-FPDP be released as future VITA 49.x versions.
Prototype V49
The Digital IF Data Representation
Digital IF serves as a strategic technology that will
enable OEMs to select processor-based
Ethernet Interface
Control/ Ethernet
Figure 5 In the configuration for demonstrating the prototype Vita 49 Digital Looking For More?
IF interface, DRS-SS supplied the RF signal generation and source Visit www.rtcmagazine.com
equipment, while Mercury supplied the destination equipment. to download additional technical
information related to this article.
22 January 2006
TechnologyInContext
Data Acquisition
by R
alph Barrera, Curtiss-Wright Controls
d Embedded Computing, Data Communications
exploration
A
her your goal
peak directly big change has taken place in data ac- measurements, such as the displacement A/D converters and discrete I/Os. A single
al page, the quisition systems over the last several of an object, the object’s acceleration or processor could easily handle and operate on
t resource.
chnology,
years, as the growing use of video its temperature. the low-level data rates required to measure
and products data in imaging systems has increased the Because throughput requirements were the object’s position, movement or size.
size and speed of the data streams these fairly low, the data could be brought into the
systems need to deliver and process. processing system via normal I/O channels Enter Video
The challenge for design engineers is using analog or discrete signals. The process- Increasingly, however, the trend in
to support the low-latency requirements ing system itself could be built using simple today’s data acquisition systems is to
of real-time data acquisition and distribu-
tion, as well as provide the high through- Control, Monitor, Analysis
put required to handle large video streams and Storage System
mpanies providing solutions now
without dropping frames or diminish-
oration into products, technologies and companies. Whether your goal is to research the latest
ingor jump
plication Engineer, the quality of the
to a company's data.page,
technical A distributed
the goal of Get Connected is to put you
vice you requireshared memory
for whatever network can provide the
type of technology,
nies and products low latency
you are and
searching for. high throughput speed
needed for these more demanding imag- V V
ing systems, while at the same time en- A
able remote placement of the processing
system away from the frequently harsh Process
A A
factory floor environment. A D Being
Controlled
Formerly, traditional data acquisition V Video Sensor
End of Article
with relatively slow data rates of perhaps
10 to 100 measurements/second at the high
D Discrete Sensor
A A
end, since they conducted simple types of Figure 1 Since traditional, centralized data acquisition systems conducted simple
types of measurements, they were typically tasked with relatively slow
Get Connected data rates of, at most, 10 to 100 measurements/second. These systems
with companies mentioned in this article. could be easily built with a single processor, simple A/D converters and
www.rtcmagazine.com/getconnected discrete I/Os.
24 January 2006
Get Connected with companies mentioned in this article.
www.rtcmagazine.com/getconnected
TechnologyInContext
facturing process. For example, the system Figure 2 The high data throughput needed for video imaging data acquisition
may divert the bad object to a dump bin. produces challenges not adequately addressed by the point-to-point
Unfortunately, factory floors can topology of an Ethernet network. Ethernet has enough bandwidth to handle
be hot, noisy, dirty and prone to large one video stream but lacks the low latency required to handle closed loop
amounts of shock and vibration. This is process control, and requires substantial processing power to provide the
a less than ideal environment for video- communications protocol.
based processing systems. Another
problem is the amount of cabling often provide the communications protocol. munications any time a new destination
required by video imaging systems. An The source node must know all of the des- node is added, or if a task using the data
imaging system that monitors multiple tination nodes and must specifically send is moved to another node. These interde-
stages of a process with multiple cameras, a message to each of them. This requires pendencies produce a network that is not
distributed over tens or even hundreds of modification of the source node’s com- easily scalable and does not easily accom-
feet, can require significant amounts of
physical cabling, which can create a po- Data Processing / Control / Data Processing / Control / Data Processing / Control /
tential hazard on a factory floor. and/or Monitor Terminal and/or Monitor Terminal and/or Monitor Terminal
dle one video data stream, it lacks the low D Discrete Sensor
January 2006 25
TechnologyInContext
26 January 2006
TechnologyInContext
distance between nodes can be quite high: transfer data around the network ring at sending task doesn’t need to know where
using standard shortwave laser transceiv- 20 Mbytes/s, SCRAMNet GT supports the receiving task is located.
ers, this distance can be as high as 200 to 2.5 Gbit/s data rates and a throughput As video-based imaging systems are
300 meters. With longwave transceivers, of 210 Mbytes/s with a latency of less being more widely deployed, and video
the distance between nodes can reach up than 0.5 microseconds per node. It fea- resolution and speed increase, it is essen-
to 10 kilometers. Cabling is also reduced tures a one-to-many and many-to-many tial to ensure that enough bandwidth is
because only a single fiber optic cable runs built-in broadcast capability and ensures available. In December 2005, at the I/IT-
between each computer, as compared to a that all nodes receive updated informa- SEC Conference in Orlando, Florida, Cur-
point-to-point network, where every cam- tion without intervention from either host tiss-Wright exhibited a SCRAMNet GT
era and sensor must be wired individually or user. The original system supported a system with a total throughput load of 190
back to a panel connected to the process- maximum of 8 Mbytes of memory. Each Mbytes/s. Four video sources—two DVD
ing computers. SCRAMNet GT board, whether VME, players and two video cameras—were run
Because the processing can be han- PCI or PMC, comes with 128 Mbytes of from four nodes generating 50 Mbytes/s
dled remotely, the transceivers on the fac- memory (Figure 4). of streaming video data. Another task
tory floor do not need to be integrated in Another advantage of using shared generating 120 Mbytes/s of additional
high-speed processors, since all that is re- memory is the low programming cost as- data throughput was added to burden the
quired is the simple process of pulling in sociated with application programs. The system. The result was no video data deg-
the data and putting it in shared memory. system designer must assign data only radation or lost frames.
