The magazine of record for the embedded computing industry
COM Express: SMALL MODULES TAKE ON
BIG JOBS Ethernet Switching Brings Speed and Density Home Small Modules Move DAQ Closer to the Signals IP–The Soul of Programmable Logic
An RTC Group Publication
Kontron and the Kontron logo are registered trademarks of Kontron AG. All other trademarks are the property of their respective owners. ÂŠ2008 Kontron America, Inc.
1-888-294-4558 - firstname.lastname@example.org â€” EMEA: +49 0 800 7253756 - Asia: +886 2 2910 3532
COM Express: SMALL MODULES TAKE ON BIG JOBS
21 Desktop Chassis, Industrial Computer (CompactRIO) and USB Single-Module Carrier
26 Some of the form-factors specified in the JTRS HMS inventory
38 GHz Digital Receiver PMC/XMC Module and Dual Multiband Tranceiver with FPGA
Technology in Context
COM Express Solutions 6 COM Express− A Living Standard for Modular Solutions 12 Industry Insider 9Latest Developments in the Embedded Marketplace Editorial The Illusion of Security
Christine Van De Graaf, Kontron
Products & Technology Newest Embedded Technology Used by Industry Leaders
News, Views and Comment Embedded Computer Business on Fast Track Despite Economic Forecasts
The Best Network Seems Like No Network Iain Kenney, SMC Networks
Industry Insight Data Acquisition
More with Less 20 Doing (Hardware, That Is)
Brett Burger, National Instruments
Data Acquisition: Harnessing Small Modules with Flexible I/O Scott Hames, GE Fanuc
SYSTEM INTEGRATION IP for Programmable Logic
the Use of FPGAs with Optimized IP 28 Enabling Tom Hill, Xilinx
Developers Empowered by the FPGA Revolution 34 Software Eric Schneider, Eridon
Digital Receiver PMC/XMC Module offers 2- or 4- Channel 38GHz Oppreation GE Fanuc Intelligent Platforms
Multiband Transceiver with 38Dual FPGA Delivers Improved A/D Performance Pentek
Industry Watch Data-Oriented Architecture
Oriented Architecture: 40Data Loosely Coupling Systems into “Systems of Systems”
Rajive Joshi, Ph.D, Real-Time Innovations Inc.
Digital Subscriptions Avaliable at www.rtcmagazine.com/digital
C OTS Your product
cPCI PMC VME PCI ATX choice of OS
January 2008 Publisher PRESIDENT John Reardon, johnr@r tcgroup.com EDITORIAL DIRECTOR/ASSOCIATE PUBLISHER Warren Andrews, warrena@r tcgroup.com
Editorial EDITOR-IN - CHIEF Tom Williams, tomw@r tcgroup.com MANAGING EDITOR Marina Tringali, marinat@r tcgroup.com
Standard & custom designs
COPY EDITOR Rochelle Cohn
Extended temp. conduction or convection cooled SBCs
Fast, flexible, reliable
CREATIVE DIRECTOR Jason Van Dorn, jasonv@r tcgroup.com ART DIRECTOR Kirsten Wyatt, kirstenw@r tcgroup.com GRAPHIC DESIGNER Christopher Saucier, chriss@r tcgroup.com DIRECTOR OF WEB DEVELOPMENT Marke Hallowell, markeh@r tcgroup.com WEB DEVELOPER Brian Hubbell, brianh@r tcgroup.com
ISO 9001:2000 Certified 321-452-1670 Visit us: www.OTIsolutions.com Untitled-1 1
11/8/07 9:42:51 AM
Advertising/Web Advertising WESTERN REGIONAL ADVERTISING MANAGER Stacy Gandre, stacyg@r tcgroup.com (949) 226 -2024 WESTERN REGIONAL ADVERTISING MANAGER Lauren Hintze, laurenh@r tcgroup.com (949) 226 -2014 EASTERN REGIONAL ADVERTISING MANAGER Nancy Vanderslice, nancy v@r tcgroup.com (978) 443 -2402 INSIDE SALES MANAGER Carrie Bowers, carrieb@r tcgroup.com (949) 226 -2029
Billing Mag gie McAuley, mag giem@r tcgroup.com (949) 226 -2024
To Contact RTC magazine: HOME OFFICE The RTC Group, 905 Calle Amanecer, Suite 250, San Clemente, CA 92673 Phone: (949) 226-2000 Fax: (949) 226-2050, www.rtcgroup.com EASTERN SALES OFFICE The RTC Group, 96 Dudley Road, Sudbury, MA 01776 Phone: (978) 443-2402 Fax: (978) 443-4844 Editorial Office Warren Andrews, Editorial Director/Associate Publisher 39 Southport Cove, Bonita, FL 34134 Phone: (239) 992-4537 Fax: (239) 992-2396 Tom Williams, Editor-in-Chief 245-M Mt. Hermon Rd., PMB#F, Scotts Valley, CA 95066 Phone: (831) 335-1509 Fax: (408) 904-7214
Published by The RTC Group Copyright 2007, The RTC Group. Printed in the United States. All rights reserved. All related graphics are trademarks of The RTC Group. All other brand and product names are the property of their holders.
The Illusion of Security by Tom Williams, Editor-in-Chief
t a recent conference, I was treated to an in-depth examination of computer security issues, accompanied by excerpts from the recent Bruce Willis film, “Live Free or Die Hard,” which depicts how the U.S., along with its economic and physical infrastructure, can be brought to a complete halt by a concerted attack on our computer networks. The upshot of the conference was the realization that our computer networks are presently woefully insecure and it is possible to make them much more secure than they currently are. But—at least in my own interpretation—the idea of total security is an illusion. I am reminded of what is possibly the last short story written by the early 20th-century author, Franz Kafka called “The Burrow.” The story is about a non-specific animal that has built itself a huge, labyrinthine burrow in which to safely store its food and to keep itself safe from enemies and intruders. At first, all seems well, but the animal frets constantly about the entrance, which, though camouflaged with moss, cannot be completely concealed, especially when the animal enters or leaves the burrow. Then one day the animal awakens inside its burrow and becomes aware of a barely audible hissing sound coming from no specific location. After that, all the animal’s efforts are directed at abating that sound, but to no avail. The last sentence of the (interestingly) unfinished story is, “but everything remained unchanged.” Note that the animal is never actually attacked by an enemy; it is a story about self absorption and insecurity. In contrast, there have been real attacks on our computer networks, some of which appear to have been probing exercises that originated in China. The threat of identity theft is also very real. I will not go into a litany of threats, which can be found elsewhere, but it is now clear that there are enormous efforts underway by both government and private entities to produce more secure networked systems. These include advanced encryption, secure operating systems and hardware-based security technology. Currently, however, most networked systems, which are commonly based on Windows or Linux operating systems, are typically rated by what are known as the “common criteria” evaluation assurance levels (EAL) at EAL4 on a scale from 1 to 7. Level 4 is designated for users who “require a moderate to high level of independently assured security in a conventional commodity.” They are NOT secure against a determined or sophisticated attack, nor can they economically be made so. Period. This does not mean that there are not critical points, such as power plants and military installations that are not being made very much more secure than your PC. But given the enormous deployed base of moderately secured systems, the uncertainty about how they are all actually connected and how the various networks are managed, how will we ever know how secure ev-
erything is? The answer, of course, is that we cannot. However, over time, it is quite possible that the situation will improve with much more secure “separation kernel” style operating systems being made commercially available as OEM products that can be integrated with commercial operating systems to provide much higher levels of security—perhaps as high as EAL6+. Still, the only realistic view of security is how much effort a determined enemy is willing to devote to hacking into a given site. You want to keep your data secure? Great. Lock your system in a vault, don’t connect it to a network and don’t turn it on. Beyond that, it’s a relative game. In addition, entities like the NSA don’t really want your or my system to be absolutely secure because they still want to have some option for access. Whatever the merits of that, if there is some kind of back door, it can also be discovered by others. According to security lore, there are PhDlevel experts in certain countries constantly studying our systems and looking for vulnerabilities, which they catalog against the day they may want to attack. Whenever one is discovered and fixed, it gets crossed off the list. But it’s a long list. As our lives become ever more intertwined with networked systems ranging from telecom systems to small devices deployed in industry, businesses and homes, we make a compromise with security. There is much that can, should and is being done to improve this situation and bring not only better operational security but more peace of mind. But it requires constant effort, analysis and innovation and no one will ever be able to certify with absolute assurance that it is completely secure. There is still that annoying hissing sound. . .
g Embedded World in Nurember ld show
Wor I will be attending the Embedded ruar y. Feb of k wee last the rg in Nurembe press to allow Unfortunately, the show does not t means Tha ite. on-s only pre- register online, but ple peo s pres h whic w kno exhibitors do not ointments are attending and can’t make app g at or bitin in advance. Any companies exhi ld like to set wou who ld attending Embedded Wor at: me tact con se up meetings, plea (831) 335 -1509 or at: tomw@r tcgroup.com.
