Hot Topics

Michael Thompson on the 40-Gbit challenge

May, 2010
By: dl

 Michael Thompson, Principal Engineer at Pentair Technical Products, has left his imprint on innumerable embedded computing standards over the past decades as part of the VSO and PICMG efforts. What’s Thompson working on now? OAR asked. 

OAR: So, what have you been working on lately, Mike?

Thompson: I’ve been working on lots of ATCA and uTCA projects related to PICMG specification enhancements. One item in particular should be quite a challenge: supplying more than 600 Watts to double-slot ATCA blades that support 40 Gbit Ethernet. There are challenges here in performance, cooling, power distribution and mechanics. 

OAR: Whew!

Thompson: The original ATCA systems were intended for use with 1 Gbps Ethernet fabrics on the backplane. This was not difficult to achieve with careful PCB design. The second generation of ATCA systems supported 10 Gbps fabrics. This required lots of simulation and empirical testing to determine the optimum PCB design for a reliable fabric. Now the goal is 40 Gbps fabrics. That will yield a really impressive theoretical max bandwidth through a full mesh backplane of over 10 terabits/second! 

OAR: That’s a pretty good leap.

Thompson: Getting 40GBase-KR4 running on a backplane has been very challenging. 

OAR: We’re really talking about 10 Gbits/sec on a differential pair, not 40 Gbits/sec, right?

Thompson: That’s correct, 10GBase-KR is 10 Gbits/sec on a single pair, and 40GBase-KR4 is four of these pairs. The signal integrity requirements are very stringent and very difficult to meet. The signaling rate is significantly higher than what the ATCA specification originally planned for and what the backplane connectors were originally designed for. There are higher performance connectors in the works, though, that will mate with the existing connectors. 

OAR: So have you been successful running 40GBase-KR4 on ATCA? 

Thompson: Yes, through a combination of very careful PCB design for the blades and backplanes, even more simulations and empirical testing, the improved backplane connectors, and some real magic in the transceivers, we have demonstrated a backplane working reliably at 10 Gbits/sec per pair. 40GBase-KR4 is just four of these pairs. Our backplane wizard, Andreas Lenkisch, spent a lot of time simulating different layouts to determine the best design, then we built and tested different versions of backplanes to prove that the simulations reflected reality.

OAR: Who do you mean when you say “we”? 

Thompson: Pentair/Schroff simulated, designed, and built backplanes and then verified their behavior with Agilent signal integrity tools and a 10GBase-KR4 test set from Vitesse. We then sent the backplanes to Intel who tested their behavior with 10GBase-KR transceivers. They reported excellent behavior and a BER of less than the required 10-12.

OAR: How important were chips and connectors in reaching the goal? 

Thompson: Both were important. Because of the way the connectors are manufactured, the performance of the right-angle connector that goes on the blades was easier to improve. Both Tyco and Erni have developed an improved blade connector that can handle the higher frequencies. They are backwards compatible with the original connectors but have reduced crosstalk, better impedance control, reduced intra-pair skew and improved insertion loss characteristics. We would really like to see the via size for the backplane connector reduced, though. That would further reduce impedance discontinuities and signal reflections.

OAR: And chips? 

Thompson: Chip manufacturers have improved their designs to the point where I would call their behavior “magic.” I have seen images of eye patterns on backplanes (not ours) that were completely closed, yet the data transmission worked flawlessly. With all of these improvements in the interconnect, we can demonstrate–and we have demonstrated–systems that work reliably at 40 Gbps. 

OAR: What were the cooling issues you faced? 

Thompson: When ATCA was originally developed, cooling 200 Watts per single-slot blade was a stretch goal for existing technology. We are now working towards much higher blade power dissipation levels. Fortunately, the target market is the Enterprise and not the Central Office. This means that the maximum ambient temperature will be about 20°C less than it is for existing telecom ATCA equipment.

OAR: What is the cooling burden of these megablades? 

Thompson: We will need to move massive amounts of air to cool the proposed blades while keeping the acoustic noise from the chassis at reasonable levels. Fan manufacturers have really increased the flow rates for the fan sizes that we use. Just as with the backplanes, we are performing simulations and empirical testing to improve the air flow rates.We also offer liquid cooled cabinets to house the ATCA chassis. These cabinets reduce the ambient temperature by about 20°C, reduce the acoustic noise from the fans and increase the system’s MTBF. This allows for very high power densities without the associated noise. In cold climates, you can even recover the heat from the ATCA blades and use it to help heat the building. 

OAR: That’s interesting. How would you get the heat from the blades into a building’s heating system?

Thompson: Many office buildings circulate water for cooling and heating. During the winter, furnaces heat the circulating water and then heat exchangers use the water to warm the air in the rooms. Instead of dumping the thermal energy from the blades into the atmosphere, you put it into the building’s circulating water. Then you can use the blade’s energy to heat the building. 

OAR: What are the power distribution issues you face?

Thompson: An existing ATCA system has a maximum possible dissipation of less than 4 kW, or about 80A of -48V. We are now considering systems with a maximum dissipation of more than 10 kW. That means we need to distribute more than 200A of -48V. This is a huge amount of power to move through a backplane that supports redundant hot-swappable FRUs. 

OAR: And the solution is…?

Thompson: The solution is always lots of copper. The power connector manufacturers are retesting their products using a procedure that is closer to reality than traditional testing. The results show that the existing power connector can handle the increased current as long as there is sufficient copper in the PCB.

OAR: You also mentioned that there are some mechanical challenges. What are they?

Thompson: One of these chassis could be over two feet deep and weigh about 75 kg empty. Supporting the weight of 14 high-power blades will require a substantial mechanical structure. We can’t just add massive amounts of metal to the structure because it would increase the cost and possibly impede the air flow. So we are back to simulations and empirical testing again. 

OAR: Mike, you noted that this hypothetical 40 Gbit/sec-capable system is taking aim at the enterprise market. Does that mean that the T in ATCA doesn’t stand for “telecommunications” anymore? I remember a few historical precedents such as when RAID changed from redundant array of inexpensive disks to redundant array of individual disks.

Thompson: Even though the original target market for ATCA was telecom, we quickly found that our customers were using ATCA systems in the enterprise and even the military market. Maybe we should change the “T” to “technology?”

OAR: What attracted these two other market segments to a telecom-oriented architecture?

Thompson: I would say that the primary interests are threefold: management, very high performance fabrics and flexibility. The comprehensive shelf management capability of ATCA will let you know when sensor readings are abnormal long before they generate an alarm. For example, the fan tachometer sensors can tell you that one fan is running 500 RPM less than the other fans in the chassis. This slow fan will fail but not for quite a while. This will give you ample time to schedule a fan tray replacement when it is convenient, not after cooling is compromised.

There are also System Management solutions available from companies like ENEA and GoAhead that can automatically reconfigure the blades in the chassis based on abnormal sensor readings or if a blade fails. These management tools can really increase the MTBF of the chassis. Achieving high MTBF is not just a telecom goal.

OAR: What about performance?

Thompson: The potential bandwidth of a full-mesh backplane running at 40 Gbps is really impressive. It is much simpler to run 40G between blades over the copper in the backplane than it is through a nest of optical cables between rack- mounted servers.

OAR: And flexibility?

Thompson: The fabric agnostic backplane allows you to run Ethernet, PCIe and Fibre Channel on the same backplane at the same time. This allows the user to reconfigure the chassis architecture as needs change, instead of replacing the whole chassis. Everyone likes flexibility. .

 *****************************************

ARMs control and design paucity

May, 2010
By: lw

The biggest sigh of relief coming out of Embedded Systems Conference this year may have been the revelation that Apple Inc. really didn’t want to go after ARM Holdings PLC in an $8 billion bid for dominance, but instead was content to acquire Intrinsity Inc. to augment its existing PA Semi semiconductor holdings. Why should this matter to chip and board developers attending ESC? Because ARM cores were dominating single-board computers to an extent never before seen, and there was a real concern that Apple might be eyeing ARM to take the architecture off the market.

Make no mistake, ARM did not have a decades-long nefarious strategy to wipe out competing RISC and CISC cores through undermining their viability, unlike certain OEMs that may have been wanting to acquire ARM. No, ARM has achieved dominance the old-fashioned way. It has delivered on single-core and multicore embedded RISC processors at a variety of performance and price points, and has devised a way of working with fabless semiconductor players that can fit ASSP, ASIC, and FPGA models. As a result, its market ubiquity has achieved the type of level where any vendor that didn’t have an ARM license already was destined to add one this year. At ESC, Xilinx’s offering of an Extensible Processing Platform based on ARM had “foregone conclusion” written all over it.

Is there a problem here? Aren’t ASSP and FPGA vendors giving designers what they ask for? Well, yes, and there are still some decent independent core vendors out there, including CEVA, Tensilica, and Wintegra. We mustn’t forget MIPS, who took pride in announcing at ESC that it had signed a pact with an “anchor Asian licensee” – unannounced, though analysts were betting on Renesas or Spreadtrum.

Still, a VPX or PCIe market dominated by ARM with a slice of Intel Atom left for diversity seems a little emptier than one in which PowerPC, MIPS, and TI or ADI DSPs played a major role. If ARM was only taking over FPGAs, the processor jungle still might be a rich ecosystem, but the ARM core dominates ASSPs in baseband (Qualcomm, TI) and IT (Marvell) as well. Sure, Freescale launched a new StarCore DSP family at ESC, but let’s remember Freescale also is seeing success in its ARM-based i.MX family.

The world has not yet dwindled to one-ARMed bandit status. Designs using unique cores from the likes of CEVA and Tensilica still are prevalent, and the MIPS and Power architectures could yet score big wins. Still, it’s hard to come to the conclusion reached by EE Times that “MIPS Elbows ARM.” More like MIPS gasps for a few breaths to prove it is still alive. And good for MIPS. When ARM has such an overweening presence in embedded systems that it serves as a hand that rocks many cradles, an effort by Apple or anyone else to take over ARM Holdings would be deadlier than anticipated.

********************************************

What’s new at ESC this year? It’s Dell.

April, 2010
By: eg 

Many industry pundits are forecasting that this year’s Embedded Systems Conference (ESC) Silicon Valley will be an “A” show … because ARM, Android and mobile Apps are creating much of the buzz. I, on the other hand, think this will be remembered as the first ESC that Dell participated in as an exhibitor. The conference, co-located with the fifth annual Multicore Expo, runs from April 26-29. 

Dell’s corporate headquarters is in Round Rock, TX, just a short drive from my office outside of Austin, TX, and earlier this week Rick Froehlich, the VP and General Manager of Dell OEM Solutions, was kind enough to meet with me and give me some insight into why Dell will have a booth at ESC. 

For starters, Dell is not new to the embedded business; in fact, they’ve been at it for over 13 years. And this year, the Dell OEM Solutions business (which includes their embedded business) will total over $1 billion. A key question, of course, is, How does Dell define embedded? Interestingly enough, Dell defines embedded exactly the same way you might. For Dell, an embedded customer is one who puts a Dell computer inside some other type of machine. Medical diagnostic machines (as in CAT Scan/CT and MRI), process control equipment, security systems and retail kiosks were just some of the vertical markets that Froehlich mentioned during our conversation. And he made a point of telling me that Dell is actively tracking 40 vertical embedded computing markets. (I say he made a point of it because he mentioned that 40 number twice in less than an hour.) 

