Test Engineers In Very Short Supply

Semiconductor design, verification, manufacturing, and test requires an army of engineers, with each playing a special role. But increasingly, these disciplines also require additional training to be able to understand the context around their jobs, and that is making it harder to fill different positions at a time when the chip industry already is severely short-staffed.

This is particularly true for engineers whose jobs traditionally came in later in the flow — the test, process, and yield engineers. Increasingly these disciplines are shifting further left, and playing a more vital role in critical applications. Knowledge in all of those areas, as well as others such as inspection, metrology and analytics, is now required to get chips out the door more quickly and to improve reliability throughout a device’s expected lifetime.

“As a test engineer, you’re likely to pick up on characteristics, trends and anomalies that the design and production teams won’t,” says Marie Ryan, a marketing executive at yieldHUB. “It is vital that test and yield are considered early in design now, but considerations such as power and security are morphing into the test engineer’s job by default.

Much of this has to be learned on the job for a couple reasons. First, curriculums at most universities are relatively fixed, so gathering expertise often means taking additional courses outside of the required classes. And second, there are so many changes underway with the slowdown in scaling, the resulting increase in heterogeneous advanced packaging, as well as the introduction of novel architectures for new markets, that the chip industry is far outpacing the trainers.

“Test requires a lot of multidisciplinary knowledge, and it’s not something that’s only breadth — it requires depth, as well,” said Rob Knoth, product management director at Cadence. “That’s what makes it particularly challenging. Some of the better test engineers that we work with are people who maybe didn’t start off as test people. There are people who were just regular RTL designers or regular semiconductor engineers, and they broadened into test. That allows them to bring a pretty powerful toolkit to bear on the problem.”

Test engineering is a highly valued expertise, and that discipline is becoming even more valuable as chips are expected to last longer in the field, and as they are used in mission- and safety-critical markets such as servers, automotive, and medical applications. “If you don’t do that job well, it can result in a pretty big end cost to the company. And so, that’s driving it, as well,” said Knoth.

Hard to train, hard to fill
Hiring is made more difficult by the fact that for years, the chip industry has been competing for engineers against the likes of Apple, Google, Facebook and Amazon. Software was considered the future until the past several years, when AI/ML entered into the picture and Moore’s Law began to slow down. Suddenly, intelligence needed to be added into everything, and some of the most interesting engineering challenges moved into hardware rather than software.

“There is a lot of industry competition for the high-demand skill sets required for machine learning, for example, and therefore can be a challenge to fill,” said Andee Nieto, senior vice president and chief people officer at Xilinx. However, “the hard-to-fill jobs are what we call combo jobs, where companies have merged job skills together to form a new ‘hybrid’ type of role. When they combine these skill sets into one position, it makes it a lot harder to find the right person.”

Test, which for years was stuck somewhere in the middle of the pack when it came to hardware engineering, suddenly became much more challenging over the past few years. Test budgets were increased, up from a flat 2% of manufacturing costs, and analytics were added into the mix in order to make sure chips would be able to withstand environmental, electrical and even unplanned events and continue working according to spec.

“The biggest difficulty in hiring is everything related to test,” said Cadence’s Knoth. “That’s the full gamut from test architecture, down to test implementation, to post-silicon, and debug. It’s a natural outgrowth of a lot of different trends. One of the big ones that was an early driver of that was the tremendous rise of semiconductors in safety-critical, high-reliability types of situations, and the pool of candidates is pretty small. You can never have enough engineers, but you always have to make do with the budget and headcount that you have.”

Test engineering isn’t just one thing, either. There are many types of test, from lab to fab. It plays a role at the very front of the design cycle, where a test strategy needs to be incorporated at the architecture stage, or very soon afterward. If a device cannot be tested, it will never make it to market.

“Definitely, it is a skill set where it gets pushed a lot to be very productive, but engineering isn’t unique in this,” said Knoth. “This is pretty much across the board. There’s a high degree of efficiency and a big attention paid on margin, overall. And so people always have been struggling with how to do more with less. We could definitely use more test engineers, but the bigger question is whether companies can hire as much as they want. Looking at some of our partners who are trying to hire for these positions, though, it takes a long time to fill. It takes them a really long time to fill.”

The new stuff
What most engineering managers didn’t account for is just how much of an impact machine learning and data analytics would play in the test world. Test data always has been important for quality control, but there are many ways to use that data more effectively than in the past. Understanding how to utilize that data has become a focal point throughout the manufacturing flow, with an increasing amount of it being looped back to other processes in real-time.

