As production volumes grow and competition increases in the IC and electronics industries, switching is becoming an increasingly critical component for the automated test equipment (ATE) system makers. Simply put, test costs are growing and therefore produce a negative affect for the profitability of manufacturers. MEMS based switches are now becoming a viable and advantageous alternative to traditional technologies such as electromechanical relays, reed relays and semiconductor switches. We recently spoke with Dr. Dan Hyman, a leading industry expert and President of XCOM Wireless, about the current state of the ATE industry and his expectations for the future. In this comprehensive interview, Dr. Hyman presents his views on the current status of ATE switching including main industry trends, device costs as well as upcoming technology developments.
MEMS Investor Journal: What is your view of the current state of the ATE business?
Dr. Dan Hyman: The ATE business overall is highly fragmented, specialized, and lagging behind the recovery of their own customers after the recession. Fortunately, the wireless and high-speed digital ATE market segments have not seen the same level of slow-down as compared to the ATE industry as a whole in recent years, although they have suffered a real flattening of growth that is only recently on the uptick again.
Meanwhile, the continual increase in semiconductor and wireless test costs (as a percentage of product cost) is testament to the fact that this testing infrastructure has not kept pace with Moore’s Law or the wireless revolution. We anticipate that the newly recovering consumer confidence will give the industry the impetus to invest in emerging technologies and get back on track to control testing costs and improve uptime.
MEMS Investor Journal: What do you see as the current technology trends and upcoming technology milestones in 2011?
Dr. Dan Hyman: The test industry as a whole is bouncing back strong, so I foresee many players investing in new technologies at many levels. Software reconfigureability is probably the most important technology trend in the industry, allowing system manufacturers and users to squeeze much more function and cross-platform value out of each new system. Most players in the industry are cash-flush, so I expect to see continued investment in new infrastructure that can be redeployed to support multiple product lines without the large amount of development and equipment overhead that used to constrain the industry. We are glad to be on the reconfigurable hardware side of that movement, as load boards and switching subsystems need to be adaptive and programmable to keep up with emerging systems.
The other big trend related to software reconfigurability is continued movement towards having one large system driving many more load boards and workstations. To enable the next step in this trend, the load boards are already becoming much smarter and more capable of a variety of functions. The miniaturization of the control electronics has made this possible, and engineers at the end product companies are more engaged in driving innovation at this operational level. This trend means component developers like XCOM need to be more engaged at a technical level with the end users of ATE systems. This is also where more risks can be taken and new technologies adopted at a faster rate so their firms and emerging product lines can get (and stay) ahead!
MEMS Investor Journal: What are the current non-MEMS technologies used for switching? How do these technologies compare to each other?
Dr. Dan Hyman: I can speak to the component markets of load boards and other switching subsystems, rather than the system-level markets. There are several technologies which dominate the marketplace due to their long histories and, in some cases, a commoditized price point.
Electromechanical relays are usually considered some of the highest quality components available in the market, and are used primarily where high repeatability is needed, such as analyzing low-power RF. These components have a millimeter-scale armature that engages or disengages metal contacts, and are powered by electromagnetic actuators pulsed with drive current. These parts are large, slow, power-hungry, and have a short lifespan, but they are extremely repeatable and reliable. Rotational coaxial versions of these relays can have extremely low loss and wide bandwidth, but they can get very expensive, and replacing these components when they wear out is often a major factor in system downtime and overall cost of RF and high-speed digital product testing.
Reed relays are inexpensive “little brothers” of electromechanical relays, having achieved commodity component price points in the past decade. They can be moderately compact, moderately fast, and hit reasonable operating specifications, so they tend to dominate load board products in high-speed digital ATE applications. This is especially true for testing consumer products, where every penny counts, and many load boards are typically run simultaneously from a single (comparatively expensive) ATE system in order to lower test costs. Reed relays suffer from poor RF performance, however, as well as poor lifetime, but the low cost and high operational reliability keep them attractive.
Semiconductor switching components of a variety of technologies (Si, SiGe, GaAs, etc.) are used in many locations in ATE and instrumentation systems. These use transistor or diode-based switching circuits, which can have very high bandwidths, low losses, high isolation, etc.; semiconductor switches are not capable of achieving all these “desirable performance characteristics” at the same time. Previous limitations of poor linearity have been overcome by RF CMOS component vendors, and they are now claiming insertion points held by electromechanical components just five or ten years ago. Other than overall RF performance, their biggest limitation for many applications is repeatability and assorted operational reliability concerns (other than lifetime, which is exceptionally high).
MEMS Investor Journal: What are the pros and cons with respect to using MEMS switches in ATE applications?
