Speaker
Description
A real-time Radio Frequency Interference (RFI) filtering system has been commissioned and released for the Upgraded GMRT (uGMRT) users. Over the last couple of years, there has been a significant increase in the usage of the system. The system helps filter broadband powerline RFI, a dominant source of interference at frequencies below 800 MHz. This system is the first to be implemented in a radio telescope backend and released for astronomical observations. Engineering results show up to 10 dB improvement in the signal-to-noise ratio, whereas we see up to 3 dB improvement for continuum imaging and pulsar observations. Here, we describe the challenges faced and solutions adopted in implementing the RFI filtering system in real-time on the FPGA platform, testing it, understanding the effects of filtering on astronomical data, and releasing it for uGMRT observations. We would also describe the RFI statistics gathered over the last couple of years, including the possibility of using the filtering system for long-term monitoring of powerline RFI.
As a next step towards refining the filtering system, we are working towards a machine learning-based model which could adapt to the variation in the duration, amplitude, and occurrence of powerline RFI. Concerning this, we are studying the statistical properties of the composite RFI signal using high-resolution time-domain data collected over the last few years in various uGMRT observing bands. We give a glimpse of the results from this analysis and the scope for going towards a more intelligent filtering algorithm using a combination of statistical learning and signal processing.
We are also working on understanding the overall computational requirements and porting the system on modern FPGA, CPU, and GPU platforms. We would present initial results from this ongoing development which would help make the system more amenable for implementation in other radio telescope backends.
| Presentation type | Oral |
|---|