For years, portions of the low-end embedded market have been migrating to various off-shore motherboard and white box vendors. But traditional suppliers in the embedded computing market have been quick to rationalize these losses because “these embedded customers do require the rigorous configuration control and guaranteed source of supply that real embedded customers demand. “ 

But Dell is in the process of changing much of this. The $65+ billion company that supplies laptops, desktops and servers to individual consumers and businesses alike has perfected the build-to-order model. So it was not a matter of the OEM Solutions group teaching Dell how to do configuration control, but rather the OEM Solutions group had only to adapt to the embedded market the robust Dell configuration processes that were already in place. Thus, providing the configuration control that much of the embedded market demands did not require a huge shift in corporate culture at Dell. 

That’s not to suggest that they are simply selling their consumer (or even their commercial) products into the embedded space because they’re not. But they are leveraging everything they can from the Dell giant, including rigorous configuration control, global service and support, etc., and to that they have added hardware specifically designed for the embedded market. The flagship of the current Dell embedded product offering is the Dell OptiPlex XE platform, offered in two sizes: a “desktop“ configuration and a “small form factor” box. Note that Dell’s embedded products are available ONLY as box level products. Froehlich was adamant that Dell has no intention of offering board level products, not even motherboards. He maintains that the company will not compete with its customers or, for that matter, with its suppliers. 

So, if you’re wandering around the ESC show floor this week, I suggest you stop by the Dell booth (#629) and take at look at what they’re offering to the embedded market. If you’re putting together a new embedded system, you may be pleasantly surprised at the performance, flexibility, lifecycle support, global service and price of the various OptiPlex XE solutions. If you’re a board-level vendor that has PCIe I/O hardware for a specific vertical market, then it might be well worth your while to start exploring how you can work with Dell to provide complete solutions into that vertical you know so well. And if you’re a motherboard or box-level vendor in the embedded computing space, then you might want to drop by to see what you’re up against. Obviously the Dell solutions won’t satisfy every customer or be a good fit for every niche. But as a motherboard or box-level vendor, you’re going to have to get very focused on how to differentiate your products. You’re going to have to do better than the old saws of “configuration control” and “lifecycle support.” And vague references to “performance” or “open architecture” won’t win the day, either. 

Earlier this week, the WSJ reported that industrial companies are enjoying a surge of new orders. They didn’t mention the embedded market specifically but, as you well know, embedded computers are everywhere. The world economy is looking up. This is not a zero sum game. This bodes well for Dell …and for the rest of us. 

Disclosures for the FTC, SEC, FCC and anyone else who wants to know: I own two Dell computers, as well as numerous other computers from a variety of other vendors. I paid full retail price for both of the Dell machines I own. And Dell has not furnished me with any financial inducement, either direct or indirect, in exchange for the comments and opinions expressed in this article. And neither has there been any promise of any future compensation from Dell (either direct or indirect) as the result of the comments and opinions expressed in this article.

 **********************************************

M&A: Everybody’s panacea

April, 2010
By: ra 

It’s true that there are quite a few M&A discussions going on in the embedded computing community (see A whiff of M&A), particularly involving Mil/Aero companies. Although difficult economic conditions are a major factor in this phenomenon, other issues are also coming into play. 

Q1 of 2010 held some surprises for the board and box industry. Based on collected reports and conversations, we saw a dip in sales in Q1 2010 over Q4 2009 and also experienced a dip in bookings. This can probably be attributed to the general 2009 financial problems, the lack of liquidity, and the hoarding of cash by most customers.A recent VITA white paper  chronicles the big dip we saw in sales and bookings in 2009: down 70% in telecom, down 30% or more in industrial and down somewhat in medical, in contrast to some growth or a flat market in MIL/Aero. We did, however, start to see some softness and uncertainty in Mil/Aero bookings in late 2009, the leading edge of the Q-1 dip. 

The general economy is now picking up, and it looks like Mil/Aero will lag the turnaround by about six months. That suggests that things might start improving by Q2 of 2010, though events are unforeseeable and Q3 may be a better bet. So, if companies cannot find organic growth in our industry, they’re considering M&A.A macro-perspective might be in order here. If you take the top ten traditional companies in the board and box biz, and add their sales up, you get to about $2.8 billion in 2009. If you assume the Pareto Principle applies (i.e., the 20-80 rule), and you add about 20%, you get a worldwide embedded board and box market of around $3.4 billion.  

There are eight companies in the top ten that do more than $100 million in sales, the largest being over $600 million. No company in this industry has ever achieved more than about $800 million in sales. That was accomplished by the old Motorola Computer Group during the Great Telecom Party back in 2000. Intel never got much bigger than about $500 million back in their heyday of Multibus and Multibus II in the ‘80s and ‘90s (and they never achieved much over $50 million in their big telecom push back in the early 2000s). So, there are a few large companies trying to get to $1 billion in sales, and the only way to get there is through M&A. Now, if you take those top ten companies, and break-out what they ship to the Mil/Aero market, you get about $750 million. If the Pareto Principle is still valid, you get $900 million as the size of the board/box market for the Mil/Aero segment worldwide. That means the Mil/Aero segment is about 26% of the total worldwide board market.

We all know that the US is the largest of the Mil/Aero markets, the others being the UK, Spain, Canada Australia, Russia, Israel, and a few more small countries. So, there are non-US board companies that want a foothold in the US Mil/Aero market, and they can accomplish that goal through M&A.

Additionally, there are $200-$300 million companies in the Mil/Aero market that want to enhance their position, especially in the US market. For them, smaller $30-$50 million companies with contracts on existing programs make for attractive acquisition targets. So, there are several companies seeking to expand their footprint in the Mil/Aero through M&A.

If you look at the major board market segments, telecom has the worst prospects for any improvement in 2010. Any company with a large footprint in that segment is looking for acquisitions, particularly in the Mil/Aero segment, to find growth and stability. So, there are also companies looking to M&A for the sake of diversification.

A look at the companies bought and sold over the past few years with commodity products or a telecom focus suggests that they are making between 10 and 15% GPM. They have sold for 0.5 to 0.8 times sales over the past ten years. 

As sales and shipments have declined in this economic downturn, commodity-oriented companies have lost economies of scale. So, some of the companies in the commodity product segments are looking for fire sales of similar commodity-oriented companies to increase their volumes and get back some cost leverage.All the Mil/Aero acquisitions tracked for the past ten years suggest those companies are making 50-70% GPM and they sell for 2.2 to 3.5 times sales. Commodity-oriented companies who have lost economies of scale need to raise their margins on their overall business.

So, there are some commodity-oriented companies looking to raise their overall margins by buying Mil/Aero companies. Once again, you guessed it, they’re looking to M&A, and it’s no surprise that the most appealing targets are companies that have strong Mil/Aero-oriented product lines and a solid position in the Mil/Aero segment.

 I do see some activity involving acquisitions of some commodity and telecom product companies, but those will go down at “firesale” prices. Visibility into the DOD budgets for 2010 and 2011 is fairly clear. Past that, things get fuzzy, especially with the present administration’s priorities on social programs. Even with that caveat, Mil/Aero companies are still the most valuable targets for acquisitions in 2010.

*******************************************

A whiff of M&A

April, 2010
By: dl

I suffered a bit of nostalgia a few months ago when one embedded computing company rolled out a big bunch of products, some of which looked very familiar. I realized they were familiar to me because they were continuations of board families, but board families from companies that had disappeared as independent entities many years ago. These companies had, of course, been acquired. It was great to see the different legacies being pursued and a diversity of customers served over the long haul.

It’s now spring, and a whiff of mergers and acquisitions (M&A) is in the air. Sources are saying that the current line-up of embedded computing companies may look quite different later this year, particularly in the mil/aero segment. Some big embedded computing companies might expand or contract their empires, some medium size companies might get gobbled up, and a dark horse just might ride in from outside the embedded world.

Several of the larger embedded computing companies are themselves the product of acquisitions in an earlier age. The Embedded Computing group of Curtiss-Wright Controls , for instance, was stitched together from Vmetro, Synergy Microsystems, DY-4, Vista Controls, Pentland and Peritek, plus a couple of verticals. For its part, GE Intelligent Platforms (formerly GE Fanuc) was built up from Computer Dynamics, VMIC, Radstone and others, including SBS. That company had itself taken a stab at empire building in an earlier era, bringing Greenspring, Logical Design Group, Bit 3, VI Computer, OR Computer and others under the SBS umbrella.

Elsewhere, empires have been built by Kontron, with its acquisition of PEP Modular Computers, Thales Computer, Teknor, Jumptec,  Matrix, Intel’s rackmount server operation and others; and by Emerson Network Systems, which gobbled up Motorola’s embedded computer group, Force Computers, Heurikon, Blue Wave Systems, Mizar, Prolog and others. Some of the empires have worked out very well, and some have not.

What’s behind the next M&A wave? The pundits will certainly come up with their reasons, but perhaps it’s all just anxiety driven. Waves come in and waves go out, while companies shrink and grow when they’re afraid that they’re too big or too small to survive, that they’re not sufficiently diversified or have stretched themselves too thin, etc., or they can see from their market numbers that they’ve clearly bet on the wrong horse.

Is a new board vendor shuffle a good thing for the marketplace? Yes and no. We’ll have to wait and see how good the acquirers are at empire building.

*******************************************

Waveguides as backplanes
………..and other optical musings

March, 2010
By: lw

VITA executive director Ray Alderman was a bit overwhelmed last week, after the trade group announced formation of the VITA Architectures for Optical (VAO) Study Group.  Despite the work on signal conditioning for 10-Gbit copper links, there’s a growing assumption that systems will be making the inevitable turn to optical backplanes as aggregate speeds move to tens of gigabits per second.  Alderman said that everyone from IBM to ITT to Boeing has been kicking the tires on 40-Gbit/sec backplanes they could implement as soon as possible.

The depth of the recession has masked the effects of the lag in board upgrades as users move from copper to optical.  Alderman said some figures show declines in line card and motherboard sales as great as 70 percent in telecom market segments, and 30 to 35 percent in industrial sectors.  But such pauses also represent a good time to think about fundamental shifts.

Plenty of concepts for using free-space optics, active optical cable or polymer waveguides have been floated in academia, but Alderman said what doesn’t seem to carry a lot of hope is the traditional fiber backplane in either glass or plastic.  Plastic will go opaque to signals at high speeds, while glass fiber cannot offer the flexibility or bend at right angles, as many footprint-sensitive board designs require.  VAO likely will spend a good deal of time examining the potential for reflective-polymer waveguides, which could be expanded to dozens of waveguides on a single backplane, capable of achieving Terabit speeds.