GE Fanuc Intelligent Platforms
There’s life in the old dog yet. You can even get him to do things he didn’t used to be able to do. VMEbus technology may be 25 years old - which, by industry standards, makes it a pretty old dog - but at GE Fanuc Intelligent Platforms, you’d never know. That’s because our commitment to VME has seen us continually push the boundaries of what it can do, to the point where it is still capable of fulfilling the requirements of today’s most demanding applications. Take, for example, what we’re doing with VXS – the way forward for VME users who want to leverage their investment in legacy systems to the maximum and take advantage of what serial switched fabrics bring to the table. The DSP220 VXS multiprocessor and CRX800 VXS Serial RapidIO managed switch bring industry-leading performance: the DSP220 features four single- or dual
core Freescale™ PowerPC® 8641 system-on-chip nodes and an XMC slot, while the CRX800 provides 22 ports of x4 Serial RapidIO at speeds up to 3.125Gbits/second per lane. Both are available in six ruggedization levels. And whether you’re planning to deploy on VME, on VXS or on VPX, GE Fanuc Intelligent Platforms’ AXIS Advanced Multiprocessor Integrated Software development environment means you develop just once – not three times. AT GE Fanuc Intelligent Platforms, our state of the art VPX, VXS and VME solutions mean that you can indeed teach an old dog new tricks.
© 2008 GE Fanuc Intelligent Platforms, Inc. All rights reserved.
G e t o n t h e F a s t Tr a c k Donâ€™t let custom requirements slow you down. Our existing products and custom design expertise help our clients to quickly meet complex design requirements. VadaTech is a leader in AdvancedTCA, AMC, cPCIe, VXS, XMC and custom board-level technology. Dedicated to leading-edge technology for the most demanding applications; we pride ourselves on our extraordinary capability to quickly create designs in response to our customer requirements. Our service to the aerospace and telecommunications markets has resulted in a quickly growing line of AdvancedTCA, AMC, cPCIe, VXS, XMC and custom boards for processing, communication, storage, and platform interoperability.
Sprint past the competition with VadaTech as your teammate!
W W W . V A D A T E C H . C O M T E L 1 . 7 0 2 . 8 9 6 . 3 3 3 7
IndustryInsider january 2008
Nanowire Battery Holds Ten Times the Charge of Existing Ones Researchers at Stanford University have found a way to use silicon nanowires to reinvent the rechargeable lithium-ion batteries that power laptops, iPods, video cameras, cell phones and countless other devices including, soon, hybrid and electric vehicles. The new version, developed through research led by Yi Cui, assistant professor of materials science and engineering, produces 10 times the amount of electricity of existing lithium-ion, known as Li-ion, batteries. A laptop that now runs on battery for two hours could operate for 20 hours, a boon to ocean-hopping travelers. “It’s not a small improvement,” Cui said. “It’s a revolutionary development.” The breakthrough is described in a paper, “High-Performance Lithium Battery Anodes Using Silicon Nanowires,” published online Dec. 16 in Nature Nanotechnology, written by Cui, his graduate chemistry student Candace Chan, and five others. The greatly expanded storage capacity could make Li-ion batteries attractive to electric car manufacturers. Cui suggested that they could also be used in homes or offices to store electricity generated by rooftop solar panels. “Given the mature infrastructure behind silicon, this new technology can be pushed to real life quickly,” Cui said. The electrical storage capacity of a Li-ion battery is limited by how much lithium can be held in the battery’s anode, which is typically made of carbon. Silicon has a much higher capacity than carbon, but also has a drawback. Silicon placed in a battery swells as it absorbs positively charged lithium atoms during charging and then shrinks during use as the lithium is drawn out of the silicon. This expand/shrink cycle typically causes the silicon (often in the form of particles or a thin film) to pulverize, degrading the performance of the battery. Cui’s battery gets around this problem with nanotechnology. The lithium is stored in a forest of tiny silicon nanowires, each with a diameter one-thousandth the thickness of a sheet of paper. The nanowires inflate four times their normal size as they soak up lithium. But, unlike other silicon shapes, they do not fracture. Research on silicon in batteries began three decades ago. Chan explained: “The people kind of gave up on it because the capacity wasn’t high enough and the cycle life wasn’t good enough. And it was just because of the shape they were using. It was just too big, and they couldn’t undergo the volume changes.” Then, along came silicon nanowires. “We just kind of put them together,” Chan said. For their experiments, Chan grew the nanowires on a stainless steel substrate, providing an excellent electrical connection. “It was a fantastic moment when Candace told me it was working,” Cui said. Cui said that a patent application has been filed. He is considering formation of a company or an agreement with a battery manufacturer. Manufacturing the nanowire batteries would require “one or two different steps, but the process can certainly be scaled up,” he added. “It’s a well understood process.”
Report Shows Developers Consider Embedded Linux as Dependable as RTOSs
Embedded Market Forecasters has released a new research report entitled, “Embedded Linux Total Cost of Development Analyzed.” Based on interviews with more than 1,300 embedded developers, the new study compares the outcomes of hundreds of design projects that used embedded Linux to those that used commercial RTOSs, and further compares project outcomes using commercial embedded Linux with non-commercial “roll your own” embedded Linux. The findings are intended to provide device manufacturers with empirical information to support decision making on embedded OS selection. According to the new EMF report: • Embedded Linux has achieved design parity with commercial RTOSs for most projects. Embedded Linux design outcomes are consistent with the outcomes of projects using operating systems from commercial RTOS vendors. • Use of a commercial embedded Linux OS is more effective than a non-commercial “in-house” Linux development undertaking. In spite of the fact that in-house Linux development projects typically involve much less complexity than projects that use commercial embedded Linux and RTOSs, 15.9% fewer inhouse embedded Linux projects meet the pre-design expectation levels achieved by developers using commercial Linux and RTOSs. • Embedded Linux can be used in a mission-critical environment that requires MILS (Multiple Independent Levels of Security) or EAL (Evaluation Assurance Level) certification January 2008
AFCEA West 2008 San Diego, CA www.afcea.org
VON.x San Jose, CA www.von.com
So. California Linux Expo Los Angeles, CA www.socallinuxexpo.org
Real-Time & Embedded Computing Conference Dallas, TX www.rtecc.com
02/12/08 Real-Time & Embedded Computing Conference Atlanta, GA www.rtecc.com/atlanta2008
02/19/08 Real-Time & Embedded Computing Conference Huntsville, AL www.rtecc.com/huntsville2008
02/21/08 Real-Time & Embedded Computing Conference Melbourne, FL www.rtecc.com/melbourne2008
03/03-07/08 SD West 2008-01-14 Santa Clara, CA www.sdexpo.com
03/11-12/08 Mountain View Alliance Communications Ecosystem Conference San Francisco, CA www.mvacec.com
03/27/08 Real-Time & Embedded Computing Conference Houston, TX www.rtecc.com
04/08-10/08 ROBOBusiness Conference & Expo Pittsburgh, PA www.robobusiness.com
04/14-18/08 Embedded Systems Conference San Jose, CA www.com-egevents.com
04/29/08 Real-Time & Embedded Computing Conference Chicago, IL www.rtecc.com
04/30 – 05/02/08 Small Fuel Cells for Commercial & Military Applications Atlanta, GA www.knowledgefoundation.com
VoiceCon Orlando 2008 Orlando, FL www.voicecon.com
If your company produces any type of industry event, you can get your event listed by contacting email@example.com. This is a FREE industry-wide listing.
or POSIX conformance, when used in protected memory under a certified RTOS. A stable and application-proven embedded Linux design could be directly incorporated within a mission-critical application that requires MILS or EAL certification. Embedded Market Forecasters’ “Embedded Linux Total Cost of Development Analyzed” report can be downloaded at www.embeddedforecast.com.
PICMG and CP-TA Complete Second Interoperability Workshop
The Communications Platforms Trade Association (CPTA) and PICMG have teamed up again in Chicago to host a joint Interoperability Workshop, demonstrating structured testing to industry profiles. PICMG has hosted several Interoperability Workshops in the past, and this is the second time they’ve partnered with CP-TA. Open to members of PICMG and CP-TA, a total of 19 different companies participated in the workshop to test products, designed to PICMG specifications, for interoperability issues. While some AdvancedTCA testing was done, the workshop focused mainly on MicroTCA and AMC. The results of the workshops are confidential to the companies participating so that the testing can take place in an environment focused on learning and resolving issues. “We are encouraged to see the continued maturation of AMC and MicroTCA products at these workshops and see how the interop testing helps companies address interoperability issues early in the design process,” said Joe Pavlat, PICMG President. “With the anticipated release of the PICMG Release 3.0 AdvancedTCA specification, we will offer testing for AdvancedTCA, AMC and MicroTCA at our next event.” “At this event, CP-TA demonstrated for the first time new
thermal and manageability test tools for AMC and MicroTCA,” said Nirlay Kundu, CP-TA Compliance Work Group Chair. “We definitely saw the most interest for MicroTCA testing. Many vendors in the early stage of the development process were able to take advantage of the structured testing. It was exciting to use the tools to find bugs early in the design process and enable companies to address interoperability issues right away.” “CP-TA demonstrated that it is on track on its current roadmap of delivering a MicroTCA interoperability specification in Q1 of 2008,” said Todd Keaffaber, CP-TA Technical Work Group Chair. During the event CP-TA and Polaris Networks demonstrated the new MicroTCA Tester. This test tool, to be released in the near future, incorporates the IPMB analyzer. DegreeC also demonstrated prototype thermal test tools for AMC and MicroTCA. PICMG and CP-TA will hold future events together and provide testing for products based on the PICMG 3.0 Release 3.0.
