In April, Salland Engineering opened a new test centre for chips in Zwolle. CEO Paul van Ulsen expects the company’s production testing business to grow by 20 to 30 per cent in the next few years. A chat on what it’s like doing business in a low-profile yet vital corner of the semiconductor industry.
He’s been testing ICs since 1986, when he wrote code for pattern generators during his college internship. After buying out PPM Oost and other shareholders last year, Paul van Ulsen is now the sole owner of Salland Engineering. With the investors’ departure, Van Ulsen closed out a turbulent period that had recently ended in bankruptcy for Salland Software. He seems content: ‘Now I can move forward faster and more effectively.’
After thirty-plus years of testing experience and a failed venture in automating test centres, Van Ulsen doesn’t seem to have lost an ounce of his corporate drive. His latest project is production testing for European and American chip manufacturers – the impetus for the five-hundred-square-metre facility he’s recently completed. The venture propels tiny Salland into a market that’s dominated by Asian players.
But Van Ulsen sees room to grow this line of business in niche markets. Salland isn’t targeting the segment where the Asians are lord and master: competing on price in volume markets. The company plans to serve chip manufacturers through quality, specialized engineering, and proximity. Van Ulsen believes his company can grow its production-testing business by 20 to 30 per cent in the next few years.
Salland Engineering is rooted in test technology, a domain ruled by Advantest, LTX-Credence and Teradyne. Together, these three giants hold 90 per cent of the nearly four-billion dollar market for automated test equipment (ATE). With forty-five developers and nine million in revenue in 2017, Salland’s success depends on the innovation and added value it delivers.
Roughly speaking, test technology addresses two main concerns. As a first step, Salland’s engineers provide the knowledge that semiconductor fabs use to validate a design. Once it’s been determined that the chip design is 100 per cent correct, volume testing can begin. This is the second step, and it means devising smart methods to herd chips through testing as quickly as possible. ‘The buzzword is test coverage,’ Van Ulsen says. ‘It can’t take too long or cost too much. Otherwise cell phones and cars will become too expensive. That’s the perpetual tradeoff: finding the right balance between quality, throughput, capacity and last but not least cost.’
That tension gives rise to intellectual challenges, because devising a strategy to test chips means solving puzzles and thinking outside the box – with cheap and reliable as hard requirements. ‘Good test strategies place high demands on both hardware and software. It’s always about that combination. And that’s exactly what we’re good at: making those disciplines our own and using our firmware expertise to create a complete solution. The hardware-software mix has to be optimal – so our people have an affinity for both. Our software people also know firmware and hardware. And vice versa: our electronic engineers understand software. It’s about the interplay.’
Traditionally, chip manufacturers develop their test technology in-house. They call on Salland Engineering to provide new ideas or if they’re short on staff. But Salland also serves the major ATE suppliers. The test cells that Advantest, LTXC and Teradyne sell – with price tags from seven hundred thousand to two million dollars – last for fifteen to twenty years, making it vital to continually upgrade them with the latest technology. High-tech renovation, as it were. ‘Chip testers basically use yesterday’s technology to test tomorrow’s components. Test technology is always running to catch up.’ The ATE equipment generates analog and digital signals, but you also need power supplies and test procedures. ‘We provide those. ATE suppliers open up their software environments to us for that.’
By way of example, Van Ulsen holds up a complex PCB jam-packed with semiconductor devices and cooling fins. ‘To keep the cost down, we do a lot of parallel testing. Four at a time for complex chips, and a couple hundred at a time for simple ones. For that, you need this kind of high-density instrument. This one has seventy-two channels that can simultaneously supply seventy-two components with power.’
A lasting advantage
Paul van Ulsen studied electrical engineering with a minor in computer science. After an internship at Rood Testhouse, he stayed on for another six years as a test applications engineer. When Rood closed its doors in 1992, he and four colleagues moved to Salland Engineering. Two former Rood employees had started the company a few months earlier, with a focus on mechatronics development. The five new engineers turned out to be an asset. From day one, Salland Engineering placed them at customers for a healthy hourly rate. That gave the young company a flying start, but also shifted its focus to delivering chip-testing technology and services.
In the six years that followed, Van Ulsen’s test applications work was joined by more and more additional tasks, from HR to training and sales. ‘It eventually became too much of a good thing,’ he laughs. ‘So I decided to focus on the commercial side and got into business development.’ He joined the company’s management team around 2000.
Van Ulsen’s technical background is a lasting advantage. In an interview on the PPM Oost website, he notes that it enables him to easily understand the customer’s point of view. ‘They start off viewing me as a salesman. But after I ask a few questions, they discover that I know a lot about the topic. That always surprises them.’
Soon after 2000, the company’s mechatronics and non-testing activities split off as Salland Electronics. Since then, Salland Engineering and Salland Electronics have operated fully independently, though they still work closely together. ‘We still have a strong relationship,’ Van Ulsen says. ‘We develop our test boards in-house, but they jump in to help when we need them. That goes for both production and development. When we have more people than we need, we send them over to Electronics, and vice versa.’ In 2001 Van Ulsen and his two fellow executives bought the company from its investors, and Van Ulsen took over as CEO.
After twenty-plus years in IC testing, Van Ulsen embarked on a new adventure seven years ago. He’d long since spotted an untapped market in test facility automation. Salland was already helping customers set up better data management and more efficient operations in their test centres. When the opportunity to take over the software companies Pintail and Compat arose in 2011, it seemed like a no-brainer.
