Executive Test Summit Tackles the Issues
Rick Nelson, chief editor for T&M World and EDN
The Executive Test Summit at SEMICON West 2009 featured panelists from Advantest, LTX Credence, Teradyne, and Verigy. Rick Nelson, chief editor for Test & Measurement World and EDN moderated the panel.
To kick off the discussion, moderator Rick Nelson asked, “How are your companies continuing to innovate during the downturn?” Consensus: Everyone has been forced to cut their R&D budget, but Mark Jagiela (Teradyne) pointed out that collectively and individually, all of the companies are still spending a “… tremendous amount of money on R&D for this industry.” Innovating through any economic time is tough, but necessary. All agreed that a strong R&D budget is critical, and that companies need to weather the downturn and emerge with a strong product portfolio for the coming upturn.
Innovating during the Downturn
Dave Tacelli (LTX-Credence), R. Keith Lee (Advantest America), Mark E. Jagiela (Teradyne), and Keith Barnes (Verigy)
In previous downturns, Keith Barnes (Verigy) noted that “What led us out of the downturns were new products that we never envisioned, and that it’s happening today as well.” He said to look at EDA companies and the new products that are coming out which will drive the industry in the future. Verigy focuses “… on our customer’s customer and what’s going to be new and next.” Panelists agreed that the industry needs to be prepared the next phase which means a total commitment to innovation.
Major Vendors: True Competition or Niche Focus
Next, the panel tackled Nelson’s question, “Will the major vendors continue to compete with each other, or will they focus on niche areas in which they can excel? For example, flash, DRAM test, SOC test, mixed-signal, RF, analog test?” Nelson mentioned an example: In March, Yokogawa said it would freeze its SOC ATE development and focus on memory test ATE.
Barnes said that Verigy is going in the opposite direction, quoting Warren Buffet who said, “Be fearful when others are greedy. Be greedy when others are fearful.” Today, Barnes said, “People are fearful, so it’s time to take a look at the opportunities that are out there.” He stated that Verigy is looking at opportunities and just expanded into DRAM, introduced a product that services the low-cost part of the marketplace, and bought a probe-card company. He also mentioned organic and inorganic additions to the Verigy product portfolio.
Both Jagiela and Keith Lee (Advantest) foresaw a future serving all markets and that the top companies (Advantest, Verigy, and Teradyne) will compete in common markets across a broad set of products. Jagiela said, “Company product portfolios are becoming more and more overlapped over time in terms of pure test and companies will look at ways to differentiate and focus on ‘time-to-market’, tools to get the silicon out, yield enhancement, and interface technologies. All of those are fertile ground.” Then Lee discussed the issue from the customer perspective in the OSAT space, stressing, “There’s a requirement for a comprehensive platform than includes digital, memory and RF mixing….That drives the test suppliers to provide a more comprehensive test solution across multiple market segments.”
Dave Tacelli (LTX Credence) took a different tack with “When you look at the four companies (on the panel) and the companies that are no longer here due to consolidation, the whole goal is differentiation… By the nature of that, you look at what you do better than someone else.” He said that the four companies (on the panel) don’t always compete with each other, only competing in areas of strength— RF, or digital, or high-performance analog. Tacelli believes that what will separate the companies from each other is focusing on areas of strength and then developing technology in those areas now.
Embed Test IP in Devices— or Buy External Testers?
A SanDisk attendee asked the panel, “Should we embed test IP in devices rather than buying external testers?” Jagiela pointed out that most tested parts have some form of embedded IP and “The nirvana is to embed enough IP so that test is completely irrelevant.” He said that so far, no one has accomplished that because the key issue is preserving the quality of the semiconductor device which is critical due to use in medical, automotive safety, etc. He summed up his views with “We’ll continue to see evolution in that direction, but it’s not over for test.”
Next Barnes took the stance that tester companies are not motivated to do embedded test, and that embedded test probably has “… another five years before really gaining momentum.” He said that the companies that have been successful so far have put huge teams on the effort (memory companies, some processor companies) and only concentrated efforts can push the issue forward. He summed up his position with “The general SOC customer doesn’t have the potential to do this with today’s tools. A lot of tools need to be developed to make this a reality. Digital alone… in analog, there’s nothing… Very little in RF. There’s a lot of work to be done.”