The high-performance processors can be to specific areas of shared memory. Ap-
placed remotely in a safe lab or control plication writers then use the data vari- Curtiss-Wright Controls
room environment. ables corresponding to these addresses Embedded Computing
SCRAMNet GT is the latest version and use the variable names as they would Data Communications
of the popular shared memory archi- normally. Tasks can be moved to other Dayton, OH.
tecture that was first introduced about processors without any changes to the ap- (937) 252-5601.
15 years ago. It is the highest band- plication itself. In actual practice, a task [www.cwcembedded.com]
width shared memory system available. could be talking to another task within
Although the original SCRAMNet the same computer, or to a task on the far
had a 150 Mbit/s data rate and could side of the ring. With shared memory, the
See
CURTISS-WRIGHTat CONTROLS
the RTECC Melbourne
February 28, 2006 Hilton Melbourne Rialto Place
Join us as we discuss how physical layer switches can make interop and
test labs more efficient and cost effective. Discover from the experts at
Curtiss-Wright Controls Embedded Computing how, by automating the
connections, these switches can lower capital equipment and opera-
tion costs, decrease configuration & reconfiguration times by wiring
once, improve efficiency, and decrease time-to-market on products/
projects. Go to www.rtecc.com/melbourne to pre-register and have
your badge waiting at the door. Event Hours are 8am - 3pm.
www.rtecc.com
rtecc_new.indd 3 January 200610:32:53 AM27
1/11/06
TechnologyInContext
Data Aquisition
A
al page, the s processing capability continues to system can use either embedded storage acquisition systems use RACE++, which
t resource. grow, signal processing systems are technology or a PC-based data recorder. If offers up to 533 Mbytes/s per 6U VME
chnology,
and products using ever larger amounts of sen- the application requires multiple channels slot. Newer systems being developed to-
sor data—in resolution, bandwidth and to disk, from 200 Mbytes/s up to several day use VITA 41 (VXS) technology to
number of channels—to perform their Gbytes/s, the system will typically use a scale up to 2.5 Gbytes/s per 6U slot. VXS
functions. Data acquisition and record- switched fabric interconnect to provide systems can use fabrics such as PCI Ex-
ing systems are required that can test and both scalability and modularity. press or Serial RapidIO, or point-to-point
support these advanced signal processors’ A variety of switched fabrics are avail- links based on the Xilinx Aurora proto-
capabilities. Fortunately, the same tools able with off-the-shelf support for modu- col. The choice of protocol depends on the
and technologies that enable faster signal lar data recorders. Many legacy radar data interoperability requirements within the
mpanies providing solutions now
processing—switched fabric interconnect
and FPGA-based processing—can also be
oration into products, technologies and companies. Whether your goal is to research the latest
plication Engineer, or jump to a company's technical page, the goal of Get Connected is to put you
vice you requireused to implement
for whatever advanced data acquisi-
type of technology, GigE SDRAM
nies and products tion
you systems forfor.a wide range of applica-
are searching
PowerPC CPU
tions, including radar.
Console Flash
Figure 1 Using a network model for the data acquisition system as a whole lets the
Get Connected system be viewed as a loosely coupled set of processing nodes, each with
with companies mentioned in this article. a PowerPC processor, local memory, I/O module site and bridge to the
www.rtcmagazine.com/getconnected
fabric.
28 January 2006
system and the complexity of the endpoint provides four independent fiber optic in-
solution, which is typically implemented terfaces connected to an onboard FPGA.
in an FPGA on each VXS card. The module also includes two banks of
One benefit of using a switched fabric DDR buffer memory to support wire-
is the built-in support for a network model speed buffering of all four data channels.
for the system as a whole. The system can When installed on a PCI-X carrier, the
be viewed as a loosely coupled set of pro- PMC module supports full throughput, 1
cessing nodes, each with a PowerPC pro- Gbyte/s transfers between all four chan-
cessor, local memory, I/O module site and nels and the host.
bridge to the fabric (Figure 1). The FPGA can be used to implement Figure 2 In many radar data acquisition
Nodes can be configured as either stor- a wide range of protocols, including Serial systems, a PMC module,
age or I/O nodes, depending on the type FPDP, Fibre Channel and Gigabit Ether- such as TEK Microsystems’
of I/O module installed. Because each I/O net, allowing the same module to sup- JazzFiber, delivers high-
module has its own dedicated processor, port different types of interfaces through speed fiber interfaces.
the software model is very simple. If a Fi- FPGA reconfiguration. Each processing Four independent fiber
bre Channel module is installed, the node chain is independent in the FPGA, allow- optic interfaces connect to
acts as a storage server, responding to cli- ing a single module to support a mix of an onboard FPGA and two
banks of DDR buffer memory
ent requests through the fabric network. protocols if required (Figure 3).
support wirespeed buffering
Alternately, if an I/O module is installed,
of all four data channels.