“To control the quality of the metadata, it really has to be done at the test, by the tester, and with some kind of interface to the company’s MES system,” said Keith Arnold, senior director of solutions at PDF Solutions. “You can take as much of that away from the operator as possible, and the test programmer, and just say the test program doesn’t need all this information. But we do need to have this information. It has to get populated properly because it is too free-form at this point. You want to be able to control the test flow, you want to be able to control it even dynamically while the tester is testing.”

Data is playing a larger role everywhere, from design through manufacturing. “Today we are producing inordinate amounts of data, driven by billions of devices connected to the Internet. The future is really about extracting value from this data,” said Nieto.

As a result, a whole new skill set is being layered upon other layers of skills. And with much of the advanced manufacturing being done in places like China, Taiwan and South Korea, finding qualified people outside of those areas is becoming much more difficult than in the past.

Crossovers and adjacencies

Test is starting to push beyond its traditional boundaries in other ways, as well. This is particularly true for security, which has been intricately bound with test in the past, particularly around counterfeiting.

“Cybersecurity has been really neglected in semiconductor design,” said Andreas Kuehlmann, CEO of Tortuga Logic. “There are very limited skills in security. How do you develop secure chips? How do you test secure chips? And how do you prove it is good silicon? This is same trend I’ve seen in software. Years ago we were running around saying software developers don’t know anything about security or how to develop secure software.”

To get from point A to a secure point B, Kuehlmann said that security expertise needs to be embedded with the design, verification, test, and manufacturing teams in a federated system of security responsibility. The team has to understand who is responsible for security — someone has to have it in their title — yet everyone needs to understand security.

“We created a new job title security application engineer that’s really dedicated for the field, working with customers,” said Kuehlmann. “Security is actually an organizational and people problem. A tiny mistake design can have catastrophic consequences. You just get a few lines of code wrong and you have a big opening for cybersecurity. But security was a side job of the testing team to do security testing. We are seeing really an exponential growth right now.”

Test engineers need security expertise. But so do verification engineers. And both verification and security engineers need a working knowledge of test. “This affects everything from just regular beta testing to actually doing penetration testing on the chip of the systems,” he said.

Another area that intersects more closely with test engineering is low power. “We’re at a beautiful place where we’re able to do something about power,” said Knoth. “That’s just one subset of this broader topic of data analytics, where we create reams and reams of data in the process of designing semiconductor products. The whole point is now building the more efficient software and systems to do something smart with all that data.”

The power area has its dedicated job titles. While none of them specifically crosses into test, increasingly there is a focus on making sure a device can be proven to work as expected. “I’ve seen a lot of different titles like, power architect, power convergence lead, powers czar,” he said. “It all varies depending on how creative the hiring managers are, but essentially what it gets down to is an engineer who has one foot in the implementation space and one foot in the verification space. They have to be able to understand the challenges of the people who are writing the RTL, who are actually designing what’s going to become the semiconductor from a functional standpoint, as well as the people who are writing the software that’s going to be on that semiconductor device, because what sort of software they’re running on the product will greatly impact how much power it’s going to burn. They have to understand simulation, they have to understand emulation, they have to understand the RTL to GDSII flows for actually implementing the design to make sure that all the good intentions that were set up by the architects actually translate themselves into silicon. It’s a very broad skill set that requires depth in certain areas, so that it’s more than just understanding the high-level concepts, but it’s actually knowing how the second order effects could manifest themselves in bigger problems down the line.”

A good job
While many sectors continue to experience ups and downs, the semiconductor industry has been remarkably stable, in part because of all the new markets that rely on ICs. Chip demand is forecast to increase 8.4% in 2021, according to the World Semiconductor Trade Statistics (WSTS) organization. Yet a shortage of engineers combined with technological challenges — more data analytics, automation, smaller nodes, more than Moore techniques, safety critical, security, 5G mmWave — adds complexity to jobs, along with lots of new opportunities engineers and technicians.

The industry has been talking about a shortage of talent for years. And as the chip industry expands across new markets, that shortage is becoming more acute.

“On a bigger-picture level, our industry is hard to staff in general due to, 1) people not knowing about it or having incorrect assumptions about the work; 2) losing engineers to software and social media companies, as those companies and that industry is much better known; and 3) the nature of the work in our industry being more inflexible and physically challenging (e.g. working in clean rooms, working very specific shifts, etc.),” said Shari Liss, executive director of the SEMI Foundation. “But our advantage is our jobs are steady, the pay is good, and the industry offers strong career paths. Also, our industry is high-specialized. From entry-level positions like equipment technicians all the way to engineering/project management can require training and experience that is not commonly found. We are an industry operating on the edge of innovation,” said Liss.

Related Stories

Jobs

Engineering Talent Shortage Now Top Risk Factor

Silo Busting In The Design Flow

The Next Big Leap: Energy Optimization