Agreeing with Barnes, Tacelli added that “The content in that IC is such a complex problem, and when you compound that with the analog circuitry and what has to be done, I don’t see that being overcome for a long period of time.” However, he sees tester companies embracing more and more test embedded in the chips “because we have to.” He said tester companies will be “… around for a while, as it’s a very complex problem, especially on the analog side.”
“Holy Grail” BIST
Intel has had some success with built-in semiconductor test (BIST). The question, “Would it take an Intel-like company to drive BIST or could it be adopted by smaller fabless companies?” The consensus? Not one panelist knew of any major semiconductor device market where the quality check with a tester has been eliminated. Although there’s some experimentation, apparently BIST is not in any real applications yet. Lee said that “100% BIST is somewhat of the ’Holy Grail’ with research in this area going on for a number of decades— a trend will continue and become more pervasive. But the demise of the ATE industry? A little early to predict at this point. We will see testers get more simplified, with tighter integration between test external box and embedded test. To see it go away wholly, I wouldn’t anticipate that.”
Embedding Test IP: Ad Hoc Efforts
An attendee from DA-Test asked, “Would your company embrace building an application department to support IP on the die to perform test functions?” First, Barnes said that people want built-in self test as a way to reduce overall cost of test and to get better quality out of their IC, but the need for at least three different environments complicates matters. He emphasized, “We have to change our R&D structure to focus on that, and that’s not going to happen for several years.” He said that companies that are building IP that would work for these products will be interesting to test companies in the short term due to lack of internal capability.
Then Tacelli brought up the issue of customer demand, noting that for his company to invest a significant amount of R&D, customer demand is crucial. “Today, customers are really focused on the overall cost of test… The other thing they’re very keen about, with devices changing so rapidly, is ‘time to market’… how to get them from pure silicon into production as fast as possible.” He said that when more customers demand it, then companies will have to focus more on this area. Lee agreed, stating, “The degree and the frequency at which test companies will adopt and work with others to embed test IP will really be driven by market and customer demand.”
Jagiela addressed the complexity of the issue by saying, “We’ve developed multiple interfaces to both commercial and customer custom-embedded test IP whether it’s diagnostics or development tools to support them. We’ve even co-developed instruments in conjunction with embedded IP that must work in concert— that cannot work separately. It all boils down to such a proliferation of differences that it’s hard to economically address it as a broad subject. So it’s been to date an ad hoc effort. ”
Who’s Responsible for Internal Test?
“Who’s responsible for internal test— tester companies or EDA companies?” was the next question poised. Barnes responded, “If you look at each of the EDA vendors, each has a partial solution to the problem.” He said that the industry could work with the EDA industry to move toward a standards-based approach and gave the example of LogicVision. This company met with some success but it was not broadly accepted as it was unable to do it on its own. Barnes stressed that “It has to be our companies along with the EDA companies and a few very specific customers that drive a standard base of products that can be used— maybe first in the digital area and then eventually in the analog area.”
The Days of “Big Iron”: Over or Relegated to the OSATs
A Pragmatics Technologies attendee asked, “Are the Days of “Big Iron” Over—or Will it be Relegated to the OSATs only?” Barnes responded, “Around the late 1990s, an inflection point occurred where the bigness of the Iron and the ASP of the Iron inverted and it started getting smaller and smaller… and we’re still on that trend. That’s not going to stop because the technology available to enable the building of the tester is racing farther ahead of Moore’s Law than the actual semiconductor devices themselves which is why our industry actually grows slower than the semiconductor industry. We’re providing more innovation ahead of the curve.” The number of customers who decide to deploy custom in-house test solutions goes up and down over the years. “Today, (Big Iron) is a lower percent of the solution than it was even ten years ago.” He said that high-complexity, lower-end products need more attention.
Tacelli said that in the last 5 to 7 years, a lot of focus has been on “Just enough test.” In other words, focus on how to test the device adequately, but not add test when it’s not really needed. He emphasized that the cost of test must be reduced, and that innovation is necessary because the ASPs are going down. He said “We have to learn as companies how to make money to supply “just enough test” for markets through creative innovation— ‘Where else do we add to the value chain?’”
Success Drivers: How Can Suppliers Help in that Battle?