Cavium Networks Teams with Broadcom to Offer Set of Switching Reference Designs
Cavium Networks has announced a set of reference designs targeted at intelligent switching and broadband applications. These reference designs will utilize Cavium’s market-leading Octeon Multi-core MIPS64 processor family along with Broadcom’s widely adopted enterprise class StrataXGS and home/SMB RoboSwitch Gigabit Ethernet switches, and Fastpath networking software. The reference designs aim to reduce engineering development cost and time-tomarket for intelligent enterprise switches, Layer 4+ switches and SOHO/SMB routers. The scalable set of reference designs using Cavium and Broadcom silicon addresses
the need for pre-integrated, optimized and validated reference designs addressing enterprise and SMB/home markets. Enterprise Class Reference Designs To enable tightly integrated design with seamless interoperability, Broadcom has licensed its HiGig protocol to Cavium Networks. The HiGig protocol enhances Ethernet switching beyond industry standards with features such as stacking and QoS. Cavium Networks recently introduced Octeon processors and follow-on processors with 10 Gigabit interfaces integrate hardware acceleration that supports HiGig protocol and enable intelligent switching applications. Additionally, Broadcom will utilize Cavium Octeon processors in its enterprise reference designs as a control plane and services processor, supporting the complete portfolio of StrataXGS Ethernet switches, which are industry-leading, widely deployed merchant silicon switching solutions. SMB/Home Reference Designs Broadcom is also utilizing Cavium Octeon processors in its SMB switching reference designs as a gateway host processor combined with Broadcom’s RoboSwitch Gigabit Ethernet Switches and Fastpath networking software.
Curtiss-Wright and Sigma Partner to Integrate Physical Layer Switches and Test Automation
Curtiss-Wright Controls Embedded Computing and Sigma Technologies & Resources have announced that they have partnered to offer integrated physical layer switch and test automation software solutions for enterprise network lab automation applications. Sigma announced that it has developed software libraries to integrate its SigmationTF test automation software with Curtiss-Wright’s GLX family of switches, includ-
ing the GLX4000 series of nonblocking, multi-protocol physical layer switches. Sigma’s SigmationTF software is a comprehensive test automation framework designed for testing TCP/IP-based network products. It provides automated regression test methodologies to free test engineers from doing costly, time-consuming and error-prone manual testing. SigmationTF automates the large number of repetitive tests typically required during the test phase of the product life cycle, and later, the regression tests needed as software versions evolve. The GLX4000 can support up to 288 ports and up to 2.5 Tb/s in switching capacity. The company offers SFP- and XFP-based interface cards for flexibility and support of a wide range of serial digital protocols including 1/2/4 Fibre Channel, 10G Ethernet, 10/100/1000 Ethernet, SONET and Firewire.
New SCOPE Profiles Cover AdvancedMC and MicroTCA
The SCOPE Alliance recently announced the availability of the following two new documents: • AdvancedMC Profile, version 1.0 • MicroTCA Profile, version 1.0. These profiles describe the preferred implementation options for their respective technologies. They represent the consensus view of the SCOPE Alliance, (which consists of seven major network equipment providers and many suppliers of standard platform elements). SCOPE hopes that by providing implementation guidance in the form of these profiles, it can reduce the variability that currently exists in AdvancedMC and MicroTCA elements, improve interoperability and facilitate high-volume ecosystems.
Each profile has an overview, a set of definitions and a profile table describing SCOPE’s preferred implementation requirements. There are also some gaps identified. A gap is a technical issue that in SCOPE’s view could benefit from some additional refinement in the AdvancedMC or MicroTCA specifications. SCOPE believes that these profiles could serve as guidelines for many different users of the PICMG specifications. Suppliers of elements (like a chassis, modules, or software) may find the profiles to be a good source of consensus marketing requirements. Network equipment providers may use these profiles as guidelines as they refine their platform architectures and prepare RFXs for suppliers. Network operators may find the profiles valuable as they plan the capabilities of their network elements. The documents can be found at: www.picmgeu.org/pages/t_interesting_articles.htm.
SDR Forum Launches Task Group on Transceiver Subsystem Interfaces
The Software Defined Radio Forum has announced the launch of a worldwide task group on transceiver subsystem interfaces (TSI). Created under the umbrella of the SDR Forum Technical Committee, the new task group will be dedicated to facilitating industry convergence on open specifications supporting the interface between the radio frequency (RF) front end and baseband processing/modem subsystems in reconfigurable radio products. The TSI task group’s initial objective will be to expand upon the work done by the European End-to-End Reconfigurability program and other related efforts to define standard transceiver application programming interfaces (API) for use by radio equipment manufacturers in the commercial, public safety and
defense domains. These APIs will act as an abstraction layer for the transceiver hardware, facilitating the insertion of new or updated physical-layer waveform applications and air interface standards while the radio is in operation and, at the same time, improving the portability of these applications and standards from radio to radio. As a part of this effort, the TSI task group will—to ensure support—evaluate the proposed transceiver API against the various lower-level transceiver interface specifications and standards that have been developed for the commercial and defense wireless markets over the past several years. This evaluation will include the Reference Point 3 specification developed under the OBSAI (Open Base Station Architecture Initiative), the DigRF Interface specification under development by the MIPI (Mobile Industry Processor Interface) Alliance, the VITA Radio Transport draft standard developed by the VITA Standards Organization as VITA 49, and the CPRI (Common Public Radio Interface). One important outcome of this activity will be the revised submission of the transceiver API in response to the OMG’s (Object Management Group) Digital IF request-forproposal for additional follow-on standardization.
COM Express Solutions
COM Express – A Living Standard for Modular Solutions Finally, an industrial standard that doesn’t obsolete itself when silicon changes and is still solid enough to meet longevity requirements of critical embedded applications. by C hristine Van De Graaf Kontron
here are few embedded computing technology people these days who have not heard the words COM Express mentioned at least once when discussing what platform to develop around for the next-generation solution. In the unlikely event that this is the first time you are hearing about COM Express, here is the concept in a nutshell: The COM Express (COM.0) specification is controlled by the PCI Industrial Computer Manufacturers’ Group (PICMG). The standard defines a computer-on-module based on serial differential signaling technology that incorporates interfaces including PCI Express, Serial ATA, USB 2.0, LVDS and Serial DVO. Paired with advanced processors and chipsets, COM Express modules provide the highest performance and I/O bandwidth available in computer-on-modules (COMs). Some key features of COM Express to keep in mind as we discuss it are as follows: • PCI Express – the elemental data path • Gigabit Ethernet – for high connectivity • USB 2.0 – for fast peripherals • Serial ATA – for fast drives • ACPI – for optimized power management
CPU NorthBridge SouthBridge
Video Audio LAN USB
Analog Power Full Custom Headers Board
Video Audio LAN USB
NorthBridge South- Embedded Bridge Module Analog Power
Custom Motherboard vs. Computer-on-Module.
Building on the successes of ETX (Embedded Technology eXtended) computer-on-modules that hit the embedded scene over five years ago, the intention of the COM Express standard is to give embedded designers a highly integrated standard computing core to insert at the heart of their systems, paired with a customized baseboard where expansion and application-specific features are designed in. This is the optimal choice, where as before computer-on-modules, embedded system designers more often than not found themselves engaged in the development of full custom solutions in order to meet their unique requirements. This allows designers to focus more on their core competencies and the custom solution design rather than on the features in the pro-
cessing engine. The small, rugged designs of computer-on-modules allow them to fit where other solutions don’t, mechanically, economically and functionally. Additionally, COM-based solutions allow designers to respond more quickly to demand fluctuations, competitive forces and new technologies (Figure 1). But, why is COM Express needed when other computer-on-modules such as ETX, XTX, SOM, etc. already are available? The initiators of the COM Express standard (Kontron, Intel and others) recognized that the market wanted to take advantage of newer technologies such as PCI Express and Serial ATA that were coming onto the scene. COM Express modules allow for a smooth transition from PCI, ISA and IDE (legacy technology) to PCI Express, SATA and more. Support for Gigabit Ethernet and USB 2.0 (advanced communications technologies) allows the use of data more immediately. An example of this would be where medical personnel at a hospital can prepare to care for a patient while that patient is being transported to the hospital, because the emergency medical team is able to quickly transmit 3-D scans and vital stats. Additionally, the support of PCI Express graphics rather than older
AGP graphics greatly improves the images that doctors have to review when evaluating a patient’s condition. More advanced treatment while en route also may become possible. There is no bottleneck in the transmission of data to the user nor in the development of the product in the first place, as much of the hard work of the hardware development, design and testing is done by the module manufacturer leaving the medical application developer to focus his or her attention on the software and special features of the tool. Today the offering of COM Expresscompatible modules encompasses various sizes and pin-outs. There are five defined by PICMG (Table 1) as well as processor and chipset variations. The most recent addition, a nano-sized (55 mm x 84 mm) COM Express module as proposed by Kontron, is intended to bring the benefits of the COM Express standard to advanced and highly mobile applications whose lowpower and size requirements are not met by the other small COMs such as X-board, XTX, etc. Although these open standards offer a good range of features and some with low-power variations, they have been limited with respect to interoperability with existing systems and scalability. Type 1 Gigabit Ethernet Serial ATA AC97 USB 2.0 PCI Express LVDS 5v Standby VGA Power LPC PCI IDE Note:
Dimensions PCI Express Lanes Serial ATA Memory
95mm x 95mm
55mm x 84mm
95mm x 125mm
155mm x 110mm
0 to 2 lanes
2 or more lanes
2 or more lanes
2 or more lanes
1x Socket DDR2 SO-DIMM
Up to 2x Socket Up to 2x Socket DDR2 SO-DIMM or Desktop DIMM Mini-DIMM
USB 2.0 Ethernet PICMG COM Express Standard
COM Express comparison chart.