“We are seeing designs today in radar and sigint [signals intelligence] where fast channels can be merged with mezzanine concepts like FMC (FPGA on Mezzanine Card, VITA 57),” Alderman said.  “I’ve seen a sigint concept where FMC is used to implement 80 channels of 10 Gbits/sec per channel.”

It’s also possible that some of the back-of-the-box fiber interconnect topologies used in data center clustering could become de facto backplanes, albeit without the traces laid down on system boards.  VAO will study a line-replaceable unit (LRU) as a substitute for a traditional line card.

While VITA will be offering its VAO finding to ANSI as it has done with other standards efforts, the organization won’t be operating in isolation.  The Optical Internetworking Forum has been working for the past two years on bringing coherency to the growing number of transceiver modules and connector types for 40G and 100G fiber networks, aiming for standards that span server links, LANs, MANs and WANs.  And on March 10, the Air Force Research Labs’ Manufacturing Technology Division at Wright-Patterson AFB announced a $2.4 million Photonics Manufacturing Program to help speed development of fiber interconnects.

The Air Force and OIF efforts can leverage the IEEE 802.3ba work on 40G and 100G Ethernet, which is helping to reduce connector and very short- reach cable costs for the data center.  Coalitions like Interlaken Alliance feed into the IEEE work to make emerging modules, including CFP and CXP, more cost-effective than they otherwise might be.

What goes on in the LAN and the server cluster is not automatically reflected in the backplane or LRU, though Alderman said it certainly helps.  As network planners become more familiar with signal conditioning in very short-reach clusters, their work can ease the transition to waveguides or other optical alternatives within the confines of a single chassis, or even a single motherboard.

*******************************************

Minimalist I/O scheme struts its stuff

SPEC WATCH
March, 2010
By: dl

Basically, the FeaturePak interface is power, ground, one or two x1 PCI Express links, one or two USB 2.0 ports, a serial port, a chunk of reserved lines, a few management and utility lines, and up to 100 user-definable I/O lines, some of them electrically isolated. The form factor is 2.55 x 1.70-inches, just under 4.5 square inches, which is small indeed even in the small-board world. The form factor of the popular PC/104, in comparison, comes in at 13.4 inches.

Fathered by Diamond Systems, the spec is now supported by seven other companies, most of them European. “Following the FeaturePak Initiative’s initial launch,” said Diamond founder and president Jonathan Miller, “we intend to turn the FeaturePak specification, trademark, and logo over to a suitable standards organization so it can become an industry-wide, open-architecture, embedded standard.”

The charter members of the FeaturePak Initiative, besides Diamond, are Arbor Technology, Cogent Computer Systems, congatec,  Connect Tech, Douglas Electronics, Hectronic AB, and IXXAT Automation.

The membership includes three companies who are also members of the Qseven consortium. That consortium took a COM standard developed by congatec in 2007 and rallied around it as a group in 2008. The Qseven form factor is also very small at 2.8 x 2.8 inches and about 7.8 square inches. Today, about two dozen embedded computing companies support the Qseven spec, with eight of them selling ten different Qseven boards, and ten of them offering Qseven carriers.

FeaturePak uses the same basic fast (2.5 GHz), high density (0.5 mm pitch), connector as Qseven: the 230-pin MXM from Foxconn, Speedtech or Lotes, originally developed for use in notebook computers. The MXM is a novel, swiveling low-profile connector: a FeaturePak board sits only 2.7 mm or 5 mm above a baseboard.

At last week’s conference, Diamond introduced its maiden FeaturePak products: a carrier and a DAC board the company says integrates its “newest and fastest analog I/O technology.” Connect Tech, in turn, demo’d a serial I/O module for FeaturePak and a PCI Express to FeaturePak adapter.

Why no LPC (Low Pin-Count) bus in FeaturePak’s interface mix, a bus that’s common to other contemporary  mix and match interface architecture schemes such as Qseven, SUMIT and COMIT?

The FeaturePak Initiative explains: “It was decided that 2 PCI Express x1 lanes, 2 USB 2.0 channels, SMBus, and a serial port were sufficient to cover the full range of current requirements, particularly in light of LPC being an Intel-specific bus that targets a diminishing number of legacy ISA peripherals. This decision allowed us to reserve more connector pins for future purposes, thereby helping to ensure that the FeaturePak specification will be able to adapt to evolving host interface requirements.”

“FeaturePak” may seem a strange name to give a minimalist interface architecture that packs no bells and whistles whatsoever, but that’s irrelevant to whether it will succeed in the marketplace or fail. Perhaps the times are ripe for a modular expansion scheme offering very fine-grained I/O functionality, for small boards and simple interfaces that slip neatly into baseboards across a range of different architectures. Perhaps not.

www.diamondsystems.com/files/binaries/featurepak-launch-pr.pdf

**************************************

Consumer mezzanine needs its own parade

January, 2010
By: lw

Casual attendees at the January Consumer Electronics Show in Las Vegas may have been surprised to see an FPGA mezzanine card (FMC) on display in the consumer TV pavilion, courtesy of Xilinx Inc.’s Targeted Design Platform. VITA hoped for great things with VITA 57 FMC, but consumer products weren’t necessarily high on the agenda. If FMC could move beyond automotive, military, and factory-floor roots and into the digital den, the economies of scale would be significant.

In theory, a modular card in a retail product might represent a greater advantage than it already shows in mil/aero and industrial markets. Interfaces requiring programmable logic for implementation, such as High Definition Multimedia Interface (HDMI) or DisplayPort, could be implemented on the FMC. At the very least, the OEM could swap out cards to differentiate product lines before moving into the distribution chain. Ideally, the FMC could be exchanged by the retail dealer or perhaps the end customer.

To some extent, this is precisely what Xilinx had in mind. When it introduced the Targeted Design Platform (TDP) concept in early 2009, the theory went far beyond evaluation boards for vertical markets. The Xilinx TDPs included middleware, soft IP cores and license rights, development and debugging software, and sample application software. When Xilinx partnered with Tokyo Electron Device Inc. for the digital TV market, the FMC-based TDP was meant to make customization easy and cheap. Harry Raftopoulos, director of consumer segment marketing at Xilinx, said that the smaller budgets and tighter schedules experienced by consumer-product OEMs in the post-2008 economy mandated that they could not spend months studying LVDS, HDMI, and DisplayPort interfaces. Hence, the arrival of FMC allowed strapped vendors to accelerate the design cycle.

But before we get too excited about VITA 57 gaining a market dwarfing any other, let’s realize that Xilinx and TED are wonderful drum majors for this parade – but they’re missing the multitudes of followers behind them. The FMC standard still must overcome public perceptions regarding previous mezzanine standards such as PMC and XMC (VITA 42). Sure, the earlier mezzanines did have and continue to have devout adherents, but the use of the cards hasn’t radicalized product delivery methods moving all the way out to the end user. There is an analogy here to the evolution of fiber optic interfaces on networking boards. New Multi-Source Agreement modules like SFP, XFP. and SFP+ certainly are much easier to use than a 300-pin optical module. But no MSA form factor has achieved the ubiquity of an RJ-11 or RJ-45 Ethernet jack.

Think of the years of development efforts that have gone into turning USB from an interface for printers and peripherals to a veritable fashion accessory when integrated in a memory stick or a communications device. A USB stick played a starring role in the movie Cloudy with a Chance of Meatballs, and USBs now represent an alternative way of releasing a music album, next to LPs and CDs and digital files. The proponents of HDMI or DisplayPort interfaces would like to see their own interfaces become USB equivalents, but the FMC advocates insist that it’s easier to make the mezzanine module a universal “great equalizer.”

Let’s look at what has happened in networking and enterprise IT markets. PMC and XMC have gained popularity as a means of increasing Ethernet, Fibre Channel or T1/T3 port density on network equipment, but the message regarding ease of use has only migrated as far as the OEM. Network distributors and system integrators do not stock PMC or XMC cards the way they would stock PCI Express NICs, for example. PIMG was hoping that the overall standards for ATCA and MicroTCA would drag the AMC mezzanine in their wake, but these standards have not been widely adopted. Truth be told, the networking and telecom equipment markets are not healthy enough these days, in terms of numbers of OEMs, to support common backplanes and subsystem interfaces.

For mil/aero and industrial control markets, there are solid reasons to consider OpenVPX, and solid reasons to adopt mezzanine board form factors that emphasize reprogrammability. In consumer worlds, such efforts have to fit with an end-user “plug and play” perception, and must be driven beyond the tier of OEM to have staying power. The VPX Marketing Alliance, working on its own, can’t promote FMC, but neither can a single vendor alliance like Xilinx/TED. The Xilinx demonstration represents a good preview as to what is possible if FMC is used in consumer devices. Now, the trick is to get the IEEE, the trade associations for HDMI and DisplayPort, and similar coalitions behind this effort. FMC needs a cheering section of thousands to gain the traction it deserves.

**************************************

New interface designs are taking off

January, 2010
By: ra

Just when you thought we were done creating “bus specifications” for all the serial fabrics, two more initiatives show-up on our radar here at OAR. Actually, these are not “buses” as we have defined them in the past. They are, instead, glorified wiring harness specifications for connecting subsystem components into a system architecture: in one case for satellites; in another for airplanes.

Some of the connections in these specifications, such as power feeds to the LRU (line replaceable units), are likely to be multidropped like a classic bus. The data connections, however, will almost certainly be point-to-point. The systems management and monitoring connections could be either multidropped or point-to-point. It’s unclear as yet whether the external connections to these subsystems will be proprietary or open, but the internal connections will probably be proprietary.

One of the exciting efforts on this front is a satellite plug-and-play bus being developed and funded by the U.S. Air Force with the goal of creating COTS-based satellite components that can be put together in days and launched at a very low cost.  (See  Northrop Grumman to Design Air Force Plug-and-Play Spacecraft Bus and Air Force extends plug-and-play spacecraft. ) As might be expected, NASA has also been deeply involved in this effort, and Northrop Grumman recently received a contract to further its development.

The objective of this project is to create a basic satellite platform–with all the power, data, control, and monitoring lines defined and wired-up–so that small mission-specific LRU’s can be designed and easily plugged together. This would reduce the time-to-launch of a new satellite down to days or weeks. Data connections for the LRUs, which consist of electronics in a standardized, enclosed metal can (similar to VITA’s V-58 LRU specification), will be accomplished with a TCP/IP compliant router. NASA seems to have plans for launching many “nanosats,” small satellites weighing between 11 and 110 pounds, in communications and networking apps in space.

What’s the motivation here? you might ask. Consider this scenario. Back in 2007, China shot-down one of its failed satellites, proving that the Chinese can take-out intelligence satellites and, thus, render the military of an opponent blind by shutting off critical communications networks. In 2008, the U.S. Navy shot-down an errant U.S. satellite, proving that we can take out an opponent’s satellites, too.

Given this scenario, the rationale behind this new standardized satellite bus may be as follows. That the only way for us to counter an opponent’s threat to our intelligence and communications satellites is to put more of them up there faster and cheaper than the other country can build missiles to take them down. If the cost of the missile to shoot down a satellite is higher than the cost of the standardized COTS-based electronics package in the satellite itself, that makes a star wars race a bad business deal for any opponent.

