| Name: | Kaushal Buch |
| Affiliation: | Giant Metrewave Radio Telescope, NCRA-TIFR, Pune, India |
| Conference ID : | ASI2023_701 |
| Title : | A Study of the Statistical Properties of Powerline RFI at GMRT |
| Authors : | Kaushal D. Buch, Giant Metrewave Radio Telescope, National Centre for Radio Astrophysics,
Pune, India
Avantika Iyengar, Giant Metrewave Radio Telescope, National Centre for Radio Astrophysics,
Pune, India
Jafar Ali Habshee, Giant Metrewave Radio Telescope, National Centre for Radio Astrophysics,
Pune, India
Divya Oberoi, Giant Metrewave Radio Telescope, National Centre for Radio Astrophysics,
Pune, India
|
| Mode of Presentation: | Poster |
| Abstract Category : | Instrumentation and Techniques |
| Abstract : | Radio Frequency Interference (RFI) is a significant challenge ground-based radio telescopes face. Giant Metrewave Radio Telescope (GMRT), which has recently undergone an upgrade, has a real-time RFI mitigation system to tackle powerline RFI. This system, a part of the GMRT Wideband Backend (GWB), is released for the upgraded GMRT (uGMRT) user community. The real-time mitigation system can detect and mitigate most of the strong powerline RFI, which manifests as outliers in the Gaussian distribution of the astronomical signal. To further optimize the mitigation system's performance and explore learning-based techniques, a more detailed study of the statistical properties of the RFI is required. We carried out an analysis on high-time-resolution digitized time series at different observing bands.
Powerline RFI occurs as a bunch of impulses repeating every 10ms and is caused due to sparking on power lines and electrical power distribution equipment. The RFI properties change broadly across the antennas, observing frequency, and observation time. We broadly classified the powerline RFI as individual sparking events, a bunch of impulses, and bunches. We modelled the randomness in the occurrence of impulses, their inter-arrival time, and amplitude distribution. The density of impulses in a bunch and the average on and off durations of RFI were analyzed using Median Absolute Deviation (MAD) based estimation and threshold detection. Individual spark morphology was analyzed using curve fitting. We would present analysis results and inferences from various GMRT data acquired at different times and observing bands.
Finally, to implement automated detection and learning-based approaches, we used a combination of Kurtosis (fourth statistical moment) and MAD to detect spark bunches. We are fine-tuning this approach to cover the variability of bunch boundaries across different datasets. |
