SKA Non-imaging Processing Software 2025 - 2027
Lead Research Organisation:
UNIVERSITY OF OXFORD
Department Name: Oxford Physics (Keble Road)
Abstract
This project is to support the component of the SKA that will enable the search of the entire sky visible from each telescope site for pulsars and time-domain transients. The SKA is made up of two telescopes, MID and LOW, each being an interferometer comprising a large number of receiving elements: in the case of MID, reflective parabolic dishes, and in the case of LOW arrays of dipoles.
Obtaining the maximum survey sensitivity requires the coherent sum of as many of elements of the telescope as possible; this generates narrow sensitivity "beams" on the sky. To cover the entire visible sky from the two SKA sites within a reasonable period of time requires simultaneously generating large numbers of beams. For MID, 1500 tied-array beams will be searched. For LOW, 500 tied-array beams will be searched. Each beam will be searched for pulsar candidates independently, and the output candidates will be brought together to exclude spurious signals.
This proposal supports the construction of this component of the telescope, which involves delivery highly efficient software for the discovery of pulsars and time-domain fast radio transients.
Obtaining the maximum survey sensitivity requires the coherent sum of as many of elements of the telescope as possible; this generates narrow sensitivity "beams" on the sky. To cover the entire visible sky from the two SKA sites within a reasonable period of time requires simultaneously generating large numbers of beams. For MID, 1500 tied-array beams will be searched. For LOW, 500 tied-array beams will be searched. Each beam will be searched for pulsar candidates independently, and the output candidates will be brought together to exclude spurious signals.
This proposal supports the construction of this component of the telescope, which involves delivery highly efficient software for the discovery of pulsars and time-domain fast radio transients.
Organisations
Publications
Hombal R
(2026)
Evaluating the effectiveness of radio frequency interference removal algorithms for single pulse searches
in RAS Techniques and Instruments
| Title | Candidate Folding and Optimization |
| Description | We have written a software module that takes candidates from the acceleration search (FDAS) and performs folding and optimization to confirm detections. This is described in SKAO objective tickets with tags: SPO-3558 , SPO-3795 , and SPO-4215. |
| Type Of Material | Data analysis technique |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | FLDO is a critical component of the pulsar discovery capabilities of SKAO. Successful tests showed the optimizer could obtain correct periods from test vectors. Analysis of FLDO performance was critical in determining software and hardware architecture for PSS processing nodes. |
| Title | Candidate Sifting Algorithms |
| Description | We have developed and implemented algorithms used to sift pulsar candidates from the harmonic summing stage of the pulsar periodicity search before they are exported for further processing. This work relates to SKAO objectives with tag SPO-3554. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | These algorithms are crucial in reducing the number of duplicate and spurious candidates, significantly lowering the processing load for the subsequent folding and optimization module. |
| Title | Fourier Domain Acceleration Search (FDAS) |
| Description | We have implemented and optimized the FDAS algorithm in both CPU and FPGA environments to detect pulsars in binary systems by searching for acceleration-induced frequency shifts. This work relates to SKAO objectives with tags SPO-3558, SPO-3793, SPO-3794, SPO-3953, SPO-3954, and SPO-4212. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | Successfully demonstrated on production-representative hardware and verified using real MeerKAT data, this implementation is critical for the SKA's ability to discover complex binary pulsar systems. |
| Title | HDBSCAN Clustering |
| Description | We have implemented the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm within the Cheetah software framework to group single pulse search candidates. This work relates to SKAO objective with tag SPO-3553. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | This algorithm improves the efficiency of identifying legitimate astronomical signals amidst background noise, with recent updates providing significant speed improvements to the processing pipeline. |
| Title | Harmonic Summing |
| Description | We have developed the Harmonic Summing component of the periodicity search, which increases the signal-to-noise ratio of pulsar signals by summing harmonically related rotational frequencies. This work relates to SKAO objective with tag SPO-3793. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | A core part of the candidate generation process, the current implementation includes internal sifting that improves overall processing efficiency and reduces data redundancy early in the pipeline. |
| Title | Parametric Occupancy Model of the Pulsar and Fast Transient Search Software |
| Description | We have produced a tool used to model pipelines and measure resource usage (RAM, data flow) using realistic search parameters. It can simulate single pulse and periodicity search pipelines simultaneously. This relates to Objectives SPO-3552 , SPO-3791 , SPO-3951 , and SPO-4214 in the SAFe engineering process for SKAO construction. |
| Type Of Material | Data analysis technique |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | Provided the first estimates of minimum RAM requirements for PSS compute nodes and allowed for the comparison of different algorithm implementations. Is vital to converge on specifications of PSS hardware. |
| Title | Radio Frequency Interference Mitigation |
| Description | We have written software designed to excise RFI from data streams, implemented for both Single Pulse Search (SPS) and CPU-based Fourier Domain Acceleration Search (FDAS) pipelines. The work has been captured in SKAO Objective tickets with tags: SPO-3798 and SPO-3954. |
| Type Of Material | Data analysis technique |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | This module is integrated into our product testing framework to ensure sensitivity compliance under realistic RFI conditions at the SKAO telescope sites. |
| Title | Random Forest Classification |
| Description | We have integrated a machine learning Random Forest algorithm within the SPSIFT module of Cheetah to automatically classify candidate clusters based on their distribution features. This work relates to SKAO objective with tag SPO-3974. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2025 |
| Provided To Others? | Yes |
| Impact | This automated classification significantly reduces the volume of candidates requiring manual inspection or further downstream computational processing, streamlining the search for new pulsars. |
| Title | The pulsar and fast radio transient search software |
| Description | The PSS software, centered around the Cheetah pipelines and Panda framework, is a real-time data processing pipeline designed for the SKAO to detect pulsars and fast transients. It integrates various modules for data reception, RFI mitigation, and search algorithms (such as Single Pulse Search and Fourier Domain Acceleration Search). Recent work has focused on containerizing the software, implementing a minimum viable control interface, and enabling the system to handle multiple scans and telescope beam configurations. In addition to the processing software, we have developped the control interface responsible for configuring and managing PSS pipelines based on telescope schemas and JSON-based configurations. It coordinates with the Central Signal Processor (CSP) LMC to create subarrays and manage node controller instances. |
| Type Of Technology | Detection Devices |
| Year Produced | 2025 |
| Impact | The software has been successfully integrated and tested at the Digital Signal PSI and Low PSI, where it demonstrated the ability to capture real-time UDP streams from the Correlator Beam Former (CBF) and execute end-to-end search pipelines. Its robustness was recently highlighted by the successful performance of a single pulsar search on real data from the early Array Assembly of the Low Telescope. The implementation of a minimum viable control interface has allowed for successful integration tests, enabling automated deployment and state management of containerized Cheetah instances during search operations. |