RESEARCH GROUP Cognitronics & Sensor Systems

Designing efficient reconfigurable systems for cutting-edge compute and embedded applications

The reconfigurable systems team bundles all key competences for the design of efficient compute and embedded applications. State-of-the-art system design expertise is used to integrate FPGA, CPU and GPU technology from various vendors to fulfil the requires from a wide range of applications, covering IoT and cyber-physical systems, cloud computing, machine learning, image processing, as well as HPC. Using FPGA technology, we cover the complete range from regular HDL-based design, dynamic run-time reconfiguration, low-level design techniques (e.g. manual routing) as well as high-level synthesis (e.g. block-set, C, or OpenCL based). The team has developed and applied specialized low-level placement and routing tools which are used in the design of fault tolerant, run-time reconfigurable FPGA-based systems, e.g. for space applications.

For prototyping and FPGA computing, the RAPTOR FPGA prototyping platform has been developed by the team. The modular and scalable architecture, integrating a flexible, high-bandwidth communication infrastructure allows easy realization even of large, complex multi-FPGA designs. The team has expertise in the design and application of high-speed serial links, which are used not only between FPGAs, but widely throughout the industry, e.g. for PCIe, 40G Ethernet, SuperSpeed USB, or Infiniband.

On PCB level, the team constantly uses state-of-the-art HDI multilayer technology (> 24 layers). Sophisticated buried/micro VIA technology, small pitch (< 0.4 mm) BGAs as well as line/space structures of less than 50 um are enablers for realization of dense microelectronic systems. The PCB is considered not just as an interconnection element between the different components, but as an essential part of the overall system design. During design and verification of a PCB, different tools are used to ensure the correct behaviour of the final system under all environmental conditions covering thermal, as well as signal and power integrity aspects.