ISSCC2018 – Future Mobile Society Enabled by Semiconductor Technology


Good morning ladies and gentlemen. I am Yukihiro Kato of Denso corporation. I’m very honored to be here with you at ISSCC. You all know this image on the screen, this
is the well-known QR code. This is how the code appears on my business
card along with my profile illustration but did you know that QR code was invented by
Denso? That�s correct, Denso invented the QR code
in 1994. This is one of many of our company’s invention. Today, I’d like to share with you some others,
today. At first, let me introduce you to Denso corporation. We became an independent company in 1949 after
leaving Toyota Motor Corporation. Since that time, we have grown to become a
leading supplier of automotive systems and components for major automakers. We have 150,000 associates in 200 group companies
around the world with sales of 40 billion dollars. Importantly, we have devoted significant time,
attention, and dollars to R&D, research and development. In fact, in 2016, 9% of all Denso�s revenue
went to R&D. Denso takes innovation seriously. In early hours, Denso made aggressive decision
to our company, to set our company up for success, including an important technical
cooperation with Robert Bosch in 1953 and the establishment of semiconductor research
center in 1968. The semiconductor research center is the origin
of the in-house IC fabrication at Denso and the start of Denso semiconductor technology;
today it�s one of the core competencies at Denso. I have spent 30 years in the field of semiconductor
at Denso and I also have been working in the field of connected vehicle and human-machine
interfaces and automated driving technology for several years at Denso. As you know, automotive industry is immersed
in once-in-a-century transformation. At the core of transformation is focused on
future mobility; specifically, in three key areas: efficient driving by electrification,
automated driving, and connected driving. First let’s explore efficient driving by electrification. As you can see here, electrification is making
progress. The reality is that Denso has been working
on electrification for very long time. We have been contributing to the development
of HEV and PHEV in the late 1990s. If you believe what you hear on TV, or the
great internet, or in news articles, electrification is moving slowly. The reason for slow progress is that the massive
changes are needed to accommodate the changing infrastructure for EV and we still have several
battery concerns such as cost, battery charge capacity, and charging time and recycling. If we consider eco-driving from Well to Wheel,
which means total eco analysis from oil-field to fuel-economy of vehicles, EV is not a perfect
solution. All of our thermal power plants uses fossil
fuel; the situation [MOU1]and the process cannot be changed easily. In our quest to solve some of these challenges,
we have looked to other industries. Some of these electrifications comes from
railroad industry. Railroad industry has been able to improve
the efficiency by replacing GTO, gate turn-off thyristor power devices with IGBT, insulated-gate
bipolar transistor. Automotive inverter must be compact and lightweight,
so we needed to reduce the size of inverter by increasing power density. Fortunately, with the help of consumer electronics
and railroad technologies, power density of inverter has been increasing. Water cooling technology accelerate the progress,
and dual-sided cooling and unified modules also improve the power density. We expect to see more progress by SiC and
I believe corundum structure, ?-Ga2O3 has potential and will gain popularity. We have lowered on-resistance to shrink IGBT,
which in turn able unification of power modules and dual-side cooling has proven to be effective
for downsizing of inverter. We call the power module the power cards because
of its thin character, which is well suited for dual-side cooling. SiC has several excellent material properties,
which makes it an ideal option for next generation of power devices. One challenge is that SiC material is difficult
to process and its high crystal defect density has adverse effect on the yield of power MOSFET
fabrication and quality. We can solve this issue with RAF method, RAF
stands for repeated a-face. RAF method is excellent wafer growth process
that dramatically reduce the defect density and enables power MOSFET fabrication. I’d like to show you RAF wafer growth process. [Video] These dark straight lines are defects. So, this is SiC power MOSFET in 6-inch wafer. Next, let’s take a look at automated driving. For automated driving to be successful, vehicle
technology must recognize and assist multiple objects simultaneously and with high precision. Road intersections are always dangerous, even
when humans were driving. Automated driving requires the technology
that can detect multiple object in a variety of conditions. Main components for recognition are cameras,
which include image sensors, image processors, and microcomputer that calculates the output
signal. The millimeter-wave radar is used to detect
object in difficult conditions, such as complete darkness and dense fog. The radars are well suited to accurately assess
the distance. Let’s compare human driving and automated
