Accelerating Faceting Wide-Field Imaging Algorithm with FPGA for SKA Radio Telescope as a Vast Sensor Array

Area of Science:

• Astronomy and Astrophysics
• Computer Engineering
• High-Performance Computing

Background:

• The Square Kilometer Array (SKA) requires advanced algorithms like faceting for image construction from vast sensor data.
• Traditional computing struggles with the immense computational demands of SKA's wide-field imaging, particularly when accounting for Earth's curvature.

Purpose of the Study:

• To design and implement a high-efficiency Field-Programmable Gate Array (FPGA)-based hardware accelerator for the faceting algorithm in the SKA.
• To address the computational bottlenecks in the phase rotation and gridding phases of the faceting algorithm.

Main Methods:

• Algorithm analysis to identify performance bottlenecks in phase rotation and gridding.
• Optimization of memory architecture and computing logic for FPGA implementation.
• Design and simulation of loop unrolling and pipelining microstructures for the gridding accelerator.

Main Results:

• Achieved a 20x acceleration for phase rotation compared to previous work.
• Demonstrated a 5.48x acceleration ratio for gridding compared to software implementations.
• Validated the high-performance acceleration of the faceting algorithm on FPGAs.

Conclusions:

• FPGA-based hardware acceleration provides a viable solution for meeting SKA's computational constraints.
• The optimized FPGA design significantly enhances the efficiency of the faceting algorithm for radio astronomy applications.
• This approach enables high-performance image construction for vast sensor arrays like the SKA.