the node acts as both an autonomous I/O Adjunct Data Processing
server and a storage client, managing its Channels While the adjunct data channels tend
own I/O module and requesting storage to While the primary mission of a data to be lower speed, they also tend to require
disk through the fabric network. acquisition system is to record high-speed some processing for interpretation and
sensor data, most systems also require formatting of the data. Adjunct data chan-
High-Speed Fiber Optic Data some amount of adjunct low-speed data to nels can use off-the-shelf interfaces—such
Transfer be recorded as well. This low-speed data as Ethernet, 1553, SCRAMNet and the
In many radar applications, the sen- typically gives information about the plat- like—or they may require tailored low-
sor data being recorded is converted from form itself, which provides the operating level interfaces such as serial or parallel
analog to digital outside the recorder and context necessary for analysis of the high- TTL, ECL, EIA-485 or LVDS. As long
is transferred using high-speed fiber optic speed data. In some cases, the adjunct as the interface can be implemented on
interfaces. This approach makes it easy data affects the high-speed data record- a PMC or XMC I/O module, it can eas-
to insert a data recorder into the system ing process directly, modifying the type ily be integrated into the data acquisition
without degrading the signal integrity of or amount of data being recorded in real system. The network model enforces a
the data being acquired. The data recorder time as the platform state changes. modular approach to adjunct channels,
typically implements a copy mode that re-
broadcasts the input data, allowing the re-
corder to be inserted between the sensor DDR
January 2006 29
TechnologyInContext
30 January 2006
TechnologyInContext
One approach to data formatting is acquisition systems and largely indepen- RAID disk arrays. This makes it possible
to use a real-time implementation of the dent of the specific type of data being to develop very high-performance data
standard FAT32 file system for all data recorded. acquisition and playback systems, with
recording and playback. This file system Signal processing systems today application-specific tailoring where nec-
is directly supported by Windows, Linux make use of switched fabrics to create essary, while reusing existing hardware,
and Solaris workstations, allowing RAID modular, scalable solutions, using FPGA- software and FPGA components for the
storage arrays to be directly accessed by based processors to perform processing majority of the system.
standard workstations without requiring at higher densities than is possible with
special software or drivers. general-purpose processors. The use of
Each channel of data is written to the same tools and techniques supports TEK Microsystems
its own file on the disk, with a common a scalable and flexible approach to build- Chelmsford, MA.
format for headers and other adjunct in- ing data acquisition systems. Leverag- (978) 244-9200.
formation such as timestamps. The use of ing off-the-shelf hardware and software, [www.tekmicro.com].
a standard file system and a common for- along with tailoring when necessary
mat enables the development of a body through FPGA-based processing, allows
of transcription and verification soft- the use of industry standard components,
ware that is common to a range of data enclosures, backplanes, I/O modules and
We design solutions.
Mercury customers span multiple industries and face unique computing challenges – whether improving
yields for semiconductor wafer inspection, increasing throughput in medical imaging, rendering high-quality
animation within a defined budget, or optimizing image exploitation to identify IEDs. Why are we so driven to
tackle these difficult computing problems? Because at Mercury your challenges drive our innovation.
Our new Dual Cell-Based Blade solution incorporating IBM® Cell technology delivers
unprecedented performance and processing power. Contact us to learn how our
products and services can optimize your challenging applications.
Let us design an innovative solution for you.
www.mc.com/rtc12 | 866-627-6951
TM
January 2006 31
SolutionsEngineering
Switched Fabric Update
by T om Cox
RapidIO Trade Association
exploration
R
her your goal
peak directly apidIO technology is a fast growing
al page, the interconnect and fabric standard for Logical Specification
Part I Part II Part V
t resource.
embedded systems, providing in- Information necessary for the end point
chnology,
and products creased performance, improved efficiency to process the transaction. (i.e., transaction I/O Message Globally Future
System Passing Shared Logical
and lower cost. RapidIO technology is type, size, physical address)
Memory Spec
supported by a broad ecosystem of leading
vendors with multiple vendors shipping Part III
production switches, endpoints, FPGAs, Transport Specification
boards, software and systems. Serial RapidIO Information to transport packet from end Common
to end in the system. (i.e., routing address)
technology offers a high-speed physical layer Transport
Spec
that can be configured to match bandwidth
mpanies providing solutions now
requirements with different speed variants Part IV Part VI
oration into products, technologies and companies. Whether your goal is to research
and numbers of lanes.
the latestSpecification
Physical
plication Engineer, or jump to a company's technical page, the goal of Get Connected is to necessary
put you to move packet
Serialtype
vice you require for whatever RapidIO builds on the commu-
of technology,
Information
between two physical devices. 8/16 LP-LVDS 1x/4x LP Future
nies and products nication industry’s
you are searching for. common roadmap at (i.e., electrical interface, flow cntl) Serial Physical
the serial physical layer, using a variant of Specs
IEEE 802.3 Xilinx 10 Gigabit Attachment
Unit Interface (XAUI) today for 3.215
Gbits/s. For future 5 and 6 Gbit/s versions, Inter-Operability
it is using a variant of the work done on Specification
the Optical Internetworking Forum’s (OIF)
Common Electrical Interface (CEI).
End of Article
RapidIO architecture has no inher-
ent limitations preventing it from scaling
Compliance
Checklist
indefinitely into the future, following or
Figure 1 The RapidIO protocol can be over both serial and parallel interfaces
and is media-agnostic. Therefore, the serial specification is defined
Get Connected at the physical electrical layer and the rest of the specification is
with companies mentioned in this article.
www.rtcmagazine.com/getconnected preserved at the higher levels.
32 January 2006
6U VME64 engine with up to two IBM® 1.8 GHz PowerPC® 970FX CPUs
Up to 4GB of DDR400 SDRAM
VME interface - Universe II VME64x interface
Two 10/100/1000 Ethernet ports
Six USB ports
AC97 audio, two serial ports
Support for Linux™ OS
Ruggedized for up to 30G shock
NEW! TA64™
(AMD Turion 64 Mobil-based SBC)
So when your mission takes you in harm’s way, you can rely on Themis.
Transformational.
© 2005. Themis Computer, Themis, Themis logo, TA64, TGA3D+, USPIIe-USB and USPIIIi are trademarks or registered trademarks of Themis Computer. All other trademarks are property of their respective owners.