Another new bus is being developed by the U.S. Navy in conjunction with Honeywell: an avionics bus for cockpit instruments in military aircraft. (See Insider: Honeywell proceeding with caution and U.S. Navy avionics systems integrators embrace open architectures to combat parts obsolescence .) It’s designed to allow easy insertion of new technologies in military aircraft avionics systems to reduce the tremendous obsolescence problems endured by the military for decades. Avionics sales have suffered terribly in this recession, and a new standardized avionics bus may be the key to increased sales in the future. As with the satellite bus discussed above, the new avionics bus is LRU based. It’s expected that once proven in military aircraft, it will rapidly migrate to commercial aircraft.

The avionics bus concept has, of course, been around for decades. Way back when, King made a lot of function boxes (AirNav, Transponders, Artificial Horizons, compass, GPS, Direction Finders, weather radar, COMM, digital displays, etc.) which it interconnected with a proprietary avionics bus. King merged with Bendix, Bendix merged with Allied Signal, and Allied Signal merged with Honeywell.

Remember when GE announced their purchase of Honeywell back in 2000 ? The US FTC approved the merger, but the EU nixed it, saying it would give GE a monopoly in the avionics market. Some very large companies have their eye on this market, and a new standardized avionics bus that mitigates those pesky obsolescence problems might be the elixir that increases sales and opportunities in the depressed aerospace market.

Despite the ascendance of open-architecture buses in recent decades, I have some concern that the new avionics bus and satellite bus may be proprietary. Hopefully, depressed market conditions for aircraft and avionics components have changed the old proprietary thinking, and these will truly be open standards. The fact that both rely on enclosed LRU boxes, however, suggests that the interfaces inside the box are not standardized, although the boxes may communicate over an instrument panel bus that will be standardized and open. We’ll have to wait and see.

**************************************

Signal processors go head over heels
for Intel’s Core i7 family

January, 2010
By: dl

The new members of Intel’s Core i7 microprocessor family  introduced at this month’s Consumer Electronics Show have created quite a stir in the embedded computing community. In his blog, John Keller, editor-in-chief of Military & Aerospace Electronics magazine, suggests that there “may be a tectonic shift” under way that would displace PowerPCs with i7s.

“While Intel sees the floating point capability of its Core i7 processor as the gateway to a new generation of complex graphics and fast streaming video,” Keller writes, “military systems designers see it as the latest and greatest way to implement signal processing for advanced radar, sonar, electronic warfare, and electro-optical applications with commercial off-the-shelf (COTS) single-board computers.”

He notes that Curtiss-Wright Controls Embedded Computing, GE Intelligent Platforms, and Extreme Engineering Solutions Inc. all announced i7-based SBCs “within hours of Intel’s introduction.” I’m not quite sure of the exact timing, but Concurrent Technologies, Kontron and Emerson Network Power also bellied up to the i7 bar in short order, and others are sure to follow.

The SBC makers’ praise of the new Intel entry is enthusiastic, indeed, perhaps even a bit over the top. GE Intelligent Platforms, for example, claims “remarkable performance” for its i7-based boards. “The increased integration and increased density of this new family of processors from Intel offers us astonishing new opportunities,” said Peter Cavill, GM for military and aerospace products at the company.

And, according to Dirk Finstel, CTO of Kontron AG, that company’s i7 entry represents “the ultimate computing tool MAG HPEC users have been waiting for, allowing them to finally walk away from 10 years of PowerPC Altivec dominance in radar, sonar and imaging applications.”

Indeed, Freescale and the PowerPC architecture may have become vulnerable in high-end signal processing applications that hear the clarion cry of Intel’s Streaming SIMD Extensions. And Freescale may have lost some of its fans by leaving Altivec out of its newest CPU core (See End of Altivec PowerPC digital signal processing chip spells headache for Serial RapidIO designers), while in some signal processing arenas, PowerPCs have been replaced by designs based on FPGAs.

On the other hand, old timers in the arena recall Intel’s on-again, off-again love affair with the embedded world, with buses such as Multibus II that went from prince to frog in short order, microprocessors such as the i960 that quickly went from star to red-headed stepchild within a decade and other such items. Is Intel in it for the long haul this time?

Who knows. But in the meantime, let the i7 games begin!

**************************************

How long can backplanes hold out?

December, 2009
By: ra

The demise of backplane-based computing systems has long been predicted, but the horizon is now coming into view. For the past decade or more, we have seen one embedded board segment after another drop backplane-based computer systems and move to interconnected motherboards, small form factor boards or little computing boxes. The industrial segment, using CAN, Profibus and Ethernet, quit using backplane-based systems long ago. That initiated the demise of STD, ISA, EISA and PCI backplanes. CompactPCI (cPCI) came on the scene about the time the shift away from backplanes occurred and was never heavily used in industrial applications.

The medical industry (MRI, CATscan, PETscan, CT, etc.) moved away from backplane-based VME and cPCI systems a few years ago, migrating to motherboard-based boxes interconnected with high-speed serial interconnects. Many medical apps have only a few control requirements (e.g., moving a beam into position), and image reconstruction is easily done with a commercial motherboard of some kind. These systems have no real time, shock, vibration or extended temperature requirements, so they can use cheap commercial 1U motherboard boxes interconnected with some networking technology like Ethernet.

That leaves two remaining markets for backplane-based systems: telecom and military. Telecom has been procrastinating about going to interconnected and stacked black boxes for years. They still have the old POTS mentality that favors backplane-based switch gear and have not made the mental or financial transition to a pure IP (Internet Protocol) communications system.

That change must happen in the next few years, so telecom may be the next segment to abandon backplane-based systems and go to large stacks of interconnected 1U boxes, especially as the telecom industry becomes more and more commoditized. The only remaining concern for them to resolve is how to maintain those systems: i.e, how to easily remove and replace the boxes in the rack.

That leaves us with military systems as potentially the last computing segment to rely on backplanes to carry their primary interconnect. But that, too, will change. Real-world requirements mean that the military will ultimately also move away from backplanes and go to stacked “boxes” known as LRU’s (line replaceable units), which are defined in the VITA-58 specification.

Today the military spends millions of dollars training soldiers to repair and maintain their sophisticated electronic systems. These people typically leave the military and take that extensive (and expensive) training to the civilian world after a single tour. A technician may be in a military training school for over a year, and then it takes about six months for them to become proficient in the field. They work at their job for a year or two, and then leave the service.

The upshot is that the military must use regular soldiers to maintain and repair systems in the field by simply swapping out LRU’s, while only career soldiers doing depot maintenance get computer training. Telecoms are seeing the same scenario: they spend millions on training and education and then lose those employees to other firms.

The primary people who repair and maintain systems on the battlefield must be the soldiers who operate those platforms in battle. They will simply pull-out failed LRU’s and slide-in replacements. The non-functioning box goes back to a depot where trained technicians will replace failed components and put the “box” back into the replacement LRU supply chain.

This suite of military requirements will drive computers to LRU’s.

o A critical military system must be repaired and battle-ready in 30 minutes or less.

o The people who maintain and repair that system must require no more than 10 minutes of training.

o No tools must be required to repair the system and make it battle-ready in 30 minutes or less.

o No unit can weigh more than 37 pounds for a two-man (or woman) lift.

Of course, the military has too many legacy platforms to make this change quickly. There are too many ATR boxes in aircraft, too many VME chassis in weapons platforms and helicopters, etc. Moreover, the serial interconnect technologies to connect those LRU’s (various fabrics and Ethernet) are too unstable. That is to say, their transmission frequencies keep going up, their electrical signaling protocols are changing from 8B/10B to PAM, their associated semiconductor chips have an 18- to 24-month life cycle, etc. Just looking at the number of refreshes to existing VME backplane-based systems says it will be decades before we can make the switch to LRU-based MIL systems, although a few are now going to deployment.

Down the road, the incentive to abandon backplane-based systems and even copper interconnects will grown even greater. As we move to 10G links and, as a consequence, adopt optical interconnects, the era will favor 6-foot, 19-inch racks of commercial boxes and ATR boxes of LRU’s woven together with 10G optical connections.

Ultimately, the highest costs of computing equipment in telecom and MIL applications are the life-cycle costs: installing, maintaining and keeping that gear functioning. Doing that requires highly trained, intelligent and expensive people at the repair depot. The only way out of the present cost trap is to use LRU’s as stackable boxes for the field, maintained by regular soldiers or employees.

Are we finally seeing the demise of copper backplanes? Yes. We have seen it occur in industrial and medical apps already. Telecom is next. And when 10G optical connections are inexpensive and stable, MIL apps will drop backplanes, too. But it will take the MIL segment a decade or more to make the shift. Until then, the MIL segment will be the only remaining market for copper backplane-based embedded computing systems.

**************************************

Quo Vadis, VPX?

December, 14, 2009
By: lw

It’s no surprise that in early December VITA took a critical role in tying together the loose ends from the ad hoc OpenVPX Marketing Group to create the VPX Marketing Alliance, contributing a new acronymn to the annals of embedded computing: VMA. This organization takes responsibility not only for VPX and OpenVPX, but for VPX REDI as well, plus any coming extensions to the family of standards. The near-term extensions will include optical and RF communications.

Two quotes from Sun Microsystems founder Scott McNealy come to mind here – the first on the necessity of uniting standards efforts behind a “single arrowhead,” the second on the need to temper committee enthusiasm over any paper standard it develops with the realization that “Standard equals volume shipped.”

By relying on the historical headstart VME had over the PCI-based server world, VMA can lean heavily on the concept of a single arrowhead. If only ATCA could rely on the historical work of PCI-SIG in a similar fashion, there might be a better story to tell, but there are one, two, many PCI’s. PCI does not equal PCI Express does not equal ATCA/MicroTCA, thus denying the SIG the benefit of that single arrowhead.

But that arrowhead advantage can be lost as OpenVPX enters new worlds. As VMA seeks to expand beyond backplane-module-chassis into broader definitions of interconnect in optical and wireless worlds, it would behoove the organization to remember that unified support of a suite of standards often represents merely the baseline of driving success. If we look at the history of communication alliances, marketing forums in the past decade often had to shift gears in sudden and unexpected ways:

• Optical Internetworking Forum, as a creature born of carriers in a long-haul transport environment, never intended to create physical-layer standards for optical interfaces operating at the physical layer, particularly in the enterprise and server-cluster realm. It certainly never intended to absorb and take on the work of the Network Processing Forum. But as NPF crashed and burned as designers decided to look to other options than an NPU, suddenly the OIF looked like the logical place to hammer out physical standards for microprocessors interfacing with short-reach optics. Who’d a thunk it?

• DSL Forum touting Digital Subscriber Line, and ATM Forum touting Asynchronous Transfer Mode, would not have considered their fortunes identical ten years ago, and would not have seen a reason to subsume their fates under a common header. And IP/MPLS Forum certainly approached broadband applications from an orthogonal direction to both groups, in particular as a direct competitor to ATM Forum. Yet somehow, they all seemed to coalesce into the Broadband Forum, despite the best intentions of marketing wars.