The proposed nanoETXexpress specification is targeted to deliver extremely power-saving computer-on-modules with mid- to high-performance x86 technology on a footprint that is a mere 55 mm x 84 mm. This is 39 percent of the original COM Express module Basic form-factor 125 x 95 mm footprint and 51 percent of microETXexpress (95 mm x 95 mm). This new COM form-factor follows the PICMG COM Express standard and will be 100 percent compliant with the COM.0 Type 1 connector. The locations of the identically mapped pin-outs will also be 100 percent COM.0 compliant. As such, apType 2
plications such as stationary point-of-sale (POS) systems built around one of the first COM Express modules could be updated to use a new nanoETXexpress module to achieve greater power efficiency. With a few tweaks, the unit possibly could become mobile, opening up a greater customer base for the POS application. Key features planned for inclusion in the nanoETXexpress specification include multiple PCI Express lanes and USB 2.0 ports, Serial ATA support, advanced graphics capabilities such as dual 24-bit LVDS channels and Gigabit Ethernet. It will represent a good option for mobile, Type 4
Yes • Single connector • Lowest cost pin-out • First nanoETXexpress modules will follow this pin-out
• Includes IDE, PCI and up to 22 PCI Express • The “Base-Line” pin-out • Most of today’s COM Express solutions follow this pin-out
Yes • Includes PCI support and up to 3 Gigabit Ethernet ports • No IDE supported • Example of solution following this pin-out is the Kontron ETXexpressWPM module
• Includes IDE support and up to 32 PCI Express lanes • No PCI support
• Support for up to 32 PCI Express lanes and 3 Gigabit Ethernet ports • No IDE support and no PCI support • The “Legacy Free” pinout
COM Express Five Defined Pin-out Types. January 2008
rugged applications because the specification includes onboard DRAM and flash instead of SO-DIMM sockets for ruggedness, an extreme low-power Ethernet controller for overall lowest power consumption and a wide range power supply input of 4.75 to 14 VDC. The goal of the nanoETXexpress specification is to build PCI Expressbased computer-on-modules on the smallest possible form-factor. Many new ap-
plications that will benefit from the nano size include handheld units for medical, mobile data solutions and other emerging applications that have not been possible as of yet due to size restrictions. New silicon that will be coming from the industry leaders in 2008 and 2009 is already set to support this trend (Figure 2). Additionally, work has begun at PICMG to update the COM Express standard. A subcommittee has been organized
to review the standard as a whole and jointly develop a design guide for COM Express baseboards that will ensure the continued interchangeability of modules independent of which vendor has supplied the COM Express module. The subcommittee will address issues outlined in both the PnP Design Guide from Congatec and the COM Express Extension Design Guide from the COM Industrial Group. Other key items that will be discussed by the subcommittee include: • Multiplexed graphics pin-out to support PEG, SDVO, TMDS and Display Port in order to fit the needs of all the chipsets that will appear in the next few years • Extended power supply range from 8.5V DC to 18V DC instead of 12V DC only • Onboard thermal control and BIOS implementations with smart battery support • Two dedicated GPIO pins for energy management functions • New obligatory support of legacy interfaces via super I/O chips By bringing out new interface technology, COM Express modules have met the design challenges of application segments that previously required full-custom single board computer solutions for a diverse array of embedded applications. The addition of nanoETXexpress as a compatible embedded computing technology solution following the elements of the COM Express standard will make it possible for designers to take many formerly stationary applications and make them not only more power efficient, but make them much more mobile as well. Existing COM solutions and those that will become available as the COM Express specification continues to adapt to new technologies and market demands, will fulfill the needs of existing advanced embedded applications and those that are yet to be imagined. Kontron Poway, CA. (858) 667-0877. [www.kontron.com].
One-Stop Shopping Bring Us Your List Everything you need for your CompactPCI® Applications Performance Technologies’ CompactPCI® products and solutions are the industry’s most capable and complete open-standards offerings for telecom, aerospace and defense, and commercial applications. We create innovative and robust solutions that enable you to get your products to market faster. Our integrated solutions include: Advanced Managed Platforms™ : These flexible, scalable platforms are ideal for a wide range of compute-intensive applications. Coupled with exceptional support services, we have what you need to quickly and cost-effectively develop and deploy your system. High-Performance Blades: Single and dual-core SBCs, Ethernet switches, intelligent shelf management, voice processors, serial and T1/E1 communications I/O, and RAID storage blades.
Platforms. Blades. Software. Support. Many solutions. One vendor.
NexusWare ® : Our CGL 3.2-registered and POSIX-compliant Linux distribution includes a full operating system, development tools, protocols, HA middleware, and SA Forum-compliant APIs. Best of all, NexusWare's total integration with our blade and platform solutions means you have shorter development cycles. Contact Performance Technologies today and bring us your product development shopping list.
The Best Network Seems Like No Network Strategic network optimization with 10 Gigabit Ethernet technology makes connections invisible. Network-wide or strategically placed, 10GbE is an important component of the overall Ethernet performance networking strategy, and there are multiple ways to deploy 10GbE to get the best “bang for the buck” in any network.
by Iain Kenney SMC Networks
ith so many bandwidth-intensive applications in use that reach across networks—both local and wide-area—the only downside to upgrades has been cost. Speed and bandwidth availability vary in most networks from 3 Mbits/s or less across standard Wi-Fi (802.11b) connections, to 10 Gigabits in high-performance corporate LANs. With bandwidth, though, it doesn’t need to be all-or-nothing. While some networks warrant 10 Gigabit Ethernet (10GbE) all the way, for many others a close analysis of network usage can uncover high-return locales for high-speed technology—places where pockets of 10GbE technology can affect overall network performance so that all users get the most from their connections, and the network budget doesn’t get depleted. The broad range of high-speed, high-bandwidth solutions available today make it possible for more companies to leverage standardscompliant 10GbE in their networks. It doesn’t have to be cost-prohibitive. As the 10GbE networking market continues to evolve and redistribute, a breadth of products to address today’s demands ensures that deployment of 10GbE is a practical solution, when and where it makes a difference—whether for the
PoE 1G Fiber 10G Fiber 1G Copper Wireless SMC6152L2
SMC2555-AG Building C
The most common application for expensive—to purchase and to deploy—10GbE connectivity has been building-to-building connection. New 10GBase-T technology makes it reasonable to put more 10GbE into more places for even freer bandwidth.
whole network, or in high-demand portions of it (Figure 1).
Relieving Bottlenecks at the Server
In the server room or data center, the benefits of 10 Gigabit connections are clear: interconnecting busy processing,
application and storage servers with as high a bandwidth connection as is available has a huge impact on efficiency. Relieving bottlenecks that occur at 10 or 100 Mbit/s server connections by replacing the overloaded segments with 10GbE uplinks can make the network as a whole move data faster. 10/100/1000
Which Way do You Want Your 10Gb Ethernet?
2500MB/sec 10G b 250MB/se
Software Stack Conventional NIC Technology
Silicon Stack Critical I/O XGE
Silicon Stack Technology from Critical I/O. 10Gb Ethernet at Wire Speed. [Problem] You’re expecting 10Gb Ethernet to deliver a whole lot more performance to your embedded system. But what if you invest in it and get no gain at all? The performance of nearly all existing 1Gb applications are limited by the software overhead associated with the TCP/IP protocol stack. This bottleneck is in the software stack, not the network hardware. So, simply upgrading to 10Gb pipes will not improve your system’s performance. [Solution] Unlike conventional Ethernet interfaces or processor-based “offload” products, Critical I/O’s Silicon Stack technology eliminates this inherent bottleneck by offloading protocol processing to silicon; thereby achieving sustained line-rate performance, microsecond latency, and rock-solid deterministic behavior. And, Silicon Stack is 100% compliant with Ethernet standards, allowing you to leverage existing applications and hardware.
XGE Silicon Stack Ethernet vs. Software-based Stack
Software Stack 10Gb
40 varies with protocol
1Gb Throughput max sustained rate in MBytes/sec Host Overhead
Determinism max sustained rate Reliability
Horrible ± 200 μsec Poor when under heavy load
Very Low 12 μsec
Rock Solid ± 1 μsec Excellent under all load conditions, no dropped data
SOLUTIONS Engineering network switches with optional slots for 10GbE fiber uplinks are able to accommodate the growing need for selective high-bandwidth connectivity as a means for maintaining overall network performance. Designed for network interconnection and specialty applications that benefit from maximal pipeline, these solutions bring fiber-optic 10GbE—and unmatched performance—via a variety of connector types to networks for about $2,000 to $6,000 per port, depending on
the type—SR, LR or ER—of connector that’s used (Figure 2). The level of deployment of optional 10GbE uplinks has grown immensely, confirming that the demand for 10GbE is real and immediate, so stand-alone 10 Gigabit XFP fiber-managed Ethernet switches have found their place in the network as well. And, while 10GBase-T fills an important role for many data center and NAS applications, 10GbE fiber is still unquestionably in demand, and that
demand will continue to evolve and grow. Using, for example, an 8-port 10GbE Layer 2 managed switch—a core aggregator switch designed to handle heavy traffic workgroups or power users and is capable of reaching up to 160 Gbits/s with a non-blocking switching architecture— can provide that speed and bandwidth over distances of up to 40 kilometers. The industry will continue to develop Ethernet networking products that incorporate density, efficiency and bandwidth appropriate to varying customer applications, including additional copper and fiber-optic 10GbE, Gigabit and 10/100 Fast Ethernet products. User demand for 10 Gigabit speed and bandwidth continues to grow, even ahead of budgets, so 10GBase-T solutions will present a lower-cost way to introduce 10GbE where it’s needed most. Switches that deliver 10GbE connectivity over
SMC884M SMC884M Network
Core Legend 10G Copper 10G Fiber
1 18Untitled-3 January 2008
1/8/08 4:10:35 PM
From the corporate data center, long distance links to other building and subnets are made via 10GbE fiber uplinks from the core aggregator switches. Inside the server room, 10GbE Copper links to servers reduce overall network construction costs without compromising bandwidth. The Gigabit switch’s 1G connections to internal workstations benefit from the speed gains of bottleneck-freeing 10GbE cross-network and server connections via the uplinks.
