Designing application specific ultra-low-power circuits for energy-aware applications

Application scenarios like Internet-of-Things (IoT), wireless sensor nodes or body sensors make increasing demands to modern embedded systems. Focus is no longer solely on increasing perforance, but more and more on maximizing energy efficiency.

Long time, advances in the miniaturization of integrated circuits was only used to enhance performance by increasing clock frequency or adding functionality. Despite state-of-the-art semiconductor technologies, this well-established approaches come more and more to its limits. Modern processor architectures provide the required performance by massive, fine-grained parallelism. This allows for a high energy efficiency, as clock frequency of a parallel processor can be kept comparatively low.

Focus on the research within the team Nanoelectronic is the development of new concepts for massive-parallel on-chip-multiprocessor architectures and corresponding software-toolchains. The scalable multiprocessor-framework CoreVA-MPSoC acts as a basis. CoreVA-MPSoC describes a scalable, hierarchical communication infrastructure, which connects a configurable number of VLIW-(Very-Long-Instruction-Word) processors. A corresponding compiler toolchain allows for the automatic mapping of applications to the multiprocessor architecture.

Additional to the development of the processor architecture, team Nanoelectronic develops key technologies on the further reduction of the power consumption and optimization of energy efficiency of integrated circuits. One promising approach is the operation of transistors below the switching threshold (Sub-Threshold). For the design of such circuits, we provide libraries, allowing for a robust operation of standard cells below threshold voltage. Currently, we develop a standard-cell library in a 28nm FD-SOI technology.

Architectures developed in team Nanoelectronic are used in various research and funded projects of the Cognitronics and Sensor Systems group (e.g. KogniHome, itsowl-EE, itsowl-IV).