ITC keynoter de Geus predicts the DUT will include a human

Seattle, WA. “You have tested positive—for complexity, that is,” said Aart de Geus, chairman and co-CEO of Synopsys, in his Tuesday morning keynote address at the International Test Conference. He last spoke at ITC 20 years ago, he said, adding that in looking back he is “…blown away by how incredibly well test has kept up with this enormous change we’ve seen.”

For example, he said, the arrival of the FinFET is opening doors leading from the horizontal transistor structures of the past toward vertical structures.

Means of technology improvement have remained the “same old, same old—10X improvement here, another 10X improvement there.” And technology drivers have remained the same: performance, power, area, and yield—yet today’s technology is dramatically more complex, he said. We employ IP reuse, moving up the hierarchy as systemic complexity becomes much more multidimensional.

Synopsys, de Geus said, is fortunate to have gained new insights into the complexity of today’s silicon technology by touching almost all new chips engineers are designing. Design is active at 10 nm, and work is being done at 7 and 5 nm—“That’s astounding yet incredibly promising,” he said, predicting that the FinFET generation will grow rapidly.

But the new generation will present subtle test challenges from a manufacturing point of view: metal shorts, open vias, and litho violations plus resistive issues. “Resistive is a four-letter word” he said, when it comes to testing memory. “Resistive effects of cores are exactly what we don’t like in digital world.” And for logic, “transient is a four-letter word.”

In today’s environment, he added, we must handle more defects, but run time can’t go up. “The advances to keep up are truly remarkable,” he added, explaining that we must look inside cells—not just at the terminals.

As for lithography, he said, we are trying to draw 10-nm lines with 193-nm crayons. That requires three illuminations, with triple-patterning entailing significant computational complexity to ensure that phases are aligned properly on successive illuminations.

Next, he traced the evolution of design and test. Synthesis was the design step that automated what people had been doing manually. Physical optimization could create circuits with lower congestion. And adding placement into synthesis enabled the prediction of distance and timing.

Similar breakthroughs in test included ATPG, scan testing, memory BIST (bringing test smarts on-chip), test-vector compression, core-based IEEE 1500 SoC test, and memory self-repair—an avant-garde approach that built in test as well as a fix.

Meeting past and future challenges requires an ecosystem, he said, from the foundries through the EDA world. And it’s important to keep in mind that Moore’s Law is as much an economic model as a technology model. “Cheap transistors have driven an entire field,” he said. “If we can continue to deliver smaller transistors, incredible applications are possible” through what he called “techonomics.”

To make money, de Geus said, we need to look further up to the application world of (among others) Amazon, Oracle, Facebook, Linked-in, Disney, Deutsche Bank, and Yahoo! And there is almost a religious aspect as we look to the deity in the sky we call “The Cloud.” But in pursuing new applications, the past focus on “sooner, better, cheaper” won’t be sufficient. He discussed differentiation and value, noting that value gets created only if it changes the opportunity space of your customer—making your customer more differentiated.

He traced the evolution of electronics technology from computation through mobility and convergence (the combination of computation and mobility) and on to social networking. “In my family,” he said, “SMS is only thing that brings response in seconds.” He added that video (with its bandwidth demands) is a gift to our field. So too will be the ubiquitous sensors and actuators that make up the Internet of Things, or as he prefers to call it, “smart everything.”

There are potential downsides, he said, envisioning a police report stating, “The perp came in through the toaster.”

Next on the horizon is the smart brain. Already, he said, it’s been demonstrated that a paraplegic equipped with sensors that can pick up brain waves can control an artificial hand. From a test perspective, he said, controllability and observability apply not just to the electronics, but to the DUT as a whole—and that includes the human.