The Harsher the Environment, the More You Need MEN Micro!
Marketing Legend 1G Copper 10G Copper 10G Fiber
D7 Xeon® Blade Server 6U CompactPCI®
Copper 10GbE switches with fiber 10GbE uplinks make more highbandwidth connections without the expense of fiber, while fiber uplinks take care of longer distance runs.
standard copper cabling enable short-run 10GbE connections at roughly half the per-port price of fiber switching—even before considering the savings in cabling. 10GBase-T enables implementation of flexible data centers and short distance networks at a reasonable price—and the ease of installation of Cat6A cabling adds even more to the value side of the equation (Figure 3). Of course, conformance with IEEE802.3an is a critical feature to look for in a 10GBase-T switch—as it provides an IT professional with the assurance that the 10GBase-T switch is fully standards-compliant and interoperable, and will install and operate in any standardsbased Ethernet network. Such a multi-port combination 10GBase-T and 10GbE fiber XFP network switch recently became broadly available: the SMC8724-10BT TigerSwitch 10GbE Stand-alone 24-port 10GbE switch with 20 10GBase-T ports and four XFP-based 10GbE fiber ports for flexible 10 Gigabit connectivity.
10GbE All the Way
While 10GbE fiber uplinks and aggregator switches—and even 10GBase-T—fill important roles for many data center and NAS applications, sometimes relief at the switch is not enough. Switching is central to speed and bandwidth management, but opening the bandwidth further down the line is a priority for many as well. 10GbE network adapters—both fiber and copper adapters for 10GbE connections to servers—that stress performance, thermal management and affordability make the connection down the line, and are rolling out now, with the first of them just re-
cently made available. Several key architectural features that should be included in a good 10GbE Ethernet Adapter include: cut-through architecture for low latency; virtualization of the host interface; and VNIC, for the cleanest, most efficient acceleration of I/O in virtualized operating system deployments, to deliver real performance benefits over a broad range of applications. Thermal management and overall performance must be considered as well. There are now 10GbE Ethernet Adapters for each high-bandwidth application market segment: Network-Attached Storage (NAS) Servers, Storage Area Network (SAN) Arrays, High Performance Cluster Computers, Blade Servers, Video Servers, Application Servers and Web Accelerators. Such adapters are currently becoming available on the market at price ranges starting under $1,000. 10GbE is the answer to today’s pervasive demand for more bandwidth, and deploying it in the most efficient way for a given network can optimize the cost-benefit ratio. High performance is becoming accessible to more companies as IT managers ensure that it’s deployed when and where it can have the greatest impact. Today’s solutions break down the challenges into manageable pieces so that IT managers can leverage the best and fastest technology in an optimal configuration—at a price that enables them to justify the upgrades in more cost-benefit analyses. SMC Networks Irvine, CA. (800) 762-4968. [www.smc.com].
For mission-critical applications, MEN Micro offers the D7 6U CompactPCI Blade Server: ■
■ ■ ■
■ ■ ■ ■
2 dual core Xeon® 1.66 GHz ULV processors (4 cores total) cPCI system slot with PCI 64-bit/ 66MHz or PCI-X 64-bit/133MHz 4 GB DDR2 SDRAM with ECC Non-volatile SRAM and FRAM SATA and PATA support for hard disk, CompactFlash and more I/Os include UARTs, USBs, Ethernet 2 XMC or PMC slots Up to 2 GB Ethernet channels Long-term availability: 5 years+
ICA Technology Showcase
MEN Micro, Inc. 24 North Main Street Ambler, PA 19002 Tel: 215.542.9575 E-mail: firstname.lastname@example.org
www.menmicro.com January 2008
Doing More with Less (Hardware, That Is) By using a system of modular hardware, companies are able to combine the benefits of homegrown flexibility with COTS reliability and support, thus harnessing the full power of flexible, modular I/O. They can select among multiple-channel modules designed for the signal types of interest and configure flexible instrumentation solutions using software.
by B rett Burger National Instruments
exploration er your goal eak directly al page, the resource. chnology, and products
ake a quick look around; there are one-hit wonders designed to measure a likely no fewer than half a dozen single, real-world phenomenon. A temelectronic devices that all have one perature data recorder did just that, and if thing in common. Three years ago they more measurements needed to be taken, were probably twice the size or had half more equipment, space and money was the capabilities, and in many cases both. required. One of the bigger evolutions Ideally, people want a personal assistant came with the mass adoption of computer to hold everything including a camera, technology. Suddenly, the data acquisition phone, wireless laptop, music player, system was separated into computational clock, address book and more. In lieu of components and measurement hardware that personal assistant, the consumer elec- components. This movement to PC-based panies providing solutions now tronics market is increasing functional- or virtual instrumentation brought great ration into products, technologies and companies. Whether your goal is to research the latest ity by integrating all of these devices of flexibility to the end user, because the lication Engineer, or jump to a company's technical page, the goal of Get Connected is to put you convenience one package, hence the processing and analysis capabilities were ice you require for whatever type into of technology, cell nearly endless with the power of the stanies and productsmultitasking you are searching for. phone. In pursuit of ultimate flexibility, the dard PC and programmable application instrumentation market is not unlike the software. For data acquisition hardware, consumer electronics market. Companies multifunction devices became available and consultants constantly strive to get for several PC buses including ISA, PCI, the most out of a limited amount of, usu- PCMCIA, PCIe and USB as well as a host ally expensive, measurement equipment. of stand-alone and proprietary protocols. This drive has led to several evolutions in Data acquisition devices also began to data acquisition. measure several types of signals, includEarlier data acquisition systems, or ing analog and digital, but even with even the more simple data loggers, were multiple signal measurement types, these devices were still fully assembled by vendors, and customers were often left with Get Connected extra channels, unneeded measurement with companies mentioned in this article. www.rtcmagazine.com/getconnected types, or a limited expansion path when
End of Article
January 2008 Get Connected with companies mentioned in this article. www.rtcmagazine.com/getconnected
the project grew. And with that, another evolution of the data acquisition system, modular systems, was born. Modular systems are a great approach for system construction because they allow for a reasonable amount of growth before having to reinvest in additional foundation components. Modular systems and platforms are designed and built by a variety of vendors on several standards, some of which are open such as PXI and VXI, while others are proprietary. The goal of modular I/O is to enable the end user to purchase only the channel count and signals needed and offer the ability to upgrade in the future without having to double the initial investment. Many of the benefits of a modular system are shared by the vendor and the customer. A customer can update, expand, or change a system with the purchase of new modules. A vendor can more rapidly enhance a data acquisition platform to meet the demand of the customer by adding new modules with new measurements or channel counts to their product lines. A trend has developed stemming from a single instrument that did everything to having individual components broken out
for the user to define and purchase separately at the discretion of the project needs. This trend has brought together the benefits of commercial off-the-shelf (COTS) quality, reliability and support with those of homegrown customization. As technology advances and the components used in instrumentation have gotten smaller, it has become easier to design smaller components for data acquisition systems. Modules that started as trays fitting into large card cages have evolved into smaller modules slightly larger than a deck of playing cards. In the ever-shrinking trend of hardware and modules, size has not sacrificed channel count, as some of these smaller modules offer more than 30 channels and may be placed in different chassis types for different configurations for various deployment options (Figure 1). One-channel modules may or may not be an economical solution, but either way, the disaggregated module movement is ending. This begs the question â€œwhat is next?â€? The answer can be found by looking at some higher-level benefits of a modular platform. Modular systems have reduced the number of systems needed and components in a test departmentâ€™s equipment library by using exchangeable modules in a single chassis or carrier. One of the remaining problems is that channel count and signal type are only two characteristics that can change. Selfcontained measurement modules solve the problem of changing channel count and signal type, but the market demands more flexibility. The next evolution of data acquisition systems is to exchange not only modules, but chassis for deployment as well. The test process, depending on the device, has many stages including design validation, controlled environment testing, subsystem component tests, hardware simulation tests, prototype tests, end-of-line manufacturing test and more. By creating a data acquisition system comprised of interchangeable modules and deployment options, test designers can further reduce their hardware needs and thus expenditures, storage space, vendor contacts and toolchain training for employees. To fully harness the power of modular I/O, it is important to maximize reusability. Just as there are different modules
The C Series family of hardware is composed of over 40 modules and multiple deployment options. Shown here: desktop chassis, industrial computer (CompactRIO) and USB single-module carrier.