SolutionsEngineering
Parallel RapidIO
Clock Rate 8-bit Mode 16-bit Mode
Sustained Sustained Sustained Sustained
PEAK PEAK
32 byte Op 256 byte Op 32 byte Op 256 byte Op
250 MHz 8 Gb/s 4 Gb/s 7.5 Gb/s 16 Gb/s 8 Gb/s 15 Gb/s
500 MHz 16 Gb/s 8 Gb/s 15 Gb/s 32 Gb/s 16 Gb/s 30 Gb/s
750 MHz 24 Gb/s 12 Gb/s 22.5 Gb/s 48 Gb/s 24 Gb/s 45 Gb/s
1GHz 32 Gb/s 16 Gb/s 30 Gb/s 64 Gb/s 32 Gb/s 60 Gb/s
Serial RapidIO
Clock Rate 1-bit Wide 4-bit Wide
Sustained Sustained Sustained Sustained
PEAK PEAK
32 byte Op 256 byte Op 32 byte Op 256 byte Op
1.25 GHz 2 Gb 1 Gb 1.8 Gb 8 Gb 4 Gb 7.2 Gb
2.5 GHz 4 Gb 2 Gb 3.6 Gb 16 Gb 8 Gb 14.4 Gb
3.125 GHz 5 Gb 2.5 Gb 4.5 Gb 20 Gb 10 Gb 18 Gb
Table 1 Comparison of data rates in different modes and clock frequencies between Parallel and Serial RapidIO.
anticipating industry requirements. RapidIO source-synchronous interface. This means Comprehensive Link Protocol
technology has evolved over the past five that a clock is transmitted along with the A unique feature of RapidIO technol-
years to a full system dataplane fabric, with associated data. Source synchronous clock- ogy is that packet transmission is managed
extensions completed and in progress for: ing allows longer transmission distances at on a link-by-link basis. In the past, with
• RapidIO Flow Control Logical Layer higher frequencies. Two clock pairs are pro- synchronous buses, a mastering device
Extensions Specification vided for the 16-bit interface to help control had to exchange handshake signals with
• RapidIO Data Streaming Logical skew. The receiving logic is able to use the the target device. These signals indicated
Layer Extension Specifications receive clock for re-synchronization of the whether a transaction was acknowledged
- Phase I: Encapsulation and Traffic data into its local clock domain. and accepted by the target device. With an
Management Framework Since the Serial RapidIO specification interface such as the RapidIO specifica-
- Phase II: Advanced Traffic is only defined in the physical layer (Rapi- tion defines, it is not practical to rely on a
Management dIO technology defines the physical layer as synchronous handshake since the receive
• RapidIO Multicast Extensions Speci- the electrical interface and device-to-device port of a link is decoupled from the send-
fications link protocol), most of the controller remains ing port. Therefore, many interconnects
• RapidIO Next Generation Physical the same. As a result, much of the design have ignored this issue and rely on an
Layer Specifications knowledge and verification infrastructure end-to-end handshake to guarantee deliv-
are preserved (Figure 1). This eases system- ery. However, this has the disadvantage of
A comparison of the data rates be- level switching between parallel and serial preventing precise detection and recovery
tween the different modes of Parallel and links. During the initial development stages of errors and forces far longer feedback
Serial RapidIO is given in Table 1. of the Serial RapidIO specifications the loops for flow control.
designers decided to preserve as many of To address this issue RapidIO uses
Flexible Physical Interface the concepts found in the RapidIO parallel embedded control symbols for link-level
The RapidIO logical packet descrip- specification as feasible. The parallel speci- communication between devices. Pack-
tion is defined to be physical-layer-inde- fication includes the concept of packets and ets are explicitly tagged between each
pendent. This means that the RapidIO pro- in-band control symbols. link with a sequence number otherwise
tocol could be transmitted over anything These were delineated and differenti- known as AckID. The AckID is inde-
from serial to parallel interfaces, from ated by both a separate frame signal and pendent of the end-to-end transaction ID.
copper to fiber media. The first physical an “S” bit in the header. In the serial link Using control symbols, the receiving de-
interface considered and defined is known specification this delineation is accom- vice indicates for each packet whether it
as the 8- or 16-bit link protocol end point plished using spare characters (“K-codes”) has been received along with additional
specification (8/16 LP-LVDS). This speci- found in the 8B/10B encoding technique. buffer status information. Receiving de-
fication is defined as having 8 or 16 data In this way, the sending device indicates vices can immediately detect a lost packet,
bits in each direction along with clock and to the receiving link partner the start of a and through control symbols, can re-syn-
frame signals in each direction. packet, end of packet or embedded control chronize with the sender and recover
The 8/16 LP-LVDS interface is a symbol using these codes. it without software intervention. The
34 January 2006
SolutionsEngineering
receiving device then forwards the packet Robust Electrical Interface reduced performance, non-standard bridg-
to the next switch in the fabric, and so on, Serial RapidIO uses differential cur- ing and more complex system design than
until the packet reaches its final target. rent steering drivers based on those de- the adoption of an application-appropriate
Serial RapidIO allows longer trans- fined in the 802.3 XAUI specifications. standard, such as RapidIO technology.
mission distances and thus involves lon- This signaling technology was developed Ethernet developed for system-to-system
ger loop latencies in providing feedback to drive long distances over backplanes. local networks requires heavy over pro-
between the receiver and transmitter on For Serial RapidIO technology, two visioning for embedded applications and
a link. Consequently, the Serial physical transmitter specifications were designated: lacks determinism, reliability and robust
layer specification increases the number a short run transmitter and a long run error handling.