• Trusted Computing Group began life as an embedded processor and module alliance specializing in the tamper-proof component known as the Trusted Platform Module. But the lack of an effective organization promoting higher-layer IT security prodded TCG into expanding into such unexpected realms as secure storage, secure mobile phones, and distributed network access control.

Does this imply that VMA will one day absorb or seek peace with the RapidIO Trade Association, or even the PCI-SIG? Not necessarily, but such a marriage of convenience cannot be ruled out. The lesson to be learned from OIF, TCG and Broadband Forum is that the physical pinout and form factor, even the Layer 2 and 3 protocol details, do not matter nearly as much as bringing together engineering and programming groups who are exploring the same application space, looking for common methods that work across media types. Given that VITA wants to aid in the implementation of ANSI standards, it should look to the example of ITU-T study groups, such as SG 11 (covering signaling methods) and SG 15 (covering optical transport standards on lower layers). The study groups are seen as neutral realms where developers from IEEE and ANSI can sit down and break bread.

VMA need not overtly extend a peace pipe to any particular group like RTA or PCI-SIG. Rather, it should drive the hell out of VPX standards, while keeping an open mind about bringing others into the tents of common backplanes, high-speed serial and parallel communication options, as well as unique form factors to serve future embedded markets. The necessary elements of the equation are the McNealy Factors of common arrowhead and volumes shipped. The sufficient elements of a larger VMA involve being open to new members and new ideas, while being flexible enough to rapidly morph or expand its entire raison d’être.

**************************************

MontaVista and the binary ties that bind

November 16, 2009
By: lw

Our fearless leader Ray Alderman already has weighed in illustriously on reasons to fear the Intel-Wind River merger. On Nov. 11, Cavium Networks announced its acquisition of MontaVista Software, an event the open-source community might see as just as frightening, albeit at smaller levels. Now, the MontaVista brand has just as great a chance of standing as a separate entity as the Wind River brand (Good? Next to nil?), so the acquisition itself is not a cause for alarm. It’s just that M&A deals like these show that many corporations still do not grasp the necessary elements that must be preserved for open systems to work on both hardware and software levels.

Don’t get us wrong. When a large vertically integrated software house like Microsoft, Google or EMC-VMWare (maybe even Red Hat, I suppose) acquires a small software developer to add a missing module to a framework, the larger company is simply expanding its portfolio and playing the big-fish-eating-little-fish game. When a hardware OEM snaps up a company making a new type of product or a software house developing products unique for the hardware, a rational expansion has occurred.

But when a semiconductor specialist or OEM picks up an RTOS, development tool, firmware or communication-protocol company that develops products for several hardware binary interfaces, the acquiring company is playing a market elimination game. And when this involves companies who develop for Linux or open-source worlds, the problem is magnified.

Sometimes, an open-source company’s own strategy to monetize can doom an open product.  NextHop Technologies Inc., for example, commercialized the popular GateD routing software from the University of Michigan.  Large OEMs like Huawei ended up licensing GateD.  But when open-source routing failed to be profitable, NextHop focused on 802.11 Wi-Fi software and ended up being acquired by the relatively unkn0wn wireless network company U4EA Technologies.  We still haven’t learned what happened to all that open-source route code!

We have seen the Intel and Cavium strategies before, to be sure. Intel itself picked up Trillium Digital Systems at the end of the 1990s to corner a communication protocol market. Broadcom snapped up the popular LVL7 Systems software company, and NetPlane Systems was traded (batted around, actually) among Conexant, Mindspeed and Motorola Computer Systems for years. In all cases, the acquisition removed the effectiveness and breadth of the underlying software.

No, we’re not about to call for laws that limit the ability of companies to acquire players that offer software for several hardware binaries. That’s capitalism. But it’s important to point out that a true commitment to open systems means, at the very least, not trying to corner a market for a software company that works in several binary architectures. In general, keeping software truly open requires allowing as much software independence in the ISV community as possible.

**************************************

Reconfigurability: an empty promise?

November 2, 2009
By: lw

For board-level form factors and large IP cores, this is the best of times for FPGAs in open systems. But it is also a time of over-hype and under-delivery, particularly in the oft-promised realm of reconfigurability. Elements on the mezzanine are in place, not only in the realm of the very popular VITA 57 ANSI standard, but also in the less-certain Advanced Mezzanine Card realms of ATCA and MicroTCA. The month of October witnessed the addition of 32-bit MIPS cores to the Altera Corp. family, plus ARM and AMBA cores for Xilinx. But what is the end product of all those elements?

If we were to rely on the marketing history of board-level programmability over the last decade or two, we might believe that the long-promised era of “reconfigurability” has arrived. From the earliest days of SRAM-based FPGA cells in the late 1980s, reliance on complex high-end programmable logic was touted as eventually bringing us to the day when hardware would be characterized on boot-up. In essence, a configuration image stored in SRAM or flash memory would allow a device to take on a different identity at each reboot.

In 1990, Lattice Semiconductor Corp assured us that “in-system programmability” was right around the corner. A decade later, startups like Silicon Spice and Improv Systems promised us that the new generation of “ASIP” (Application-Specific Instruction Processors) would change our lives within months. Since then, the battlefield has been littered with the bones of Improv, Silicon Spice, Quicksilver, Adaptive Systems and Chameleon Systems. Isn’t this all beginning to sound like the mainstreaming of gallium arsenide process technology?

Now that FPGAs have met and exceeded many ASICs in gate count and complexity, and the arrival of FMC/AMC has simplified interfaces to board-level systems, a number of existing companies and new startups have used that “reconfigurability” marketing term again. There’s only one problem: system-level designers seem unusually reticent to actually use the programming tools that might buy them that advantage. As a result, some board-level startups in this field have tested the water, looked at some specific vertical markets and re-spun their “reconfigurable” architectures for a single vertical configurable application.  A good example emerged this week, as Stone Ridge Technology re-spun its RDX-11 for specific applications in gene sequencing.

Kevin Morris of FPGA and Structured ASIC Journal sees the same result from the addition of MIPS and ARM cores to FPGA architectures. In an Oct. 27 essay, Morris made comparisons to similar announcements Altera and Xilinx made ten years ago and said he expected few practical results to emerge from the licensees. In essence, popular RISC cores are added to portfolios for marketing reasons, but real FPGA users require a lot more hand-holding, and many more tools surrounding the cores.

Morris may have overstated the case, in that Xilinx’s licensing of ARM’s AMBA interconnect technology would suggest some on-chip communication value similar to HyperTransport or RapidIO. But his overall point is well taken and resembles the problem we’re seeing in the endless promises of reconfigurability. Yes, there may be OEMs who want a “soft base station,” a “soft handset,” or a “soft electronic warfare platform.” But if we look carefully at the issue of mezzanine FPGA interfaces, large IP cores in an FPGA block and reconfigurable memory structures, we come back to the same notion seen in the flash vs. EEPROM debates of yore. Many people talk about “true” reprogrammability or reconfigurability. When it comes down to designing configuration images or paying for devices, however, they end up opting for a special device that is configured once and then forgotten.

Will it always be so? Maybe this time, we are standing on the threshold of the reconfigurable era. Maybe this time, users will ask for ARM cores in their FPGAs, use FPGAs in an FMC form factor and demand true reconfigurability on reboot. But to get there, OEMs will have to offer goof-proof programming tools that use object-oriented methodologies, XML programming interfaces and Wizard-like graphic tools for dummies. Morris has pointed to a real problem, and that problem is similar to one we will find in FMC/AMC cards employing FPGAs. It’s a problem that can only be solved by stopping the over-promising in the industry, giving users simple tools and letting them discover the benefits of reconfigurability for themselves.

**************************************

Moving to optical connections at light speed

October, 2009
By: ra

For the past twenty years, we have been only three years away from optical backplanes and architectures taking over the computing industry. And for those twenty years, it never happened. Like the perpetuation of Moore’s Law, the signal engineers made Megahertz and Gigahertz signals run effectively on copper wires and copper connectors. But, as Moore’s Law nears the end of its ability to double the number on transistors in a given space on a die, copper may be nearing the end of its ability to reliably pass signals at higher frequencies.

It’s finally time we dropped copper connections and moved to optics. Why? Well, it just gets too hard to make copper work reliably at 10G and above, and it’s too expensive too. How do I know this? Because, the first two generations of PCI Express used 8b/10b signaling (at 2.5 GHz and 5 GHz respectively). In Gen-3 silicon, they moved to PAM (Pulse Amplitude Modulation) and 8 GHz so they could move the same amount of data as a 10G connection that used 8b/10b, but not have to deal with the 10G signal integrity problems. If you look back at the high-speed connection work over the years, you will see that every time the frequency of the signal doubled, the distance that signal could reliably travel was cut in half. The parallel PCI interface enhancements over time prove this point. So, all the advancement in backplane and cable signal frequencies over the years came at the cost of shorter and shorter runs. “Other than that, Mrs. Lincoln, how did you like the play?”

Another reason we have refrained from optical connections has been the limitations of the optical architectures. Optics do not like to be multidropped (like a bus). They do not like to be switched (like the old POTS telephone network). Optics like to run from one point to another, without any interference. Market acceptance of the high-speed serial fabrics (and the dying-off of the bus-thinker engineers) have removed that impediment. A number of the fabric-based machines are built with Meshes, which are point-to-point serial connections to and from each board in the rack. Pretty soon, the copper-thinker engineers will die-off, and we can make a wholesale move to optical.

Speed and signal integrity are not the only reasons to move to optical connections. Noise immunity is a “raison primaire” for going optical. Data can be moved in any electrically-noisy environment at GHz speeds without any concern for the ever-shrinking signal-to-noise ratio. In the Military applications, optical connections are immune to EMP (Electro Magnetic Pulse) from nuclear explosions or “pulse bombs”. But in those situations, you must use glass fiber: plastic fiber will undergo a crystalline change, turn opaque, and no longer pass light when exposed to EMP. Optical connections are lighter and more resilient to shock and vibration than copper connections. That is why we are seeing them used frequently in UGV (Unmanned Ground Vehicles) and avionics platforms (including Unmanned Aerial Vehicles and fighter planes). Anywhere weight is a factor, you find optical connections.

Many of VITA’s members have been marketing some optical connections over the years, under product names like “reflective memory”. The Holy Grail for optical connections, until now, was 10 Gigahertz Ethernet. But, Apple and Intel just announced a new optical concept, called “Light Peak”. LP also runs at 10 GHz, and is targeted to replace all the different cables now found in PCs, servers, and other computing devices (techresearch.intel.com/articles/None/1813.htm).

In the VITA Standards Organization we already have two efforts (VITA-66 and VITA-67) that are adding optical connections on VPX backplanes and XMC mezzanine cards with the existing copper connectors. Obviously, we need to sort-out the connector schemes and architectural implications for pure-optical VPX backplane systems. I think it is time we started moving to optical connections rapidly, especially in MIL/Aero and ground vehicle applications that use VPX.

