The Need for Speed: Strategies for Data and Testing in Semiconductor Manufacturing

Despite recent supply-chain disruptions, semiconductor buyers are increasingly spoiled for choice when it comes to chip manufacturers. From a chipmaker’s perspective competing for “sockets” in the next gadget — whether it be a mobile phone a VR headset, or an EV — is a game of speed. As a result, competition among producers continued to reduce product development and delivery times — once years to now a few short months.

Now, clients expect faster delivery times as a prerequisite for their chip providers. At the same time, the chips themselves, and the uses for chips, have become exponentially more complex. Chip integration — packing more features into the same silicon — is often a key differentiator for features. However, more complex chips have more complex development, test, and manufacturing challenges.

To respond to these challenges, semiconductor manufacturers must deliver speed and manage complexity. Both hinge on having solid data and testing strategies. Companies that streamline their data and testing strategies the right way can signifi cantly speed up the time it takes to get complex chips to market.

The companies that get it wrong could see delays, with chips returning for redesign after a lengthy testing process. Longer term, these firms could be edged out by faster and more efficient competitors.

The World Has Changed

In addition to buyers demanding faster turnaround times, the testing requirements are becoming more complex. Especially in the transportation and wireless sectors, test complexity is increasing because:

• Chips are packing more and more functionality into the same footprint as designers look to conserve space in designs.

• There is a heavy focus on reliability in the transportation sector. (The stakes are higher for chip failure in a vehicle traveling 80 miles per hour than, say, a smart home speaker.) This, in turn, requires more test cases and larger test sample sizes.

• There are also more potential uses of chips. For example, consider that the same modem in your 5G phone might also be in the dashboard of a car. This means that the chips need to be tested for multiple applications and environments.

Despite these challenges, there are ways for semiconductor companies to deliver finished chips faster, in a way that increases scalability and efficiency.

Still, it’s not just a matter of doing more of the same things faster. Trying to increase throughput using the old methods is a recipe for pushing manual testing to the breaking point.

How Manufacturers Can Adapt

Semiconductor manufacturers need new strategies and approaches to survive, and thrive, amidst the emphasis on speed:

• Test Automation. This involves moving away from manual testing and toward automating repetitive measurements using software-defined functionality. Chipmakers utilizing automated testing typically cover more test cases than organizations doing testing manually — and they also do it faster.

• Standardization. Companies must ensure that comparing test results from one organization to the next isn’t like comparing apples and oranges. Standardizing on common platforms of test hardware and software not only makes these comparisons easier but also makes testing go faster. Organizations who have standardized test systems report that new teams can go from receiving a part to producing a test report much more quickly.

• Centralization of Data and Analysis. Data is worthless if it can’t be found or if it doesn’t have the necessary context. Centrally storing and managing data is vital to compare measurement results from simulation to validation to manufacturing data. Centrally managing test data is often a key step that enables using test results to improve product designs.

The move toward automation has obvious, immediate benefits by speeding up testing processes and adding to scalability and efficiency. But this is just the first step.

To remain competitive, chip manufacturers need to take the next two steps to transform into data-centric organizations, leverage data to speed up design approvals, and identify errors more quickly, earlier in the process.

The bottom line is that, for leaders in this space, testing is increasingly viewed not as a necessary evil in the design process but as a business value driver, where measurement data can be used to improve product design.

Common Challenges and Solutions

To be sure, adopting new data management and test standardization strategies can pose challenges for many chip manufacturers, especially those that have grown their businesses rapidly through acquisitions. We often find in large chipmakers that each site will use a different approach to validation. This is especially true with test software as software choices are almost religious in nature. One can imagine that adopting a new test software or programming language is about as difficult as convincing someone to change their religion.

To overcome this, companies need to develop a testing framework that can connect different languages and systems to provide a standard across the organization, hopefully one that allows test code to be used and re-used.

The need for a measurement framework is why we built the MeasurementLink capability at NI. This makes it easier for engineers to stitch code in LabVIEW, Python, and C# within a single measurement framework — allowing for software re-use without requiring a change in religion.

Investing in centralized test tools like a measurement framework isn’t trivial. The cost of investment means management needs to be sold on their potential before any change can occur.

While the semiconductor industry has been a hero driving data-centric approaches in key areas like manufacturing, the shoemaker’s children sometimes go barefoot. This is because the move to data-centric semiconductor validation isn’t always an easy sell.

That’s why it’s often best to start with test automation solutions that provide software-agnostic approaches to lay the groundwork for standardization and data centralization down the line.

Speed Up, or Get Run Over

Engineering teams are changing their approach to validation tests and becoming more organized. Conversely, the era of engineers doing manual tests and writing a report three weeks later will soon be behind us.

To survive, companies will need to embrace intention-built testing, architecting the measurement hardware and software in the lab to drive more scalable, lower-cost, and faster processes.

Speed has gone from a way of attracting new customers to a prerequisite for keeping them. Firms that don’t embrace speed won’t come in second in this new semiconductor race. They’ll be left behind. Or run over.

This article was written by David Hall, Global Go-to-Market Leader, Semiconductor, NI (Austin, TX). For more information, visit here  .

---