argosim : a Python package for radio interferometric simulations
Ezequiel Centofanti ,, Emma Ayçoberry, Samuel Farrens, Samuel Gullin, Manal Bensahli, Jean-Luc Starck, John Antoniadis · 2026
Read argosim : a Python package for radio interferometric simulations by Ezequiel Centofanti ,, Emma Ayçoberry, Samuel Farrens, Samuel Gullin, Manal Bensahli, Jean-Luc Starck, John Antoniadis (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-25573.txt.
Introduction
Radio telescopes with thousands of antennas produce data faster than it can be stored. Reconstruction methods must therefore be proven in simulation before precious observations are trusted to them. Ezequiel Centofanti and colleagues present argosim in 2026, a Python package that simulates radio interferometric observations end to end. Antenna positions go in one side; synthetic measurements and a cleaned image come out the other. The package stays modular and lightweight across major operating systems.
Its computing backend is written in JAX, which brings two advantages at once. Hardware acceleration handles large simulation volumes. Full differentiability lets gradients flow through a simulated observation into the methods being tested. That second property fits the coming era of giant arrays, where reconstruction is increasingly learned rather than hand tuned. The project is fully open source, with code public on GitHub.
Catalog record
| CREATOR | Ezequiel Centofanti ,, Emma Ayçoberry, Samuel Farrens, Samuel Gullin, Manal Bensahli, Jean-Luc Starck, John Antoniadis |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.25573
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
argosim : a Python package for radio interferometric simulations Ezequiel Centofanti ,, Emma Ayçoberry, Samuel Farrens, Samuel Gullin, Manal Bensahli, Jean-Luc Starck, John Antoniadis Abstract In this paper, we present argosim , a Python package for simulating radio interferometric observations. The argosim package is modular, lightweight and compatible with all major operating systems. Its computational backend is written in JAX , which allows for greatly accelerated performance as well as the advantage of being fully differentiable. We detail the main argosim modules and describe how to use them to generate an observation, from the antenna positions to the cleaned image. The package is a fully open-source project, and its code is publicly available on GitHub. 1 Introduction Modern radio interferometers deploy thousands of antennas that collect data at high temporal and spectral resolution, resulting in billions of uv -samples per observation ( 14 ; 9 ; 23 ; 8 ) , and reaching data rates of several Tbits/s. This volume of data presents significant processing challenges, requiring substantial computational resources, and is nearly impossible to store in its raw form. For example, the Square Kilometre Array (SKA) ( 8 ) will be forced to discard the majority of its raw data, yet it will still need to archive over several hundred petabytes of processed data per year. Given the unprecedented volume and rate of collected data, it is essential to optimise image reconstruction algorithms to the point that observations can be processed in near real time. Furthermore, since raw data will not be stored, all analysis steps must be as accurate and robust as possible, as reprocessing will not be feasible. As the next generation of radio interferometers moves through their design and construction phases, simulation tools are essential for developing and testing methods in image reconstruction, calibration, data processing, and storage. Several software packages exist for processing interferometric data ( 2 ; 25 ; 24 ; 30 ; 34 ; 26 ) , with one of the most widely used being CASA (Common Astronomy Software Applications for Radio Astronomy), originally developed for the Atacama Large Millimeter/submillimeter Array (ALMA) ( 41 ) and Very Large Array (VLA) ( 22 ; 35 ) telescopes. CASA is written in C++ and features a Python-based interface. It is a comprehensive suite that offers tools for imaging, calibration, data reduction and image deconvolution. CASA is well supported by the radio astronomy community and receives regular updates. Although highly capable, CASA may be less suitable for lightweight or experimental workflows as it can be resource-intensive, requiring significant CPU and memory. Its installation can also be challenging due to the large number of dependencies and the fact that it is written in C++, especially on machines with newer CPU architectures. Furthermore, its steep learning curve and reliance on multiple configuration files can pose barriers to quick prototyping and streamlined simulation tasks. To address the need for lightweight and flexible simulation tools, we developed argosim , a