Embedded Processor Watch

Editor: Cary D. Snyder

Contributors to this issue: Max Baron, Cary D. Snyder, and Marlene Bourne

In This Issue:

• Special Note: EPW Schedule
• EPF 2001 Flash
• Editorial: Some Assembly Required
• Narrowed I/O Bus Options Expand
• MEMS Industry to Undergo Major Changes
  

Special Note: EPW Schedule

EPW issue #135 hung up in cyberspace, causing a delay, but this issue should put it back on schedule now. A lot is happening around the offices of MicroDesign Resources, especially with Embedded Processor Forum (EPF 2001) a little over a week away.

I imagine most of you are like me in being totally overloaded with making the best out of all the new technology. The new electronics development wheel is one thing that hasn't slowed in the dot com meltdown. Case in point are all the new announcements and products making their grand debut at EPF the week of June 11th at the San Jose Fairmont. EPF is a series of events, chock-full of technical detail with direct access to architects and engineers that normally stay hidden in the background.

There is so much happening during the week that lots of good information won't make it into EPW or MPR for some time. EPF includes a three-day technical conference with sessions on network processors, processor cores, and embedded products. There is also a EPF Expo and two full-day seminar on Information Appliances, DSP in Communications, Network Processors, and Digital Audio.

The neat part about most of these events is the large number of brand-new technology announcements and product rollouts. For example, the Network Processor Forum will focus on new devices from XStream Logic and Lexra that achieve speeds of up to 10Gbps, and a new scalable architecture from Cognigine handles packet processing speeds of 40Gbps and beyond. In the same session, Brecis will be disclosing the first details of its multi service processor, which makes innovative use of DMA, clocking, and queuing to handle multi-service applications.

The afternoon session will focus on specialized silicon for networking, like Hi/fn’s latest gigabit security processor, Mapletree Networks’ new MTN3000 series access processor, and PMC-Sierra’s RM9000x2, the first multi-core processor from its MIPS processor division (formerly QED). In a demonstration of the high-speed interconnect technology essential to faster networks of Infinibridge, Mellanox Technologies will introduce a integrated 10Gbps Infiniband channel adapter and switch. There will be other first time disclosures from Motorola, LSI Logic, Zucotto and STMicro. TriMedia Technologies will mark its first EPF appearance since spinning off from Philips with the unveiling of system development details for its TM32 VLIW core. Corrent Corp has selected the first annual Network Processor Forum to make the first public disclosure of its new, full-duplex OC-24 IPSec security processor, the CR7120. MIPS, ARC Cores, ARM, Tensilica and others are giving key updates and additions to their product portfolios.

I could go on describing every that will appear at EPF 2001, but words alone can't express the high level technical content given out in these oral presentations; especially when the audience comes prepared with difficult questions for the industries best architects and engineers. It is a great opportunity to get that critical information you need for that project you're working on. I know, as I count on this myself. I look forward to meeting EPW subscribers at EPF to share more information.

— Cary D. Snyder, Editor Embedded Processor Watch

Embedded Processor Forum 2001 News Flash

We have more than one surprise planned for you at this year’s Embedded Processor Forum!

Let’s see how your swing measures up to mine.

This year’s MDR Embedded Processor Expo will include a demonstration of the Swing Solutions GVA500 Golf Video Analyzer—the most advanced portable golf training system in the world. (Yes, but will it help my golf game?)

We all can have a look at this intriguing application for embedded processor technology at the close of our afternoon sessions on Tuesday, and on Wednesday night you’ll have a chance to beat me in a game of golf. Fore!

For registration info, Complete program details and fast on-line registration, see www.mdronline.com/epf, or call us toll-free at 800.527.0288 (408.328.3900 outside North America). Tell them Cary told you to come over and have some fun while working hard.

Some Assembly Required
By Max Baron {5/29/01-02}

Most of us have had the pleasure, and sometimes frustration, of purchasing an inexpensive piece of furniture and assembling it at home. You have to be just a bit mechanically inclined: most of the time the hardest task will require you to tighten a screw or tap a wooden doohickey into place without hitting your thumb too many times. You are helping the manufacturer reduce the cost of the product. The factory delivers a package of furniture that, unassembled, is more compact and therefore less expensive to ship. You, the customer, are virtually helping transport the furniture; your work is also reducing the cost of labor and use of space at the factory.

Most assembly directions are simple; your work is soon done. You dispose of the cardboard and plastic packaging and masterfully discard the two pages of drawings and directions. You look proudly at the completed product, the saved money safe in your pocket.

