The goal in microelectronic systems design is to develop an optimal realization under given constraints. These can generally be quite different in nature; for instance they can cover specifications regarding functionality, limited energy resources or a fixed target technology. To master the vast number of possible design decisions, we investigate systems with respect to application, design methodology, and architecture.
In the beginning of the design the application has to be defined and analyzed in detail. Commonly, an exact specification is not available in this early stage (partly due to short product cycles) or is subject to change throughout the process. Our focus on applications is primarily on communication systems (wireless communication, coding standards, MIMO systems), which pose the designer with a large challenge w.r.t throughput and energy efficiency, especially in mobile application scenarios. Furthermore we investigate ambient systems and the acceleration of computationally demanding simulations from the financial mathematics and scientific domains.
The implementation of a system can be done in various styles and using different architectures. We develop application specific hardware and application specific instruction set processors (ASIPs) in field programmable gate arrays (FPGAs) and application specific integrated circuits (ASICs). Besides the optimization on a microarchitectural level we deal with heterogeneous systems comprising components with differing implementation styles, as well as with networks-on-chip and 3D-DRAM.
We use different methodologies, which allow for early estimation of implementation complexity, performance, reliability or other aspects of the implemented hardware. By means of rapid prototyping we can do exhaustive design-space-explorations, partially leveraging high-level synthesis tools but also manual refinement of a system from the functional model to the register transfer description of the hardware.
Digital communication systems are essential in today's globally connected world. We investigate new systems for signal processing in the baseband / PHY layer. Our research includes the whole range from communication theory and algorithms to hardware architectures and their implementation in chips. Therefore, in our department, communication engineers and hardware designers work closely together.
3D-DRAMs (WIDE I/O) are a promising approach for mobile devices like Tablet-PCs and Smartphones. In the area of servers exists a high potential as well due to the high bandwidth and energy efficiency of this new memory architecture. In our research group, we develop and evaluate novel DRAM architectures, we implement DRAM controllers and we investigate the thermal behaviour of Wide-I/O DRAM by means of virtual platforms.
Technology scaling has reached a point at which process and environmental variability are no longer negligible. We focus our research on hardware-reliability in wireless communication systems, as well as memories (SRAM, DRAM). We follow a cross-layer approach, which exploits the mutual trade-offs of system performance, hardware reliability, and implementation complexity. Wireless systems are especially suited for this approach because they have inherent algorithmic error resilience.
Energy Efficient High Performance Computing
Our research group develops and evaluates new methodologies that are helpful to design sophisticated heterogeneous platforms for high-performance computing applications. In particular we focus on accelerating application-specific arithmetic problems, so-called "kernels". For the final design, we analyze the possible acceleration factors as well as the energy saving potentials, for example when using FPGAs in financial simulations. We employ the latest development tools (High-level synthesis, Maxeler design flow, Vivado design flow) as well as the most recent hardware platforms like the Xilinx Zynq EPP.
Wireless Sensor Networks collect information from the real world and form the backbone of future services (Internet of Things, Cyber Physical Systems), which will help us in our daily private and work life. We view this topic from different perspectives in an interdisciplinary way in our research, develop solutions and evaluate their relevance in practical use.
Today's companies have to deal with complex hardware architectures such as heterogeneous multi-core systems. Therefore, new development tools and approaches such as Virtual Prototyping are needed for efficient and fast design on electronic system level. In our research, we use SystemC and gem5 based virtual platforms for a thorough design space exploration on software and hardware level.