Dr. Dan Hyman: MEMS are small – in RF and high-speed digital testing applications, testing is improved when you get the switching subsystem as close as possible to the devices under test. The physical size of most load boards is dominated by EM and reed relays. Smaller footprint is important, as is reduced thickness when testing ultra-dense I/O chipsets. MEMS relays are essentially identical in size to surface mount semiconductor switching products.
MEMS are fast and consume no or low power – the time and available power to switch reed and EM relays is a significant limitation for test engineers to consider how and when they reconfigure their test programs. A complex load board might have hundreds of parts, and it would be impossible to switch all of them at the same time if using reed or EM relays. With MEMS, every part on the board could be reconfigured at any time on the fly using effectively no power. Overall, the testing time can be reduced somewhat (for mild operational benefit), but the ease of setting up or changing test programs is greatly improved. Note that MEMS are considered slow by semiconductor standards, so they are inappropriate for certain instrumentation, radar, and communications insertion points. Power consumption is high when compared to FET-based switches, but low compared to most other semiconductor switching circuits.
MEMS last a long time – operating lifetimes under typical load conditions are 5x-20x that of reed or EM relays (although still lower than semiconductor relays). One million or 10 million cycles no longer lasts an entire production run in the consumer electronics industry, so load boards are often replaced mid-run, resulting in substantial down time. With MEMS relays, down time due to end-of-life can be essentially eliminated. All else being equal, this alone presents a compelling advantage to many ATE system users, and would result in substantial test cost savings.
MEMS are repeatable – MEMS relays are essentially electromechanical by nature, so they are significantly more repeatable than semiconductor switching devices. Repeatability at low load is based on organic and water contamination, so cleanliness and hermeticity are critical. When clean and packaged, MEMS are as repeatable as high-end EM relays, so packaging has been the most substantial development effort for all MEMS relay vendors.
MEMS are inexpensive for their performance – high-end relays can be made in other technologies at the expense of cost and other performance metrics. With MEMS, there is very little sacrifice of any performance metric; you get a high level of all performance metrics (high, not necessarily highest), and it comes with a mid-end relay price tag. If you want a lot of them, then you get them at a low-end relay price!
MEMS Investor Journal: Are MEMS currently replacing non-MEMS technologies for ATE applications? Why or why not?
Dr. Dan Hyman: Yes, but very slowly. The ATE industry is fairly conservative when it comes to adopting new technologies, as they are part of a critical infrastructure for many industries. The risk of change can be high, so new technologies must be tested and hardened very carefully to ensure each new insertion point is both “helpful” in the cost/efficiency direction and “not harmful” in any other direction.
The longest running MEMS parts in test equipment are, of course, those used by Advantest. They have their own proprietary MEMS components for switching critical RF and wireless paths in many of their systems. These have been in the field over a decade, with a likely total of hundreds of thousands of MEMS components clicking away at this moment inside their systems worldwide.
XCOM is a relative newcomer, having developed our parts over the course of the last 11 years. That's a long time for a semiconductor business, but not in the world of relays and switching components. Although our part was essentially completed several years ago, it takes time to perform the necessary testing and qualification required by ATE customers, and each new package or post-process modification requires new testing!
We have two small design wins in place in ATE systems used in the defense industry for high-frequency applications. We are in qualification for several others in similar applications that I cannot discuss. We are excited to have two major semiconductor companies looking to use our parts to replace existing load boards and racks, but we have not completed qualification with those customers, nor has the exact location of our first insertion been defined.
MEMS Investor Journal: If MEMS based switching in ATE application proves that it is, after all, not to be the best solution, what do you think are the main limitations?
Dr. Dan Hyman: ATE systems are highly technical and widly varying beasts – there are many types of components already used inside each system, and no single technology will ever displace all other technologies. There are always going to be niche insertion points for the various technologies, and MEMS are simply newer offerings in a long list of parts available to the switching subsystem designer.
A more precise question to answer is whether or not MEMS will be a dominant technology in the industry, essentially displacing electromechanical, reed, and semiconductor relays in many or most systems. In order to achieve this level of market penetration, the cost of MEMS components needs to come down towards commodity levels (e.g., $1-4 per part) to compete with reed and semiconductor relays. Fortunately, because they are essentially semiconductor products, we all know this can be done, but the question is when and at what cost to MEMS suppliers and manufacturers? Further investment to drive down costs and expand product lines is dependent on achieving a level of return on past investment, and this is true for larger suppliers (e.g., OMRON) as well as small (e.g., XCOM). If a moderate level of commercial success is not achieved within the next few years, then it is unlikely that we will see the volumes ramp, costs plummet, and the type of significant proliferation recently projected by analysts for 2015 and beyond.