for different measurements, the next evolution of modular data acquisition systems will have different deployment options for the same set of modules. Deployments will vary between industries, but will include options such as size, portability, PC-connectivity, ability to run in a standalone mode, ruggedness, reliability and so on. And, as with modules, no vendor will cover every possible use-case, but a family of modular, flexible hardware should be able cover a wide array of the aforementioned deployments. To better illustrate this, consider a scenario where a fictional company deploys a single-vendor flexible deployment hardware platform. Assume this company is a consultantfocused engineering house specializing in rotating equipment monitoring and maintenance in the machine condition monitoring (MCM) industry. This company uses accelerometers for vibration measurements. Accelerometer measurements are on the higher end of sensor measurements due to the high sample rate, resolution and bandwidth required. In addition, such integrated electronic piezo-electric (IEPE) sensors require current excitation to power the sensor. Anti-alias filters are a benefit to remove any traces of highfrequency noise in the system, and due to the speed of acquisition and nature of phenomena, simultaneous sampling
ADCs are preferred to ensure signals are in phase. Using a modular platform, all of this signal conditioning and data acquisition circuitry must fit into one module. Measurements performed by an MCM specialist are often from proximity sensors for shaft alignment, tachometers for shaft rotational speed, power load by the motors, and accelerometers for vibration analysis on bearing housings. These measurements are performed in different deployments. One application is the small, low-channel-count, portable unit that a consultant travels with for spot-checks on systems. These spot-checks can be done routinely or when a noise is heard by an operator and called in as a problem. Portable deployments require portable display, storage, reporting and easy setup. More complicated deployments involve a larger channel count for complete machine testing and mixed sensor types, since a complete machine check-up will have tachometers, proximity probes and accelerometers. This larger-scale monitoring system may have bigger storage and processing requirements due to the vast amounts of data involved. The system needs to be moveable, but not as portable as the spot-check system. With the temporary install location often in an industrial environment, the equipment must also be rugged. The final level of machine maintenance that this company offers is January 2008
the permanently mounted monitoring system. These are online systems installed on rotating equipment to constantly monitor the health of the system and, when the limits are exceeded, notify managers of necessary repairs, or sound alarms or initiate emergency shutdown procedures in hazardous situations. The company in this scenario is comprised of engineers and service technicians with significant industry experience in machine condition monitoring, but limited understanding of hardware design. In today’s market, the company must purchase several devices: a one-box solution originally designed a decade ago and is slightly large for full-system monitoring, a newer PC-based solution for the portable system that runs off of a laptop (Figure 2), and a more expensive, ruggedized brick-style system for the permanently mounted system. With these choices, the company must buy, learn and maintain a lot of different equipment, or try to find one vertical vendor that makes all three types of systems. But even with one vendor there may be a
The single module carrier connects to a PC over USB for a low-channel-count, portable solution that can connect to a laptop for field applications. Individual modules can be easily swapped out of the carrier.
concession in quality or performance for a one-vendor solution. The other, preferable choice is to design and build their own systems. This ensures that the specs needed are met exactly and all functions are available. However, this company has more technical knowledge in collecting, reading and interpreting the results than hardware design.
The solution is to use a family of hardware with modular I/O and flexible deployment such as the C Series family of hardware from National Instruments. This collection of hardware offers more than 40 measurement modules that are a little larger than a deck of playing cards. These modules can be used in any one of several different chassis for multiple deployment options. In our example, an accelerometer module could be used in the singlemodule carrier with a rugged laptop for a portable vibration measurement system. For the full system monitoring, the same module could be inserted into an 8-slot chassis along with other modules for proximity probes and tachometers. For larger machines, multiple chassis can be linked together and synchronized for in-phase measurements. Finally, the same set of modules can be inserted into an industrial computer chassis for a permanently mounted monitoring and alarming system. The CompactRIO chassis, for example, offers either four or eight slots and has a builtin storage and processing capability in an extremely rugged housing, and operates in environments ranging from -40° to 70°C and up to 50g shock. Hardware is just half of the equation; software must still be considered. Functions such as FFT, order analysis, orbital analysis and waterfall plots are needed for a complete MCM system. This analysis and experience is where the consulting company’s expertise lies, and using software from the same vendor, NI LabVIEW, a company can write and reuse analysis code on all three application deployment platforms. With this one-vendor flexible solution, this consulting house can better manage the training and hardware needed for complete customer service. Should the company decide to resell the systems, the COTS benefits kick in again, because of the manufacturing resources of the selected big name vendor. National Instruments Austin, TX. (512) 683-9300. [www.ni.com].
1 22Untitled-6 January 2008
1/8/08 4:17:39 PM
with standards. with Wind River.
Mike Deliman Here’s a guy who appreciates the view from above. When he’s not trekking the high plains and mountain passes of Tibet, Mike Deliman is working on aerospace and defense projects for Wind River. He’s fond of Mars rovers, solar panels, and astronauts; and his real-life heroes are Albert Einstein and the Dalai Lama. He’s aiming high.
Regional Developer Conference, Aerospace and Defense Edition Register now to join Wind River and our partners at one of three , full-day conferences dedicated to Wind River’s aerospace and defense solutions including: Wind River's new VxWorks 653 Platform 2.2, Wind River Real-Time Core for Linux (with FSMLabs technology), Wind River's VxWorks AMP/SMP multicore support, ARINC 653 and Integrated Modular Avionics (IMA) , and DO-178B safety certiﬁcation http://www.windriver.com/announces/ad_conference/index.html
March 12, 2008
March 27, 2008
April 2, 2008
Hilton Santa Clara 4949 Great America Parkway Santa Clara, CA 95054 Tel.: 408-330-0001
Turf Valley Resort 2700 Turf Valley Road Ellicott City, MD 21042
Four Seasons Resort & Club Dallas at Las Colinas 4150 North MacArthur Boulevard Irving, TX 75038 Tel.: 972-717-0700
® 2008 Wind River Systems, Inc. The Wind RIver logo is a trademark, and Wind River is a registered trademark of Wind River Systems, Inc. Others marks are the property of their respective owners.
Data Acquisition: Harnessing Small Modules with Flexible I/O Techniques used for effective implementation of remotely deployed small form-factor data acquisition modules are increasingly looking to wireless sensor networks due to their dynamic reconfigurability, robustness and ease of deployment of small form-factor sensor nodes.
by S cott Hames GE Fanuc
mall form-factor data acquisition devices can be thought of as the sharp end of the stick in a system of remotely deployed sensors. Often the reason they are small form-factors (SFF) is because the details of the deployment (size, weight, cost, etc.) prevent the use of COTS form-factors such as PMC, AMC, etc. Regardless, remotely deployed sensors of any form-factor are in fact data producers for a centralized data consumer process. The fact that they are not connected to their hosts by a data bus requires them to use unconventional I/O methods. There are various objectives of remote sensor deployment. Among them are the need to make measurements at remote locations, to simultaneously measure a variable in multiple widespread locations, and the need to measure a variable where the sensor will be lost or destroyed in a relatively short time. Additionally, remote deployments often carry with them restrictive space, weight and power requirements. Air-launched sonobuoys are possibly the ultimate example of remote deployment. Because they operate on batteries, power consumption must be frugal.
Airborne equipment performs signal processing, target tracking, recording, etc.
Deployed sonobuoys transmit sonar returns. 99 channels 136 - 174MHz
Legacy wireless sensor architecture.
They are deployed once and are likely to never be recovered, so they must be cheap. Additionally, sonobuoys are usually deployed in significant numbers (99 channels are theoretically available) from patrol aircraft or helicopters, so size and weight must be kept low. With the special requirements of remote sensing in mind, SFF DAQ modules are deployed in one of three ways: fixed sensors with a hardwired connection, fixed sensors with a wireless connection and mobile sensors with a wireless connection.
Fixed sensor arrays are often encountered in industrial process control applications, where a number of sensors transmit data back to a central control system. Early systems used analog parameters, such as electric current in a control loop, to represent physical conditions such as flow rates, pressures, or temperatures. Modern systems represent the process variables as digital data and communicate with the controller using commodity communication links such as USB. The rapid acceptance of USB 2.0 in industrial control provides an example of fixed sensors and hardwired connections. Over the last few years, USB has become a ubiquitous mechanism for connecting computers and various I/O devices. It is readily available on virtually all styles of PCs, and also on most single board computers and stand-alone instrumentation, regardless of the host OS. USB also offers the user a combination of low price, good performance, reliability and â€œplug and playâ€? installation. Because differential signaling is used, USB has strong resistance to noise typically encountered in factory environments.
The easy installation capabilities of USB also allow industrial sensor systems to be quickly modified. Hot swap/hot plug capability means that in many cases USB modules can be unplugged and reinstalled in new locations without interrupting system operations. USB 2.0 offers this easy connectivity for up to 127 devices, enough to monitor several complex industrial processes from a single PC. And with commercially available USB hubs and extenders, the reach of USB connections can be extended to well over 100 meters. One of the biggest benefits of USB for remote deployments is the fact that the same cable that provides communication also provides the sensor power. Although the total power available from a USB host is low (mandated at 0.5A @ 5V), many commercially available hubs provide additional power to USB devices.
signal processing and software defined radio technology now allow use of modern waveforms. Typical legacy airborne antisubmarine warfare (ASW) systems support 8 to 32 concurrent sonobuoy receiver channels selected from any of 99 channels in the VHF band 136 to 174 MHz.
