Over the past months, the Radioblocks project produced several new reports detailing significant advances in simulation, data transport, and correlator applications for next-generation radio telescope systems.
A report on the overall simulation work was recently submitted. At the core of this work is a full PAF frontend simulation system developed at the Max Planck Institute for Radioastronomy. The report demonstrates the capabilities of the simulation system as a whole. It does so through a variety of simulations and setups. A detailed description of the software can be found in the documentation, available publicly in the project's open research repository.
Simulation is performed in frequency space using a fixed wavelength, with multi-wavelength studies calculated automatically in sequence. A wide variety of plotting options is available for all modules. In the left image: two examples of far-field visualisation options are shown. Left: 3D plot of a CST-simulated dielectric resonator antenna. Right: 2D polar projection of the internal dipole field (50 mm dipole length). In the right image: CST far-field simulation data of an embedded DRA antenna for six frequencies, as used in the PAF simulation.
Alongside the simulation work, a project team led by the University of Bordeaux has developed a demonstrator to test high-speed data transport in radio telescope systems. The demonstrator explores two technologies, DPDK (Data Plane Development Kit) and RDMA (Remote Direct Memory Access), aiming to achieve—and potentially exceed—400 GbE transfer rates. It also introduces the first version of the high-speed data transport prototype, planned for integration into the ALMA Wideband Sensitivity Upgrade (WSU).
Three correlator variants were implemented, all using DPDK but with different data paths. In this image: comparison of the copy behaviour for the different implementations.
Adding to these updates, a document from ASTRON, created together with JIVE, the University of Manchester, VIRAC, and other partners, offered a clear look at the project's progress on GPU-based correlator applications. These applications use GPU “radio blocks”, a collection of libraries designed for common signal-processing tasks in correlators and beam-forming. The libraries are highly optimised, each improving performance and energy efficiency while reducing the amount of code that needs to be developed and maintained. The correlator applications discussed in the report aim at handling the high data rates of upgraded instruments, taking full advantage of the efficiency offered by the shared radio blocks.
Performance and energy efficiency of the Tensor-Core Correlator (TCC) on the Grace Hopper GH200. The TCC is currently the fastest GPU correlator library available.