Some digital electronics manufacturers may have learned the wrong lesson from this model. They are not trying to lower your price. They are rushing to ship products at the beginnings of market windows, to compete with the others, to increase their revenue. The products turn out to be incomplete and may still have unresolved bugs. Whether caused by greed, a "don't care" attitude, product complexity, or all of the above, a few manufacturers are introducing products that need far more than a screwdriver and hammer to assemble. The products may not even work without a call to the product-support group and some serious computer effort. Manufacturers' names withheld, here are some examples of products that target the consumer market:

A reputable digital camera manufacturer has posted on its Web site bug fixes for the firmware of an $800 camera. Can you see how all the digital camera owners would find out about the bugs before they leave for that big expensive vacation? Search the Web, load the new software, and boot it without problems?

Another manufacturer has released cameras from which it was difficult to extract the battery. Yet another digital camera manufacturer's camera-to-computer transfer software warns you that upon installation, the software will disable some of the dynamic linked library files (DLL) on your system and, to make life more interesting, fails to tell you which files will be removed. Should it go ahead? Oh sure! Please click the "yes" button, or, better yet, give it an axe; it'll be easier and faster! Makes you really anxious for the next batch of fixes—after the reputable manufacturer's gurus learn to program under the operating system. But as far as the consumer is concerned, by the time all the fixes are in, the camera will probably have been leapfrogged several times by new models, none of them truly complete.

A reputable answering machine provider has shipped a flash-memory-based "no-moving parts" unit that wipes out its boot section if power goes off during the time the machine is using its memory. The consumer is left with a silent, "broken" answering machine. Solution? Buy another one. Buy several if you live in California.

A well-known home security vendor sells a color video camera and a VCR controller. A motion sensor can simultaneously turn on the camera and VCR. Sound good? Just wait until you find out that the VCR controller has four buttons that must be programmed in something close to Morse code! But there's more fun: depending on how persevering you are, 15 minutes or two hours later, you'll give up and call support. The nice person there will tell you that for best results, you should use only two of the buttons (forget the other features, ha, ha) and you'll be sent an email with the new, debugged reduced instruction-set (RISC?) instructions. Or is it reduced button-set instructions?

Partial functionality and post-purchase bug fixes may seem natural to many of us. We buy software. We search the Web for the latest install and driver software for a hard disk, Zip drive, or graphics board. I suspect some of us even enjoy it. We pay for licenses for partly debugged software. Some software won't even work until you call support to get a patch via email. Bug fixes dressed up as new revisions abound. Look for revisions that read like this: Rev 4.01. One of the most prestigious companies makes you download compatibility fixes if you want a development tool that you licensed today to work with the one you acquired from it just a few weeks ago. A well-known software collection of bug fixes, compatibility patches, and additions has reached 133MB. Download this!

But the work that some of us may even enjoy is a major disincentive for those who express their feelings by saying that their four-year-old programs the VCR. Many products that manufacturers are building today are intended to engage people that until now have not purchased a computer. These people are expected to assemble partly functioning products with an unfamiliar screwdriver: a computer. And it's not enough merely to have a computer. For incomplete products, they must also be computer literate—and beyond. They must look for parts of hardware-software products on the Web, install them, and boot the new fixes at their own risk. When they call for support, in some cases they are put on hold for 20-30 minutes, and then they don't always get an answer to their problem. The manufacturer may have just licensed the product from some other source and hasn't had the time or training to debug and support it adequately.

Sound engineering and good business minds have invented hardware-software co-design to reduce time-to-market and product-development risk. We need to believe that these idealists were hoping to create complete, working products. Some vendors, however, are paralleling co-design and debug with sales. Co-design and co-sell with resulting co-disappointment and return-for-refund. Are they really expecting the consumer to both assemble the product and help the factory debug it? How can this approach help sell products into the broad consumer market, even in good-time economies?

It may be argued that very complex embedded systems are impossible to completely debug before shipment. But an embedded system should work as promised when taken out of the box and powered up. Programmable logic devices (PLD) and flash memory can be used to fix the one or two remaining non-important bugs as well as to upgrade the product. An ideal system, when first turned on, should not require a computer. It should periodically use the telephone or broadband connectivity to look for upgrades on the Web in order to deliver the best headache-free experience to its owner.

The industry is still missing a large portion of the infrastructure and attitude required to realize the dream: support for embedded Internet and self-sufficient systems that doesn't need a computer and expert human intervention. The industry needs standards of quality, guidelines for user interfaces, and failure-proof design—like a boot section in memory that doesn't get wiped out when you upgrade the firmware. And the consumer has to be comfortable paying the extra cost. Until then, more assembly is required. At the factory. - Max Baron

Narrowed I/O Options Expand
Part 1: Is HyperTransport and RapidIO Enough?
By Cary D. Snyder {5/29/01-03}

The article explores the technical advantages and weaknesses of common high-bandwidth on-board or internal interconnect schemes. The backers of HyperTransport, RapidIO, and the yet-to-appear third-generation I/O (3GIO) are all pushing narrow I/O buses as the chip-to-chip, internal in-the-box standard to replace or supplement PCI. The push results from the need to increase I/O interface efficiency in pin count and power consumption. API Networks Inc. has announced it will soon begin shipping its AP 1011 HyperTransport-to-PCI bridge chip. However, the AMD chip set that uses HyperTransport Technology may be delayed, together with the next-generation processor.