of AckID values from 8 to 32. transmitter. The short run transmitter is The RapidIO fabric provides a robust
Additionally, the Serial RapidIO used mainly for chip-to-chip connections packet-switched system level intercon-
specification now defines a transmitter- either on the same printed circuit board or nect. It provides a partitioned architecture
controlled flow control scheme whereby across a single connector such as that for that can be enhanced in the future. It en-
the receiving port provides information to a mezzanine card. The minimum swings ables higher levels of system performance
its link partner about the amount of buf- of the short run specification reduce the while maintaining or reducing implemen-
fer space it has available. With this infor- overall power used by the transceivers. A tation costs. A RapidIO end point can
mation, the sending port can allocate the user can further reduce the power by low- be implemented in a small silicon foot-
use of the receive buffers of the receiving ering the termination voltages. print. Proven industry-standard signaling
port. The sending port does not have to be The long run transmitter uses larger schemes (LVDS, XAUI) are used for the
concerned that one or more of the packets “voltage swings” that are capable of driv- physical interfaces. Error management in-
shall be forced to retry. ing across backplanes. This allows a user cludes the ability
Looking For More?to detect multi-bit errors to download
Visit www.rtcmagazine.com
additional technical information related to this article.
to drive signals across two connectors and and survive most multi-bit and all single
PCS and PMA Layers common printed circuit board material. bit errors. Even with all these capabili-
The Serial RapidIO specification To ensure interoperability between driv- ties, the RapidIO protocol overhead and
uses a physical coding sublayer (PCS) and ers and receivers of different vendors and latency are comparable to current bus
physical media attachment (PMA) sub- technologies, AC coupling must be used technologies and significantly better than
layer to organize packets into a serial bit at the receiver input. local area network-based fabric technolo-
stream at the sending side and to extract The engineer’s interconnect choices gies such as Ethernet.
the bit stream at the receiving side. This may include use of proprietary, home-
terminology is adopted from IEEE 802.3. grown technologies, legacy interfaces or
Besides encoding for transmission application-appropriate emerging stan- Looking For More?
and decoding for reception, the PCS func- dard technologies. The three leading Visit www.rtcmagazine.com
tion is also responsible for idle sequence choices are Ethernet, PCI Express and to download additional technical
information related to this article.
generation, lane striping, lane alignment RapidIO technology. While the three in-
and de-striping on reception. The PCS terconnect technologies have some simi-
uses 8B/10B encoding for transmission larities, they are quite different in terms The RapidIO Trade Association
over the link. of technical merit. In many cases they can [www.RapidIO.org].
The PCS layer also provides the be highly complementary in the overall
mechanisms for automatically deter- system architecture landscape.
mining the operational mode of the RapidIO was designed specifically
port as either 1-lane or 4-lane, and pro- as a widely applicable, flexible, extensible
vides for clock difference tolerance be- system fabric for embedded infrastructure
tween the sender and receiver without equipment including networking, storage
requiring flow control. The PMA func- and communication systems. PCI Express
tion is responsible for serializing 10- was formulated as an improvement on the
bit parallel code-groups to/from a se- Peripheral Component Interconnect bus,
rial bit stream on a lane-by-lane basis. primarily for the commercial comput-
Upon receiving data, the PMA function ing market. Historically, PCI, because of
provides alignment of the received bit its ubiquitous nature and the consequent
stream to 10-bit code-group boundar- economies of scale, has been adopted
ies, independently on a lane-by-lane within embedded systems despite not nec-
basis. It then provides a continuous essarily providing optimum functionality.
stream of 10-bit code-groups to the There may be a similar desire to force-fit
PCS—one stream for each lane. The PCI Express into applications beyond the
10-bit code-groups are not observable intent of the architectural scope of that in-
by layers higher than the PCS. terconnect. However, this is likely to be
at the expense of inferior functionality,
January 2006 35
/ %"&"'"%#"%'"!
%'&%&%)
, ) )"#%& "%' "!! % %!&&! '
#"*%"%"&"'.!"*&
'"(%')&
''&'!(&'%%""'&%" )%'""'$(
!
)%'""'$(%!
-' ' *'
"%
""
"
/52 'IG% 4/%
$/%3 (!2$
2%!, 4)-% $UAL #HANNEL 4/% !CHIEVES
,ATENCY «SEC ¢«SEC -"YTESSEC AT #05
&./+$-)+#-*.#"-/"* 4+ ( 0./&*"!-*.#"-/"
,,(& /&+*/+")+/",,(& /&+*
,+-/#0((!0,("3+,"-/&+*."*!*!-" "&1"
/"* 4//&./& .
4/".."
)&*
6."
4/".,"-." +*!
2&/%0."!
)3 6."
)"* 6."
./!!"1 6."
,+-/%(#!0,("3+,"-/&+*."*!+*(4
4/".."
2&/%0."!