Python package for radio interferometric simulations. This software was created in the context of the ARGOS project 1 1 1 https://argos-telescope.eu/ , however, it remains generalisable to many other radio interferometers and can be used for different science cases. ARGOS is a European initiative aimed at designing a cutting-edge, low-cost, and sustainable radio interferometer in Crete, Greece. The project envisions the deployment of an instrument composed of approximately one thousand antennas, operating in the 1-3 GHz frequency range with baselines extending up to 4 km. It will address a wide range of science cases, including the detection of fast radio bursts (FRBs), pulsars, and gravitational waves (GWs). ARGOS is expected to detect a few strongly lensed FRBs per year, opening new avenues for precision cosmology, such as enabling a 1% measurement of the Hubble constant H_{0} ( 42 ) . Additionally, ARGOS will detect and characterise radio transients through imaging, providing insight into accreting compact objects, the nature of supernova progenitors and the physics of cosmic explosions. The ARGOS pathfinder, currently under construction, consists of five six-meter parabolic antennas and serves as a proof-of-concept for both the hardware and software of ARGOS. An imaging example with the ARGOS pathfinder is presented in Section 4.7 . In this paper, we present the first release of argosim , an open-source, lightweight and modular interferometric simulation package written in Python with a JAX computational backend. This paper is structured as follows. The following section introduces the fundamentals of radio interferometric imaging. Sect. 3 outlines the software design principles, with a focus on accessibility, reproducibility, and maintainability. Sect. 4 presents the argosim simulation pipeline, from antenna configuration to the generation of synthetic observations, and the computation of image reconstruction metrics. Sect. 5 discusses the advantages of accelerated array computation within argosim . Sect. 6 outlines plans for future development, and the paper concludes with final remarks in Sect. 7 . 2 Radio interferometric imaging Radio interferometers correlate signals between pairs of antennas, which by the principle of interferometry, measure the magnitude of sinusoidal spatial components of the observed sky. These components are called visibilities and their relation to the observed sky is given by the van Cittert-Zernike theorem ( 36 ) . \mathcal{V}(u,v,w)=\iint\frac{i_{{\rm sky}}(l,m)}{n}e^{-2\pi j(ul+vm+w(n-1))}{\rm d}l{\rm d}m, (1) where \mathcal{V}(u,v,w) is the complex visibility function, i_{{\rm sky}} is the observed sky intensity as a function of the two direction cosines (l,m) and n is the third direction cosine n=\sqrt{1-l^{2}-m^{2}}. (2) If we assume that the observed sources do not subtend large angles in the sky, the small sky approximation is fulfilled l^{2}+m^{2}\rightarrow 0, (3) which simplifies Eq. 1 yielding \mathcal{V}(u,v)=\iint i_{{\rm sky}}(l,m)e^{-2\pi j(ul+vm)}{\rm d}l{\rm d}m. (4) Neglecting the w term introduces a distortion associated with projecting the celestial sphere onto a plane ( 33 ) . Nevertheless, if the imaged region is sufficiently small and centred on the direction of observation, this effect can be ignored ( 4 ) . We can rewrite Eq. 4 in terms of the two-dimensional (continuous) Fourier transform \mathcal{V}(u,v)=\mathcal{F}\left[i_{{\rm sky}}(l,m)\right](u,v). (5) However, since each pair of antennas samples the complex visibility function at a specific position in the uv -plane, which depends on their separation, known as the baseline, only a limited number of noisy uv -points are measured \mathcal{V}_{i}=I_{{\rm sky}}(u_{i},v_{i})\;+\;N(u_{i},v_{i}). (6) If we represent the set of uv -sampling positions by a mask in Fourier space, we can describe the measured visibilities as follows: V=M\;\odot\;\left(F\,i_{{\rm sky}}+N\right), (7) where V are the measured visibilities, F is the discrete Fourier transform, M is the sampling mask (containing ones at uv -positions with samples and zeros elsewhere), \odot is the Hadamard product which represents the element-wise matrix multiplication, and N is the additive Gaussian noise. Note that the noise, which is white in uv -space, is correlated through the sampling mask in real space. 2.1 Aperture synthesis Eq. 7 defines the observational model in radio interferometry, in which a finite number of uv -samples, defined by the sampling mask M , are measured. Each uv -point corresponds to a measurement of a spatial component of the observed sky. The frequency and orientation of the measured components depend on the relative location of each pair of antennas. The vector of separation between a pair of antennas is called the baseline and has