Xilinx has announced production release of the first physical layer RapidIO core. IBM has announced it will soon have next-generation 1GHz PowerPC chips with integrated RapidIO interfaces. This is good news to Motorola, which has been promoting RapidIO from the beginning as the narrow I/O standard that can easily be adapted to support other physical interfaces demonstrated by the rapidity with which a Serial RapidIO physical standard was created.

A bigger push is coming from Intel, which has promised a 3GIO specification by fall 2001, claiming that this specification will have longer-lasting attributes than those of other proposals. It's unclear how the new 3GIO technology presented and backed by Intel will affect HyperTransport's or RapidIO's acceptance (see MPR 3/26/01-03, "Intel Undercuts HyperTransport".

Both HyperTransport and RapidIO are positioned to replace PCI as a chip interface, but neither yet addresses a low-cost connector or attributes that made PCI a ubiquitous I/O standard (see MPR 5/08/00-01, "RapidIO Expands Narrow-Bus Options").

(The full version of this article is available online to Microprocessor Report subscribers at http://www.mdronline.com/mpr/h/2001/0529/152203.html)

MEMS Industry to Undergo Major Changes
By Marlene Bourne {5/21/01-01}

As MEMS (microelectromechanical systems) enter their second decade of commercialization, the industry is about to take on an entirely new look, owing to significant levels of venture capital funding, increased collaboration among companies, and the emergence of new products and applications.

MEMS are chip-level devices that can either sense or control the physical environment. Typically made out of silicon, most have moving parts. Created using various micromachining processes, MEMS use a number of manufacturing steps derived from basic integrated circuit (IC) techniques. As a result, they can be fabricated in batches; in some instances, up to several thousand devices can be made on a single wafer. Consequently, under high-volume circumstances, they can be very low in cost.

While the scope and uses of MEMS are vast, ranging from medicine and transportation to industrial and consumer markets, only a few products have found large-scale commercial success to date. As a result, some people have begun to regard MEMS as a flash-in-the-pan technology. No one disputes MEMS' possibilities, but, to this point, the technology has come up short with regard to commercial deployment beyond a few niche markets. Over the next two years, an incredibly diverse array of products will move into volume production, and the impact will be far-reaching. Not only are current markets expanding, but new ones are emerging. This industry will therefore be very different in five years' time.

Currently, the MEMS market largely consists of sensors having an average selling price (ASP) of less than $20. Some, such as pressure sensors for use in blood pressure monitoring, cost about $1. Others, like accelerometers for use in airbag deployment, cost $34. In both cases, we're talking about unit volumes in the tens of millions. But these are the exceptions. Most sensor applications require volumes of less than 100,000 and have a higher price.

Please visit http://www.instat.com/pr/2001/ea0102mf_pr.htm to view a related graph.

By 2005, actuators, not sensors, will account for the lion's share of sales. Some, such as biochips for use in drug discovery and micorelays for use in cell phones, will see unit sales in the hundreds of millions and will have an ASP of around $5. But most will be much more costly. The best example is the use of mirror arrays in applications like optical networking and projection systems. These are complex and costly products having an ASP greater than $1,000.

Mirror arrays for use in photonic switches are currently being priced at about $1,000 per port. For example, a 1 x 2 array costs about $1,000, with larger arrays, such as 8 x 8s, exceeding $8,000. Despite the fact that unit sales of these devices as a whole will probably reach no more than hundreds of thousands, their price will make a significant difference to the overall MEMS market.

As a result, worldwide revenues for MEMS will almost quadruple by 2005, to nearly $12 billion. In addition, the current market share ratio (79% sensors/21% actuators) will abruptly about-face, resulting in a market share split of 63% actuators/37% sensors by 2005.

End-use markets will also change dramatically, moving from those that are primarily sensor driven to those that provide much more opportunity for actuators. In 2000, sales of MEMS to the commercial/industrial market accounted for nearly half of all revenues, with automotive sensor applications leading.

By 2005, MEMS revenues will be almost evenly distributed among three markets: consumer, commercial/industrial, and telecom. The consumer and telecom sectors will become especially strong over the next few years as a variety of new products for high-volume applications go into production. This change is significant, given that telecom accounted for less than 1% of the total market in 2000.

Add to this situation the tremendous amounts of venture capital being showered upon MEMS companies, a particularly encouraging sign for MEMS as a whole. It appears that after a decade of struggling to take its place as a breakout technology, MEMS is finally coming of age.

Marlene Bourne is a senior analyst with Cahners In-Stat Group. She can be
reached at mbourne@instat.com.

Cahners In-Stat Group (http://www.instat.com) covers the full spectrum of digital communications research from vendor to end-user, providing the analysis and perspective that allows technology vendors and service providers worldwide to make more informed business decisions.