We can’t wait for telecom to drive the optical revolution: telecoms are having trouble controlling their own bowel movements these days, and cannot drive anything except their journey to bankruptcy. So, it’s up to the Military applications segment to initiate this move away from copper and to a new generation of pure-optical architectures and technologies. And it’s up to VITA and its members to lead the way.

**************************************

Is it RapidIO or PCI Express? IDT will be pushing both.

October, 2009
By: lw

When Integrated Device Technology Inc. completed its acquisition of Tundra Semiconductor three months ago, many analysts assumed that a rationalization of the Tundra product line would mean that either RapidIO or PCI Express interfaces would fall victim to aggressive editing. So far, there appears to be no sign of slacking on either front.

To some who say that standard RapidIO and its Serial RIO equivalent seem to be in the biggest danger of losing market acceptance, IDT is launching a new RapidIO Gen2 suite of control and switch products through the end of the year and early 2010. It also announced on Oct. 6 that Freescale Semiconductor would be using the IDT CPS-10Q switch to link three MSC8156DSP processors in a new development platform.

Trevor Hiatt, product and applications manager for RapidIO products at IDT, said that even if opportunities in wireline communications were softening, the use of Serial RapidIO switches in VITA 41 and 46 boards, and in the new generation of 3G, LTE, and WiMAX wireless base stations, will assure the standard’s popularity.

The Gen2 spec is recognized in the industry for being faster than its predecessor (6.25-Gbaud per line) and supporting longer reach (100 cm across two connectors, which means RapidIO can be used across boards in a single system). What is less recognized, Hiatt said, is that Gen2 supports advanced Quality of Service methods, including the support of real-time streaming of isochronous traffic. RapidIO Gen2 specifies the creation of Virtual Channels, to allow for more dynamic allocation of bandwidth resources than was possible in RapidIO 1.3.

Where these features will come to the most immediate use is in applications such as peer-to-peer clusters in server farms, said senior product manager Devashish Paul. The QoS mechanisms allow for more efficient use of embedded processors in peer-to-peer designs.

In reality, speeds and capabilities of RapidIO Gen2 and PCI Express are tracking each other in parallel timely paths.  Choice of one serial standard over another largely boils down to personal preference of a designer.  PCIe can offer legacy system discovery for earlier generations of PCI, for example, while RapidIO offers no legacy support. At the same time, PCIe is limited to a tree topology, while RapidIO can be implemented in star, dual-star, or point-to-point topologies.

The bulk of high-speed serial switching product introductions in the next few months from IDT will revolve around RapidIO Gen2, though the company is not discussing specific device features or switch fanouts yet. But these launches will be followed relatively quickly by more PCIe products. In both realms IDT is expecting significant new design opportunities from China-based OEMs who have not been strong players in either vertical market to date.

**************************************

SPECWATCH
OpenVPX: boon or bust?

October, 2009
By: dl

The OpenVPX Working Group expects to turn a document over to VITA later this week, and it is scheduled to present that document to the world at a MILCOM 2009 press briefing on October 19.

Sources differ over whether the draft spec, reported to be between 400 and 600 pages long, will be the workhorse that carries VPX into a long and successful future or a camel that will only increase short-term confusion. According to the ad hoc Group, now over 30 members strong, OpenVPX is a “key milestone in [defining] system-level interoperability for…the VPX (VITA 46) architecture.”

The mission of the Group, which was initiated by Mercury Computer Systems in January, has several dimensions. The core of the issue is that the base VITA 46.0 specification takes a minimalist approach, defining mechanical standards and a few utility signals, while leaving it to auxiliary specifications (or “dot” specs) to fill in the particulars. And, as VITA’s John Rynearson explains, “Each dot specification defined a different pin configuration as if it was the only fabric that would be used.”  (See chart below.)

VITA 46 Connector Assignments

Source: OpenVPX Working Group

That’s a problem. Out in the real world, board and system designers are finding it advantageous to use more than one fabric, typically a high-speed data interface plus a separate control interface; and implementing two mutually exclusive fabrics requires a custom design. “Everyone has their own flavor of VPX,” sources lamented. “Everything’s custom.”

Moreover, the DOD has reportedly become very serious about interoperability, a concern that has trickled down from primes to subs and board, component and system companies big time. The DOD is, of course, the largest end customer for VPX-based equipment.

The great flexibility of VPX is causing interoperability problems in other realms. One reason for the variability in VPX implementations, according to Doug Patterson of Aitech Defense Systems, is that “it’s a case of technology driving a market vs a market demanding a technology.”

“In 3U, you can use half of the P1 connector for the data fabric and map some I/O from a mezzanine board to the other half,” says Bob Sullivan of Hybricon. “But guess what? If somebody maps something different into that space, they’re not interoperable.”

In VPX, that same P1 connector is also specified for use by VITA 46.3 Serial Rapid I/O (SRIO), 46.4 PCI Express, 46.6 Ethernet and other dot specs. Unless the same fabric resides at both ends of the point-to-point interface, or both devices understand the same pin mapping scheme, there’s going to be trouble. And there’s great variability in VPX systems overall. Systems differ in their fabric count, for example, as well as how those fabrics are woven together–in a host-slave organization, for example, a mesh, a ring, a switch, a star, a cube or some combination or hierarchy of these topologies.

One source characterized VPX’s interoperability problem as its “dirty little secret” and one which prevents customers from using off-the-shelf system components from multiple manufacturers and forces them to pay unexpected NRE. “We have six or eight ‘standard’ VPX backplanes,” says Sullivan, “but we must have built 50 customs!”

What’s in a profile?

OpenVPX attacks the VPX interoperability problem by defining a set of so-called “system-level” profiles. According to sources, the original intent was to define “a handful” of these profiles, but the draft spec contains far more than that.

“Nobody wanted their current VPX implementations to be orphaned, so they were all grandfathered in,” one source says. It has not yet been revealed exactly how many profiles OpenVPX defines, but sources put the number as high as 50.

“There will be even more ‘standard’ backplanes than before!” lamented one industry curmudgeon. “Each of the four major VPX card manufacturers will now have their own three or four pet backplane topologies that will be a bit different from the other three. They will each hope that their own favorite will have universal appeal but the end result could be twelve different ‘standards,’ each offered in at least two different slot counts.”

Doug Patterson dislikes the term “profiles” because of the association with Futurebus and its failure to get off the ground. But he thinks the profiles “go a long way towards solving things” by codifying practical alternatives for different kinds of systems in different application spaces. “People can go in [to the document] and readily see which profiles fit their medical application or military application and make a choice,” he says.

With “a set of fixed VPX implementations” (profiles) defined by the Group, Mark Littlefield of Curtiss-Wright Controls Embedded Computing says there is “a reasonably good expectation of interoperability.” Other sources noted, however, that the boon will not descend automatically, but will take some time to settle in.

The OpenVPX “framework” defines at least three different classes of profiles: backplane, slot and module. A backplane profile can accommodate different slot profiles and, likewise, a slot can accommodate different module profiles. Different backplane profiles are defined for different topologies: host-slave, for example, and both centralized and distributed switching. Homogeneous and heterogeneous computing environments are also reflected in different backplane profiles.

The slot profile as presented by OpenVPX is informed by the concepts of planes and pipes. “Planes” segment a system into functional environments. In addition to data and control planes, OpenVPX defines a management plane, a user I/O plane and an expansion plane. The expansion plane is intended to provide an alternate, secondary communications link that’s isolated from the main data plane, Littlefield explains.

OpenVPX defines at least three kinds of pipes, a fabric-agnostic term that refers to different size groups of differential pairs, These are the fat, thin and ultrathin pipes, containing 8, 4 and 2 differential pairs, respectively.

When the dust settles

No one has yet suggested that OpenVPX defines an ideal number of profiles, but at least “the dogies have been corralled” in that the VPX implementations of the leading companies have now all been documented. “At least people will be able to point to different topologies and pick one,” even the curmudgeon grants.

One source noted a fair amount of politicking in the early days of the OpenVPX effort, but another said this fell away as the committees got down to work. Neil Peterson of Hybricon calls the amount of time and energy put into OpenVPX by member companies “incredible.”

I’ve never seen anything like this, says Peterson, “and I’ve been in the industry for 20 years.” He notes a lack of “the normal BS and games being played… just competitors working towards a common goal.” With the completion of the spec, he says, “the competition’s back on.”

Several sources suggested that the completion of OpenVPX will inject new energy and purpose into the VITA 46 dot specs, which have progressed to completion rather slowly. Rynearson, for example, expects VITA 46.10 Rear Transition Modules, important to all the fabrics, to be done by the end of the year.  PCI Express, SRIO and Ethernet dot specs are also expected to be completed relatively soon, although the attention given to the OpenVPX effort over the past nine months has contributed to their delay.

It’s worth noting that no one expects the long slate of profiles defined by OpenVPX to last for very long. “People will build subsystems and then the marketplace will coalesce around certain aspects,” says Patterson, “as it always does.”

Rynearson believes that as many as four of the now standard VPX implementations will gain traction over the next couple of years, and then things should “dwindle to just a few choices,” he predicts.

The OpenVPX Working Group notes that user-defined pins are reserved in all of its slot profiles, which to a degree “limits full interoperability and interchangeability.” The rationale here, according to the Group, is that “full plug-and-play is considered less critical than customer and vendor differentiation to meet critical application, functional and SWaP [size, weight and power] requirements.”

**************************************

Optical interconnects pose future challenges

By: tc
September, 2009

Optical technologies have been on the verge of takeoff within computer systems for several years now, but optics still doesn’t play a significant role in most systems. That may change, however, as communication speeds exceed 10 Gbits/second. But if it does, the supply chain is going to have to undergo some major changes. The board manufacturers who provide the bare FR4 boards that are loaded up by CPU and other board suppliers, for example, will have to make dramatic changes in their factories to handle optical components.

“Handling optical is almost impossible for the average board company,” said Jack Fisher, chairman of the Roadmap Committee of the IPC, formerly known as the Institute for Interconnecting and Packaging Electronic Circuits. The IPC members who use the organization’s roadmap data to upgrade their factories play a key role in the board market, putting together laminate and copper layers to make bare boards. Optics “requires a big change in the business model,” said Fisher. “You’ve got to have a clean room without any dust, for example. The operating environment is significantly different for board assemblers and wave guide manufacturers.”

IPC statistics show that North American suppliers churned out $3.73 billion worth of bare boards in 2008. That’s a small percentage of the $40 billion produced worldwide, but it’s important domestically because military applications buy nearly one third of U.S. production.

The IPC expects to see more demand for optical near the end of the 10-year period they address in their projections. That’s also what they said to expect in previous projections, but it didn’t happen. Fisher noted that the International Electronics Manufacturing Initiative (iNEMI), a related consortium that focuses on the global electronics manufacturing supply chain, saw similar results in its latest roadmap project.

“None of the emulators [OEM customers] for IPC or iNEMI showed optical interconnects as a requirement,” Fisher said. “Copper always seems to get better and better.”

Having spent much of his career in research, Fisher keeps wondering when that’s going to change. He noted that tricks in semiconductors have helped overcome the limits of copper traces so far, but chipmakers find it increasingly difficult to reshape transmission to overcome limits like skin effect.

