Dr. Hans Rijns, Vice President, Head of Research, for NXP Semiconductors recently spoke to EDN Asia about some of the challenges and issues in semiconductor design, and how design and R&D are changing as they evolve past the traditional definition of Moore’s Law, and go to “More than Moore”. Excerpts:
Moore’s law—until when do you think it will hold?
The famous question. As we all know, it’s been challenged every two or three years mainly based on arguments of cost and affordability. A different take on this question is the aspect of reliability of which I’ve recently given a key note presentation at ASQED. In today’s 45nm or 40nm technology, we have devices with gate oxide thickness of only six to seven or eight atom layers and with diffusion areas containing only about 70 to 90 doping atoms. Its clear that any random fluctuation in these key transistor characteristics is going to kill the matching properties of these devices. So this means that reliability, affected by these variability effects, is going to be a dominant challenge if we keep on reducing the feature sizes of our process technologies. My point is the discreteness of the charge particles is going to be an intrinsic source of inaccuracy and therefore reliability of these devices.
Let us think about truly dependable electronics that we can use in life-critical applications; for instance automotive drive-by-wire functions or biomedical applications, where our lives are critically dependent on the reliability of those semiconductor-based products. There is of course a part of the circuitry that executes the basic functions, but we will also need more and more supporting functionality in there—all types of auto-compensation, auto-correction circuitry—just to make sure that those devices are operating reliably in these deep sub-micron technologies. Obviously this supporting circuitry adds to the silicon cost and power consumption of the overall product and is essentially a negative offset against the positive benefits of the process technology shrink path.
From a technology perspective, process technology scientists will definitely continue to find solutions in the years to come to maintain the “more Moore” roadmap, driven by the demand for high-compute density SoCs, but at skyrocketing costs. But from a reliability point of view, it’s going to be an increasing challenge to make reliable products in those nodes.
What do you think should semicon companies do to address the skyrocketing cost levels of SoC development in advanced CMOS technologies?The only thing that the industry can still do is to try to make sure that we do not have a sequential product development cycle, wherein semiconductors are being designed, then validated in the system, and then released. We have to understand that those types of product cycles are simply not meeting the consumer markets’ expectations. So it’s primarily a matter of making sure that the semiconductor design teams, the applications teams and also the EDA world is working more concurrently together—that is “parallel” in time, rather than sequential.
On the other hand, I think there is little room for fundamental change to the intrinsic costs that we are talking about. The explosion of cost is there, so it’s a matter of those teams who play there to try to optimize the time axis.
I believe on the other side, wherein we focus on applications where we see a wider technology mix—high-voltage, analog mixed signal, RF, sensors and actuators—there we will see a very broad spectrum of semiconductor based technologies. But not in the most aggressive technology nodes per se but in technology choices optimized for cost and performance
Please briefly discuss NXP’s “More than Moore”.The advanced CMOS technologies—those are really only applicable for the sheer-compute-power-type SoC applications. Those applications are typically consumer electronics related —mobile phones, portable devices, digital video processing, etc.
On the other hand, we see ‘more than Moore’ for the need for devices that connect us human beings to the world outside and improve the quality of our lives. For applications like environmental aware systems, personal healthcare, automotive safety and traffic management systems, we need devices with sensors and actuators with incorporated security and safety measures, needing a mix of process technologies. It’s more about high-performance mixed signal functions than purely digital compute systems.
NXP is one of the foreground players in that field. We are very actively stepping into those application areas, where possible with an industry co-development perspective in mind. For example, we just finished a very interesting pilot in the Netherlands, an intelligent traffic management system we have developed together with IBM. This is a device built into your cars, in a secure way and with respect of one’s privacy that is able to track the car for security purposes and / or for tolling purposes. That device integrates GPS, GSM, NFC and dedicated crypto / security technology into one single product.
Another example is a very early activity that we have in solar energy, in which NXP is looking not at the power efficiency of the panel’s individual cells, but to the power efficiency—and power management—of the overall PV system.
Perhaps, a third example is this: for many years, we have been working very intensively on embedded power management, mainly in the mobile area. We brought that into a new application area which, in cooperation with Swiss company Phonak, resulted in world-leading hearing aid solutions now in the market.
Do you think Asian IC design continues to be in the low-end compared to Europe or the U.S.?Historically, the Western world looks at Asia for low-cost test, assembly, and qualification work. The prime reason to be here is cost, but personally I’d like to take a different perspective to Asia. In fact, we also would like to recognize that some of the applications that I mentioned are simply driven more strongly out of Asia than out of the western part of the world. Not always of course, because there are still fundamentally different end-consumer needs across the globe. These initiatives also are not always driven from the perspective of consumer requirements, but in some areas local governments have different sets of arguments to stimulate innovation of new applications in a particular field.
My point is to let us have a global perspective to where those types of new applications are being pioneered, and after we have identified that, to connect, also the R&D, innovation, and design, to where the action is. For that reason, and being responsible for NXP Research, I am now looking into Asia where those types of opportunities exist.