VISIT OUR WEBSITE AT WWWCRITICALIOCOM
SolutionsEngineering
Switched Fabric Update
S
Get Connected with technology and
torage systems large and small have companies is PCIe’sproviding
scalability of bandwidth,
solutions now robust- doubled once more by moving from PCI’s
begun adopting PCI Express (PCIe) Getness and integrity
Connected is a newofresource
data transfer, reduced
for further exploration 64-bit, 66 MHz to the 64-bit, 133 MHz
technology as the interconnect stan- into products,
pin count of ASICs
technologies and simplified
and companies. Whether circuit
your goal performance of PCI-X. This worked well
dard at the card-to-card level, reflecting board
is to research thelayout—all
latest datasheetoffrom
which add up
a company, speaktodirectly
sys- enough in getting throughput to one giga-
a natural evolution from last year’s
withchip- tems
an Application that are
Engineer, faster
or jump and less technical
to a company's expensive.page, the byte per second. However, the next jump
goal of Get Connected is to put you in touch with the right resource.
to-chip level interconnect deployment Extending
Whichever level of service you requirethe bandwidth
for whatever type ofof storage
technology, in performance meant either doubling the
of PCIe. There are many reasonsGet forConnected
the controllers
will help youwas
connecteasy back
with the in theand32-bit
companies products bus width to 128 bits or doubling the bus
transition, foremost among themyou being
are searchingPCIfor. days. Throughput could be doubled rate to 266 MHz. Both have serious draw-
the availability of PCIe-based chipsets for by moving from PCI’s 33 MHz to its next-
www.rtcmagazine.com/getconnected backs. First, 128-bit bus signals would re-
storage-system controller applications. But generation 66 MHz. It was doubled again quire excessive board space for the traces
beyond that, what is driving the transition by going from 32-bit to 64-bit PCI, then and make for high-pin-count chips, adding
cost, footprint, power dissipation and noise
with all those I/Os switching. Secondly,
CPU CPU
Get Connected with technology and companies providing cranking the bus frequency up to 266 MHz
solutions now
increases the effect of clock skews and
Get Connected is a new resource for further exploration into products, technologies and companies. Whether your goal is to research the l
datasheet from a company, speak directly with an Application Engineer,makes it aextremely
or jump to difficult
company's technical page,tothemeet
goal of the
Get Connected
in
MEM
system
touch with the right resource. Whichever level of service you require for timing
whatever type ofrequirements.
technology, This is where
Get Connected will help Root a serialized architecture makes sense.
Complex with the companies and products you are searching for.
you connect
www.rtcmagazine.com/getconnected
Scalability and Reliability with
PCI Express
PCI Express PCI Express Storage interconnects such as U320
Bridge Switch SCSI, SAS, SATA2 and Multi-Gigabit Fi-
bre Channel are delivering performance
improvements with increased connection
Products End of Article
speeds. PCIe offers the ability to deploy
PCI Express multi-port adapters and RAID controllers
without creating a local I/O bottleneck.
PCI Express
Switch
End Point
January 2006 39
Get Connected with companies mentioned in this arti
www.rtcmagazine.com/getconnected
SolutionsEngineering
PCI-X Slots drivers are required to accommodate PCIe includes a hot plug capability,
PCIe. From the viewpoint of the system allowing users to replace add-in cards
model, each PCIe port is a virtual PCI-to- and other hardware modules to perform
PCIe device and has its own set of PCIe maintenance without powering down the
PCI-X
PCI-X Bus
configuration registers. It is through the system. Each downstream port includes a
Chipset
upstream port that the BIOS or host can standard hot plug controller. If the PCIe
configure the other ports using standard switch used in an application where one or
PCI Express PCI enumeration. The virtual PCI-to- more of its downstream ports connect to
Add-On Slot
PEX PCIe bridges within PCIe switches and PCIe slots, each port’s hot plug controller
bridges are compliant with the PCI and can be used to manage the hot plug event
8114
Host CPU
PCI-X to-PCI Express PCIe system models. The Configura- of its associated slot. Furthermore, its up-
Bridge tion Space Registers (CSRs) in a virtual stream port is a hot plug client, allowing
Figure 2 Adding PCI Express to PCI-X primary/secondary PCI-to PCIe bridge it to be used on hot-plug-capable adapter
System Boards Requires are accessible by type 0 configuration cards, backplanes and fabric modules.
Reverse Bridging. cycles through the virtual primary bus
interface—matching bus number, device System Board Uses
number and function number. PCI Express
PCIe provides a more robust inter- The earliest deployment of PCIe has
connect compared with PCI-X. This en- been for chip-to-chip interconnect on sys-
hances system reliability and data avail- tem boards, as shown in Figure 1. In this
ability. Since PCIe is a point-to-point example, a dual-host system’s CPUs and
architecture, it eliminates the shared bus memory chips are interconnected via a
that is used in PCI-X. With a shared bus, PCIe root complex. A root complex is a
it is never clear how many devices may re- specialized PCIe switch. The root com-
side and what impact the various devices plex is also connected to a standard PCIe
may have on the bus bandwidth. But with switch to provide several channels of I/O
a dedicated channel for each endpoint, fan-out. In this example, the standard
quality of service and throughput can be PCIe switch fans out the root complex to a
Figure 3 This U320 SCSI host adapter
from PLX has a PCI-to-PCI deterministic. PCIe bridge, switch and a native endpoint.
Express Bridge added to a In the PCIe architecture, the Data The PCIe bridge allows the creation of
PCI-X HBA. Link Layer (Layer 2) provides link man- PCI/PCI-X slots for various I/O functions
agement and ensures data integrity us- including legacy adapter cards and com-
As storage interconnects transition to 10 ing error detection and correction. This munications ports. The second-level stan-
Gbit/s speeds, PCIe can provide the basis layer calculates and appends a Cyclic dard PCIe switch provides fan-out for sev-
for a long-term roadmap for storage-per- Redundancy Check (CRC) and assigns eral PCIe slots. The PCIe endpoint could
formance capability. InfiniBand connec- a sequence number to the information be any number of PCIe devices, such as
tions enabled for PCIe deliver data centers sent from the data packet. The sequence PCIe-native Gigabit Ethernet controllers
a full 10 Gbits/s of clustering connectiv- number allows proper ordering of the or adapters.