An Analog Devices attendee asked, “What will drive success and how can suppliers help in that battle?” Tacelli responded, “The greatest thing that you can do during a difficult period of time is make tough decisions and stay the course, and then hope that those decisions bear fruit.” Building companies to be as flexible as possible is important to minimize problems with ramping production. He said that some issues like keeping skilled people are now easier to deal with due to outsourcing manufacturing, etc. Lee stated, “The big challenge is to ensure that your R&D is being invested in the right areas so that when you hit the inevitable upturn, you can enjoy the market ramp. ‘What can analog devices do as a supplier do to help?’ Investing in R&D in the pin electronics technology area and in those ASICS is very important to make sure that we hit the technology points as we come out of the downturn.”
Jagiela added, “The push and the innovation around reducing the cost of test will continue… The real innovation will likely occur outside of that dimension…. Reducing ’time to market’ to develop complex test programs, that’s one area of innovation that will be a differentiator going forward.” Another example he talked about involves “using the tester to be more of a diagnostic tool of the design or of the process.” In test, he concluded, these are the areas outside of the realm of analog devices that will be different in the next five years.
Systems-Level Test: A Parallel Industry?
An Amkor attendee asked, “Do you see systems-level test as a parallel industry or an industry that may converge again?” noting the Agilent/Verigy split. Jagiela said, “Some of (our) customers want to see more systems-level test capability— integrating multiple die onto substrates, or packaging them in 3-D or whatever. We’re seeing lots of different MCP packages. Taking the logical extension, it moves them towards a system-test approach. Everyone is aware of the protocol capabilities that are being working on. So there is a move toward that (systems-level test), but it’s not a move that is specifically going into the domain of Agilent or Techtronics… it’s pretty specific to higher levels of integration of the devices and packages that we see today.”
Lee said that customers want to move system-level test back into package test, and reduce the need for system-level test. He noted that conversely, new architectures are coming out that drive test in “… a different direction like 3-D packaging, through complex die stacks requiring some level of system-like test.” Also, he noted that there is a “convergence of automatic thermal control technology that is a parallel you see in both package test as well as system-level test. Soon, you’ll see some of the ATE companies participate in systems level perhaps in that area as well.”
Anticipating Customer Needs
Nelson then noted, “At our summit two years ago, keynote speaker Ashoke Seth (Intel) urged test equipment vendors to be proactive in anticipating customer needs.” He asked, “Have you been able to anticipate what your customers want?” Tacelli said that he wants to work with all his customers “… on where their designs are headed, not only for the next generation but for multiple generations beyond that. It’s not just about focusing on what you think is important but really about focusing on where their designs are headed… How much technology they’re trying to combine on one piece of silicon or multiple pieces of silicon in a single package…” He also mentioned DFT and how it gets put more and more on the analog side. He said that companies always need to test things ahead of the technology so it’s good to work with customers on short-term plans, but critical to know where customers are taking their designs in the future.
Collaboration with Partners in the EDA Industry
“How important is it to collaborate with partners in the EDA industry on DFM and yield-learning solutions?” was the next question thrown at the panel. Barnes emphasized how critical working with partners is, saying, “ … there’s so much work that can be done in that space. If you think about how people design ICs today…ask yourself, ‘Why do you do it that way?’ Case in point, when you build an IC and then simulate it to see if it’s working correctly or not, then extract a VCD file, it’s flattened and all the timing is taken out. It’s a dumb file with respect to what the chip really looked like. Then we put it in the tester and reinsert the timing. It’s not a very well thought-out process.” Barnes suggested that an improved process would keep the data at a better level of integrity. He summed it up with, “The industry needs to work more closely with them (EDA companies) on developing standards… to make it easier for them to get the data in and out.”
Next, Jagiela added his opinion with, “One of the innovations is starting to program the devices of the native protocol of the device is a big departure from how testers have been used in the past….The need to translate from EDA tools into tester-specific vectors and then back again. It’s going to be something that five years from now, won’t be solved but will be tremendously eased. Going back to anticipating customer needs, market share is a good proxy for that. Our design teams are faced with this challenge which is that we spend 2 to 3 years developing a product and then it has to be useful for 10 years in this dynamic world of semiconductors. How do we envision, in working with customer insights, what that’s supposed to look like? Because it’s a 13+-year time frame that we have to design to.” He thinks that the economics of test are good, but challenges facing the industry now are a bit different.
Changing the Mindset: Where is the Value?