“Copper runs out of steam not because of copper costs but because of the processes silicon designers can use to support copper,” he explained.

Fisher also noted that a few other challenges face the overall electronics industry as it ponders a shift to optics. “Standards in optical are loose,” he noted. “There aren’t any trade associations or standards groups.”

There’s also another challenge facing the board industry. “There are difficulties for manufacturing,” he noted. “If you want to bury a waveguide, will it take the stresses of board manufacturing?” But while that’s a significant issue, he doesn’t see it as a show stopper.

Click here to see the IPC’s International Technology Roadmap for Electronic Interconnections (2008-2009).

Click here for a look at iNEMI’s research agenda.

**************************************

CONFERENCE REPORT:
SSIF shows expanding ambitions at IDF

September, 2009

By: lw

One of the most intriguing things to observe at the recent Intel Developer Forum was the expanded mission and intent of the Server System Infrastructure (SSI) Forum. Launched in 2000 under Intel Corp.’s auspices, the SSI Forum was tasked to find a common ground in power supplies and enclosures for corporate data-center servers. Now, however, it has taken form factor specifications and system-level blade specifications under its wing.

Kurt Lender, who heads  blade marketing and business development in Intel’s enterprise segment, said that the addition of full interoperability tests, possibly including an academic link for an external lab, will increase the Forum’s visibility. At the same time, the Forum does not want to get directly into bus architectures or electrical interfaces. This leaves it  as an organization that could challenge PICMG’s role in cPCI or ATCA – in theory. But in practice, Lender sees the corporate-oriented server blade world as occupying a different universe than a five-nines communication server.

“Our members are not as concerned as telcos with high reliability and specific board-level standards at the 1U or 3U level,” Lender said. “But they are interested in applying modular design methods to upgrading existing blade server architectures.”

The Forum has received switch-level intellectual property from IBM’s Blade Center group and now has such NIC heavy-hitter members as Neterion and H3C in high-speed Ethernet, and Mellanox and Voltaire in Infiniband. It has published specs for compute modules, mezzanine cards, chassis management modules and switch modules, as well as design guides for the midplane and for system management features.

The Forum brought a representative of Dawning Information Industry Co. Ltd. in China to the Intel Developer Forum to provide a testimonial on the use of SSI Forum specs to cut design times. Chaoqun Sha, product center general manager at Dawning, said his company’s previous server family had taken two years to design, while the move to the new TC2600 family, in contrast, took only a year. The open module approach not only allowed Dawning to design its own modular elements faster,he said, but it also allowed the company to contract out switch, chassis management and power supply functions to outside companies, based on common size and interface specs.

The Forum continues to treat its blade-center partners as a closed universe, at least in comparison to groups like VITA and PICMG. While the latter groups pledge to pay attention to global standards and look for opportunities to work with official standards organizations, SSI Forum still operates as a more informal community of interest. But its move at the end of this year to offer official compliance testing should make the fuzzy enterprise blade server world move a little closer to its equivalent groups in industrial and mil/aero realms.

**************************************

DISPLAYWATCH: OLED displays are on a slow roll

September, 2009

By: dl

OLED (organic light-emitting diode) displays, the new kid on the block in flat-panel displays, figure to play a larger and larger role in many applications over time. It’s no wonder. These are the first low-voltage emissive FPDs in history to reach market, they are inherently thin and lightweight, and they lay claim to higher efficiency than today’s king of the hill in FPDs: the AM LCD (active-matrix liquid-crystal display).

Following Sony’s introduction of the first OLED television back in December, 2008–a very expensive 11 incher—OLED displays have received more attention this year than in previous years, when they were limited mostly to car audio equipment, cellphones and a fair number of handheld gadgets touting a next-gen display technology with real pizzazz.

Recently,  the Zune HD has pitted an AM OLED against the AM LCD of the iPod (andof  the vast majority of handheld electronic products). The Zune MP3 player is reportedly on backlog, although it’s not outselling the iPod. Click here to see a trade show video of the HD OLED display and comparably sized iPhone LCD going head to head. www.oled-info.com/zune-hd-vs-iphone-display-quality

Most embedded and other markets today are well served by AM LCDs, which are well down the maturity, performance and cost curve and will remain the default choice for most small-, medium- and large-screen applications for some time to come. OLED displays are, however, on the rise, and at least one display vendor has pointed to 15 inchers coming for laptops by some time next year.

According to iSuppli Corporation, AM OLEDs are taking a bigger and bigger nibble out of applications now owned by AM LCDs. A recent report from the company indicates that the market for cellphone-class AM OLEDs, typically between 2 and 3 inches in diagonal, is “set to explode in the coming years.” The only thing holding them back, the company says, is a limited number of OLED suppliers with a limited number of factories, but this will change.

iSuppli predicts that the number of AM OLEDs used as the main display in cellphones will jump by a factor of eight from today’s 22.2 million displays (2% of all cellphone main displays) to 178 million displays in 2015 (over 6%), for a 41.4 % CAGR from 2009 to 2015, compared to 8% for handset displays as a whole.

In the meantime, though, OLED microdisplays are already becoming a favorite for HMD applications in some quarters, competing with LCOS (liquid-crystal-on-silicon) displays, laser+MEMS technologies and others. The DOD continues to fund an OLED night-vision goggles project at eMagin, an East Fishkill, NY, specialist in OLED-on-silicon microdisplays, with a new $3-million grant expected on top of the $2.4-million awarded in 2007, as reported by www.oled-info.com. www.oled-info.com/congress-approves-3m-emagin-further-develop-oled-night-vision-goggles

At this year’s SPIE Defense, Security and Sensing conference in April, eMagin demonstrated more than a dozen new systems from customers, including Intevac Photonics Inc., Elbit Systems of America, Liteye, Oasys, Thales and Qioptiq, as reported by Reuters (www.reuters.com/article/pressRelease/idUS151725+15-Apr-2009+BW20090415). Most of these systems are targeted at military, law enforcement and emergency personnel. A typical eMagin OLED microdisplay measures 0.77-inch on the diagonal, provides a 100-nit-bright virtual image in an 800 x 600- or 1280 x 1024-pixel full-color format and dissipates less than 200 mW under typical operating conditions. The size of the virtual image produced by such a display depends on the associated optics. To describe just a few of the products shown:

• Intevac’s offerings included Night Port, a night vision monocular whose low-light level viewing is “comparable to GEN III Night Vision Goggles,” the company says, “with the added benefit of having the ability to record the scene as video or still frames;” the I-Port EX, a display mounted on lightweight military-approved eyewear; and the iPort 50 binocular system, with a new eMagin SXGA integrated with Intevac’s ILROC (in-line reflective optical collimator).
• Elbit’s QuadEye, in turn, is a panoramic night-vision goggle with what the company claims are “advanced features and performance so pilots and aircrew can accomplish difficult night missions successfully and safely.” Besides a binocular field of view of 40°, QuadEye provides an additional 30° of monocular field of view to each side, “similar to the eye`s peripheral vision,” the company says, which “reduces the need and degree to which panning the head is needed when wearing goggles.” The company also demonstrated a long-range thermal imager based on an eMagin OLED.
• As for Oasys, that company is using OLEDs-on-silicon in a HMD system which “attaches directly to standard-issue military helmets and protective eyewear, is compatible with both new and legacy Thermal Weapon Sight (TWS) systems, is thin enough to fit behind the blast visor of the cupola protective ensemble, and is backwards compatible with previous versions of the system.” Up to 20,000 of these HMDs are slated to go to the U.S. Army for the TWS Remote Viewer program.

Despite OLED displays’ clear appeal for near-eye microdisplay applications, as well as for the high end of  cellphones and MP3 players, it will take some time before they’ll hit big in midsize- and large-screen applications where cost is a concern. Perhaps their greatest day awaits the successful commercialization of flexible AM OLEDs, which will attract real attention, especially where display breakage is a major concern. Stay tuned, too, for OLED lighting, expected to make a play for markets in specialty lighting and backlighting applications in the very near term.

**************************************

FPGA-centrism feeds open inventories

September, 2009

By: lw

At first glance, Pentek’s debut of a data-acquisition module based on dual FPGAs and four different bus architectures would seem to underscore the case for my previous editorial warning of the death of DSP in MIL/Aero platforms. Not so fast. What Pentek did with its 7158 architecture could be implemented in DSP or integer processors, as well. The real revolution going on is one involving inventory and its relation to open buses, open chips and open software.

To put it simply, Pentek has designed Xilinx Virtex-5 FPGAs in an asymmetric fashion, dedicating one to control-plane duties and one to DSP algorithms. Designers can choose from three different Virtex-5 devices, including one with an embedded PowerPC for faster signals-intelligence acquisition. Those FPGAs are programmed with pre-assigned algorithmic library elements from Pentek’s GateFlow library. To top it all off, the FPGA-based design has real legs and hardware independence: Pentek can place the thing on a module designed for PMC/XMC, PCI, PCI Express, Compact PCI or, conceivably, any other particular bus or fabric architecture in which a customer has an interest.

Block diagram

The customer benefits of such FPGA-centric design here are threefold: flexibility of architecture, the ability to rapidly add and upgrade algorithms and the ability to add high-level digital filtering blocks without tangling with lines of code. While some of these advantages exist in extant DSP tools, the way Pentek has put them together will no doubt be emulated by others in days to come.

That’s because the advantages are not limited to the customer. The OEM can design a subsystem once, order specific types of FPGA from a vendor like Xilinx on a JIT basis and then place them on particular bus-based modules, depending on what is in highest demand. Proponents of open-bus architectures and open-source software have been touting this vision for decades. Now we’re seeing it come to fruition and extend all the way to the level of the individual chip.

**************************************

Seeking the next big opportunities in embedded computing

September, 2009

By: Ray Alderman

Company CEOs and VPs are always asking: What are the next big opportunities that will drive our industry? There are a few strong candidates, as well as some probables and a few dark horse possibilities. Let’s consider five of the next potential “big things” coming to the embedded board and systems industry in the next few years:

• unmanned military platforms,
• the air traffic control system,
• the smart power grid,
• home medical monitoring equipment and
• telecommunications.

Let’s take a brief look at each of these sectors.

1. Unmanned Military Platforms

The Unmanned Military Platform market has been developing for many years, and is now taking off like a rocket. Our military is automating warfare at a rapid pace. This market includes

• UAVs (unmanned aerial vehicles) like Global Hawk, Predator, Reaper, etc.,
• UGVs (unmanned ground vehicles) like MRAP (mine resistant ambush protected),
• robotic soldiers and unmanned robotic soldiers like Warrior and Talon, and
• UUVs (unmanned underwater and surface vehicles) like the Spartan Scout and REMUS (remote environmental measuring units).

In his recent book, Wired For War, P. W. Singer notes that “the number of unmanned ground systems in Iraq and Afghanistan went from almost zero to five thousand by the end of 2006 and was targeted to reach twelve thousand by the end of 2008.”

He further states that between June 2005 and June 2006, Predators conducted “2,073 missions, flew 33,833 hours, surveyed 18,490 targets, and participated in 242 separate raids.”

