Two new 6G professorships as a corporate donation to the University of Oulu
The Professors have started in their positions. Lauri Koskinen has been hired to Circuits and Systems research unit (CAS), and Zaheer Khan to the Centre for Wireless Communications (CWC). Both professors’ background is briefly presented below.
The need for SoC specific professors in Oulu is immense.
“The Oulu ICT ecosystem has been among the world leaders in the creation of wireless technology for 3G, 4G and 5G. Now we are pioneering in 6G that will change the world again, probably even more than the previous cellular generations have done. The system on chip technology is the workhorse enabling 5G to wide use now, and it will be doing the same for 6G in the 2030s”, says Professor Markku Juntti from the Centre for Wireless Communications, University of Oulu.
The new professorships are established with the support of Nokia, Nordic Semiconductor, MediaTek and the City of Oulu.
Improving computing platforms for wireless algorithms is key
Zaheer Khan, the newly appointed professor in System on Chip Technologies (SoC) at the University of Oulu, wants to make the most of the connection between academia and industry.
The professorship has two goals, as Khan sees it.
”The first is you make people in research more familiar with SoC architectures and intellectual property core designs used in SoCs. The second is that we train people in research in a way that fits the needs of industry,” he says.
Khan is looking to nurture a group of experts in the University who have a strong connection to electronics, computing, computer science and wireless technology. Algorithms are only a part of the equation, says Khan, and he wants to emphasise the importance of platforms now and in the future.
Back in the 4G era, Khan's primary research interest was working on how to make networks more adaptive. He received two PhD offers, one from Paris and the other from Oulu.
”I had lived in Sweden for three years when I got my Master's degree there in electrical engineering, so I knew about life in a Nordic country. In the end we decided to come to Oulu,” Khan says.
Khan's main research interest was working on how to make networks more adaptive. Game theory, a concept from economics, was a hot topic, which meant looking at networks as strategic entities that can intelligently adjust to their environment. This was the topic of his doctoral thesis, which he completed in 2008 in Oulu.
After his doctoral degree, he got funding from the Finnish Academy and collaborated with professor Matti Latva-aho. By this time, 4G research was going forward to the next generation, 5G.
”During this period I changed my research focus because I saw that there is a lot more potential on how we use different computing platforms for wireless algorithms. Theoretically we can hypothesise about wireless algorithms, but how could we design and prototype them for real systems? So, I started working on embedded systems, because when you go to a real algorithm which is designed and implemented on a chip, there are real constraints such as area and energy in addition to performance.”
Khan started playing around with FPGAs, or Field Programmable Gate Arrays. These are semiconductor devices which can be reprogrammed to a desired application or functionality and can be used to simulate how wireless algorithms perform. Eventually, the project he was working on came to a close, and he took a leave of absence from the University of Oulu to join a Tenure Track Position at the University of Liverpool, England. During his work in Britain, he moved from working with Wireless Access Research Platforms (WARP) to more advanced Zynq-based wireless platforms for his work.
During his time in Britain, Brexit happened, and Khan and his Finnish wife had difficulty adjusting to the change. They started figuring out their next step and eventually decided to move back to Oulu.
When Khan returned, he got a project grant with one of his Finnish colleagues, Janne Lehtomäki, and they spent a lot of time working with FPGAs again. However, Khan felt that besides funding and academic research, he should gain a lot of direct industry experience.
”There can be a strong bias between academia and industry. For instance, from an academic point of view you will look at algorithms from the perspective of performance. In reality you have to look at the chip, the area that's available, the energy that is available, and all those things add up,” Khan explains.
He applied for a position at Nokia and started working as a Technical Lead in Nokia's SoC Architecture and Specification Group. While working full-time for Nokia for more than a year and a half so far, he has also contributed to research at the University of Oulu.
”It is really fortunate that both people in Nokia and people at the University realise the need for this two-track approach. Both parties have been very understanding, and I feel that in general the collaboration between industry and academia is improving,” Khan says.
For 6G, the low-hanging fruit of technology have already been picked
While at Nokia, he was working on issues with practical constraints. He as well as academia and industry are shifting their gaze to 6G, the next generation of telecommunications technology. Khan is keen to stress the computing platform and its significance.
”For 6G, my question is how we can make algorithms more efficient given that computing platforms are changing very fast. This is extremely important. To keep the pace with platform development, we need to design algorithms from the platform's perspective and improve the computing platform for wireless algorithms.”
For every new 'G', two things have happened, says Khan. You can have more antennas and you can have more bandwidth. Both of these things mean you have to be able to handle much more data than before.
