2026-11-02 –, Conference Foyer 2
The four-year, 10M€, RADIOBLOCKS proposal was submitted by a consortium of 33 partners on 15 April 2022 and already accepted by 26 July. Work started 1 March 2023, which means the project formally ends 28 February 2027 - just a few months after this ADASS conference.
This timeline allows me to discuss the actual outputs realised by the project because all the work is realistically already finished by now.
RADIOBLOCKS aimed to improve key European radio-astronomical facilities (EVN, LOFAR, NOEMA, the Effelsberg-, Sardinia-, and Yebes radio telescopes, with a side-order of ALMA and SKA) along ALL the parts of the signal chain with the aim of improving field-of-view, sensitivity, and post-processing. A part of the project dealt with real hardware (actual receivers and hard-core semiconductor development), but with all signal chains moving from the analog to the digital domain asap, the project improves Digital Signal Processing across the board - in the widest possible meaning of the concept.
Of particular interest to this audience would be the (extremely) high throughput data transport (200-400 Gbps) straight into GPU memory, GPU-based processing (amongst many others replicating the unique science capabilities of the CPU-based VLBI correlator SFXC), but also touching on the aspect of graph-based algorithm execution (Dask), cloud-native data formats (MeasurementSet v4 based on xarrays w/ zarr backend) and the impact of that on processing, and, finally, bringing state-of-the-art Event Horizon Telescope-specific advanced Bayesian-inference based calibration and imaging algorithms (including dynamic imaging!) to the centimeter-wavelength VLBI community.
My colleagues Mark Kettenis and Aard Keimpema have submitted abstracts about the results of (highly) specific tasks in the project, but in my talk I can put those into wider perspective.
Head of Technical Operations and R&D at the Joint Institute for VLBI European Research Infrastructure Consortium