One look at the DOD budget (now in Congress) will show the massive funding for the platforms mentioned above, as well as new ones like UCAV (unmanned combat aerial vehicle), the unmanned fighter jet. The UCAS-N (unmanned combat air system – Navy) is the Navy’s version of that. Robotic combat medics now being developed for the battlefield will find wounded troops, drag them into a protective container and transport them back to safety while simultaneously starting trauma treatment for their wounds.

The UUVs being developed are not only being crafted as weapons platforms (i.e., smart torpedoes), but also to duplicate the mission of the UAVs, monitoring maritime traffic and gathering intelligence information while remaining undetected by the enemy.

This market segment will grow rapidly for many years to come, as we deploy many thousands of these systems every year. Also, you can expect to see a rapid pace of upgrades to these platforms over their operational lives—-turning this segment into a good long-term opportunity, as well as a near-term boost. The first upgrades for the existing Global Hawk UAVs were awarded in June ‘09.

2. The Air Traffic Control System (ATCS)

Before the recession hit, the ATCS was overloaded. Airlines, with expectations of increasing air travel, leased new planes and opened new routes at a record pace, clogging the old and antiquated system with flights. Since 2008 and the declining economic conditions, airlines have done a turnaround, eliminating unprofitable routes, reducing flights to many destinations and pulling planes out of their fleets. (Most of these mothballed planes are parked in the desert south of Phoenix).

The pilot is always responsible for his aircraft’s distance from the ground. The flight tower ATCS and air traffic controllers are responsible for the distance between that aircraft and other aircraft in the air. The IFR (Instrument Flight Rules) of today’s ATCS requires pilots to fly on predetermined routes in the sky, from IFR point to IFR point–typically not a fuel efficient way to travel. Centralized IFR also directs aircraft to increase/decrease altitude to avoid collisions at the intersections of those routes.

A more distributed approach would allow all aircraft to fly direct routes to their destinations, saving fuel and time. Embedded computers with a GPS (Global Positioning System) on board could easily automate this function–an opportunity for computers that can meet the ruggedness and reliability requirements. Automating some safety issues can also be effectively done on board aircraft, and could actually increase safety without requiring human intervention to avoid collisions.

At the current time, as previously mentioned, the ATCS is well within its capacity, which will reduce the urgency to revise and revamp for a number of years. An economic comeback, however, with an increase in travel would be fertile ground for airworthy computer boxes and boards, plus new gear in the flight tower to interact with it.

3. The Smart Power Grid

With the present power grid, it seems that a tree falling on a power line in Oregon can shut-off all power to the State of Texas for a week. The grid is overloaded in the West in summer, and in the Northeast in winter. Installing millions of embedded computers in homes, substations, and power generation facilities will make the grid more efficient, more predictable, and more easily managed, or so the theory goes. But what’s the opportunity?

The present transformer and generation equipment makers (GE and Emerson) both have embedded computing divisions, so this market could be dicey for board makers on the outside. Nonetheless, the smart grid revamp is underway, and it will need millions of embedded computing devices.

4. Home Medical Monitoring Equipment (Homedics)

People with chronic conditions sometimes go to the doctor several times per week for basic tests, running-up insurance bills. What if they had a medical station in their home, connected to the Internet? They could come to that station, hook themselves up to an EKG pad, put on a blood pressure cuff, and plug-in their glucose meter. The data would be extracted and sent to a server at their doctor’s office and compared to a baseline established by their physician for a monthly charge that costs far less than all those doctor visits.

GE Medical is already involved in this type of activity, relying on its own internal embedded computing group. The other traditional medical equipment vendors have either their own design and manufacturing groups or strong relationships with contract electronic manufacturers. If the concept of the home medical station takes hold, however, there could be some business here for focused embedded board vendors.

5. Telecommunications

Despite much hype to the contrary, there just aren’t any major opportunities in telecom, and there won’t be until consumers start spending on frivolous telecom services and devices again. By 2013, the market for cellphones is predicted to saturate in all developed nations. The same goes for cellular services.

It’s a sign of the times that Verizon Wireless apparently sees its future as dim unless it can find a way to sell the bandwidth of its existing cellular networks. The company is, thus, visiting all the power utilities and trying to convince them to use the Verizon network to control all those millions of embedded computers that will go into the smart grid outlined above, instead of installing their own RF networks.

While this might make some economic sense, it would involve forming an alliance between the telcos (some of whose executives are either in jail or under indictment for financial fraud) and the power industry executives (some of whom have a state-granted monopoly and continue to demand rate increases on consumers). Strange bedfellows, indeed!

Moreover, there would be tremendous security problems involved in using the Verizon open network. Open networks get hacked. Hackers could get into an open smart grid, shut down nuclear reactors and shut off power to cities before a terrorist attack. They can already get into the commodity telecom systems and shut those down. Note that Russian hackers shut down all the telecom systems and government computers just before they invaded Georgia last year. Building the smart grid on an open network is clearly not in the cards.

**************************************

Seeking directions for driverless technologies

August, 2009

By: tc

Unmanned vehicles are one of the hottest growth areas in the military world these days. Unmanned aerial vehicles, for example, are transforming war zones, providing significant amounts of intelligence and then acting on it “with extreme prejudice,” to use a great euphemism for a lethal strike. On the ground, sophisticated unmanned rovers are saving plenty of lives and gathering tons of information.

These driverless vehicles already provide a hot growth market for many board suppliers. What’s more, as military technology trickles down into other application areas, driverless vehicles will provide a real opportunity for those with relevant intellectual property that can easily be adapted for new venues.

Autonomous vehicle techniques are seeing growth in off-highway markets, especially construction and agriculture. Off-highway equipment designers have the benefit of being able to tap into existing know-how and well proven hardware. The technologies and IP used to navigate military aircraft or land rovers, for example, could aptly be applied to the huge ore-carrying mining trucks which continually traverse the same path. They’d be equally at home in tractors that go up and down rows. Both applications are already well along the path to autonomy.

The market numbers for these and similar autonomous-vehicle applications may not be huge, but the ruggedness requirements and similarity of the technologies used could make off-highway equipment an easy market to attack for military UAV suppliers.

**************************************

Wind River becomes Intel. Watch out below!

August, 2009

By: ra

The question of why Intel bought Wind River Systems has come across my desk numerous times since the acquisition announcement back in June ‘09 (www.intel.com/pressroom/archive/releases/20090604corp.htm). My desk is also littered with all the theories suggested for why this event occurred.

Some speculate, for example, that Intel bought Wind River Systems because it wants to get into the military business, but this makes no sense for several reasons. First, the volume of microprocessors used in MIL applications is very low relative to other market segments, especially consumer electronics. Further, Intel has a history of avoiding any direct responsibility for critical systems that could injure, maim or kill people if they fail. (They leave that market, and those liabilities, to the third parties who use their chips.) The same story goes for the medical equipment markets, too.

Others speculate that the rationale behind the acquisition was that Intel wants Wind River’s customer base, but this makes no sense, either. WRS has a very scattered customer base, which cuts across legacy industrial apps, military apps (mostly based on PowerPC processors), medical apps and some apps in consumer product segments such as smart phones and netbooks using ARM and MIPS processors. For the likes of Intel, industrial is too fragmented, military is too small, and medical is both small and fraught with liabilities. So, by elimination, it would seem that smart phones and netbooks would be the only attraction here under this theory.

Everything Intel has done over the past decade to enter embedded markets has been a largely ineffectual passive program, attacking these markets through third-party product makers. Perhaps the WRS purchase means that the company intends to take a more active role in embedded computing, as it did (intermittently) in the ‘70s, ‘80s and ‘90s.

At the announcement, Intel said that WRS “will become part of [our] strategy to grow [our] processor and software presence outside the traditional PC and server market segments into embedded systems and mobile handheld devices.” According to Renee James, Intel vice president and general manager of the company’s Software and Services Group, the WRS acquisition will bring “complementary, market-leading software assets and an incredibly talented group of people to help us continue to grow our embedded systems and mobile device capabilities.”

Intel proclaimed at the announcement that embedded systems and mobile devices are important growth areas, pointing specifically to “smart phones, mobile Internet devices, other consumer electronics (CE) devices, in-car ‘info-tainment’ systems and other automotive areas, networking equipment, aerospace and defense, energy and thousands of other devices.” Labeling this a “multi-billion dollar market opportunity,” the company noted that it is “increasingly becoming connected and more intelligent, requiring supporting applications and services as well as full Internet functionality”–and thus, no doubt, also increasingly in need of Intel silicon.

Intel could have ridden on the cellphone/netbook software infrastructure to support its chips in these applications but that arena is getting more fragmented every day: Nokia/Symbian, Google/Android, Mobilan/Lunix, and Microsoft, for example. Riding this infrastructure would be just another passive third-party approach. All these operating systems are running on cheap microcontrollers from small but successful companies (ARM, MIPS, Renesas, etc.). With VxWorks in their product lineup, Intel controls both the silicon and the API.

Intel completely missed the boat on cellphones and smart phones. As PC and server sales decline, and those markets become mature, they must take a direct active approach, not a third-party passive approach, to smart phones and netbooks.

One open question is how long Intel will continue in its enthusiasm for embedded markets. In the past, it created its own board product divisions (Multibus, Multibus II, ATCA, motherboards) and then sold them off after a few years. The company also spent billions of dollars buying telecom board and chip companies (Ziatech, Dialogic, DEC’s embedded processor fab and X-scale architecture, etc.), only to sell them off piecemeal for pennies on the dollar a few years later. Will they have staying power this time around? Who knows.

UNANSWERED QUESTIONS

Beyond the question of why Intel bought WRS lie a number of far more important questions. Whatever Intel’s motivation, its acquisition of WRS is a concern for embedded board makers. Many of our applications use VxWorks, especially in MIL/COTS. Will Freescale (PowerPC) and IBM (PowerPC) feel comfortable sending their first-silicon of a new processor to an Intel division for porting of VxWorks? Will MIPS and ARM feel comfortable sending their first silicon for a port? I have my doubts.

One also wonders whether Intel will continue WRS’s old business model of selling tool seats and run-time licenses. I have my doubts about this, too. Wind River sold out because that old business model was failing. Note that Intel bought the firm for 2.46 times sales. In 2004, Harmon Automotive bought QNX Software Systems for 5 times sales. Compare the two and it’s clear the price for WRS was a firesale price.

Does Intel want to sell software development tools and licenses, perpetuate that antiquated business model, and then assume the heavy support burden for all that software? No. Intel just wants to sell silicon.

Other open questions include, How tightly will Intel marry VxWorks to the Atom processor? And how long will they support ports to other CPUs such as PowerPC, MIPS and ARM? I believe that Intel will integrate VxWorks so tightly into the Atom architecture that it will take a crowbar to pry them apart.

At the announcement, Ken Klein–WRS chairman, president and CEO–noted that as a wholly owned subsidiary, Wind River “will more tightly align its software expertise to Intel’s platforms to speed the pace of progress and software innovation.” You bet!