ity today. Table 1 shows the performance data packets. The CRC verifies that Some designs may involve a system
scalability as a function of lane count data from link to link has been correctly board with plenty of performance, but fea-
for 2.5 GHz (Gen 1) PCIe. With Gen 2 transmitted. In addition, the PCIe speci- ture only a PCI-X interconnect and need to
(5 GHz) on the horizon, these bandwidths fication allows for providing end-to-end be able to connect to PCIe add-in cards. A
will double, from 8 Gbytes/s to 16 Gbytes/s CRC protection (ECRC) and poison-bit PCIe-to-PCI-X bridge allows the migra-
per direction, for an aggregate maximum support to enable designs to guarantee tion of an existing PCI-X storage system
bandwidth of 32 Gbytes/s. error-free packets. While these features board to accept PCIe add-in cards. This
When the system software finds a are optional in the PCIe specification, allows the creation of a PCIe system board
PCIe bridge or switch, it looks just like they are already integrated in some ven- without having to qualify a completely new
a PCI bridge; no software changes or dors’ PCIe devices. chipset. The catch here is that the bridge
Table 1 PCI Express Scalable bandwidth (Gen 1, 2.5 GHz PHY). The yellow box indicates the equivalent maximum bandwidth for a
32-bit, 33 MHz PCI bus; the green, 32-bit, 66 MHz and the magenta, 64-bit, 133 MHz PCI-X functionality.
40 January 2006
SolutionsEngineering
the secondary processor assumes control list of features that build on the PCI and
PCI-X without bringing down the system. This PCI-X legacy—hot plug, ECRC, quality
Device failover operation requires a non-transpar- of service, deterministic bandwidth and
ent bridge in a PCI-X environment. With a generically more robust interconnect
CPU
PLX
PCIe, a non-transparent switch such as the based on a point-to-point topology—are
Bridge
PLX PEX 8532 provides the high-speed enabling robust storage systems for real-
interconnection and domain isolation re- time and mainstream application.
quired for dual-host operation.
PCIe has emerged as the foremost PLX Technology
interconnect standard for chip-to-chip Sunnyvale CA.
Figure 4 Non-Transparent Bridging connections and add-in cards. It provides (408) 331-6400.
between PCI and PCIe. several advantages over its predecessors [www.plxtech.com].
while maintaining software compatibility
must be capable of reverse-bridging mode, and performance scalability. PCIe’s long
which isn’t found on all PCIe bridges. With
reverse mode, the PCI-X port is the up-
stream port, and the PCIe port is the down-
1 2 3
D
D
stream port (Figure 2). DESIGN DEVELOP DEPLOY
Most of today’s high-performance
storage cards use PCI-X. Many such D
January 2006 41
IndustryInsight
Dual-Core Processors
D
chnology,
and products ual-core CPUs have been com- performance with each CPU generation, transmission. At the highest frequencies,
mercially available since 2000 but the current level of miniaturization of tunneling can become so extreme that it
when IBM first introduced the feature sizes is forcing IC manufacturers totally negates signal recognition.
IBM POWER4. They provide a method to look to more innovative solutions. To drive high performance across
for gaining greater performance while The problems caused by extreme smaller, more powerful transistors re-
avoiding the increases in form-factor, in- feature density are interrelated. Electri- quires more power. In turn, higher power
cremental heat and power requirements cal features in extreme proximity produce results in unacceptable levels of waste
associated with the higher feature density quantum effects, i.e., electrons that ran- heat as power (wattage) increases and
mpanies providing solutions now
of fast single-core processors. In pursuing domly tunnel across the CPU’s features produces more unwanted quantum ef-
oration into products, technologies and companies. Whether your goal is to research the latest
dual-core
plication Engineer, or jump to aCPUs, thetechnical
company's major IC manufactur-
page, causing isinterference
the goal of Get Connected to put you with normal signal fects. Machines with dense CPUs run-
vice you require ers have acknowledged
for whatever type of technology,that the historical ning at higher wattage are noisier, be-
nies and products approach of gaining
you are searching for. performance by sim- Semiconductor Technology Generations cause they require additional, more pow-
ply increasing CPU feature density has by Feature Size erful fans for cooling. Fan motors add yet
reached diminishing practical returns. (averaged across vendors) more electrical noise.
The current generation of high-per- Dual-core processors, such as the
Feature size
formance CPUs (Table 1) is 90 nanome- Year (in nanometers) Comments AMD Opteron, can mitigate these prob-
ters (nm) between surface features, thus 1982 1,500 lems while at the same time enabling sig-
entering the realm of bona fide nano- nificant increases in performance.
1993 600
technology, which is 100 nm and below.
End of Article
However, at this extreme density, there
are many unwanted effects. The industry
1998
1999
250
180
The AMD Opteron Dual-Core
Processor
has grown accustomed to ever improving 2001 130 The AMD Opteron processor is a
2003 90 current generation
high-density, 90-nm CPU, packing 233
Get Connected million transistors on a 199 mm2 die. The
2005 65 just beginning to appear
with companies mentioned in this article. chip is microarchitected to lessen un-
www.rtcmagazine.com/getconnected Table 1 wanted effects, principally through thread-
42 January 2006
Get Connected with companies mentioned in this article.
www.rtcmagazine.com/getconnected
IndustryInsight
medical and security imaging, storage and 2x AMD Operton 265 [2 Core] 1.8GHz 2x1ML2
telecommunications. The dual-core AMD 2x AMD Operton 280 [2 Core] 2.4GHz 2x1ML2
Opteron processor enables basic reference Figure 1 In floating-point performance tests (iterations per second) conducted at
designs that can be modified to meet these WIN Enterprises, the 2.4 GHz dual-core AMD Opteron 280 shows an 85%
systems’ needs for compactness, design improvement over the 2.6 GHz single-core Opteron 252. This improvement
longevity, lower power consumption, low can be attributed to the dual-core processor’s ability to multithread its tasks.
latency and high reliability, often in harsh
environments.