Mobile Ad-hoc Networks (MANETs) represent one of the most challenging applications of small form-factor data acquisition modules. However, creative implementation of MANETs is yielding some very efficient ways to collect data from
Fixed wireless connections are used in applications where physically connecting the sensor nodes is inconvenient, undesirable, or impossible. In factory applications similar to the fixed and hardwired example, a wireless architecture will enable rapid deployment of a new sensor network with little or no additional infrastructure. An excellent example of a legacy “fixed” (to use a term loosely) wireless SFF DAQ system is the sonobuoy. These are typically either active or passive airlaunched sensors that are used in various marine applications from submarine detection to sea state monitoring, to animal tracking. Each sonobuoy contains its own radio transmitter, which relays data back to a receiver located in a patrol aircraft or helicopter. The buoys themselves do little or no processing, merely relaying data back to a host system. Each sonobuoy typically transmits on a preset carrier frequency, and the host is responsible for receiving all the deployed channels, or at least those of interest. Figure 1 shows a typical airborne acoustic processing system. The basic signal acquisition system architecture has changed little since the Second World War. This type of signal distribution was originally based on analog FM radio technology, but advances in digital
Some of the form-factors specified in the JTRS HMS inventory.
remote, sometimes dangerous, environments. Ad-hoc wireless sensor networks consist of large numbers of small, cheap and often single-use devices, each capable of limited computation, wireless communication and sensing. The actual sensor nodes are deployed as close to the signals of interest as possible, and ideally the sensors are actually mobile. Rather than routing signals of interest directly to a centralized location for sampling and processing, information moves among the nodes en route to its destination. Wireless sensor networking technologies are heavily employed in the U.S. DoD JTRS HMS specification, which defines radio equipment characteristics for a wide range of applications, and re-defines the problem as wireless tactical networking.
The JTRS HMS Requirement
The U.S. Department of Defense Joint Tactical Radio System (JTRS) has placed new space, weight and power chal-
lenges on the handheld, man pack and small form fit (HMS) radios. These radios are intended for deployment in a wide variety of systems and platforms such as unattended ground sensors, UAVs, robotic vehicles, and on the soldiers themselves (Figure 2). The objectives of the JTRS HMS include networked communications among various levels of command and the ability to support technology insertion upgrades while maintaining interoperability with existing equipment. With this in mind, wireless tactical networking is one of the most critical capabilities the JTRS program will deliver. Like commercial wireless sensor networks, wireless tactical network nodes (soldiers) often move and the bandwidth of the links can be very limited. MANET protocols mentioned above are designed to handle these wireless environments. The MANET protocols will be used in conjunction with new JTRS networking waveforms, to permit soldiers in the field to connect to the DoD’s Global Information Grid. Software Defined Radio technologies are being employed to meet the performance objectives within the aggressive new space, weight and power limits. WSNs (and WTNs) have their roots in packet radio systems. Military packet radio networks from their very beginning developed hardware, software and protocols that could adapt to the changing topologies and environments that were expected on a battlefield. Growing out of the University of Hawaii’s ALOHANET, the DARPA-sponsored Packet Radio Network (PRNET) project extended the single-hop packet radio into a multi-hop packet radio network. Unlike most amateur packet radio networks, the PRNET project designed and tested protocols in environments where the nodes were expected to be mounted on mobile platforms, such as trucks. As a result, the protocols had to adapt automatically to changes in topology, although routes were expected to remain stable for at least a few minutes. Mobility of the sensor nodes increases the complexity of the networking problem considerably. In contrast to traditional wired networks, routing in a WSN is both highly dynamic and ad hoc. The idea of ad
P2P Over Internet
Sensor Network Relay Broker Broker Node w/ Event Correlation Cluster Head (Aggregator)
A ZigBee network can incorporate a variety of topologies.
hoc networking is sometimes also called “infrastructureless” networking, since the mobile nodes in the network dynamically establish routing among themselves to form their own network on the fly. For example, after the initial deployment, sensors may fail due to damage or battery depletion. Also, mobile radios or sensors are likely to experience changes in their position, available energy and task details. Changes in the environment can dramatically affect radio propagation, causing frequent network topology changes and network partitions. Wireless sensor networks often deploy large numbers of sensors. Although they don’t require high bandwidth, they usually require low latency. And because the sensors themselves are usually battery powered, minimal power consumption is a must. These requirements are similar to those imposed by the JTRS HMS radio specifications, which are heavily based on networked communications. Because technologies such as Bluetooth and ZigBee are aimed at providing short range wireless ad-hoc networking capability to low-cost, battery-powered devices, techniques and algorithms developed for commercial markets could well be adapted for military use. For commercial applications, ZigBee (IEEE 802.15.4) in particular has great potential in the area of wireless sensor networks (Figure 3).
ZigBee Wireless Networks
ZigBee’s general characteristics include use of dual PHY (2.4 GHz and 868/915 MHz) with Data rates of 250
Kbits/s (@ 2.4 GHz), 40 Kbits/s (@ 915 MHz) and 20 Kbits/s (@ 868 MHz). The protocol is optimized for low duty cycle applications (<0.1%). CSMA-CA channel access yields high throughput and low latency for low duty cycle devices like sensors and controls. It also provides for an optional guaranteed time slot for applications requiring low latency as well as for low power usage with battery life ranging from multi-month to years. ZigBee allows for multiple topologies including star, peer-to-peer and mesh, and is a full handshake protocol for transfer reliability with a typical range of 50m (5-500m based on environment). ZigBee uses spread-spectrum technologies to avoid multi-path fading and increase robustness. This approach allows for improved signal immunity in the presence of radio interference. The IEEE 802.15.4 standard defines two PHYs representing three licensefree frequency bands that include sixteen channels at 2.4 GHz, ten channels at 902 to 928 MHz, and one channel at 868 to 870 MHz. The maximum data rates for each band are 250 Kbits/s, 40 Kbits/s and 20 Kbits/s, respectively. The 2.4 GHz band operates worldwide while the sub1 GHz band operates in North America, Europe and Australia/New Zealand. The IEEE standard is intended to conform to established regulations in Europe, Japan, Canada and the United States. GE Fanuc Charlottesville, VA. [www.gefanuc.com].
1/8/08 4:16:02 PM
ip for programmable logic
Enabling the Use of FPGAs with Optimized IP The combination of FPGA-optimized IP with a domain-specific development environment enables the non-traditional FPGA designer to realize the full benefits an FPGA has to offer.
by T om Hill Xilinx
exploration er your goal eak directly al page, the resource. chnology, and products
he benefits of an FPGA are clear: Hardware they can deliver a significant perAccelerated Custom Local formance boost for applications that Functions Memory benefit from parallelism, or dramatic cost savings through consolidation of the funcMicroBlaze PowerPC tions of multiple chips onto a single device. Often, these benefits are perceived to External Memory Interrupt be out of reach to developers whose priController Controller mary design experience is with DSPs or DMA Timer/PWM general-purpose processors. Ethernet MAC PC/SP1 Intellectual property, tightly intepanies providing solutions now grated into domain-specific design flows, UART PCI ration into products, technologies and companies. Whether your goal is to research the latest enables non-traditional FPGA designers to lication Engineer, or jump to a company's technical page, the goal of Get Connected is to putCntlr you CAN/MOST Peripheral quickly create hardware sysice you require for whatever type ofsophisticated technology, using FPGAs. Unlike the processor, ies and productstems you are searching for. Custom IO the FPGA does not come pre-equipped Peripherals with a hardware architecture, so rather than simply programming the device, the Figure 1 FPGA-based Embedded developer must also create the device. This System. implies a basic knowledge of hardware architectures for complex functions and in- System Overview terface standards. If these operations are FPGAs allow users to preserve the not designed to efficiently use the FPGA easy-to-use programming model of a hardware resources, then the performance processor while optimizing hardware and cost benefits are compromised. for a specific application. The fixed architecture of a processor removes a degree of freedom in the development Get Connected process that facilitates rapid developwith companies mentioned in this article. www.rtcmagazine.com/getconnected ment of an application. Performance
End of Article
January 2008 Get Connected with companies mentioned in this article. www.rtcmagazine.com/getconnected
issues for hardware relying on standard processors tend to be addressed through additional devices. Common solutions typically involve using multiple processors for parallelism or an FPGA coprocessor to accelerate performancecritical functions. FPGAs, however, are uniquely positioned to provide a singlechip solution that can be tailored to an application’s specific requirements. Implementing an embedded system on an FPGA lets developers customize the system to include multiple hard or soft core processors, select a unique set of peripherals and include hardware accelerated custom functions implemented on the FPGA fabric. This flexibility allows performance issues to be addressed while minimizing cost and power. Figure 1 provides a block diagram of a typical FPGAbased embedded system. Can such a system be created by a traditional processor developer with limited FPGA design expertise? Will the implementation of this system use the available FPGA resources sufficiently for production hardware? The answer to both these questions is “yes,” through the use of FPGA-optimized IP
COM EXPRESS SHOWCASE
Featuring the latest in COM Express technology Intel® Pentium® M-915GM COM Express Module (SOM-5780)
Intel® Core™2 Duo-945GM COM Express Module (SOM-5782)
Embedded Intel® Pentium® M/ Celeron® M processor Intel’s new graphics core based on GMA 900 & external PCIe x16 graphics interface Supports eight host USB 2.0 ports and two SATA ports Supports dual channel LVDS panels
Embedded Intel® Core™2 Duo/Core™ Duo/Celeron® M processor Intel’s new graphics core based on GMA 950 & external PCIe x16 graphics interface Supports eight host USB 2.0 ports and two SATA ports Supports dual channel LVDS panels
Advantech Corporation Phone: (949) 789-7178 Fax: (949) 789-7179
E-mail: ECGinfo@advantech.com Web: www.advantech.com/applied
Phone: (949) 789-7178 Fax: (949) 789-7179
E-mail: ECGinfo@advantech.com Web: www.advantech.com/applied
Ampro COM 840
Ampro COM 830
Core™ 2 Duo COM Express Intel GME 965 chipset in Type 2 pinout of the official PICMG COM.0 Up to 4GB DDR2 Memory World’s first Extreme Rugged COM Express product ACPI power management with S3 for next generation portable applications High performance graphics engine and dual core computing for imaging, gaming, signage and streaming media applications
Ampro Computers, Inc. Phone: (408) 360-0200 Fax: (408) 360-0222
Ampro Computers, Inc.
