Looking back 20 years hence, how will this period be viewed? The question was posed by the CEO of imec, Luc Van de hove, during his opening talk at a day event imec organised in Tel-Aviv.
For Van den hove, this period will be seen as one of turbulent technological change. “The world is changing at an incredible rate,” he says. “The era of digital disruption is changing our industry and this disruption is not going to stop.”
It was the Belgium nonoelectronics R&D centre’s second visit to Israel to promote its chip and systems expertise as it seeks to expand its links with Israel’s high-tech industry. And what most excites imec is the Internet of Things (IoT), the advent of connected smart devices that turn data into information and adapt the environment to our needs.
The world is changing at an incredible rate. The era of digital disruption is changing our industry and this disruption is not going to stop
Internet of Things
Imec is focussing on five IoT areas: Smart Health - wearable and diagnostic devices, Smart Mobility which includes technologies for autonomous cars, drones and robots, Smart Cities, Smart Industry and Smart Energy. “In all these areas we look at how we can leverage our semiconductor know-how,” says Van den hove. “How we can bring innovative solutions by using our microchip technology.”
The broad nature of the IoT means imec must form partnerships across industries while strengthening its systems expertise. In healthcare, for example, imec is working with John Hopkins University, while last October, imec completed the acquisition of iMinds, a Belgium research centre specialising in systems software and security.
“One of the challenges of IoT is that there is not one big killer application,” says Van den hove. “How to bring these technologies to market is a challenge.” And this is where start-ups can play a role and explains why imec is visiting Israel, to build its partnerships with local high-tech firms.
Imec also wants to bring its own technologies to market through start-ups and has established a €100 million investment fund to incubate new ideas and spin-offs.
Imec’s expertise ranges from fundamental semiconductor research to complex systems-on-chip. It is focussing on advanced sensor designs for IoT as this is where it feels it can bring an advantage. Imec detailed a radar chip design for cars that operates at 79GHz yet is implemented in CMOS. It is also developing a Light Detection and Ranging (LIDAR) chip for cars based on integrated photonics. Future cars will have between 50 to 100 sensors including cameras, LIDAR, radar and ultrasound.
The data generated from these sensors must be processed, fused and acted upon. Imec is doing work in the areas of artificial intelligence and machine learning. In particular, it is developing neuromorphic computing devices that use analogue circuits to mimic the biological circuitry of the brain. Quantum computing is another area imec has begun to explore.
One of the challenges of IoT is that there is not one big killer application
“There is going to be so much data generated,” says Van den Hove. “And it is better to do it [processing] locally because computation is cheaper than bandwidth.”
Imec envisages a network with layers of intelligence, from the sensors all the way to the cloud core. As much of the data as possible will be processed by the sensor so that it can pass on more intelligent information to the network edge, also known as fog computing. Meanwhile, the cloud will be used for long-term data storage, for historical trending and for prediction using neuromorphic algorithms, says Van den Hove.
But to perform intensive processing on-chip and send the results off-chip in a power-efficient manner will require advances in semiconductor technology and the continuation of Moore’s law.
Imec remains confident that Moore’s law will continue to advance for some years yet but notes it is getting harder. In the past, semiconductor technology had a predictable roadmap such that chip designers could plan ahead and know their design goals would be met. Now chip technologists and designers must work together, a process dubbed technology-design co-optimisation.
Van den hove cites the example of imec’s work with ARM Holdings to develop a 7nm CMOS process node. “You can create some circuit density improvement just by optimising the design, but you need some specific technology features to do that,” he says. For example, by using a self-alignment technique, fewer metal tracks can be used when designing a chip's standard cell circuitry. "Using the same pitch you get an enormous shrink," he says. But even that is not going to be enough and techniques such as system-technology co-optimisation will be needed.
Imec is working on FinFETs, a style of transistor, to extend CMOS processes down to 5nm and then sees the use of silicon nanowire technology - first horizontal and then vertical designs - to extend the roadmap to 3nm, 2.5nm and even 1.8nm feature sizes.
Imec is also working on 3D chip stacking techniques that will enable multi-layer circuits to be built. “You can use specific technologies for the SRAM, processing cores and the input-output.” Imec is an active silicon photonics player, seeing the technology playing an important role for optical interconnect.
Imec awarded Gordon Moore a lifetime of innovation award last year, and Van den hove spent an afternoon at Moore’s home in Hawaii. Van den hove was struck with Moore’s humility and sharpness: “He was still so interested in the technology and how things were going.”