That said, on the technology front, I do not see any major technical limitations of MEMS switching components as compared to the dominant technologies in the ATE industry. In most quantitative categories, MEMS are superior, but that does not mean they are perfect or ideal for most switching needs. I would say that they are very competitive and compelling for many insertion points, but not all, and if a particular insertion point requires an enhanced performance specification (e.g., power handling or bandwidth) or an expanded product configuration (e.g., 2x2 matrix or 4x4 module), then it will cost someone real money to get it.
In this sense, it can be summarized that the most substantial limitations holding back MEMS proliferation are very real business and accounting matters rather than technical ones.
MEMS Investor Journal: What are the current cost ranges and cost trends for switches used in ATE applications? Where are MEMS switches in this spectrum?
Dr. Dan Hyman: Each technology has a range of costs and trends that are different. In general, parts with higher performance in one or more performance characteristics (e.g., RF bandwidth, loss, isolation, lifetime, repeatability, size, speed, power consumption, etc.) will cost more. The relay industry is highly competitive, and new product line entrants typically result in a bit of a shakeup for older, more established lines. In the RF and high-speed digital component markets, some product lines can become entrenched and be highly profitable for many years.
Electromechanical components range from a few dollars to many thousands of dollars depending on the configuration and performance of the part. They can be value priced based on the landscape of competing products, but most RF-capable ATE parts are in the $5-$50 range for surface mount(ish) or through-hole parts and in the $50-500 range for coaxial boxes. Because these parts tend to be used in higher-end RF applications, the trend on pricing is to remain relatively flat over time until a new product shakes up the curve.
Reed relays are at a lower cost tier, which has resulted in a rapidly growing market share for testing high-speed digital chipsets and boards. Reed relays are mostly surface mountable components in the $1-10 range, and have higher volumes that help vendors achieve a lower variable cost. Because of commoditization of the technologies and their proliferation in test systems for consumer electronics, reed relays have stronger downward price and size pressure than EM relays. I expect this downward price pressure to ease during the economic recovery, and then pick up again in a few years as manufacturing lines refill.
Semiconductor relays have perhaps the widest cost tier, but the strongest downward price pressure. High end semiconductor switching devices can run into the hundreds or thousands of dollars for coaxial boxes, and commodity FET-based switches can be just a few cents. The semiconductor relays in the test industry tend to be mid-level products in the $1-20 range, selected for high speed and very high lifetime applications, or in areas where hot switching of substantial load signals is required (i.e., instrumentation).
At their core, MEMS relays are semiconductor products, so their cost structure closely matches that of the semiconductor industry. The manufacturing has not been standardized and ramped like the VLSI industry, however, so cost/volume curves for MEMS more closely match GaAs rather than silicon CMOS, but with higher packaging costs due to hermeticity requirements. Present MEMS relays from the few vendors available are in the $20-50 range at low volumes for surface mount parts and $100-1000 for coaxial boards (price rises with frequency). This is comparable to (though less expensive than) other high-end switching products, with pricing based on actual costs of manufacturing and qualification testing in low-volumes. Fortunately, because they are batch-fabricated, MEMS relays scale down to $2-6 in higher volumes, and can compete with reed and semiconductor relays in cost sensitive applications, especially where higher performance can add real value.
MEMS Investor Journal: Can you provide examples of one or two of your customers and why they are using your MEMS switches for their ATE applications?
Dr. Dan Hyman: Our defense ATE customers like our parts because they are the single highest combination of bandwidth, insertion loss, and isolation available on any commercial component available anywhere in the world. In ultra-wide bandwidth applications, this is a major selling point, as our one part can replace and outperform a family of 4-6 larger, more expensive parts. Other defense customers specifically enjoy one or more specifications such as loss, linearity, size, etc., but ATE applications focus on the RF performance as a whole.
One of our commercial customers likes our parts because they have a much higher lifetime than the much larger, costlier electromechanical components they are presently using, and because our parts are small semiconductor-style components instead of unwieldy connectorized boxes. We are working with them to define a larger module or board-level switch matrix that can increase their ATE system “uptime” from 70% all the way to 90%, while simultaneously slashing their yearly relay costs down to a quarter (or less) of what they are presently paying. This kind of application makes great business sense for everyone, and will directly result in increased profitability for a significant semiconductor product line.
*********************************************
This article is a part of MEMS Investor Journal's ongoing market research project in the area of MEMS and non-MEMS switching devices for ATE applications. If you would like to receive our comprehensive market research report on this topic, please contact John Williamson at [email protected] for more information about rates and report contents.
Copyright 2011 MEMS Investor Journal, Inc.
Comments