Feasibility study of PERCIVAL Data Acquisition Backend Architecture

The PERCIVAL soft-X-ray (250 eV 1 keV) image detector project is a collaboration between DESY, STFC, Elettra Sincrotrone Trieste and Diamond Light Source. The objective of the project is to develop a back-thinned CMOS detector which outperforms present soft-X ray image detector technology, in terms of sensor size, noise, dynamic range and frame rate. The size of this 13M pixel imager associated with its 120 frames per second frame rate impose very challenging requirements to the Data Acquisition Backend of the system. A DAQ backend system architecture, using a commercial deep-buffer switch to rearrange image data streams coming from different regions of the sensor via several 10Gbps Ethernet links has been proposed to reassemble image frames. Real-time data processing is to be performed on multiple, parallel commodity compute nodes. This contribution to the conference reports on benchmarking tests performed as a feasibility study, and presents the resulting recommendations for the system architecture of the PERCIVAL detector DAQ backend. The feasibility study covered three key issues: Reliably moving data in UDP packets from multiple 10Gbps Ethernet links from the DAQ front-end electronics to commodity compute nodes; Real-time processing on the compute nodes; and finally streaming data to a central parallel storage system.