”But from an SoC perspective you don't have actually more margin in terms of energy in 6G,” Khan explains. ”In the past, when we transitioned toward 7 nanometer (nm) process technology, we saw anticipated gains in terms of area and energy. However, 5 nm chips do not seem to perform in terms of area and energy efficiency as well as they should. As chip makers are trying to transition from 5nm to 3nm technology, transistors are reaching a tipping point and the gains in terms of energy efficiency tend to decrease. The same can be said about different generations of memory.”
As the easy gains seem to have been reaped already, 6G presents new challenges, Khan muses.
”We need to come up with very clever strategies to process more and more data efficiently in 6G SoCs. Because we can't expect a smaller area of the chip to carry more and more transistors efficiently.”
The power of more with less: Energy efficiency still has room to develop
Lauri Koskinen, the freshly appointed System on Chip (SoC) technologies professor at the University of Oulu, takes energy saving very seriously. Koskinen has worked on making processors use as little energy as possible for nearly 20 years as a researcher and in the industry.
Processors play a big part in devices such as mobile phones or in-ear noise-cancelling headphones, and there is still significant energy savings to be made. According to Koskinen, this is surprisingly easy.
”Saving energy in microchips is very simple”, Koskinen says. “You cut the voltage by half, and you spend only a quarter of the energy you were using previously.”
Koskinen is making light of the situation, of course, but his comment illustrates a fundamental issue in energy efficiency at the processor level. At a very basic level, it is that simple, but numerous other issues come into play: outside variations in temperature, variations in the semiconductor substrate, and so on. These variations mean that while the chip could be made more energy efficient in a simple manner, there is no guarantee it will still do what it needs to do reliably and quickly.
Feedback loop for energy management
While Koskinen was working on this problem in 2008, he landed on his first big project to develop a technology that would save processor energy and still have it operate as designed. Having secured funding from the Academy of Finland and Business Finland, he made strides in developing this technology with his research group.
“Simply put, our solution is a feedback loop. It gauges the chip's performance constantly and makes adjustments as necessary. The processor always uses as little energy as possible but is supplied more energy if the situation requires it. It's kind of like the traction control system in cars: when the road is straight and dry, you can drive fast, but when you're approaching an icy bend, the system will recognise the situation and react accordingly,” Koskinen says.
The technology was so sound that Koskinen founded a company, Minima Processor, with his partners Toni Soini and Jani Mäkipää in 2015 to take it to market. During this time, he gained substantial experience in business and how the industry worked. Koskinen was a visiting Fulbright researcher at the UC Berkeley Wireless Research Center in the early 2010s when one of the professors there said something that resonated with Koskinen later.
“The professor founded a start-up company and worked as a CTO himself. Very quickly, he said that he did not 'speak accounting', which is a very familiar situation many new technology entrepreneurs find themselves in. Technology is their strong suit, but running a business is another thing, and I am still perplexed as to why this transition is not addressed more in our education. Every start-up will face difficult times that people could be more prepared for,” Koskinen muses.
Microelectronics do not exist in a vacuum
Koskinen says that the most important lesson he learned at Berkeley – even more important than technology – was collaboration and openness. This is why in his professorship in Oulu, he wants his students to be able to utilise their knowledge across different courses and disciplines.
“Microelectronics in general and university courses in particular should not exist in a vacuum. I want students to be able to use the information they gain, say, in health technology, to develop integrated circuits in my courses. After you learn the basics, it is about applying and integrating your knowledge across the board.”
Koskinen is splitting his time 40-60 between the University of Oulu and Bosch Sensortec, a world leader in micro electrical, mechanical systems, or MEMS. For example, a simple MEMS sensor will sense if a device such as a cell phone is horizontal or vertical and adjust the display orientation accordingly. More complex sensors detect, for example, rotation and gas properties.
Sensors will only become more intelligent in the future, which means they will need calculating power instead of dumping raw data into the cloud. And this is where energy efficiency becomes critical yet again.
“Even if data transfer rates are increasing significantly, we will need local computing power. If a sensor senses data 10 times a second, it will not result in a large amount of data being transmitted to the cloud. Sending small batches of data over the network in real-time is very energy expensive compared to sending large amounts of data at once, like with video. Also, if we look at 6G and beyond, the algorithms will only grow more complicated. This means we need high computing capabilities with very little energy.”
Koskinen thinks a lot about the human brain to develop energy-saving technologies further.
“The brain is not an exact 'computer' but constantly deals with noise. It is highly energy efficient, however. A human brain can recognise a cat reliably using a light bulb's worth of energy when a server farm will use megawatts to do the same.”
So, while we are developing new technological strides to enable a vision of an interconnected future with superior energy-efficient computing power, nature remains a source of inspiration and awe.
“A housefly in flight is constantly changing and correcting its course using what can only be milliwatts of energy and doing it far better than any human-built autonomous system so far,” Koskinen says.