driving. In automated driving, sensors replace our
eyesight, our brains are replaced by semiconductors, with operation conducted by actuator. First, let’s look at sense and recognition. I’d like to show you camera recognition example
using our deep neural network hardware. The top image is input picture, the middle
image is depth estimation, the bottom is semantic segmentation. This location is Akihabara in Tokyo, the holy
land of anime. This location is always crowded. The camera with deep neural network hardware
can identify the object: pedestrian, automobile, roads, and so on. Next let’s go to judgment. For example, route calculation for path-planning,
which is an important yet difficult task for automated driving. What we found is that both CPUs and GPUs are
not ideal solution for judgment of automated driving. CPUs are well suited to complex serial processing
but they are relatively slow, and although GPUs can quickly perform large-scale parallel
processing, they are not suitable for complex processing. Denso DFP is ideal for high-speed processing
with pace of natural human reflexes. GPU is good at attaining [MOU2]uniform work. On the other hand, DFP is good at mix work
because of variable border / calculation content and variable calculation scale. DFP provides exceptional processing without
wasting time. DFP handles multiple processes simultaneously
while consuming less than one tenth of GPU power with speeds higher than GPU. For route calculation, GPU analyzes all possible
directions. With DFP, we can eliminate unnecessary paths
and achieve a more efficient calculation. You can see that, an added benefit of efficient
DFP calculation is reduction of heat generation. So, last, let’s go to the connected driving. We expect connected driving to deliver many
excellent new services and values. Connected driving enhance driving support
services including mobility for elderly people, automated driving support, and traffic control
capabilities. Collective driving also offers mobility services
such as car sharing, ride sharing, multi-modal mobility. In addition, connected driving offers convenient
services including pay-in car, efficient delivery, remote update, and failure prediction. If we consider cyber side in connected driving,
we know that as the number of connected vehicles increases, the computational time will explode. This is an example calculation result of combinations
of vehicles and their many possible routes. How will we solve these issues created by
explosion of computational time? With quantum computer. We are now planning demonstration with traffic
control in Thailand using D-wave quantum computer together with Toyota Tsusho Nexty Electronics. If we see the physical side in connected driving,
vehicle structure should follow dynamic evolution of connected vehicle services. Connected driving requires connected application. To facilitate this connected application,
we must first activate plug & play function, similar to smartphones. Vehicle structure should support dynamic application
installation even in the field such as software update over the air, SOTA. Currently, application connection through
a dedicated interface allows little flexibility. We have developed service bus control mechanism
to facilitate application installation. By adopting service-oriented architecture,
including service bus control and common interface, we can enable plug & play function. This in turn allows us to enjoy flexibility
and several important services. As connected driving and automatic driving
become common, the importance of HMI will increase. The
system becomes more complex, the information that the system provide will increase. So, let’s look at Denso HMI. [Video] All information should be controlled by HMI-ECU. Safety content is very important; it should
be timely and robust. Safety content should never freeze as we sometimes
see with smartphones. This can’t happen in a vehicle, not even for
one second. Adopting hypervisor, which is virtualization
technology developed by IT industry for automotive application enables the magic of robust OS
for safety content and open OS. We will be in mass production of our hypervisor
SoC for HMI-ECU next year. This new technology is a result of collaboration
between Intel, Blackberry, and Denso. What are the important takeaways today? First, information technology is essential
for future mobility. Our challenge is solving the transition of
information technology to automotive-grade application. Automotive grade information technology will
require quality, durability, functional safety, quick correction, and maintenance services
for lives of people and long-term use. Second, Denso is aggressively developing automotive
grade semiconductor technology and systems for the benefit of future mobility. I’d like to leave you with one final thought. To effectively continue our pursuit of new
technologies and to ensure our long-term supply of parts and solution, we need to increase
long-term collaboration and partnership with IT companies and with semiconductor companies. This is critical to our long-range success. It was pleasure to be here with you today. Thank you for attention. [Applause]
[MOU1] [MOU2]

Leave a Reply

Your email address will not be published. Required fields are marked *