E-mail: email@example.com Web: www.ampro.com
Kontron ETXexpress-MC - a COM Express solution Robust and power-efficient COM Express module Intel® Core™ 2 Duo processor and Intel® GME965 chipset Up to 4GB Dual-Channel Memory On-board TPM 1.2 Ready to evaluate today!
Kontron Phone: (888) 294-4558 Fax: (858) 677-0898
Core™ Duo COM Up to 4GB DDR2 Memory True PICMG COM Express COM.0 Compliance PCI Express, SATA ACPI 2.0 with S3 support x16 graphics on the baseboard 0°C to +60°C temperature rating
E-mail: firstname.lastname@example.org Web: www.kontron.com
Phone: (408) 360-0200 Fax: (408) 360-0222
E-mail: email@example.com Web: www.ampro.com
Spartan - 3A DSP 250 250
73% Performance Improvement
175 Spartan - 3A
100 Non-symmetric 366 tap single rate
Non-symmetric 20 tap single rate
Non-symmetric 4 tap single rate
Symmetric 31 tap single rate interpolate by 5 Symmetric 32 Non-symmetric 31 tap single rate halfband decimate by 2
Symmetric 69 tap decimate by 6
Symmetric 34 tap decimate by 6
FIR Filter Performance Comparison between Spartan-3A DSP and Spartan3A.
integrated into domain-specific design flows now available for embedded and DSP development.
Device-optimized IP reduces the barrier to adoption of FPGAs for non-traditional developers of production systems. Vendor-independent IP is an understandable choice when the FPGA is being used to prototype an ASIC. This retargetability, however, comes at a cost in silicon efficiency. When the FPGA is included in the production hardware, the quality of results will have a direct effect on the cost and performance of the overall system. In these situations, IP that has been specifically optimized for the FPGA is a preferable option. FPGAs have grown in recent years into sophisticated devices that provide a variety of specialized hardware resources. These enhancements to the original “sea of gates” architecture are what allow these devices to achieve their compelling performance, cost and power metrics. For example, today’s largest FPGAs include, in addition to over 51,000 slices of configurable logic, up to 640 dedicated DSP blocks that can
be programmed to perform a variety of operations, 10 Mbytes of block RAM, hardened embedded processors and dedicated interconnect IP such as PCI Express and Ethernet. RTL synthesis, in the hands of seasoned FPGA design experts, is a proven flow for developing hardware-efficient implementations. This flow, however, is not always viable for developers with programming backgrounds. Production quality results are achievable without RTL design skills using IP that has been optimized for the specific FPGA. IP provided by FPGA vendors is painstakingly optimized by in-house design experts for each supported technology. For example, an IP core optimized for a Xilinx Spartan-3A device would have to be re-optimized for a Spartan-3A DSP device to take advantage of new features such as the DSP48A slice. Figure 2 shows the effects of this type of optimization by comparing the results of the Xilinx FIR Compiler IP generator between the Spartan-3A and Spartan-3A DSP device families. Developing an implementation that targets these slices increases the filter performance by 73% while reducing device utilization.
Wizard-based Embedded System Development.
Creating the Embedded System
The most natural migration from processor-based hardware to an FPGA is to develop an embedded system that is sufficiently optimized to meet the system performance requirements. This preserves the programming model for application development while allowing the architecture customization to address performance issues. Easy-to-use wizard-based flows, such as Xilinx Base System Builder (Figure 3), are available from FPGA vendors for creating the initial embedded hardware. A series of forms allows a user to configure an embedded processor, connect it to the processor bus and select a unique set of peripherals in a few minutes. Once the initial embedded system has been created it can be further optimized using an embedded hardware development environment, provided by the FPGA vendors (Figure 4), that provides a bus-oriented view of the system along with a catalog of additional embedded IP and automatic generation of hardware and software driver files. IP integrated into a processor-oriented development environment brings this type of architecture flexibility within reach of designers with varied backgrounds. Fully exploiting the advantages
ETX form factor single board computer NANO-9452
FPGA Embedded Development Environment showing IP and Buses.
data_out 1 din
err_found err_cnt fail
blk_strt vout blk_end reset
Socket M Intel® Celeron M 1.0GHz CPU
Intel® 852GM + ICH4 chipset / DDR2 memory up to 1GB / 10/100 Ethernet / Dual channel 18-bit LVDS LCD panel / 4 x PCI support / ISA support / 2x IDE / 4 x USB/ 2 x RS-232
3 info 5 err_cnt 4 fail
Fanless AMD LX-800 Universal Controller
rrfd RS Decoder
Simulink-based FPGA Design. AMD LX-800 500MHz
Spartan 3-DSP / Virtex-5 Platform Studio
AMD LX + AMD CS5536 chipset / 8 x COM / Ethernet / 4 x USB / Windows CE 6.0 OPC Modbus support library provide by SDK/ ideal to work with PLC, industrial communication converter/gateway, and factory automation
Intel Core 2 Quad Mini-ITX KINO-9654G4 Mini ITX Figure 6
Embedded Interface Generation from System Generator.
of an FPGA-based embedded system, however, requires the creation of FPGA accelerated custom functions for use by the embedded system.
Turbo-Charging the Embedded System
Developing an embedded system on an FPGA allows performance-critical functions to be “turbocharged” through hardware acceleration on the FPGA fabric. Profiling the application code allows identification of “hot
spots” that diminish the performance of the entire system. Calls to functions executing on the processor can then be replaced with calls to functions that execute on a dedicated hardware accelerator implemented in the FPGA fabric. These custom instructions can dramatically improve the performance of the entire system. Creating efficient hardware accelerators requires optimal use of the available FPGA resources. Traditional processor programmers often don’t have
Ideal for High performance computing or network appliance
Intel Core 2 Extreme, Core 2 Quad CPU / DDR 2 memory / 4 Gigabit Network / 6x USB / 2xGbE / 4x RS-232 / 1x PCI express expansion / 7.1 Channel Audio
1-866-276-6754 Tel :1-909-595-2819
10/18/07 10:50:24 AM
the background to accomplish this type of low-level FPGA design. Automated â€œC-to-gatesâ€? flows have emerged as a possible solution. These flows are software centric and allow performance bottlenecks, identified through profiling, to be mapped to the FPGA fabric using a pushbutton approach. Modest performance gains are achievable but the hardware results tend to be more suitable for prototyping than for production hardware. An alternative approach would be to use a domain-specific development environment with FPGA-optimized IP. Such environments exist today for DSP applications that enable the use of The Mathworks DSP-friendly Simulink / Matlab modeling environments for FPGA development (Figure 5). Vendor-optimized libraries are provided for use within Simulink that include up to 100 DSP building blocks. Creating the design within Simulink allows these blocks to be abstractly connected and verified within the context of a DSP simulation.
Integration now exists to allow DSP development environments to automatically convert designs into custom peripherals for a corresponding embedded development environment. Integration includes the ability to automatically generate the hardware interface and software driver files and can often save weeks of development time (Figure 6). It is the combination of FPGA-optimized IP tightly integrated into a domain-specific development environment that enables developers with programming backgrounds to leverage the flexibility of an FPGA-based hardware system for their applications. These development environments are able to abstract away low-level implementation details of the interconnects and present the IP as abstract system building blocks in an appropriate context. Embedded development environments present a bus-centric view of the systemâ€”including masters, slaves and peripheralsâ€”which makes sense for em-
bedded development, while DSP development environments present a dataflow view of the system that includes fixedpoint bit growth and key DSP operations such as filtering or transforms, which is intuitive for DSP development. Integration between these two environments makes including a hardware-accelerated customer function for DSP a straightforward process. Because the IP supporting these environments is optimized for the target FPGA, production quality results are achievable, and the performance, cost and power benefits of the FPGA are fully realizable. Xilinx San Jose, CA. (408) 559-7778. [www.xilinx.com].
DIAMOND SYSTEMS IS DATA ACQUISITION! When you need data acquisition for your embedded application, we offer many patented, high performance solutions. Choose from either our industry leading single board computers with integrated data acquisition, or our stand-alone PC/104 I/O modules.
NEW! ATHENA II
0# !NALOG )/ -ODULE 0ATENTED AUTO AUTOCALIBRATION FOR MAXIMUM ACCURACY s BIT ANALOG INPUTS AT +(Z s BIT ANALOG OUTPUTS AT +(Z s DIGITAL )/ LINES s
-(Z OR -(Z COMPACT FORM FACTOR 3"# s )NTEGRATED DATA ACQUISITION WITH AUTOCALIBRATION s /N BOARD -"