An question from the audience was next, “ASP is falling, and the industry is not in a growth phase. Can adaptive test, yield analysis, diagnostics, etc. provide growth to this industry when none of the ATEs are set up to sell or charge for software?” Barnes stated “The market is shrinking, but the market will grow again starting next year. It won’t be back to the rates we had in 2007 for some time but it will grow again.” He said that everyone has twice as many software engineers as hardware engineers, and that the industry is used to integrating it all and selling it as a product. He feels that “Some of these analysis tools will be sold separately, whether there’re on perpetual licenses or annual subscriptions.” He stated that Verigy is moving in this direction with their yield vision product. He says that it is important to help customers increase their yield, but the industry also needs to refine the model so the gain can be shared.
Lee said “What will drive growth in the ATE industry is growth in the semiconductor industry. We’re seeing some early signs of that in DDR3 growth and in next generation microprocessor growth, which is a positive trend for the ATE industry….but the question remains ‘Is that sustainable?’” Next, Tacelli brought up the issue of value: “Customers will pay if they see value. If they don’t see value, they will not pay… It’s up to us to educate them on the value that we’ve delivered. If we can’t do that, they absolutely won’t pay— for application support, software enhancements…” He said that the ATE industry needs DFT tools, yield enhancement, whatever our differentiators are, we need to show value through “… more productivity, lower test time, faster time to market.”
Different Approaches on How to Succeed
Nelson quoted Frost & Sullivan, “Some of the key criteria for succeeding in the semiconductor test market are to provide comprehensive testing solutions, reduce the overall cost of test, reduce ‘time to market,’ and increase production yields.” Then he asked if panelists agree, and if so, “How are they addressing these criteria for success?” The panelists agreed that these were the key areas, but made it clear that priorities have to be set, especially with the downturn.
Lee discussed Advantest as being unique because it is vertically integrated with its own in-house manufacturing. He believes that in-house manufacturing makes for better reliability and allows the company to control their own inventory and just-in-time pipeline. He said that that they are not “… burdened with sharing subcontracted manufacturing” so they have the ability to deliver a high quantity of product in the coming upturn.
Taking the opposite approach, Tacelli said that he believes that focusing on core competencies is critical. “For most of us, what we do is we innovate things, we solve problems. Part of solving the problem isn’t building something. It’s not putting the screws in a frame or cabling. It’s not loading the pallets.” He says that it’s better to “Spend money on the things that will earn you money and leave these other things to someone where it’s their core competency… For LTX Credence, it’s been a significant benefit to be as flexible as possible so that we could invest in R&D—not in inventory.”
What’s the Next Big Thing?
To wrap up the Test Summit, Nelson asked the panelists, “What’s the next big thing? What are the challenges you’ll face in the future as process geometries shrink and start thinking about testing 3-D structures with TSV?” Lee said, “Although automotive is suffering now as an industry, but the semiconductor content for automotive is increasing and will continue to do so. Power management applications, mixed signal applications for automotive are a growth area over the long term. Same thing applies for power management devices for hybrid cars and other ’green’ products. I also see a long-term trend for RF proliferation in devices, anything mobile or wireless connected— those are positive trends for the industry. Then there’s the solid state drive potential for the flash market as a growth driver.”
Barnes agreed with Lee and said that customers are exploring new areas every day. Mobile computing, consumer mixed signal areas, medical, no matter what… It’s exciting… We learn by working with customers. There are lots of really cool technologies that are being worked on today that will drag us out of this downturn.” He is optimistic that the next couple years will move the industry in a positive direction. Regarding TSVs and other technologies, he thinks that they are interesting challenges for us but that the industry will have to be able to test those technologies differently than today. “Once we crack those challenges, whether it’s remote sensing, redundant TSVs, and trying to work through that, that’ll help push the technology from a packaging standpoint even further and that will help our industry.”
Summarizing, Jagiela added, “The test industry gets much maligned, but I’m proud of being in the industry, and what it’s accomplished in conjunction with all you who work with our customers in test in bringing down the cost of semiconductors for consumers worldwide. It’s an intriguing industry, and we’ve been successful at it. It’s hard to be optimistic in this environment that we’re in right now… but the technical challenges keep on. Technology challenges are as prolific as ever.”
For more information, please visit http://www.semi.org/en/IndustrySegments/Test/index.htm or contact Karl Stuber at firstname.lastname@example.org.
August 3, 2009