AMIGA 7. Entornos extremos de galaxias con los precursores de SKA: desde el diseño del flujo de datos hacia su construcción. Procesado de datos en hardware

Project details

Year
2019
Leader
Carlos Carreras Vaquer
Funding Entity
Ministerio de Economía y Competitividad, Proyecto Retos de la Sociedad

Project detail

The AMIGA 7 Hardware Data Processing subproject investigates circuit synthesis and optimization techniques that implement both signal processing algorithms from the Central Signal Processor (CSP) and scientific algorithms from the Science Data Processor (SDP) and the Square Kilometre Array (SKA) Regional Radio Interferometer Centers in reconfigurable hardware. AMIGA 7 will span a key period for SKA, following the completion of consortia designs and the success of their Critical Design Reviews, leading up to the start of construction.

Reconfigurable devices are firmly embedded in the architecture of SKA’s CSP modules, where they are widely used. DSP circuits typically support the use of fixed-point formats that offer higher performance and lower cost and power consumption than their floating-point counterparts. The project proposes the application of signal sensitivity analysis techniques based on their dependencies to optimize automated methods. The application of these methods will enable new levels of optimization of CSP modules as they evolve toward SKA construction.

Furthermore, the potential use of highly energy-efficient FPGA-based accelerators for scientific computing at SKA Regional Centers is of great interest to reduce their high energy costs. In this context, the project continues the development of the Witelo environment for the automatic synthesis and generation of VHDL for high-performance FPGA implementations, thereby overcoming the limitations of existing

On a second front, the project evaluates the feasibility and efficiency of FPGA designs for SKA scientific algorithms. Specifically, the priority target is the artificial intelligence algorithms used in the analysis of images obtained with SKA, given their potentially high level of parallelism.

Source: Observatorio de I+D+i UPM 

Related Projects

View all projects