发表机构
Commonwealth Scientific and Industrial Research Organisation (CSIRO)（联邦科学与工业研究组织）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对大口径阵列的实时成像需求，开发基于FPGA的EPIC原型，解决传统架构的复杂度问题，为下一代射电望远镜提供高效方案。
AI 中文摘要
亚毫秒级时间尺度的实时处理对探测快速天体瞬变事件（如快速射电暴FRBs）以及引力波并合的瞬时电磁对应体至关重要。即时定位可实现快速多波段后续观测，最大化科学回报，因此实时能力已成为现代宽视场口径阵列的关键要求，这类阵列正以数千至数万个天线单元的规模部署。然而，传统相关器架构的计算复杂度为$\boldsymbol{\text{O}}(N^2)$，给大$N$阵列带来显著的计算、带宽和功耗挑战，这些限制往往会缩小视场、降低角分辨率或缩短观测占空比，最终减少瞬变事件的发现潜力。采用电场并行成像“相关器”（EPIC）对全视场进行直接成像提供了一种有前景的替代方案，它避免了显式形成成对相关，对于密集口径阵列可实现$\boldsymbol{\text{O}}(N \text{ log } N)$的计算复杂度，且已在美国塞维利亚的长波阵列（LWA）台站通过图形处理单元（GPU）得到验证。不过，GPU实现受限于低比特率网格化和快速傅里叶变换（FFT）工作负载中的内存带宽。因此，我们正在开发一种采用现场可编程门阵列（FPGA）的EPIC原型实现，它利用高度并行的FFT流水线来提高计算效率并降低功耗。我们利用初步结果从可扩展性角度评估其性能，并讨论其对下一代仪器（包括SKA-Low和中频口径阵列）的适用性。
英文摘要
Real-time processing on sub-millisecond timescales is essential for detecting fast astrophysical transients such as fast radio bursts (FRBs) and prompt electromagnetic counterparts to gravitational-wave mergers. Immediate localisation enables rapid multi-wavelength follow-up and maximises scientific return. As a result, real-time capability has become a key requirement for modern wide-field aperture arrays, which are increasingly being deployed at scales of thousands to tens of thousands of antenna elements. However, traditional correlator architectures scale as $\mathcal{O}(N^2)$, creating significant computational, bandwidth, and power challenges for large-$N$ arrays. These constraints often limit field of view, angular resolution, or observing duty cycle, ultimately reducing transient discovery potential. Direct imaging of the full field of view using E-field Parallel Imaging ``Correlator'' (EPIC) offers a promising alternative by avoiding the explicit formation of pairwise correlations. EPIC can achieve $\mathcal{O}(N \log N)$ computational scaling for dense aperture arrays and has been demonstrated on the Long Wavelength Array (LWA) station in Sevilleta (USA) using Graphics Processing Unit (GPU). However, GPU implementations can be limited by memory bandwidth in low-bitwidth gridding and Fast Fourier Transform (FFT) workloads. We, therefore, are developing a prototype implementation of EPIC using Field Programmable Gate Arrays (FPGA) that exploits highly parallel FFT pipelines to improve computational efficiency and reduce power consumption. Using preliminary results, we evaluate its performance particularly from the viewpoint of scalability, and discuss its suitability for next-generation instruments including SKA-Low and mid-frequency aperture arrays.
Comments9 pages including references, 5 figures, published in proceedings of SPIE Astronomical Telescopes + Instrumentation 2026
Journal refProc. SPIE 14153, Radio Telescopes, Technologies, and Methods, 141530G (17 Aug 2026)