Pintail delivered realtime adaptive test technology to major semiconductor manufacturers. Salland already had a several-year strategic alliance with the Austin-based company. Closer to home, Compat built software solutions for wafer map storage systems and automation in the semiconductor industry. Van Ulsen saw his chance and absorbed both companies with financing from PPM Oost and others. The move seemed to equip Salland with all the technology it needed to start developing an entirely new market.
The ambitious CEO christened the new venture Salland Software, and the seventeen software engineers from Pintail and Compat hit the ground running. But three years later he had to throw in the towel and file for bankruptcy. ‘We were too far ahead of the market,’ he says about the venture now. ‘The test industry wasn’t ready for the next evolution in automation. There was just too little support among upper management.’
He goes into more detail: ‘You’re talking about automating global chip companies’ entire facilities at multiple locations in multiple countries. To do that you absolutely have to have the support of senior management. We had an installed base, customers who used it. We also had ample resources, but at some point we didn’t have the cash flow to keep paying a staff that had grown by then to thirty highly educated engineers. It took too long for the big orders to come in. We couldn’t hold out.’
Van Ulsen is convinced that his vision was sound, but he was just too early. ‘The market just wasn’t ready. Right now two top-tier companies are trying to sell this kind of software, but I don’t think anyone’s making money with it yet. I think it’s a lack of attention from upper management. The major chip manufacturers and foundries could really save a huge amount of money this way.’
SerDes, MEMS, photonics
For Salland Engineering, swelling to Asian proportions in production testing isn’t realistic. But there is room for the company in the market for high-end components such as very complex digital and mixed-signal chips. Its customers are primarily specialized European companies, though Salland also has one customer in San Diego. In addition to serving suppliers of complex chips for the automotive market, the company also tests relatively small series such as those used in ASML’s machines and Thales’ radar systems.
SerDes (serializer/deserializer) chips are an interesting example. In data centres, these chips divide the massive information stream across the centre’s servers and then at the end recombine the data into a serial bitstream for the optical fibre. This market requires highly advanced chips with data signals at thirty gigabits per second and higher, and Salland Engineering is able to test and validate them. Most of the chips the company tests in Zwolle come from European fabs. But it also receives ICs from the major Taiwanese foundries – usually for designs from fabless companies in Europe.
The expertise to test at several-dozen-gigahertz frequencies comes in handy for a few emerging markets that Salland Engineering has its eye on. The company is currently developing automated test solutions for micromechanical components and photonics. MEMS chips such as accelerometers for airbags and chip-based gyroscopes are currently still tested by exposing them to physical acceleration or rotation. ‘We want to start testing and validating those components fully electrically, that is, by sending them a signal and instantly measuring their response,’ Van Ulsen says. ‘That will let us cut production costs by at least 50 per cent.’
To enable the use of the technology for MEMS and photonics in its production testing centre in Zwolle, Salland Engineering has partnered with the University of Twente to explore the possibility of setting up a test technology lab. ‘If we want to build the measuring instruments for MEMS and photonics, we’re going to have to take things to the next level,’ says Van Ulsen. ‘We’d love for other companies in the Netherlands to join us.’ He’s optimistic about the market’s potential: it involves a significant technical challenge and interesting projects where Salland and its innovative team can stand out from the crowd. ‘There’s demand for new technologies, especially if you’re leading the market; we’ve proven that time and again.’
How’s photonics coming along?
‘It’s still a long way off. MEMS is closer. It’s already part of the foundation that underlies the Internet of Things, though it’ll take three or four years before it reaches truly high volumes. Compare it to Bluetooth. It took quite a while for that to hit high volumes in the market. The current generation of MEMS chips are still tested using physical stimuli. All the testing is custom work. But there are more and more integrated solutions where manufacturers are making the MEMS sensor with a microcontroller and a wireless interface in a single standard semiconductor process. That’s when things will really start rolling and an obvious opportunity will arise to cut costs by testing it fully electrically.’
How far along is Salland in that regard?
‘We’re developing the technology, but it’s also a question of timing. There are already a few manufacturers with MEMS testers, but even those can’t yet handle complex application-specific chips with integrated MEMS functionality. Right now, our technology is closer to a lab-based setup with measuring equipment than a truly complete MEMS tester. But we have the technology available and we’re looking at the right time to go to market. We’re already discussing it with customers, too. I estimate it’ll take another year or two before the market is really ready. So we’ll be focusing on development for the next two years.’
Roughly speaking, creating new test solutions means constructing a measurement channel and developing a measuring instrument. ‘We have to create a design, add in a PCB, develop firmware for FPGAs and then validate the whole shebang. We have to make sure the instrument gives the same values today as two years from now and that it tests system number one exactly the same as system one hundred. That’s why validation takes twice as long as development.’
That calls to mind your adventure in test automation – so it’s round two of convincing the market?
‘Indeed, it means evangelizing and being ready to go the instant the market starts to invest. In photonics, the horizon is further away. We aren’t standing still there, but we do have a few years to decide on the direction we’ll take for test solutions. This year or next we’ll present proofs of concept, and then we’ll move on to production-ready test solutions. In the coming months we’ll be looking for partners to help us make all this happen.’