Single-Core Opteron Dual-Core Opteron
In response to market forces and
evolving technology in x86 processors,
such as the AMD Opteron and Pentium M, CPU0 CPU0 CPU1
many designers of high-performance em-
bedded systems are turning from highly
1MB 1MB 1MB
specialized platforms to x86-based solu- L2 Cache L2 Cache L2 Cache
tions. These systems typically run either
Windows Embedded XP or Linux. System Request Interface System Request Interface
Regardless of the operating system
chosen, it should be dual-core-aware in Crossbar Switch Crossbar Switch
order to provide the benefit of multi-
threading. The dual-core AMD Opteron
provides improved 32-bit legacy appli-
HyperTransport HyperTransport
cation support, in addition to concur- Memory
0 1 2
Memory
0 1 2
rent 64-bit performance. This ability to Controller Controller
support legacy applications enables a
smooth upgrade path in the enterprise
market and expands the dual-core Opter-
on’s flexibility in the high-performance Figure 2 The dual-core AMD Opteron processor utilizes the same basic architecture
embedded market. as the single-core Opteron, but reduces board-level footprint. The two
Terascala, which manufactures stor- cores connect to a common crossbar that manages processing tasks and
age appliances for Linux-based clusters, a dedicated L2 cache for each core provides scalability.
January 2006 43
IndustryInsight
44 January 2006
IndustryInsight
cluded dual-core Opteron processors Dual-core CPU technology is at the WIN Enterprises
with HyperTransport, PCI Express, leading edge of high-performance em- N. Andover, MA.
USB 2.0, ALC850 audio and Gigabit bedded designs. Dual-core and Hyper- (978) 688-2000.
[www.win-ent.com].
Ethernet. Transport technology enable a significant
In designing this board, the nVidia advance as a standardized x86-based plat-
nForce 2200 chipset was chosen to work form that fulfills the requirements of low
with the AMD CPU. However, the two latency and high performance in a small
had never been used together in a small form-factor. This approach is seeing a
form-factor, which presented some de- high level of OEM interest, evaluation and
sign challenges. The successful mating application.
of dual-core and small form-factor was
a breakthrough for the embedded OEM
market.
Other challenges were overcome by
designing a 10-layer motherboard rather
than utilizing the traditional 6 layers.
Nextcom, a leading manufacturer
of extreme performance, mobile, small-
footprint computing products, is utiliz-
ing the Opteron dual-core CPUs on a
related design, the MB-06048, which
is a PICMG 1.3 form-factor. This is be-
ing used in a field-rugged, mobile data
communications server used by military
and government agencies. The advanced
SBC enabled Nextcom to respond to
market requirements for a distributed
computing appliance that integrates
legacy technology, performs multiple
processes simultaneously, utilizes the
advantages of COTS technology and al-
lows application customization as mar-
ket needs evolve.
The high-performance comput-
ing power of both single- and dual-core
AMD Opteron processors is being lever-
aged in Nextcom’s field-deployable units,
the FleXtreme Vigor and NextDimension
products. These small units top out at 2.6
GHz per processor. They use the stack-
able HyperTransport extension boards to
offer quad-core CPU processing capabil-
ity to military, government agency and
other customers.
A WIN PICMG 1.3 reference design
is also being used by a major workstation
vendor in its medical imaging solution.
Software Considerations
Software is increasingly a concern
in high-performance embedded designs,
and that usually means Linux. To com-
plement its efforts in high-performance
small form-factor designs, WIN devel-
oped its own standard BIOS, as well as
a downloadable Linux image for product
testing and a Linux SDK.
January 2006 45
IndustryInsight
Dual-Core Processors
exploration
er your goal by D
errick Keefe and David Inglis
peak directly QNX Software Systems
al page, the
t resource.
F
chnology, MPC8641D
and products aced with growing energy consump-
tion and excessive operating tem-
e600 Core e600 Core
peratures caused by high CPU clock 1 MB
L2
1 MB
L2
speeds, microprocessor vendors have ad- 32 KB 32 KB (ECC) 32 KB 32 KB (ECC)
opted a new approach to boosting system D-Cache I-Cache D-Cache I-Cache
performance: integrating multiple, inde- MSS MSS
pendent processor cores on a single chip. MPX Bus
Intel, for example, has proclaimed that all
mpanies providing solutions now
of its new CPUs will use multicore archi-
oration into products, technologies and companies. Whether your goal is to research the latest 64-bit DDR/ 64-bit DDR/
tectures
or jump toand recentlytechnical
produced a roadmap
QUEUE
QUEUE
Get Connected Figure 1 Multicore chips such as the Freescale MPC8641D dual-core PowerPC
with companies mentioned in this article.
www.rtcmagazine.com/getconnected
processor offer much better performance per watt than existing
uniprocessor designs, as well as truly concurrent multi-tasking.
46 January 2006
Get Connected with companies mentioned in this article.
www.rtcmagazine.com/getconnected
"7"*-"#-
&
$0..&3 '30.
%1.'&"563&4 $*"-
50 '6--.*
s .4 80 ,INUX 6X7ORKS AND 3OLARIS ALL RUN ON
BOARD AND
THERES 08% SUPPORT FOR DISKLESS BOOTING IN SECURE SYSTEMS
-
s 4WO 'B %THERNET PORTS IN COMPLIANCE WITH 6)4! FOR
BACKPLANE FABRIC SWITCHING